aboutsummaryrefslogtreecommitdiffstats
path: root/arch
diff options
context:
space:
mode:
Diffstat (limited to 'arch')
-rw-r--r--arch/alpha/kernel/Makefile2
-rw-r--r--arch/alpha/kernel/alpha_ksyms.c9
-rw-r--r--arch/alpha/kernel/semaphore.c224
-rw-r--r--arch/arm/kernel/Makefile2
-rw-r--r--arch/arm/kernel/semaphore.c221
-rw-r--r--arch/avr32/kernel/Makefile2
-rw-r--r--arch/avr32/kernel/semaphore.c148
-rw-r--r--arch/blackfin/Kconfig4
-rw-r--r--arch/blackfin/kernel/bfin_ksyms.c5
-rw-r--r--arch/cris/kernel/Makefile3
-rw-r--r--arch/cris/kernel/crisksyms.c7
-rw-r--r--arch/cris/kernel/semaphore.c129
-rw-r--r--arch/frv/kernel/Makefile2
-rw-r--r--arch/frv/kernel/frv_ksyms.c1
-rw-r--r--arch/frv/kernel/semaphore.c155
-rw-r--r--arch/h8300/kernel/Makefile2
-rw-r--r--arch/h8300/kernel/h8300_ksyms.c1
-rw-r--r--arch/h8300/kernel/semaphore.c132
-rw-r--r--arch/ia64/kernel/Makefile2
-rw-r--r--arch/ia64/kernel/ia64_ksyms.c6
-rw-r--r--arch/ia64/kernel/semaphore.c165
-rw-r--r--arch/m32r/kernel/Makefile2
-rw-r--r--arch/m32r/kernel/m32r_ksyms.c5
-rw-r--r--arch/m32r/kernel/semaphore.c185
-rw-r--r--arch/m68k/kernel/Makefile2
-rw-r--r--arch/m68k/kernel/m68k_ksyms.c6
-rw-r--r--arch/m68k/kernel/semaphore.c132
-rw-r--r--arch/m68k/lib/Makefile2
-rw-r--r--arch/m68k/lib/semaphore.S53
-rw-r--r--arch/m68knommu/kernel/Makefile2
-rw-r--r--arch/m68knommu/kernel/m68k_ksyms.c6
-rw-r--r--arch/m68knommu/kernel/semaphore.c133
-rw-r--r--arch/m68knommu/lib/Makefile2
-rw-r--r--arch/m68knommu/lib/semaphore.S66
-rw-r--r--arch/mips/kernel/Makefile2
-rw-r--r--arch/mips/kernel/semaphore.c168
-rw-r--r--arch/mn10300/kernel/Makefile2
-rw-r--r--arch/mn10300/kernel/semaphore.c149
-rw-r--r--arch/parisc/kernel/Makefile2
-rw-r--r--arch/parisc/kernel/parisc_ksyms.c5
-rw-r--r--arch/parisc/kernel/semaphore.c102
-rw-r--r--arch/powerpc/kernel/Makefile2
-rw-r--r--arch/powerpc/kernel/ppc_ksyms.c1
-rw-r--r--arch/powerpc/kernel/semaphore.c135
-rw-r--r--arch/ppc/kernel/semaphore.c131
-rw-r--r--arch/s390/Kconfig33
-rw-r--r--arch/s390/crypto/aes_s390.c8
-rw-r--r--arch/s390/crypto/des_s390.c8
-rw-r--r--arch/s390/crypto/sha1_s390.c8
-rw-r--r--arch/s390/crypto/sha256_s390.c8
-rw-r--r--arch/s390/defconfig1
-rw-r--r--arch/s390/kernel/Makefile4
-rw-r--r--arch/s390/kernel/compat_linux.h73
-rw-r--r--arch/s390/kernel/compat_signal.c11
-rw-r--r--arch/s390/kernel/debug.c53
-rw-r--r--arch/s390/kernel/early.c1
-rw-r--r--arch/s390/kernel/entry.h60
-rw-r--r--arch/s390/kernel/entry64.S2
-rw-r--r--arch/s390/kernel/ipl.c2
-rw-r--r--arch/s390/kernel/kprobes.c2
-rw-r--r--arch/s390/kernel/process.c77
-rw-r--r--arch/s390/kernel/ptrace.c1
-rw-r--r--arch/s390/kernel/s390_ext.c14
-rw-r--r--arch/s390/kernel/s390_ksyms.c7
-rw-r--r--arch/s390/kernel/semaphore.c108
-rw-r--r--arch/s390/kernel/setup.c15
-rw-r--r--arch/s390/kernel/signal.c16
-rw-r--r--arch/s390/kernel/smp.c91
-rw-r--r--arch/s390/kernel/sys_s390.c2
-rw-r--r--arch/s390/kernel/time.c259
-rw-r--r--arch/s390/kernel/topology.c314
-rw-r--r--arch/s390/kernel/traps.c17
-rw-r--r--arch/s390/lib/delay.c14
-rw-r--r--arch/s390/lib/uaccess_pt.c59
-rw-r--r--arch/s390/mm/extmem.c67
-rw-r--r--arch/s390/mm/fault.c21
-rw-r--r--arch/s390/mm/init.c1
-rw-r--r--arch/sh/kernel/Makefile_322
-rw-r--r--arch/sh/kernel/Makefile_642
-rw-r--r--arch/sh/kernel/semaphore.c139
-rw-r--r--arch/sh/kernel/sh_ksyms_32.c7
-rw-r--r--arch/sh/kernel/sh_ksyms_64.c4
-rw-r--r--arch/sparc/kernel/Makefile2
-rw-r--r--arch/sparc/kernel/semaphore.c155
-rw-r--r--arch/sparc/kernel/sparc_ksyms.c5
-rw-r--r--arch/sparc64/kernel/Makefile2
-rw-r--r--arch/sparc64/kernel/semaphore.c254
-rw-r--r--arch/sparc64/kernel/sparc64_ksyms.c6
-rw-r--r--arch/um/Kconfig.i3864
-rw-r--r--arch/um/Kconfig.x86_644
-rw-r--r--arch/um/sys-i386/ksyms.c12
-rw-r--r--arch/um/sys-ppc/Makefile8
-rw-r--r--arch/um/sys-x86_64/ksyms.c13
-rw-r--r--arch/v850/kernel/Makefile2
-rw-r--r--arch/v850/kernel/semaphore.c166
-rw-r--r--arch/v850/kernel/v850_ksyms.c7
-rw-r--r--arch/x86/Kconfig73
-rw-r--r--arch/x86/Kconfig.cpu2
-rw-r--r--arch/x86/Kconfig.debug26
-rw-r--r--arch/x86/Makefile7
-rw-r--r--arch/x86/boot/Makefile16
-rw-r--r--arch/x86/boot/boot.h5
-rw-r--r--arch/x86/boot/compressed/Makefile2
-rw-r--r--arch/x86/boot/compressed/misc.c202
-rw-r--r--arch/x86/boot/cpucheck.c20
-rw-r--r--arch/x86/boot/header.S6
-rw-r--r--arch/x86/boot/pm.c2
-rw-r--r--arch/x86/boot/tools/build.c88
-rw-r--r--arch/x86/boot/video-bios.c6
-rw-r--r--arch/x86/boot/video-mode.c173
-rw-r--r--arch/x86/boot/video-vesa.c8
-rw-r--r--arch/x86/boot/video-vga.c12
-rw-r--r--arch/x86/boot/video.c157
-rw-r--r--arch/x86/ia32/ia32_signal.c2
-rw-r--r--arch/x86/ia32/ia32entry.S12
-rw-r--r--arch/x86/ia32/sys_ia32.c31
-rw-r--r--arch/x86/kernel/Makefile19
-rw-r--r--arch/x86/kernel/acpi/Makefile9
-rw-r--r--arch/x86/kernel/acpi/boot.c67
-rw-r--r--arch/x86/kernel/acpi/realmode/Makefile57
-rw-r--r--arch/x86/kernel/acpi/realmode/copy.S1
-rw-r--r--arch/x86/kernel/acpi/realmode/video-bios.c1
-rw-r--r--arch/x86/kernel/acpi/realmode/video-mode.c1
-rw-r--r--arch/x86/kernel/acpi/realmode/video-vesa.c1
-rw-r--r--arch/x86/kernel/acpi/realmode/video-vga.c1
-rw-r--r--arch/x86/kernel/acpi/realmode/wakemain.c81
-rw-r--r--arch/x86/kernel/acpi/realmode/wakeup.S113
-rw-r--r--arch/x86/kernel/acpi/realmode/wakeup.h36
-rw-r--r--arch/x86/kernel/acpi/realmode/wakeup.lds.S61
-rw-r--r--arch/x86/kernel/acpi/sleep.c73
-rw-r--r--arch/x86/kernel/acpi/sleep.h16
-rw-r--r--arch/x86/kernel/acpi/sleep_32.c40
-rw-r--r--arch/x86/kernel/acpi/wakeup_32.S247
-rw-r--r--arch/x86/kernel/acpi/wakeup_64.S313
-rw-r--r--arch/x86/kernel/acpi/wakeup_rm.S10
-rw-r--r--arch/x86/kernel/alternative.c103
-rw-r--r--arch/x86/kernel/aperture_64.c6
-rw-r--r--arch/x86/kernel/apic_32.c224
-rw-r--r--arch/x86/kernel/apic_64.c140
-rw-r--r--arch/x86/kernel/apm_32.c5
-rw-r--r--arch/x86/kernel/asm-offsets_32.c2
-rw-r--r--arch/x86/kernel/bugs_64.c14
-rw-r--r--arch/x86/kernel/cpu/Makefile4
-rw-r--r--arch/x86/kernel/cpu/amd.c121
-rw-r--r--arch/x86/kernel/cpu/centaur.c490
-rw-r--r--arch/x86/kernel/cpu/common.c180
-rw-r--r--arch/x86/kernel/cpu/cpu.h26
-rw-r--r--arch/x86/kernel/cpu/cyrix.c136
-rw-r--r--arch/x86/kernel/cpu/feature_names.c2
-rw-r--r--arch/x86/kernel/cpu/intel.c106
-rw-r--r--arch/x86/kernel/cpu/mcheck/mce_32.c50
-rw-r--r--arch/x86/kernel/cpu/mcheck/non-fatal.c21
-rw-r--r--arch/x86/kernel/cpu/mcheck/p5.c16
-rw-r--r--arch/x86/kernel/cpu/mcheck/p6.c48
-rw-r--r--arch/x86/kernel/cpu/mcheck/winchip.c8
-rw-r--r--arch/x86/kernel/cpu/mtrr/generic.c139
-rw-r--r--arch/x86/kernel/cpu/mtrr/if.c7
-rw-r--r--arch/x86/kernel/cpu/mtrr/main.c2
-rw-r--r--arch/x86/kernel/cpu/mtrr/state.c14
-rw-r--r--arch/x86/kernel/cpu/nexgen.c15
-rw-r--r--arch/x86/kernel/cpu/proc.c170
-rw-r--r--arch/x86/kernel/cpu/transmeta.c30
-rw-r--r--arch/x86/kernel/cpu/umc.c19
-rw-r--r--arch/x86/kernel/crash.c4
-rw-r--r--arch/x86/kernel/ds.c8
-rw-r--r--arch/x86/kernel/e820_32.c27
-rw-r--r--arch/x86/kernel/e820_64.c156
-rw-r--r--arch/x86/kernel/early_printk.c24
-rw-r--r--arch/x86/kernel/entry_32.S39
-rw-r--r--arch/x86/kernel/entry_64.S8
-rw-r--r--arch/x86/kernel/genapic_64.c47
-rw-r--r--arch/x86/kernel/genapic_flat_64.c7
-rw-r--r--arch/x86/kernel/genx2apic_uv_x.c245
-rw-r--r--arch/x86/kernel/head32.c14
-rw-r--r--arch/x86/kernel/head64.c78
-rw-r--r--arch/x86/kernel/head_32.S2
-rw-r--r--arch/x86/kernel/head_64.S28
-rw-r--r--arch/x86/kernel/i386_ksyms_32.c5
-rw-r--r--arch/x86/kernel/i387.c88
-rw-r--r--arch/x86/kernel/io_apic_32.c164
-rw-r--r--arch/x86/kernel/io_apic_64.c63
-rw-r--r--arch/x86/kernel/ipi.c178
-rw-r--r--arch/x86/kernel/irq_32.c2
-rw-r--r--arch/x86/kernel/kprobes.c14
-rw-r--r--arch/x86/kernel/mca_32.c96
-rw-r--r--arch/x86/kernel/microcode.c16
-rw-r--r--arch/x86/kernel/mpparse.c (renamed from arch/x86/kernel/mpparse_32.c)850
-rw-r--r--arch/x86/kernel/mpparse_64.c867
-rw-r--r--arch/x86/kernel/msr.c4
-rw-r--r--arch/x86/kernel/nmi_32.c14
-rw-r--r--arch/x86/kernel/nmi_64.c2
-rw-r--r--arch/x86/kernel/paravirt.c18
-rw-r--r--arch/x86/kernel/pci-dma_64.c20
-rw-r--r--arch/x86/kernel/process_32.c41
-rw-r--r--arch/x86/kernel/process_64.c27
-rw-r--r--arch/x86/kernel/ptrace.c1
-rw-r--r--arch/x86/kernel/reboot.c5
-rw-r--r--arch/x86/kernel/relocate_kernel_32.S30
-rw-r--r--arch/x86/kernel/relocate_kernel_64.S40
-rw-r--r--arch/x86/kernel/rtc.c33
-rw-r--r--arch/x86/kernel/setup.c113
-rw-r--r--arch/x86/kernel/setup64.c81
-rw-r--r--arch/x86/kernel/setup_32.c95
-rw-r--r--arch/x86/kernel/setup_64.c224
-rw-r--r--arch/x86/kernel/sigframe.h (renamed from arch/x86/kernel/sigframe_32.h)14
-rw-r--r--arch/x86/kernel/signal_32.c303
-rw-r--r--arch/x86/kernel/signal_64.c108
-rw-r--r--arch/x86/kernel/smp.c343
-rw-r--r--arch/x86/kernel/smp_32.c712
-rw-r--r--arch/x86/kernel/smpboot.c (renamed from arch/x86/kernel/smpboot_32.c)1345
-rw-r--r--arch/x86/kernel/smpboot_64.c1108
-rw-r--r--arch/x86/kernel/smpcommon.c83
-rw-r--r--arch/x86/kernel/smpcommon_32.c81
-rw-r--r--arch/x86/kernel/srat_32.c10
-rw-r--r--arch/x86/kernel/step.c2
-rw-r--r--arch/x86/kernel/summit_32.c57
-rw-r--r--arch/x86/kernel/syscall_64.c13
-rw-r--r--arch/x86/kernel/test_nx.c2
-rw-r--r--arch/x86/kernel/tlb_32.c243
-rw-r--r--arch/x86/kernel/tlb_64.c (renamed from arch/x86/kernel/smp_64.c)266
-rw-r--r--arch/x86/kernel/trampoline.c18
-rw-r--r--arch/x86/kernel/trampoline_64.S5
-rw-r--r--arch/x86/kernel/traps_32.c612
-rw-r--r--arch/x86/kernel/traps_64.c2
-rw-r--r--arch/x86/kernel/tsc_32.c16
-rw-r--r--arch/x86/kernel/tsc_64.c3
-rw-r--r--arch/x86/kernel/vm86_32.c213
-rw-r--r--arch/x86/kernel/vmlinux_32.lds.S5
-rw-r--r--arch/x86/kernel/vmlinux_64.lds.S11
-rw-r--r--arch/x86/kernel/vsmp_64.c131
-rw-r--r--arch/x86/kernel/x8664_ksyms_64.c18
-rw-r--r--arch/x86/lguest/boot.c2
-rw-r--r--arch/x86/lib/memcpy_32.c2
-rw-r--r--arch/x86/lib/memmove_64.c8
-rw-r--r--arch/x86/lib/mmx_32.c197
-rw-r--r--arch/x86/lib/semaphore_32.S83
-rw-r--r--arch/x86/lib/string_32.c60
-rw-r--r--arch/x86/lib/strstr_32.c4
-rw-r--r--arch/x86/lib/thunk_64.S5
-rw-r--r--arch/x86/lib/usercopy_32.c122
-rw-r--r--arch/x86/mach-generic/bigsmp.c30
-rw-r--r--arch/x86/mach-generic/default.c8
-rw-r--r--arch/x86/mach-generic/probe.c37
-rw-r--r--arch/x86/mach-generic/summit.c8
-rw-r--r--arch/x86/mach-rdc321x/Makefile2
-rw-r--r--arch/x86/mach-rdc321x/wdt.c275
-rw-r--r--arch/x86/mach-voyager/voyager_smp.c18
-rw-r--r--arch/x86/math-emu/fpu_entry.c1
-rw-r--r--arch/x86/math-emu/reg_ld_str.c13
-rw-r--r--arch/x86/mm/Makefile16
-rw-r--r--arch/x86/mm/Makefile_329
-rw-r--r--arch/x86/mm/Makefile_649
-rw-r--r--arch/x86/mm/discontig_32.c2
-rw-r--r--arch/x86/mm/dump_pagetables.c354
-rw-r--r--arch/x86/mm/fault.c6
-rw-r--r--arch/x86/mm/init_32.c35
-rw-r--r--arch/x86/mm/init_64.c222
-rw-r--r--arch/x86/mm/ioremap.c152
-rw-r--r--arch/x86/mm/numa_64.c6
-rw-r--r--arch/x86/mm/pageattr.c105
-rw-r--r--arch/x86/mm/pat.c421
-rw-r--r--arch/x86/mm/pgtable_32.c8
-rw-r--r--arch/x86/mm/srat_64.c7
-rw-r--r--arch/x86/oprofile/init.c13
-rw-r--r--arch/x86/oprofile/nmi_timer_int.c6
-rw-r--r--arch/x86/oprofile/op_model_athlon.c46
-rw-r--r--arch/x86/oprofile/op_model_ppro.c52
-rw-r--r--arch/x86/pci/i386.c73
-rw-r--r--arch/x86/pci/irq.c12
-rw-r--r--arch/x86/pci/numa.c32
-rw-r--r--arch/x86/power/cpu_32.c41
-rw-r--r--arch/x86/vdso/Makefile2
-rw-r--r--arch/x86/vdso/vdso32-setup.c11
-rw-r--r--arch/x86/xen/enlighten.c7
-rw-r--r--arch/x86/xen/multicalls.c4
-rw-r--r--arch/x86/xen/smp.c6
-rw-r--r--arch/x86/xen/xen-asm.S11
-rw-r--r--arch/x86/xen/xen-ops.h2
-rw-r--r--arch/xtensa/kernel/Makefile2
-rw-r--r--arch/xtensa/kernel/semaphore.c226
-rw-r--r--arch/xtensa/kernel/xtensa_ksyms.c9
281 files changed, 8951 insertions, 12096 deletions
diff --git a/arch/alpha/kernel/Makefile b/arch/alpha/kernel/Makefile
index dccf05245d4d..ac706c1d7ada 100644
--- a/arch/alpha/kernel/Makefile
+++ b/arch/alpha/kernel/Makefile
@@ -7,7 +7,7 @@ EXTRA_AFLAGS := $(KBUILD_CFLAGS)
7EXTRA_CFLAGS := -Werror -Wno-sign-compare 7EXTRA_CFLAGS := -Werror -Wno-sign-compare
8 8
9obj-y := entry.o traps.o process.o init_task.o osf_sys.o irq.o \ 9obj-y := entry.o traps.o process.o init_task.o osf_sys.o irq.o \
10 irq_alpha.o signal.o setup.o ptrace.o time.o semaphore.o \ 10 irq_alpha.o signal.o setup.o ptrace.o time.o \
11 alpha_ksyms.o systbls.o err_common.o io.o 11 alpha_ksyms.o systbls.o err_common.o io.o
12 12
13obj-$(CONFIG_VGA_HOSE) += console.o 13obj-$(CONFIG_VGA_HOSE) += console.o
diff --git a/arch/alpha/kernel/alpha_ksyms.c b/arch/alpha/kernel/alpha_ksyms.c
index e9762a33b043..d96e742d4dc2 100644
--- a/arch/alpha/kernel/alpha_ksyms.c
+++ b/arch/alpha/kernel/alpha_ksyms.c
@@ -77,15 +77,6 @@ EXPORT_SYMBOL(__do_clear_user);
77EXPORT_SYMBOL(__strncpy_from_user); 77EXPORT_SYMBOL(__strncpy_from_user);
78EXPORT_SYMBOL(__strnlen_user); 78EXPORT_SYMBOL(__strnlen_user);
79 79
80/* Semaphore helper functions. */
81EXPORT_SYMBOL(__down_failed);
82EXPORT_SYMBOL(__down_failed_interruptible);
83EXPORT_SYMBOL(__up_wakeup);
84EXPORT_SYMBOL(down);
85EXPORT_SYMBOL(down_interruptible);
86EXPORT_SYMBOL(down_trylock);
87EXPORT_SYMBOL(up);
88
89/* 80/*
90 * SMP-specific symbols. 81 * SMP-specific symbols.
91 */ 82 */
diff --git a/arch/alpha/kernel/semaphore.c b/arch/alpha/kernel/semaphore.c
deleted file mode 100644
index 8d2982aa1b8d..000000000000
--- a/arch/alpha/kernel/semaphore.c
+++ /dev/null
@@ -1,224 +0,0 @@
1/*
2 * Alpha semaphore implementation.
3 *
4 * (C) Copyright 1996 Linus Torvalds
5 * (C) Copyright 1999, 2000 Richard Henderson
6 */
7
8#include <linux/errno.h>
9#include <linux/sched.h>
10#include <linux/init.h>
11
12/*
13 * This is basically the PPC semaphore scheme ported to use
14 * the Alpha ll/sc sequences, so see the PPC code for
15 * credits.
16 */
17
18/*
19 * Atomically update sem->count.
20 * This does the equivalent of the following:
21 *
22 * old_count = sem->count;
23 * tmp = MAX(old_count, 0) + incr;
24 * sem->count = tmp;
25 * return old_count;
26 */
27static inline int __sem_update_count(struct semaphore *sem, int incr)
28{
29 long old_count, tmp = 0;
30
31 __asm__ __volatile__(
32 "1: ldl_l %0,%2\n"
33 " cmovgt %0,%0,%1\n"
34 " addl %1,%3,%1\n"
35 " stl_c %1,%2\n"
36 " beq %1,2f\n"
37 " mb\n"
38 ".subsection 2\n"
39 "2: br 1b\n"
40 ".previous"
41 : "=&r" (old_count), "=&r" (tmp), "=m" (sem->count)
42 : "Ir" (incr), "1" (tmp), "m" (sem->count));
43
44 return old_count;
45}
46
47/*
48 * Perform the "down" function. Return zero for semaphore acquired,
49 * return negative for signalled out of the function.
50 *
51 * If called from down, the return is ignored and the wait loop is
52 * not interruptible. This means that a task waiting on a semaphore
53 * using "down()" cannot be killed until someone does an "up()" on
54 * the semaphore.
55 *
56 * If called from down_interruptible, the return value gets checked
57 * upon return. If the return value is negative then the task continues
58 * with the negative value in the return register (it can be tested by
59 * the caller).
60 *
61 * Either form may be used in conjunction with "up()".
62 */
63
64void __sched
65__down_failed(struct semaphore *sem)
66{
67 struct task_struct *tsk = current;
68 DECLARE_WAITQUEUE(wait, tsk);
69
70#ifdef CONFIG_DEBUG_SEMAPHORE
71 printk("%s(%d): down failed(%p)\n",
72 tsk->comm, task_pid_nr(tsk), sem);
73#endif
74
75 tsk->state = TASK_UNINTERRUPTIBLE;
76 wmb();
77 add_wait_queue_exclusive(&sem->wait, &wait);
78
79 /*
80 * Try to get the semaphore. If the count is > 0, then we've
81 * got the semaphore; we decrement count and exit the loop.
82 * If the count is 0 or negative, we set it to -1, indicating
83 * that we are asleep, and then sleep.
84 */
85 while (__sem_update_count(sem, -1) <= 0) {
86 schedule();
87 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
88 }
89 remove_wait_queue(&sem->wait, &wait);
90 tsk->state = TASK_RUNNING;
91
92 /*
93 * If there are any more sleepers, wake one of them up so
94 * that it can either get the semaphore, or set count to -1
95 * indicating that there are still processes sleeping.
96 */
97 wake_up(&sem->wait);
98
99#ifdef CONFIG_DEBUG_SEMAPHORE
100 printk("%s(%d): down acquired(%p)\n",
101 tsk->comm, task_pid_nr(tsk), sem);
102#endif
103}
104
105int __sched
106__down_failed_interruptible(struct semaphore *sem)
107{
108 struct task_struct *tsk = current;
109 DECLARE_WAITQUEUE(wait, tsk);
110 long ret = 0;
111
112#ifdef CONFIG_DEBUG_SEMAPHORE
113 printk("%s(%d): down failed(%p)\n",
114 tsk->comm, task_pid_nr(tsk), sem);
115#endif
116
117 tsk->state = TASK_INTERRUPTIBLE;
118 wmb();
119 add_wait_queue_exclusive(&sem->wait, &wait);
120
121 while (__sem_update_count(sem, -1) <= 0) {
122 if (signal_pending(current)) {
123 /*
124 * A signal is pending - give up trying.
125 * Set sem->count to 0 if it is negative,
126 * since we are no longer sleeping.
127 */
128 __sem_update_count(sem, 0);
129 ret = -EINTR;
130 break;
131 }
132 schedule();
133 set_task_state(tsk, TASK_INTERRUPTIBLE);
134 }
135
136 remove_wait_queue(&sem->wait, &wait);
137 tsk->state = TASK_RUNNING;
138 wake_up(&sem->wait);
139
140#ifdef CONFIG_DEBUG_SEMAPHORE
141 printk("%s(%d): down %s(%p)\n",
142 current->comm, task_pid_nr(current),
143 (ret < 0 ? "interrupted" : "acquired"), sem);
144#endif
145 return ret;
146}
147
148void
149__up_wakeup(struct semaphore *sem)
150{
151 /*
152 * Note that we incremented count in up() before we came here,
153 * but that was ineffective since the result was <= 0, and
154 * any negative value of count is equivalent to 0.
155 * This ends up setting count to 1, unless count is now > 0
156 * (i.e. because some other cpu has called up() in the meantime),
157 * in which case we just increment count.
158 */
159 __sem_update_count(sem, 1);
160 wake_up(&sem->wait);
161}
162
163void __sched
164down(struct semaphore *sem)
165{
166#ifdef WAITQUEUE_DEBUG
167 CHECK_MAGIC(sem->__magic);
168#endif
169#ifdef CONFIG_DEBUG_SEMAPHORE
170 printk("%s(%d): down(%p) <count=%d> from %p\n",
171 current->comm, task_pid_nr(current), sem,
172 atomic_read(&sem->count), __builtin_return_address(0));
173#endif
174 __down(sem);
175}
176
177int __sched
178down_interruptible(struct semaphore *sem)
179{
180#ifdef WAITQUEUE_DEBUG
181 CHECK_MAGIC(sem->__magic);
182#endif
183#ifdef CONFIG_DEBUG_SEMAPHORE
184 printk("%s(%d): down(%p) <count=%d> from %p\n",
185 current->comm, task_pid_nr(current), sem,
186 atomic_read(&sem->count), __builtin_return_address(0));
187#endif
188 return __down_interruptible(sem);
189}
190
191int
192down_trylock(struct semaphore *sem)
193{
194 int ret;
195
196#ifdef WAITQUEUE_DEBUG
197 CHECK_MAGIC(sem->__magic);
198#endif
199
200 ret = __down_trylock(sem);
201
202#ifdef CONFIG_DEBUG_SEMAPHORE
203 printk("%s(%d): down_trylock %s from %p\n",
204 current->comm, task_pid_nr(current),
205 ret ? "failed" : "acquired",
206 __builtin_return_address(0));
207#endif
208
209 return ret;
210}
211
212void
213up(struct semaphore *sem)
214{
215#ifdef WAITQUEUE_DEBUG
216 CHECK_MAGIC(sem->__magic);
217#endif
218#ifdef CONFIG_DEBUG_SEMAPHORE
219 printk("%s(%d): up(%p) <count=%d> from %p\n",
220 current->comm, task_pid_nr(current), sem,
221 atomic_read(&sem->count), __builtin_return_address(0));
222#endif
223 __up(sem);
224}
diff --git a/arch/arm/kernel/Makefile b/arch/arm/kernel/Makefile
index 00d44c6fbfe9..6235f72a14f0 100644
--- a/arch/arm/kernel/Makefile
+++ b/arch/arm/kernel/Makefile
@@ -7,7 +7,7 @@ AFLAGS_head.o := -DTEXT_OFFSET=$(TEXT_OFFSET)
7# Object file lists. 7# Object file lists.
8 8
9obj-y := compat.o entry-armv.o entry-common.o irq.o \ 9obj-y := compat.o entry-armv.o entry-common.o irq.o \
10 process.o ptrace.o semaphore.o setup.o signal.o \ 10 process.o ptrace.o setup.o signal.o \
11 sys_arm.o stacktrace.o time.o traps.o 11 sys_arm.o stacktrace.o time.o traps.o
12 12
13obj-$(CONFIG_ISA_DMA_API) += dma.o 13obj-$(CONFIG_ISA_DMA_API) += dma.o
diff --git a/arch/arm/kernel/semaphore.c b/arch/arm/kernel/semaphore.c
deleted file mode 100644
index 981fe5c6ccbe..000000000000
--- a/arch/arm/kernel/semaphore.c
+++ /dev/null
@@ -1,221 +0,0 @@
1/*
2 * ARM semaphore implementation, taken from
3 *
4 * i386 semaphore implementation.
5 *
6 * (C) Copyright 1999 Linus Torvalds
7 *
8 * Modified for ARM by Russell King
9 *
10 * This program is free software; you can redistribute it and/or modify
11 * it under the terms of the GNU General Public License version 2 as
12 * published by the Free Software Foundation.
13 */
14#include <linux/module.h>
15#include <linux/sched.h>
16#include <linux/errno.h>
17#include <linux/init.h>
18
19#include <asm/semaphore.h>
20
21/*
22 * Semaphores are implemented using a two-way counter:
23 * The "count" variable is decremented for each process
24 * that tries to acquire the semaphore, while the "sleeping"
25 * variable is a count of such acquires.
26 *
27 * Notably, the inline "up()" and "down()" functions can
28 * efficiently test if they need to do any extra work (up
29 * needs to do something only if count was negative before
30 * the increment operation.
31 *
32 * "sleeping" and the contention routine ordering is
33 * protected by the semaphore spinlock.
34 *
35 * Note that these functions are only called when there is
36 * contention on the lock, and as such all this is the
37 * "non-critical" part of the whole semaphore business. The
38 * critical part is the inline stuff in <asm/semaphore.h>
39 * where we want to avoid any extra jumps and calls.
40 */
41
42/*
43 * Logic:
44 * - only on a boundary condition do we need to care. When we go
45 * from a negative count to a non-negative, we wake people up.
46 * - when we go from a non-negative count to a negative do we
47 * (a) synchronize with the "sleeper" count and (b) make sure
48 * that we're on the wakeup list before we synchronize so that
49 * we cannot lose wakeup events.
50 */
51
52void __up(struct semaphore *sem)
53{
54 wake_up(&sem->wait);
55}
56
57static DEFINE_SPINLOCK(semaphore_lock);
58
59void __sched __down(struct semaphore * sem)
60{
61 struct task_struct *tsk = current;
62 DECLARE_WAITQUEUE(wait, tsk);
63 tsk->state = TASK_UNINTERRUPTIBLE;
64 add_wait_queue_exclusive(&sem->wait, &wait);
65
66 spin_lock_irq(&semaphore_lock);
67 sem->sleepers++;
68 for (;;) {
69 int sleepers = sem->sleepers;
70
71 /*
72 * Add "everybody else" into it. They aren't
73 * playing, because we own the spinlock.
74 */
75 if (!atomic_add_negative(sleepers - 1, &sem->count)) {
76 sem->sleepers = 0;
77 break;
78 }
79 sem->sleepers = 1; /* us - see -1 above */
80 spin_unlock_irq(&semaphore_lock);
81
82 schedule();
83 tsk->state = TASK_UNINTERRUPTIBLE;
84 spin_lock_irq(&semaphore_lock);
85 }
86 spin_unlock_irq(&semaphore_lock);
87 remove_wait_queue(&sem->wait, &wait);
88 tsk->state = TASK_RUNNING;
89 wake_up(&sem->wait);
90}
91
92int __sched __down_interruptible(struct semaphore * sem)
93{
94 int retval = 0;
95 struct task_struct *tsk = current;
96 DECLARE_WAITQUEUE(wait, tsk);
97 tsk->state = TASK_INTERRUPTIBLE;
98 add_wait_queue_exclusive(&sem->wait, &wait);
99
100 spin_lock_irq(&semaphore_lock);
101 sem->sleepers ++;
102 for (;;) {
103 int sleepers = sem->sleepers;
104
105 /*
106 * With signals pending, this turns into
107 * the trylock failure case - we won't be
108 * sleeping, and we* can't get the lock as
109 * it has contention. Just correct the count
110 * and exit.
111 */
112 if (signal_pending(current)) {
113 retval = -EINTR;
114 sem->sleepers = 0;
115 atomic_add(sleepers, &sem->count);
116 break;
117 }
118
119 /*
120 * Add "everybody else" into it. They aren't
121 * playing, because we own the spinlock. The
122 * "-1" is because we're still hoping to get
123 * the lock.
124 */
125 if (!atomic_add_negative(sleepers - 1, &sem->count)) {
126 sem->sleepers = 0;
127 break;
128 }
129 sem->sleepers = 1; /* us - see -1 above */
130 spin_unlock_irq(&semaphore_lock);
131
132 schedule();
133 tsk->state = TASK_INTERRUPTIBLE;
134 spin_lock_irq(&semaphore_lock);
135 }
136 spin_unlock_irq(&semaphore_lock);
137 tsk->state = TASK_RUNNING;
138 remove_wait_queue(&sem->wait, &wait);
139 wake_up(&sem->wait);
140 return retval;
141}
142
143/*
144 * Trylock failed - make sure we correct for
145 * having decremented the count.
146 *
147 * We could have done the trylock with a
148 * single "cmpxchg" without failure cases,
149 * but then it wouldn't work on a 386.
150 */
151int __down_trylock(struct semaphore * sem)
152{
153 int sleepers;
154 unsigned long flags;
155
156 spin_lock_irqsave(&semaphore_lock, flags);
157 sleepers = sem->sleepers + 1;
158 sem->sleepers = 0;
159
160 /*
161 * Add "everybody else" and us into it. They aren't
162 * playing, because we own the spinlock.
163 */
164 if (!atomic_add_negative(sleepers, &sem->count))
165 wake_up(&sem->wait);
166
167 spin_unlock_irqrestore(&semaphore_lock, flags);
168 return 1;
169}
170
171/*
172 * The semaphore operations have a special calling sequence that
173 * allow us to do a simpler in-line version of them. These routines
174 * need to convert that sequence back into the C sequence when
175 * there is contention on the semaphore.
176 *
177 * ip contains the semaphore pointer on entry. Save the C-clobbered
178 * registers (r0 to r3 and lr), but not ip, as we use it as a return
179 * value in some cases..
180 * To remain AAPCS compliant (64-bit stack align) we save r4 as well.
181 */
182asm(" .section .sched.text,\"ax\",%progbits \n\
183 .align 5 \n\
184 .globl __down_failed \n\
185__down_failed: \n\
186 stmfd sp!, {r0 - r4, lr} \n\
187 mov r0, ip \n\
188 bl __down \n\
189 ldmfd sp!, {r0 - r4, pc} \n\
190 \n\
191 .align 5 \n\
192 .globl __down_interruptible_failed \n\
193__down_interruptible_failed: \n\
194 stmfd sp!, {r0 - r4, lr} \n\
195 mov r0, ip \n\
196 bl __down_interruptible \n\
197 mov ip, r0 \n\
198 ldmfd sp!, {r0 - r4, pc} \n\
199 \n\
200 .align 5 \n\
201 .globl __down_trylock_failed \n\
202__down_trylock_failed: \n\
203 stmfd sp!, {r0 - r4, lr} \n\
204 mov r0, ip \n\
205 bl __down_trylock \n\
206 mov ip, r0 \n\
207 ldmfd sp!, {r0 - r4, pc} \n\
208 \n\
209 .align 5 \n\
210 .globl __up_wakeup \n\
211__up_wakeup: \n\
212 stmfd sp!, {r0 - r4, lr} \n\
213 mov r0, ip \n\
214 bl __up \n\
215 ldmfd sp!, {r0 - r4, pc} \n\
216 ");
217
218EXPORT_SYMBOL(__down_failed);
219EXPORT_SYMBOL(__down_interruptible_failed);
220EXPORT_SYMBOL(__down_trylock_failed);
221EXPORT_SYMBOL(__up_wakeup);
diff --git a/arch/avr32/kernel/Makefile b/arch/avr32/kernel/Makefile
index e4b6d122b033..18229d0d1861 100644
--- a/arch/avr32/kernel/Makefile
+++ b/arch/avr32/kernel/Makefile
@@ -6,7 +6,7 @@ extra-y := head.o vmlinux.lds
6 6
7obj-$(CONFIG_SUBARCH_AVR32B) += entry-avr32b.o 7obj-$(CONFIG_SUBARCH_AVR32B) += entry-avr32b.o
8obj-y += syscall_table.o syscall-stubs.o irq.o 8obj-y += syscall_table.o syscall-stubs.o irq.o
9obj-y += setup.o traps.o semaphore.o ocd.o ptrace.o 9obj-y += setup.o traps.o ocd.o ptrace.o
10obj-y += signal.o sys_avr32.o process.o time.o 10obj-y += signal.o sys_avr32.o process.o time.o
11obj-y += init_task.o switch_to.o cpu.o 11obj-y += init_task.o switch_to.o cpu.o
12obj-$(CONFIG_MODULES) += module.o avr32_ksyms.o 12obj-$(CONFIG_MODULES) += module.o avr32_ksyms.o
diff --git a/arch/avr32/kernel/semaphore.c b/arch/avr32/kernel/semaphore.c
deleted file mode 100644
index 1e2705a05016..000000000000
--- a/arch/avr32/kernel/semaphore.c
+++ /dev/null
@@ -1,148 +0,0 @@
1/*
2 * AVR32 sempahore implementation.
3 *
4 * Copyright (C) 2004-2006 Atmel Corporation
5 *
6 * Based on linux/arch/i386/kernel/semaphore.c
7 * Copyright (C) 1999 Linus Torvalds
8 *
9 * This program is free software; you can redistribute it and/or modify
10 * it under the terms of the GNU General Public License version 2 as
11 * published by the Free Software Foundation.
12 */
13
14#include <linux/sched.h>
15#include <linux/errno.h>
16#include <linux/module.h>
17
18#include <asm/semaphore.h>
19#include <asm/atomic.h>
20
21/*
22 * Semaphores are implemented using a two-way counter:
23 * The "count" variable is decremented for each process
24 * that tries to acquire the semaphore, while the "sleeping"
25 * variable is a count of such acquires.
26 *
27 * Notably, the inline "up()" and "down()" functions can
28 * efficiently test if they need to do any extra work (up
29 * needs to do something only if count was negative before
30 * the increment operation.
31 *
32 * "sleeping" and the contention routine ordering is protected
33 * by the spinlock in the semaphore's waitqueue head.
34 *
35 * Note that these functions are only called when there is
36 * contention on the lock, and as such all this is the
37 * "non-critical" part of the whole semaphore business. The
38 * critical part is the inline stuff in <asm/semaphore.h>
39 * where we want to avoid any extra jumps and calls.
40 */
41
42/*
43 * Logic:
44 * - only on a boundary condition do we need to care. When we go
45 * from a negative count to a non-negative, we wake people up.
46 * - when we go from a non-negative count to a negative do we
47 * (a) synchronize with the "sleeper" count and (b) make sure
48 * that we're on the wakeup list before we synchronize so that
49 * we cannot lose wakeup events.
50 */
51
52void __up(struct semaphore *sem)
53{
54 wake_up(&sem->wait);
55}
56EXPORT_SYMBOL(__up);
57
58void __sched __down(struct semaphore *sem)
59{
60 struct task_struct *tsk = current;
61 DECLARE_WAITQUEUE(wait, tsk);
62 unsigned long flags;
63
64 tsk->state = TASK_UNINTERRUPTIBLE;
65 spin_lock_irqsave(&sem->wait.lock, flags);
66 add_wait_queue_exclusive_locked(&sem->wait, &wait);
67
68 sem->sleepers++;
69 for (;;) {
70 int sleepers = sem->sleepers;
71
72 /*
73 * Add "everybody else" into it. They aren't
74 * playing, because we own the spinlock in
75 * the wait_queue_head.
76 */
77 if (atomic_add_return(sleepers - 1, &sem->count) >= 0) {
78 sem->sleepers = 0;
79 break;
80 }
81 sem->sleepers = 1; /* us - see -1 above */
82 spin_unlock_irqrestore(&sem->wait.lock, flags);
83
84 schedule();
85
86 spin_lock_irqsave(&sem->wait.lock, flags);
87 tsk->state = TASK_UNINTERRUPTIBLE;
88 }
89 remove_wait_queue_locked(&sem->wait, &wait);
90 wake_up_locked(&sem->wait);
91 spin_unlock_irqrestore(&sem->wait.lock, flags);
92 tsk->state = TASK_RUNNING;
93}
94EXPORT_SYMBOL(__down);
95
96int __sched __down_interruptible(struct semaphore *sem)
97{
98 int retval = 0;
99 struct task_struct *tsk = current;
100 DECLARE_WAITQUEUE(wait, tsk);
101 unsigned long flags;
102
103 tsk->state = TASK_INTERRUPTIBLE;
104 spin_lock_irqsave(&sem->wait.lock, flags);
105 add_wait_queue_exclusive_locked(&sem->wait, &wait);
106
107 sem->sleepers++;
108 for (;;) {
109 int sleepers = sem->sleepers;
110
111 /*
112 * With signals pending, this turns into the trylock
113 * failure case - we won't be sleeping, and we can't
114 * get the lock as it has contention. Just correct the
115 * count and exit.
116 */
117 if (signal_pending(current)) {
118 retval = -EINTR;
119 sem->sleepers = 0;
120 atomic_add(sleepers, &sem->count);
121 break;
122 }
123
124 /*
125 * Add "everybody else" into it. They aren't
126 * playing, because we own the spinlock in
127 * the wait_queue_head.
128 */
129 if (atomic_add_return(sleepers - 1, &sem->count) >= 0) {
130 sem->sleepers = 0;
131 break;
132 }
133 sem->sleepers = 1; /* us - see -1 above */
134 spin_unlock_irqrestore(&sem->wait.lock, flags);
135
136 schedule();
137
138 spin_lock_irqsave(&sem->wait.lock, flags);
139 tsk->state = TASK_INTERRUPTIBLE;
140 }
141 remove_wait_queue_locked(&sem->wait, &wait);
142 wake_up_locked(&sem->wait);
143 spin_unlock_irqrestore(&sem->wait.lock, flags);
144
145 tsk->state = TASK_RUNNING;
146 return retval;
147}
148EXPORT_SYMBOL(__down_interruptible);
diff --git a/arch/blackfin/Kconfig b/arch/blackfin/Kconfig
index 589c6aca4803..2dd1f300a5cf 100644
--- a/arch/blackfin/Kconfig
+++ b/arch/blackfin/Kconfig
@@ -31,10 +31,6 @@ config ZONE_DMA
31 bool 31 bool
32 default y 32 default y
33 33
34config SEMAPHORE_SLEEPERS
35 bool
36 default y
37
38config GENERIC_FIND_NEXT_BIT 34config GENERIC_FIND_NEXT_BIT
39 bool 35 bool
40 default y 36 default y
diff --git a/arch/blackfin/kernel/bfin_ksyms.c b/arch/blackfin/kernel/bfin_ksyms.c
index 0bfbb269e350..053edff6c0d8 100644
--- a/arch/blackfin/kernel/bfin_ksyms.c
+++ b/arch/blackfin/kernel/bfin_ksyms.c
@@ -42,11 +42,6 @@ EXPORT_SYMBOL(ip_fast_csum);
42 42
43EXPORT_SYMBOL(kernel_thread); 43EXPORT_SYMBOL(kernel_thread);
44 44
45EXPORT_SYMBOL(__up);
46EXPORT_SYMBOL(__down);
47EXPORT_SYMBOL(__down_trylock);
48EXPORT_SYMBOL(__down_interruptible);
49
50EXPORT_SYMBOL(is_in_rom); 45EXPORT_SYMBOL(is_in_rom);
51EXPORT_SYMBOL(bfin_return_from_exception); 46EXPORT_SYMBOL(bfin_return_from_exception);
52 47
diff --git a/arch/cris/kernel/Makefile b/arch/cris/kernel/Makefile
index c8e8ea570989..ee7bcd4d20b2 100644
--- a/arch/cris/kernel/Makefile
+++ b/arch/cris/kernel/Makefile
@@ -5,8 +5,7 @@
5 5
6extra-y := vmlinux.lds 6extra-y := vmlinux.lds
7 7
8obj-y := process.o traps.o irq.o ptrace.o setup.o \ 8obj-y := process.o traps.o irq.o ptrace.o setup.o time.o sys_cris.o
9 time.o sys_cris.o semaphore.o
10 9
11obj-$(CONFIG_MODULES) += crisksyms.o 10obj-$(CONFIG_MODULES) += crisksyms.o
12obj-$(CONFIG_MODULES) += module.o 11obj-$(CONFIG_MODULES) += module.o
diff --git a/arch/cris/kernel/crisksyms.c b/arch/cris/kernel/crisksyms.c
index 62f0e752915a..7ac000f6a888 100644
--- a/arch/cris/kernel/crisksyms.c
+++ b/arch/cris/kernel/crisksyms.c
@@ -9,7 +9,6 @@
9#include <linux/string.h> 9#include <linux/string.h>
10#include <linux/tty.h> 10#include <linux/tty.h>
11 11
12#include <asm/semaphore.h>
13#include <asm/processor.h> 12#include <asm/processor.h>
14#include <asm/uaccess.h> 13#include <asm/uaccess.h>
15#include <asm/checksum.h> 14#include <asm/checksum.h>
@@ -49,12 +48,6 @@ EXPORT_SYMBOL(__negdi2);
49EXPORT_SYMBOL(__ioremap); 48EXPORT_SYMBOL(__ioremap);
50EXPORT_SYMBOL(iounmap); 49EXPORT_SYMBOL(iounmap);
51 50
52/* Semaphore functions */
53EXPORT_SYMBOL(__up);
54EXPORT_SYMBOL(__down);
55EXPORT_SYMBOL(__down_interruptible);
56EXPORT_SYMBOL(__down_trylock);
57
58/* Userspace access functions */ 51/* Userspace access functions */
59EXPORT_SYMBOL(__copy_user_zeroing); 52EXPORT_SYMBOL(__copy_user_zeroing);
60EXPORT_SYMBOL(__copy_user); 53EXPORT_SYMBOL(__copy_user);
diff --git a/arch/cris/kernel/semaphore.c b/arch/cris/kernel/semaphore.c
deleted file mode 100644
index f137a439041f..000000000000
--- a/arch/cris/kernel/semaphore.c
+++ /dev/null
@@ -1,129 +0,0 @@
1/*
2 * Generic semaphore code. Buyer beware. Do your own
3 * specific changes in <asm/semaphore-helper.h>
4 */
5
6#include <linux/sched.h>
7#include <asm/semaphore-helper.h>
8
9/*
10 * Semaphores are implemented using a two-way counter:
11 * The "count" variable is decremented for each process
12 * that tries to sleep, while the "waking" variable is
13 * incremented when the "up()" code goes to wake up waiting
14 * processes.
15 *
16 * Notably, the inline "up()" and "down()" functions can
17 * efficiently test if they need to do any extra work (up
18 * needs to do something only if count was negative before
19 * the increment operation.
20 *
21 * waking_non_zero() (from asm/semaphore.h) must execute
22 * atomically.
23 *
24 * When __up() is called, the count was negative before
25 * incrementing it, and we need to wake up somebody.
26 *
27 * This routine adds one to the count of processes that need to
28 * wake up and exit. ALL waiting processes actually wake up but
29 * only the one that gets to the "waking" field first will gate
30 * through and acquire the semaphore. The others will go back
31 * to sleep.
32 *
33 * Note that these functions are only called when there is
34 * contention on the lock, and as such all this is the
35 * "non-critical" part of the whole semaphore business. The
36 * critical part is the inline stuff in <asm/semaphore.h>
37 * where we want to avoid any extra jumps and calls.
38 */
39void __up(struct semaphore *sem)
40{
41 wake_one_more(sem);
42 wake_up(&sem->wait);
43}
44
45/*
46 * Perform the "down" function. Return zero for semaphore acquired,
47 * return negative for signalled out of the function.
48 *
49 * If called from __down, the return is ignored and the wait loop is
50 * not interruptible. This means that a task waiting on a semaphore
51 * using "down()" cannot be killed until someone does an "up()" on
52 * the semaphore.
53 *
54 * If called from __down_interruptible, the return value gets checked
55 * upon return. If the return value is negative then the task continues
56 * with the negative value in the return register (it can be tested by
57 * the caller).
58 *
59 * Either form may be used in conjunction with "up()".
60 *
61 */
62
63#define DOWN_VAR \
64 struct task_struct *tsk = current; \
65 wait_queue_t wait; \
66 init_waitqueue_entry(&wait, tsk);
67
68#define DOWN_HEAD(task_state) \
69 \
70 \
71 tsk->state = (task_state); \
72 add_wait_queue(&sem->wait, &wait); \
73 \
74 /* \
75 * Ok, we're set up. sem->count is known to be less than zero \
76 * so we must wait. \
77 * \
78 * We can let go the lock for purposes of waiting. \
79 * We re-acquire it after awaking so as to protect \
80 * all semaphore operations. \
81 * \
82 * If "up()" is called before we call waking_non_zero() then \
83 * we will catch it right away. If it is called later then \
84 * we will have to go through a wakeup cycle to catch it. \
85 * \
86 * Multiple waiters contend for the semaphore lock to see \
87 * who gets to gate through and who has to wait some more. \
88 */ \
89 for (;;) {
90
91#define DOWN_TAIL(task_state) \
92 tsk->state = (task_state); \
93 } \
94 tsk->state = TASK_RUNNING; \
95 remove_wait_queue(&sem->wait, &wait);
96
97void __sched __down(struct semaphore * sem)
98{
99 DOWN_VAR
100 DOWN_HEAD(TASK_UNINTERRUPTIBLE)
101 if (waking_non_zero(sem))
102 break;
103 schedule();
104 DOWN_TAIL(TASK_UNINTERRUPTIBLE)
105}
106
107int __sched __down_interruptible(struct semaphore * sem)
108{
109 int ret = 0;
110 DOWN_VAR
111 DOWN_HEAD(TASK_INTERRUPTIBLE)
112
113 ret = waking_non_zero_interruptible(sem, tsk);
114 if (ret)
115 {
116 if (ret == 1)
117 /* ret != 0 only if we get interrupted -arca */
118 ret = 0;
119 break;
120 }
121 schedule();
122 DOWN_TAIL(TASK_INTERRUPTIBLE)
123 return ret;
124}
125
126int __down_trylock(struct semaphore * sem)
127{
128 return waking_non_zero_trylock(sem);
129}
diff --git a/arch/frv/kernel/Makefile b/arch/frv/kernel/Makefile
index e8f73ed28b52..c36f70b6699a 100644
--- a/arch/frv/kernel/Makefile
+++ b/arch/frv/kernel/Makefile
@@ -9,7 +9,7 @@ extra-y:= head.o init_task.o vmlinux.lds
9 9
10obj-y := $(heads-y) entry.o entry-table.o break.o switch_to.o kernel_thread.o \ 10obj-y := $(heads-y) entry.o entry-table.o break.o switch_to.o kernel_thread.o \
11 kernel_execve.o process.o traps.o ptrace.o signal.o dma.o \ 11 kernel_execve.o process.o traps.o ptrace.o signal.o dma.o \
12 sys_frv.o time.o semaphore.o setup.o frv_ksyms.o \ 12 sys_frv.o time.o setup.o frv_ksyms.o \
13 debug-stub.o irq.o sleep.o uaccess.o 13 debug-stub.o irq.o sleep.o uaccess.o
14 14
15obj-$(CONFIG_GDBSTUB) += gdb-stub.o gdb-io.o 15obj-$(CONFIG_GDBSTUB) += gdb-stub.o gdb-io.o
diff --git a/arch/frv/kernel/frv_ksyms.c b/arch/frv/kernel/frv_ksyms.c
index f772704b3d28..0316b3c50eff 100644
--- a/arch/frv/kernel/frv_ksyms.c
+++ b/arch/frv/kernel/frv_ksyms.c
@@ -12,7 +12,6 @@
12#include <asm/pgalloc.h> 12#include <asm/pgalloc.h>
13#include <asm/irq.h> 13#include <asm/irq.h>
14#include <asm/io.h> 14#include <asm/io.h>
15#include <asm/semaphore.h>
16#include <asm/checksum.h> 15#include <asm/checksum.h>
17#include <asm/hardirq.h> 16#include <asm/hardirq.h>
18#include <asm/cacheflush.h> 17#include <asm/cacheflush.h>
diff --git a/arch/frv/kernel/semaphore.c b/arch/frv/kernel/semaphore.c
deleted file mode 100644
index 7ee3a147b471..000000000000
--- a/arch/frv/kernel/semaphore.c
+++ /dev/null
@@ -1,155 +0,0 @@
1/* semaphore.c: FR-V semaphores
2 *
3 * Copyright (C) 2003 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
5 * - Derived from lib/rwsem-spinlock.c
6 *
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License
9 * as published by the Free Software Foundation; either version
10 * 2 of the License, or (at your option) any later version.
11 */
12
13#include <linux/sched.h>
14#include <linux/module.h>
15#include <asm/semaphore.h>
16
17struct sem_waiter {
18 struct list_head list;
19 struct task_struct *task;
20};
21
22#ifdef CONFIG_DEBUG_SEMAPHORE
23void semtrace(struct semaphore *sem, const char *str)
24{
25 if (sem->debug)
26 printk("[%d] %s({%d,%d})\n",
27 current->pid,
28 str,
29 sem->counter,
30 list_empty(&sem->wait_list) ? 0 : 1);
31}
32#else
33#define semtrace(SEM,STR) do { } while(0)
34#endif
35
36/*
37 * wait for a token to be granted from a semaphore
38 * - entered with lock held and interrupts disabled
39 */
40void __down(struct semaphore *sem, unsigned long flags)
41{
42 struct task_struct *tsk = current;
43 struct sem_waiter waiter;
44
45 semtrace(sem, "Entering __down");
46
47 /* set up my own style of waitqueue */
48 waiter.task = tsk;
49 get_task_struct(tsk);
50
51 list_add_tail(&waiter.list, &sem->wait_list);
52
53 /* we don't need to touch the semaphore struct anymore */
54 spin_unlock_irqrestore(&sem->wait_lock, flags);
55
56 /* wait to be given the semaphore */
57 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
58
59 for (;;) {
60 if (list_empty(&waiter.list))
61 break;
62 schedule();
63 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
64 }
65
66 tsk->state = TASK_RUNNING;
67 semtrace(sem, "Leaving __down");
68}
69
70EXPORT_SYMBOL(__down);
71
72/*
73 * interruptibly wait for a token to be granted from a semaphore
74 * - entered with lock held and interrupts disabled
75 */
76int __down_interruptible(struct semaphore *sem, unsigned long flags)
77{
78 struct task_struct *tsk = current;
79 struct sem_waiter waiter;
80 int ret;
81
82 semtrace(sem,"Entering __down_interruptible");
83
84 /* set up my own style of waitqueue */
85 waiter.task = tsk;
86 get_task_struct(tsk);
87
88 list_add_tail(&waiter.list, &sem->wait_list);
89
90 /* we don't need to touch the semaphore struct anymore */
91 set_task_state(tsk, TASK_INTERRUPTIBLE);
92
93 spin_unlock_irqrestore(&sem->wait_lock, flags);
94
95 /* wait to be given the semaphore */
96 ret = 0;
97 for (;;) {
98 if (list_empty(&waiter.list))
99 break;
100 if (unlikely(signal_pending(current)))
101 goto interrupted;
102 schedule();
103 set_task_state(tsk, TASK_INTERRUPTIBLE);
104 }
105
106 out:
107 tsk->state = TASK_RUNNING;
108 semtrace(sem, "Leaving __down_interruptible");
109 return ret;
110
111 interrupted:
112 spin_lock_irqsave(&sem->wait_lock, flags);
113
114 if (!list_empty(&waiter.list)) {
115 list_del(&waiter.list);
116 ret = -EINTR;
117 }
118
119 spin_unlock_irqrestore(&sem->wait_lock, flags);
120 if (ret == -EINTR)
121 put_task_struct(current);
122 goto out;
123}
124
125EXPORT_SYMBOL(__down_interruptible);
126
127/*
128 * release a single token back to a semaphore
129 * - entered with lock held and interrupts disabled
130 */
131void __up(struct semaphore *sem)
132{
133 struct task_struct *tsk;
134 struct sem_waiter *waiter;
135
136 semtrace(sem,"Entering __up");
137
138 /* grant the token to the process at the front of the queue */
139 waiter = list_entry(sem->wait_list.next, struct sem_waiter, list);
140
141 /* We must be careful not to touch 'waiter' after we set ->task = NULL.
142 * It is allocated on the waiter's stack and may become invalid at
143 * any time after that point (due to a wakeup from another source).
144 */
145 list_del_init(&waiter->list);
146 tsk = waiter->task;
147 mb();
148 waiter->task = NULL;
149 wake_up_process(tsk);
150 put_task_struct(tsk);
151
152 semtrace(sem,"Leaving __up");
153}
154
155EXPORT_SYMBOL(__up);
diff --git a/arch/h8300/kernel/Makefile b/arch/h8300/kernel/Makefile
index 874f6aefee65..6c248c3c5c3b 100644
--- a/arch/h8300/kernel/Makefile
+++ b/arch/h8300/kernel/Makefile
@@ -5,7 +5,7 @@
5extra-y := vmlinux.lds 5extra-y := vmlinux.lds
6 6
7obj-y := process.o traps.o ptrace.o irq.o \ 7obj-y := process.o traps.o ptrace.o irq.o \
8 sys_h8300.o time.o semaphore.o signal.o \ 8 sys_h8300.o time.o signal.o \
9 setup.o gpio.o init_task.o syscalls.o \ 9 setup.o gpio.o init_task.o syscalls.o \
10 entry.o 10 entry.o
11 11
diff --git a/arch/h8300/kernel/h8300_ksyms.c b/arch/h8300/kernel/h8300_ksyms.c
index d1b15267ac81..6866bd9c7fb4 100644
--- a/arch/h8300/kernel/h8300_ksyms.c
+++ b/arch/h8300/kernel/h8300_ksyms.c
@@ -12,7 +12,6 @@
12#include <asm/pgalloc.h> 12#include <asm/pgalloc.h>
13#include <asm/irq.h> 13#include <asm/irq.h>
14#include <asm/io.h> 14#include <asm/io.h>
15#include <asm/semaphore.h>
16#include <asm/checksum.h> 15#include <asm/checksum.h>
17#include <asm/current.h> 16#include <asm/current.h>
18#include <asm/gpio.h> 17#include <asm/gpio.h>
diff --git a/arch/h8300/kernel/semaphore.c b/arch/h8300/kernel/semaphore.c
deleted file mode 100644
index d12cbbfe6ebd..000000000000
--- a/arch/h8300/kernel/semaphore.c
+++ /dev/null
@@ -1,132 +0,0 @@
1/*
2 * Generic semaphore code. Buyer beware. Do your own
3 * specific changes in <asm/semaphore-helper.h>
4 */
5
6#include <linux/sched.h>
7#include <linux/init.h>
8#include <asm/semaphore-helper.h>
9
10#ifndef CONFIG_RMW_INSNS
11spinlock_t semaphore_wake_lock;
12#endif
13
14/*
15 * Semaphores are implemented using a two-way counter:
16 * The "count" variable is decremented for each process
17 * that tries to sleep, while the "waking" variable is
18 * incremented when the "up()" code goes to wake up waiting
19 * processes.
20 *
21 * Notably, the inline "up()" and "down()" functions can
22 * efficiently test if they need to do any extra work (up
23 * needs to do something only if count was negative before
24 * the increment operation.
25 *
26 * waking_non_zero() (from asm/semaphore.h) must execute
27 * atomically.
28 *
29 * When __up() is called, the count was negative before
30 * incrementing it, and we need to wake up somebody.
31 *
32 * This routine adds one to the count of processes that need to
33 * wake up and exit. ALL waiting processes actually wake up but
34 * only the one that gets to the "waking" field first will gate
35 * through and acquire the semaphore. The others will go back
36 * to sleep.
37 *
38 * Note that these functions are only called when there is
39 * contention on the lock, and as such all this is the
40 * "non-critical" part of the whole semaphore business. The
41 * critical part is the inline stuff in <asm/semaphore.h>
42 * where we want to avoid any extra jumps and calls.
43 */
44void __up(struct semaphore *sem)
45{
46 wake_one_more(sem);
47 wake_up(&sem->wait);
48}
49
50/*
51 * Perform the "down" function. Return zero for semaphore acquired,
52 * return negative for signalled out of the function.
53 *
54 * If called from __down, the return is ignored and the wait loop is
55 * not interruptible. This means that a task waiting on a semaphore
56 * using "down()" cannot be killed until someone does an "up()" on
57 * the semaphore.
58 *
59 * If called from __down_interruptible, the return value gets checked
60 * upon return. If the return value is negative then the task continues
61 * with the negative value in the return register (it can be tested by
62 * the caller).
63 *
64 * Either form may be used in conjunction with "up()".
65 *
66 */
67
68
69#define DOWN_HEAD(task_state) \
70 \
71 \
72 current->state = (task_state); \
73 add_wait_queue(&sem->wait, &wait); \
74 \
75 /* \
76 * Ok, we're set up. sem->count is known to be less than zero \
77 * so we must wait. \
78 * \
79 * We can let go the lock for purposes of waiting. \
80 * We re-acquire it after awaking so as to protect \
81 * all semaphore operations. \
82 * \
83 * If "up()" is called before we call waking_non_zero() then \
84 * we will catch it right away. If it is called later then \
85 * we will have to go through a wakeup cycle to catch it. \
86 * \
87 * Multiple waiters contend for the semaphore lock to see \
88 * who gets to gate through and who has to wait some more. \
89 */ \
90 for (;;) {
91
92#define DOWN_TAIL(task_state) \
93 current->state = (task_state); \
94 } \
95 current->state = TASK_RUNNING; \
96 remove_wait_queue(&sem->wait, &wait);
97
98void __sched __down(struct semaphore * sem)
99{
100 DECLARE_WAITQUEUE(wait, current);
101
102 DOWN_HEAD(TASK_UNINTERRUPTIBLE)
103 if (waking_non_zero(sem))
104 break;
105 schedule();
106 DOWN_TAIL(TASK_UNINTERRUPTIBLE)
107}
108
109int __sched __down_interruptible(struct semaphore * sem)
110{
111 DECLARE_WAITQUEUE(wait, current);
112 int ret = 0;
113
114 DOWN_HEAD(TASK_INTERRUPTIBLE)
115
116 ret = waking_non_zero_interruptible(sem, current);
117 if (ret)
118 {
119 if (ret == 1)
120 /* ret != 0 only if we get interrupted -arca */
121 ret = 0;
122 break;
123 }
124 schedule();
125 DOWN_TAIL(TASK_INTERRUPTIBLE)
126 return ret;
127}
128
129int __down_trylock(struct semaphore * sem)
130{
131 return waking_non_zero_trylock(sem);
132}
diff --git a/arch/ia64/kernel/Makefile b/arch/ia64/kernel/Makefile
index 33e5a598672d..13fd10e8699e 100644
--- a/arch/ia64/kernel/Makefile
+++ b/arch/ia64/kernel/Makefile
@@ -6,7 +6,7 @@ extra-y := head.o init_task.o vmlinux.lds
6 6
7obj-y := acpi.o entry.o efi.o efi_stub.o gate-data.o fsys.o ia64_ksyms.o irq.o irq_ia64.o \ 7obj-y := acpi.o entry.o efi.o efi_stub.o gate-data.o fsys.o ia64_ksyms.o irq.o irq_ia64.o \
8 irq_lsapic.o ivt.o machvec.o pal.o patch.o process.o perfmon.o ptrace.o sal.o \ 8 irq_lsapic.o ivt.o machvec.o pal.o patch.o process.o perfmon.o ptrace.o sal.o \
9 salinfo.o semaphore.o setup.o signal.o sys_ia64.o time.o traps.o unaligned.o \ 9 salinfo.o setup.o signal.o sys_ia64.o time.o traps.o unaligned.o \
10 unwind.o mca.o mca_asm.o topology.o 10 unwind.o mca.o mca_asm.o topology.o
11 11
12obj-$(CONFIG_IA64_BRL_EMU) += brl_emu.o 12obj-$(CONFIG_IA64_BRL_EMU) += brl_emu.o
diff --git a/arch/ia64/kernel/ia64_ksyms.c b/arch/ia64/kernel/ia64_ksyms.c
index 8e7193d55528..6da1f20d7372 100644
--- a/arch/ia64/kernel/ia64_ksyms.c
+++ b/arch/ia64/kernel/ia64_ksyms.c
@@ -19,12 +19,6 @@ EXPORT_SYMBOL_GPL(empty_zero_page);
19EXPORT_SYMBOL(ip_fast_csum); /* hand-coded assembly */ 19EXPORT_SYMBOL(ip_fast_csum); /* hand-coded assembly */
20EXPORT_SYMBOL(csum_ipv6_magic); 20EXPORT_SYMBOL(csum_ipv6_magic);
21 21
22#include <asm/semaphore.h>
23EXPORT_SYMBOL(__down);
24EXPORT_SYMBOL(__down_interruptible);
25EXPORT_SYMBOL(__down_trylock);
26EXPORT_SYMBOL(__up);
27
28#include <asm/page.h> 22#include <asm/page.h>
29EXPORT_SYMBOL(clear_page); 23EXPORT_SYMBOL(clear_page);
30 24
diff --git a/arch/ia64/kernel/semaphore.c b/arch/ia64/kernel/semaphore.c
deleted file mode 100644
index 2724ef3fbae2..000000000000
--- a/arch/ia64/kernel/semaphore.c
+++ /dev/null
@@ -1,165 +0,0 @@
1/*
2 * IA-64 semaphore implementation (derived from x86 version).
3 *
4 * Copyright (C) 1999-2000, 2002 Hewlett-Packard Co
5 * David Mosberger-Tang <davidm@hpl.hp.com>
6 */
7
8/*
9 * Semaphores are implemented using a two-way counter: The "count"
10 * variable is decremented for each process that tries to acquire the
11 * semaphore, while the "sleepers" variable is a count of such
12 * acquires.
13 *
14 * Notably, the inline "up()" and "down()" functions can efficiently
15 * test if they need to do any extra work (up needs to do something
16 * only if count was negative before the increment operation.
17 *
18 * "sleeping" and the contention routine ordering is protected
19 * by the spinlock in the semaphore's waitqueue head.
20 *
21 * Note that these functions are only called when there is contention
22 * on the lock, and as such all this is the "non-critical" part of the
23 * whole semaphore business. The critical part is the inline stuff in
24 * <asm/semaphore.h> where we want to avoid any extra jumps and calls.
25 */
26#include <linux/sched.h>
27#include <linux/init.h>
28
29#include <asm/errno.h>
30#include <asm/semaphore.h>
31
32/*
33 * Logic:
34 * - Only on a boundary condition do we need to care. When we go
35 * from a negative count to a non-negative, we wake people up.
36 * - When we go from a non-negative count to a negative do we
37 * (a) synchronize with the "sleepers" count and (b) make sure
38 * that we're on the wakeup list before we synchronize so that
39 * we cannot lose wakeup events.
40 */
41
42void
43__up (struct semaphore *sem)
44{
45 wake_up(&sem->wait);
46}
47
48void __sched __down (struct semaphore *sem)
49{
50 struct task_struct *tsk = current;
51 DECLARE_WAITQUEUE(wait, tsk);
52 unsigned long flags;
53
54 tsk->state = TASK_UNINTERRUPTIBLE;
55 spin_lock_irqsave(&sem->wait.lock, flags);
56 add_wait_queue_exclusive_locked(&sem->wait, &wait);
57
58 sem->sleepers++;
59 for (;;) {
60 int sleepers = sem->sleepers;
61
62 /*
63 * Add "everybody else" into it. They aren't
64 * playing, because we own the spinlock in
65 * the wait_queue_head.
66 */
67 if (!atomic_add_negative(sleepers - 1, &sem->count)) {
68 sem->sleepers = 0;
69 break;
70 }
71 sem->sleepers = 1; /* us - see -1 above */
72 spin_unlock_irqrestore(&sem->wait.lock, flags);
73
74 schedule();
75
76 spin_lock_irqsave(&sem->wait.lock, flags);
77 tsk->state = TASK_UNINTERRUPTIBLE;
78 }
79 remove_wait_queue_locked(&sem->wait, &wait);
80 wake_up_locked(&sem->wait);
81 spin_unlock_irqrestore(&sem->wait.lock, flags);
82 tsk->state = TASK_RUNNING;
83}
84
85int __sched __down_interruptible (struct semaphore * sem)
86{
87 int retval = 0;
88 struct task_struct *tsk = current;
89 DECLARE_WAITQUEUE(wait, tsk);
90 unsigned long flags;
91
92 tsk->state = TASK_INTERRUPTIBLE;
93 spin_lock_irqsave(&sem->wait.lock, flags);
94 add_wait_queue_exclusive_locked(&sem->wait, &wait);
95
96 sem->sleepers ++;
97 for (;;) {
98 int sleepers = sem->sleepers;
99
100 /*
101 * With signals pending, this turns into
102 * the trylock failure case - we won't be
103 * sleeping, and we* can't get the lock as
104 * it has contention. Just correct the count
105 * and exit.
106 */
107 if (signal_pending(current)) {
108 retval = -EINTR;
109 sem->sleepers = 0;
110 atomic_add(sleepers, &sem->count);
111 break;
112 }
113
114 /*
115 * Add "everybody else" into it. They aren't
116 * playing, because we own the spinlock in
117 * wait_queue_head. The "-1" is because we're
118 * still hoping to get the semaphore.
119 */
120 if (!atomic_add_negative(sleepers - 1, &sem->count)) {
121 sem->sleepers = 0;
122 break;
123 }
124 sem->sleepers = 1; /* us - see -1 above */
125 spin_unlock_irqrestore(&sem->wait.lock, flags);
126
127 schedule();
128
129 spin_lock_irqsave(&sem->wait.lock, flags);
130 tsk->state = TASK_INTERRUPTIBLE;
131 }
132 remove_wait_queue_locked(&sem->wait, &wait);
133 wake_up_locked(&sem->wait);
134 spin_unlock_irqrestore(&sem->wait.lock, flags);
135
136 tsk->state = TASK_RUNNING;
137 return retval;
138}
139
140/*
141 * Trylock failed - make sure we correct for having decremented the
142 * count.
143 */
144int
145__down_trylock (struct semaphore *sem)
146{
147 unsigned long flags;
148 int sleepers;
149
150 spin_lock_irqsave(&sem->wait.lock, flags);
151 sleepers = sem->sleepers + 1;
152 sem->sleepers = 0;
153
154 /*
155 * Add "everybody else" and us into it. They aren't
156 * playing, because we own the spinlock in the
157 * wait_queue_head.
158 */
159 if (!atomic_add_negative(sleepers, &sem->count)) {
160 wake_up_locked(&sem->wait);
161 }
162
163 spin_unlock_irqrestore(&sem->wait.lock, flags);
164 return 1;
165}
diff --git a/arch/m32r/kernel/Makefile b/arch/m32r/kernel/Makefile
index e97e26e87c9e..09200d4886e3 100644
--- a/arch/m32r/kernel/Makefile
+++ b/arch/m32r/kernel/Makefile
@@ -5,7 +5,7 @@
5extra-y := head.o init_task.o vmlinux.lds 5extra-y := head.o init_task.o vmlinux.lds
6 6
7obj-y := process.o entry.o traps.o align.o irq.o setup.o time.o \ 7obj-y := process.o entry.o traps.o align.o irq.o setup.o time.o \
8 m32r_ksyms.o sys_m32r.o semaphore.o signal.o ptrace.o 8 m32r_ksyms.o sys_m32r.o signal.o ptrace.o
9 9
10obj-$(CONFIG_SMP) += smp.o smpboot.o 10obj-$(CONFIG_SMP) += smp.o smpboot.o
11obj-$(CONFIG_MODULES) += module.o 11obj-$(CONFIG_MODULES) += module.o
diff --git a/arch/m32r/kernel/m32r_ksyms.c b/arch/m32r/kernel/m32r_ksyms.c
index 41a4c95e06d6..e6709fe950ba 100644
--- a/arch/m32r/kernel/m32r_ksyms.c
+++ b/arch/m32r/kernel/m32r_ksyms.c
@@ -7,7 +7,6 @@
7#include <linux/interrupt.h> 7#include <linux/interrupt.h>
8#include <linux/string.h> 8#include <linux/string.h>
9 9
10#include <asm/semaphore.h>
11#include <asm/processor.h> 10#include <asm/processor.h>
12#include <asm/uaccess.h> 11#include <asm/uaccess.h>
13#include <asm/checksum.h> 12#include <asm/checksum.h>
@@ -22,10 +21,6 @@ EXPORT_SYMBOL(dump_fpu);
22EXPORT_SYMBOL(__ioremap); 21EXPORT_SYMBOL(__ioremap);
23EXPORT_SYMBOL(iounmap); 22EXPORT_SYMBOL(iounmap);
24EXPORT_SYMBOL(kernel_thread); 23EXPORT_SYMBOL(kernel_thread);
25EXPORT_SYMBOL(__down);
26EXPORT_SYMBOL(__down_interruptible);
27EXPORT_SYMBOL(__up);
28EXPORT_SYMBOL(__down_trylock);
29 24
30/* Networking helper routines. */ 25/* Networking helper routines. */
31/* Delay loops */ 26/* Delay loops */
diff --git a/arch/m32r/kernel/semaphore.c b/arch/m32r/kernel/semaphore.c
deleted file mode 100644
index 940c2d37cfd1..000000000000
--- a/arch/m32r/kernel/semaphore.c
+++ /dev/null
@@ -1,185 +0,0 @@
1/*
2 * linux/arch/m32r/semaphore.c
3 * orig : i386 2.6.4
4 *
5 * M32R semaphore implementation.
6 *
7 * Copyright (c) 2002 - 2004 Hitoshi Yamamoto
8 */
9
10/*
11 * i386 semaphore implementation.
12 *
13 * (C) Copyright 1999 Linus Torvalds
14 *
15 * Portions Copyright 1999 Red Hat, Inc.
16 *
17 * This program is free software; you can redistribute it and/or
18 * modify it under the terms of the GNU General Public License
19 * as published by the Free Software Foundation; either version
20 * 2 of the License, or (at your option) any later version.
21 *
22 * rw semaphores implemented November 1999 by Benjamin LaHaise <bcrl@kvack.org>
23 */
24#include <linux/sched.h>
25#include <linux/err.h>
26#include <linux/init.h>
27#include <asm/semaphore.h>
28
29/*
30 * Semaphores are implemented using a two-way counter:
31 * The "count" variable is decremented for each process
32 * that tries to acquire the semaphore, while the "sleeping"
33 * variable is a count of such acquires.
34 *
35 * Notably, the inline "up()" and "down()" functions can
36 * efficiently test if they need to do any extra work (up
37 * needs to do something only if count was negative before
38 * the increment operation.
39 *
40 * "sleeping" and the contention routine ordering is protected
41 * by the spinlock in the semaphore's waitqueue head.
42 *
43 * Note that these functions are only called when there is
44 * contention on the lock, and as such all this is the
45 * "non-critical" part of the whole semaphore business. The
46 * critical part is the inline stuff in <asm/semaphore.h>
47 * where we want to avoid any extra jumps and calls.
48 */
49
50/*
51 * Logic:
52 * - only on a boundary condition do we need to care. When we go
53 * from a negative count to a non-negative, we wake people up.
54 * - when we go from a non-negative count to a negative do we
55 * (a) synchronize with the "sleeper" count and (b) make sure
56 * that we're on the wakeup list before we synchronize so that
57 * we cannot lose wakeup events.
58 */
59
60asmlinkage void __up(struct semaphore *sem)
61{
62 wake_up(&sem->wait);
63}
64
65asmlinkage void __sched __down(struct semaphore * sem)
66{
67 struct task_struct *tsk = current;
68 DECLARE_WAITQUEUE(wait, tsk);
69 unsigned long flags;
70
71 tsk->state = TASK_UNINTERRUPTIBLE;
72 spin_lock_irqsave(&sem->wait.lock, flags);
73 add_wait_queue_exclusive_locked(&sem->wait, &wait);
74
75 sem->sleepers++;
76 for (;;) {
77 int sleepers = sem->sleepers;
78
79 /*
80 * Add "everybody else" into it. They aren't
81 * playing, because we own the spinlock in
82 * the wait_queue_head.
83 */
84 if (!atomic_add_negative(sleepers - 1, &sem->count)) {
85 sem->sleepers = 0;
86 break;
87 }
88 sem->sleepers = 1; /* us - see -1 above */
89 spin_unlock_irqrestore(&sem->wait.lock, flags);
90
91 schedule();
92
93 spin_lock_irqsave(&sem->wait.lock, flags);
94 tsk->state = TASK_UNINTERRUPTIBLE;
95 }
96 remove_wait_queue_locked(&sem->wait, &wait);
97 wake_up_locked(&sem->wait);
98 spin_unlock_irqrestore(&sem->wait.lock, flags);
99 tsk->state = TASK_RUNNING;
100}
101
102asmlinkage int __sched __down_interruptible(struct semaphore * sem)
103{
104 int retval = 0;
105 struct task_struct *tsk = current;
106 DECLARE_WAITQUEUE(wait, tsk);
107 unsigned long flags;
108
109 tsk->state = TASK_INTERRUPTIBLE;
110 spin_lock_irqsave(&sem->wait.lock, flags);
111 add_wait_queue_exclusive_locked(&sem->wait, &wait);
112
113 sem->sleepers++;
114 for (;;) {
115 int sleepers = sem->sleepers;
116
117 /*
118 * With signals pending, this turns into
119 * the trylock failure case - we won't be
120 * sleeping, and we* can't get the lock as
121 * it has contention. Just correct the count
122 * and exit.
123 */
124 if (signal_pending(current)) {
125 retval = -EINTR;
126 sem->sleepers = 0;
127 atomic_add(sleepers, &sem->count);
128 break;
129 }
130
131 /*
132 * Add "everybody else" into it. They aren't
133 * playing, because we own the spinlock in
134 * wait_queue_head. The "-1" is because we're
135 * still hoping to get the semaphore.
136 */
137 if (!atomic_add_negative(sleepers - 1, &sem->count)) {
138 sem->sleepers = 0;
139 break;
140 }
141 sem->sleepers = 1; /* us - see -1 above */
142 spin_unlock_irqrestore(&sem->wait.lock, flags);
143
144 schedule();
145
146 spin_lock_irqsave(&sem->wait.lock, flags);
147 tsk->state = TASK_INTERRUPTIBLE;
148 }
149 remove_wait_queue_locked(&sem->wait, &wait);
150 wake_up_locked(&sem->wait);
151 spin_unlock_irqrestore(&sem->wait.lock, flags);
152
153 tsk->state = TASK_RUNNING;
154 return retval;
155}
156
157/*
158 * Trylock failed - make sure we correct for
159 * having decremented the count.
160 *
161 * We could have done the trylock with a
162 * single "cmpxchg" without failure cases,
163 * but then it wouldn't work on a 386.
164 */
165asmlinkage int __down_trylock(struct semaphore * sem)
166{
167 int sleepers;
168 unsigned long flags;
169
170 spin_lock_irqsave(&sem->wait.lock, flags);
171 sleepers = sem->sleepers + 1;
172 sem->sleepers = 0;
173
174 /*
175 * Add "everybody else" and us into it. They aren't
176 * playing, because we own the spinlock in the
177 * wait_queue_head.
178 */
179 if (!atomic_add_negative(sleepers, &sem->count)) {
180 wake_up_locked(&sem->wait);
181 }
182
183 spin_unlock_irqrestore(&sem->wait.lock, flags);
184 return 1;
185}
diff --git a/arch/m68k/kernel/Makefile b/arch/m68k/kernel/Makefile
index a806208c7fb5..7a62a718143b 100644
--- a/arch/m68k/kernel/Makefile
+++ b/arch/m68k/kernel/Makefile
@@ -10,7 +10,7 @@ endif
10extra-y += vmlinux.lds 10extra-y += vmlinux.lds
11 11
12obj-y := entry.o process.o traps.o ints.o signal.o ptrace.o module.o \ 12obj-y := entry.o process.o traps.o ints.o signal.o ptrace.o module.o \
13 sys_m68k.o time.o semaphore.o setup.o m68k_ksyms.o devres.o 13 sys_m68k.o time.o setup.o m68k_ksyms.o devres.o
14 14
15devres-y = ../../../kernel/irq/devres.o 15devres-y = ../../../kernel/irq/devres.o
16 16
diff --git a/arch/m68k/kernel/m68k_ksyms.c b/arch/m68k/kernel/m68k_ksyms.c
index 6fc69c74fe2e..d900e77e5363 100644
--- a/arch/m68k/kernel/m68k_ksyms.c
+++ b/arch/m68k/kernel/m68k_ksyms.c
@@ -1,5 +1,4 @@
1#include <linux/module.h> 1#include <linux/module.h>
2#include <asm/semaphore.h>
3 2
4asmlinkage long long __ashldi3 (long long, int); 3asmlinkage long long __ashldi3 (long long, int);
5asmlinkage long long __ashrdi3 (long long, int); 4asmlinkage long long __ashrdi3 (long long, int);
@@ -15,8 +14,3 @@ EXPORT_SYMBOL(__ashrdi3);
15EXPORT_SYMBOL(__lshrdi3); 14EXPORT_SYMBOL(__lshrdi3);
16EXPORT_SYMBOL(__muldi3); 15EXPORT_SYMBOL(__muldi3);
17 16
18EXPORT_SYMBOL(__down_failed);
19EXPORT_SYMBOL(__down_failed_interruptible);
20EXPORT_SYMBOL(__down_failed_trylock);
21EXPORT_SYMBOL(__up_wakeup);
22
diff --git a/arch/m68k/kernel/semaphore.c b/arch/m68k/kernel/semaphore.c
deleted file mode 100644
index d12cbbfe6ebd..000000000000
--- a/arch/m68k/kernel/semaphore.c
+++ /dev/null
@@ -1,132 +0,0 @@
1/*
2 * Generic semaphore code. Buyer beware. Do your own
3 * specific changes in <asm/semaphore-helper.h>
4 */
5
6#include <linux/sched.h>
7#include <linux/init.h>
8#include <asm/semaphore-helper.h>
9
10#ifndef CONFIG_RMW_INSNS
11spinlock_t semaphore_wake_lock;
12#endif
13
14/*
15 * Semaphores are implemented using a two-way counter:
16 * The "count" variable is decremented for each process
17 * that tries to sleep, while the "waking" variable is
18 * incremented when the "up()" code goes to wake up waiting
19 * processes.
20 *
21 * Notably, the inline "up()" and "down()" functions can
22 * efficiently test if they need to do any extra work (up
23 * needs to do something only if count was negative before
24 * the increment operation.
25 *
26 * waking_non_zero() (from asm/semaphore.h) must execute
27 * atomically.
28 *
29 * When __up() is called, the count was negative before
30 * incrementing it, and we need to wake up somebody.
31 *
32 * This routine adds one to the count of processes that need to
33 * wake up and exit. ALL waiting processes actually wake up but
34 * only the one that gets to the "waking" field first will gate
35 * through and acquire the semaphore. The others will go back
36 * to sleep.
37 *
38 * Note that these functions are only called when there is
39 * contention on the lock, and as such all this is the
40 * "non-critical" part of the whole semaphore business. The
41 * critical part is the inline stuff in <asm/semaphore.h>
42 * where we want to avoid any extra jumps and calls.
43 */
44void __up(struct semaphore *sem)
45{
46 wake_one_more(sem);
47 wake_up(&sem->wait);
48}
49
50/*
51 * Perform the "down" function. Return zero for semaphore acquired,
52 * return negative for signalled out of the function.
53 *
54 * If called from __down, the return is ignored and the wait loop is
55 * not interruptible. This means that a task waiting on a semaphore
56 * using "down()" cannot be killed until someone does an "up()" on
57 * the semaphore.
58 *
59 * If called from __down_interruptible, the return value gets checked
60 * upon return. If the return value is negative then the task continues
61 * with the negative value in the return register (it can be tested by
62 * the caller).
63 *
64 * Either form may be used in conjunction with "up()".
65 *
66 */
67
68
69#define DOWN_HEAD(task_state) \
70 \
71 \
72 current->state = (task_state); \
73 add_wait_queue(&sem->wait, &wait); \
74 \
75 /* \
76 * Ok, we're set up. sem->count is known to be less than zero \
77 * so we must wait. \
78 * \
79 * We can let go the lock for purposes of waiting. \
80 * We re-acquire it after awaking so as to protect \
81 * all semaphore operations. \
82 * \
83 * If "up()" is called before we call waking_non_zero() then \
84 * we will catch it right away. If it is called later then \
85 * we will have to go through a wakeup cycle to catch it. \
86 * \
87 * Multiple waiters contend for the semaphore lock to see \
88 * who gets to gate through and who has to wait some more. \
89 */ \
90 for (;;) {
91
92#define DOWN_TAIL(task_state) \
93 current->state = (task_state); \
94 } \
95 current->state = TASK_RUNNING; \
96 remove_wait_queue(&sem->wait, &wait);
97
98void __sched __down(struct semaphore * sem)
99{
100 DECLARE_WAITQUEUE(wait, current);
101
102 DOWN_HEAD(TASK_UNINTERRUPTIBLE)
103 if (waking_non_zero(sem))
104 break;
105 schedule();
106 DOWN_TAIL(TASK_UNINTERRUPTIBLE)
107}
108
109int __sched __down_interruptible(struct semaphore * sem)
110{
111 DECLARE_WAITQUEUE(wait, current);
112 int ret = 0;
113
114 DOWN_HEAD(TASK_INTERRUPTIBLE)
115
116 ret = waking_non_zero_interruptible(sem, current);
117 if (ret)
118 {
119 if (ret == 1)
120 /* ret != 0 only if we get interrupted -arca */
121 ret = 0;
122 break;
123 }
124 schedule();
125 DOWN_TAIL(TASK_INTERRUPTIBLE)
126 return ret;
127}
128
129int __down_trylock(struct semaphore * sem)
130{
131 return waking_non_zero_trylock(sem);
132}
diff --git a/arch/m68k/lib/Makefile b/arch/m68k/lib/Makefile
index 6bbf19f96007..a18af095cd7c 100644
--- a/arch/m68k/lib/Makefile
+++ b/arch/m68k/lib/Makefile
@@ -5,4 +5,4 @@
5EXTRA_AFLAGS := -traditional 5EXTRA_AFLAGS := -traditional
6 6
7lib-y := ashldi3.o ashrdi3.o lshrdi3.o muldi3.o \ 7lib-y := ashldi3.o ashrdi3.o lshrdi3.o muldi3.o \
8 checksum.o string.o semaphore.o uaccess.o 8 checksum.o string.o uaccess.o
diff --git a/arch/m68k/lib/semaphore.S b/arch/m68k/lib/semaphore.S
deleted file mode 100644
index 0215624c1602..000000000000
--- a/arch/m68k/lib/semaphore.S
+++ /dev/null
@@ -1,53 +0,0 @@
1/*
2 * linux/arch/m68k/lib/semaphore.S
3 *
4 * Copyright (C) 1996 Linus Torvalds
5 *
6 * m68k version by Andreas Schwab
7 */
8
9#include <linux/linkage.h>
10#include <asm/semaphore.h>
11
12/*
13 * The semaphore operations have a special calling sequence that
14 * allow us to do a simpler in-line version of them. These routines
15 * need to convert that sequence back into the C sequence when
16 * there is contention on the semaphore.
17 */
18ENTRY(__down_failed)
19 moveml %a0/%d0/%d1,-(%sp)
20 movel %a1,-(%sp)
21 jbsr __down
22 movel (%sp)+,%a1
23 moveml (%sp)+,%a0/%d0/%d1
24 rts
25
26ENTRY(__down_failed_interruptible)
27 movel %a0,-(%sp)
28 movel %d1,-(%sp)
29 movel %a1,-(%sp)
30 jbsr __down_interruptible
31 movel (%sp)+,%a1
32 movel (%sp)+,%d1
33 movel (%sp)+,%a0
34 rts
35
36ENTRY(__down_failed_trylock)
37 movel %a0,-(%sp)
38 movel %d1,-(%sp)
39 movel %a1,-(%sp)
40 jbsr __down_trylock
41 movel (%sp)+,%a1
42 movel (%sp)+,%d1
43 movel (%sp)+,%a0
44 rts
45
46ENTRY(__up_wakeup)
47 moveml %a0/%d0/%d1,-(%sp)
48 movel %a1,-(%sp)
49 jbsr __up
50 movel (%sp)+,%a1
51 moveml (%sp)+,%a0/%d0/%d1
52 rts
53
diff --git a/arch/m68knommu/kernel/Makefile b/arch/m68knommu/kernel/Makefile
index 1524b39ad63f..f0eab3dedb5a 100644
--- a/arch/m68knommu/kernel/Makefile
+++ b/arch/m68knommu/kernel/Makefile
@@ -5,7 +5,7 @@
5extra-y := vmlinux.lds 5extra-y := vmlinux.lds
6 6
7obj-y += dma.o entry.o init_task.o irq.o m68k_ksyms.o process.o ptrace.o \ 7obj-y += dma.o entry.o init_task.o irq.o m68k_ksyms.o process.o ptrace.o \
8 semaphore.o setup.o signal.o syscalltable.o sys_m68k.o time.o traps.o 8 setup.o signal.o syscalltable.o sys_m68k.o time.o traps.o
9 9
10obj-$(CONFIG_MODULES) += module.o 10obj-$(CONFIG_MODULES) += module.o
11obj-$(CONFIG_COMEMPCI) += comempci.o 11obj-$(CONFIG_COMEMPCI) += comempci.o
diff --git a/arch/m68knommu/kernel/m68k_ksyms.c b/arch/m68knommu/kernel/m68k_ksyms.c
index 53fad1490282..39fe0a7aec32 100644
--- a/arch/m68knommu/kernel/m68k_ksyms.c
+++ b/arch/m68knommu/kernel/m68k_ksyms.c
@@ -13,7 +13,6 @@
13#include <asm/pgalloc.h> 13#include <asm/pgalloc.h>
14#include <asm/irq.h> 14#include <asm/irq.h>
15#include <asm/io.h> 15#include <asm/io.h>
16#include <asm/semaphore.h>
17#include <asm/checksum.h> 16#include <asm/checksum.h>
18#include <asm/current.h> 17#include <asm/current.h>
19 18
@@ -39,11 +38,6 @@ EXPORT_SYMBOL(csum_partial_copy_nocheck);
39EXPORT_SYMBOL(memcpy); 38EXPORT_SYMBOL(memcpy);
40EXPORT_SYMBOL(memset); 39EXPORT_SYMBOL(memset);
41 40
42EXPORT_SYMBOL(__down_failed);
43EXPORT_SYMBOL(__down_failed_interruptible);
44EXPORT_SYMBOL(__down_failed_trylock);
45EXPORT_SYMBOL(__up_wakeup);
46
47/* 41/*
48 * libgcc functions - functions that are used internally by the 42 * libgcc functions - functions that are used internally by the
49 * compiler... (prototypes are not correct though, but that 43 * compiler... (prototypes are not correct though, but that
diff --git a/arch/m68knommu/kernel/semaphore.c b/arch/m68knommu/kernel/semaphore.c
deleted file mode 100644
index bce2bc7d87c6..000000000000
--- a/arch/m68knommu/kernel/semaphore.c
+++ /dev/null
@@ -1,133 +0,0 @@
1/*
2 * Generic semaphore code. Buyer beware. Do your own
3 * specific changes in <asm/semaphore-helper.h>
4 */
5
6#include <linux/sched.h>
7#include <linux/err.h>
8#include <linux/init.h>
9#include <asm/semaphore-helper.h>
10
11#ifndef CONFIG_RMW_INSNS
12spinlock_t semaphore_wake_lock;
13#endif
14
15/*
16 * Semaphores are implemented using a two-way counter:
17 * The "count" variable is decremented for each process
18 * that tries to sleep, while the "waking" variable is
19 * incremented when the "up()" code goes to wake up waiting
20 * processes.
21 *
22 * Notably, the inline "up()" and "down()" functions can
23 * efficiently test if they need to do any extra work (up
24 * needs to do something only if count was negative before
25 * the increment operation.
26 *
27 * waking_non_zero() (from asm/semaphore.h) must execute
28 * atomically.
29 *
30 * When __up() is called, the count was negative before
31 * incrementing it, and we need to wake up somebody.
32 *
33 * This routine adds one to the count of processes that need to
34 * wake up and exit. ALL waiting processes actually wake up but
35 * only the one that gets to the "waking" field first will gate
36 * through and acquire the semaphore. The others will go back
37 * to sleep.
38 *
39 * Note that these functions are only called when there is
40 * contention on the lock, and as such all this is the
41 * "non-critical" part of the whole semaphore business. The
42 * critical part is the inline stuff in <asm/semaphore.h>
43 * where we want to avoid any extra jumps and calls.
44 */
45void __up(struct semaphore *sem)
46{
47 wake_one_more(sem);
48 wake_up(&sem->wait);
49}
50
51/*
52 * Perform the "down" function. Return zero for semaphore acquired,
53 * return negative for signalled out of the function.
54 *
55 * If called from __down, the return is ignored and the wait loop is
56 * not interruptible. This means that a task waiting on a semaphore
57 * using "down()" cannot be killed until someone does an "up()" on
58 * the semaphore.
59 *
60 * If called from __down_interruptible, the return value gets checked
61 * upon return. If the return value is negative then the task continues
62 * with the negative value in the return register (it can be tested by
63 * the caller).
64 *
65 * Either form may be used in conjunction with "up()".
66 *
67 */
68
69
70#define DOWN_HEAD(task_state) \
71 \
72 \
73 current->state = (task_state); \
74 add_wait_queue(&sem->wait, &wait); \
75 \
76 /* \
77 * Ok, we're set up. sem->count is known to be less than zero \
78 * so we must wait. \
79 * \
80 * We can let go the lock for purposes of waiting. \
81 * We re-acquire it after awaking so as to protect \
82 * all semaphore operations. \
83 * \
84 * If "up()" is called before we call waking_non_zero() then \
85 * we will catch it right away. If it is called later then \
86 * we will have to go through a wakeup cycle to catch it. \
87 * \
88 * Multiple waiters contend for the semaphore lock to see \
89 * who gets to gate through and who has to wait some more. \
90 */ \
91 for (;;) {
92
93#define DOWN_TAIL(task_state) \
94 current->state = (task_state); \
95 } \
96 current->state = TASK_RUNNING; \
97 remove_wait_queue(&sem->wait, &wait);
98
99void __sched __down(struct semaphore * sem)
100{
101 DECLARE_WAITQUEUE(wait, current);
102
103 DOWN_HEAD(TASK_UNINTERRUPTIBLE)
104 if (waking_non_zero(sem))
105 break;
106 schedule();
107 DOWN_TAIL(TASK_UNINTERRUPTIBLE)
108}
109
110int __sched __down_interruptible(struct semaphore * sem)
111{
112 DECLARE_WAITQUEUE(wait, current);
113 int ret = 0;
114
115 DOWN_HEAD(TASK_INTERRUPTIBLE)
116
117 ret = waking_non_zero_interruptible(sem, current);
118 if (ret)
119 {
120 if (ret == 1)
121 /* ret != 0 only if we get interrupted -arca */
122 ret = 0;
123 break;
124 }
125 schedule();
126 DOWN_TAIL(TASK_INTERRUPTIBLE)
127 return ret;
128}
129
130int __down_trylock(struct semaphore * sem)
131{
132 return waking_non_zero_trylock(sem);
133}
diff --git a/arch/m68knommu/lib/Makefile b/arch/m68knommu/lib/Makefile
index e051a7913987..d94d709665aa 100644
--- a/arch/m68knommu/lib/Makefile
+++ b/arch/m68knommu/lib/Makefile
@@ -4,4 +4,4 @@
4 4
5lib-y := ashldi3.o ashrdi3.o lshrdi3.o \ 5lib-y := ashldi3.o ashrdi3.o lshrdi3.o \
6 muldi3.o mulsi3.o divsi3.o udivsi3.o modsi3.o umodsi3.o \ 6 muldi3.o mulsi3.o divsi3.o udivsi3.o modsi3.o umodsi3.o \
7 checksum.o semaphore.o memcpy.o memset.o delay.o 7 checksum.o memcpy.o memset.o delay.o
diff --git a/arch/m68knommu/lib/semaphore.S b/arch/m68knommu/lib/semaphore.S
deleted file mode 100644
index 87c746034376..000000000000
--- a/arch/m68knommu/lib/semaphore.S
+++ /dev/null
@@ -1,66 +0,0 @@
1/*
2 * linux/arch/m68k/lib/semaphore.S
3 *
4 * Copyright (C) 1996 Linus Torvalds
5 *
6 * m68k version by Andreas Schwab
7 *
8 * MAR/1999 -- modified to support ColdFire (gerg@snapgear.com)
9 */
10
11#include <linux/linkage.h>
12#include <asm/semaphore.h>
13
14/*
15 * "down_failed" is called with the eventual return address
16 * in %a0, and the address of the semaphore in %a1. We need
17 * to increment the number of waiters on the semaphore,
18 * call "__down()", and then eventually return to try again.
19 */
20ENTRY(__down_failed)
21#ifdef CONFIG_COLDFIRE
22 subl #12,%sp
23 moveml %a0/%d0/%d1,(%sp)
24#else
25 moveml %a0/%d0/%d1,-(%sp)
26#endif
27 movel %a1,-(%sp)
28 jbsr __down
29 movel (%sp)+,%a1
30 movel (%sp)+,%d0
31 movel (%sp)+,%d1
32 rts
33
34ENTRY(__down_failed_interruptible)
35 movel %a0,-(%sp)
36 movel %d1,-(%sp)
37 movel %a1,-(%sp)
38 jbsr __down_interruptible
39 movel (%sp)+,%a1
40 movel (%sp)+,%d1
41 rts
42
43ENTRY(__up_wakeup)
44#ifdef CONFIG_COLDFIRE
45 subl #12,%sp
46 moveml %a0/%d0/%d1,(%sp)
47#else
48 moveml %a0/%d0/%d1,-(%sp)
49#endif
50 movel %a1,-(%sp)
51 jbsr __up
52 movel (%sp)+,%a1
53 movel (%sp)+,%d0
54 movel (%sp)+,%d1
55 rts
56
57ENTRY(__down_failed_trylock)
58 movel %a0,-(%sp)
59 movel %d1,-(%sp)
60 movel %a1,-(%sp)
61 jbsr __down_trylock
62 movel (%sp)+,%a1
63 movel (%sp)+,%d1
64 movel (%sp)+,%a0
65 rts
66
diff --git a/arch/mips/kernel/Makefile b/arch/mips/kernel/Makefile
index 9e78e1a4ca17..6fcdb6fda2e2 100644
--- a/arch/mips/kernel/Makefile
+++ b/arch/mips/kernel/Makefile
@@ -5,7 +5,7 @@
5extra-y := head.o init_task.o vmlinux.lds 5extra-y := head.o init_task.o vmlinux.lds
6 6
7obj-y += cpu-probe.o branch.o entry.o genex.o irq.o process.o \ 7obj-y += cpu-probe.o branch.o entry.o genex.o irq.o process.o \
8 ptrace.o reset.o semaphore.o setup.o signal.o syscall.o \ 8 ptrace.o reset.o setup.o signal.o syscall.o \
9 time.o topology.o traps.o unaligned.o 9 time.o topology.o traps.o unaligned.o
10 10
11obj-$(CONFIG_CEVT_BCM1480) += cevt-bcm1480.o 11obj-$(CONFIG_CEVT_BCM1480) += cevt-bcm1480.o
diff --git a/arch/mips/kernel/semaphore.c b/arch/mips/kernel/semaphore.c
deleted file mode 100644
index 1265358cdca1..000000000000
--- a/arch/mips/kernel/semaphore.c
+++ /dev/null
@@ -1,168 +0,0 @@
1/*
2 * MIPS-specific semaphore code.
3 *
4 * Copyright (C) 1999 Cort Dougan <cort@cs.nmt.edu>
5 * Copyright (C) 2004 Ralf Baechle <ralf@linux-mips.org>
6 *
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License
9 * as published by the Free Software Foundation; either version
10 * 2 of the License, or (at your option) any later version.
11 *
12 * April 2001 - Reworked by Paul Mackerras <paulus@samba.org>
13 * to eliminate the SMP races in the old version between the updates
14 * of `count' and `waking'. Now we use negative `count' values to
15 * indicate that some process(es) are waiting for the semaphore.
16 */
17
18#include <linux/module.h>
19#include <linux/sched.h>
20#include <linux/init.h>
21#include <asm/atomic.h>
22#include <asm/cpu-features.h>
23#include <asm/errno.h>
24#include <asm/semaphore.h>
25#include <asm/war.h>
26/*
27 * Atomically update sem->count.
28 * This does the equivalent of the following:
29 *
30 * old_count = sem->count;
31 * tmp = MAX(old_count, 0) + incr;
32 * sem->count = tmp;
33 * return old_count;
34 *
35 * On machines without lld/scd we need a spinlock to make the manipulation of
36 * sem->count and sem->waking atomic. Scalability isn't an issue because
37 * this lock is used on UP only so it's just an empty variable.
38 */
39static inline int __sem_update_count(struct semaphore *sem, int incr)
40{
41 int old_count, tmp;
42
43 if (cpu_has_llsc && R10000_LLSC_WAR) {
44 __asm__ __volatile__(
45 " .set mips3 \n"
46 "1: ll %0, %2 # __sem_update_count \n"
47 " sra %1, %0, 31 \n"
48 " not %1 \n"
49 " and %1, %0, %1 \n"
50 " addu %1, %1, %3 \n"
51 " sc %1, %2 \n"
52 " beqzl %1, 1b \n"
53 " .set mips0 \n"
54 : "=&r" (old_count), "=&r" (tmp), "=m" (sem->count)
55 : "r" (incr), "m" (sem->count));
56 } else if (cpu_has_llsc) {
57 __asm__ __volatile__(
58 " .set mips3 \n"
59 "1: ll %0, %2 # __sem_update_count \n"
60 " sra %1, %0, 31 \n"
61 " not %1 \n"
62 " and %1, %0, %1 \n"
63 " addu %1, %1, %3 \n"
64 " sc %1, %2 \n"
65 " beqz %1, 1b \n"
66 " .set mips0 \n"
67 : "=&r" (old_count), "=&r" (tmp), "=m" (sem->count)
68 : "r" (incr), "m" (sem->count));
69 } else {
70 static DEFINE_SPINLOCK(semaphore_lock);
71 unsigned long flags;
72
73 spin_lock_irqsave(&semaphore_lock, flags);
74 old_count = atomic_read(&sem->count);
75 tmp = max_t(int, old_count, 0) + incr;
76 atomic_set(&sem->count, tmp);
77 spin_unlock_irqrestore(&semaphore_lock, flags);
78 }
79
80 return old_count;
81}
82
83void __up(struct semaphore *sem)
84{
85 /*
86 * Note that we incremented count in up() before we came here,
87 * but that was ineffective since the result was <= 0, and
88 * any negative value of count is equivalent to 0.
89 * This ends up setting count to 1, unless count is now > 0
90 * (i.e. because some other cpu has called up() in the meantime),
91 * in which case we just increment count.
92 */
93 __sem_update_count(sem, 1);
94 wake_up(&sem->wait);
95}
96
97EXPORT_SYMBOL(__up);
98
99/*
100 * Note that when we come in to __down or __down_interruptible,
101 * we have already decremented count, but that decrement was
102 * ineffective since the result was < 0, and any negative value
103 * of count is equivalent to 0.
104 * Thus it is only when we decrement count from some value > 0
105 * that we have actually got the semaphore.
106 */
107void __sched __down(struct semaphore *sem)
108{
109 struct task_struct *tsk = current;
110 DECLARE_WAITQUEUE(wait, tsk);
111
112 __set_task_state(tsk, TASK_UNINTERRUPTIBLE);
113 add_wait_queue_exclusive(&sem->wait, &wait);
114
115 /*
116 * Try to get the semaphore. If the count is > 0, then we've
117 * got the semaphore; we decrement count and exit the loop.
118 * If the count is 0 or negative, we set it to -1, indicating
119 * that we are asleep, and then sleep.
120 */
121 while (__sem_update_count(sem, -1) <= 0) {
122 schedule();
123 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
124 }
125 remove_wait_queue(&sem->wait, &wait);
126 __set_task_state(tsk, TASK_RUNNING);
127
128 /*
129 * If there are any more sleepers, wake one of them up so
130 * that it can either get the semaphore, or set count to -1
131 * indicating that there are still processes sleeping.
132 */
133 wake_up(&sem->wait);
134}
135
136EXPORT_SYMBOL(__down);
137
138int __sched __down_interruptible(struct semaphore * sem)
139{
140 int retval = 0;
141 struct task_struct *tsk = current;
142 DECLARE_WAITQUEUE(wait, tsk);
143
144 __set_task_state(tsk, TASK_INTERRUPTIBLE);
145 add_wait_queue_exclusive(&sem->wait, &wait);
146
147 while (__sem_update_count(sem, -1) <= 0) {
148 if (signal_pending(current)) {
149 /*
150 * A signal is pending - give up trying.
151 * Set sem->count to 0 if it is negative,
152 * since we are no longer sleeping.
153 */
154 __sem_update_count(sem, 0);
155 retval = -EINTR;
156 break;
157 }
158 schedule();
159 set_task_state(tsk, TASK_INTERRUPTIBLE);
160 }
161 remove_wait_queue(&sem->wait, &wait);
162 __set_task_state(tsk, TASK_RUNNING);
163
164 wake_up(&sem->wait);
165 return retval;
166}
167
168EXPORT_SYMBOL(__down_interruptible);
diff --git a/arch/mn10300/kernel/Makefile b/arch/mn10300/kernel/Makefile
index ef07c956170a..23f2ab67574c 100644
--- a/arch/mn10300/kernel/Makefile
+++ b/arch/mn10300/kernel/Makefile
@@ -3,7 +3,7 @@
3# 3#
4extra-y := head.o init_task.o vmlinux.lds 4extra-y := head.o init_task.o vmlinux.lds
5 5
6obj-y := process.o semaphore.o signal.o entry.o fpu.o traps.o irq.o \ 6obj-y := process.o signal.o entry.o fpu.o traps.o irq.o \
7 ptrace.o setup.o time.o sys_mn10300.o io.o kthread.o \ 7 ptrace.o setup.o time.o sys_mn10300.o io.o kthread.o \
8 switch_to.o mn10300_ksyms.o kernel_execve.o 8 switch_to.o mn10300_ksyms.o kernel_execve.o
9 9
diff --git a/arch/mn10300/kernel/semaphore.c b/arch/mn10300/kernel/semaphore.c
deleted file mode 100644
index 9153c4039fd2..000000000000
--- a/arch/mn10300/kernel/semaphore.c
+++ /dev/null
@@ -1,149 +0,0 @@
1/* MN10300 Semaphore implementation
2 *
3 * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public Licence
8 * as published by the Free Software Foundation; either version
9 * 2 of the Licence, or (at your option) any later version.
10 */
11#include <linux/sched.h>
12#include <linux/module.h>
13#include <asm/semaphore.h>
14
15struct sem_waiter {
16 struct list_head list;
17 struct task_struct *task;
18};
19
20#if SEMAPHORE_DEBUG
21void semtrace(struct semaphore *sem, const char *str)
22{
23 if (sem->debug)
24 printk(KERN_DEBUG "[%d] %s({%d,%d})\n",
25 current->pid,
26 str,
27 atomic_read(&sem->count),
28 list_empty(&sem->wait_list) ? 0 : 1);
29}
30#else
31#define semtrace(SEM, STR) do { } while (0)
32#endif
33
34/*
35 * wait for a token to be granted from a semaphore
36 * - entered with lock held and interrupts disabled
37 */
38void __down(struct semaphore *sem, unsigned long flags)
39{
40 struct task_struct *tsk = current;
41 struct sem_waiter waiter;
42
43 semtrace(sem, "Entering __down");
44
45 /* set up my own style of waitqueue */
46 waiter.task = tsk;
47 get_task_struct(tsk);
48
49 list_add_tail(&waiter.list, &sem->wait_list);
50
51 /* we don't need to touch the semaphore struct anymore */
52 spin_unlock_irqrestore(&sem->wait_lock, flags);
53
54 /* wait to be given the semaphore */
55 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
56
57 for (;;) {
58 if (!waiter.task)
59 break;
60 schedule();
61 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
62 }
63
64 tsk->state = TASK_RUNNING;
65 semtrace(sem, "Leaving __down");
66}
67EXPORT_SYMBOL(__down);
68
69/*
70 * interruptibly wait for a token to be granted from a semaphore
71 * - entered with lock held and interrupts disabled
72 */
73int __down_interruptible(struct semaphore *sem, unsigned long flags)
74{
75 struct task_struct *tsk = current;
76 struct sem_waiter waiter;
77 int ret;
78
79 semtrace(sem, "Entering __down_interruptible");
80
81 /* set up my own style of waitqueue */
82 waiter.task = tsk;
83 get_task_struct(tsk);
84
85 list_add_tail(&waiter.list, &sem->wait_list);
86
87 /* we don't need to touch the semaphore struct anymore */
88 set_task_state(tsk, TASK_INTERRUPTIBLE);
89
90 spin_unlock_irqrestore(&sem->wait_lock, flags);
91
92 /* wait to be given the semaphore */
93 ret = 0;
94 for (;;) {
95 if (!waiter.task)
96 break;
97 if (unlikely(signal_pending(current)))
98 goto interrupted;
99 schedule();
100 set_task_state(tsk, TASK_INTERRUPTIBLE);
101 }
102
103 out:
104 tsk->state = TASK_RUNNING;
105 semtrace(sem, "Leaving __down_interruptible");
106 return ret;
107
108 interrupted:
109 spin_lock_irqsave(&sem->wait_lock, flags);
110 list_del(&waiter.list);
111 spin_unlock_irqrestore(&sem->wait_lock, flags);
112
113 ret = 0;
114 if (!waiter.task) {
115 put_task_struct(current);
116 ret = -EINTR;
117 }
118 goto out;
119}
120EXPORT_SYMBOL(__down_interruptible);
121
122/*
123 * release a single token back to a semaphore
124 * - entered with lock held and interrupts disabled
125 */
126void __up(struct semaphore *sem)
127{
128 struct task_struct *tsk;
129 struct sem_waiter *waiter;
130
131 semtrace(sem, "Entering __up");
132
133 /* grant the token to the process at the front of the queue */
134 waiter = list_entry(sem->wait_list.next, struct sem_waiter, list);
135
136 /* We must be careful not to touch 'waiter' after we set ->task = NULL.
137 * It is an allocated on the waiter's stack and may become invalid at
138 * any time after that point (due to a wakeup from another source).
139 */
140 list_del_init(&waiter->list);
141 tsk = waiter->task;
142 smp_mb();
143 waiter->task = NULL;
144 wake_up_process(tsk);
145 put_task_struct(tsk);
146
147 semtrace(sem, "Leaving __up");
148}
149EXPORT_SYMBOL(__up);
diff --git a/arch/parisc/kernel/Makefile b/arch/parisc/kernel/Makefile
index 27827bc3717e..1f6585a56f97 100644
--- a/arch/parisc/kernel/Makefile
+++ b/arch/parisc/kernel/Makefile
@@ -9,7 +9,7 @@ AFLAGS_pacache.o := -traditional
9 9
10obj-y := cache.o pacache.o setup.o traps.o time.o irq.o \ 10obj-y := cache.o pacache.o setup.o traps.o time.o irq.o \
11 pa7300lc.o syscall.o entry.o sys_parisc.o firmware.o \ 11 pa7300lc.o syscall.o entry.o sys_parisc.o firmware.o \
12 ptrace.o hardware.o inventory.o drivers.o semaphore.o \ 12 ptrace.o hardware.o inventory.o drivers.o \
13 signal.o hpmc.o real2.o parisc_ksyms.o unaligned.o \ 13 signal.o hpmc.o real2.o parisc_ksyms.o unaligned.o \
14 process.o processor.o pdc_cons.o pdc_chassis.o unwind.o \ 14 process.o processor.o pdc_cons.o pdc_chassis.o unwind.o \
15 topology.o 15 topology.o
diff --git a/arch/parisc/kernel/parisc_ksyms.c b/arch/parisc/kernel/parisc_ksyms.c
index 7aca704e96f0..5b7fc4aa044d 100644
--- a/arch/parisc/kernel/parisc_ksyms.c
+++ b/arch/parisc/kernel/parisc_ksyms.c
@@ -69,11 +69,6 @@ EXPORT_SYMBOL(memcpy_toio);
69EXPORT_SYMBOL(memcpy_fromio); 69EXPORT_SYMBOL(memcpy_fromio);
70EXPORT_SYMBOL(memset_io); 70EXPORT_SYMBOL(memset_io);
71 71
72#include <asm/semaphore.h>
73EXPORT_SYMBOL(__up);
74EXPORT_SYMBOL(__down_interruptible);
75EXPORT_SYMBOL(__down);
76
77extern void $$divI(void); 72extern void $$divI(void);
78extern void $$divU(void); 73extern void $$divU(void);
79extern void $$remI(void); 74extern void $$remI(void);
diff --git a/arch/parisc/kernel/semaphore.c b/arch/parisc/kernel/semaphore.c
deleted file mode 100644
index ee806bcc3726..000000000000
--- a/arch/parisc/kernel/semaphore.c
+++ /dev/null
@@ -1,102 +0,0 @@
1/*
2 * Semaphore implementation Copyright (c) 2001 Matthew Wilcox, Hewlett-Packard
3 */
4
5#include <linux/sched.h>
6#include <linux/spinlock.h>
7#include <linux/errno.h>
8#include <linux/init.h>
9
10/*
11 * Semaphores are complex as we wish to avoid using two variables.
12 * `count' has multiple roles, depending on its value. If it is positive
13 * or zero, there are no waiters. The functions here will never be
14 * called; see <asm/semaphore.h>
15 *
16 * When count is -1 it indicates there is at least one task waiting
17 * for the semaphore.
18 *
19 * When count is less than that, there are '- count - 1' wakeups
20 * pending. ie if it has value -3, there are 2 wakeups pending.
21 *
22 * Note that these functions are only called when there is contention
23 * on the lock, and as such all this is the "non-critical" part of the
24 * whole semaphore business. The critical part is the inline stuff in
25 * <asm/semaphore.h> where we want to avoid any extra jumps and calls.
26 */
27void __up(struct semaphore *sem)
28{
29 sem->count--;
30 wake_up(&sem->wait);
31}
32
33#define wakers(count) (-1 - count)
34
35#define DOWN_HEAD \
36 int ret = 0; \
37 DECLARE_WAITQUEUE(wait, current); \
38 \
39 /* Note that someone is waiting */ \
40 if (sem->count == 0) \
41 sem->count = -1; \
42 \
43 /* protected by the sentry still -- use unlocked version */ \
44 wait.flags = WQ_FLAG_EXCLUSIVE; \
45 __add_wait_queue_tail(&sem->wait, &wait); \
46 lost_race: \
47 spin_unlock_irq(&sem->sentry); \
48
49#define DOWN_TAIL \
50 spin_lock_irq(&sem->sentry); \
51 if (wakers(sem->count) == 0 && ret == 0) \
52 goto lost_race; /* Someone stole our wakeup */ \
53 __remove_wait_queue(&sem->wait, &wait); \
54 current->state = TASK_RUNNING; \
55 if (!waitqueue_active(&sem->wait) && (sem->count < 0)) \
56 sem->count = wakers(sem->count);
57
58#define UPDATE_COUNT \
59 sem->count += (sem->count < 0) ? 1 : - 1;
60
61
62void __sched __down(struct semaphore * sem)
63{
64 DOWN_HEAD
65
66 for(;;) {
67 set_task_state(current, TASK_UNINTERRUPTIBLE);
68 /* we can _read_ this without the sentry */
69 if (sem->count != -1)
70 break;
71 schedule();
72 }
73
74 DOWN_TAIL
75 UPDATE_COUNT
76}
77
78int __sched __down_interruptible(struct semaphore * sem)
79{
80 DOWN_HEAD
81
82 for(;;) {
83 set_task_state(current, TASK_INTERRUPTIBLE);
84 /* we can _read_ this without the sentry */
85 if (sem->count != -1)
86 break;
87
88 if (signal_pending(current)) {
89 ret = -EINTR;
90 break;
91 }
92 schedule();
93 }
94
95 DOWN_TAIL
96
97 if (!ret) {
98 UPDATE_COUNT
99 }
100
101 return ret;
102}
diff --git a/arch/powerpc/kernel/Makefile b/arch/powerpc/kernel/Makefile
index c1baf9d5903f..b9dbfff9afe9 100644
--- a/arch/powerpc/kernel/Makefile
+++ b/arch/powerpc/kernel/Makefile
@@ -12,7 +12,7 @@ CFLAGS_prom_init.o += -fPIC
12CFLAGS_btext.o += -fPIC 12CFLAGS_btext.o += -fPIC
13endif 13endif
14 14
15obj-y := semaphore.o cputable.o ptrace.o syscalls.o \ 15obj-y := cputable.o ptrace.o syscalls.o \
16 irq.o align.o signal_32.o pmc.o vdso.o \ 16 irq.o align.o signal_32.o pmc.o vdso.o \
17 init_task.o process.o systbl.o idle.o \ 17 init_task.o process.o systbl.o idle.o \
18 signal.o 18 signal.o
diff --git a/arch/powerpc/kernel/ppc_ksyms.c b/arch/powerpc/kernel/ppc_ksyms.c
index 9c98424277a8..65d14e6ddc3c 100644
--- a/arch/powerpc/kernel/ppc_ksyms.c
+++ b/arch/powerpc/kernel/ppc_ksyms.c
@@ -15,7 +15,6 @@
15#include <linux/bitops.h> 15#include <linux/bitops.h>
16 16
17#include <asm/page.h> 17#include <asm/page.h>
18#include <asm/semaphore.h>
19#include <asm/processor.h> 18#include <asm/processor.h>
20#include <asm/cacheflush.h> 19#include <asm/cacheflush.h>
21#include <asm/uaccess.h> 20#include <asm/uaccess.h>
diff --git a/arch/powerpc/kernel/semaphore.c b/arch/powerpc/kernel/semaphore.c
deleted file mode 100644
index 2f8c3c951394..000000000000
--- a/arch/powerpc/kernel/semaphore.c
+++ /dev/null
@@ -1,135 +0,0 @@
1/*
2 * PowerPC-specific semaphore code.
3 *
4 * Copyright (C) 1999 Cort Dougan <cort@cs.nmt.edu>
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
10 *
11 * April 2001 - Reworked by Paul Mackerras <paulus@samba.org>
12 * to eliminate the SMP races in the old version between the updates
13 * of `count' and `waking'. Now we use negative `count' values to
14 * indicate that some process(es) are waiting for the semaphore.
15 */
16
17#include <linux/sched.h>
18#include <linux/init.h>
19#include <linux/module.h>
20
21#include <asm/atomic.h>
22#include <asm/semaphore.h>
23#include <asm/errno.h>
24
25/*
26 * Atomically update sem->count.
27 * This does the equivalent of the following:
28 *
29 * old_count = sem->count;
30 * tmp = MAX(old_count, 0) + incr;
31 * sem->count = tmp;
32 * return old_count;
33 */
34static inline int __sem_update_count(struct semaphore *sem, int incr)
35{
36 int old_count, tmp;
37
38 __asm__ __volatile__("\n"
39"1: lwarx %0,0,%3\n"
40" srawi %1,%0,31\n"
41" andc %1,%0,%1\n"
42" add %1,%1,%4\n"
43 PPC405_ERR77(0,%3)
44" stwcx. %1,0,%3\n"
45" bne 1b"
46 : "=&r" (old_count), "=&r" (tmp), "=m" (sem->count)
47 : "r" (&sem->count), "r" (incr), "m" (sem->count)
48 : "cc");
49
50 return old_count;
51}
52
53void __up(struct semaphore *sem)
54{
55 /*
56 * Note that we incremented count in up() before we came here,
57 * but that was ineffective since the result was <= 0, and
58 * any negative value of count is equivalent to 0.
59 * This ends up setting count to 1, unless count is now > 0
60 * (i.e. because some other cpu has called up() in the meantime),
61 * in which case we just increment count.
62 */
63 __sem_update_count(sem, 1);
64 wake_up(&sem->wait);
65}
66EXPORT_SYMBOL(__up);
67
68/*
69 * Note that when we come in to __down or __down_interruptible,
70 * we have already decremented count, but that decrement was
71 * ineffective since the result was < 0, and any negative value
72 * of count is equivalent to 0.
73 * Thus it is only when we decrement count from some value > 0
74 * that we have actually got the semaphore.
75 */
76void __sched __down(struct semaphore *sem)
77{
78 struct task_struct *tsk = current;
79 DECLARE_WAITQUEUE(wait, tsk);
80
81 __set_task_state(tsk, TASK_UNINTERRUPTIBLE);
82 add_wait_queue_exclusive(&sem->wait, &wait);
83
84 /*
85 * Try to get the semaphore. If the count is > 0, then we've
86 * got the semaphore; we decrement count and exit the loop.
87 * If the count is 0 or negative, we set it to -1, indicating
88 * that we are asleep, and then sleep.
89 */
90 while (__sem_update_count(sem, -1) <= 0) {
91 schedule();
92 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
93 }
94 remove_wait_queue(&sem->wait, &wait);
95 __set_task_state(tsk, TASK_RUNNING);
96
97 /*
98 * If there are any more sleepers, wake one of them up so
99 * that it can either get the semaphore, or set count to -1
100 * indicating that there are still processes sleeping.
101 */
102 wake_up(&sem->wait);
103}
104EXPORT_SYMBOL(__down);
105
106int __sched __down_interruptible(struct semaphore * sem)
107{
108 int retval = 0;
109 struct task_struct *tsk = current;
110 DECLARE_WAITQUEUE(wait, tsk);
111
112 __set_task_state(tsk, TASK_INTERRUPTIBLE);
113 add_wait_queue_exclusive(&sem->wait, &wait);
114
115 while (__sem_update_count(sem, -1) <= 0) {
116 if (signal_pending(current)) {
117 /*
118 * A signal is pending - give up trying.
119 * Set sem->count to 0 if it is negative,
120 * since we are no longer sleeping.
121 */
122 __sem_update_count(sem, 0);
123 retval = -EINTR;
124 break;
125 }
126 schedule();
127 set_task_state(tsk, TASK_INTERRUPTIBLE);
128 }
129 remove_wait_queue(&sem->wait, &wait);
130 __set_task_state(tsk, TASK_RUNNING);
131
132 wake_up(&sem->wait);
133 return retval;
134}
135EXPORT_SYMBOL(__down_interruptible);
diff --git a/arch/ppc/kernel/semaphore.c b/arch/ppc/kernel/semaphore.c
deleted file mode 100644
index 2fe429b27c14..000000000000
--- a/arch/ppc/kernel/semaphore.c
+++ /dev/null
@@ -1,131 +0,0 @@
1/*
2 * PowerPC-specific semaphore code.
3 *
4 * Copyright (C) 1999 Cort Dougan <cort@cs.nmt.edu>
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
10 *
11 * April 2001 - Reworked by Paul Mackerras <paulus@samba.org>
12 * to eliminate the SMP races in the old version between the updates
13 * of `count' and `waking'. Now we use negative `count' values to
14 * indicate that some process(es) are waiting for the semaphore.
15 */
16
17#include <linux/sched.h>
18#include <linux/init.h>
19#include <asm/atomic.h>
20#include <asm/semaphore.h>
21#include <asm/errno.h>
22
23/*
24 * Atomically update sem->count.
25 * This does the equivalent of the following:
26 *
27 * old_count = sem->count;
28 * tmp = MAX(old_count, 0) + incr;
29 * sem->count = tmp;
30 * return old_count;
31 */
32static inline int __sem_update_count(struct semaphore *sem, int incr)
33{
34 int old_count, tmp;
35
36 __asm__ __volatile__("\n"
37"1: lwarx %0,0,%3\n"
38" srawi %1,%0,31\n"
39" andc %1,%0,%1\n"
40" add %1,%1,%4\n"
41 PPC405_ERR77(0,%3)
42" stwcx. %1,0,%3\n"
43" bne 1b"
44 : "=&r" (old_count), "=&r" (tmp), "=m" (sem->count)
45 : "r" (&sem->count), "r" (incr), "m" (sem->count)
46 : "cc");
47
48 return old_count;
49}
50
51void __up(struct semaphore *sem)
52{
53 /*
54 * Note that we incremented count in up() before we came here,
55 * but that was ineffective since the result was <= 0, and
56 * any negative value of count is equivalent to 0.
57 * This ends up setting count to 1, unless count is now > 0
58 * (i.e. because some other cpu has called up() in the meantime),
59 * in which case we just increment count.
60 */
61 __sem_update_count(sem, 1);
62 wake_up(&sem->wait);
63}
64
65/*
66 * Note that when we come in to __down or __down_interruptible,
67 * we have already decremented count, but that decrement was
68 * ineffective since the result was < 0, and any negative value
69 * of count is equivalent to 0.
70 * Thus it is only when we decrement count from some value > 0
71 * that we have actually got the semaphore.
72 */
73void __sched __down(struct semaphore *sem)
74{
75 struct task_struct *tsk = current;
76 DECLARE_WAITQUEUE(wait, tsk);
77
78 tsk->state = TASK_UNINTERRUPTIBLE;
79 add_wait_queue_exclusive(&sem->wait, &wait);
80 smp_wmb();
81
82 /*
83 * Try to get the semaphore. If the count is > 0, then we've
84 * got the semaphore; we decrement count and exit the loop.
85 * If the count is 0 or negative, we set it to -1, indicating
86 * that we are asleep, and then sleep.
87 */
88 while (__sem_update_count(sem, -1) <= 0) {
89 schedule();
90 tsk->state = TASK_UNINTERRUPTIBLE;
91 }
92 remove_wait_queue(&sem->wait, &wait);
93 tsk->state = TASK_RUNNING;
94
95 /*
96 * If there are any more sleepers, wake one of them up so
97 * that it can either get the semaphore, or set count to -1
98 * indicating that there are still processes sleeping.
99 */
100 wake_up(&sem->wait);
101}
102
103int __sched __down_interruptible(struct semaphore * sem)
104{
105 int retval = 0;
106 struct task_struct *tsk = current;
107 DECLARE_WAITQUEUE(wait, tsk);
108
109 tsk->state = TASK_INTERRUPTIBLE;
110 add_wait_queue_exclusive(&sem->wait, &wait);
111 smp_wmb();
112
113 while (__sem_update_count(sem, -1) <= 0) {
114 if (signal_pending(current)) {
115 /*
116 * A signal is pending - give up trying.
117 * Set sem->count to 0 if it is negative,
118 * since we are no longer sleeping.
119 */
120 __sem_update_count(sem, 0);
121 retval = -EINTR;
122 break;
123 }
124 schedule();
125 tsk->state = TASK_INTERRUPTIBLE;
126 }
127 tsk->state = TASK_RUNNING;
128 remove_wait_queue(&sem->wait, &wait);
129 wake_up(&sem->wait);
130 return retval;
131}
diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig
index 1831833c430e..f6a68e178fc5 100644
--- a/arch/s390/Kconfig
+++ b/arch/s390/Kconfig
@@ -3,6 +3,10 @@
3# see Documentation/kbuild/kconfig-language.txt. 3# see Documentation/kbuild/kconfig-language.txt.
4# 4#
5 5
6config SCHED_MC
7 def_bool y
8 depends on SMP
9
6config MMU 10config MMU
7 def_bool y 11 def_bool y
8 12
@@ -39,6 +43,9 @@ config GENERIC_HWEIGHT
39config GENERIC_TIME 43config GENERIC_TIME
40 def_bool y 44 def_bool y
41 45
46config GENERIC_CLOCKEVENTS
47 def_bool y
48
42config GENERIC_BUG 49config GENERIC_BUG
43 bool 50 bool
44 depends on BUG 51 depends on BUG
@@ -69,6 +76,8 @@ menu "Base setup"
69 76
70comment "Processor type and features" 77comment "Processor type and features"
71 78
79source "kernel/time/Kconfig"
80
72config 64BIT 81config 64BIT
73 bool "64 bit kernel" 82 bool "64 bit kernel"
74 help 83 help
@@ -301,10 +310,7 @@ config QDIO
301 tristate "QDIO support" 310 tristate "QDIO support"
302 ---help--- 311 ---help---
303 This driver provides the Queued Direct I/O base support for 312 This driver provides the Queued Direct I/O base support for
304 IBM mainframes. 313 IBM System z.
305
306 For details please refer to the documentation provided by IBM at
307 <http://www10.software.ibm.com/developerworks/opensource/linux390>
308 314
309 To compile this driver as a module, choose M here: the 315 To compile this driver as a module, choose M here: the
310 module will be called qdio. 316 module will be called qdio.
@@ -486,25 +492,6 @@ config APPLDATA_NET_SUM
486 492
487source kernel/Kconfig.hz 493source kernel/Kconfig.hz
488 494
489config NO_IDLE_HZ
490 bool "No HZ timer ticks in idle"
491 help
492 Switches the regular HZ timer off when the system is going idle.
493 This helps z/VM to detect that the Linux system is idle. VM can
494 then "swap-out" this guest which reduces memory usage. It also
495 reduces the overhead of idle systems.
496
497 The HZ timer can be switched on/off via /proc/sys/kernel/hz_timer.
498 hz_timer=0 means HZ timer is disabled. hz_timer=1 means HZ
499 timer is active.
500
501config NO_IDLE_HZ_INIT
502 bool "HZ timer in idle off by default"
503 depends on NO_IDLE_HZ
504 help
505 The HZ timer is switched off in idle by default. That means the
506 HZ timer is already disabled at boot time.
507
508config S390_HYPFS_FS 495config S390_HYPFS_FS
509 bool "s390 hypervisor file system support" 496 bool "s390 hypervisor file system support"
510 select SYS_HYPERVISOR 497 select SYS_HYPERVISOR
diff --git a/arch/s390/crypto/aes_s390.c b/arch/s390/crypto/aes_s390.c
index a3f67f8b5427..e33f32b54c08 100644
--- a/arch/s390/crypto/aes_s390.c
+++ b/arch/s390/crypto/aes_s390.c
@@ -499,7 +499,7 @@ static struct crypto_alg cbc_aes_alg = {
499 } 499 }
500}; 500};
501 501
502static int __init aes_init(void) 502static int __init aes_s390_init(void)
503{ 503{
504 int ret; 504 int ret;
505 505
@@ -542,15 +542,15 @@ aes_err:
542 goto out; 542 goto out;
543} 543}
544 544
545static void __exit aes_fini(void) 545static void __exit aes_s390_fini(void)
546{ 546{
547 crypto_unregister_alg(&cbc_aes_alg); 547 crypto_unregister_alg(&cbc_aes_alg);
548 crypto_unregister_alg(&ecb_aes_alg); 548 crypto_unregister_alg(&ecb_aes_alg);
549 crypto_unregister_alg(&aes_alg); 549 crypto_unregister_alg(&aes_alg);
550} 550}
551 551
552module_init(aes_init); 552module_init(aes_s390_init);
553module_exit(aes_fini); 553module_exit(aes_s390_fini);
554 554
555MODULE_ALIAS("aes"); 555MODULE_ALIAS("aes");
556 556
diff --git a/arch/s390/crypto/des_s390.c b/arch/s390/crypto/des_s390.c
index ea22707f435f..4aba83b31596 100644
--- a/arch/s390/crypto/des_s390.c
+++ b/arch/s390/crypto/des_s390.c
@@ -550,7 +550,7 @@ static struct crypto_alg cbc_des3_192_alg = {
550 } 550 }
551}; 551};
552 552
553static int init(void) 553static int des_s390_init(void)
554{ 554{
555 int ret = 0; 555 int ret = 0;
556 556
@@ -612,7 +612,7 @@ des_err:
612 goto out; 612 goto out;
613} 613}
614 614
615static void __exit fini(void) 615static void __exit des_s390_fini(void)
616{ 616{
617 crypto_unregister_alg(&cbc_des3_192_alg); 617 crypto_unregister_alg(&cbc_des3_192_alg);
618 crypto_unregister_alg(&ecb_des3_192_alg); 618 crypto_unregister_alg(&ecb_des3_192_alg);
@@ -625,8 +625,8 @@ static void __exit fini(void)
625 crypto_unregister_alg(&des_alg); 625 crypto_unregister_alg(&des_alg);
626} 626}
627 627
628module_init(init); 628module_init(des_s390_init);
629module_exit(fini); 629module_exit(des_s390_fini);
630 630
631MODULE_ALIAS("des"); 631MODULE_ALIAS("des");
632MODULE_ALIAS("des3_ede"); 632MODULE_ALIAS("des3_ede");
diff --git a/arch/s390/crypto/sha1_s390.c b/arch/s390/crypto/sha1_s390.c
index 5a834f6578ab..9cf9eca22747 100644
--- a/arch/s390/crypto/sha1_s390.c
+++ b/arch/s390/crypto/sha1_s390.c
@@ -137,7 +137,7 @@ static struct crypto_alg alg = {
137 .dia_final = sha1_final } } 137 .dia_final = sha1_final } }
138}; 138};
139 139
140static int __init init(void) 140static int __init sha1_s390_init(void)
141{ 141{
142 if (!crypt_s390_func_available(KIMD_SHA_1)) 142 if (!crypt_s390_func_available(KIMD_SHA_1))
143 return -EOPNOTSUPP; 143 return -EOPNOTSUPP;
@@ -145,13 +145,13 @@ static int __init init(void)
145 return crypto_register_alg(&alg); 145 return crypto_register_alg(&alg);
146} 146}
147 147
148static void __exit fini(void) 148static void __exit sha1_s390_fini(void)
149{ 149{
150 crypto_unregister_alg(&alg); 150 crypto_unregister_alg(&alg);
151} 151}
152 152
153module_init(init); 153module_init(sha1_s390_init);
154module_exit(fini); 154module_exit(sha1_s390_fini);
155 155
156MODULE_ALIAS("sha1"); 156MODULE_ALIAS("sha1");
157 157
diff --git a/arch/s390/crypto/sha256_s390.c b/arch/s390/crypto/sha256_s390.c
index ccf8633c4f65..2a3d756b35d4 100644
--- a/arch/s390/crypto/sha256_s390.c
+++ b/arch/s390/crypto/sha256_s390.c
@@ -133,7 +133,7 @@ static struct crypto_alg alg = {
133 .dia_final = sha256_final } } 133 .dia_final = sha256_final } }
134}; 134};
135 135
136static int init(void) 136static int sha256_s390_init(void)
137{ 137{
138 if (!crypt_s390_func_available(KIMD_SHA_256)) 138 if (!crypt_s390_func_available(KIMD_SHA_256))
139 return -EOPNOTSUPP; 139 return -EOPNOTSUPP;
@@ -141,13 +141,13 @@ static int init(void)
141 return crypto_register_alg(&alg); 141 return crypto_register_alg(&alg);
142} 142}
143 143
144static void __exit fini(void) 144static void __exit sha256_s390_fini(void)
145{ 145{
146 crypto_unregister_alg(&alg); 146 crypto_unregister_alg(&alg);
147} 147}
148 148
149module_init(init); 149module_init(sha256_s390_init);
150module_exit(fini); 150module_exit(sha256_s390_fini);
151 151
152MODULE_ALIAS("sha256"); 152MODULE_ALIAS("sha256");
153 153
diff --git a/arch/s390/defconfig b/arch/s390/defconfig
index 62f6b5a606dd..dcc3ec2ef643 100644
--- a/arch/s390/defconfig
+++ b/arch/s390/defconfig
@@ -3,6 +3,7 @@
3# Linux kernel version: 2.6.25-rc4 3# Linux kernel version: 2.6.25-rc4
4# Wed Mar 5 11:22:59 2008 4# Wed Mar 5 11:22:59 2008
5# 5#
6CONFIG_SCHED_MC=y
6CONFIG_MMU=y 7CONFIG_MMU=y
7CONFIG_ZONE_DMA=y 8CONFIG_ZONE_DMA=y
8CONFIG_LOCKDEP_SUPPORT=y 9CONFIG_LOCKDEP_SUPPORT=y
diff --git a/arch/s390/kernel/Makefile b/arch/s390/kernel/Makefile
index 4d3e38392cb1..77051cd27925 100644
--- a/arch/s390/kernel/Makefile
+++ b/arch/s390/kernel/Makefile
@@ -11,7 +11,7 @@ CFLAGS_smp.o := -Wno-nonnull
11 11
12obj-y := bitmap.o traps.o time.o process.o base.o early.o \ 12obj-y := bitmap.o traps.o time.o process.o base.o early.o \
13 setup.o sys_s390.o ptrace.o signal.o cpcmd.o ebcdic.o \ 13 setup.o sys_s390.o ptrace.o signal.o cpcmd.o ebcdic.o \
14 semaphore.o s390_ext.o debug.o irq.o ipl.o dis.o diag.o 14 s390_ext.o debug.o irq.o ipl.o dis.o diag.o
15 15
16obj-y += $(if $(CONFIG_64BIT),entry64.o,entry.o) 16obj-y += $(if $(CONFIG_64BIT),entry64.o,entry.o)
17obj-y += $(if $(CONFIG_64BIT),reipl64.o,reipl.o) 17obj-y += $(if $(CONFIG_64BIT),reipl64.o,reipl.o)
@@ -19,7 +19,7 @@ obj-y += $(if $(CONFIG_64BIT),reipl64.o,reipl.o)
19extra-y += head.o init_task.o vmlinux.lds 19extra-y += head.o init_task.o vmlinux.lds
20 20
21obj-$(CONFIG_MODULES) += s390_ksyms.o module.o 21obj-$(CONFIG_MODULES) += s390_ksyms.o module.o
22obj-$(CONFIG_SMP) += smp.o 22obj-$(CONFIG_SMP) += smp.o topology.o
23 23
24obj-$(CONFIG_AUDIT) += audit.o 24obj-$(CONFIG_AUDIT) += audit.o
25compat-obj-$(CONFIG_AUDIT) += compat_audit.o 25compat-obj-$(CONFIG_AUDIT) += compat_audit.o
diff --git a/arch/s390/kernel/compat_linux.h b/arch/s390/kernel/compat_linux.h
index e89f8c0c42a0..20723a062017 100644
--- a/arch/s390/kernel/compat_linux.h
+++ b/arch/s390/kernel/compat_linux.h
@@ -162,4 +162,77 @@ struct ucontext32 {
162 compat_sigset_t uc_sigmask; /* mask last for extensibility */ 162 compat_sigset_t uc_sigmask; /* mask last for extensibility */
163}; 163};
164 164
165struct __sysctl_args32;
166struct stat64_emu31;
167struct mmap_arg_struct_emu31;
168struct fadvise64_64_args;
169struct old_sigaction32;
170struct old_sigaction32;
171
172long sys32_chown16(const char __user * filename, u16 user, u16 group);
173long sys32_lchown16(const char __user * filename, u16 user, u16 group);
174long sys32_fchown16(unsigned int fd, u16 user, u16 group);
175long sys32_setregid16(u16 rgid, u16 egid);
176long sys32_setgid16(u16 gid);
177long sys32_setreuid16(u16 ruid, u16 euid);
178long sys32_setuid16(u16 uid);
179long sys32_setresuid16(u16 ruid, u16 euid, u16 suid);
180long sys32_getresuid16(u16 __user *ruid, u16 __user *euid, u16 __user *suid);
181long sys32_setresgid16(u16 rgid, u16 egid, u16 sgid);
182long sys32_getresgid16(u16 __user *rgid, u16 __user *egid, u16 __user *sgid);
183long sys32_setfsuid16(u16 uid);
184long sys32_setfsgid16(u16 gid);
185long sys32_getgroups16(int gidsetsize, u16 __user *grouplist);
186long sys32_setgroups16(int gidsetsize, u16 __user *grouplist);
187long sys32_getuid16(void);
188long sys32_geteuid16(void);
189long sys32_getgid16(void);
190long sys32_getegid16(void);
191long sys32_ipc(u32 call, int first, int second, int third, u32 ptr);
192long sys32_truncate64(const char __user * path, unsigned long high,
193 unsigned long low);
194long sys32_ftruncate64(unsigned int fd, unsigned long high, unsigned long low);
195long sys32_sched_rr_get_interval(compat_pid_t pid,
196 struct compat_timespec __user *interval);
197long sys32_rt_sigprocmask(int how, compat_sigset_t __user *set,
198 compat_sigset_t __user *oset, size_t sigsetsize);
199long sys32_rt_sigpending(compat_sigset_t __user *set, size_t sigsetsize);
200long sys32_rt_sigqueueinfo(int pid, int sig, compat_siginfo_t __user *uinfo);
201long sys32_execve(void);
202long sys32_init_module(void __user *umod, unsigned long len,
203 const char __user *uargs);
204long sys32_delete_module(const char __user *name_user, unsigned int flags);
205long sys32_gettimeofday(struct compat_timeval __user *tv,
206 struct timezone __user *tz);
207long sys32_settimeofday(struct compat_timeval __user *tv,
208 struct timezone __user *tz);
209long sys32_pause(void);
210long sys32_pread64(unsigned int fd, char __user *ubuf, size_t count,
211 u32 poshi, u32 poslo);
212long sys32_pwrite64(unsigned int fd, const char __user *ubuf,
213 size_t count, u32 poshi, u32 poslo);
214compat_ssize_t sys32_readahead(int fd, u32 offhi, u32 offlo, s32 count);
215long sys32_sendfile(int out_fd, int in_fd, compat_off_t __user *offset,
216 size_t count);
217long sys32_sendfile64(int out_fd, int in_fd, compat_loff_t __user *offset,
218 s32 count);
219long sys32_sysctl(struct __sysctl_args32 __user *args);
220long sys32_stat64(char __user * filename, struct stat64_emu31 __user * statbuf);
221long sys32_lstat64(char __user * filename,
222 struct stat64_emu31 __user * statbuf);
223long sys32_fstat64(unsigned long fd, struct stat64_emu31 __user * statbuf);
224long sys32_fstatat64(unsigned int dfd, char __user *filename,
225 struct stat64_emu31 __user* statbuf, int flag);
226unsigned long old32_mmap(struct mmap_arg_struct_emu31 __user *arg);
227long sys32_mmap2(struct mmap_arg_struct_emu31 __user *arg);
228long sys32_read(unsigned int fd, char __user * buf, size_t count);
229long sys32_write(unsigned int fd, char __user * buf, size_t count);
230long sys32_clone(void);
231long sys32_fadvise64(int fd, loff_t offset, size_t len, int advise);
232long sys32_fadvise64_64(struct fadvise64_64_args __user *args);
233long sys32_sigaction(int sig, const struct old_sigaction32 __user *act,
234 struct old_sigaction32 __user *oact);
235long sys32_rt_sigaction(int sig, const struct sigaction32 __user *act,
236 struct sigaction32 __user *oact, size_t sigsetsize);
237long sys32_sigaltstack(const stack_t32 __user *uss, stack_t32 __user *uoss);
165#endif /* _ASM_S390X_S390_H */ 238#endif /* _ASM_S390X_S390_H */
diff --git a/arch/s390/kernel/compat_signal.c b/arch/s390/kernel/compat_signal.c
index a5692c460bad..c7f02e777af2 100644
--- a/arch/s390/kernel/compat_signal.c
+++ b/arch/s390/kernel/compat_signal.c
@@ -29,6 +29,7 @@
29#include <asm/lowcore.h> 29#include <asm/lowcore.h>
30#include "compat_linux.h" 30#include "compat_linux.h"
31#include "compat_ptrace.h" 31#include "compat_ptrace.h"
32#include "entry.h"
32 33
33#define _BLOCKABLE (~(sigmask(SIGKILL) | sigmask(SIGSTOP))) 34#define _BLOCKABLE (~(sigmask(SIGKILL) | sigmask(SIGSTOP)))
34 35
@@ -428,6 +429,10 @@ get_sigframe(struct k_sigaction *ka, struct pt_regs * regs, size_t frame_size)
428 /* Default to using normal stack */ 429 /* Default to using normal stack */
429 sp = (unsigned long) A(regs->gprs[15]); 430 sp = (unsigned long) A(regs->gprs[15]);
430 431
432 /* Overflow on alternate signal stack gives SIGSEGV. */
433 if (on_sig_stack(sp) && !on_sig_stack((sp - frame_size) & -8UL))
434 return (void __user *) -1UL;
435
431 /* This is the X/Open sanctioned signal stack switching. */ 436 /* This is the X/Open sanctioned signal stack switching. */
432 if (ka->sa.sa_flags & SA_ONSTACK) { 437 if (ka->sa.sa_flags & SA_ONSTACK) {
433 if (! sas_ss_flags(sp)) 438 if (! sas_ss_flags(sp))
@@ -461,6 +466,9 @@ static int setup_frame32(int sig, struct k_sigaction *ka,
461 if (!access_ok(VERIFY_WRITE, frame, sizeof(sigframe32))) 466 if (!access_ok(VERIFY_WRITE, frame, sizeof(sigframe32)))
462 goto give_sigsegv; 467 goto give_sigsegv;
463 468
469 if (frame == (void __user *) -1UL)
470 goto give_sigsegv;
471
464 if (__copy_to_user(&frame->sc.oldmask, &set->sig, _SIGMASK_COPY_SIZE32)) 472 if (__copy_to_user(&frame->sc.oldmask, &set->sig, _SIGMASK_COPY_SIZE32))
465 goto give_sigsegv; 473 goto give_sigsegv;
466 474
@@ -514,6 +522,9 @@ static int setup_rt_frame32(int sig, struct k_sigaction *ka, siginfo_t *info,
514 if (!access_ok(VERIFY_WRITE, frame, sizeof(rt_sigframe32))) 522 if (!access_ok(VERIFY_WRITE, frame, sizeof(rt_sigframe32)))
515 goto give_sigsegv; 523 goto give_sigsegv;
516 524
525 if (frame == (void __user *) -1UL)
526 goto give_sigsegv;
527
517 if (copy_siginfo_to_user32(&frame->info, info)) 528 if (copy_siginfo_to_user32(&frame->info, info))
518 goto give_sigsegv; 529 goto give_sigsegv;
519 530
diff --git a/arch/s390/kernel/debug.c b/arch/s390/kernel/debug.c
index 1b2f5ce45320..1e7d4ac7068b 100644
--- a/arch/s390/kernel/debug.c
+++ b/arch/s390/kernel/debug.c
@@ -73,7 +73,7 @@ static ssize_t debug_input(struct file *file, const char __user *user_buf,
73static int debug_open(struct inode *inode, struct file *file); 73static int debug_open(struct inode *inode, struct file *file);
74static int debug_close(struct inode *inode, struct file *file); 74static int debug_close(struct inode *inode, struct file *file);
75static debug_info_t* debug_info_create(char *name, int pages_per_area, 75static debug_info_t* debug_info_create(char *name, int pages_per_area,
76 int nr_areas, int buf_size); 76 int nr_areas, int buf_size, mode_t mode);
77static void debug_info_get(debug_info_t *); 77static void debug_info_get(debug_info_t *);
78static void debug_info_put(debug_info_t *); 78static void debug_info_put(debug_info_t *);
79static int debug_prolog_level_fn(debug_info_t * id, 79static int debug_prolog_level_fn(debug_info_t * id,
@@ -157,7 +157,7 @@ struct debug_view debug_sprintf_view = {
157}; 157};
158 158
159/* used by dump analysis tools to determine version of debug feature */ 159/* used by dump analysis tools to determine version of debug feature */
160unsigned int debug_feature_version = __DEBUG_FEATURE_VERSION; 160static unsigned int __used debug_feature_version = __DEBUG_FEATURE_VERSION;
161 161
162/* static globals */ 162/* static globals */
163 163
@@ -327,7 +327,8 @@ debug_info_free(debug_info_t* db_info){
327 */ 327 */
328 328
329static debug_info_t* 329static debug_info_t*
330debug_info_create(char *name, int pages_per_area, int nr_areas, int buf_size) 330debug_info_create(char *name, int pages_per_area, int nr_areas, int buf_size,
331 mode_t mode)
331{ 332{
332 debug_info_t* rc; 333 debug_info_t* rc;
333 334
@@ -336,6 +337,8 @@ debug_info_create(char *name, int pages_per_area, int nr_areas, int buf_size)
336 if(!rc) 337 if(!rc)
337 goto out; 338 goto out;
338 339
340 rc->mode = mode & ~S_IFMT;
341
339 /* create root directory */ 342 /* create root directory */
340 rc->debugfs_root_entry = debugfs_create_dir(rc->name, 343 rc->debugfs_root_entry = debugfs_create_dir(rc->name,
341 debug_debugfs_root_entry); 344 debug_debugfs_root_entry);
@@ -676,23 +679,30 @@ debug_close(struct inode *inode, struct file *file)
676} 679}
677 680
678/* 681/*
679 * debug_register: 682 * debug_register_mode:
680 * - creates and initializes debug area for the caller 683 * - Creates and initializes debug area for the caller
681 * - returns handle for debug area 684 * The mode parameter allows to specify access rights for the s390dbf files
685 * - Returns handle for debug area
682 */ 686 */
683 687
684debug_info_t* 688debug_info_t *debug_register_mode(char *name, int pages_per_area, int nr_areas,
685debug_register (char *name, int pages_per_area, int nr_areas, int buf_size) 689 int buf_size, mode_t mode, uid_t uid,
690 gid_t gid)
686{ 691{
687 debug_info_t *rc = NULL; 692 debug_info_t *rc = NULL;
688 693
694 /* Since debugfs currently does not support uid/gid other than root, */
695 /* we do not allow gid/uid != 0 until we get support for that. */
696 if ((uid != 0) || (gid != 0))
697 printk(KERN_WARNING "debug: Warning - Currently only uid/gid "
698 "= 0 are supported. Using root as owner now!");
689 if (!initialized) 699 if (!initialized)
690 BUG(); 700 BUG();
691 mutex_lock(&debug_mutex); 701 mutex_lock(&debug_mutex);
692 702
693 /* create new debug_info */ 703 /* create new debug_info */
694 704
695 rc = debug_info_create(name, pages_per_area, nr_areas, buf_size); 705 rc = debug_info_create(name, pages_per_area, nr_areas, buf_size, mode);
696 if(!rc) 706 if(!rc)
697 goto out; 707 goto out;
698 debug_register_view(rc, &debug_level_view); 708 debug_register_view(rc, &debug_level_view);
@@ -705,6 +715,20 @@ out:
705 mutex_unlock(&debug_mutex); 715 mutex_unlock(&debug_mutex);
706 return rc; 716 return rc;
707} 717}
718EXPORT_SYMBOL(debug_register_mode);
719
720/*
721 * debug_register:
722 * - creates and initializes debug area for the caller
723 * - returns handle for debug area
724 */
725
726debug_info_t *debug_register(char *name, int pages_per_area, int nr_areas,
727 int buf_size)
728{
729 return debug_register_mode(name, pages_per_area, nr_areas, buf_size,
730 S_IRUSR | S_IWUSR, 0, 0);
731}
708 732
709/* 733/*
710 * debug_unregister: 734 * debug_unregister:
@@ -1073,15 +1097,16 @@ debug_register_view(debug_info_t * id, struct debug_view *view)
1073 int rc = 0; 1097 int rc = 0;
1074 int i; 1098 int i;
1075 unsigned long flags; 1099 unsigned long flags;
1076 mode_t mode = S_IFREG; 1100 mode_t mode;
1077 struct dentry *pde; 1101 struct dentry *pde;
1078 1102
1079 if (!id) 1103 if (!id)
1080 goto out; 1104 goto out;
1081 if (view->prolog_proc || view->format_proc || view->header_proc) 1105 mode = (id->mode | S_IFREG) & ~S_IXUGO;
1082 mode |= S_IRUSR; 1106 if (!(view->prolog_proc || view->format_proc || view->header_proc))
1083 if (view->input_proc) 1107 mode &= ~(S_IRUSR | S_IRGRP | S_IROTH);
1084 mode |= S_IWUSR; 1108 if (!view->input_proc)
1109 mode &= ~(S_IWUSR | S_IWGRP | S_IWOTH);
1085 pde = debugfs_create_file(view->name, mode, id->debugfs_root_entry, 1110 pde = debugfs_create_file(view->name, mode, id->debugfs_root_entry,
1086 id , &debug_file_ops); 1111 id , &debug_file_ops);
1087 if (!pde){ 1112 if (!pde){
diff --git a/arch/s390/kernel/early.c b/arch/s390/kernel/early.c
index 01832c440636..540a67f979b6 100644
--- a/arch/s390/kernel/early.c
+++ b/arch/s390/kernel/early.c
@@ -21,6 +21,7 @@
21#include <asm/setup.h> 21#include <asm/setup.h>
22#include <asm/cpcmd.h> 22#include <asm/cpcmd.h>
23#include <asm/sclp.h> 23#include <asm/sclp.h>
24#include "entry.h"
24 25
25/* 26/*
26 * Create a Kernel NSS if the SAVESYS= parameter is defined 27 * Create a Kernel NSS if the SAVESYS= parameter is defined
diff --git a/arch/s390/kernel/entry.h b/arch/s390/kernel/entry.h
new file mode 100644
index 000000000000..6b1896345eda
--- /dev/null
+++ b/arch/s390/kernel/entry.h
@@ -0,0 +1,60 @@
1#ifndef _ENTRY_H
2#define _ENTRY_H
3
4#include <linux/types.h>
5#include <linux/signal.h>
6#include <asm/ptrace.h>
7
8typedef void pgm_check_handler_t(struct pt_regs *, long);
9extern pgm_check_handler_t *pgm_check_table[128];
10pgm_check_handler_t do_protection_exception;
11pgm_check_handler_t do_dat_exception;
12
13extern int sysctl_userprocess_debug;
14
15void do_single_step(struct pt_regs *regs);
16void syscall_trace(struct pt_regs *regs, int entryexit);
17void kernel_stack_overflow(struct pt_regs * regs);
18void do_signal(struct pt_regs *regs);
19int handle_signal32(unsigned long sig, struct k_sigaction *ka,
20 siginfo_t *info, sigset_t *oldset, struct pt_regs *regs);
21
22void do_extint(struct pt_regs *regs, unsigned short code);
23int __cpuinit start_secondary(void *cpuvoid);
24void __init startup_init(void);
25void die(const char * str, struct pt_regs * regs, long err);
26
27struct new_utsname;
28struct mmap_arg_struct;
29struct fadvise64_64_args;
30struct old_sigaction;
31struct sel_arg_struct;
32
33long sys_pipe(unsigned long __user *fildes);
34long sys_mmap2(struct mmap_arg_struct __user *arg);
35long old_mmap(struct mmap_arg_struct __user *arg);
36long sys_ipc(uint call, int first, unsigned long second,
37 unsigned long third, void __user *ptr);
38long s390x_newuname(struct new_utsname __user *name);
39long s390x_personality(unsigned long personality);
40long s390_fadvise64(int fd, u32 offset_high, u32 offset_low,
41 size_t len, int advice);
42long s390_fadvise64_64(struct fadvise64_64_args __user *args);
43long s390_fallocate(int fd, int mode, loff_t offset, u32 len_high, u32 len_low);
44long sys_fork(void);
45long sys_clone(void);
46long sys_vfork(void);
47void execve_tail(void);
48long sys_execve(void);
49int sys_sigsuspend(int history0, int history1, old_sigset_t mask);
50long sys_sigaction(int sig, const struct old_sigaction __user *act,
51 struct old_sigaction __user *oact);
52long sys_sigaltstack(const stack_t __user *uss, stack_t __user *uoss);
53long sys_sigreturn(void);
54long sys_rt_sigreturn(void);
55long sys32_sigreturn(void);
56long sys32_rt_sigreturn(void);
57long old_select(struct sel_arg_struct __user *arg);
58long sys_ptrace(long request, long pid, long addr, long data);
59
60#endif /* _ENTRY_H */
diff --git a/arch/s390/kernel/entry64.S b/arch/s390/kernel/entry64.S
index efde6e178f6c..cd959c0b2e16 100644
--- a/arch/s390/kernel/entry64.S
+++ b/arch/s390/kernel/entry64.S
@@ -475,6 +475,7 @@ pgm_check_handler:
475pgm_no_vtime: 475pgm_no_vtime:
476#endif 476#endif
477 lg %r9,__LC_THREAD_INFO # load pointer to thread_info struct 477 lg %r9,__LC_THREAD_INFO # load pointer to thread_info struct
478 mvc SP_ARGS(8,%r15),__LC_LAST_BREAK
478 TRACE_IRQS_OFF 479 TRACE_IRQS_OFF
479 lgf %r3,__LC_PGM_ILC # load program interruption code 480 lgf %r3,__LC_PGM_ILC # load program interruption code
480 lghi %r8,0x7f 481 lghi %r8,0x7f
@@ -847,6 +848,7 @@ stack_overflow:
847 je 0f 848 je 0f
848 la %r1,__LC_SAVE_AREA+32 849 la %r1,__LC_SAVE_AREA+32
8490: mvc SP_R12(32,%r15),0(%r1) # move %r12-%r15 to stack 8500: mvc SP_R12(32,%r15),0(%r1) # move %r12-%r15 to stack
851 mvc SP_ARGS(8,%r15),__LC_LAST_BREAK
850 xc __SF_BACKCHAIN(8,%r15),__SF_BACKCHAIN(%r15) # clear back chain 852 xc __SF_BACKCHAIN(8,%r15),__SF_BACKCHAIN(%r15) # clear back chain
851 la %r2,SP_PTREGS(%r15) # load pt_regs 853 la %r2,SP_PTREGS(%r15) # load pt_regs
852 jg kernel_stack_overflow 854 jg kernel_stack_overflow
diff --git a/arch/s390/kernel/ipl.c b/arch/s390/kernel/ipl.c
index 375232c46c7a..532542447d66 100644
--- a/arch/s390/kernel/ipl.c
+++ b/arch/s390/kernel/ipl.c
@@ -655,7 +655,7 @@ static struct kobj_attribute reipl_type_attr =
655 655
656static struct kset *reipl_kset; 656static struct kset *reipl_kset;
657 657
658void reipl_run(struct shutdown_trigger *trigger) 658static void reipl_run(struct shutdown_trigger *trigger)
659{ 659{
660 struct ccw_dev_id devid; 660 struct ccw_dev_id devid;
661 static char buf[100]; 661 static char buf[100];
diff --git a/arch/s390/kernel/kprobes.c b/arch/s390/kernel/kprobes.c
index c5549a206284..ed04d1372d5d 100644
--- a/arch/s390/kernel/kprobes.c
+++ b/arch/s390/kernel/kprobes.c
@@ -360,7 +360,7 @@ no_kprobe:
360 * - When the probed function returns, this probe 360 * - When the probed function returns, this probe
361 * causes the handlers to fire 361 * causes the handlers to fire
362 */ 362 */
363void kretprobe_trampoline_holder(void) 363static void __used kretprobe_trampoline_holder(void)
364{ 364{
365 asm volatile(".global kretprobe_trampoline\n" 365 asm volatile(".global kretprobe_trampoline\n"
366 "kretprobe_trampoline: bcr 0,0\n"); 366 "kretprobe_trampoline: bcr 0,0\n");
diff --git a/arch/s390/kernel/process.c b/arch/s390/kernel/process.c
index ce203154d8ce..c1aff194141d 100644
--- a/arch/s390/kernel/process.c
+++ b/arch/s390/kernel/process.c
@@ -36,6 +36,8 @@
36#include <linux/module.h> 36#include <linux/module.h>
37#include <linux/notifier.h> 37#include <linux/notifier.h>
38#include <linux/utsname.h> 38#include <linux/utsname.h>
39#include <linux/tick.h>
40#include <linux/elfcore.h>
39#include <asm/uaccess.h> 41#include <asm/uaccess.h>
40#include <asm/pgtable.h> 42#include <asm/pgtable.h>
41#include <asm/system.h> 43#include <asm/system.h>
@@ -44,6 +46,7 @@
44#include <asm/irq.h> 46#include <asm/irq.h>
45#include <asm/timer.h> 47#include <asm/timer.h>
46#include <asm/cpu.h> 48#include <asm/cpu.h>
49#include "entry.h"
47 50
48asmlinkage void ret_from_fork(void) asm ("ret_from_fork"); 51asmlinkage void ret_from_fork(void) asm ("ret_from_fork");
49 52
@@ -76,6 +79,7 @@ unsigned long thread_saved_pc(struct task_struct *tsk)
76 * Need to know about CPUs going idle? 79 * Need to know about CPUs going idle?
77 */ 80 */
78static ATOMIC_NOTIFIER_HEAD(idle_chain); 81static ATOMIC_NOTIFIER_HEAD(idle_chain);
82DEFINE_PER_CPU(struct s390_idle_data, s390_idle);
79 83
80int register_idle_notifier(struct notifier_block *nb) 84int register_idle_notifier(struct notifier_block *nb)
81{ 85{
@@ -89,9 +93,33 @@ int unregister_idle_notifier(struct notifier_block *nb)
89} 93}
90EXPORT_SYMBOL(unregister_idle_notifier); 94EXPORT_SYMBOL(unregister_idle_notifier);
91 95
92void do_monitor_call(struct pt_regs *regs, long interruption_code) 96static int s390_idle_enter(void)
97{
98 struct s390_idle_data *idle;
99 int nr_calls = 0;
100 void *hcpu;
101 int rc;
102
103 hcpu = (void *)(long)smp_processor_id();
104 rc = __atomic_notifier_call_chain(&idle_chain, S390_CPU_IDLE, hcpu, -1,
105 &nr_calls);
106 if (rc == NOTIFY_BAD) {
107 nr_calls--;
108 __atomic_notifier_call_chain(&idle_chain, S390_CPU_NOT_IDLE,
109 hcpu, nr_calls, NULL);
110 return rc;
111 }
112 idle = &__get_cpu_var(s390_idle);
113 spin_lock(&idle->lock);
114 idle->idle_count++;
115 idle->in_idle = 1;
116 idle->idle_enter = get_clock();
117 spin_unlock(&idle->lock);
118 return NOTIFY_OK;
119}
120
121void s390_idle_leave(void)
93{ 122{
94#ifdef CONFIG_SMP
95 struct s390_idle_data *idle; 123 struct s390_idle_data *idle;
96 124
97 idle = &__get_cpu_var(s390_idle); 125 idle = &__get_cpu_var(s390_idle);
@@ -99,10 +127,6 @@ void do_monitor_call(struct pt_regs *regs, long interruption_code)
99 idle->idle_time += get_clock() - idle->idle_enter; 127 idle->idle_time += get_clock() - idle->idle_enter;
100 idle->in_idle = 0; 128 idle->in_idle = 0;
101 spin_unlock(&idle->lock); 129 spin_unlock(&idle->lock);
102#endif
103 /* disable monitor call class 0 */
104 __ctl_clear_bit(8, 15);
105
106 atomic_notifier_call_chain(&idle_chain, S390_CPU_NOT_IDLE, 130 atomic_notifier_call_chain(&idle_chain, S390_CPU_NOT_IDLE,
107 (void *)(long) smp_processor_id()); 131 (void *)(long) smp_processor_id());
108} 132}
@@ -113,61 +137,30 @@ extern void s390_handle_mcck(void);
113 */ 137 */
114static void default_idle(void) 138static void default_idle(void)
115{ 139{
116 int cpu, rc;
117 int nr_calls = 0;
118 void *hcpu;
119#ifdef CONFIG_SMP
120 struct s390_idle_data *idle;
121#endif
122
123 /* CPU is going idle. */ 140 /* CPU is going idle. */
124 cpu = smp_processor_id();
125 hcpu = (void *)(long)cpu;
126 local_irq_disable(); 141 local_irq_disable();
127 if (need_resched()) { 142 if (need_resched()) {
128 local_irq_enable(); 143 local_irq_enable();
129 return; 144 return;
130 } 145 }
131 146 if (s390_idle_enter() == NOTIFY_BAD) {
132 rc = __atomic_notifier_call_chain(&idle_chain, S390_CPU_IDLE, hcpu, -1,
133 &nr_calls);
134 if (rc == NOTIFY_BAD) {
135 nr_calls--;
136 __atomic_notifier_call_chain(&idle_chain, S390_CPU_NOT_IDLE,
137 hcpu, nr_calls, NULL);
138 local_irq_enable(); 147 local_irq_enable();
139 return; 148 return;
140 } 149 }
141
142 /* enable monitor call class 0 */
143 __ctl_set_bit(8, 15);
144
145#ifdef CONFIG_HOTPLUG_CPU 150#ifdef CONFIG_HOTPLUG_CPU
146 if (cpu_is_offline(cpu)) { 151 if (cpu_is_offline(smp_processor_id())) {
147 preempt_enable_no_resched(); 152 preempt_enable_no_resched();
148 cpu_die(); 153 cpu_die();
149 } 154 }
150#endif 155#endif
151
152 local_mcck_disable(); 156 local_mcck_disable();
153 if (test_thread_flag(TIF_MCCK_PENDING)) { 157 if (test_thread_flag(TIF_MCCK_PENDING)) {
154 local_mcck_enable(); 158 local_mcck_enable();
155 /* disable monitor call class 0 */ 159 s390_idle_leave();
156 __ctl_clear_bit(8, 15);
157 atomic_notifier_call_chain(&idle_chain, S390_CPU_NOT_IDLE,
158 hcpu);
159 local_irq_enable(); 160 local_irq_enable();
160 s390_handle_mcck(); 161 s390_handle_mcck();
161 return; 162 return;
162 } 163 }
163#ifdef CONFIG_SMP
164 idle = &__get_cpu_var(s390_idle);
165 spin_lock(&idle->lock);
166 idle->idle_count++;
167 idle->in_idle = 1;
168 idle->idle_enter = get_clock();
169 spin_unlock(&idle->lock);
170#endif
171 trace_hardirqs_on(); 164 trace_hardirqs_on();
172 /* Wait for external, I/O or machine check interrupt. */ 165 /* Wait for external, I/O or machine check interrupt. */
173 __load_psw_mask(psw_kernel_bits | PSW_MASK_WAIT | 166 __load_psw_mask(psw_kernel_bits | PSW_MASK_WAIT |
@@ -177,9 +170,10 @@ static void default_idle(void)
177void cpu_idle(void) 170void cpu_idle(void)
178{ 171{
179 for (;;) { 172 for (;;) {
173 tick_nohz_stop_sched_tick();
180 while (!need_resched()) 174 while (!need_resched())
181 default_idle(); 175 default_idle();
182 176 tick_nohz_restart_sched_tick();
183 preempt_enable_no_resched(); 177 preempt_enable_no_resched();
184 schedule(); 178 schedule();
185 preempt_disable(); 179 preempt_disable();
@@ -201,6 +195,7 @@ void show_regs(struct pt_regs *regs)
201 /* Show stack backtrace if pt_regs is from kernel mode */ 195 /* Show stack backtrace if pt_regs is from kernel mode */
202 if (!(regs->psw.mask & PSW_MASK_PSTATE)) 196 if (!(regs->psw.mask & PSW_MASK_PSTATE))
203 show_trace(NULL, (unsigned long *) regs->gprs[15]); 197 show_trace(NULL, (unsigned long *) regs->gprs[15]);
198 show_last_breaking_event(regs);
204} 199}
205 200
206extern void kernel_thread_starter(void); 201extern void kernel_thread_starter(void);
diff --git a/arch/s390/kernel/ptrace.c b/arch/s390/kernel/ptrace.c
index 6e036bae9875..58a064296987 100644
--- a/arch/s390/kernel/ptrace.c
+++ b/arch/s390/kernel/ptrace.c
@@ -41,6 +41,7 @@
41#include <asm/system.h> 41#include <asm/system.h>
42#include <asm/uaccess.h> 42#include <asm/uaccess.h>
43#include <asm/unistd.h> 43#include <asm/unistd.h>
44#include "entry.h"
44 45
45#ifdef CONFIG_COMPAT 46#ifdef CONFIG_COMPAT
46#include "compat_ptrace.h" 47#include "compat_ptrace.h"
diff --git a/arch/s390/kernel/s390_ext.c b/arch/s390/kernel/s390_ext.c
index acf93dba7727..e019b419efc6 100644
--- a/arch/s390/kernel/s390_ext.c
+++ b/arch/s390/kernel/s390_ext.c
@@ -13,11 +13,12 @@
13#include <linux/errno.h> 13#include <linux/errno.h>
14#include <linux/kernel_stat.h> 14#include <linux/kernel_stat.h>
15#include <linux/interrupt.h> 15#include <linux/interrupt.h>
16 16#include <asm/cpu.h>
17#include <asm/lowcore.h> 17#include <asm/lowcore.h>
18#include <asm/s390_ext.h> 18#include <asm/s390_ext.h>
19#include <asm/irq_regs.h> 19#include <asm/irq_regs.h>
20#include <asm/irq.h> 20#include <asm/irq.h>
21#include "entry.h"
21 22
22/* 23/*
23 * ext_int_hash[index] is the start of the list for all external interrupts 24 * ext_int_hash[index] is the start of the list for all external interrupts
@@ -119,13 +120,10 @@ void do_extint(struct pt_regs *regs, unsigned short code)
119 120
120 old_regs = set_irq_regs(regs); 121 old_regs = set_irq_regs(regs);
121 irq_enter(); 122 irq_enter();
122 asm volatile ("mc 0,0"); 123 s390_idle_check();
123 if (S390_lowcore.int_clock >= S390_lowcore.jiffy_timer) 124 if (S390_lowcore.int_clock >= S390_lowcore.clock_comparator)
124 /** 125 /* Serve timer interrupts first. */
125 * Make sure that the i/o interrupt did not "overtake" 126 clock_comparator_work();
126 * the last HZ timer interrupt.
127 */
128 account_ticks(S390_lowcore.int_clock);
129 kstat_cpu(smp_processor_id()).irqs[EXTERNAL_INTERRUPT]++; 127 kstat_cpu(smp_processor_id()).irqs[EXTERNAL_INTERRUPT]++;
130 index = ext_hash(code); 128 index = ext_hash(code);
131 for (p = ext_int_hash[index]; p; p = p->next) { 129 for (p = ext_int_hash[index]; p; p = p->next) {
diff --git a/arch/s390/kernel/s390_ksyms.c b/arch/s390/kernel/s390_ksyms.c
index 7234c737f825..48238a114ce9 100644
--- a/arch/s390/kernel/s390_ksyms.c
+++ b/arch/s390/kernel/s390_ksyms.c
@@ -27,13 +27,6 @@ EXPORT_SYMBOL(_zb_findmap);
27EXPORT_SYMBOL(_sb_findmap); 27EXPORT_SYMBOL(_sb_findmap);
28 28
29/* 29/*
30 * semaphore ops
31 */
32EXPORT_SYMBOL(__up);
33EXPORT_SYMBOL(__down);
34EXPORT_SYMBOL(__down_interruptible);
35
36/*
37 * binfmt_elf loader 30 * binfmt_elf loader
38 */ 31 */
39extern int dump_fpu (struct pt_regs * regs, s390_fp_regs *fpregs); 32extern int dump_fpu (struct pt_regs * regs, s390_fp_regs *fpregs);
diff --git a/arch/s390/kernel/semaphore.c b/arch/s390/kernel/semaphore.c
deleted file mode 100644
index 191303f6c1d8..000000000000
--- a/arch/s390/kernel/semaphore.c
+++ /dev/null
@@ -1,108 +0,0 @@
1/*
2 * linux/arch/s390/kernel/semaphore.c
3 *
4 * S390 version
5 * Copyright (C) 1998-2000 IBM Corporation
6 * Author(s): Martin Schwidefsky
7 *
8 * Derived from "linux/arch/i386/kernel/semaphore.c
9 * Copyright (C) 1999, Linus Torvalds
10 *
11 */
12#include <linux/sched.h>
13#include <linux/errno.h>
14#include <linux/init.h>
15
16#include <asm/semaphore.h>
17
18/*
19 * Atomically update sem->count. Equivalent to:
20 * old_val = sem->count.counter;
21 * new_val = ((old_val >= 0) ? old_val : 0) + incr;
22 * sem->count.counter = new_val;
23 * return old_val;
24 */
25static inline int __sem_update_count(struct semaphore *sem, int incr)
26{
27 int old_val, new_val;
28
29 asm volatile(
30 " l %0,0(%3)\n"
31 "0: ltr %1,%0\n"
32 " jhe 1f\n"
33 " lhi %1,0\n"
34 "1: ar %1,%4\n"
35 " cs %0,%1,0(%3)\n"
36 " jl 0b\n"
37 : "=&d" (old_val), "=&d" (new_val), "=m" (sem->count)
38 : "a" (&sem->count), "d" (incr), "m" (sem->count)
39 : "cc");
40 return old_val;
41}
42
43/*
44 * The inline function up() incremented count but the result
45 * was <= 0. This indicates that some process is waiting on
46 * the semaphore. The semaphore is free and we'll wake the
47 * first sleeping process, so we set count to 1 unless some
48 * other cpu has called up in the meantime in which case
49 * we just increment count by 1.
50 */
51void __up(struct semaphore *sem)
52{
53 __sem_update_count(sem, 1);
54 wake_up(&sem->wait);
55}
56
57/*
58 * The inline function down() decremented count and the result
59 * was < 0. The wait loop will atomically test and update the
60 * semaphore counter following the rules:
61 * count > 0: decrement count, wake up queue and exit.
62 * count <= 0: set count to -1, go to sleep.
63 */
64void __sched __down(struct semaphore * sem)
65{
66 struct task_struct *tsk = current;
67 DECLARE_WAITQUEUE(wait, tsk);
68
69 __set_task_state(tsk, TASK_UNINTERRUPTIBLE);
70 add_wait_queue_exclusive(&sem->wait, &wait);
71 while (__sem_update_count(sem, -1) <= 0) {
72 schedule();
73 set_task_state(tsk, TASK_UNINTERRUPTIBLE);
74 }
75 remove_wait_queue(&sem->wait, &wait);
76 __set_task_state(tsk, TASK_RUNNING);
77 wake_up(&sem->wait);
78}
79
80/*
81 * Same as __down() with an additional test for signals.
82 * If a signal is pending the count is updated as follows:
83 * count > 0: wake up queue and exit.
84 * count <= 0: set count to 0, wake up queue and exit.
85 */
86int __sched __down_interruptible(struct semaphore * sem)
87{
88 int retval = 0;
89 struct task_struct *tsk = current;
90 DECLARE_WAITQUEUE(wait, tsk);
91
92 __set_task_state(tsk, TASK_INTERRUPTIBLE);
93 add_wait_queue_exclusive(&sem->wait, &wait);
94 while (__sem_update_count(sem, -1) <= 0) {
95 if (signal_pending(current)) {
96 __sem_update_count(sem, 0);
97 retval = -EINTR;
98 break;
99 }
100 schedule();
101 set_task_state(tsk, TASK_INTERRUPTIBLE);
102 }
103 remove_wait_queue(&sem->wait, &wait);
104 __set_task_state(tsk, TASK_RUNNING);
105 wake_up(&sem->wait);
106 return retval;
107}
108
diff --git a/arch/s390/kernel/setup.c b/arch/s390/kernel/setup.c
index 290e504061a3..7141147e6b63 100644
--- a/arch/s390/kernel/setup.c
+++ b/arch/s390/kernel/setup.c
@@ -39,6 +39,7 @@
39#include <linux/pfn.h> 39#include <linux/pfn.h>
40#include <linux/ctype.h> 40#include <linux/ctype.h>
41#include <linux/reboot.h> 41#include <linux/reboot.h>
42#include <linux/topology.h>
42 43
43#include <asm/ipl.h> 44#include <asm/ipl.h>
44#include <asm/uaccess.h> 45#include <asm/uaccess.h>
@@ -427,7 +428,7 @@ setup_lowcore(void)
427 lc->io_new_psw.mask = psw_kernel_bits; 428 lc->io_new_psw.mask = psw_kernel_bits;
428 lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler; 429 lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
429 lc->ipl_device = S390_lowcore.ipl_device; 430 lc->ipl_device = S390_lowcore.ipl_device;
430 lc->jiffy_timer = -1LL; 431 lc->clock_comparator = -1ULL;
431 lc->kernel_stack = ((unsigned long) &init_thread_union) + THREAD_SIZE; 432 lc->kernel_stack = ((unsigned long) &init_thread_union) + THREAD_SIZE;
432 lc->async_stack = (unsigned long) 433 lc->async_stack = (unsigned long)
433 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) + ASYNC_SIZE; 434 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) + ASYNC_SIZE;
@@ -687,7 +688,7 @@ static __init unsigned int stfl(void)
687 return S390_lowcore.stfl_fac_list; 688 return S390_lowcore.stfl_fac_list;
688} 689}
689 690
690static __init int stfle(unsigned long long *list, int doublewords) 691static int __init __stfle(unsigned long long *list, int doublewords)
691{ 692{
692 typedef struct { unsigned long long _[doublewords]; } addrtype; 693 typedef struct { unsigned long long _[doublewords]; } addrtype;
693 register unsigned long __nr asm("0") = doublewords - 1; 694 register unsigned long __nr asm("0") = doublewords - 1;
@@ -697,6 +698,13 @@ static __init int stfle(unsigned long long *list, int doublewords)
697 return __nr + 1; 698 return __nr + 1;
698} 699}
699 700
701int __init stfle(unsigned long long *list, int doublewords)
702{
703 if (!(stfl() & (1UL << 24)))
704 return -EOPNOTSUPP;
705 return __stfle(list, doublewords);
706}
707
700/* 708/*
701 * Setup hardware capabilities. 709 * Setup hardware capabilities.
702 */ 710 */
@@ -741,7 +749,7 @@ static void __init setup_hwcaps(void)
741 * HWCAP_S390_DFP bit 6. 749 * HWCAP_S390_DFP bit 6.
742 */ 750 */
743 if ((elf_hwcap & (1UL << 2)) && 751 if ((elf_hwcap & (1UL << 2)) &&
744 stfle(&facility_list_extended, 1) > 0) { 752 __stfle(&facility_list_extended, 1) > 0) {
745 if (facility_list_extended & (1ULL << (64 - 43))) 753 if (facility_list_extended & (1ULL << (64 - 43)))
746 elf_hwcap |= 1UL << 6; 754 elf_hwcap |= 1UL << 6;
747 } 755 }
@@ -823,6 +831,7 @@ setup_arch(char **cmdline_p)
823 831
824 cpu_init(); 832 cpu_init();
825 __cpu_logical_map[0] = S390_lowcore.cpu_data.cpu_addr; 833 __cpu_logical_map[0] = S390_lowcore.cpu_data.cpu_addr;
834 s390_init_cpu_topology();
826 835
827 /* 836 /*
828 * Setup capabilities (ELF_HWCAP & ELF_PLATFORM). 837 * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
diff --git a/arch/s390/kernel/signal.c b/arch/s390/kernel/signal.c
index 4449bf32cbf1..b97682040215 100644
--- a/arch/s390/kernel/signal.c
+++ b/arch/s390/kernel/signal.c
@@ -27,6 +27,7 @@
27#include <asm/ucontext.h> 27#include <asm/ucontext.h>
28#include <asm/uaccess.h> 28#include <asm/uaccess.h>
29#include <asm/lowcore.h> 29#include <asm/lowcore.h>
30#include "entry.h"
30 31
31#define _BLOCKABLE (~(sigmask(SIGKILL) | sigmask(SIGSTOP))) 32#define _BLOCKABLE (~(sigmask(SIGKILL) | sigmask(SIGSTOP)))
32 33
@@ -235,6 +236,10 @@ get_sigframe(struct k_sigaction *ka, struct pt_regs * regs, size_t frame_size)
235 /* Default to using normal stack */ 236 /* Default to using normal stack */
236 sp = regs->gprs[15]; 237 sp = regs->gprs[15];
237 238
239 /* Overflow on alternate signal stack gives SIGSEGV. */
240 if (on_sig_stack(sp) && !on_sig_stack((sp - frame_size) & -8UL))
241 return (void __user *) -1UL;
242
238 /* This is the X/Open sanctioned signal stack switching. */ 243 /* This is the X/Open sanctioned signal stack switching. */
239 if (ka->sa.sa_flags & SA_ONSTACK) { 244 if (ka->sa.sa_flags & SA_ONSTACK) {
240 if (! sas_ss_flags(sp)) 245 if (! sas_ss_flags(sp))
@@ -270,6 +275,9 @@ static int setup_frame(int sig, struct k_sigaction *ka,
270 if (!access_ok(VERIFY_WRITE, frame, sizeof(sigframe))) 275 if (!access_ok(VERIFY_WRITE, frame, sizeof(sigframe)))
271 goto give_sigsegv; 276 goto give_sigsegv;
272 277
278 if (frame == (void __user *) -1UL)
279 goto give_sigsegv;
280
273 if (__copy_to_user(&frame->sc.oldmask, &set->sig, _SIGMASK_COPY_SIZE)) 281 if (__copy_to_user(&frame->sc.oldmask, &set->sig, _SIGMASK_COPY_SIZE))
274 goto give_sigsegv; 282 goto give_sigsegv;
275 283
@@ -327,6 +335,9 @@ static int setup_rt_frame(int sig, struct k_sigaction *ka, siginfo_t *info,
327 if (!access_ok(VERIFY_WRITE, frame, sizeof(rt_sigframe))) 335 if (!access_ok(VERIFY_WRITE, frame, sizeof(rt_sigframe)))
328 goto give_sigsegv; 336 goto give_sigsegv;
329 337
338 if (frame == (void __user *) -1UL)
339 goto give_sigsegv;
340
330 if (copy_siginfo_to_user(&frame->info, info)) 341 if (copy_siginfo_to_user(&frame->info, info))
331 goto give_sigsegv; 342 goto give_sigsegv;
332 343
@@ -474,11 +485,6 @@ void do_signal(struct pt_regs *regs)
474 int ret; 485 int ret;
475#ifdef CONFIG_COMPAT 486#ifdef CONFIG_COMPAT
476 if (test_thread_flag(TIF_31BIT)) { 487 if (test_thread_flag(TIF_31BIT)) {
477 extern int handle_signal32(unsigned long sig,
478 struct k_sigaction *ka,
479 siginfo_t *info,
480 sigset_t *oldset,
481 struct pt_regs *regs);
482 ret = handle_signal32(signr, &ka, &info, oldset, regs); 488 ret = handle_signal32(signr, &ka, &info, oldset, regs);
483 } 489 }
484 else 490 else
diff --git a/arch/s390/kernel/smp.c b/arch/s390/kernel/smp.c
index 8f894d380a62..0dfa988c1b26 100644
--- a/arch/s390/kernel/smp.c
+++ b/arch/s390/kernel/smp.c
@@ -44,6 +44,7 @@
44#include <asm/lowcore.h> 44#include <asm/lowcore.h>
45#include <asm/sclp.h> 45#include <asm/sclp.h>
46#include <asm/cpu.h> 46#include <asm/cpu.h>
47#include "entry.h"
47 48
48/* 49/*
49 * An array with a pointer the lowcore of every CPU. 50 * An array with a pointer the lowcore of every CPU.
@@ -67,13 +68,12 @@ enum s390_cpu_state {
67 CPU_STATE_CONFIGURED, 68 CPU_STATE_CONFIGURED,
68}; 69};
69 70
70#ifdef CONFIG_HOTPLUG_CPU 71DEFINE_MUTEX(smp_cpu_state_mutex);
71static DEFINE_MUTEX(smp_cpu_state_mutex); 72int smp_cpu_polarization[NR_CPUS];
72#endif
73static int smp_cpu_state[NR_CPUS]; 73static int smp_cpu_state[NR_CPUS];
74static int cpu_management;
74 75
75static DEFINE_PER_CPU(struct cpu, cpu_devices); 76static DEFINE_PER_CPU(struct cpu, cpu_devices);
76DEFINE_PER_CPU(struct s390_idle_data, s390_idle);
77 77
78static void smp_ext_bitcall(int, ec_bit_sig); 78static void smp_ext_bitcall(int, ec_bit_sig);
79 79
@@ -298,7 +298,7 @@ static void smp_ext_bitcall(int cpu, ec_bit_sig sig)
298/* 298/*
299 * this function sends a 'purge tlb' signal to another CPU. 299 * this function sends a 'purge tlb' signal to another CPU.
300 */ 300 */
301void smp_ptlb_callback(void *info) 301static void smp_ptlb_callback(void *info)
302{ 302{
303 __tlb_flush_local(); 303 __tlb_flush_local();
304} 304}
@@ -456,6 +456,7 @@ static int smp_rescan_cpus_sigp(cpumask_t avail)
456 if (cpu_known(cpu_id)) 456 if (cpu_known(cpu_id))
457 continue; 457 continue;
458 __cpu_logical_map[logical_cpu] = cpu_id; 458 __cpu_logical_map[logical_cpu] = cpu_id;
459 smp_cpu_polarization[logical_cpu] = POLARIZATION_UNKNWN;
459 if (!cpu_stopped(logical_cpu)) 460 if (!cpu_stopped(logical_cpu))
460 continue; 461 continue;
461 cpu_set(logical_cpu, cpu_present_map); 462 cpu_set(logical_cpu, cpu_present_map);
@@ -489,6 +490,7 @@ static int smp_rescan_cpus_sclp(cpumask_t avail)
489 if (cpu_known(cpu_id)) 490 if (cpu_known(cpu_id))
490 continue; 491 continue;
491 __cpu_logical_map[logical_cpu] = cpu_id; 492 __cpu_logical_map[logical_cpu] = cpu_id;
493 smp_cpu_polarization[logical_cpu] = POLARIZATION_UNKNWN;
492 cpu_set(logical_cpu, cpu_present_map); 494 cpu_set(logical_cpu, cpu_present_map);
493 if (cpu >= info->configured) 495 if (cpu >= info->configured)
494 smp_cpu_state[logical_cpu] = CPU_STATE_STANDBY; 496 smp_cpu_state[logical_cpu] = CPU_STATE_STANDBY;
@@ -846,6 +848,7 @@ void __init smp_prepare_boot_cpu(void)
846 S390_lowcore.percpu_offset = __per_cpu_offset[0]; 848 S390_lowcore.percpu_offset = __per_cpu_offset[0];
847 current_set[0] = current; 849 current_set[0] = current;
848 smp_cpu_state[0] = CPU_STATE_CONFIGURED; 850 smp_cpu_state[0] = CPU_STATE_CONFIGURED;
851 smp_cpu_polarization[0] = POLARIZATION_UNKNWN;
849 spin_lock_init(&(&__get_cpu_var(s390_idle))->lock); 852 spin_lock_init(&(&__get_cpu_var(s390_idle))->lock);
850} 853}
851 854
@@ -897,15 +900,19 @@ static ssize_t cpu_configure_store(struct sys_device *dev, const char *buf,
897 case 0: 900 case 0:
898 if (smp_cpu_state[cpu] == CPU_STATE_CONFIGURED) { 901 if (smp_cpu_state[cpu] == CPU_STATE_CONFIGURED) {
899 rc = sclp_cpu_deconfigure(__cpu_logical_map[cpu]); 902 rc = sclp_cpu_deconfigure(__cpu_logical_map[cpu]);
900 if (!rc) 903 if (!rc) {
901 smp_cpu_state[cpu] = CPU_STATE_STANDBY; 904 smp_cpu_state[cpu] = CPU_STATE_STANDBY;
905 smp_cpu_polarization[cpu] = POLARIZATION_UNKNWN;
906 }
902 } 907 }
903 break; 908 break;
904 case 1: 909 case 1:
905 if (smp_cpu_state[cpu] == CPU_STATE_STANDBY) { 910 if (smp_cpu_state[cpu] == CPU_STATE_STANDBY) {
906 rc = sclp_cpu_configure(__cpu_logical_map[cpu]); 911 rc = sclp_cpu_configure(__cpu_logical_map[cpu]);
907 if (!rc) 912 if (!rc) {
908 smp_cpu_state[cpu] = CPU_STATE_CONFIGURED; 913 smp_cpu_state[cpu] = CPU_STATE_CONFIGURED;
914 smp_cpu_polarization[cpu] = POLARIZATION_UNKNWN;
915 }
909 } 916 }
910 break; 917 break;
911 default: 918 default:
@@ -919,6 +926,34 @@ out:
919static SYSDEV_ATTR(configure, 0644, cpu_configure_show, cpu_configure_store); 926static SYSDEV_ATTR(configure, 0644, cpu_configure_show, cpu_configure_store);
920#endif /* CONFIG_HOTPLUG_CPU */ 927#endif /* CONFIG_HOTPLUG_CPU */
921 928
929static ssize_t cpu_polarization_show(struct sys_device *dev, char *buf)
930{
931 int cpu = dev->id;
932 ssize_t count;
933
934 mutex_lock(&smp_cpu_state_mutex);
935 switch (smp_cpu_polarization[cpu]) {
936 case POLARIZATION_HRZ:
937 count = sprintf(buf, "horizontal\n");
938 break;
939 case POLARIZATION_VL:
940 count = sprintf(buf, "vertical:low\n");
941 break;
942 case POLARIZATION_VM:
943 count = sprintf(buf, "vertical:medium\n");
944 break;
945 case POLARIZATION_VH:
946 count = sprintf(buf, "vertical:high\n");
947 break;
948 default:
949 count = sprintf(buf, "unknown\n");
950 break;
951 }
952 mutex_unlock(&smp_cpu_state_mutex);
953 return count;
954}
955static SYSDEV_ATTR(polarization, 0444, cpu_polarization_show, NULL);
956
922static ssize_t show_cpu_address(struct sys_device *dev, char *buf) 957static ssize_t show_cpu_address(struct sys_device *dev, char *buf)
923{ 958{
924 return sprintf(buf, "%d\n", __cpu_logical_map[dev->id]); 959 return sprintf(buf, "%d\n", __cpu_logical_map[dev->id]);
@@ -931,6 +966,7 @@ static struct attribute *cpu_common_attrs[] = {
931 &attr_configure.attr, 966 &attr_configure.attr,
932#endif 967#endif
933 &attr_address.attr, 968 &attr_address.attr,
969 &attr_polarization.attr,
934 NULL, 970 NULL,
935}; 971};
936 972
@@ -1075,11 +1111,48 @@ static ssize_t __ref rescan_store(struct sys_device *dev,
1075out: 1111out:
1076 put_online_cpus(); 1112 put_online_cpus();
1077 mutex_unlock(&smp_cpu_state_mutex); 1113 mutex_unlock(&smp_cpu_state_mutex);
1114 if (!cpus_empty(newcpus))
1115 topology_schedule_update();
1078 return rc ? rc : count; 1116 return rc ? rc : count;
1079} 1117}
1080static SYSDEV_ATTR(rescan, 0200, NULL, rescan_store); 1118static SYSDEV_ATTR(rescan, 0200, NULL, rescan_store);
1081#endif /* CONFIG_HOTPLUG_CPU */ 1119#endif /* CONFIG_HOTPLUG_CPU */
1082 1120
1121static ssize_t dispatching_show(struct sys_device *dev, char *buf)
1122{
1123 ssize_t count;
1124
1125 mutex_lock(&smp_cpu_state_mutex);
1126 count = sprintf(buf, "%d\n", cpu_management);
1127 mutex_unlock(&smp_cpu_state_mutex);
1128 return count;
1129}
1130
1131static ssize_t dispatching_store(struct sys_device *dev, const char *buf,
1132 size_t count)
1133{
1134 int val, rc;
1135 char delim;
1136
1137 if (sscanf(buf, "%d %c", &val, &delim) != 1)
1138 return -EINVAL;
1139 if (val != 0 && val != 1)
1140 return -EINVAL;
1141 rc = 0;
1142 mutex_lock(&smp_cpu_state_mutex);
1143 get_online_cpus();
1144 if (cpu_management == val)
1145 goto out;
1146 rc = topology_set_cpu_management(val);
1147 if (!rc)
1148 cpu_management = val;
1149out:
1150 put_online_cpus();
1151 mutex_unlock(&smp_cpu_state_mutex);
1152 return rc ? rc : count;
1153}
1154static SYSDEV_ATTR(dispatching, 0644, dispatching_show, dispatching_store);
1155
1083static int __init topology_init(void) 1156static int __init topology_init(void)
1084{ 1157{
1085 int cpu; 1158 int cpu;
@@ -1093,6 +1166,10 @@ static int __init topology_init(void)
1093 if (rc) 1166 if (rc)
1094 return rc; 1167 return rc;
1095#endif 1168#endif
1169 rc = sysfs_create_file(&cpu_sysdev_class.kset.kobj,
1170 &attr_dispatching.attr);
1171 if (rc)
1172 return rc;
1096 for_each_present_cpu(cpu) { 1173 for_each_present_cpu(cpu) {
1097 rc = smp_add_present_cpu(cpu); 1174 rc = smp_add_present_cpu(cpu);
1098 if (rc) 1175 if (rc)
diff --git a/arch/s390/kernel/sys_s390.c b/arch/s390/kernel/sys_s390.c
index fefee99f28aa..988d0d64c2c8 100644
--- a/arch/s390/kernel/sys_s390.c
+++ b/arch/s390/kernel/sys_s390.c
@@ -29,8 +29,8 @@
29#include <linux/personality.h> 29#include <linux/personality.h>
30#include <linux/unistd.h> 30#include <linux/unistd.h>
31#include <linux/ipc.h> 31#include <linux/ipc.h>
32
33#include <asm/uaccess.h> 32#include <asm/uaccess.h>
33#include "entry.h"
34 34
35/* 35/*
36 * sys_pipe() is the normal C calling standard for creating 36 * sys_pipe() is the normal C calling standard for creating
diff --git a/arch/s390/kernel/time.c b/arch/s390/kernel/time.c
index cb232c155360..7aec676fefd5 100644
--- a/arch/s390/kernel/time.c
+++ b/arch/s390/kernel/time.c
@@ -30,7 +30,7 @@
30#include <linux/timex.h> 30#include <linux/timex.h>
31#include <linux/notifier.h> 31#include <linux/notifier.h>
32#include <linux/clocksource.h> 32#include <linux/clocksource.h>
33 33#include <linux/clockchips.h>
34#include <asm/uaccess.h> 34#include <asm/uaccess.h>
35#include <asm/delay.h> 35#include <asm/delay.h>
36#include <asm/s390_ext.h> 36#include <asm/s390_ext.h>
@@ -39,6 +39,7 @@
39#include <asm/irq_regs.h> 39#include <asm/irq_regs.h>
40#include <asm/timer.h> 40#include <asm/timer.h>
41#include <asm/etr.h> 41#include <asm/etr.h>
42#include <asm/cio.h>
42 43
43/* change this if you have some constant time drift */ 44/* change this if you have some constant time drift */
44#define USECS_PER_JIFFY ((unsigned long) 1000000/HZ) 45#define USECS_PER_JIFFY ((unsigned long) 1000000/HZ)
@@ -57,16 +58,16 @@
57 58
58static ext_int_info_t ext_int_info_cc; 59static ext_int_info_t ext_int_info_cc;
59static ext_int_info_t ext_int_etr_cc; 60static ext_int_info_t ext_int_etr_cc;
60static u64 init_timer_cc;
61static u64 jiffies_timer_cc; 61static u64 jiffies_timer_cc;
62static u64 xtime_cc; 62
63static DEFINE_PER_CPU(struct clock_event_device, comparators);
63 64
64/* 65/*
65 * Scheduler clock - returns current time in nanosec units. 66 * Scheduler clock - returns current time in nanosec units.
66 */ 67 */
67unsigned long long sched_clock(void) 68unsigned long long sched_clock(void)
68{ 69{
69 return ((get_clock() - jiffies_timer_cc) * 125) >> 9; 70 return ((get_clock_xt() - jiffies_timer_cc) * 125) >> 9;
70} 71}
71 72
72/* 73/*
@@ -95,162 +96,40 @@ void tod_to_timeval(__u64 todval, struct timespec *xtime)
95#define s390_do_profile() do { ; } while(0) 96#define s390_do_profile() do { ; } while(0)
96#endif /* CONFIG_PROFILING */ 97#endif /* CONFIG_PROFILING */
97 98
98/* 99void clock_comparator_work(void)
99 * Advance the per cpu tick counter up to the time given with the
100 * "time" argument. The per cpu update consists of accounting
101 * the virtual cpu time, calling update_process_times and calling
102 * the profiling hook. If xtime is before time it is advanced as well.
103 */
104void account_ticks(u64 time)
105{ 100{
106 __u32 ticks; 101 struct clock_event_device *cd;
107 __u64 tmp;
108
109 /* Calculate how many ticks have passed. */
110 if (time < S390_lowcore.jiffy_timer)
111 return;
112 tmp = time - S390_lowcore.jiffy_timer;
113 if (tmp >= 2*CLK_TICKS_PER_JIFFY) { /* more than two ticks ? */
114 ticks = __div(tmp, CLK_TICKS_PER_JIFFY) + 1;
115 S390_lowcore.jiffy_timer +=
116 CLK_TICKS_PER_JIFFY * (__u64) ticks;
117 } else if (tmp >= CLK_TICKS_PER_JIFFY) {
118 ticks = 2;
119 S390_lowcore.jiffy_timer += 2*CLK_TICKS_PER_JIFFY;
120 } else {
121 ticks = 1;
122 S390_lowcore.jiffy_timer += CLK_TICKS_PER_JIFFY;
123 }
124
125#ifdef CONFIG_SMP
126 /*
127 * Do not rely on the boot cpu to do the calls to do_timer.
128 * Spread it over all cpus instead.
129 */
130 write_seqlock(&xtime_lock);
131 if (S390_lowcore.jiffy_timer > xtime_cc) {
132 __u32 xticks;
133 tmp = S390_lowcore.jiffy_timer - xtime_cc;
134 if (tmp >= 2*CLK_TICKS_PER_JIFFY) {
135 xticks = __div(tmp, CLK_TICKS_PER_JIFFY);
136 xtime_cc += (__u64) xticks * CLK_TICKS_PER_JIFFY;
137 } else {
138 xticks = 1;
139 xtime_cc += CLK_TICKS_PER_JIFFY;
140 }
141 do_timer(xticks);
142 }
143 write_sequnlock(&xtime_lock);
144#else
145 do_timer(ticks);
146#endif
147
148 while (ticks--)
149 update_process_times(user_mode(get_irq_regs()));
150 102
103 S390_lowcore.clock_comparator = -1ULL;
104 set_clock_comparator(S390_lowcore.clock_comparator);
105 cd = &__get_cpu_var(comparators);
106 cd->event_handler(cd);
151 s390_do_profile(); 107 s390_do_profile();
152} 108}
153 109
154#ifdef CONFIG_NO_IDLE_HZ
155
156#ifdef CONFIG_NO_IDLE_HZ_INIT
157int sysctl_hz_timer = 0;
158#else
159int sysctl_hz_timer = 1;
160#endif
161
162/*
163 * Stop the HZ tick on the current CPU.
164 * Only cpu_idle may call this function.
165 */
166static void stop_hz_timer(void)
167{
168 unsigned long flags;
169 unsigned long seq, next;
170 __u64 timer, todval;
171 int cpu = smp_processor_id();
172
173 if (sysctl_hz_timer != 0)
174 return;
175
176 cpu_set(cpu, nohz_cpu_mask);
177
178 /*
179 * Leave the clock comparator set up for the next timer
180 * tick if either rcu or a softirq is pending.
181 */
182 if (rcu_needs_cpu(cpu) || local_softirq_pending()) {
183 cpu_clear(cpu, nohz_cpu_mask);
184 return;
185 }
186
187 /*
188 * This cpu is going really idle. Set up the clock comparator
189 * for the next event.
190 */
191 next = next_timer_interrupt();
192 do {
193 seq = read_seqbegin_irqsave(&xtime_lock, flags);
194 timer = ((__u64) next) - ((__u64) jiffies) + jiffies_64;
195 } while (read_seqretry_irqrestore(&xtime_lock, seq, flags));
196 todval = -1ULL;
197 /* Be careful about overflows. */
198 if (timer < (-1ULL / CLK_TICKS_PER_JIFFY)) {
199 timer = jiffies_timer_cc + timer * CLK_TICKS_PER_JIFFY;
200 if (timer >= jiffies_timer_cc)
201 todval = timer;
202 }
203 set_clock_comparator(todval);
204}
205
206/* 110/*
207 * Start the HZ tick on the current CPU. 111 * Fixup the clock comparator.
208 * Only cpu_idle may call this function.
209 */ 112 */
210static void start_hz_timer(void) 113static void fixup_clock_comparator(unsigned long long delta)
211{ 114{
212 if (!cpu_isset(smp_processor_id(), nohz_cpu_mask)) 115 /* If nobody is waiting there's nothing to fix. */
116 if (S390_lowcore.clock_comparator == -1ULL)
213 return; 117 return;
214 account_ticks(get_clock()); 118 S390_lowcore.clock_comparator += delta;
215 set_clock_comparator(S390_lowcore.jiffy_timer + CPU_DEVIATION); 119 set_clock_comparator(S390_lowcore.clock_comparator);
216 cpu_clear(smp_processor_id(), nohz_cpu_mask);
217}
218
219static int nohz_idle_notify(struct notifier_block *self,
220 unsigned long action, void *hcpu)
221{
222 switch (action) {
223 case S390_CPU_IDLE:
224 stop_hz_timer();
225 break;
226 case S390_CPU_NOT_IDLE:
227 start_hz_timer();
228 break;
229 }
230 return NOTIFY_OK;
231} 120}
232 121
233static struct notifier_block nohz_idle_nb = { 122static int s390_next_event(unsigned long delta,
234 .notifier_call = nohz_idle_notify, 123 struct clock_event_device *evt)
235};
236
237static void __init nohz_init(void)
238{ 124{
239 if (register_idle_notifier(&nohz_idle_nb)) 125 S390_lowcore.clock_comparator = get_clock() + delta;
240 panic("Couldn't register idle notifier"); 126 set_clock_comparator(S390_lowcore.clock_comparator);
127 return 0;
241} 128}
242 129
243#endif 130static void s390_set_mode(enum clock_event_mode mode,
244 131 struct clock_event_device *evt)
245/*
246 * Set up per cpu jiffy timer and set the clock comparator.
247 */
248static void setup_jiffy_timer(void)
249{ 132{
250 /* Set up clock comparator to next jiffy. */
251 S390_lowcore.jiffy_timer =
252 jiffies_timer_cc + (jiffies_64 + 1) * CLK_TICKS_PER_JIFFY;
253 set_clock_comparator(S390_lowcore.jiffy_timer + CPU_DEVIATION);
254} 133}
255 134
256/* 135/*
@@ -259,7 +138,26 @@ static void setup_jiffy_timer(void)
259 */ 138 */
260void init_cpu_timer(void) 139void init_cpu_timer(void)
261{ 140{
262 setup_jiffy_timer(); 141 struct clock_event_device *cd;
142 int cpu;
143
144 S390_lowcore.clock_comparator = -1ULL;
145 set_clock_comparator(S390_lowcore.clock_comparator);
146
147 cpu = smp_processor_id();
148 cd = &per_cpu(comparators, cpu);
149 cd->name = "comparator";
150 cd->features = CLOCK_EVT_FEAT_ONESHOT;
151 cd->mult = 16777;
152 cd->shift = 12;
153 cd->min_delta_ns = 1;
154 cd->max_delta_ns = LONG_MAX;
155 cd->rating = 400;
156 cd->cpumask = cpumask_of_cpu(cpu);
157 cd->set_next_event = s390_next_event;
158 cd->set_mode = s390_set_mode;
159
160 clockevents_register_device(cd);
263 161
264 /* Enable clock comparator timer interrupt. */ 162 /* Enable clock comparator timer interrupt. */
265 __ctl_set_bit(0,11); 163 __ctl_set_bit(0,11);
@@ -270,8 +168,6 @@ void init_cpu_timer(void)
270 168
271static void clock_comparator_interrupt(__u16 code) 169static void clock_comparator_interrupt(__u16 code)
272{ 170{
273 /* set clock comparator for next tick */
274 set_clock_comparator(S390_lowcore.jiffy_timer + CPU_DEVIATION);
275} 171}
276 172
277static void etr_reset(void); 173static void etr_reset(void);
@@ -316,8 +212,9 @@ static struct clocksource clocksource_tod = {
316 */ 212 */
317void __init time_init(void) 213void __init time_init(void)
318{ 214{
215 u64 init_timer_cc;
216
319 init_timer_cc = reset_tod_clock(); 217 init_timer_cc = reset_tod_clock();
320 xtime_cc = init_timer_cc + CLK_TICKS_PER_JIFFY;
321 jiffies_timer_cc = init_timer_cc - jiffies_64 * CLK_TICKS_PER_JIFFY; 218 jiffies_timer_cc = init_timer_cc - jiffies_64 * CLK_TICKS_PER_JIFFY;
322 219
323 /* set xtime */ 220 /* set xtime */
@@ -342,10 +239,6 @@ void __init time_init(void)
342 /* Enable TOD clock interrupts on the boot cpu. */ 239 /* Enable TOD clock interrupts on the boot cpu. */
343 init_cpu_timer(); 240 init_cpu_timer();
344 241
345#ifdef CONFIG_NO_IDLE_HZ
346 nohz_init();
347#endif
348
349#ifdef CONFIG_VIRT_TIMER 242#ifdef CONFIG_VIRT_TIMER
350 vtime_init(); 243 vtime_init();
351#endif 244#endif
@@ -699,53 +592,49 @@ static int etr_aib_follows(struct etr_aib *a1, struct etr_aib *a2, int p)
699} 592}
700 593
701/* 594/*
702 * The time is "clock". xtime is what we think the time is. 595 * The time is "clock". old is what we think the time is.
703 * Adjust the value by a multiple of jiffies and add the delta to ntp. 596 * Adjust the value by a multiple of jiffies and add the delta to ntp.
704 * "delay" is an approximation how long the synchronization took. If 597 * "delay" is an approximation how long the synchronization took. If
705 * the time correction is positive, then "delay" is subtracted from 598 * the time correction is positive, then "delay" is subtracted from
706 * the time difference and only the remaining part is passed to ntp. 599 * the time difference and only the remaining part is passed to ntp.
707 */ 600 */
708static void etr_adjust_time(unsigned long long clock, unsigned long long delay) 601static unsigned long long etr_adjust_time(unsigned long long old,
602 unsigned long long clock,
603 unsigned long long delay)
709{ 604{
710 unsigned long long delta, ticks; 605 unsigned long long delta, ticks;
711 struct timex adjust; 606 struct timex adjust;
712 607
713 /* 608 if (clock > old) {
714 * We don't have to take the xtime lock because the cpu
715 * executing etr_adjust_time is running disabled in
716 * tasklet context and all other cpus are looping in
717 * etr_sync_cpu_start.
718 */
719 if (clock > xtime_cc) {
720 /* It is later than we thought. */ 609 /* It is later than we thought. */
721 delta = ticks = clock - xtime_cc; 610 delta = ticks = clock - old;
722 delta = ticks = (delta < delay) ? 0 : delta - delay; 611 delta = ticks = (delta < delay) ? 0 : delta - delay;
723 delta -= do_div(ticks, CLK_TICKS_PER_JIFFY); 612 delta -= do_div(ticks, CLK_TICKS_PER_JIFFY);
724 init_timer_cc = init_timer_cc + delta;
725 jiffies_timer_cc = jiffies_timer_cc + delta;
726 xtime_cc = xtime_cc + delta;
727 adjust.offset = ticks * (1000000 / HZ); 613 adjust.offset = ticks * (1000000 / HZ);
728 } else { 614 } else {
729 /* It is earlier than we thought. */ 615 /* It is earlier than we thought. */
730 delta = ticks = xtime_cc - clock; 616 delta = ticks = old - clock;
731 delta -= do_div(ticks, CLK_TICKS_PER_JIFFY); 617 delta -= do_div(ticks, CLK_TICKS_PER_JIFFY);
732 init_timer_cc = init_timer_cc - delta; 618 delta = -delta;
733 jiffies_timer_cc = jiffies_timer_cc - delta;
734 xtime_cc = xtime_cc - delta;
735 adjust.offset = -ticks * (1000000 / HZ); 619 adjust.offset = -ticks * (1000000 / HZ);
736 } 620 }
621 jiffies_timer_cc += delta;
737 if (adjust.offset != 0) { 622 if (adjust.offset != 0) {
738 printk(KERN_NOTICE "etr: time adjusted by %li micro-seconds\n", 623 printk(KERN_NOTICE "etr: time adjusted by %li micro-seconds\n",
739 adjust.offset); 624 adjust.offset);
740 adjust.modes = ADJ_OFFSET_SINGLESHOT; 625 adjust.modes = ADJ_OFFSET_SINGLESHOT;
741 do_adjtimex(&adjust); 626 do_adjtimex(&adjust);
742 } 627 }
628 return delta;
743} 629}
744 630
631static struct {
632 int in_sync;
633 unsigned long long fixup_cc;
634} etr_sync;
635
745static void etr_sync_cpu_start(void *dummy) 636static void etr_sync_cpu_start(void *dummy)
746{ 637{
747 int *in_sync = dummy;
748
749 etr_enable_sync_clock(); 638 etr_enable_sync_clock();
750 /* 639 /*
751 * This looks like a busy wait loop but it isn't. etr_sync_cpus 640 * This looks like a busy wait loop but it isn't. etr_sync_cpus
@@ -753,7 +642,7 @@ static void etr_sync_cpu_start(void *dummy)
753 * __udelay will stop the cpu on an enabled wait psw until the 642 * __udelay will stop the cpu on an enabled wait psw until the
754 * TOD is running again. 643 * TOD is running again.
755 */ 644 */
756 while (*in_sync == 0) { 645 while (etr_sync.in_sync == 0) {
757 __udelay(1); 646 __udelay(1);
758 /* 647 /*
759 * A different cpu changes *in_sync. Therefore use 648 * A different cpu changes *in_sync. Therefore use
@@ -761,14 +650,14 @@ static void etr_sync_cpu_start(void *dummy)
761 */ 650 */
762 barrier(); 651 barrier();
763 } 652 }
764 if (*in_sync != 1) 653 if (etr_sync.in_sync != 1)
765 /* Didn't work. Clear per-cpu in sync bit again. */ 654 /* Didn't work. Clear per-cpu in sync bit again. */
766 etr_disable_sync_clock(NULL); 655 etr_disable_sync_clock(NULL);
767 /* 656 /*
768 * This round of TOD syncing is done. Set the clock comparator 657 * This round of TOD syncing is done. Set the clock comparator
769 * to the next tick and let the processor continue. 658 * to the next tick and let the processor continue.
770 */ 659 */
771 setup_jiffy_timer(); 660 fixup_clock_comparator(etr_sync.fixup_cc);
772} 661}
773 662
774static void etr_sync_cpu_end(void *dummy) 663static void etr_sync_cpu_end(void *dummy)
@@ -783,8 +672,8 @@ static void etr_sync_cpu_end(void *dummy)
783static int etr_sync_clock(struct etr_aib *aib, int port) 672static int etr_sync_clock(struct etr_aib *aib, int port)
784{ 673{
785 struct etr_aib *sync_port; 674 struct etr_aib *sync_port;
786 unsigned long long clock, delay; 675 unsigned long long clock, old_clock, delay, delta;
787 int in_sync, follows; 676 int follows;
788 int rc; 677 int rc;
789 678
790 /* Check if the current aib is adjacent to the sync port aib. */ 679 /* Check if the current aib is adjacent to the sync port aib. */
@@ -799,9 +688,9 @@ static int etr_sync_clock(struct etr_aib *aib, int port)
799 * successfully synced the clock. smp_call_function will 688 * successfully synced the clock. smp_call_function will
800 * return after all other cpus are in etr_sync_cpu_start. 689 * return after all other cpus are in etr_sync_cpu_start.
801 */ 690 */
802 in_sync = 0; 691 memset(&etr_sync, 0, sizeof(etr_sync));
803 preempt_disable(); 692 preempt_disable();
804 smp_call_function(etr_sync_cpu_start,&in_sync,0,0); 693 smp_call_function(etr_sync_cpu_start, NULL, 0, 0);
805 local_irq_disable(); 694 local_irq_disable();
806 etr_enable_sync_clock(); 695 etr_enable_sync_clock();
807 696
@@ -809,6 +698,7 @@ static int etr_sync_clock(struct etr_aib *aib, int port)
809 __ctl_set_bit(14, 21); 698 __ctl_set_bit(14, 21);
810 __ctl_set_bit(0, 29); 699 __ctl_set_bit(0, 29);
811 clock = ((unsigned long long) (aib->edf2.etv + 1)) << 32; 700 clock = ((unsigned long long) (aib->edf2.etv + 1)) << 32;
701 old_clock = get_clock();
812 if (set_clock(clock) == 0) { 702 if (set_clock(clock) == 0) {
813 __udelay(1); /* Wait for the clock to start. */ 703 __udelay(1); /* Wait for the clock to start. */
814 __ctl_clear_bit(0, 29); 704 __ctl_clear_bit(0, 29);
@@ -817,16 +707,17 @@ static int etr_sync_clock(struct etr_aib *aib, int port)
817 /* Adjust Linux timing variables. */ 707 /* Adjust Linux timing variables. */
818 delay = (unsigned long long) 708 delay = (unsigned long long)
819 (aib->edf2.etv - sync_port->edf2.etv) << 32; 709 (aib->edf2.etv - sync_port->edf2.etv) << 32;
820 etr_adjust_time(clock, delay); 710 delta = etr_adjust_time(old_clock, clock, delay);
821 setup_jiffy_timer(); 711 etr_sync.fixup_cc = delta;
712 fixup_clock_comparator(delta);
822 /* Verify that the clock is properly set. */ 713 /* Verify that the clock is properly set. */
823 if (!etr_aib_follows(sync_port, aib, port)) { 714 if (!etr_aib_follows(sync_port, aib, port)) {
824 /* Didn't work. */ 715 /* Didn't work. */
825 etr_disable_sync_clock(NULL); 716 etr_disable_sync_clock(NULL);
826 in_sync = -EAGAIN; 717 etr_sync.in_sync = -EAGAIN;
827 rc = -EAGAIN; 718 rc = -EAGAIN;
828 } else { 719 } else {
829 in_sync = 1; 720 etr_sync.in_sync = 1;
830 rc = 0; 721 rc = 0;
831 } 722 }
832 } else { 723 } else {
@@ -834,7 +725,7 @@ static int etr_sync_clock(struct etr_aib *aib, int port)
834 __ctl_clear_bit(0, 29); 725 __ctl_clear_bit(0, 29);
835 __ctl_clear_bit(14, 21); 726 __ctl_clear_bit(14, 21);
836 etr_disable_sync_clock(NULL); 727 etr_disable_sync_clock(NULL);
837 in_sync = -EAGAIN; 728 etr_sync.in_sync = -EAGAIN;
838 rc = -EAGAIN; 729 rc = -EAGAIN;
839 } 730 }
840 local_irq_enable(); 731 local_irq_enable();
diff --git a/arch/s390/kernel/topology.c b/arch/s390/kernel/topology.c
new file mode 100644
index 000000000000..12b39b3d9c38
--- /dev/null
+++ b/arch/s390/kernel/topology.c
@@ -0,0 +1,314 @@
1/*
2 * Copyright IBM Corp. 2007
3 * Author(s): Heiko Carstens <heiko.carstens@de.ibm.com>
4 */
5
6#include <linux/kernel.h>
7#include <linux/mm.h>
8#include <linux/init.h>
9#include <linux/device.h>
10#include <linux/bootmem.h>
11#include <linux/sched.h>
12#include <linux/workqueue.h>
13#include <linux/cpu.h>
14#include <linux/smp.h>
15#include <asm/delay.h>
16#include <asm/s390_ext.h>
17#include <asm/sysinfo.h>
18
19#define CPU_BITS 64
20#define NR_MAG 6
21
22#define PTF_HORIZONTAL (0UL)
23#define PTF_VERTICAL (1UL)
24#define PTF_CHECK (2UL)
25
26struct tl_cpu {
27 unsigned char reserved0[4];
28 unsigned char :6;
29 unsigned char pp:2;
30 unsigned char reserved1;
31 unsigned short origin;
32 unsigned long mask[CPU_BITS / BITS_PER_LONG];
33};
34
35struct tl_container {
36 unsigned char reserved[8];
37};
38
39union tl_entry {
40 unsigned char nl;
41 struct tl_cpu cpu;
42 struct tl_container container;
43};
44
45struct tl_info {
46 unsigned char reserved0[2];
47 unsigned short length;
48 unsigned char mag[NR_MAG];
49 unsigned char reserved1;
50 unsigned char mnest;
51 unsigned char reserved2[4];
52 union tl_entry tle[0];
53};
54
55struct core_info {
56 struct core_info *next;
57 cpumask_t mask;
58};
59
60static void topology_work_fn(struct work_struct *work);
61static struct tl_info *tl_info;
62static struct core_info core_info;
63static int machine_has_topology;
64static int machine_has_topology_irq;
65static struct timer_list topology_timer;
66static void set_topology_timer(void);
67static DECLARE_WORK(topology_work, topology_work_fn);
68
69cpumask_t cpu_coregroup_map(unsigned int cpu)
70{
71 struct core_info *core = &core_info;
72 cpumask_t mask;
73
74 cpus_clear(mask);
75 if (!machine_has_topology)
76 return cpu_present_map;
77 mutex_lock(&smp_cpu_state_mutex);
78 while (core) {
79 if (cpu_isset(cpu, core->mask)) {
80 mask = core->mask;
81 break;
82 }
83 core = core->next;
84 }
85 mutex_unlock(&smp_cpu_state_mutex);
86 if (cpus_empty(mask))
87 mask = cpumask_of_cpu(cpu);
88 return mask;
89}
90
91static void add_cpus_to_core(struct tl_cpu *tl_cpu, struct core_info *core)
92{
93 unsigned int cpu;
94
95 for (cpu = find_first_bit(&tl_cpu->mask[0], CPU_BITS);
96 cpu < CPU_BITS;
97 cpu = find_next_bit(&tl_cpu->mask[0], CPU_BITS, cpu + 1))
98 {
99 unsigned int rcpu, lcpu;
100
101 rcpu = CPU_BITS - 1 - cpu + tl_cpu->origin;
102 for_each_present_cpu(lcpu) {
103 if (__cpu_logical_map[lcpu] == rcpu) {
104 cpu_set(lcpu, core->mask);
105 smp_cpu_polarization[lcpu] = tl_cpu->pp;
106 }
107 }
108 }
109}
110
111static void clear_cores(void)
112{
113 struct core_info *core = &core_info;
114
115 while (core) {
116 cpus_clear(core->mask);
117 core = core->next;
118 }
119}
120
121static union tl_entry *next_tle(union tl_entry *tle)
122{
123 if (tle->nl)
124 return (union tl_entry *)((struct tl_container *)tle + 1);
125 else
126 return (union tl_entry *)((struct tl_cpu *)tle + 1);
127}
128
129static void tl_to_cores(struct tl_info *info)
130{
131 union tl_entry *tle, *end;
132 struct core_info *core = &core_info;
133
134 mutex_lock(&smp_cpu_state_mutex);
135 clear_cores();
136 tle = info->tle;
137 end = (union tl_entry *)((unsigned long)info + info->length);
138 while (tle < end) {
139 switch (tle->nl) {
140 case 5:
141 case 4:
142 case 3:
143 case 2:
144 break;
145 case 1:
146 core = core->next;
147 break;
148 case 0:
149 add_cpus_to_core(&tle->cpu, core);
150 break;
151 default:
152 clear_cores();
153 machine_has_topology = 0;
154 return;
155 }
156 tle = next_tle(tle);
157 }
158 mutex_unlock(&smp_cpu_state_mutex);
159}
160
161static void topology_update_polarization_simple(void)
162{
163 int cpu;
164
165 mutex_lock(&smp_cpu_state_mutex);
166 for_each_present_cpu(cpu)
167 smp_cpu_polarization[cpu] = POLARIZATION_HRZ;
168 mutex_unlock(&smp_cpu_state_mutex);
169}
170
171static int ptf(unsigned long fc)
172{
173 int rc;
174
175 asm volatile(
176 " .insn rre,0xb9a20000,%1,%1\n"
177 " ipm %0\n"
178 " srl %0,28\n"
179 : "=d" (rc)
180 : "d" (fc) : "cc");
181 return rc;
182}
183
184int topology_set_cpu_management(int fc)
185{
186 int cpu;
187 int rc;
188
189 if (!machine_has_topology)
190 return -EOPNOTSUPP;
191 if (fc)
192 rc = ptf(PTF_VERTICAL);
193 else
194 rc = ptf(PTF_HORIZONTAL);
195 if (rc)
196 return -EBUSY;
197 for_each_present_cpu(cpu)
198 smp_cpu_polarization[cpu] = POLARIZATION_UNKNWN;
199 return rc;
200}
201
202void arch_update_cpu_topology(void)
203{
204 struct tl_info *info = tl_info;
205 struct sys_device *sysdev;
206 int cpu;
207
208 if (!machine_has_topology) {
209 topology_update_polarization_simple();
210 return;
211 }
212 stsi(info, 15, 1, 2);
213 tl_to_cores(info);
214 for_each_online_cpu(cpu) {
215 sysdev = get_cpu_sysdev(cpu);
216 kobject_uevent(&sysdev->kobj, KOBJ_CHANGE);
217 }
218}
219
220static void topology_work_fn(struct work_struct *work)
221{
222 arch_reinit_sched_domains();
223}
224
225void topology_schedule_update(void)
226{
227 schedule_work(&topology_work);
228}
229
230static void topology_timer_fn(unsigned long ignored)
231{
232 if (ptf(PTF_CHECK))
233 topology_schedule_update();
234 set_topology_timer();
235}
236
237static void set_topology_timer(void)
238{
239 topology_timer.function = topology_timer_fn;
240 topology_timer.data = 0;
241 topology_timer.expires = jiffies + 60 * HZ;
242 add_timer(&topology_timer);
243}
244
245static void topology_interrupt(__u16 code)
246{
247 schedule_work(&topology_work);
248}
249
250static int __init init_topology_update(void)
251{
252 int rc;
253
254 if (!machine_has_topology) {
255 topology_update_polarization_simple();
256 return 0;
257 }
258 init_timer_deferrable(&topology_timer);
259 if (machine_has_topology_irq) {
260 rc = register_external_interrupt(0x2005, topology_interrupt);
261 if (rc)
262 return rc;
263 ctl_set_bit(0, 8);
264 }
265 else
266 set_topology_timer();
267 return 0;
268}
269__initcall(init_topology_update);
270
271void __init s390_init_cpu_topology(void)
272{
273 unsigned long long facility_bits;
274 struct tl_info *info;
275 struct core_info *core;
276 int nr_cores;
277 int i;
278
279 if (stfle(&facility_bits, 1) <= 0)
280 return;
281 if (!(facility_bits & (1ULL << 52)) || !(facility_bits & (1ULL << 61)))
282 return;
283 machine_has_topology = 1;
284
285 if (facility_bits & (1ULL << 51))
286 machine_has_topology_irq = 1;
287
288 tl_info = alloc_bootmem_pages(PAGE_SIZE);
289 if (!tl_info)
290 goto error;
291 info = tl_info;
292 stsi(info, 15, 1, 2);
293
294 nr_cores = info->mag[NR_MAG - 2];
295 for (i = 0; i < info->mnest - 2; i++)
296 nr_cores *= info->mag[NR_MAG - 3 - i];
297
298 printk(KERN_INFO "CPU topology:");
299 for (i = 0; i < NR_MAG; i++)
300 printk(" %d", info->mag[i]);
301 printk(" / %d\n", info->mnest);
302
303 core = &core_info;
304 for (i = 0; i < nr_cores; i++) {
305 core->next = alloc_bootmem(sizeof(struct core_info));
306 core = core->next;
307 if (!core)
308 goto error;
309 }
310 return;
311error:
312 machine_has_topology = 0;
313 machine_has_topology_irq = 0;
314}
diff --git a/arch/s390/kernel/traps.c b/arch/s390/kernel/traps.c
index 60f728aeaf12..57b607b61100 100644
--- a/arch/s390/kernel/traps.c
+++ b/arch/s390/kernel/traps.c
@@ -42,11 +42,8 @@
42#include <asm/s390_ext.h> 42#include <asm/s390_ext.h>
43#include <asm/lowcore.h> 43#include <asm/lowcore.h>
44#include <asm/debug.h> 44#include <asm/debug.h>
45#include "entry.h"
45 46
46/* Called from entry.S only */
47extern void handle_per_exception(struct pt_regs *regs);
48
49typedef void pgm_check_handler_t(struct pt_regs *, long);
50pgm_check_handler_t *pgm_check_table[128]; 47pgm_check_handler_t *pgm_check_table[128];
51 48
52#ifdef CONFIG_SYSCTL 49#ifdef CONFIG_SYSCTL
@@ -59,7 +56,6 @@ int sysctl_userprocess_debug = 0;
59 56
60extern pgm_check_handler_t do_protection_exception; 57extern pgm_check_handler_t do_protection_exception;
61extern pgm_check_handler_t do_dat_exception; 58extern pgm_check_handler_t do_dat_exception;
62extern pgm_check_handler_t do_monitor_call;
63extern pgm_check_handler_t do_asce_exception; 59extern pgm_check_handler_t do_asce_exception;
64 60
65#define stack_pointer ({ void **sp; asm("la %0,0(15)" : "=&d" (sp)); sp; }) 61#define stack_pointer ({ void **sp; asm("la %0,0(15)" : "=&d" (sp)); sp; })
@@ -138,7 +134,6 @@ void show_trace(struct task_struct *task, unsigned long *stack)
138 else 134 else
139 __show_trace(sp, S390_lowcore.thread_info, 135 __show_trace(sp, S390_lowcore.thread_info,
140 S390_lowcore.thread_info + THREAD_SIZE); 136 S390_lowcore.thread_info + THREAD_SIZE);
141 printk("\n");
142 if (!task) 137 if (!task)
143 task = current; 138 task = current;
144 debug_show_held_locks(task); 139 debug_show_held_locks(task);
@@ -166,6 +161,15 @@ void show_stack(struct task_struct *task, unsigned long *sp)
166 show_trace(task, sp); 161 show_trace(task, sp);
167} 162}
168 163
164#ifdef CONFIG_64BIT
165void show_last_breaking_event(struct pt_regs *regs)
166{
167 printk("Last Breaking-Event-Address:\n");
168 printk(" [<%016lx>] ", regs->args[0] & PSW_ADDR_INSN);
169 print_symbol("%s\n", regs->args[0] & PSW_ADDR_INSN);
170}
171#endif
172
169/* 173/*
170 * The architecture-independent dump_stack generator 174 * The architecture-independent dump_stack generator
171 */ 175 */
@@ -739,6 +743,5 @@ void __init trap_init(void)
739 pgm_check_table[0x15] = &operand_exception; 743 pgm_check_table[0x15] = &operand_exception;
740 pgm_check_table[0x1C] = &space_switch_exception; 744 pgm_check_table[0x1C] = &space_switch_exception;
741 pgm_check_table[0x1D] = &hfp_sqrt_exception; 745 pgm_check_table[0x1D] = &hfp_sqrt_exception;
742 pgm_check_table[0x40] = &do_monitor_call;
743 pfault_irq_init(); 746 pfault_irq_init();
744} 747}
diff --git a/arch/s390/lib/delay.c b/arch/s390/lib/delay.c
index 70f2a862b670..eae21a8ac72d 100644
--- a/arch/s390/lib/delay.c
+++ b/arch/s390/lib/delay.c
@@ -34,7 +34,7 @@ void __delay(unsigned long loops)
34 */ 34 */
35void __udelay(unsigned long usecs) 35void __udelay(unsigned long usecs)
36{ 36{
37 u64 end, time, jiffy_timer = 0; 37 u64 end, time, old_cc = 0;
38 unsigned long flags, cr0, mask, dummy; 38 unsigned long flags, cr0, mask, dummy;
39 int irq_context; 39 int irq_context;
40 40
@@ -43,8 +43,8 @@ void __udelay(unsigned long usecs)
43 local_bh_disable(); 43 local_bh_disable();
44 local_irq_save(flags); 44 local_irq_save(flags);
45 if (raw_irqs_disabled_flags(flags)) { 45 if (raw_irqs_disabled_flags(flags)) {
46 jiffy_timer = S390_lowcore.jiffy_timer; 46 old_cc = S390_lowcore.clock_comparator;
47 S390_lowcore.jiffy_timer = -1ULL - (4096 << 12); 47 S390_lowcore.clock_comparator = -1ULL;
48 __ctl_store(cr0, 0, 0); 48 __ctl_store(cr0, 0, 0);
49 dummy = (cr0 & 0xffff00e0) | 0x00000800; 49 dummy = (cr0 & 0xffff00e0) | 0x00000800;
50 __ctl_load(dummy , 0, 0); 50 __ctl_load(dummy , 0, 0);
@@ -55,8 +55,8 @@ void __udelay(unsigned long usecs)
55 55
56 end = get_clock() + ((u64) usecs << 12); 56 end = get_clock() + ((u64) usecs << 12);
57 do { 57 do {
58 time = end < S390_lowcore.jiffy_timer ? 58 time = end < S390_lowcore.clock_comparator ?
59 end : S390_lowcore.jiffy_timer; 59 end : S390_lowcore.clock_comparator;
60 set_clock_comparator(time); 60 set_clock_comparator(time);
61 trace_hardirqs_on(); 61 trace_hardirqs_on();
62 __load_psw_mask(mask); 62 __load_psw_mask(mask);
@@ -65,10 +65,10 @@ void __udelay(unsigned long usecs)
65 65
66 if (raw_irqs_disabled_flags(flags)) { 66 if (raw_irqs_disabled_flags(flags)) {
67 __ctl_load(cr0, 0, 0); 67 __ctl_load(cr0, 0, 0);
68 S390_lowcore.jiffy_timer = jiffy_timer; 68 S390_lowcore.clock_comparator = old_cc;
69 } 69 }
70 if (!irq_context) 70 if (!irq_context)
71 _local_bh_enable(); 71 _local_bh_enable();
72 set_clock_comparator(S390_lowcore.jiffy_timer); 72 set_clock_comparator(S390_lowcore.clock_comparator);
73 local_irq_restore(flags); 73 local_irq_restore(flags);
74} 74}
diff --git a/arch/s390/lib/uaccess_pt.c b/arch/s390/lib/uaccess_pt.c
index 5efdfe9f5e76..d66215b0fde9 100644
--- a/arch/s390/lib/uaccess_pt.c
+++ b/arch/s390/lib/uaccess_pt.c
@@ -302,6 +302,10 @@ static size_t copy_in_user_pt(size_t n, void __user *to,
302 pte_t *pte_from, *pte_to; 302 pte_t *pte_from, *pte_to;
303 int write_user; 303 int write_user;
304 304
305 if (segment_eq(get_fs(), KERNEL_DS)) {
306 memcpy((void __force *) to, (void __force *) from, n);
307 return 0;
308 }
305 done = 0; 309 done = 0;
306retry: 310retry:
307 spin_lock(&mm->page_table_lock); 311 spin_lock(&mm->page_table_lock);
@@ -361,18 +365,10 @@ fault:
361 : "0" (-EFAULT), "d" (oparg), "a" (uaddr), \ 365 : "0" (-EFAULT), "d" (oparg), "a" (uaddr), \
362 "m" (*uaddr) : "cc" ); 366 "m" (*uaddr) : "cc" );
363 367
364int futex_atomic_op_pt(int op, int __user *uaddr, int oparg, int *old) 368static int __futex_atomic_op_pt(int op, int __user *uaddr, int oparg, int *old)
365{ 369{
366 int oldval = 0, newval, ret; 370 int oldval = 0, newval, ret;
367 371
368 spin_lock(&current->mm->page_table_lock);
369 uaddr = (int __user *) __dat_user_addr((unsigned long) uaddr);
370 if (!uaddr) {
371 spin_unlock(&current->mm->page_table_lock);
372 return -EFAULT;
373 }
374 get_page(virt_to_page(uaddr));
375 spin_unlock(&current->mm->page_table_lock);
376 switch (op) { 372 switch (op) {
377 case FUTEX_OP_SET: 373 case FUTEX_OP_SET:
378 __futex_atomic_op("lr %2,%5\n", 374 __futex_atomic_op("lr %2,%5\n",
@@ -397,17 +393,17 @@ int futex_atomic_op_pt(int op, int __user *uaddr, int oparg, int *old)
397 default: 393 default:
398 ret = -ENOSYS; 394 ret = -ENOSYS;
399 } 395 }
400 put_page(virt_to_page(uaddr)); 396 if (ret == 0)
401 *old = oldval; 397 *old = oldval;
402 return ret; 398 return ret;
403} 399}
404 400
405int futex_atomic_cmpxchg_pt(int __user *uaddr, int oldval, int newval) 401int futex_atomic_op_pt(int op, int __user *uaddr, int oparg, int *old)
406{ 402{
407 int ret; 403 int ret;
408 404
409 if (!current->mm) 405 if (segment_eq(get_fs(), KERNEL_DS))
410 return -EFAULT; 406 return __futex_atomic_op_pt(op, uaddr, oparg, old);
411 spin_lock(&current->mm->page_table_lock); 407 spin_lock(&current->mm->page_table_lock);
412 uaddr = (int __user *) __dat_user_addr((unsigned long) uaddr); 408 uaddr = (int __user *) __dat_user_addr((unsigned long) uaddr);
413 if (!uaddr) { 409 if (!uaddr) {
@@ -416,13 +412,40 @@ int futex_atomic_cmpxchg_pt(int __user *uaddr, int oldval, int newval)
416 } 412 }
417 get_page(virt_to_page(uaddr)); 413 get_page(virt_to_page(uaddr));
418 spin_unlock(&current->mm->page_table_lock); 414 spin_unlock(&current->mm->page_table_lock);
419 asm volatile(" cs %1,%4,0(%5)\n" 415 ret = __futex_atomic_op_pt(op, uaddr, oparg, old);
420 "0: lr %0,%1\n" 416 put_page(virt_to_page(uaddr));
421 "1:\n" 417 return ret;
422 EX_TABLE(0b,1b) 418}
419
420static int __futex_atomic_cmpxchg_pt(int __user *uaddr, int oldval, int newval)
421{
422 int ret;
423
424 asm volatile("0: cs %1,%4,0(%5)\n"
425 "1: lr %0,%1\n"
426 "2:\n"
427 EX_TABLE(0b,2b) EX_TABLE(1b,2b)
423 : "=d" (ret), "+d" (oldval), "=m" (*uaddr) 428 : "=d" (ret), "+d" (oldval), "=m" (*uaddr)
424 : "0" (-EFAULT), "d" (newval), "a" (uaddr), "m" (*uaddr) 429 : "0" (-EFAULT), "d" (newval), "a" (uaddr), "m" (*uaddr)
425 : "cc", "memory" ); 430 : "cc", "memory" );
431 return ret;
432}
433
434int futex_atomic_cmpxchg_pt(int __user *uaddr, int oldval, int newval)
435{
436 int ret;
437
438 if (segment_eq(get_fs(), KERNEL_DS))
439 return __futex_atomic_cmpxchg_pt(uaddr, oldval, newval);
440 spin_lock(&current->mm->page_table_lock);
441 uaddr = (int __user *) __dat_user_addr((unsigned long) uaddr);
442 if (!uaddr) {
443 spin_unlock(&current->mm->page_table_lock);
444 return -EFAULT;
445 }
446 get_page(virt_to_page(uaddr));
447 spin_unlock(&current->mm->page_table_lock);
448 ret = __futex_atomic_cmpxchg_pt(uaddr, oldval, newval);
426 put_page(virt_to_page(uaddr)); 449 put_page(virt_to_page(uaddr));
427 return ret; 450 return ret;
428} 451}
diff --git a/arch/s390/mm/extmem.c b/arch/s390/mm/extmem.c
index 880b0ebf894b..ed2af0a3303b 100644
--- a/arch/s390/mm/extmem.c
+++ b/arch/s390/mm/extmem.c
@@ -289,22 +289,8 @@ __segment_load (char *name, int do_nonshared, unsigned long *addr, unsigned long
289 289
290 rc = add_shared_memory(seg->start_addr, seg->end - seg->start_addr + 1); 290 rc = add_shared_memory(seg->start_addr, seg->end - seg->start_addr + 1);
291 291
292 switch (rc) { 292 if (rc)
293 case 0:
294 break;
295 case -ENOSPC:
296 PRINT_WARN("segment_load: not loading segment %s - overlaps "
297 "storage/segment\n", name);
298 goto out_free;
299 case -ERANGE:
300 PRINT_WARN("segment_load: not loading segment %s - exceeds "
301 "kernel mapping range\n", name);
302 goto out_free;
303 default:
304 PRINT_WARN("segment_load: not loading segment %s (rc: %d)\n",
305 name, rc);
306 goto out_free; 293 goto out_free;
307 }
308 294
309 seg->res = kzalloc(sizeof(struct resource), GFP_KERNEL); 295 seg->res = kzalloc(sizeof(struct resource), GFP_KERNEL);
310 if (seg->res == NULL) { 296 if (seg->res == NULL) {
@@ -582,8 +568,59 @@ out:
582 mutex_unlock(&dcss_lock); 568 mutex_unlock(&dcss_lock);
583} 569}
584 570
571/*
572 * print appropriate error message for segment_load()/segment_type()
573 * return code
574 */
575void segment_warning(int rc, char *seg_name)
576{
577 switch (rc) {
578 case -ENOENT:
579 PRINT_WARN("cannot load/query segment %s, "
580 "does not exist\n", seg_name);
581 break;
582 case -ENOSYS:
583 PRINT_WARN("cannot load/query segment %s, "
584 "not running on VM\n", seg_name);
585 break;
586 case -EIO:
587 PRINT_WARN("cannot load/query segment %s, "
588 "hardware error\n", seg_name);
589 break;
590 case -ENOTSUPP:
591 PRINT_WARN("cannot load/query segment %s, "
592 "is a multi-part segment\n", seg_name);
593 break;
594 case -ENOSPC:
595 PRINT_WARN("cannot load/query segment %s, "
596 "overlaps with storage\n", seg_name);
597 break;
598 case -EBUSY:
599 PRINT_WARN("cannot load/query segment %s, "
600 "overlaps with already loaded dcss\n", seg_name);
601 break;
602 case -EPERM:
603 PRINT_WARN("cannot load/query segment %s, "
604 "already loaded in incompatible mode\n", seg_name);
605 break;
606 case -ENOMEM:
607 PRINT_WARN("cannot load/query segment %s, "
608 "out of memory\n", seg_name);
609 break;
610 case -ERANGE:
611 PRINT_WARN("cannot load/query segment %s, "
612 "exceeds kernel mapping range\n", seg_name);
613 break;
614 default:
615 PRINT_WARN("cannot load/query segment %s, "
616 "return value %i\n", seg_name, rc);
617 break;
618 }
619}
620
585EXPORT_SYMBOL(segment_load); 621EXPORT_SYMBOL(segment_load);
586EXPORT_SYMBOL(segment_unload); 622EXPORT_SYMBOL(segment_unload);
587EXPORT_SYMBOL(segment_save); 623EXPORT_SYMBOL(segment_save);
588EXPORT_SYMBOL(segment_type); 624EXPORT_SYMBOL(segment_type);
589EXPORT_SYMBOL(segment_modify_shared); 625EXPORT_SYMBOL(segment_modify_shared);
626EXPORT_SYMBOL(segment_warning);
diff --git a/arch/s390/mm/fault.c b/arch/s390/mm/fault.c
index ed13d429a487..2650f46001d0 100644
--- a/arch/s390/mm/fault.c
+++ b/arch/s390/mm/fault.c
@@ -28,11 +28,11 @@
28#include <linux/hardirq.h> 28#include <linux/hardirq.h>
29#include <linux/kprobes.h> 29#include <linux/kprobes.h>
30#include <linux/uaccess.h> 30#include <linux/uaccess.h>
31
32#include <asm/system.h> 31#include <asm/system.h>
33#include <asm/pgtable.h> 32#include <asm/pgtable.h>
34#include <asm/s390_ext.h> 33#include <asm/s390_ext.h>
35#include <asm/mmu_context.h> 34#include <asm/mmu_context.h>
35#include "../kernel/entry.h"
36 36
37#ifndef CONFIG_64BIT 37#ifndef CONFIG_64BIT
38#define __FAIL_ADDR_MASK 0x7ffff000 38#define __FAIL_ADDR_MASK 0x7ffff000
@@ -50,8 +50,6 @@
50extern int sysctl_userprocess_debug; 50extern int sysctl_userprocess_debug;
51#endif 51#endif
52 52
53extern void die(const char *,struct pt_regs *,long);
54
55#ifdef CONFIG_KPROBES 53#ifdef CONFIG_KPROBES
56static inline int notify_page_fault(struct pt_regs *regs, long err) 54static inline int notify_page_fault(struct pt_regs *regs, long err)
57{ 55{
@@ -245,11 +243,6 @@ static void do_sigbus(struct pt_regs *regs, unsigned long error_code,
245} 243}
246 244
247#ifdef CONFIG_S390_EXEC_PROTECT 245#ifdef CONFIG_S390_EXEC_PROTECT
248extern long sys_sigreturn(struct pt_regs *regs);
249extern long sys_rt_sigreturn(struct pt_regs *regs);
250extern long sys32_sigreturn(struct pt_regs *regs);
251extern long sys32_rt_sigreturn(struct pt_regs *regs);
252
253static int signal_return(struct mm_struct *mm, struct pt_regs *regs, 246static int signal_return(struct mm_struct *mm, struct pt_regs *regs,
254 unsigned long address, unsigned long error_code) 247 unsigned long address, unsigned long error_code)
255{ 248{
@@ -270,15 +263,15 @@ static int signal_return(struct mm_struct *mm, struct pt_regs *regs,
270#ifdef CONFIG_COMPAT 263#ifdef CONFIG_COMPAT
271 compat = test_tsk_thread_flag(current, TIF_31BIT); 264 compat = test_tsk_thread_flag(current, TIF_31BIT);
272 if (compat && instruction == 0x0a77) 265 if (compat && instruction == 0x0a77)
273 sys32_sigreturn(regs); 266 sys32_sigreturn();
274 else if (compat && instruction == 0x0aad) 267 else if (compat && instruction == 0x0aad)
275 sys32_rt_sigreturn(regs); 268 sys32_rt_sigreturn();
276 else 269 else
277#endif 270#endif
278 if (instruction == 0x0a77) 271 if (instruction == 0x0a77)
279 sys_sigreturn(regs); 272 sys_sigreturn();
280 else if (instruction == 0x0aad) 273 else if (instruction == 0x0aad)
281 sys_rt_sigreturn(regs); 274 sys_rt_sigreturn();
282 else { 275 else {
283 current->thread.prot_addr = address; 276 current->thread.prot_addr = address;
284 current->thread.trap_no = error_code; 277 current->thread.trap_no = error_code;
@@ -424,7 +417,7 @@ no_context:
424} 417}
425 418
426void __kprobes do_protection_exception(struct pt_regs *regs, 419void __kprobes do_protection_exception(struct pt_regs *regs,
427 unsigned long error_code) 420 long error_code)
428{ 421{
429 /* Protection exception is supressing, decrement psw address. */ 422 /* Protection exception is supressing, decrement psw address. */
430 regs->psw.addr -= (error_code >> 16); 423 regs->psw.addr -= (error_code >> 16);
@@ -440,7 +433,7 @@ void __kprobes do_protection_exception(struct pt_regs *regs,
440 do_exception(regs, 4, 1); 433 do_exception(regs, 4, 1);
441} 434}
442 435
443void __kprobes do_dat_exception(struct pt_regs *regs, unsigned long error_code) 436void __kprobes do_dat_exception(struct pt_regs *regs, long error_code)
444{ 437{
445 do_exception(regs, error_code & 0xff, 0); 438 do_exception(regs, error_code & 0xff, 0);
446} 439}
diff --git a/arch/s390/mm/init.c b/arch/s390/mm/init.c
index 8053245fe259..202c952a29b4 100644
--- a/arch/s390/mm/init.c
+++ b/arch/s390/mm/init.c
@@ -50,7 +50,6 @@ void show_mem(void)
50 50
51 printk("Mem-info:\n"); 51 printk("Mem-info:\n");
52 show_free_areas(); 52 show_free_areas();
53 printk("Free swap: %6ldkB\n", nr_swap_pages << (PAGE_SHIFT - 10));
54 i = max_mapnr; 53 i = max_mapnr;
55 while (i-- > 0) { 54 while (i-- > 0) {
56 if (!pfn_valid(i)) 55 if (!pfn_valid(i))
diff --git a/arch/sh/kernel/Makefile_32 b/arch/sh/kernel/Makefile_32
index 62bf373266f7..4bbdce36b92b 100644
--- a/arch/sh/kernel/Makefile_32
+++ b/arch/sh/kernel/Makefile_32
@@ -5,7 +5,7 @@
5extra-y := head_32.o init_task.o vmlinux.lds 5extra-y := head_32.o init_task.o vmlinux.lds
6 6
7obj-y := debugtraps.o io.o io_generic.o irq.o machvec.o process_32.o \ 7obj-y := debugtraps.o io.o io_generic.o irq.o machvec.o process_32.o \
8 ptrace_32.o semaphore.o setup.o signal_32.o sys_sh.o sys_sh32.o \ 8 ptrace_32.o setup.o signal_32.o sys_sh.o sys_sh32.o \
9 syscalls_32.o time_32.o topology.o traps.o traps_32.o 9 syscalls_32.o time_32.o topology.o traps.o traps_32.o
10 10
11obj-y += cpu/ timers/ 11obj-y += cpu/ timers/
diff --git a/arch/sh/kernel/Makefile_64 b/arch/sh/kernel/Makefile_64
index e01283d49cbf..6edf53b93d94 100644
--- a/arch/sh/kernel/Makefile_64
+++ b/arch/sh/kernel/Makefile_64
@@ -1,7 +1,7 @@
1extra-y := head_64.o init_task.o vmlinux.lds 1extra-y := head_64.o init_task.o vmlinux.lds
2 2
3obj-y := debugtraps.o io.o io_generic.o irq.o machvec.o process_64.o \ 3obj-y := debugtraps.o io.o io_generic.o irq.o machvec.o process_64.o \
4 ptrace_64.o semaphore.o setup.o signal_64.o sys_sh.o sys_sh64.o \ 4 ptrace_64.o setup.o signal_64.o sys_sh.o sys_sh64.o \
5 syscalls_64.o time_64.o topology.o traps.o traps_64.o 5 syscalls_64.o time_64.o topology.o traps.o traps_64.o
6 6
7obj-y += cpu/ timers/ 7obj-y += cpu/ timers/
diff --git a/arch/sh/kernel/semaphore.c b/arch/sh/kernel/semaphore.c
deleted file mode 100644
index 184119eeae56..000000000000
--- a/arch/sh/kernel/semaphore.c
+++ /dev/null
@@ -1,139 +0,0 @@
1/*
2 * Just taken from alpha implementation.
3 * This can't work well, perhaps.
4 */
5/*
6 * Generic semaphore code. Buyer beware. Do your own
7 * specific changes in <asm/semaphore-helper.h>
8 */
9
10#include <linux/errno.h>
11#include <linux/sched.h>
12#include <linux/wait.h>
13#include <linux/init.h>
14#include <asm/semaphore.h>
15#include <asm/semaphore-helper.h>
16
17DEFINE_SPINLOCK(semaphore_wake_lock);
18
19/*
20 * Semaphores are implemented using a two-way counter:
21 * The "count" variable is decremented for each process
22 * that tries to sleep, while the "waking" variable is
23 * incremented when the "up()" code goes to wake up waiting
24 * processes.
25 *
26 * Notably, the inline "up()" and "down()" functions can
27 * efficiently test if they need to do any extra work (up
28 * needs to do something only if count was negative before
29 * the increment operation.
30 *
31 * waking_non_zero() (from asm/semaphore.h) must execute
32 * atomically.
33 *
34 * When __up() is called, the count was negative before
35 * incrementing it, and we need to wake up somebody.
36 *
37 * This routine adds one to the count of processes that need to
38 * wake up and exit. ALL waiting processes actually wake up but
39 * only the one that gets to the "waking" field first will gate
40 * through and acquire the semaphore. The others will go back
41 * to sleep.
42 *
43 * Note that these functions are only called when there is
44 * contention on the lock, and as such all this is the
45 * "non-critical" part of the whole semaphore business. The
46 * critical part is the inline stuff in <asm/semaphore.h>
47 * where we want to avoid any extra jumps and calls.
48 */
49void __up(struct semaphore *sem)
50{
51 wake_one_more(sem);
52 wake_up(&sem->wait);
53}
54
55/*
56 * Perform the "down" function. Return zero for semaphore acquired,
57 * return negative for signalled out of the function.
58 *
59 * If called from __down, the return is ignored and the wait loop is
60 * not interruptible. This means that a task waiting on a semaphore
61 * using "down()" cannot be killed until someone does an "up()" on
62 * the semaphore.
63 *
64 * If called from __down_interruptible, the return value gets checked
65 * upon return. If the return value is negative then the task continues
66 * with the negative value in the return register (it can be tested by
67 * the caller).
68 *
69 * Either form may be used in conjunction with "up()".
70 *
71 */
72
73#define DOWN_VAR \
74 struct task_struct *tsk = current; \
75 wait_queue_t wait; \
76 init_waitqueue_entry(&wait, tsk);
77
78#define DOWN_HEAD(task_state) \
79 \
80 \
81 tsk->state = (task_state); \
82 add_wait_queue(&sem->wait, &wait); \
83 \
84 /* \
85 * Ok, we're set up. sem->count is known to be less than zero \
86 * so we must wait. \
87 * \
88 * We can let go the lock for purposes of waiting. \
89 * We re-acquire it after awaking so as to protect \
90 * all semaphore operations. \
91 * \
92 * If "up()" is called before we call waking_non_zero() then \
93 * we will catch it right away. If it is called later then \
94 * we will have to go through a wakeup cycle to catch it. \
95 * \
96 * Multiple waiters contend for the semaphore lock to see \
97 * who gets to gate through and who has to wait some more. \
98 */ \
99 for (;;) {
100
101#define DOWN_TAIL(task_state) \
102 tsk->state = (task_state); \
103 } \
104 tsk->state = TASK_RUNNING; \
105 remove_wait_queue(&sem->wait, &wait);
106
107void __sched __down(struct semaphore * sem)
108{
109 DOWN_VAR
110 DOWN_HEAD(TASK_UNINTERRUPTIBLE)
111 if (waking_non_zero(sem))
112 break;
113 schedule();
114 DOWN_TAIL(TASK_UNINTERRUPTIBLE)
115}
116
117int __sched __down_interruptible(struct semaphore * sem)
118{
119 int ret = 0;
120 DOWN_VAR
121 DOWN_HEAD(TASK_INTERRUPTIBLE)
122
123 ret = waking_non_zero_interruptible(sem, tsk);
124 if (ret)
125 {
126 if (ret == 1)
127 /* ret != 0 only if we get interrupted -arca */
128 ret = 0;
129 break;
130 }
131 schedule();
132 DOWN_TAIL(TASK_INTERRUPTIBLE)
133 return ret;
134}
135
136int __down_trylock(struct semaphore * sem)
137{
138 return waking_non_zero_trylock(sem);
139}
diff --git a/arch/sh/kernel/sh_ksyms_32.c b/arch/sh/kernel/sh_ksyms_32.c
index 45bb333fd9ec..6d405462cee8 100644
--- a/arch/sh/kernel/sh_ksyms_32.c
+++ b/arch/sh/kernel/sh_ksyms_32.c
@@ -9,7 +9,6 @@
9#include <linux/pci.h> 9#include <linux/pci.h>
10#include <linux/irq.h> 10#include <linux/irq.h>
11#include <asm/sections.h> 11#include <asm/sections.h>
12#include <asm/semaphore.h>
13#include <asm/processor.h> 12#include <asm/processor.h>
14#include <asm/uaccess.h> 13#include <asm/uaccess.h>
15#include <asm/checksum.h> 14#include <asm/checksum.h>
@@ -48,12 +47,6 @@ EXPORT_SYMBOL(__copy_user);
48EXPORT_SYMBOL(get_vm_area); 47EXPORT_SYMBOL(get_vm_area);
49#endif 48#endif
50 49
51/* semaphore exports */
52EXPORT_SYMBOL(__up);
53EXPORT_SYMBOL(__down);
54EXPORT_SYMBOL(__down_interruptible);
55EXPORT_SYMBOL(__down_trylock);
56
57EXPORT_SYMBOL(__udelay); 50EXPORT_SYMBOL(__udelay);
58EXPORT_SYMBOL(__ndelay); 51EXPORT_SYMBOL(__ndelay);
59EXPORT_SYMBOL(__const_udelay); 52EXPORT_SYMBOL(__const_udelay);
diff --git a/arch/sh/kernel/sh_ksyms_64.c b/arch/sh/kernel/sh_ksyms_64.c
index b6410ce4bd1d..a310c9707f03 100644
--- a/arch/sh/kernel/sh_ksyms_64.c
+++ b/arch/sh/kernel/sh_ksyms_64.c
@@ -16,7 +16,6 @@
16#include <linux/in6.h> 16#include <linux/in6.h>
17#include <linux/interrupt.h> 17#include <linux/interrupt.h>
18#include <linux/screen_info.h> 18#include <linux/screen_info.h>
19#include <asm/semaphore.h>
20#include <asm/processor.h> 19#include <asm/processor.h>
21#include <asm/uaccess.h> 20#include <asm/uaccess.h>
22#include <asm/checksum.h> 21#include <asm/checksum.h>
@@ -37,9 +36,6 @@ EXPORT_SYMBOL(csum_partial_copy_nocheck);
37EXPORT_SYMBOL(screen_info); 36EXPORT_SYMBOL(screen_info);
38#endif 37#endif
39 38
40EXPORT_SYMBOL(__down);
41EXPORT_SYMBOL(__down_trylock);
42EXPORT_SYMBOL(__up);
43EXPORT_SYMBOL(__put_user_asm_l); 39EXPORT_SYMBOL(__put_user_asm_l);
44EXPORT_SYMBOL(__get_user_asm_l); 40EXPORT_SYMBOL(__get_user_asm_l);
45EXPORT_SYMBOL(copy_page); 41EXPORT_SYMBOL(copy_page);
diff --git a/arch/sparc/kernel/Makefile b/arch/sparc/kernel/Makefile
index bf1b15d3f6f5..2712bb166f6f 100644
--- a/arch/sparc/kernel/Makefile
+++ b/arch/sparc/kernel/Makefile
@@ -12,7 +12,7 @@ obj-y := entry.o wof.o wuf.o etrap.o rtrap.o traps.o $(IRQ_OBJS) \
12 sys_sparc.o sunos_asm.o systbls.o \ 12 sys_sparc.o sunos_asm.o systbls.o \
13 time.o windows.o cpu.o devices.o sclow.o \ 13 time.o windows.o cpu.o devices.o sclow.o \
14 tadpole.o tick14.o ptrace.o sys_solaris.o \ 14 tadpole.o tick14.o ptrace.o sys_solaris.o \
15 unaligned.o una_asm.o muldiv.o semaphore.o \ 15 unaligned.o una_asm.o muldiv.o \
16 prom.o of_device.o devres.o 16 prom.o of_device.o devres.o
17 17
18devres-y = ../../../kernel/irq/devres.o 18devres-y = ../../../kernel/irq/devres.o
diff --git a/arch/sparc/kernel/semaphore.c b/arch/sparc/kernel/semaphore.c
deleted file mode 100644
index 0c37c1a7cd7e..000000000000
--- a/arch/sparc/kernel/semaphore.c
+++ /dev/null
@@ -1,155 +0,0 @@
1/* $Id: semaphore.c,v 1.7 2001/04/18 21:06:05 davem Exp $ */
2
3/* sparc32 semaphore implementation, based on i386 version */
4
5#include <linux/sched.h>
6#include <linux/errno.h>
7#include <linux/init.h>
8
9#include <asm/semaphore.h>
10
11/*
12 * Semaphores are implemented using a two-way counter:
13 * The "count" variable is decremented for each process
14 * that tries to acquire the semaphore, while the "sleeping"
15 * variable is a count of such acquires.
16 *
17 * Notably, the inline "up()" and "down()" functions can
18 * efficiently test if they need to do any extra work (up
19 * needs to do something only if count was negative before
20 * the increment operation.
21 *
22 * "sleeping" and the contention routine ordering is
23 * protected by the semaphore spinlock.
24 *
25 * Note that these functions are only called when there is
26 * contention on the lock, and as such all this is the
27 * "non-critical" part of the whole semaphore business. The
28 * critical part is the inline stuff in <asm/semaphore.h>
29 * where we want to avoid any extra jumps and calls.
30 */
31
32/*
33 * Logic:
34 * - only on a boundary condition do we need to care. When we go
35 * from a negative count to a non-negative, we wake people up.
36 * - when we go from a non-negative count to a negative do we
37 * (a) synchronize with the "sleeper" count and (b) make sure
38 * that we're on the wakeup list before we synchronize so that
39 * we cannot lose wakeup events.
40 */
41
42void __up(struct semaphore *sem)
43{
44 wake_up(&sem->wait);
45}
46
47static DEFINE_SPINLOCK(semaphore_lock);
48
49void __sched __down(struct semaphore * sem)
50{
51 struct task_struct *tsk = current;
52 DECLARE_WAITQUEUE(wait, tsk);
53 tsk->state = TASK_UNINTERRUPTIBLE;
54 add_wait_queue_exclusive(&sem->wait, &wait);
55
56 spin_lock_irq(&semaphore_lock);
57 sem->sleepers++;
58 for (;;) {
59 int sleepers = sem->sleepers;
60
61 /*
62 * Add "everybody else" into it. They aren't
63 * playing, because we own the spinlock.
64 */
65 if (!atomic24_add_negative(sleepers - 1, &sem->count)) {
66 sem->sleepers = 0;
67 break;
68 }
69 sem->sleepers = 1; /* us - see -1 above */
70 spin_unlock_irq(&semaphore_lock);
71
72 schedule();
73 tsk->state = TASK_UNINTERRUPTIBLE;
74 spin_lock_irq(&semaphore_lock);
75 }
76 spin_unlock_irq(&semaphore_lock);
77 remove_wait_queue(&sem->wait, &wait);
78 tsk->state = TASK_RUNNING;
79 wake_up(&sem->wait);
80}
81
82int __sched __down_interruptible(struct semaphore * sem)
83{
84 int retval = 0;
85 struct task_struct *tsk = current;
86 DECLARE_WAITQUEUE(wait, tsk);
87 tsk->state = TASK_INTERRUPTIBLE;
88 add_wait_queue_exclusive(&sem->wait, &wait);
89
90 spin_lock_irq(&semaphore_lock);
91 sem->sleepers ++;
92 for (;;) {
93 int sleepers = sem->sleepers;
94
95 /*
96 * With signals pending, this turns into
97 * the trylock failure case - we won't be
98 * sleeping, and we* can't get the lock as
99 * it has contention. Just correct the count
100 * and exit.
101 */
102 if (signal_pending(current)) {
103 retval = -EINTR;
104 sem->sleepers = 0;
105 atomic24_add(sleepers, &sem->count);
106 break;
107 }
108
109 /*
110 * Add "everybody else" into it. They aren't
111 * playing, because we own the spinlock. The
112 * "-1" is because we're still hoping to get
113 * the lock.
114 */
115 if (!atomic24_add_negative(sleepers - 1, &sem->count)) {
116 sem->sleepers = 0;
117 break;
118 }
119 sem->sleepers = 1; /* us - see -1 above */
120 spin_unlock_irq(&semaphore_lock);
121
122 schedule();
123 tsk->state = TASK_INTERRUPTIBLE;
124 spin_lock_irq(&semaphore_lock);
125 }
126 spin_unlock_irq(&semaphore_lock);
127 tsk->state = TASK_RUNNING;
128 remove_wait_queue(&sem->wait, &wait);
129 wake_up(&sem->wait);
130 return retval;
131}
132
133/*
134 * Trylock failed - make sure we correct for
135 * having decremented the count.
136 */
137int __down_trylock(struct semaphore * sem)
138{
139 int sleepers;
140 unsigned long flags;
141
142 spin_lock_irqsave(&semaphore_lock, flags);
143 sleepers = sem->sleepers + 1;
144 sem->sleepers = 0;
145
146 /*
147 * Add "everybody else" and us into it. They aren't
148 * playing, because we own the spinlock.
149 */
150 if (!atomic24_add_negative(sleepers, &sem->count))
151 wake_up(&sem->wait);
152
153 spin_unlock_irqrestore(&semaphore_lock, flags);
154 return 1;
155}
diff --git a/arch/sparc/kernel/sparc_ksyms.c b/arch/sparc/kernel/sparc_ksyms.c
index c1025e551650..97b1de0e9094 100644
--- a/arch/sparc/kernel/sparc_ksyms.c
+++ b/arch/sparc/kernel/sparc_ksyms.c
@@ -107,11 +107,6 @@ EXPORT_SYMBOL(___rw_read_try);
107EXPORT_SYMBOL(___rw_read_exit); 107EXPORT_SYMBOL(___rw_read_exit);
108EXPORT_SYMBOL(___rw_write_enter); 108EXPORT_SYMBOL(___rw_write_enter);
109#endif 109#endif
110/* semaphores */
111EXPORT_SYMBOL(__up);
112EXPORT_SYMBOL(__down);
113EXPORT_SYMBOL(__down_trylock);
114EXPORT_SYMBOL(__down_interruptible);
115 110
116EXPORT_SYMBOL(sparc_valid_addr_bitmap); 111EXPORT_SYMBOL(sparc_valid_addr_bitmap);
117EXPORT_SYMBOL(phys_base); 112EXPORT_SYMBOL(phys_base);
diff --git a/arch/sparc64/kernel/Makefile b/arch/sparc64/kernel/Makefile
index 1bf5b187de49..459462e80a12 100644
--- a/arch/sparc64/kernel/Makefile
+++ b/arch/sparc64/kernel/Makefile
@@ -10,7 +10,7 @@ extra-y := head.o init_task.o vmlinux.lds
10obj-y := process.o setup.o cpu.o idprom.o \ 10obj-y := process.o setup.o cpu.o idprom.o \
11 traps.o auxio.o una_asm.o sysfs.o iommu.o \ 11 traps.o auxio.o una_asm.o sysfs.o iommu.o \
12 irq.o ptrace.o time.o sys_sparc.o signal.o \ 12 irq.o ptrace.o time.o sys_sparc.o signal.o \
13 unaligned.o central.o pci.o starfire.o semaphore.o \ 13 unaligned.o central.o pci.o starfire.o \
14 power.o sbus.o sparc64_ksyms.o chmc.o \ 14 power.o sbus.o sparc64_ksyms.o chmc.o \
15 visemul.o prom.o of_device.o hvapi.o sstate.o mdesc.o 15 visemul.o prom.o of_device.o hvapi.o sstate.o mdesc.o
16 16
diff --git a/arch/sparc64/kernel/semaphore.c b/arch/sparc64/kernel/semaphore.c
deleted file mode 100644
index 9974a6899551..000000000000
--- a/arch/sparc64/kernel/semaphore.c
+++ /dev/null
@@ -1,254 +0,0 @@
1/* semaphore.c: Sparc64 semaphore implementation.
2 *
3 * This is basically the PPC semaphore scheme ported to use
4 * the sparc64 atomic instructions, so see the PPC code for
5 * credits.
6 */
7
8#include <linux/sched.h>
9#include <linux/errno.h>
10#include <linux/init.h>
11
12/*
13 * Atomically update sem->count.
14 * This does the equivalent of the following:
15 *
16 * old_count = sem->count;
17 * tmp = MAX(old_count, 0) + incr;
18 * sem->count = tmp;
19 * return old_count;
20 */
21static inline int __sem_update_count(struct semaphore *sem, int incr)
22{
23 int old_count, tmp;
24
25 __asm__ __volatile__("\n"
26" ! __sem_update_count old_count(%0) tmp(%1) incr(%4) &sem->count(%3)\n"
27"1: ldsw [%3], %0\n"
28" mov %0, %1\n"
29" cmp %0, 0\n"
30" movl %%icc, 0, %1\n"
31" add %1, %4, %1\n"
32" cas [%3], %0, %1\n"
33" cmp %0, %1\n"
34" membar #StoreLoad | #StoreStore\n"
35" bne,pn %%icc, 1b\n"
36" nop\n"
37 : "=&r" (old_count), "=&r" (tmp), "=m" (sem->count)
38 : "r" (&sem->count), "r" (incr), "m" (sem->count)
39 : "cc");
40
41 return old_count;
42}
43
44static void __up(struct semaphore *sem)
45{
46 __sem_update_count(sem, 1);
47 wake_up(&sem->wait);
48}
49
50void up(struct semaphore *sem)
51{
52 /* This atomically does:
53 * old_val = sem->count;
54 * new_val = sem->count + 1;
55 * sem->count = new_val;
56 * if (old_val < 0)
57 * __up(sem);
58 *
59 * The (old_val < 0) test is equivalent to
60 * the more straightforward (new_val <= 0),
61 * but it is easier to test the former because
62 * of how the CAS instruction works.
63 */
64
65 __asm__ __volatile__("\n"
66" ! up sem(%0)\n"
67" membar #StoreLoad | #LoadLoad\n"
68"1: lduw [%0], %%g1\n"
69" add %%g1, 1, %%g7\n"
70" cas [%0], %%g1, %%g7\n"
71" cmp %%g1, %%g7\n"
72" bne,pn %%icc, 1b\n"
73" addcc %%g7, 1, %%g0\n"
74" membar #StoreLoad | #StoreStore\n"
75" ble,pn %%icc, 3f\n"
76" nop\n"
77"2:\n"
78" .subsection 2\n"
79"3: mov %0, %%g1\n"
80" save %%sp, -160, %%sp\n"
81" call %1\n"
82" mov %%g1, %%o0\n"
83" ba,pt %%xcc, 2b\n"
84" restore\n"
85" .previous\n"
86 : : "r" (sem), "i" (__up)
87 : "g1", "g2", "g3", "g7", "memory", "cc");
88}
89
90static void __sched __down(struct semaphore * sem)
91{
92 struct task_struct *tsk = current;
93 DECLARE_WAITQUEUE(wait, tsk);
94
95 tsk->state = TASK_UNINTERRUPTIBLE;
96 add_wait_queue_exclusive(&sem->wait, &wait);
97
98 while (__sem_update_count(sem, -1) <= 0) {
99 schedule();
100 tsk->state = TASK_UNINTERRUPTIBLE;
101 }
102 remove_wait_queue(&sem->wait, &wait);
103 tsk->state = TASK_RUNNING;
104
105 wake_up(&sem->wait);
106}
107
108void __sched down(struct semaphore *sem)
109{
110 might_sleep();
111 /* This atomically does:
112 * old_val = sem->count;
113 * new_val = sem->count - 1;
114 * sem->count = new_val;
115 * if (old_val < 1)
116 * __down(sem);
117 *
118 * The (old_val < 1) test is equivalent to
119 * the more straightforward (new_val < 0),
120 * but it is easier to test the former because
121 * of how the CAS instruction works.
122 */
123
124 __asm__ __volatile__("\n"
125" ! down sem(%0)\n"
126"1: lduw [%0], %%g1\n"
127" sub %%g1, 1, %%g7\n"
128" cas [%0], %%g1, %%g7\n"
129" cmp %%g1, %%g7\n"
130" bne,pn %%icc, 1b\n"
131" cmp %%g7, 1\n"
132" membar #StoreLoad | #StoreStore\n"
133" bl,pn %%icc, 3f\n"
134" nop\n"
135"2:\n"
136" .subsection 2\n"
137"3: mov %0, %%g1\n"
138" save %%sp, -160, %%sp\n"
139" call %1\n"
140" mov %%g1, %%o0\n"
141" ba,pt %%xcc, 2b\n"
142" restore\n"
143" .previous\n"
144 : : "r" (sem), "i" (__down)
145 : "g1", "g2", "g3", "g7", "memory", "cc");
146}
147
148int down_trylock(struct semaphore *sem)
149{
150 int ret;
151
152 /* This atomically does:
153 * old_val = sem->count;
154 * new_val = sem->count - 1;
155 * if (old_val < 1) {
156 * ret = 1;
157 * } else {
158 * sem->count = new_val;
159 * ret = 0;
160 * }
161 *
162 * The (old_val < 1) test is equivalent to
163 * the more straightforward (new_val < 0),
164 * but it is easier to test the former because
165 * of how the CAS instruction works.
166 */
167
168 __asm__ __volatile__("\n"
169" ! down_trylock sem(%1) ret(%0)\n"
170"1: lduw [%1], %%g1\n"
171" sub %%g1, 1, %%g7\n"
172" cmp %%g1, 1\n"
173" bl,pn %%icc, 2f\n"
174" mov 1, %0\n"
175" cas [%1], %%g1, %%g7\n"
176" cmp %%g1, %%g7\n"
177" bne,pn %%icc, 1b\n"
178" mov 0, %0\n"
179" membar #StoreLoad | #StoreStore\n"
180"2:\n"
181 : "=&r" (ret)
182 : "r" (sem)
183 : "g1", "g7", "memory", "cc");
184
185 return ret;
186}
187
188static int __sched __down_interruptible(struct semaphore * sem)
189{
190 int retval = 0;
191 struct task_struct *tsk = current;
192 DECLARE_WAITQUEUE(wait, tsk);
193
194 tsk->state = TASK_INTERRUPTIBLE;
195 add_wait_queue_exclusive(&sem->wait, &wait);
196
197 while (__sem_update_count(sem, -1) <= 0) {
198 if (signal_pending(current)) {
199 __sem_update_count(sem, 0);
200 retval = -EINTR;
201 break;
202 }
203 schedule();
204 tsk->state = TASK_INTERRUPTIBLE;
205 }
206 tsk->state = TASK_RUNNING;
207 remove_wait_queue(&sem->wait, &wait);
208 wake_up(&sem->wait);
209 return retval;
210}
211
212int __sched down_interruptible(struct semaphore *sem)
213{
214 int ret = 0;
215
216 might_sleep();
217 /* This atomically does:
218 * old_val = sem->count;
219 * new_val = sem->count - 1;
220 * sem->count = new_val;
221 * if (old_val < 1)
222 * ret = __down_interruptible(sem);
223 *
224 * The (old_val < 1) test is equivalent to
225 * the more straightforward (new_val < 0),
226 * but it is easier to test the former because
227 * of how the CAS instruction works.
228 */
229
230 __asm__ __volatile__("\n"
231" ! down_interruptible sem(%2) ret(%0)\n"
232"1: lduw [%2], %%g1\n"
233" sub %%g1, 1, %%g7\n"
234" cas [%2], %%g1, %%g7\n"
235" cmp %%g1, %%g7\n"
236" bne,pn %%icc, 1b\n"
237" cmp %%g7, 1\n"
238" membar #StoreLoad | #StoreStore\n"
239" bl,pn %%icc, 3f\n"
240" nop\n"
241"2:\n"
242" .subsection 2\n"
243"3: mov %2, %%g1\n"
244" save %%sp, -160, %%sp\n"
245" call %3\n"
246" mov %%g1, %%o0\n"
247" ba,pt %%xcc, 2b\n"
248" restore\n"
249" .previous\n"
250 : "=r" (ret)
251 : "0" (ret), "r" (sem), "i" (__down_interruptible)
252 : "g1", "g2", "g3", "g7", "memory", "cc");
253 return ret;
254}
diff --git a/arch/sparc64/kernel/sparc64_ksyms.c b/arch/sparc64/kernel/sparc64_ksyms.c
index 51fa773f38c9..051b8d9cb989 100644
--- a/arch/sparc64/kernel/sparc64_ksyms.c
+++ b/arch/sparc64/kernel/sparc64_ksyms.c
@@ -130,12 +130,6 @@ EXPORT_SYMBOL(_mcount);
130 130
131EXPORT_SYMBOL(sparc64_get_clock_tick); 131EXPORT_SYMBOL(sparc64_get_clock_tick);
132 132
133/* semaphores */
134EXPORT_SYMBOL(down);
135EXPORT_SYMBOL(down_trylock);
136EXPORT_SYMBOL(down_interruptible);
137EXPORT_SYMBOL(up);
138
139/* RW semaphores */ 133/* RW semaphores */
140EXPORT_SYMBOL(__down_read); 134EXPORT_SYMBOL(__down_read);
141EXPORT_SYMBOL(__down_read_trylock); 135EXPORT_SYMBOL(__down_read_trylock);
diff --git a/arch/um/Kconfig.i386 b/arch/um/Kconfig.i386
index 3cd8a04d66d8..e09edfa560da 100644
--- a/arch/um/Kconfig.i386
+++ b/arch/um/Kconfig.i386
@@ -19,10 +19,6 @@ config 64BIT
19 bool 19 bool
20 default n 20 default n
21 21
22config SEMAPHORE_SLEEPERS
23 bool
24 default y
25
26config 3_LEVEL_PGTABLES 22config 3_LEVEL_PGTABLES
27 bool "Three-level pagetables (EXPERIMENTAL)" 23 bool "Three-level pagetables (EXPERIMENTAL)"
28 default n 24 default n
diff --git a/arch/um/Kconfig.x86_64 b/arch/um/Kconfig.x86_64
index 6533b349f061..3fbe69e359ed 100644
--- a/arch/um/Kconfig.x86_64
+++ b/arch/um/Kconfig.x86_64
@@ -11,10 +11,6 @@ config RWSEM_GENERIC_SPINLOCK
11 bool 11 bool
12 default y 12 default y
13 13
14config SEMAPHORE_SLEEPERS
15 bool
16 default y
17
18config 3_LEVEL_PGTABLES 14config 3_LEVEL_PGTABLES
19 bool 15 bool
20 default y 16 default y
diff --git a/arch/um/sys-i386/ksyms.c b/arch/um/sys-i386/ksyms.c
index 2a1eac1859ce..bfbefd30db8f 100644
--- a/arch/um/sys-i386/ksyms.c
+++ b/arch/um/sys-i386/ksyms.c
@@ -1,17 +1,5 @@
1#include "linux/module.h" 1#include "linux/module.h"
2#include "linux/in6.h"
3#include "linux/rwsem.h"
4#include "asm/byteorder.h"
5#include "asm/delay.h"
6#include "asm/semaphore.h"
7#include "asm/uaccess.h"
8#include "asm/checksum.h" 2#include "asm/checksum.h"
9#include "asm/errno.h"
10
11EXPORT_SYMBOL(__down_failed);
12EXPORT_SYMBOL(__down_failed_interruptible);
13EXPORT_SYMBOL(__down_failed_trylock);
14EXPORT_SYMBOL(__up_wakeup);
15 3
16/* Networking helper routines. */ 4/* Networking helper routines. */
17EXPORT_SYMBOL(csum_partial); 5EXPORT_SYMBOL(csum_partial);
diff --git a/arch/um/sys-ppc/Makefile b/arch/um/sys-ppc/Makefile
index 08901526e893..b8bc844fd2c4 100644
--- a/arch/um/sys-ppc/Makefile
+++ b/arch/um/sys-ppc/Makefile
@@ -3,7 +3,7 @@ OBJ = built-in.o
3.S.o: 3.S.o:
4 $(CC) $(KBUILD_AFLAGS) -D__ASSEMBLY__ -D__UM_PPC__ -c $< -o $*.o 4 $(CC) $(KBUILD_AFLAGS) -D__ASSEMBLY__ -D__UM_PPC__ -c $< -o $*.o
5 5
6OBJS = ptrace.o sigcontext.o semaphore.o checksum.o miscthings.o misc.o \ 6OBJS = ptrace.o sigcontext.o checksum.o miscthings.o misc.o \
7 ptrace_user.o sysrq.o 7 ptrace_user.o sysrq.o
8 8
9EXTRA_AFLAGS := -DCONFIG_PPC32 -I. -I$(srctree)/arch/ppc/kernel 9EXTRA_AFLAGS := -DCONFIG_PPC32 -I. -I$(srctree)/arch/ppc/kernel
@@ -20,10 +20,6 @@ ptrace_user.o: ptrace_user.c
20sigcontext.o: sigcontext.c 20sigcontext.o: sigcontext.c
21 $(CC) $(USER_CFLAGS) $(EXTRA_CFLAGS) -c -o $@ $< 21 $(CC) $(USER_CFLAGS) $(EXTRA_CFLAGS) -c -o $@ $<
22 22
23semaphore.c:
24 rm -f $@
25 ln -s $(srctree)/arch/ppc/kernel/$@ $@
26
27checksum.S: 23checksum.S:
28 rm -f $@ 24 rm -f $@
29 ln -s $(srctree)/arch/ppc/lib/$@ $@ 25 ln -s $(srctree)/arch/ppc/lib/$@ $@
@@ -66,4 +62,4 @@ misc.o: misc.S ppc_defs.h
66 $(CC) $(EXTRA_AFLAGS) $(KBUILD_AFLAGS) -D__ASSEMBLY__ -D__UM_PPC__ -c $< -o $*.o 62 $(CC) $(EXTRA_AFLAGS) $(KBUILD_AFLAGS) -D__ASSEMBLY__ -D__UM_PPC__ -c $< -o $*.o
67 rm -f asm 63 rm -f asm
68 64
69clean-files := $(OBJS) ppc_defs.h checksum.S semaphore.c mk_defs.c 65clean-files := $(OBJS) ppc_defs.h checksum.S mk_defs.c
diff --git a/arch/um/sys-x86_64/ksyms.c b/arch/um/sys-x86_64/ksyms.c
index 12c593607c59..4d7d1a812d8f 100644
--- a/arch/um/sys-x86_64/ksyms.c
+++ b/arch/um/sys-x86_64/ksyms.c
@@ -1,16 +1,5 @@
1#include "linux/module.h" 1#include "linux/module.h"
2#include "linux/in6.h" 2#include "asm/string.h"
3#include "linux/rwsem.h"
4#include "asm/byteorder.h"
5#include "asm/semaphore.h"
6#include "asm/uaccess.h"
7#include "asm/checksum.h"
8#include "asm/errno.h"
9
10EXPORT_SYMBOL(__down_failed);
11EXPORT_SYMBOL(__down_failed_interruptible);
12EXPORT_SYMBOL(__down_failed_trylock);
13EXPORT_SYMBOL(__up_wakeup);
14 3
15/*XXX: we need them because they would be exported by x86_64 */ 4/*XXX: we need them because they would be exported by x86_64 */
16EXPORT_SYMBOL(__memcpy); 5EXPORT_SYMBOL(__memcpy);
diff --git a/arch/v850/kernel/Makefile b/arch/v850/kernel/Makefile
index 3930482bddc4..da5889c53576 100644
--- a/arch/v850/kernel/Makefile
+++ b/arch/v850/kernel/Makefile
@@ -11,7 +11,7 @@
11 11
12extra-y := head.o init_task.o vmlinux.lds 12extra-y := head.o init_task.o vmlinux.lds
13 13
14obj-y += intv.o entry.o process.o syscalls.o time.o semaphore.o setup.o \ 14obj-y += intv.o entry.o process.o syscalls.o time.o setup.o \
15 signal.o irq.o mach.o ptrace.o bug.o 15 signal.o irq.o mach.o ptrace.o bug.o
16obj-$(CONFIG_MODULES) += module.o v850_ksyms.o 16obj-$(CONFIG_MODULES) += module.o v850_ksyms.o
17# chip-specific code 17# chip-specific code
diff --git a/arch/v850/kernel/semaphore.c b/arch/v850/kernel/semaphore.c
deleted file mode 100644
index fc89fd661c99..000000000000
--- a/arch/v850/kernel/semaphore.c
+++ /dev/null
@@ -1,166 +0,0 @@
1/*
2 * arch/v850/kernel/semaphore.c -- Semaphore support
3 *
4 * Copyright (C) 1998-2000 IBM Corporation
5 * Copyright (C) 1999 Linus Torvalds
6 *
7 * This file is subject to the terms and conditions of the GNU General
8 * Public License. See the file COPYING in the main directory of this
9 * archive for more details.
10 *
11 * This file is a copy of the s390 version, arch/s390/kernel/semaphore.c
12 * Author(s): Martin Schwidefsky
13 * which was derived from the i386 version, linux/arch/i386/kernel/semaphore.c
14 */
15
16#include <linux/errno.h>
17#include <linux/sched.h>
18#include <linux/init.h>
19
20#include <asm/semaphore.h>
21
22/*
23 * Semaphores are implemented using a two-way counter:
24 * The "count" variable is decremented for each process
25 * that tries to acquire the semaphore, while the "sleeping"
26 * variable is a count of such acquires.
27 *
28 * Notably, the inline "up()" and "down()" functions can
29 * efficiently test if they need to do any extra work (up
30 * needs to do something only if count was negative before
31 * the increment operation.
32 *
33 * "sleeping" and the contention routine ordering is
34 * protected by the semaphore spinlock.
35 *
36 * Note that these functions are only called when there is
37 * contention on the lock, and as such all this is the
38 * "non-critical" part of the whole semaphore business. The
39 * critical part is the inline stuff in <asm/semaphore.h>
40 * where we want to avoid any extra jumps and calls.
41 */
42
43/*
44 * Logic:
45 * - only on a boundary condition do we need to care. When we go
46 * from a negative count to a non-negative, we wake people up.
47 * - when we go from a non-negative count to a negative do we
48 * (a) synchronize with the "sleeper" count and (b) make sure
49 * that we're on the wakeup list before we synchronize so that
50 * we cannot lose wakeup events.
51 */
52
53void __up(struct semaphore *sem)
54{
55 wake_up(&sem->wait);
56}
57
58static DEFINE_SPINLOCK(semaphore_lock);
59
60void __sched __down(struct semaphore * sem)
61{
62 struct task_struct *tsk = current;
63 DECLARE_WAITQUEUE(wait, tsk);
64 tsk->state = TASK_UNINTERRUPTIBLE;
65 add_wait_queue_exclusive(&sem->wait, &wait);
66
67 spin_lock_irq(&semaphore_lock);
68 sem->sleepers++;
69 for (;;) {
70 int sleepers = sem->sleepers;
71
72 /*
73 * Add "everybody else" into it. They aren't
74 * playing, because we own the spinlock.
75 */
76 if (!atomic_add_negative(sleepers - 1, &sem->count)) {
77 sem->sleepers = 0;
78 break;
79 }
80 sem->sleepers = 1; /* us - see -1 above */
81 spin_unlock_irq(&semaphore_lock);
82
83 schedule();
84 tsk->state = TASK_UNINTERRUPTIBLE;
85 spin_lock_irq(&semaphore_lock);
86 }
87 spin_unlock_irq(&semaphore_lock);
88 remove_wait_queue(&sem->wait, &wait);
89 tsk->state = TASK_RUNNING;
90 wake_up(&sem->wait);
91}
92
93int __sched __down_interruptible(struct semaphore * sem)
94{
95 int retval = 0;
96 struct task_struct *tsk = current;
97 DECLARE_WAITQUEUE(wait, tsk);
98 tsk->state = TASK_INTERRUPTIBLE;
99 add_wait_queue_exclusive(&sem->wait, &wait);
100
101 spin_lock_irq(&semaphore_lock);
102 sem->sleepers ++;
103 for (;;) {
104 int sleepers = sem->sleepers;
105
106 /*
107 * With signals pending, this turns into
108 * the trylock failure case - we won't be
109 * sleeping, and we* can't get the lock as
110 * it has contention. Just correct the count
111 * and exit.
112 */
113 if (signal_pending(current)) {
114 retval = -EINTR;
115 sem->sleepers = 0;
116 atomic_add(sleepers, &sem->count);
117 break;
118 }
119
120 /*
121 * Add "everybody else" into it. They aren't
122 * playing, because we own the spinlock. The
123 * "-1" is because we're still hoping to get
124 * the lock.
125 */
126 if (!atomic_add_negative(sleepers - 1, &sem->count)) {
127 sem->sleepers = 0;
128 break;
129 }
130 sem->sleepers = 1; /* us - see -1 above */
131 spin_unlock_irq(&semaphore_lock);
132
133 schedule();
134 tsk->state = TASK_INTERRUPTIBLE;
135 spin_lock_irq(&semaphore_lock);
136 }
137 spin_unlock_irq(&semaphore_lock);
138 tsk->state = TASK_RUNNING;
139 remove_wait_queue(&sem->wait, &wait);
140 wake_up(&sem->wait);
141 return retval;
142}
143
144/*
145 * Trylock failed - make sure we correct for
146 * having decremented the count.
147 */
148int __down_trylock(struct semaphore * sem)
149{
150 unsigned long flags;
151 int sleepers;
152
153 spin_lock_irqsave(&semaphore_lock, flags);
154 sleepers = sem->sleepers + 1;
155 sem->sleepers = 0;
156
157 /*
158 * Add "everybody else" and us into it. They aren't
159 * playing, because we own the spinlock.
160 */
161 if (!atomic_add_negative(sleepers, &sem->count))
162 wake_up(&sem->wait);
163
164 spin_unlock_irqrestore(&semaphore_lock, flags);
165 return 1;
166}
diff --git a/arch/v850/kernel/v850_ksyms.c b/arch/v850/kernel/v850_ksyms.c
index 93575fdc874d..8d386a5dbc4a 100644
--- a/arch/v850/kernel/v850_ksyms.c
+++ b/arch/v850/kernel/v850_ksyms.c
@@ -11,7 +11,6 @@
11#include <asm/pgalloc.h> 11#include <asm/pgalloc.h>
12#include <asm/irq.h> 12#include <asm/irq.h>
13#include <asm/io.h> 13#include <asm/io.h>
14#include <asm/semaphore.h>
15#include <asm/checksum.h> 14#include <asm/checksum.h>
16#include <asm/current.h> 15#include <asm/current.h>
17 16
@@ -34,12 +33,6 @@ EXPORT_SYMBOL (memset);
34EXPORT_SYMBOL (memcpy); 33EXPORT_SYMBOL (memcpy);
35EXPORT_SYMBOL (memmove); 34EXPORT_SYMBOL (memmove);
36 35
37/* semaphores */
38EXPORT_SYMBOL (__down);
39EXPORT_SYMBOL (__down_interruptible);
40EXPORT_SYMBOL (__down_trylock);
41EXPORT_SYMBOL (__up);
42
43/* 36/*
44 * libgcc functions - functions that are used internally by the 37 * libgcc functions - functions that are used internally by the
45 * compiler... (prototypes are not correct though, but that 38 * compiler... (prototypes are not correct though, but that
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index 5c4c8d7a46cc..2a59dbb28248 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -54,9 +54,6 @@ config STACKTRACE_SUPPORT
54config HAVE_LATENCYTOP_SUPPORT 54config HAVE_LATENCYTOP_SUPPORT
55 def_bool y 55 def_bool y
56 56
57config SEMAPHORE_SLEEPERS
58 def_bool y
59
60config FAST_CMPXCHG_LOCAL 57config FAST_CMPXCHG_LOCAL
61 bool 58 bool
62 default y 59 default y
@@ -118,7 +115,7 @@ config ARCH_HAS_CPU_RELAX
118 def_bool y 115 def_bool y
119 116
120config HAVE_SETUP_PER_CPU_AREA 117config HAVE_SETUP_PER_CPU_AREA
121 def_bool X86_64 118 def_bool X86_64 || (X86_SMP && !X86_VOYAGER)
122 119
123config ARCH_HIBERNATION_POSSIBLE 120config ARCH_HIBERNATION_POSSIBLE
124 def_bool y 121 def_bool y
@@ -172,7 +169,7 @@ config X86_64_SMP
172config X86_HT 169config X86_HT
173 bool 170 bool
174 depends on SMP 171 depends on SMP
175 depends on (X86_32 && !(X86_VISWS || X86_VOYAGER)) || (X86_64 && !MK8) 172 depends on (X86_32 && !(X86_VISWS || X86_VOYAGER)) || X86_64
176 default y 173 default y
177 174
178config X86_BIOS_REBOOT 175config X86_BIOS_REBOOT
@@ -182,7 +179,7 @@ config X86_BIOS_REBOOT
182 179
183config X86_TRAMPOLINE 180config X86_TRAMPOLINE
184 bool 181 bool
185 depends on X86_SMP || (X86_VOYAGER && SMP) 182 depends on X86_SMP || (X86_VOYAGER && SMP) || (64BIT && ACPI_SLEEP)
186 default y 183 default y
187 184
188config KTIME_SCALAR 185config KTIME_SCALAR
@@ -242,8 +239,7 @@ config X86_ELAN
242 239
243config X86_VOYAGER 240config X86_VOYAGER
244 bool "Voyager (NCR)" 241 bool "Voyager (NCR)"
245 depends on X86_32 242 depends on X86_32 && (SMP || BROKEN)
246 select SMP if !BROKEN
247 help 243 help
248 Voyager is an MCA-based 32-way capable SMP architecture proprietary 244 Voyager is an MCA-based 32-way capable SMP architecture proprietary
249 to NCR Corp. Machine classes 345x/35xx/4100/51xx are Voyager-based. 245 to NCR Corp. Machine classes 345x/35xx/4100/51xx are Voyager-based.
@@ -255,9 +251,8 @@ config X86_VOYAGER
255 251
256config X86_NUMAQ 252config X86_NUMAQ
257 bool "NUMAQ (IBM/Sequent)" 253 bool "NUMAQ (IBM/Sequent)"
258 select SMP 254 depends on SMP && X86_32
259 select NUMA 255 select NUMA
260 depends on X86_32
261 help 256 help
262 This option is used for getting Linux to run on a (IBM/Sequent) NUMA 257 This option is used for getting Linux to run on a (IBM/Sequent) NUMA
263 multiquad box. This changes the way that processors are bootstrapped, 258 multiquad box. This changes the way that processors are bootstrapped,
@@ -328,8 +323,9 @@ config X86_RDC321X
328 323
329config X86_VSMP 324config X86_VSMP
330 bool "Support for ScaleMP vSMP" 325 bool "Support for ScaleMP vSMP"
331 depends on X86_64 && PCI 326 select PARAVIRT
332 help 327 depends on X86_64
328 help
333 Support for ScaleMP vSMP systems. Say 'Y' here if this kernel is 329 Support for ScaleMP vSMP systems. Say 'Y' here if this kernel is
334 supposed to run on these EM64T-based machines. Only choose this option 330 supposed to run on these EM64T-based machines. Only choose this option
335 if you have one of these machines. 331 if you have one of these machines.
@@ -384,6 +380,35 @@ config PARAVIRT
384 380
385endif 381endif
386 382
383config MEMTEST_BOOTPARAM
384 bool "Memtest boot parameter"
385 depends on X86_64
386 default y
387 help
388 This option adds a kernel parameter 'memtest', which allows memtest
389 to be disabled at boot. If this option is selected, memtest
390 functionality can be disabled with memtest=0 on the kernel
391 command line. The purpose of this option is to allow a single
392 kernel image to be distributed with memtest built in, but not
393 necessarily enabled.
394
395 If you are unsure how to answer this question, answer Y.
396
397config MEMTEST_BOOTPARAM_VALUE
398 int "Memtest boot parameter default value (0-4)"
399 depends on MEMTEST_BOOTPARAM
400 range 0 4
401 default 0
402 help
403 This option sets the default value for the kernel parameter
404 'memtest', which allows memtest to be disabled at boot. If this
405 option is set to 0 (zero), the memtest kernel parameter will
406 default to 0, disabling memtest at bootup. If this option is
407 set to 4, the memtest kernel parameter will default to 4,
408 enabling memtest at bootup, and use that as pattern number.
409
410 If you are unsure how to answer this question, answer 0.
411
387config ACPI_SRAT 412config ACPI_SRAT
388 def_bool y 413 def_bool y
389 depends on X86_32 && ACPI && NUMA && (X86_SUMMIT || X86_GENERICARCH) 414 depends on X86_32 && ACPI && NUMA && (X86_SUMMIT || X86_GENERICARCH)
@@ -508,7 +533,7 @@ config NR_CPUS
508 533
509config SCHED_SMT 534config SCHED_SMT
510 bool "SMT (Hyperthreading) scheduler support" 535 bool "SMT (Hyperthreading) scheduler support"
511 depends on (X86_64 && SMP) || (X86_32 && X86_HT) 536 depends on X86_HT
512 help 537 help
513 SMT scheduler support improves the CPU scheduler's decision making 538 SMT scheduler support improves the CPU scheduler's decision making
514 when dealing with Intel Pentium 4 chips with HyperThreading at a 539 when dealing with Intel Pentium 4 chips with HyperThreading at a
@@ -518,7 +543,7 @@ config SCHED_SMT
518config SCHED_MC 543config SCHED_MC
519 def_bool y 544 def_bool y
520 prompt "Multi-core scheduler support" 545 prompt "Multi-core scheduler support"
521 depends on (X86_64 && SMP) || (X86_32 && X86_HT) 546 depends on X86_HT
522 help 547 help
523 Multi-core scheduler support improves the CPU scheduler's decision 548 Multi-core scheduler support improves the CPU scheduler's decision
524 making when dealing with multi-core CPU chips at a cost of slightly 549 making when dealing with multi-core CPU chips at a cost of slightly
@@ -887,7 +912,7 @@ config NUMA_EMU
887 number of nodes. This is only useful for debugging. 912 number of nodes. This is only useful for debugging.
888 913
889config NODES_SHIFT 914config NODES_SHIFT
890 int 915 int "Max num nodes shift(1-15)"
891 range 1 15 if X86_64 916 range 1 15 if X86_64
892 default "6" if X86_64 917 default "6" if X86_64
893 default "4" if X86_NUMAQ 918 default "4" if X86_NUMAQ
@@ -1011,6 +1036,21 @@ config MTRR
1011 1036
1012 See <file:Documentation/mtrr.txt> for more information. 1037 See <file:Documentation/mtrr.txt> for more information.
1013 1038
1039config X86_PAT
1040 def_bool y
1041 prompt "x86 PAT support"
1042 depends on MTRR && NONPROMISC_DEVMEM
1043 help
1044 Use PAT attributes to setup page level cache control.
1045
1046 PATs are the modern equivalents of MTRRs and are much more
1047 flexible than MTRRs.
1048
1049 Say N here if you see bootup problems (boot crash, boot hang,
1050 spontaneous reboots) or a non-working video driver.
1051
1052 If unsure, say Y.
1053
1014config EFI 1054config EFI
1015 def_bool n 1055 def_bool n
1016 prompt "EFI runtime service support" 1056 prompt "EFI runtime service support"
@@ -1079,6 +1119,7 @@ source kernel/Kconfig.hz
1079 1119
1080config KEXEC 1120config KEXEC
1081 bool "kexec system call" 1121 bool "kexec system call"
1122 depends on X86_64 || X86_BIOS_REBOOT
1082 help 1123 help
1083 kexec is a system call that implements the ability to shutdown your 1124 kexec is a system call that implements the ability to shutdown your
1084 current kernel, and to start another kernel. It is like a reboot 1125 current kernel, and to start another kernel. It is like a reboot
@@ -1380,7 +1421,7 @@ endmenu
1380menu "Bus options (PCI etc.)" 1421menu "Bus options (PCI etc.)"
1381 1422
1382config PCI 1423config PCI
1383 bool "PCI support" if !X86_VISWS 1424 bool "PCI support" if !X86_VISWS && !X86_VSMP
1384 depends on !X86_VOYAGER 1425 depends on !X86_VOYAGER
1385 default y 1426 default y
1386 select ARCH_SUPPORTS_MSI if (X86_LOCAL_APIC && X86_IO_APIC) 1427 select ARCH_SUPPORTS_MSI if (X86_LOCAL_APIC && X86_IO_APIC)
diff --git a/arch/x86/Kconfig.cpu b/arch/x86/Kconfig.cpu
index 9304bfba7d45..57072f2716f9 100644
--- a/arch/x86/Kconfig.cpu
+++ b/arch/x86/Kconfig.cpu
@@ -388,7 +388,7 @@ config X86_OOSTORE
388# 388#
389config X86_P6_NOP 389config X86_P6_NOP
390 def_bool y 390 def_bool y
391 depends on (X86_64 || !X86_GENERIC) && (M686 || MPENTIUMII || MPENTIUMIII || MPENTIUMM || MCORE2 || MPENTIUM4) 391 depends on (X86_64 || !X86_GENERIC) && (M686 || MPENTIUMII || MPENTIUMIII || MPENTIUMM || MCORE2 || MPENTIUM4 || MPSC)
392 392
393config X86_TSC 393config X86_TSC
394 def_bool y 394 def_bool y
diff --git a/arch/x86/Kconfig.debug b/arch/x86/Kconfig.debug
index 702eb39901ca..610aaecc19f8 100644
--- a/arch/x86/Kconfig.debug
+++ b/arch/x86/Kconfig.debug
@@ -54,6 +54,18 @@ config DEBUG_PER_CPU_MAPS
54 54
55 Say N if unsure. 55 Say N if unsure.
56 56
57config X86_PTDUMP
58 bool "Export kernel pagetable layout to userspace via debugfs"
59 depends on DEBUG_KERNEL
60 select DEBUG_FS
61 help
62 Say Y here if you want to show the kernel pagetable layout in a
63 debugfs file. This information is only useful for kernel developers
64 who are working in architecture specific areas of the kernel.
65 It is probably not a good idea to enable this feature in a production
66 kernel.
67 If in doubt, say "N"
68
57config DEBUG_RODATA 69config DEBUG_RODATA
58 bool "Write protect kernel read-only data structures" 70 bool "Write protect kernel read-only data structures"
59 default y 71 default y
@@ -64,6 +76,18 @@ config DEBUG_RODATA
64 data. This is recommended so that we can catch kernel bugs sooner. 76 data. This is recommended so that we can catch kernel bugs sooner.
65 If in doubt, say "Y". 77 If in doubt, say "Y".
66 78
79config DIRECT_GBPAGES
80 bool "Enable gbpages-mapped kernel pagetables"
81 depends on DEBUG_KERNEL && EXPERIMENTAL && X86_64
82 help
83 Enable gigabyte pages support (if the CPU supports it). This can
84 improve the kernel's performance a tiny bit by reducing TLB
85 pressure.
86
87 This is experimental code.
88
89 If in doubt, say "N".
90
67config DEBUG_RODATA_TEST 91config DEBUG_RODATA_TEST
68 bool "Testcase for the DEBUG_RODATA feature" 92 bool "Testcase for the DEBUG_RODATA feature"
69 depends on DEBUG_RODATA 93 depends on DEBUG_RODATA
@@ -82,8 +106,8 @@ config DEBUG_NX_TEST
82 106
83config 4KSTACKS 107config 4KSTACKS
84 bool "Use 4Kb for kernel stacks instead of 8Kb" 108 bool "Use 4Kb for kernel stacks instead of 8Kb"
85 depends on DEBUG_KERNEL
86 depends on X86_32 109 depends on X86_32
110 default y
87 help 111 help
88 If you say Y here the kernel will use a 4Kb stacksize for the 112 If you say Y here the kernel will use a 4Kb stacksize for the
89 kernel stack attached to each process/thread. This facilitates 113 kernel stack attached to each process/thread. This facilitates
diff --git a/arch/x86/Makefile b/arch/x86/Makefile
index f1e739a43d41..3cff3c894cf3 100644
--- a/arch/x86/Makefile
+++ b/arch/x86/Makefile
@@ -151,7 +151,6 @@ mflags-y += -Iinclude/asm-x86/mach-default
151# 64 bit does not support subarch support - clear sub arch variables 151# 64 bit does not support subarch support - clear sub arch variables
152fcore-$(CONFIG_X86_64) := 152fcore-$(CONFIG_X86_64) :=
153mcore-$(CONFIG_X86_64) := 153mcore-$(CONFIG_X86_64) :=
154mflags-$(CONFIG_X86_64) :=
155 154
156KBUILD_CFLAGS += $(mflags-y) 155KBUILD_CFLAGS += $(mflags-y)
157KBUILD_AFLAGS += $(mflags-y) 156KBUILD_AFLAGS += $(mflags-y)
@@ -159,9 +158,9 @@ KBUILD_AFLAGS += $(mflags-y)
159### 158###
160# Kernel objects 159# Kernel objects
161 160
162head-y := arch/x86/kernel/head_$(BITS).o 161head-y := arch/x86/kernel/head_$(BITS).o
163head-$(CONFIG_X86_64) += arch/x86/kernel/head64.o 162head-y += arch/x86/kernel/head$(BITS).o
164head-y += arch/x86/kernel/init_task.o 163head-y += arch/x86/kernel/init_task.o
165 164
166libs-y += arch/x86/lib/ 165libs-y += arch/x86/lib/
167 166
diff --git a/arch/x86/boot/Makefile b/arch/x86/boot/Makefile
index f88458e83ef0..7ee102f9c4f8 100644
--- a/arch/x86/boot/Makefile
+++ b/arch/x86/boot/Makefile
@@ -30,7 +30,7 @@ subdir- := compressed
30 30
31setup-y += a20.o cmdline.o copy.o cpu.o cpucheck.o edd.o 31setup-y += a20.o cmdline.o copy.o cpu.o cpucheck.o edd.o
32setup-y += header.o main.o mca.o memory.o pm.o pmjump.o 32setup-y += header.o main.o mca.o memory.o pm.o pmjump.o
33setup-y += printf.o string.o tty.o video.o version.o 33setup-y += printf.o string.o tty.o video.o video-mode.o version.o
34setup-$(CONFIG_X86_APM_BOOT) += apm.o 34setup-$(CONFIG_X86_APM_BOOT) += apm.o
35setup-$(CONFIG_X86_VOYAGER) += voyager.o 35setup-$(CONFIG_X86_VOYAGER) += voyager.o
36 36
@@ -94,6 +94,20 @@ $(obj)/vmlinux.bin: $(obj)/compressed/vmlinux FORCE
94 94
95SETUP_OBJS = $(addprefix $(obj)/,$(setup-y)) 95SETUP_OBJS = $(addprefix $(obj)/,$(setup-y))
96 96
97sed-offsets := -e 's/^00*/0/' \
98 -e 's/^\([0-9a-fA-F]*\) . \(input_data\|input_data_end\)$$/\#define \2 0x\1/p'
99
100quiet_cmd_offsets = OFFSETS $@
101 cmd_offsets = $(NM) $< | sed -n $(sed-offsets) > $@
102
103$(obj)/offsets.h: $(obj)/compressed/vmlinux FORCE
104 $(call if_changed,offsets)
105
106targets += offsets.h
107
108AFLAGS_header.o += -I$(obj)
109$(obj)/header.o: $(obj)/offsets.h
110
97LDFLAGS_setup.elf := -T 111LDFLAGS_setup.elf := -T
98$(obj)/setup.elf: $(src)/setup.ld $(SETUP_OBJS) FORCE 112$(obj)/setup.elf: $(src)/setup.ld $(SETUP_OBJS) FORCE
99 $(call if_changed,ld) 113 $(call if_changed,ld)
diff --git a/arch/x86/boot/boot.h b/arch/x86/boot/boot.h
index 7822a4983da2..09578070bfba 100644
--- a/arch/x86/boot/boot.h
+++ b/arch/x86/boot/boot.h
@@ -286,6 +286,11 @@ int getchar_timeout(void);
286/* video.c */ 286/* video.c */
287void set_video(void); 287void set_video(void);
288 288
289/* video-mode.c */
290int set_mode(u16 mode);
291int mode_defined(u16 mode);
292void probe_cards(int unsafe);
293
289/* video-vesa.c */ 294/* video-vesa.c */
290void vesa_store_edid(void); 295void vesa_store_edid(void);
291 296
diff --git a/arch/x86/boot/compressed/Makefile b/arch/x86/boot/compressed/Makefile
index d2b9f3bb87c0..92fdd35bd93e 100644
--- a/arch/x86/boot/compressed/Makefile
+++ b/arch/x86/boot/compressed/Makefile
@@ -22,7 +22,7 @@ $(obj)/vmlinux: $(src)/vmlinux_$(BITS).lds $(obj)/head_$(BITS).o $(obj)/misc.o $
22 $(call if_changed,ld) 22 $(call if_changed,ld)
23 @: 23 @:
24 24
25OBJCOPYFLAGS_vmlinux.bin := -O binary -R .note -R .comment -S 25OBJCOPYFLAGS_vmlinux.bin := -R .comment -S
26$(obj)/vmlinux.bin: vmlinux FORCE 26$(obj)/vmlinux.bin: vmlinux FORCE
27 $(call if_changed,objcopy) 27 $(call if_changed,objcopy)
28 28
diff --git a/arch/x86/boot/compressed/misc.c b/arch/x86/boot/compressed/misc.c
index 8182e32c1b42..dad4e699f5a3 100644
--- a/arch/x86/boot/compressed/misc.c
+++ b/arch/x86/boot/compressed/misc.c
@@ -15,6 +15,10 @@
15 * we just keep it from happening 15 * we just keep it from happening
16 */ 16 */
17#undef CONFIG_PARAVIRT 17#undef CONFIG_PARAVIRT
18#ifdef CONFIG_X86_32
19#define _ASM_DESC_H_ 1
20#endif
21
18#ifdef CONFIG_X86_64 22#ifdef CONFIG_X86_64
19#define _LINUX_STRING_H_ 1 23#define _LINUX_STRING_H_ 1
20#define __LINUX_BITMAP_H 1 24#define __LINUX_BITMAP_H 1
@@ -22,6 +26,7 @@
22 26
23#include <linux/linkage.h> 27#include <linux/linkage.h>
24#include <linux/screen_info.h> 28#include <linux/screen_info.h>
29#include <linux/elf.h>
25#include <asm/io.h> 30#include <asm/io.h>
26#include <asm/page.h> 31#include <asm/page.h>
27#include <asm/boot.h> 32#include <asm/boot.h>
@@ -53,8 +58,8 @@
53 * 1 bit (last block flag) 58 * 1 bit (last block flag)
54 * 2 bits (block type) 59 * 2 bits (block type)
55 * 60 *
56 * 1 block occurs every 32K -1 bytes or when there 50% compression has been achieved. 61 * 1 block occurs every 32K -1 bytes or when there 50% compression
57 * The smallest block type encoding is always used. 62 * has been achieved. The smallest block type encoding is always used.
58 * 63 *
59 * stored: 64 * stored:
60 * 32 bits length in bytes. 65 * 32 bits length in bytes.
@@ -90,9 +95,9 @@
90 * 95 *
91 * All of which is enough to compute an amount of extra data that is required 96 * All of which is enough to compute an amount of extra data that is required
92 * to be safe. To avoid problems at the block level allocating 5 extra bytes 97 * to be safe. To avoid problems at the block level allocating 5 extra bytes
93 * per 32767 bytes of data is sufficient. To avoind problems internal to a block 98 * per 32767 bytes of data is sufficient. To avoind problems internal to a
94 * adding an extra 32767 bytes (the worst case uncompressed block size) is 99 * block adding an extra 32767 bytes (the worst case uncompressed block size)
95 * sufficient, to ensure that in the worst case the decompressed data for 100 * is sufficient, to ensure that in the worst case the decompressed data for
96 * block will stop the byte before the compressed data for a block begins. 101 * block will stop the byte before the compressed data for a block begins.
97 * To avoid problems with the compressed data's meta information an extra 18 102 * To avoid problems with the compressed data's meta information an extra 18
98 * bytes are needed. Leading to the formula: 103 * bytes are needed. Leading to the formula:
@@ -111,58 +116,66 @@
111 * gzip declarations 116 * gzip declarations
112 */ 117 */
113 118
114#define OF(args) args 119#define OF(args) args
115#define STATIC static 120#define STATIC static
116 121
117#undef memset 122#undef memset
118#undef memcpy 123#undef memcpy
119#define memzero(s, n) memset ((s), 0, (n)) 124#define memzero(s, n) memset((s), 0, (n))
125
126typedef unsigned char uch;
127typedef unsigned short ush;
128typedef unsigned long ulg;
129
130/*
131 * Window size must be at least 32k, and a power of two.
132 * We don't actually have a window just a huge output buffer,
133 * so we report a 2G window size, as that should always be
134 * larger than our output buffer:
135 */
136#define WSIZE 0x80000000
137
138/* Input buffer: */
139static unsigned char *inbuf;
120 140
121typedef unsigned char uch; 141/* Sliding window buffer (and final output buffer): */
122typedef unsigned short ush; 142static unsigned char *window;
123typedef unsigned long ulg;
124 143
125#define WSIZE 0x80000000 /* Window size must be at least 32k, 144/* Valid bytes in inbuf: */
126 * and a power of two 145static unsigned insize;
127 * We don't actually have a window just
128 * a huge output buffer so I report
129 * a 2G windows size, as that should
130 * always be larger than our output buffer.
131 */
132 146
133static uch *inbuf; /* input buffer */ 147/* Index of next byte to be processed in inbuf: */
134static uch *window; /* Sliding window buffer, (and final output buffer) */ 148static unsigned inptr;
135 149
136static unsigned insize; /* valid bytes in inbuf */ 150/* Bytes in output buffer: */
137static unsigned inptr; /* index of next byte to be processed in inbuf */ 151static unsigned outcnt;
138static unsigned outcnt; /* bytes in output buffer */
139 152
140/* gzip flag byte */ 153/* gzip flag byte */
141#define ASCII_FLAG 0x01 /* bit 0 set: file probably ASCII text */ 154#define ASCII_FLAG 0x01 /* bit 0 set: file probably ASCII text */
142#define CONTINUATION 0x02 /* bit 1 set: continuation of multi-part gzip file */ 155#define CONTINUATION 0x02 /* bit 1 set: continuation of multi-part gz file */
143#define EXTRA_FIELD 0x04 /* bit 2 set: extra field present */ 156#define EXTRA_FIELD 0x04 /* bit 2 set: extra field present */
144#define ORIG_NAME 0x08 /* bit 3 set: original file name present */ 157#define ORIG_NAM 0x08 /* bit 3 set: original file name present */
145#define COMMENT 0x10 /* bit 4 set: file comment present */ 158#define COMMENT 0x10 /* bit 4 set: file comment present */
146#define ENCRYPTED 0x20 /* bit 5 set: file is encrypted */ 159#define ENCRYPTED 0x20 /* bit 5 set: file is encrypted */
147#define RESERVED 0xC0 /* bit 6,7: reserved */ 160#define RESERVED 0xC0 /* bit 6, 7: reserved */
148 161
149#define get_byte() (inptr < insize ? inbuf[inptr++] : fill_inbuf()) 162#define get_byte() (inptr < insize ? inbuf[inptr++] : fill_inbuf())
150 163
151/* Diagnostic functions */ 164/* Diagnostic functions */
152#ifdef DEBUG 165#ifdef DEBUG
153# define Assert(cond,msg) {if(!(cond)) error(msg);} 166# define Assert(cond, msg) do { if (!(cond)) error(msg); } while (0)
154# define Trace(x) fprintf x 167# define Trace(x) do { fprintf x; } while (0)
155# define Tracev(x) {if (verbose) fprintf x ;} 168# define Tracev(x) do { if (verbose) fprintf x ; } while (0)
156# define Tracevv(x) {if (verbose>1) fprintf x ;} 169# define Tracevv(x) do { if (verbose > 1) fprintf x ; } while (0)
157# define Tracec(c,x) {if (verbose && (c)) fprintf x ;} 170# define Tracec(c, x) do { if (verbose && (c)) fprintf x ; } while (0)
158# define Tracecv(c,x) {if (verbose>1 && (c)) fprintf x ;} 171# define Tracecv(c, x) do { if (verbose > 1 && (c)) fprintf x ; } while (0)
159#else 172#else
160# define Assert(cond,msg) 173# define Assert(cond, msg)
161# define Trace(x) 174# define Trace(x)
162# define Tracev(x) 175# define Tracev(x)
163# define Tracevv(x) 176# define Tracevv(x)
164# define Tracec(c,x) 177# define Tracec(c, x)
165# define Tracecv(c,x) 178# define Tracecv(c, x)
166#endif 179#endif
167 180
168static int fill_inbuf(void); 181static int fill_inbuf(void);
@@ -170,7 +183,7 @@ static void flush_window(void);
170static void error(char *m); 183static void error(char *m);
171static void gzip_mark(void **); 184static void gzip_mark(void **);
172static void gzip_release(void **); 185static void gzip_release(void **);
173 186
174/* 187/*
175 * This is set up by the setup-routine at boot-time 188 * This is set up by the setup-routine at boot-time
176 */ 189 */
@@ -185,7 +198,7 @@ static unsigned char *real_mode; /* Pointer to real-mode data */
185extern unsigned char input_data[]; 198extern unsigned char input_data[];
186extern int input_len; 199extern int input_len;
187 200
188static long bytes_out = 0; 201static long bytes_out;
189 202
190static void *malloc(int size); 203static void *malloc(int size);
191static void free(void *where); 204static void free(void *where);
@@ -210,7 +223,7 @@ static memptr free_mem_end_ptr;
210#define HEAP_SIZE 0x4000 223#define HEAP_SIZE 0x4000
211#endif 224#endif
212 225
213static char *vidmem = (char *)0xb8000; 226static char *vidmem;
214static int vidport; 227static int vidport;
215static int lines, cols; 228static int lines, cols;
216 229
@@ -224,8 +237,10 @@ static void *malloc(int size)
224{ 237{
225 void *p; 238 void *p;
226 239
227 if (size <0) error("Malloc error"); 240 if (size < 0)
228 if (free_mem_ptr <= 0) error("Memory error"); 241 error("Malloc error");
242 if (free_mem_ptr <= 0)
243 error("Memory error");
229 244
230 free_mem_ptr = (free_mem_ptr + 3) & ~3; /* Align */ 245 free_mem_ptr = (free_mem_ptr + 3) & ~3; /* Align */
231 246
@@ -251,19 +266,19 @@ static void gzip_release(void **ptr)
251{ 266{
252 free_mem_ptr = (memptr) *ptr; 267 free_mem_ptr = (memptr) *ptr;
253} 268}
254 269
255static void scroll(void) 270static void scroll(void)
256{ 271{
257 int i; 272 int i;
258 273
259 memcpy ( vidmem, vidmem + cols * 2, ( lines - 1 ) * cols * 2 ); 274 memcpy(vidmem, vidmem + cols * 2, (lines - 1) * cols * 2);
260 for ( i = ( lines - 1 ) * cols * 2; i < lines * cols * 2; i += 2 ) 275 for (i = (lines - 1) * cols * 2; i < lines * cols * 2; i += 2)
261 vidmem[i] = ' '; 276 vidmem[i] = ' ';
262} 277}
263 278
264static void putstr(const char *s) 279static void putstr(const char *s)
265{ 280{
266 int x,y,pos; 281 int x, y, pos;
267 char c; 282 char c;
268 283
269#ifdef CONFIG_X86_32 284#ifdef CONFIG_X86_32
@@ -274,18 +289,18 @@ static void putstr(const char *s)
274 x = RM_SCREEN_INFO.orig_x; 289 x = RM_SCREEN_INFO.orig_x;
275 y = RM_SCREEN_INFO.orig_y; 290 y = RM_SCREEN_INFO.orig_y;
276 291
277 while ( ( c = *s++ ) != '\0' ) { 292 while ((c = *s++) != '\0') {
278 if ( c == '\n' ) { 293 if (c == '\n') {
279 x = 0; 294 x = 0;
280 if ( ++y >= lines ) { 295 if (++y >= lines) {
281 scroll(); 296 scroll();
282 y--; 297 y--;
283 } 298 }
284 } else { 299 } else {
285 vidmem [(x + cols * y) * 2] = c; 300 vidmem [(x + cols * y) * 2] = c;
286 if ( ++x >= cols ) { 301 if (++x >= cols) {
287 x = 0; 302 x = 0;
288 if ( ++y >= lines ) { 303 if (++y >= lines) {
289 scroll(); 304 scroll();
290 y--; 305 y--;
291 } 306 }
@@ -303,22 +318,22 @@ static void putstr(const char *s)
303 outb(0xff & (pos >> 1), vidport+1); 318 outb(0xff & (pos >> 1), vidport+1);
304} 319}
305 320
306static void* memset(void* s, int c, unsigned n) 321static void *memset(void *s, int c, unsigned n)
307{ 322{
308 int i; 323 int i;
309 char *ss = s; 324 char *ss = s;
310 325
311 for (i=0;i<n;i++) ss[i] = c; 326 for (i = 0; i < n; i++) ss[i] = c;
312 return s; 327 return s;
313} 328}
314 329
315static void* memcpy(void* dest, const void* src, unsigned n) 330static void *memcpy(void *dest, const void *src, unsigned n)
316{ 331{
317 int i; 332 int i;
318 const char *s = src; 333 const char *s = src;
319 char *d = dest; 334 char *d = dest;
320 335
321 for (i=0;i<n;i++) d[i] = s[i]; 336 for (i = 0; i < n; i++) d[i] = s[i];
322 return dest; 337 return dest;
323} 338}
324 339
@@ -341,9 +356,9 @@ static void flush_window(void)
341 /* With my window equal to my output buffer 356 /* With my window equal to my output buffer
342 * I only need to compute the crc here. 357 * I only need to compute the crc here.
343 */ 358 */
344 ulg c = crc; /* temporary variable */ 359 unsigned long c = crc; /* temporary variable */
345 unsigned n; 360 unsigned n;
346 uch *in, ch; 361 unsigned char *in, ch;
347 362
348 in = window; 363 in = window;
349 for (n = 0; n < outcnt; n++) { 364 for (n = 0; n < outcnt; n++) {
@@ -351,7 +366,7 @@ static void flush_window(void)
351 c = crc_32_tab[((int)c ^ ch) & 0xff] ^ (c >> 8); 366 c = crc_32_tab[((int)c ^ ch) & 0xff] ^ (c >> 8);
352 } 367 }
353 crc = c; 368 crc = c;
354 bytes_out += (ulg)outcnt; 369 bytes_out += (unsigned long)outcnt;
355 outcnt = 0; 370 outcnt = 0;
356} 371}
357 372
@@ -365,9 +380,59 @@ static void error(char *x)
365 asm("hlt"); 380 asm("hlt");
366} 381}
367 382
383static void parse_elf(void *output)
384{
385#ifdef CONFIG_X86_64
386 Elf64_Ehdr ehdr;
387 Elf64_Phdr *phdrs, *phdr;
388#else
389 Elf32_Ehdr ehdr;
390 Elf32_Phdr *phdrs, *phdr;
391#endif
392 void *dest;
393 int i;
394
395 memcpy(&ehdr, output, sizeof(ehdr));
396 if (ehdr.e_ident[EI_MAG0] != ELFMAG0 ||
397 ehdr.e_ident[EI_MAG1] != ELFMAG1 ||
398 ehdr.e_ident[EI_MAG2] != ELFMAG2 ||
399 ehdr.e_ident[EI_MAG3] != ELFMAG3) {
400 error("Kernel is not a valid ELF file");
401 return;
402 }
403
404 putstr("Parsing ELF... ");
405
406 phdrs = malloc(sizeof(*phdrs) * ehdr.e_phnum);
407 if (!phdrs)
408 error("Failed to allocate space for phdrs");
409
410 memcpy(phdrs, output + ehdr.e_phoff, sizeof(*phdrs) * ehdr.e_phnum);
411
412 for (i = 0; i < ehdr.e_phnum; i++) {
413 phdr = &phdrs[i];
414
415 switch (phdr->p_type) {
416 case PT_LOAD:
417#ifdef CONFIG_RELOCATABLE
418 dest = output;
419 dest += (phdr->p_paddr - LOAD_PHYSICAL_ADDR);
420#else
421 dest = (void *)(phdr->p_paddr);
422#endif
423 memcpy(dest,
424 output + phdr->p_offset,
425 phdr->p_filesz);
426 break;
427 default: /* Ignore other PT_* */ break;
428 }
429 }
430}
431
368asmlinkage void decompress_kernel(void *rmode, memptr heap, 432asmlinkage void decompress_kernel(void *rmode, memptr heap,
369 uch *input_data, unsigned long input_len, 433 unsigned char *input_data,
370 uch *output) 434 unsigned long input_len,
435 unsigned char *output)
371{ 436{
372 real_mode = rmode; 437 real_mode = rmode;
373 438
@@ -390,12 +455,12 @@ asmlinkage void decompress_kernel(void *rmode, memptr heap,
390 inptr = 0; 455 inptr = 0;
391 456
392#ifdef CONFIG_X86_64 457#ifdef CONFIG_X86_64
393 if ((ulg)output & (__KERNEL_ALIGN - 1)) 458 if ((unsigned long)output & (__KERNEL_ALIGN - 1))
394 error("Destination address not 2M aligned"); 459 error("Destination address not 2M aligned");
395 if ((ulg)output >= 0xffffffffffUL) 460 if ((unsigned long)output >= 0xffffffffffUL)
396 error("Destination address too large"); 461 error("Destination address too large");
397#else 462#else
398 if ((u32)output & (CONFIG_PHYSICAL_ALIGN -1)) 463 if ((u32)output & (CONFIG_PHYSICAL_ALIGN - 1))
399 error("Destination address not CONFIG_PHYSICAL_ALIGN aligned"); 464 error("Destination address not CONFIG_PHYSICAL_ALIGN aligned");
400 if (heap > ((-__PAGE_OFFSET-(512<<20)-1) & 0x7fffffff)) 465 if (heap > ((-__PAGE_OFFSET-(512<<20)-1) & 0x7fffffff))
401 error("Destination address too large"); 466 error("Destination address too large");
@@ -408,6 +473,7 @@ asmlinkage void decompress_kernel(void *rmode, memptr heap,
408 makecrc(); 473 makecrc();
409 putstr("\nDecompressing Linux... "); 474 putstr("\nDecompressing Linux... ");
410 gunzip(); 475 gunzip();
476 parse_elf(output);
411 putstr("done.\nBooting the kernel.\n"); 477 putstr("done.\nBooting the kernel.\n");
412 return; 478 return;
413} 479}
diff --git a/arch/x86/boot/cpucheck.c b/arch/x86/boot/cpucheck.c
index 769065bd23d7..2462c88689ed 100644
--- a/arch/x86/boot/cpucheck.c
+++ b/arch/x86/boot/cpucheck.c
@@ -56,27 +56,27 @@ static const u32 req_flags[NCAPINTS] =
56 REQUIRED_MASK7, 56 REQUIRED_MASK7,
57}; 57};
58 58
59#define A32(a,b,c,d) (((d) << 24)+((c) << 16)+((b) << 8)+(a)) 59#define A32(a, b, c, d) (((d) << 24)+((c) << 16)+((b) << 8)+(a))
60 60
61static int is_amd(void) 61static int is_amd(void)
62{ 62{
63 return cpu_vendor[0] == A32('A','u','t','h') && 63 return cpu_vendor[0] == A32('A', 'u', 't', 'h') &&
64 cpu_vendor[1] == A32('e','n','t','i') && 64 cpu_vendor[1] == A32('e', 'n', 't', 'i') &&
65 cpu_vendor[2] == A32('c','A','M','D'); 65 cpu_vendor[2] == A32('c', 'A', 'M', 'D');
66} 66}
67 67
68static int is_centaur(void) 68static int is_centaur(void)
69{ 69{
70 return cpu_vendor[0] == A32('C','e','n','t') && 70 return cpu_vendor[0] == A32('C', 'e', 'n', 't') &&
71 cpu_vendor[1] == A32('a','u','r','H') && 71 cpu_vendor[1] == A32('a', 'u', 'r', 'H') &&
72 cpu_vendor[2] == A32('a','u','l','s'); 72 cpu_vendor[2] == A32('a', 'u', 'l', 's');
73} 73}
74 74
75static int is_transmeta(void) 75static int is_transmeta(void)
76{ 76{
77 return cpu_vendor[0] == A32('G','e','n','u') && 77 return cpu_vendor[0] == A32('G', 'e', 'n', 'u') &&
78 cpu_vendor[1] == A32('i','n','e','T') && 78 cpu_vendor[1] == A32('i', 'n', 'e', 'T') &&
79 cpu_vendor[2] == A32('M','x','8','6'); 79 cpu_vendor[2] == A32('M', 'x', '8', '6');
80} 80}
81 81
82static int has_fpu(void) 82static int has_fpu(void)
diff --git a/arch/x86/boot/header.S b/arch/x86/boot/header.S
index 64ad9016585a..6d2df8d61c54 100644
--- a/arch/x86/boot/header.S
+++ b/arch/x86/boot/header.S
@@ -22,6 +22,7 @@
22#include <asm/page.h> 22#include <asm/page.h>
23#include <asm/setup.h> 23#include <asm/setup.h>
24#include "boot.h" 24#include "boot.h"
25#include "offsets.h"
25 26
26SETUPSECTS = 4 /* default nr of setup-sectors */ 27SETUPSECTS = 4 /* default nr of setup-sectors */
27BOOTSEG = 0x07C0 /* original address of boot-sector */ 28BOOTSEG = 0x07C0 /* original address of boot-sector */
@@ -119,7 +120,7 @@ _start:
119 # Part 2 of the header, from the old setup.S 120 # Part 2 of the header, from the old setup.S
120 121
121 .ascii "HdrS" # header signature 122 .ascii "HdrS" # header signature
122 .word 0x0207 # header version number (>= 0x0105) 123 .word 0x0208 # header version number (>= 0x0105)
123 # or else old loadlin-1.5 will fail) 124 # or else old loadlin-1.5 will fail)
124 .globl realmode_swtch 125 .globl realmode_swtch
125realmode_swtch: .word 0, 0 # default_switch, SETUPSEG 126realmode_swtch: .word 0, 0 # default_switch, SETUPSEG
@@ -223,6 +224,9 @@ hardware_subarch: .long 0 # subarchitecture, added with 2.07
223 224
224hardware_subarch_data: .quad 0 225hardware_subarch_data: .quad 0
225 226
227payload_offset: .long input_data
228payload_length: .long input_data_end-input_data
229
226# End of setup header ##################################################### 230# End of setup header #####################################################
227 231
228 .section ".inittext", "ax" 232 .section ".inittext", "ax"
diff --git a/arch/x86/boot/pm.c b/arch/x86/boot/pm.c
index 1a0f936c160b..a93cb8bded4d 100644
--- a/arch/x86/boot/pm.c
+++ b/arch/x86/boot/pm.c
@@ -100,7 +100,7 @@ static void reset_coprocessor(void)
100/* 100/*
101 * Set up the GDT 101 * Set up the GDT
102 */ 102 */
103#define GDT_ENTRY(flags,base,limit) \ 103#define GDT_ENTRY(flags, base, limit) \
104 (((u64)(base & 0xff000000) << 32) | \ 104 (((u64)(base & 0xff000000) << 32) | \
105 ((u64)flags << 40) | \ 105 ((u64)flags << 40) | \
106 ((u64)(limit & 0x00ff0000) << 32) | \ 106 ((u64)(limit & 0x00ff0000) << 32) | \
diff --git a/arch/x86/boot/tools/build.c b/arch/x86/boot/tools/build.c
index b4248740ff0d..44dc1923c0e3 100644
--- a/arch/x86/boot/tools/build.c
+++ b/arch/x86/boot/tools/build.c
@@ -50,6 +50,75 @@ typedef unsigned long u32;
50u8 buf[SETUP_SECT_MAX*512]; 50u8 buf[SETUP_SECT_MAX*512];
51int is_big_kernel; 51int is_big_kernel;
52 52
53/*----------------------------------------------------------------------*/
54
55static const u32 crctab32[] = {
56 0x00000000, 0x77073096, 0xee0e612c, 0x990951ba, 0x076dc419,
57 0x706af48f, 0xe963a535, 0x9e6495a3, 0x0edb8832, 0x79dcb8a4,
58 0xe0d5e91e, 0x97d2d988, 0x09b64c2b, 0x7eb17cbd, 0xe7b82d07,
59 0x90bf1d91, 0x1db71064, 0x6ab020f2, 0xf3b97148, 0x84be41de,
60 0x1adad47d, 0x6ddde4eb, 0xf4d4b551, 0x83d385c7, 0x136c9856,
61 0x646ba8c0, 0xfd62f97a, 0x8a65c9ec, 0x14015c4f, 0x63066cd9,
62 0xfa0f3d63, 0x8d080df5, 0x3b6e20c8, 0x4c69105e, 0xd56041e4,
63 0xa2677172, 0x3c03e4d1, 0x4b04d447, 0xd20d85fd, 0xa50ab56b,
64 0x35b5a8fa, 0x42b2986c, 0xdbbbc9d6, 0xacbcf940, 0x32d86ce3,
65 0x45df5c75, 0xdcd60dcf, 0xabd13d59, 0x26d930ac, 0x51de003a,
66 0xc8d75180, 0xbfd06116, 0x21b4f4b5, 0x56b3c423, 0xcfba9599,
67 0xb8bda50f, 0x2802b89e, 0x5f058808, 0xc60cd9b2, 0xb10be924,
68 0x2f6f7c87, 0x58684c11, 0xc1611dab, 0xb6662d3d, 0x76dc4190,
69 0x01db7106, 0x98d220bc, 0xefd5102a, 0x71b18589, 0x06b6b51f,
70 0x9fbfe4a5, 0xe8b8d433, 0x7807c9a2, 0x0f00f934, 0x9609a88e,
71 0xe10e9818, 0x7f6a0dbb, 0x086d3d2d, 0x91646c97, 0xe6635c01,
72 0x6b6b51f4, 0x1c6c6162, 0x856530d8, 0xf262004e, 0x6c0695ed,
73 0x1b01a57b, 0x8208f4c1, 0xf50fc457, 0x65b0d9c6, 0x12b7e950,
74 0x8bbeb8ea, 0xfcb9887c, 0x62dd1ddf, 0x15da2d49, 0x8cd37cf3,
75 0xfbd44c65, 0x4db26158, 0x3ab551ce, 0xa3bc0074, 0xd4bb30e2,
76 0x4adfa541, 0x3dd895d7, 0xa4d1c46d, 0xd3d6f4fb, 0x4369e96a,
77 0x346ed9fc, 0xad678846, 0xda60b8d0, 0x44042d73, 0x33031de5,
78 0xaa0a4c5f, 0xdd0d7cc9, 0x5005713c, 0x270241aa, 0xbe0b1010,
79 0xc90c2086, 0x5768b525, 0x206f85b3, 0xb966d409, 0xce61e49f,
80 0x5edef90e, 0x29d9c998, 0xb0d09822, 0xc7d7a8b4, 0x59b33d17,
81 0x2eb40d81, 0xb7bd5c3b, 0xc0ba6cad, 0xedb88320, 0x9abfb3b6,
82 0x03b6e20c, 0x74b1d29a, 0xead54739, 0x9dd277af, 0x04db2615,
83 0x73dc1683, 0xe3630b12, 0x94643b84, 0x0d6d6a3e, 0x7a6a5aa8,
84 0xe40ecf0b, 0x9309ff9d, 0x0a00ae27, 0x7d079eb1, 0xf00f9344,
85 0x8708a3d2, 0x1e01f268, 0x6906c2fe, 0xf762575d, 0x806567cb,
86 0x196c3671, 0x6e6b06e7, 0xfed41b76, 0x89d32be0, 0x10da7a5a,
87 0x67dd4acc, 0xf9b9df6f, 0x8ebeeff9, 0x17b7be43, 0x60b08ed5,
88 0xd6d6a3e8, 0xa1d1937e, 0x38d8c2c4, 0x4fdff252, 0xd1bb67f1,
89 0xa6bc5767, 0x3fb506dd, 0x48b2364b, 0xd80d2bda, 0xaf0a1b4c,
90 0x36034af6, 0x41047a60, 0xdf60efc3, 0xa867df55, 0x316e8eef,
91 0x4669be79, 0xcb61b38c, 0xbc66831a, 0x256fd2a0, 0x5268e236,
92 0xcc0c7795, 0xbb0b4703, 0x220216b9, 0x5505262f, 0xc5ba3bbe,
93 0xb2bd0b28, 0x2bb45a92, 0x5cb36a04, 0xc2d7ffa7, 0xb5d0cf31,
94 0x2cd99e8b, 0x5bdeae1d, 0x9b64c2b0, 0xec63f226, 0x756aa39c,
95 0x026d930a, 0x9c0906a9, 0xeb0e363f, 0x72076785, 0x05005713,
96 0x95bf4a82, 0xe2b87a14, 0x7bb12bae, 0x0cb61b38, 0x92d28e9b,
97 0xe5d5be0d, 0x7cdcefb7, 0x0bdbdf21, 0x86d3d2d4, 0xf1d4e242,
98 0x68ddb3f8, 0x1fda836e, 0x81be16cd, 0xf6b9265b, 0x6fb077e1,
99 0x18b74777, 0x88085ae6, 0xff0f6a70, 0x66063bca, 0x11010b5c,
100 0x8f659eff, 0xf862ae69, 0x616bffd3, 0x166ccf45, 0xa00ae278,
101 0xd70dd2ee, 0x4e048354, 0x3903b3c2, 0xa7672661, 0xd06016f7,
102 0x4969474d, 0x3e6e77db, 0xaed16a4a, 0xd9d65adc, 0x40df0b66,
103 0x37d83bf0, 0xa9bcae53, 0xdebb9ec5, 0x47b2cf7f, 0x30b5ffe9,
104 0xbdbdf21c, 0xcabac28a, 0x53b39330, 0x24b4a3a6, 0xbad03605,
105 0xcdd70693, 0x54de5729, 0x23d967bf, 0xb3667a2e, 0xc4614ab8,
106 0x5d681b02, 0x2a6f2b94, 0xb40bbe37, 0xc30c8ea1, 0x5a05df1b,
107 0x2d02ef8d
108};
109
110static u32 partial_crc32_one(u8 c, u32 crc)
111{
112 return crctab32[(crc ^ c) & 0xff] ^ (crc >> 8);
113}
114
115static u32 partial_crc32(const u8 *s, int len, u32 crc)
116{
117 while (len--)
118 crc = partial_crc32_one(*s++, crc);
119 return crc;
120}
121
53static void die(const char * str, ...) 122static void die(const char * str, ...)
54{ 123{
55 va_list args; 124 va_list args;
@@ -74,6 +143,7 @@ int main(int argc, char ** argv)
74 FILE *file; 143 FILE *file;
75 int fd; 144 int fd;
76 void *kernel; 145 void *kernel;
146 u32 crc = 0xffffffffUL;
77 147
78 if (argc > 2 && !strcmp(argv[1], "-b")) 148 if (argc > 2 && !strcmp(argv[1], "-b"))
79 { 149 {
@@ -144,7 +214,8 @@ int main(int argc, char ** argv)
144 kernel = mmap(NULL, sz, PROT_READ, MAP_SHARED, fd, 0); 214 kernel = mmap(NULL, sz, PROT_READ, MAP_SHARED, fd, 0);
145 if (kernel == MAP_FAILED) 215 if (kernel == MAP_FAILED)
146 die("Unable to mmap '%s': %m", argv[2]); 216 die("Unable to mmap '%s': %m", argv[2]);
147 sys_size = (sz + 15) / 16; 217 /* Number of 16-byte paragraphs, including space for a 4-byte CRC */
218 sys_size = (sz + 15 + 4) / 16;
148 if (!is_big_kernel && sys_size > DEF_SYSSIZE) 219 if (!is_big_kernel && sys_size > DEF_SYSSIZE)
149 die("System is too big. Try using bzImage or modules."); 220 die("System is too big. Try using bzImage or modules.");
150 221
@@ -155,12 +226,27 @@ int main(int argc, char ** argv)
155 buf[0x1f6] = sys_size >> 16; 226 buf[0x1f6] = sys_size >> 16;
156 buf[0x1f7] = sys_size >> 24; 227 buf[0x1f7] = sys_size >> 24;
157 228
229 crc = partial_crc32(buf, i, crc);
158 if (fwrite(buf, 1, i, stdout) != i) 230 if (fwrite(buf, 1, i, stdout) != i)
159 die("Writing setup failed"); 231 die("Writing setup failed");
160 232
161 /* Copy the kernel code */ 233 /* Copy the kernel code */
234 crc = partial_crc32(kernel, sz, crc);
162 if (fwrite(kernel, 1, sz, stdout) != sz) 235 if (fwrite(kernel, 1, sz, stdout) != sz)
163 die("Writing kernel failed"); 236 die("Writing kernel failed");
237
238 /* Add padding leaving 4 bytes for the checksum */
239 while (sz++ < (sys_size*16) - 4) {
240 crc = partial_crc32_one('\0', crc);
241 if (fwrite("\0", 1, 1, stdout) != 1)
242 die("Writing padding failed");
243 }
244
245 /* Write the CRC */
246 fprintf(stderr, "CRC %lx\n", crc);
247 if (fwrite(&crc, 1, 4, stdout) != 4)
248 die("Writing CRC failed");
249
164 close(fd); 250 close(fd);
165 251
166 /* Everything is OK */ 252 /* Everything is OK */
diff --git a/arch/x86/boot/video-bios.c b/arch/x86/boot/video-bios.c
index ff664a117096..39e247e96172 100644
--- a/arch/x86/boot/video-bios.c
+++ b/arch/x86/boot/video-bios.c
@@ -50,6 +50,7 @@ static int set_bios_mode(u8 mode)
50 if (new_mode == mode) 50 if (new_mode == mode)
51 return 0; /* Mode change OK */ 51 return 0; /* Mode change OK */
52 52
53#ifndef _WAKEUP
53 if (new_mode != boot_params.screen_info.orig_video_mode) { 54 if (new_mode != boot_params.screen_info.orig_video_mode) {
54 /* Mode setting failed, but we didn't end up where we 55 /* Mode setting failed, but we didn't end up where we
55 started. That's bad. Try to revert to the original 56 started. That's bad. Try to revert to the original
@@ -59,13 +60,18 @@ static int set_bios_mode(u8 mode)
59 : "+a" (ax) 60 : "+a" (ax)
60 : : "ebx", "ecx", "edx", "esi", "edi"); 61 : : "ebx", "ecx", "edx", "esi", "edi");
61 } 62 }
63#endif
62 return -1; 64 return -1;
63} 65}
64 66
65static int bios_probe(void) 67static int bios_probe(void)
66{ 68{
67 u8 mode; 69 u8 mode;
70#ifdef _WAKEUP
71 u8 saved_mode = 0x03;
72#else
68 u8 saved_mode = boot_params.screen_info.orig_video_mode; 73 u8 saved_mode = boot_params.screen_info.orig_video_mode;
74#endif
69 u16 crtc; 75 u16 crtc;
70 struct mode_info *mi; 76 struct mode_info *mi;
71 int nmodes = 0; 77 int nmodes = 0;
diff --git a/arch/x86/boot/video-mode.c b/arch/x86/boot/video-mode.c
new file mode 100644
index 000000000000..748e8d06290a
--- /dev/null
+++ b/arch/x86/boot/video-mode.c
@@ -0,0 +1,173 @@
1/* -*- linux-c -*- ------------------------------------------------------- *
2 *
3 * Copyright (C) 1991, 1992 Linus Torvalds
4 * Copyright 2007-2008 rPath, Inc. - All Rights Reserved
5 *
6 * This file is part of the Linux kernel, and is made available under
7 * the terms of the GNU General Public License version 2.
8 *
9 * ----------------------------------------------------------------------- */
10
11/*
12 * arch/i386/boot/video-mode.c
13 *
14 * Set the video mode. This is separated out into a different
15 * file in order to be shared with the ACPI wakeup code.
16 */
17
18#include "boot.h"
19#include "video.h"
20#include "vesa.h"
21
22/*
23 * Common variables
24 */
25int adapter; /* 0=CGA/MDA/HGC, 1=EGA, 2=VGA+ */
26u16 video_segment;
27int force_x, force_y; /* Don't query the BIOS for cols/rows */
28
29int do_restore; /* Screen contents changed during mode flip */
30int graphic_mode; /* Graphic mode with linear frame buffer */
31
32/* Probe the video drivers and have them generate their mode lists. */
33void probe_cards(int unsafe)
34{
35 struct card_info *card;
36 static u8 probed[2];
37
38 if (probed[unsafe])
39 return;
40
41 probed[unsafe] = 1;
42
43 for (card = video_cards; card < video_cards_end; card++) {
44 if (card->unsafe == unsafe) {
45 if (card->probe)
46 card->nmodes = card->probe();
47 else
48 card->nmodes = 0;
49 }
50 }
51}
52
53/* Test if a mode is defined */
54int mode_defined(u16 mode)
55{
56 struct card_info *card;
57 struct mode_info *mi;
58 int i;
59
60 for (card = video_cards; card < video_cards_end; card++) {
61 mi = card->modes;
62 for (i = 0; i < card->nmodes; i++, mi++) {
63 if (mi->mode == mode)
64 return 1;
65 }
66 }
67
68 return 0;
69}
70
71/* Set mode (without recalc) */
72static int raw_set_mode(u16 mode, u16 *real_mode)
73{
74 int nmode, i;
75 struct card_info *card;
76 struct mode_info *mi;
77
78 /* Drop the recalc bit if set */
79 mode &= ~VIDEO_RECALC;
80
81 /* Scan for mode based on fixed ID, position, or resolution */
82 nmode = 0;
83 for (card = video_cards; card < video_cards_end; card++) {
84 mi = card->modes;
85 for (i = 0; i < card->nmodes; i++, mi++) {
86 int visible = mi->x || mi->y;
87
88 if ((mode == nmode && visible) ||
89 mode == mi->mode ||
90 mode == (mi->y << 8)+mi->x) {
91 *real_mode = mi->mode;
92 return card->set_mode(mi);
93 }
94
95 if (visible)
96 nmode++;
97 }
98 }
99
100 /* Nothing found? Is it an "exceptional" (unprobed) mode? */
101 for (card = video_cards; card < video_cards_end; card++) {
102 if (mode >= card->xmode_first &&
103 mode < card->xmode_first+card->xmode_n) {
104 struct mode_info mix;
105 *real_mode = mix.mode = mode;
106 mix.x = mix.y = 0;
107 return card->set_mode(&mix);
108 }
109 }
110
111 /* Otherwise, failure... */
112 return -1;
113}
114
115/*
116 * Recalculate the vertical video cutoff (hack!)
117 */
118static void vga_recalc_vertical(void)
119{
120 unsigned int font_size, rows;
121 u16 crtc;
122 u8 pt, ov;
123
124 set_fs(0);
125 font_size = rdfs8(0x485); /* BIOS: font size (pixels) */
126 rows = force_y ? force_y : rdfs8(0x484)+1; /* Text rows */
127
128 rows *= font_size; /* Visible scan lines */
129 rows--; /* ... minus one */
130
131 crtc = vga_crtc();
132
133 pt = in_idx(crtc, 0x11);
134 pt &= ~0x80; /* Unlock CR0-7 */
135 out_idx(pt, crtc, 0x11);
136
137 out_idx((u8)rows, crtc, 0x12); /* Lower height register */
138
139 ov = in_idx(crtc, 0x07); /* Overflow register */
140 ov &= 0xbd;
141 ov |= (rows >> (8-1)) & 0x02;
142 ov |= (rows >> (9-6)) & 0x40;
143 out_idx(ov, crtc, 0x07);
144}
145
146/* Set mode (with recalc if specified) */
147int set_mode(u16 mode)
148{
149 int rv;
150 u16 real_mode;
151
152 /* Very special mode numbers... */
153 if (mode == VIDEO_CURRENT_MODE)
154 return 0; /* Nothing to do... */
155 else if (mode == NORMAL_VGA)
156 mode = VIDEO_80x25;
157 else if (mode == EXTENDED_VGA)
158 mode = VIDEO_8POINT;
159
160 rv = raw_set_mode(mode, &real_mode);
161 if (rv)
162 return rv;
163
164 if (mode & VIDEO_RECALC)
165 vga_recalc_vertical();
166
167 /* Save the canonical mode number for the kernel, not
168 an alias, size specification or menu position */
169#ifndef _WAKEUP
170 boot_params.hdr.vid_mode = real_mode;
171#endif
172 return 0;
173}
diff --git a/arch/x86/boot/video-vesa.c b/arch/x86/boot/video-vesa.c
index 419b5c273374..5d5a3f6e8b5c 100644
--- a/arch/x86/boot/video-vesa.c
+++ b/arch/x86/boot/video-vesa.c
@@ -24,7 +24,11 @@ static struct vesa_mode_info vminfo;
24 24
25__videocard video_vesa; 25__videocard video_vesa;
26 26
27#ifndef _WAKEUP
27static void vesa_store_mode_params_graphics(void); 28static void vesa_store_mode_params_graphics(void);
29#else /* _WAKEUP */
30static inline void vesa_store_mode_params_graphics(void) {}
31#endif /* _WAKEUP */
28 32
29static int vesa_probe(void) 33static int vesa_probe(void)
30{ 34{
@@ -165,6 +169,8 @@ static int vesa_set_mode(struct mode_info *mode)
165} 169}
166 170
167 171
172#ifndef _WAKEUP
173
168/* Switch DAC to 8-bit mode */ 174/* Switch DAC to 8-bit mode */
169static void vesa_dac_set_8bits(void) 175static void vesa_dac_set_8bits(void)
170{ 176{
@@ -288,6 +294,8 @@ void vesa_store_edid(void)
288#endif /* CONFIG_FIRMWARE_EDID */ 294#endif /* CONFIG_FIRMWARE_EDID */
289} 295}
290 296
297#endif /* not _WAKEUP */
298
291__videocard video_vesa = 299__videocard video_vesa =
292{ 300{
293 .card_name = "VESA", 301 .card_name = "VESA",
diff --git a/arch/x86/boot/video-vga.c b/arch/x86/boot/video-vga.c
index 7259387b7d19..330d6589a2ad 100644
--- a/arch/x86/boot/video-vga.c
+++ b/arch/x86/boot/video-vga.c
@@ -210,6 +210,8 @@ static int vga_set_mode(struct mode_info *mode)
210 */ 210 */
211static int vga_probe(void) 211static int vga_probe(void)
212{ 212{
213 u16 ega_bx;
214
213 static const char *card_name[] = { 215 static const char *card_name[] = {
214 "CGA/MDA/HGC", "EGA", "VGA" 216 "CGA/MDA/HGC", "EGA", "VGA"
215 }; 217 };
@@ -226,12 +228,16 @@ static int vga_probe(void)
226 u8 vga_flag; 228 u8 vga_flag;
227 229
228 asm(INT10 230 asm(INT10
229 : "=b" (boot_params.screen_info.orig_video_ega_bx) 231 : "=b" (ega_bx)
230 : "a" (0x1200), "b" (0x10) /* Check EGA/VGA */ 232 : "a" (0x1200), "b" (0x10) /* Check EGA/VGA */
231 : "ecx", "edx", "esi", "edi"); 233 : "ecx", "edx", "esi", "edi");
232 234
235#ifndef _WAKEUP
236 boot_params.screen_info.orig_video_ega_bx = ega_bx;
237#endif
238
233 /* If we have MDA/CGA/HGC then BL will be unchanged at 0x10 */ 239 /* If we have MDA/CGA/HGC then BL will be unchanged at 0x10 */
234 if ((u8)boot_params.screen_info.orig_video_ega_bx != 0x10) { 240 if ((u8)ega_bx != 0x10) {
235 /* EGA/VGA */ 241 /* EGA/VGA */
236 asm(INT10 242 asm(INT10
237 : "=a" (vga_flag) 243 : "=a" (vga_flag)
@@ -240,7 +246,9 @@ static int vga_probe(void)
240 246
241 if (vga_flag == 0x1a) { 247 if (vga_flag == 0x1a) {
242 adapter = ADAPTER_VGA; 248 adapter = ADAPTER_VGA;
249#ifndef _WAKEUP
243 boot_params.screen_info.orig_video_isVGA = 1; 250 boot_params.screen_info.orig_video_isVGA = 1;
251#endif
244 } else { 252 } else {
245 adapter = ADAPTER_EGA; 253 adapter = ADAPTER_EGA;
246 } 254 }
diff --git a/arch/x86/boot/video.c b/arch/x86/boot/video.c
index 696d08f3843c..c1c47ba069ef 100644
--- a/arch/x86/boot/video.c
+++ b/arch/x86/boot/video.c
@@ -18,21 +18,6 @@
18#include "video.h" 18#include "video.h"
19#include "vesa.h" 19#include "vesa.h"
20 20
21/*
22 * Mode list variables
23 */
24static struct card_info cards[]; /* List of cards to probe for */
25
26/*
27 * Common variables
28 */
29int adapter; /* 0=CGA/MDA/HGC, 1=EGA, 2=VGA+ */
30u16 video_segment;
31int force_x, force_y; /* Don't query the BIOS for cols/rows */
32
33int do_restore = 0; /* Screen contents changed during mode flip */
34int graphic_mode; /* Graphic mode with linear frame buffer */
35
36static void store_cursor_position(void) 21static void store_cursor_position(void)
37{ 22{
38 u16 curpos; 23 u16 curpos;
@@ -107,147 +92,6 @@ static void store_mode_params(void)
107 boot_params.screen_info.orig_video_lines = y; 92 boot_params.screen_info.orig_video_lines = y;
108} 93}
109 94
110/* Probe the video drivers and have them generate their mode lists. */
111static void probe_cards(int unsafe)
112{
113 struct card_info *card;
114 static u8 probed[2];
115
116 if (probed[unsafe])
117 return;
118
119 probed[unsafe] = 1;
120
121 for (card = video_cards; card < video_cards_end; card++) {
122 if (card->unsafe == unsafe) {
123 if (card->probe)
124 card->nmodes = card->probe();
125 else
126 card->nmodes = 0;
127 }
128 }
129}
130
131/* Test if a mode is defined */
132int mode_defined(u16 mode)
133{
134 struct card_info *card;
135 struct mode_info *mi;
136 int i;
137
138 for (card = video_cards; card < video_cards_end; card++) {
139 mi = card->modes;
140 for (i = 0; i < card->nmodes; i++, mi++) {
141 if (mi->mode == mode)
142 return 1;
143 }
144 }
145
146 return 0;
147}
148
149/* Set mode (without recalc) */
150static int raw_set_mode(u16 mode, u16 *real_mode)
151{
152 int nmode, i;
153 struct card_info *card;
154 struct mode_info *mi;
155
156 /* Drop the recalc bit if set */
157 mode &= ~VIDEO_RECALC;
158
159 /* Scan for mode based on fixed ID, position, or resolution */
160 nmode = 0;
161 for (card = video_cards; card < video_cards_end; card++) {
162 mi = card->modes;
163 for (i = 0; i < card->nmodes; i++, mi++) {
164 int visible = mi->x || mi->y;
165
166 if ((mode == nmode && visible) ||
167 mode == mi->mode ||
168 mode == (mi->y << 8)+mi->x) {
169 *real_mode = mi->mode;
170 return card->set_mode(mi);
171 }
172
173 if (visible)
174 nmode++;
175 }
176 }
177
178 /* Nothing found? Is it an "exceptional" (unprobed) mode? */
179 for (card = video_cards; card < video_cards_end; card++) {
180 if (mode >= card->xmode_first &&
181 mode < card->xmode_first+card->xmode_n) {
182 struct mode_info mix;
183 *real_mode = mix.mode = mode;
184 mix.x = mix.y = 0;
185 return card->set_mode(&mix);
186 }
187 }
188
189 /* Otherwise, failure... */
190 return -1;
191}
192
193/*
194 * Recalculate the vertical video cutoff (hack!)
195 */
196static void vga_recalc_vertical(void)
197{
198 unsigned int font_size, rows;
199 u16 crtc;
200 u8 pt, ov;
201
202 set_fs(0);
203 font_size = rdfs8(0x485); /* BIOS: font size (pixels) */
204 rows = force_y ? force_y : rdfs8(0x484)+1; /* Text rows */
205
206 rows *= font_size; /* Visible scan lines */
207 rows--; /* ... minus one */
208
209 crtc = vga_crtc();
210
211 pt = in_idx(crtc, 0x11);
212 pt &= ~0x80; /* Unlock CR0-7 */
213 out_idx(pt, crtc, 0x11);
214
215 out_idx((u8)rows, crtc, 0x12); /* Lower height register */
216
217 ov = in_idx(crtc, 0x07); /* Overflow register */
218 ov &= 0xbd;
219 ov |= (rows >> (8-1)) & 0x02;
220 ov |= (rows >> (9-6)) & 0x40;
221 out_idx(ov, crtc, 0x07);
222}
223
224/* Set mode (with recalc if specified) */
225static int set_mode(u16 mode)
226{
227 int rv;
228 u16 real_mode;
229
230 /* Very special mode numbers... */
231 if (mode == VIDEO_CURRENT_MODE)
232 return 0; /* Nothing to do... */
233 else if (mode == NORMAL_VGA)
234 mode = VIDEO_80x25;
235 else if (mode == EXTENDED_VGA)
236 mode = VIDEO_8POINT;
237
238 rv = raw_set_mode(mode, &real_mode);
239 if (rv)
240 return rv;
241
242 if (mode & VIDEO_RECALC)
243 vga_recalc_vertical();
244
245 /* Save the canonical mode number for the kernel, not
246 an alias, size specification or menu position */
247 boot_params.hdr.vid_mode = real_mode;
248 return 0;
249}
250
251static unsigned int get_entry(void) 95static unsigned int get_entry(void)
252{ 96{
253 char entry_buf[4]; 97 char entry_buf[4];
@@ -486,6 +330,7 @@ void set_video(void)
486 printf("Undefined video mode number: %x\n", mode); 330 printf("Undefined video mode number: %x\n", mode);
487 mode = ASK_VGA; 331 mode = ASK_VGA;
488 } 332 }
333 boot_params.hdr.vid_mode = mode;
489 vesa_store_edid(); 334 vesa_store_edid();
490 store_mode_params(); 335 store_mode_params();
491 336
diff --git a/arch/x86/ia32/ia32_signal.c b/arch/x86/ia32/ia32_signal.c
index 5e7771a3ba2f..05e155d3fb6c 100644
--- a/arch/x86/ia32/ia32_signal.c
+++ b/arch/x86/ia32/ia32_signal.c
@@ -468,7 +468,7 @@ int ia32_setup_frame(int sig, struct k_sigaction *ka,
468 restorer = ka->sa.sa_restorer; 468 restorer = ka->sa.sa_restorer;
469 } else { 469 } else {
470 /* Return stub is in 32bit vsyscall page */ 470 /* Return stub is in 32bit vsyscall page */
471 if (current->binfmt->hasvdso) 471 if (current->mm->context.vdso)
472 restorer = VDSO32_SYMBOL(current->mm->context.vdso, 472 restorer = VDSO32_SYMBOL(current->mm->context.vdso,
473 sigreturn); 473 sigreturn);
474 else 474 else
diff --git a/arch/x86/ia32/ia32entry.S b/arch/x86/ia32/ia32entry.S
index 8022d3c695c0..ae7158bce4d6 100644
--- a/arch/x86/ia32/ia32entry.S
+++ b/arch/x86/ia32/ia32entry.S
@@ -162,12 +162,14 @@ sysenter_tracesys:
162 SAVE_REST 162 SAVE_REST
163 CLEAR_RREGS 163 CLEAR_RREGS
164 movq %r9,R9(%rsp) 164 movq %r9,R9(%rsp)
165 movq $-ENOSYS,RAX(%rsp) /* really needed? */ 165 movq $-ENOSYS,RAX(%rsp)/* ptrace can change this for a bad syscall */
166 movq %rsp,%rdi /* &pt_regs -> arg1 */ 166 movq %rsp,%rdi /* &pt_regs -> arg1 */
167 call syscall_trace_enter 167 call syscall_trace_enter
168 LOAD_ARGS32 ARGOFFSET /* reload args from stack in case ptrace changed it */ 168 LOAD_ARGS32 ARGOFFSET /* reload args from stack in case ptrace changed it */
169 RESTORE_REST 169 RESTORE_REST
170 xchgl %ebp,%r9d 170 xchgl %ebp,%r9d
171 cmpl $(IA32_NR_syscalls-1),%eax
172 ja int_ret_from_sys_call /* sysenter_tracesys has set RAX(%rsp) */
171 jmp sysenter_do_call 173 jmp sysenter_do_call
172 CFI_ENDPROC 174 CFI_ENDPROC
173ENDPROC(ia32_sysenter_target) 175ENDPROC(ia32_sysenter_target)
@@ -261,13 +263,15 @@ cstar_tracesys:
261 SAVE_REST 263 SAVE_REST
262 CLEAR_RREGS 264 CLEAR_RREGS
263 movq %r9,R9(%rsp) 265 movq %r9,R9(%rsp)
264 movq $-ENOSYS,RAX(%rsp) /* really needed? */ 266 movq $-ENOSYS,RAX(%rsp) /* ptrace can change this for a bad syscall */
265 movq %rsp,%rdi /* &pt_regs -> arg1 */ 267 movq %rsp,%rdi /* &pt_regs -> arg1 */
266 call syscall_trace_enter 268 call syscall_trace_enter
267 LOAD_ARGS32 ARGOFFSET /* reload args from stack in case ptrace changed it */ 269 LOAD_ARGS32 ARGOFFSET /* reload args from stack in case ptrace changed it */
268 RESTORE_REST 270 RESTORE_REST
269 xchgl %ebp,%r9d 271 xchgl %ebp,%r9d
270 movl RSP-ARGOFFSET(%rsp), %r8d 272 movl RSP-ARGOFFSET(%rsp), %r8d
273 cmpl $(IA32_NR_syscalls-1),%eax
274 ja int_ret_from_sys_call /* cstar_tracesys has set RAX(%rsp) */
271 jmp cstar_do_call 275 jmp cstar_do_call
272END(ia32_cstar_target) 276END(ia32_cstar_target)
273 277
@@ -325,7 +329,7 @@ ENTRY(ia32_syscall)
325 jnz ia32_tracesys 329 jnz ia32_tracesys
326ia32_do_syscall: 330ia32_do_syscall:
327 cmpl $(IA32_NR_syscalls-1),%eax 331 cmpl $(IA32_NR_syscalls-1),%eax
328 ja ia32_badsys 332 ja int_ret_from_sys_call /* ia32_tracesys has set RAX(%rsp) */
329 IA32_ARG_FIXUP 333 IA32_ARG_FIXUP
330 call *ia32_sys_call_table(,%rax,8) # xxx: rip relative 334 call *ia32_sys_call_table(,%rax,8) # xxx: rip relative
331ia32_sysret: 335ia32_sysret:
@@ -335,7 +339,7 @@ ia32_sysret:
335ia32_tracesys: 339ia32_tracesys:
336 SAVE_REST 340 SAVE_REST
337 CLEAR_RREGS 341 CLEAR_RREGS
338 movq $-ENOSYS,RAX(%rsp) /* really needed? */ 342 movq $-ENOSYS,RAX(%rsp) /* ptrace can change this for a bad syscall */
339 movq %rsp,%rdi /* &pt_regs -> arg1 */ 343 movq %rsp,%rdi /* &pt_regs -> arg1 */
340 call syscall_trace_enter 344 call syscall_trace_enter
341 LOAD_ARGS32 ARGOFFSET /* reload args from stack in case ptrace changed it */ 345 LOAD_ARGS32 ARGOFFSET /* reload args from stack in case ptrace changed it */
diff --git a/arch/x86/ia32/sys_ia32.c b/arch/x86/ia32/sys_ia32.c
index abf71d26fc2a..7cede7a9e0dc 100644
--- a/arch/x86/ia32/sys_ia32.c
+++ b/arch/x86/ia32/sys_ia32.c
@@ -26,51 +26,27 @@
26#include <linux/file.h> 26#include <linux/file.h>
27#include <linux/signal.h> 27#include <linux/signal.h>
28#include <linux/syscalls.h> 28#include <linux/syscalls.h>
29#include <linux/resource.h>
30#include <linux/times.h> 29#include <linux/times.h>
31#include <linux/utsname.h> 30#include <linux/utsname.h>
32#include <linux/smp.h>
33#include <linux/smp_lock.h> 31#include <linux/smp_lock.h>
34#include <linux/sem.h>
35#include <linux/msg.h>
36#include <linux/mm.h> 32#include <linux/mm.h>
37#include <linux/shm.h>
38#include <linux/slab.h>
39#include <linux/uio.h> 33#include <linux/uio.h>
40#include <linux/nfs_fs.h>
41#include <linux/quota.h>
42#include <linux/module.h>
43#include <linux/sunrpc/svc.h>
44#include <linux/nfsd/nfsd.h>
45#include <linux/nfsd/cache.h>
46#include <linux/nfsd/xdr.h>
47#include <linux/nfsd/syscall.h>
48#include <linux/poll.h> 34#include <linux/poll.h>
49#include <linux/personality.h> 35#include <linux/personality.h>
50#include <linux/stat.h> 36#include <linux/stat.h>
51#include <linux/ipc.h>
52#include <linux/rwsem.h> 37#include <linux/rwsem.h>
53#include <linux/binfmts.h>
54#include <linux/init.h>
55#include <linux/aio_abi.h>
56#include <linux/aio.h>
57#include <linux/compat.h> 38#include <linux/compat.h>
58#include <linux/vfs.h> 39#include <linux/vfs.h>
59#include <linux/ptrace.h> 40#include <linux/ptrace.h>
60#include <linux/highuid.h> 41#include <linux/highuid.h>
61#include <linux/vmalloc.h>
62#include <linux/fsnotify.h>
63#include <linux/sysctl.h> 42#include <linux/sysctl.h>
64#include <asm/mman.h> 43#include <asm/mman.h>
65#include <asm/types.h> 44#include <asm/types.h>
66#include <asm/uaccess.h> 45#include <asm/uaccess.h>
67#include <asm/semaphore.h> 46#include <asm/semaphore.h>
68#include <asm/atomic.h> 47#include <asm/atomic.h>
69#include <asm/ldt.h>
70
71#include <net/scm.h>
72#include <net/sock.h>
73#include <asm/ia32.h> 48#include <asm/ia32.h>
49#include <asm/vgtod.h>
74 50
75#define AA(__x) ((unsigned long)(__x)) 51#define AA(__x) ((unsigned long)(__x))
76 52
@@ -804,11 +780,6 @@ asmlinkage long sys32_execve(char __user *name, compat_uptr_t __user *argv,
804 if (IS_ERR(filename)) 780 if (IS_ERR(filename))
805 return error; 781 return error;
806 error = compat_do_execve(filename, argv, envp, regs); 782 error = compat_do_execve(filename, argv, envp, regs);
807 if (error == 0) {
808 task_lock(current);
809 current->ptrace &= ~PT_DTRACE;
810 task_unlock(current);
811 }
812 putname(filename); 783 putname(filename);
813 return error; 784 return error;
814} 785}
diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile
index 4a4260c7f672..c3920ea8ac56 100644
--- a/arch/x86/kernel/Makefile
+++ b/arch/x86/kernel/Makefile
@@ -2,8 +2,7 @@
2# Makefile for the linux kernel. 2# Makefile for the linux kernel.
3# 3#
4 4
5extra-y := head_$(BITS).o init_task.o vmlinux.lds 5extra-y := head_$(BITS).o head$(BITS).o init_task.o vmlinux.lds
6extra-$(CONFIG_X86_64) += head64.o
7 6
8CPPFLAGS_vmlinux.lds += -U$(UTS_MACHINE) 7CPPFLAGS_vmlinux.lds += -U$(UTS_MACHINE)
9 8
@@ -19,7 +18,7 @@ CFLAGS_tsc_64.o := $(nostackp)
19obj-y := process_$(BITS).o signal_$(BITS).o entry_$(BITS).o 18obj-y := process_$(BITS).o signal_$(BITS).o entry_$(BITS).o
20obj-y += traps_$(BITS).o irq_$(BITS).o 19obj-y += traps_$(BITS).o irq_$(BITS).o
21obj-y += time_$(BITS).o ioport.o ldt.o 20obj-y += time_$(BITS).o ioport.o ldt.o
22obj-y += setup_$(BITS).o i8259_$(BITS).o 21obj-y += setup_$(BITS).o i8259_$(BITS).o setup.o
23obj-$(CONFIG_X86_32) += sys_i386_32.o i386_ksyms_32.o 22obj-$(CONFIG_X86_32) += sys_i386_32.o i386_ksyms_32.o
24obj-$(CONFIG_X86_64) += sys_x86_64.o x8664_ksyms_64.o 23obj-$(CONFIG_X86_64) += sys_x86_64.o x8664_ksyms_64.o
25obj-$(CONFIG_X86_64) += syscall_64.o vsyscall_64.o setup64.o 24obj-$(CONFIG_X86_64) += syscall_64.o vsyscall_64.o setup64.o
@@ -29,6 +28,7 @@ obj-y += alternative.o i8253.o
29obj-$(CONFIG_X86_64) += pci-nommu_64.o bugs_64.o 28obj-$(CONFIG_X86_64) += pci-nommu_64.o bugs_64.o
30obj-y += tsc_$(BITS).o io_delay.o rtc.o 29obj-y += tsc_$(BITS).o io_delay.o rtc.o
31 30
31obj-$(CONFIG_X86_TRAMPOLINE) += trampoline.o
32obj-y += i387.o 32obj-y += i387.o
33obj-y += ptrace.o 33obj-y += ptrace.o
34obj-y += ds.o 34obj-y += ds.o
@@ -47,11 +47,12 @@ obj-$(CONFIG_MICROCODE) += microcode.o
47obj-$(CONFIG_PCI) += early-quirks.o 47obj-$(CONFIG_PCI) += early-quirks.o
48apm-y := apm_32.o 48apm-y := apm_32.o
49obj-$(CONFIG_APM) += apm.o 49obj-$(CONFIG_APM) += apm.o
50obj-$(CONFIG_X86_SMP) += smp_$(BITS).o smpboot_$(BITS).o tsc_sync.o 50obj-$(CONFIG_X86_SMP) += smp.o
51obj-$(CONFIG_X86_32_SMP) += smpcommon_32.o 51obj-$(CONFIG_X86_SMP) += smpboot.o tsc_sync.o ipi.o tlb_$(BITS).o
52obj-$(CONFIG_X86_64_SMP) += smp_64.o smpboot_64.o tsc_sync.o 52obj-$(CONFIG_X86_32_SMP) += smpcommon.o
53obj-$(CONFIG_X86_64_SMP) += tsc_sync.o smpcommon.o
53obj-$(CONFIG_X86_TRAMPOLINE) += trampoline_$(BITS).o 54obj-$(CONFIG_X86_TRAMPOLINE) += trampoline_$(BITS).o
54obj-$(CONFIG_X86_MPPARSE) += mpparse_$(BITS).o 55obj-$(CONFIG_X86_MPPARSE) += mpparse.o
55obj-$(CONFIG_X86_LOCAL_APIC) += apic_$(BITS).o nmi_$(BITS).o 56obj-$(CONFIG_X86_LOCAL_APIC) += apic_$(BITS).o nmi_$(BITS).o
56obj-$(CONFIG_X86_IO_APIC) += io_apic_$(BITS).o 57obj-$(CONFIG_X86_IO_APIC) += io_apic_$(BITS).o
57obj-$(CONFIG_X86_REBOOTFIXUPS) += reboot_fixups_32.o 58obj-$(CONFIG_X86_REBOOTFIXUPS) += reboot_fixups_32.o
@@ -60,7 +61,7 @@ obj-$(CONFIG_KEXEC) += relocate_kernel_$(BITS).o crash.o
60obj-$(CONFIG_CRASH_DUMP) += crash_dump_$(BITS).o 61obj-$(CONFIG_CRASH_DUMP) += crash_dump_$(BITS).o
61obj-$(CONFIG_X86_NUMAQ) += numaq_32.o 62obj-$(CONFIG_X86_NUMAQ) += numaq_32.o
62obj-$(CONFIG_X86_SUMMIT_NUMA) += summit_32.o 63obj-$(CONFIG_X86_SUMMIT_NUMA) += summit_32.o
63obj-$(CONFIG_X86_VSMP) += vsmp_64.o 64obj-y += vsmp_64.o
64obj-$(CONFIG_KPROBES) += kprobes.o 65obj-$(CONFIG_KPROBES) += kprobes.o
65obj-$(CONFIG_MODULES) += module_$(BITS).o 66obj-$(CONFIG_MODULES) += module_$(BITS).o
66obj-$(CONFIG_ACPI_SRAT) += srat_32.o 67obj-$(CONFIG_ACPI_SRAT) += srat_32.o
@@ -90,7 +91,7 @@ scx200-y += scx200_32.o
90### 91###
91# 64 bit specific files 92# 64 bit specific files
92ifeq ($(CONFIG_X86_64),y) 93ifeq ($(CONFIG_X86_64),y)
93 obj-y += genapic_64.o genapic_flat_64.o 94 obj-y += genapic_64.o genapic_flat_64.o genx2apic_uv_x.o
94 obj-$(CONFIG_X86_PM_TIMER) += pmtimer_64.o 95 obj-$(CONFIG_X86_PM_TIMER) += pmtimer_64.o
95 obj-$(CONFIG_AUDIT) += audit_64.o 96 obj-$(CONFIG_AUDIT) += audit_64.o
96 97
diff --git a/arch/x86/kernel/acpi/Makefile b/arch/x86/kernel/acpi/Makefile
index 19d3d6e9d09b..7335959b6aff 100644
--- a/arch/x86/kernel/acpi/Makefile
+++ b/arch/x86/kernel/acpi/Makefile
@@ -1,7 +1,14 @@
1subdir- := realmode
2
1obj-$(CONFIG_ACPI) += boot.o 3obj-$(CONFIG_ACPI) += boot.o
2obj-$(CONFIG_ACPI_SLEEP) += sleep.o wakeup_$(BITS).o 4obj-$(CONFIG_ACPI_SLEEP) += sleep.o wakeup_rm.o wakeup_$(BITS).o
3 5
4ifneq ($(CONFIG_ACPI_PROCESSOR),) 6ifneq ($(CONFIG_ACPI_PROCESSOR),)
5obj-y += cstate.o processor.o 7obj-y += cstate.o processor.o
6endif 8endif
7 9
10$(obj)/wakeup_rm.o: $(obj)/realmode/wakeup.bin
11
12$(obj)/realmode/wakeup.bin: FORCE
13 $(Q)$(MAKE) $(build)=$(obj)/realmode $@
14
diff --git a/arch/x86/kernel/acpi/boot.c b/arch/x86/kernel/acpi/boot.c
index 2cdc9de9371d..057ccf1d5ad4 100644
--- a/arch/x86/kernel/acpi/boot.c
+++ b/arch/x86/kernel/acpi/boot.c
@@ -39,6 +39,11 @@
39#include <asm/apic.h> 39#include <asm/apic.h>
40#include <asm/io.h> 40#include <asm/io.h>
41#include <asm/mpspec.h> 41#include <asm/mpspec.h>
42#include <asm/smp.h>
43
44#ifdef CONFIG_X86_LOCAL_APIC
45# include <mach_apic.h>
46#endif
42 47
43static int __initdata acpi_force = 0; 48static int __initdata acpi_force = 0;
44 49
@@ -52,9 +57,7 @@ EXPORT_SYMBOL(acpi_disabled);
52#ifdef CONFIG_X86_64 57#ifdef CONFIG_X86_64
53 58
54#include <asm/proto.h> 59#include <asm/proto.h>
55 60#include <asm/genapic.h>
56static inline int acpi_madt_oem_check(char *oem_id, char *oem_table_id) { return 0; }
57
58 61
59#else /* X86 */ 62#else /* X86 */
60 63
@@ -111,7 +114,7 @@ char *__init __acpi_map_table(unsigned long phys_addr, unsigned long size)
111 if (!phys_addr || !size) 114 if (!phys_addr || !size)
112 return NULL; 115 return NULL;
113 116
114 if (phys_addr+size <= (end_pfn_map << PAGE_SHIFT) + PAGE_SIZE) 117 if (phys_addr+size <= (max_pfn_mapped << PAGE_SHIFT) + PAGE_SIZE)
115 return __va(phys_addr); 118 return __va(phys_addr);
116 119
117 return NULL; 120 return NULL;
@@ -237,6 +240,16 @@ static int __init acpi_parse_madt(struct acpi_table_header *table)
237 return 0; 240 return 0;
238} 241}
239 242
243static void __cpuinit acpi_register_lapic(int id, u8 enabled)
244{
245 if (!enabled) {
246 ++disabled_cpus;
247 return;
248 }
249
250 generic_processor_info(id, 0);
251}
252
240static int __init 253static int __init
241acpi_parse_lapic(struct acpi_subtable_header * header, const unsigned long end) 254acpi_parse_lapic(struct acpi_subtable_header * header, const unsigned long end)
242{ 255{
@@ -256,8 +269,26 @@ acpi_parse_lapic(struct acpi_subtable_header * header, const unsigned long end)
256 * to not preallocating memory for all NR_CPUS 269 * to not preallocating memory for all NR_CPUS
257 * when we use CPU hotplug. 270 * when we use CPU hotplug.
258 */ 271 */
259 mp_register_lapic(processor->id, /* APIC ID */ 272 acpi_register_lapic(processor->id, /* APIC ID */
260 processor->lapic_flags & ACPI_MADT_ENABLED); /* Enabled? */ 273 processor->lapic_flags & ACPI_MADT_ENABLED);
274
275 return 0;
276}
277
278static int __init
279acpi_parse_sapic(struct acpi_subtable_header *header, const unsigned long end)
280{
281 struct acpi_madt_local_sapic *processor = NULL;
282
283 processor = (struct acpi_madt_local_sapic *)header;
284
285 if (BAD_MADT_ENTRY(processor, end))
286 return -EINVAL;
287
288 acpi_table_print_madt_entry(header);
289
290 acpi_register_lapic((processor->id << 8) | processor->eid,/* APIC ID */
291 processor->lapic_flags & ACPI_MADT_ENABLED);
261 292
262 return 0; 293 return 0;
263} 294}
@@ -300,6 +331,8 @@ acpi_parse_lapic_nmi(struct acpi_subtable_header * header, const unsigned long e
300 331
301#ifdef CONFIG_X86_IO_APIC 332#ifdef CONFIG_X86_IO_APIC
302 333
334struct mp_ioapic_routing mp_ioapic_routing[MAX_IO_APICS];
335
303static int __init 336static int __init
304acpi_parse_ioapic(struct acpi_subtable_header * header, const unsigned long end) 337acpi_parse_ioapic(struct acpi_subtable_header * header, const unsigned long end)
305{ 338{
@@ -532,7 +565,7 @@ static int __cpuinit _acpi_map_lsapic(acpi_handle handle, int *pcpu)
532 buffer.pointer = NULL; 565 buffer.pointer = NULL;
533 566
534 tmp_map = cpu_present_map; 567 tmp_map = cpu_present_map;
535 mp_register_lapic(physid, lapic->lapic_flags & ACPI_MADT_ENABLED); 568 acpi_register_lapic(physid, lapic->lapic_flags & ACPI_MADT_ENABLED);
536 569
537 /* 570 /*
538 * If mp_register_lapic successfully generates a new logical cpu 571 * If mp_register_lapic successfully generates a new logical cpu
@@ -732,6 +765,16 @@ static int __init acpi_parse_fadt(struct acpi_table_header *table)
732 * Parse LAPIC entries in MADT 765 * Parse LAPIC entries in MADT
733 * returns 0 on success, < 0 on error 766 * returns 0 on success, < 0 on error
734 */ 767 */
768
769static void __init acpi_register_lapic_address(unsigned long address)
770{
771 mp_lapic_addr = address;
772
773 set_fixmap_nocache(FIX_APIC_BASE, address);
774 if (boot_cpu_physical_apicid == -1U)
775 boot_cpu_physical_apicid = GET_APIC_ID(read_apic_id());
776}
777
735static int __init acpi_parse_madt_lapic_entries(void) 778static int __init acpi_parse_madt_lapic_entries(void)
736{ 779{
737 int count; 780 int count;
@@ -753,10 +796,14 @@ static int __init acpi_parse_madt_lapic_entries(void)
753 return count; 796 return count;
754 } 797 }
755 798
756 mp_register_lapic_address(acpi_lapic_addr); 799 acpi_register_lapic_address(acpi_lapic_addr);
800
801 count = acpi_table_parse_madt(ACPI_MADT_TYPE_LOCAL_SAPIC,
802 acpi_parse_sapic, MAX_APICS);
757 803
758 count = acpi_table_parse_madt(ACPI_MADT_TYPE_LOCAL_APIC, acpi_parse_lapic, 804 if (!count)
759 MAX_APICS); 805 count = acpi_table_parse_madt(ACPI_MADT_TYPE_LOCAL_APIC,
806 acpi_parse_lapic, MAX_APICS);
760 if (!count) { 807 if (!count) {
761 printk(KERN_ERR PREFIX "No LAPIC entries present\n"); 808 printk(KERN_ERR PREFIX "No LAPIC entries present\n");
762 /* TBD: Cleanup to allow fallback to MPS */ 809 /* TBD: Cleanup to allow fallback to MPS */
diff --git a/arch/x86/kernel/acpi/realmode/Makefile b/arch/x86/kernel/acpi/realmode/Makefile
new file mode 100644
index 000000000000..092900854acc
--- /dev/null
+++ b/arch/x86/kernel/acpi/realmode/Makefile
@@ -0,0 +1,57 @@
1#
2# arch/x86/kernel/acpi/realmode/Makefile
3#
4# This file is subject to the terms and conditions of the GNU General Public
5# License. See the file "COPYING" in the main directory of this archive
6# for more details.
7#
8
9targets := wakeup.bin wakeup.elf
10
11wakeup-y += wakeup.o wakemain.o video-mode.o copy.o
12
13# The link order of the video-*.o modules can matter. In particular,
14# video-vga.o *must* be listed first, followed by video-vesa.o.
15# Hardware-specific drivers should follow in the order they should be
16# probed, and video-bios.o should typically be last.
17wakeup-y += video-vga.o
18wakeup-y += video-vesa.o
19wakeup-y += video-bios.o
20
21targets += $(wakeup-y)
22
23bootsrc := $(src)/../../../boot
24
25# ---------------------------------------------------------------------------
26
27# How to compile the 16-bit code. Note we always compile for -march=i386,
28# that way we can complain to the user if the CPU is insufficient.
29# Compile with _SETUP since this is similar to the boot-time setup code.
30KBUILD_CFLAGS := $(LINUXINCLUDE) -g -Os -D_SETUP -D_WAKEUP -D__KERNEL__ \
31 -I$(srctree)/$(bootsrc) \
32 $(cflags-y) \
33 -Wall -Wstrict-prototypes \
34 -march=i386 -mregparm=3 \
35 -include $(srctree)/$(bootsrc)/code16gcc.h \
36 -fno-strict-aliasing -fomit-frame-pointer \
37 $(call cc-option, -ffreestanding) \
38 $(call cc-option, -fno-toplevel-reorder,\
39 $(call cc-option, -fno-unit-at-a-time)) \
40 $(call cc-option, -fno-stack-protector) \
41 $(call cc-option, -mpreferred-stack-boundary=2)
42KBUILD_CFLAGS += $(call cc-option, -m32)
43KBUILD_AFLAGS := $(KBUILD_CFLAGS) -D__ASSEMBLY__
44
45WAKEUP_OBJS = $(addprefix $(obj)/,$(wakeup-y))
46
47LDFLAGS_wakeup.elf := -T
48
49CPPFLAGS_wakeup.lds += -P -C
50
51$(obj)/wakeup.elf: $(src)/wakeup.lds $(WAKEUP_OBJS) FORCE
52 $(call if_changed,ld)
53
54OBJCOPYFLAGS_wakeup.bin := -O binary
55
56$(obj)/wakeup.bin: $(obj)/wakeup.elf FORCE
57 $(call if_changed,objcopy)
diff --git a/arch/x86/kernel/acpi/realmode/copy.S b/arch/x86/kernel/acpi/realmode/copy.S
new file mode 100644
index 000000000000..dc59ebee69d8
--- /dev/null
+++ b/arch/x86/kernel/acpi/realmode/copy.S
@@ -0,0 +1 @@
#include "../../../boot/copy.S"
diff --git a/arch/x86/kernel/acpi/realmode/video-bios.c b/arch/x86/kernel/acpi/realmode/video-bios.c
new file mode 100644
index 000000000000..7deabc144a27
--- /dev/null
+++ b/arch/x86/kernel/acpi/realmode/video-bios.c
@@ -0,0 +1 @@
#include "../../../boot/video-bios.c"
diff --git a/arch/x86/kernel/acpi/realmode/video-mode.c b/arch/x86/kernel/acpi/realmode/video-mode.c
new file mode 100644
index 000000000000..328ad209f113
--- /dev/null
+++ b/arch/x86/kernel/acpi/realmode/video-mode.c
@@ -0,0 +1 @@
#include "../../../boot/video-mode.c"
diff --git a/arch/x86/kernel/acpi/realmode/video-vesa.c b/arch/x86/kernel/acpi/realmode/video-vesa.c
new file mode 100644
index 000000000000..9dbb9672226a
--- /dev/null
+++ b/arch/x86/kernel/acpi/realmode/video-vesa.c
@@ -0,0 +1 @@
#include "../../../boot/video-vesa.c"
diff --git a/arch/x86/kernel/acpi/realmode/video-vga.c b/arch/x86/kernel/acpi/realmode/video-vga.c
new file mode 100644
index 000000000000..bcc81255f374
--- /dev/null
+++ b/arch/x86/kernel/acpi/realmode/video-vga.c
@@ -0,0 +1 @@
#include "../../../boot/video-vga.c"
diff --git a/arch/x86/kernel/acpi/realmode/wakemain.c b/arch/x86/kernel/acpi/realmode/wakemain.c
new file mode 100644
index 000000000000..883962d9eef2
--- /dev/null
+++ b/arch/x86/kernel/acpi/realmode/wakemain.c
@@ -0,0 +1,81 @@
1#include "wakeup.h"
2#include "boot.h"
3
4static void udelay(int loops)
5{
6 while (loops--)
7 io_delay(); /* Approximately 1 us */
8}
9
10static void beep(unsigned int hz)
11{
12 u8 enable;
13
14 if (!hz) {
15 enable = 0x00; /* Turn off speaker */
16 } else {
17 u16 div = 1193181/hz;
18
19 outb(0xb6, 0x43); /* Ctr 2, squarewave, load, binary */
20 io_delay();
21 outb(div, 0x42); /* LSB of counter */
22 io_delay();
23 outb(div >> 8, 0x42); /* MSB of counter */
24 io_delay();
25
26 enable = 0x03; /* Turn on speaker */
27 }
28 inb(0x61); /* Dummy read of System Control Port B */
29 io_delay();
30 outb(enable, 0x61); /* Enable timer 2 output to speaker */
31 io_delay();
32}
33
34#define DOT_HZ 880
35#define DASH_HZ 587
36#define US_PER_DOT 125000
37
38/* Okay, this is totally silly, but it's kind of fun. */
39static void send_morse(const char *pattern)
40{
41 char s;
42
43 while ((s = *pattern++)) {
44 switch (s) {
45 case '.':
46 beep(DOT_HZ);
47 udelay(US_PER_DOT);
48 beep(0);
49 udelay(US_PER_DOT);
50 break;
51 case '-':
52 beep(DASH_HZ);
53 udelay(US_PER_DOT * 3);
54 beep(0);
55 udelay(US_PER_DOT);
56 break;
57 default: /* Assume it's a space */
58 udelay(US_PER_DOT * 3);
59 break;
60 }
61 }
62}
63
64void main(void)
65{
66 /* Kill machine if structures are wrong */
67 if (wakeup_header.real_magic != 0x12345678)
68 while (1);
69
70 if (wakeup_header.realmode_flags & 4)
71 send_morse("...-");
72
73 if (wakeup_header.realmode_flags & 1)
74 asm volatile("lcallw $0xc000,$3");
75
76 if (wakeup_header.realmode_flags & 2) {
77 /* Need to call BIOS */
78 probe_cards(0);
79 set_mode(wakeup_header.video_mode);
80 }
81}
diff --git a/arch/x86/kernel/acpi/realmode/wakeup.S b/arch/x86/kernel/acpi/realmode/wakeup.S
new file mode 100644
index 000000000000..f9b77fb37e5b
--- /dev/null
+++ b/arch/x86/kernel/acpi/realmode/wakeup.S
@@ -0,0 +1,113 @@
1/*
2 * ACPI wakeup real mode startup stub
3 */
4#include <asm/segment.h>
5#include <asm/msr-index.h>
6#include <asm/page.h>
7#include <asm/pgtable.h>
8
9 .code16
10 .section ".header", "a"
11
12/* This should match the structure in wakeup.h */
13 .globl wakeup_header
14wakeup_header:
15video_mode: .short 0 /* Video mode number */
16pmode_return: .byte 0x66, 0xea /* ljmpl */
17 .long 0 /* offset goes here */
18 .short __KERNEL_CS
19pmode_cr0: .long 0 /* Saved %cr0 */
20pmode_cr3: .long 0 /* Saved %cr3 */
21pmode_cr4: .long 0 /* Saved %cr4 */
22pmode_efer: .quad 0 /* Saved EFER */
23pmode_gdt: .quad 0
24realmode_flags: .long 0
25real_magic: .long 0
26trampoline_segment: .word 0
27signature: .long 0x51ee1111
28
29 .text
30 .globl _start
31 .code16
32wakeup_code:
33_start:
34 cli
35 cld
36
37 /* Set up segments */
38 movw %cs, %ax
39 movw %ax, %ds
40 movw %ax, %es
41 movw %ax, %ss
42
43 movl $wakeup_stack_end, %esp
44
45 /* Clear the EFLAGS */
46 pushl $0
47 popfl
48
49 /* Check header signature... */
50 movl signature, %eax
51 cmpl $0x51ee1111, %eax
52 jne bogus_real_magic
53
54 /* Check we really have everything... */
55 movl end_signature, %eax
56 cmpl $0x65a22c82, %eax
57 jne bogus_real_magic
58
59 /* Call the C code */
60 calll main
61
62 /* Do any other stuff... */
63
64#ifndef CONFIG_64BIT
65 /* This could also be done in C code... */
66 movl pmode_cr3, %eax
67 movl %eax, %cr3
68
69 movl pmode_cr4, %ecx
70 jecxz 1f
71 movl %ecx, %cr4
721:
73 movl pmode_efer, %eax
74 movl pmode_efer + 4, %edx
75 movl %eax, %ecx
76 orl %edx, %ecx
77 jz 1f
78 movl $0xc0000080, %ecx
79 wrmsr
801:
81
82 lgdtl pmode_gdt
83
84 /* This really couldn't... */
85 movl pmode_cr0, %eax
86 movl %eax, %cr0
87 jmp pmode_return
88#else
89 pushw $0
90 pushw trampoline_segment
91 pushw $0
92 lret
93#endif
94
95bogus_real_magic:
961:
97 hlt
98 jmp 1b
99
100 .data
101 .balign 4
102 .globl HEAP, heap_end
103HEAP:
104 .long wakeup_heap
105heap_end:
106 .long wakeup_stack
107
108 .bss
109wakeup_heap:
110 .space 2048
111wakeup_stack:
112 .space 2048
113wakeup_stack_end:
diff --git a/arch/x86/kernel/acpi/realmode/wakeup.h b/arch/x86/kernel/acpi/realmode/wakeup.h
new file mode 100644
index 000000000000..ef8166fe8020
--- /dev/null
+++ b/arch/x86/kernel/acpi/realmode/wakeup.h
@@ -0,0 +1,36 @@
1/*
2 * Definitions for the wakeup data structure at the head of the
3 * wakeup code.
4 */
5
6#ifndef ARCH_X86_KERNEL_ACPI_RM_WAKEUP_H
7#define ARCH_X86_KERNEL_ACPI_RM_WAKEUP_H
8
9#ifndef __ASSEMBLY__
10#include <linux/types.h>
11
12/* This must match data at wakeup.S */
13struct wakeup_header {
14 u16 video_mode; /* Video mode number */
15 u16 _jmp1; /* ljmpl opcode, 32-bit only */
16 u32 pmode_entry; /* Protected mode resume point, 32-bit only */
17 u16 _jmp2; /* CS value, 32-bit only */
18 u32 pmode_cr0; /* Protected mode cr0 */
19 u32 pmode_cr3; /* Protected mode cr3 */
20 u32 pmode_cr4; /* Protected mode cr4 */
21 u32 pmode_efer_low; /* Protected mode EFER */
22 u32 pmode_efer_high;
23 u64 pmode_gdt;
24 u32 realmode_flags;
25 u32 real_magic;
26 u16 trampoline_segment; /* segment with trampoline code, 64-bit only */
27 u32 signature; /* To check we have correct structure */
28} __attribute__((__packed__));
29
30extern struct wakeup_header wakeup_header;
31#endif
32
33#define HEADER_OFFSET 0x3f00
34#define WAKEUP_SIZE 0x4000
35
36#endif /* ARCH_X86_KERNEL_ACPI_RM_WAKEUP_H */
diff --git a/arch/x86/kernel/acpi/realmode/wakeup.lds.S b/arch/x86/kernel/acpi/realmode/wakeup.lds.S
new file mode 100644
index 000000000000..22fab6c4be15
--- /dev/null
+++ b/arch/x86/kernel/acpi/realmode/wakeup.lds.S
@@ -0,0 +1,61 @@
1/*
2 * wakeup.ld
3 *
4 * Linker script for the real-mode wakeup code
5 */
6#undef i386
7#include "wakeup.h"
8
9OUTPUT_FORMAT("elf32-i386", "elf32-i386", "elf32-i386")
10OUTPUT_ARCH(i386)
11ENTRY(_start)
12
13SECTIONS
14{
15 . = HEADER_OFFSET;
16 .header : {
17 *(.header)
18 }
19
20 . = 0;
21 .text : {
22 *(.text*)
23 }
24
25 . = ALIGN(16);
26 .rodata : {
27 *(.rodata*)
28 }
29
30 .videocards : {
31 video_cards = .;
32 *(.videocards)
33 video_cards_end = .;
34 }
35
36 . = ALIGN(16);
37 .data : {
38 *(.data*)
39 }
40
41 .signature : {
42 end_signature = .;
43 LONG(0x65a22c82)
44 }
45
46 . = ALIGN(16);
47 .bss : {
48 __bss_start = .;
49 *(.bss)
50 __bss_end = .;
51 }
52
53 . = ALIGN(16);
54 _end = .;
55
56 /DISCARD/ : {
57 *(.note*)
58 }
59
60 . = ASSERT(_end <= WAKEUP_SIZE, "Wakeup too big!");
61}
diff --git a/arch/x86/kernel/acpi/sleep.c b/arch/x86/kernel/acpi/sleep.c
index 6bc815cd8cb3..afc25ee9964b 100644
--- a/arch/x86/kernel/acpi/sleep.c
+++ b/arch/x86/kernel/acpi/sleep.c
@@ -10,30 +10,72 @@
10#include <linux/dmi.h> 10#include <linux/dmi.h>
11#include <linux/cpumask.h> 11#include <linux/cpumask.h>
12 12
13#include <asm/smp.h> 13#include "realmode/wakeup.h"
14#include "sleep.h"
14 15
15/* address in low memory of the wakeup routine. */ 16unsigned long acpi_wakeup_address;
16unsigned long acpi_wakeup_address = 0;
17unsigned long acpi_realmode_flags; 17unsigned long acpi_realmode_flags;
18extern char wakeup_start, wakeup_end;
19 18
20extern unsigned long acpi_copy_wakeup_routine(unsigned long); 19/* address in low memory of the wakeup routine. */
20static unsigned long acpi_realmode;
21
22#ifdef CONFIG_64BIT
23static char temp_stack[10240];
24#endif
21 25
22/** 26/**
23 * acpi_save_state_mem - save kernel state 27 * acpi_save_state_mem - save kernel state
24 * 28 *
25 * Create an identity mapped page table and copy the wakeup routine to 29 * Create an identity mapped page table and copy the wakeup routine to
26 * low memory. 30 * low memory.
31 *
32 * Note that this is too late to change acpi_wakeup_address.
27 */ 33 */
28int acpi_save_state_mem(void) 34int acpi_save_state_mem(void)
29{ 35{
30 if (!acpi_wakeup_address) { 36 struct wakeup_header *header;
31 printk(KERN_ERR "Could not allocate memory during boot, S3 disabled\n"); 37
38 if (!acpi_realmode) {
39 printk(KERN_ERR "Could not allocate memory during boot, "
40 "S3 disabled\n");
32 return -ENOMEM; 41 return -ENOMEM;
33 } 42 }
34 memcpy((void *)acpi_wakeup_address, &wakeup_start, 43 memcpy((void *)acpi_realmode, &wakeup_code_start, WAKEUP_SIZE);
35 &wakeup_end - &wakeup_start); 44
36 acpi_copy_wakeup_routine(acpi_wakeup_address); 45 header = (struct wakeup_header *)(acpi_realmode + HEADER_OFFSET);
46 if (header->signature != 0x51ee1111) {
47 printk(KERN_ERR "wakeup header does not match\n");
48 return -EINVAL;
49 }
50
51 header->video_mode = saved_video_mode;
52
53#ifndef CONFIG_64BIT
54 store_gdt((struct desc_ptr *)&header->pmode_gdt);
55
56 header->pmode_efer_low = nx_enabled;
57 if (header->pmode_efer_low & 1) {
58 /* This is strange, why not save efer, always? */
59 rdmsr(MSR_EFER, header->pmode_efer_low,
60 header->pmode_efer_high);
61 }
62#endif /* !CONFIG_64BIT */
63
64 header->pmode_cr0 = read_cr0();
65 header->pmode_cr4 = read_cr4();
66 header->realmode_flags = acpi_realmode_flags;
67 header->real_magic = 0x12345678;
68
69#ifndef CONFIG_64BIT
70 header->pmode_entry = (u32)&wakeup_pmode_return;
71 header->pmode_cr3 = (u32)(swsusp_pg_dir - __PAGE_OFFSET);
72 saved_magic = 0x12345678;
73#else /* CONFIG_64BIT */
74 header->trampoline_segment = setup_trampoline() >> 4;
75 init_rsp = (unsigned long)temp_stack + 4096;
76 initial_code = (unsigned long)wakeup_long64;
77 saved_magic = 0x123456789abcdef0;
78#endif /* CONFIG_64BIT */
37 79
38 return 0; 80 return 0;
39} 81}
@@ -56,15 +98,20 @@ void acpi_restore_state_mem(void)
56 */ 98 */
57void __init acpi_reserve_bootmem(void) 99void __init acpi_reserve_bootmem(void)
58{ 100{
59 if ((&wakeup_end - &wakeup_start) > PAGE_SIZE*2) { 101 if ((&wakeup_code_end - &wakeup_code_start) > WAKEUP_SIZE) {
60 printk(KERN_ERR 102 printk(KERN_ERR
61 "ACPI: Wakeup code way too big, S3 disabled.\n"); 103 "ACPI: Wakeup code way too big, S3 disabled.\n");
62 return; 104 return;
63 } 105 }
64 106
65 acpi_wakeup_address = (unsigned long)alloc_bootmem_low(PAGE_SIZE*2); 107 acpi_realmode = (unsigned long)alloc_bootmem_low(WAKEUP_SIZE);
66 if (!acpi_wakeup_address) 108
109 if (!acpi_realmode) {
67 printk(KERN_ERR "ACPI: Cannot allocate lowmem, S3 disabled.\n"); 110 printk(KERN_ERR "ACPI: Cannot allocate lowmem, S3 disabled.\n");
111 return;
112 }
113
114 acpi_wakeup_address = acpi_realmode;
68} 115}
69 116
70 117
diff --git a/arch/x86/kernel/acpi/sleep.h b/arch/x86/kernel/acpi/sleep.h
new file mode 100644
index 000000000000..adbcbaa6f1df
--- /dev/null
+++ b/arch/x86/kernel/acpi/sleep.h
@@ -0,0 +1,16 @@
1/*
2 * Variables and functions used by the code in sleep.c
3 */
4
5#include <asm/trampoline.h>
6
7extern char wakeup_code_start, wakeup_code_end;
8
9extern unsigned long saved_video_mode;
10extern long saved_magic;
11
12extern int wakeup_pmode_return;
13extern char swsusp_pg_dir[PAGE_SIZE];
14
15extern unsigned long acpi_copy_wakeup_routine(unsigned long);
16extern void wakeup_long64(void);
diff --git a/arch/x86/kernel/acpi/sleep_32.c b/arch/x86/kernel/acpi/sleep_32.c
deleted file mode 100644
index 63fe5525e026..000000000000
--- a/arch/x86/kernel/acpi/sleep_32.c
+++ /dev/null
@@ -1,40 +0,0 @@
1/*
2 * sleep.c - x86-specific ACPI sleep support.
3 *
4 * Copyright (C) 2001-2003 Patrick Mochel
5 * Copyright (C) 2001-2003 Pavel Machek <pavel@suse.cz>
6 */
7
8#include <linux/acpi.h>
9#include <linux/bootmem.h>
10#include <linux/dmi.h>
11#include <linux/cpumask.h>
12
13#include <asm/smp.h>
14
15/* Ouch, we want to delete this. We already have better version in userspace, in
16 s2ram from suspend.sf.net project */
17static __init int reset_videomode_after_s3(const struct dmi_system_id *d)
18{
19 acpi_realmode_flags |= 2;
20 return 0;
21}
22
23static __initdata struct dmi_system_id acpisleep_dmi_table[] = {
24 { /* Reset video mode after returning from ACPI S3 sleep */
25 .callback = reset_videomode_after_s3,
26 .ident = "Toshiba Satellite 4030cdt",
27 .matches = {
28 DMI_MATCH(DMI_PRODUCT_NAME, "S4030CDT/4.3"),
29 },
30 },
31 {}
32};
33
34static int __init acpisleep_dmi_init(void)
35{
36 dmi_check_system(acpisleep_dmi_table);
37 return 0;
38}
39
40core_initcall(acpisleep_dmi_init);
diff --git a/arch/x86/kernel/acpi/wakeup_32.S b/arch/x86/kernel/acpi/wakeup_32.S
index f53e3277f8e5..a12e6a9fb659 100644
--- a/arch/x86/kernel/acpi/wakeup_32.S
+++ b/arch/x86/kernel/acpi/wakeup_32.S
@@ -3,178 +3,12 @@
3#include <asm/segment.h> 3#include <asm/segment.h>
4#include <asm/page.h> 4#include <asm/page.h>
5 5
6# 6# Copyright 2003, 2008 Pavel Machek <pavel@suse.cz>, distribute under GPLv2
7# wakeup_code runs in real mode, and at unknown address (determined at run-time).
8# Therefore it must only use relative jumps/calls.
9#
10# Do we need to deal with A20? It is okay: ACPI specs says A20 must be enabled
11#
12# If physical address of wakeup_code is 0x12345, BIOS should call us with
13# cs = 0x1234, eip = 0x05
14#
15
16#define BEEP \
17 inb $97, %al; \
18 outb %al, $0x80; \
19 movb $3, %al; \
20 outb %al, $97; \
21 outb %al, $0x80; \
22 movb $-74, %al; \
23 outb %al, $67; \
24 outb %al, $0x80; \
25 movb $-119, %al; \
26 outb %al, $66; \
27 outb %al, $0x80; \
28 movb $15, %al; \
29 outb %al, $66;
30
31ALIGN
32 .align 4096
33ENTRY(wakeup_start)
34wakeup_code:
35 wakeup_code_start = .
36 .code16
37
38 cli
39 cld
40
41 # setup data segment
42 movw %cs, %ax
43 movw %ax, %ds # Make ds:0 point to wakeup_start
44 movw %ax, %ss
45
46 testl $4, realmode_flags - wakeup_code
47 jz 1f
48 BEEP
491:
50 mov $(wakeup_stack - wakeup_code), %sp # Private stack is needed for ASUS board
51
52 pushl $0 # Kill any dangerous flags
53 popfl
54
55 movl real_magic - wakeup_code, %eax
56 cmpl $0x12345678, %eax
57 jne bogus_real_magic
58
59 testl $1, realmode_flags - wakeup_code
60 jz 1f
61 lcall $0xc000,$3
62 movw %cs, %ax
63 movw %ax, %ds # Bios might have played with that
64 movw %ax, %ss
651:
66
67 testl $2, realmode_flags - wakeup_code
68 jz 1f
69 mov video_mode - wakeup_code, %ax
70 call mode_set
711:
72
73 # set up page table
74 movl $swsusp_pg_dir-__PAGE_OFFSET, %eax
75 movl %eax, %cr3
76
77 testl $1, real_efer_save_restore - wakeup_code
78 jz 4f
79 # restore efer setting
80 movl real_save_efer_edx - wakeup_code, %edx
81 movl real_save_efer_eax - wakeup_code, %eax
82 mov $0xc0000080, %ecx
83 wrmsr
844:
85 # make sure %cr4 is set correctly (features, etc)
86 movl real_save_cr4 - wakeup_code, %eax
87 movl %eax, %cr4
88
89 # need a gdt -- use lgdtl to force 32-bit operands, in case
90 # the GDT is located past 16 megabytes.
91 lgdtl real_save_gdt - wakeup_code
92
93 movl real_save_cr0 - wakeup_code, %eax
94 movl %eax, %cr0
95 jmp 1f
961:
97 movl real_magic - wakeup_code, %eax
98 cmpl $0x12345678, %eax
99 jne bogus_real_magic
100
101 testl $8, realmode_flags - wakeup_code
102 jz 1f
103 BEEP
1041:
105 ljmpl $__KERNEL_CS, $wakeup_pmode_return
106
107real_save_gdt: .word 0
108 .long 0
109real_save_cr0: .long 0
110real_save_cr3: .long 0
111real_save_cr4: .long 0
112real_magic: .long 0
113video_mode: .long 0
114realmode_flags: .long 0
115real_efer_save_restore: .long 0
116real_save_efer_edx: .long 0
117real_save_efer_eax: .long 0
118
119bogus_real_magic:
120 jmp bogus_real_magic
121
122/* This code uses an extended set of video mode numbers. These include:
123 * Aliases for standard modes
124 * NORMAL_VGA (-1)
125 * EXTENDED_VGA (-2)
126 * ASK_VGA (-3)
127 * Video modes numbered by menu position -- NOT RECOMMENDED because of lack
128 * of compatibility when extending the table. These are between 0x00 and 0xff.
129 */
130#define VIDEO_FIRST_MENU 0x0000
131
132/* Standard BIOS video modes (BIOS number + 0x0100) */
133#define VIDEO_FIRST_BIOS 0x0100
134
135/* VESA BIOS video modes (VESA number + 0x0200) */
136#define VIDEO_FIRST_VESA 0x0200
137
138/* Video7 special modes (BIOS number + 0x0900) */
139#define VIDEO_FIRST_V7 0x0900
140
141# Setting of user mode (AX=mode ID) => CF=success
142
143# For now, we only handle VESA modes (0x0200..0x03ff). To handle other
144# modes, we should probably compile in the video code from the boot
145# directory.
146mode_set:
147 movw %ax, %bx
148 subb $VIDEO_FIRST_VESA>>8, %bh
149 cmpb $2, %bh
150 jb check_vesa
151
152setbad:
153 clc
154 ret
155
156check_vesa:
157 orw $0x4000, %bx # Use linear frame buffer
158 movw $0x4f02, %ax # VESA BIOS mode set call
159 int $0x10
160 cmpw $0x004f, %ax # AL=4f if implemented
161 jnz setbad # AH=0 if OK
162
163 stc
164 ret
165 7
166 .code32 8 .code32
167 ALIGN 9 ALIGN
168 10
169.org 0x800 11ENTRY(wakeup_pmode_return)
170wakeup_stack_begin: # Stack grows down
171
172.org 0xff0 # Just below end of page
173wakeup_stack:
174ENTRY(wakeup_end)
175
176.org 0x1000
177
178wakeup_pmode_return: 12wakeup_pmode_return:
179 movw $__KERNEL_DS, %ax 13 movw $__KERNEL_DS, %ax
180 movw %ax, %ss 14 movw %ax, %ss
@@ -187,7 +21,7 @@ wakeup_pmode_return:
187 lgdt saved_gdt 21 lgdt saved_gdt
188 lidt saved_idt 22 lidt saved_idt
189 lldt saved_ldt 23 lldt saved_ldt
190 ljmp $(__KERNEL_CS),$1f 24 ljmp $(__KERNEL_CS), $1f
1911: 251:
192 movl %cr3, %eax 26 movl %cr3, %eax
193 movl %eax, %cr3 27 movl %eax, %cr3
@@ -201,82 +35,41 @@ wakeup_pmode_return:
201 jne bogus_magic 35 jne bogus_magic
202 36
203 # jump to place where we left off 37 # jump to place where we left off
204 movl saved_eip,%eax 38 movl saved_eip, %eax
205 jmp *%eax 39 jmp *%eax
206 40
207bogus_magic: 41bogus_magic:
208 jmp bogus_magic 42 jmp bogus_magic
209 43
210 44
211##
212# acpi_copy_wakeup_routine
213#
214# Copy the above routine to low memory.
215#
216# Parameters:
217# %eax: place to copy wakeup routine to
218#
219# Returned address is location of code in low memory (past data and stack)
220#
221ENTRY(acpi_copy_wakeup_routine)
222 45
223 pushl %ebx 46save_registers:
224 sgdt saved_gdt 47 sgdt saved_gdt
225 sidt saved_idt 48 sidt saved_idt
226 sldt saved_ldt 49 sldt saved_ldt
227 str saved_tss 50 str saved_tss
228 51
229 movl nx_enabled, %edx
230 movl %edx, real_efer_save_restore - wakeup_start (%eax)
231 testl $1, real_efer_save_restore - wakeup_start (%eax)
232 jz 2f
233 # save efer setting
234 pushl %eax
235 movl %eax, %ebx
236 mov $0xc0000080, %ecx
237 rdmsr
238 movl %edx, real_save_efer_edx - wakeup_start (%ebx)
239 movl %eax, real_save_efer_eax - wakeup_start (%ebx)
240 popl %eax
2412:
242
243 movl %cr3, %edx
244 movl %edx, real_save_cr3 - wakeup_start (%eax)
245 movl %cr4, %edx
246 movl %edx, real_save_cr4 - wakeup_start (%eax)
247 movl %cr0, %edx
248 movl %edx, real_save_cr0 - wakeup_start (%eax)
249 sgdt real_save_gdt - wakeup_start (%eax)
250
251 movl saved_videomode, %edx
252 movl %edx, video_mode - wakeup_start (%eax)
253 movl acpi_realmode_flags, %edx
254 movl %edx, realmode_flags - wakeup_start (%eax)
255 movl $0x12345678, real_magic - wakeup_start (%eax)
256 movl $0x12345678, saved_magic
257 popl %ebx
258 ret
259
260save_registers:
261 leal 4(%esp), %eax 52 leal 4(%esp), %eax
262 movl %eax, saved_context_esp 53 movl %eax, saved_context_esp
263 movl %ebx, saved_context_ebx 54 movl %ebx, saved_context_ebx
264 movl %ebp, saved_context_ebp 55 movl %ebp, saved_context_ebp
265 movl %esi, saved_context_esi 56 movl %esi, saved_context_esi
266 movl %edi, saved_context_edi 57 movl %edi, saved_context_edi
267 pushfl ; popl saved_context_eflags 58 pushfl
268 59 popl saved_context_eflags
269 movl $ret_point, saved_eip 60
61 movl $ret_point, saved_eip
270 ret 62 ret
271 63
272 64
273restore_registers: 65restore_registers:
274 movl saved_context_ebp, %ebp 66 movl saved_context_ebp, %ebp
275 movl saved_context_ebx, %ebx 67 movl saved_context_ebx, %ebx
276 movl saved_context_esi, %esi 68 movl saved_context_esi, %esi
277 movl saved_context_edi, %edi 69 movl saved_context_edi, %edi
278 pushl saved_context_eflags ; popfl 70 pushl saved_context_eflags
279 ret 71 popfl
72 ret
280 73
281ENTRY(do_suspend_lowlevel) 74ENTRY(do_suspend_lowlevel)
282 call save_processor_state 75 call save_processor_state
diff --git a/arch/x86/kernel/acpi/wakeup_64.S b/arch/x86/kernel/acpi/wakeup_64.S
index 2e1b9e0d0767..bcc293423a70 100644
--- a/arch/x86/kernel/acpi/wakeup_64.S
+++ b/arch/x86/kernel/acpi/wakeup_64.S
@@ -7,191 +7,18 @@
7#include <asm/asm-offsets.h> 7#include <asm/asm-offsets.h>
8 8
9# Copyright 2003 Pavel Machek <pavel@suse.cz>, distribute under GPLv2 9# Copyright 2003 Pavel Machek <pavel@suse.cz>, distribute under GPLv2
10#
11# wakeup_code runs in real mode, and at unknown address (determined at run-time).
12# Therefore it must only use relative jumps/calls.
13#
14# Do we need to deal with A20? It is okay: ACPI specs says A20 must be enabled
15#
16# If physical address of wakeup_code is 0x12345, BIOS should call us with
17# cs = 0x1234, eip = 0x05
18#
19
20#define BEEP \
21 inb $97, %al; \
22 outb %al, $0x80; \
23 movb $3, %al; \
24 outb %al, $97; \
25 outb %al, $0x80; \
26 movb $-74, %al; \
27 outb %al, $67; \
28 outb %al, $0x80; \
29 movb $-119, %al; \
30 outb %al, $66; \
31 outb %al, $0x80; \
32 movb $15, %al; \
33 outb %al, $66;
34
35
36ALIGN
37 .align 16
38ENTRY(wakeup_start)
39wakeup_code:
40 wakeup_code_start = .
41 .code16
42
43# Running in *copy* of this code, somewhere in low 1MB.
44
45 cli
46 cld
47 # setup data segment
48 movw %cs, %ax
49 movw %ax, %ds # Make ds:0 point to wakeup_start
50 movw %ax, %ss
51
52 # Data segment must be set up before we can see whether to beep.
53 testl $4, realmode_flags - wakeup_code
54 jz 1f
55 BEEP
561:
57
58 # Private stack is needed for ASUS board
59 mov $(wakeup_stack - wakeup_code), %sp
60
61 pushl $0 # Kill any dangerous flags
62 popfl
63
64 movl real_magic - wakeup_code, %eax
65 cmpl $0x12345678, %eax
66 jne bogus_real_magic
67
68 testl $1, realmode_flags - wakeup_code
69 jz 1f
70 lcall $0xc000,$3
71 movw %cs, %ax
72 movw %ax, %ds # Bios might have played with that
73 movw %ax, %ss
741:
75
76 testl $2, realmode_flags - wakeup_code
77 jz 1f
78 mov video_mode - wakeup_code, %ax
79 call mode_set
801:
81
82 mov %ds, %ax # Find 32bit wakeup_code addr
83 movzx %ax, %esi # (Convert %ds:gdt to a liner ptr)
84 shll $4, %esi
85 # Fix up the vectors
86 addl %esi, wakeup_32_vector - wakeup_code
87 addl %esi, wakeup_long64_vector - wakeup_code
88 addl %esi, gdt_48a + 2 - wakeup_code # Fixup the gdt pointer
89
90 lidtl %ds:idt_48a - wakeup_code
91 lgdtl %ds:gdt_48a - wakeup_code # load gdt with whatever is
92 # appropriate
93
94 movl $1, %eax # protected mode (PE) bit
95 lmsw %ax # This is it!
96 jmp 1f
971:
98
99 ljmpl *(wakeup_32_vector - wakeup_code)
100
101 .balign 4
102wakeup_32_vector:
103 .long wakeup_32 - wakeup_code
104 .word __KERNEL32_CS, 0
105
106 .code32
107wakeup_32:
108# Running in this code, but at low address; paging is not yet turned on.
109
110 movl $__KERNEL_DS, %eax
111 movl %eax, %ds
112
113 /*
114 * Prepare for entering 64bits mode
115 */
116
117 /* Enable PAE */
118 xorl %eax, %eax
119 btsl $5, %eax
120 movl %eax, %cr4
121
122 /* Setup early boot stage 4 level pagetables */
123 leal (wakeup_level4_pgt - wakeup_code)(%esi), %eax
124 movl %eax, %cr3
125
126 /* Check if nx is implemented */
127 movl $0x80000001, %eax
128 cpuid
129 movl %edx,%edi
130
131 /* Enable Long Mode */
132 xorl %eax, %eax
133 btsl $_EFER_LME, %eax
134
135 /* No Execute supported? */
136 btl $20,%edi
137 jnc 1f
138 btsl $_EFER_NX, %eax
139
140 /* Make changes effective */
1411: movl $MSR_EFER, %ecx
142 xorl %edx, %edx
143 wrmsr
144
145 xorl %eax, %eax
146 btsl $31, %eax /* Enable paging and in turn activate Long Mode */
147 btsl $0, %eax /* Enable protected mode */
148
149 /* Make changes effective */
150 movl %eax, %cr0
151
152 /* At this point:
153 CR4.PAE must be 1
154 CS.L must be 0
155 CR3 must point to PML4
156 Next instruction must be a branch
157 This must be on identity-mapped page
158 */
159 /*
160 * At this point we're in long mode but in 32bit compatibility mode
161 * with EFER.LME = 1, CS.L = 0, CS.D = 1 (and in turn
162 * EFER.LMA = 1). Now we want to jump in 64bit mode, to do that we load
163 * the new gdt/idt that has __KERNEL_CS with CS.L = 1.
164 */
165
166 /* Finally jump in 64bit mode */
167 ljmp *(wakeup_long64_vector - wakeup_code)(%esi)
168
169 .balign 4
170wakeup_long64_vector:
171 .long wakeup_long64 - wakeup_code
172 .word __KERNEL_CS, 0
173 10
174.code64 11.code64
175
176 /* Hooray, we are in Long 64-bit mode (but still running in
177 * low memory)
178 */
179wakeup_long64:
180 /* 12 /*
181 * We must switch to a new descriptor in kernel space for the GDT 13 * Hooray, we are in Long 64-bit mode (but still running in low memory)
182 * because soon the kernel won't have access anymore to the userspace
183 * addresses where we're currently running on. We have to do that here
184 * because in 32bit we couldn't load a 64bit linear address.
185 */ 14 */
186 lgdt cpu_gdt_descr 15ENTRY(wakeup_long64)
187 16wakeup_long64:
188 movq saved_magic, %rax 17 movq saved_magic, %rax
189 movq $0x123456789abcdef0, %rdx 18 movq $0x123456789abcdef0, %rdx
190 cmpq %rdx, %rax 19 cmpq %rdx, %rax
191 jne bogus_64_magic 20 jne bogus_64_magic
192 21
193 nop
194 nop
195 movw $__KERNEL_DS, %ax 22 movw $__KERNEL_DS, %ax
196 movw %ax, %ss 23 movw %ax, %ss
197 movw %ax, %ds 24 movw %ax, %ds
@@ -208,130 +35,8 @@ wakeup_long64:
208 movq saved_rip, %rax 35 movq saved_rip, %rax
209 jmp *%rax 36 jmp *%rax
210 37
211.code32
212
213 .align 64
214gdta:
215 /* Its good to keep gdt in sync with one in trampoline.S */
216 .word 0, 0, 0, 0 # dummy
217 /* ??? Why I need the accessed bit set in order for this to work? */
218 .quad 0x00cf9b000000ffff # __KERNEL32_CS
219 .quad 0x00af9b000000ffff # __KERNEL_CS
220 .quad 0x00cf93000000ffff # __KERNEL_DS
221
222idt_48a:
223 .word 0 # idt limit = 0
224 .word 0, 0 # idt base = 0L
225
226gdt_48a:
227 .word 0x800 # gdt limit=2048,
228 # 256 GDT entries
229 .long gdta - wakeup_code # gdt base (relocated in later)
230
231real_magic: .quad 0
232video_mode: .quad 0
233realmode_flags: .quad 0
234
235.code16
236bogus_real_magic:
237 jmp bogus_real_magic
238
239.code64
240bogus_64_magic: 38bogus_64_magic:
241 jmp bogus_64_magic 39 jmp bogus_64_magic
242
243/* This code uses an extended set of video mode numbers. These include:
244 * Aliases for standard modes
245 * NORMAL_VGA (-1)
246 * EXTENDED_VGA (-2)
247 * ASK_VGA (-3)
248 * Video modes numbered by menu position -- NOT RECOMMENDED because of lack
249 * of compatibility when extending the table. These are between 0x00 and 0xff.
250 */
251#define VIDEO_FIRST_MENU 0x0000
252
253/* Standard BIOS video modes (BIOS number + 0x0100) */
254#define VIDEO_FIRST_BIOS 0x0100
255
256/* VESA BIOS video modes (VESA number + 0x0200) */
257#define VIDEO_FIRST_VESA 0x0200
258
259/* Video7 special modes (BIOS number + 0x0900) */
260#define VIDEO_FIRST_V7 0x0900
261
262# Setting of user mode (AX=mode ID) => CF=success
263
264# For now, we only handle VESA modes (0x0200..0x03ff). To handle other
265# modes, we should probably compile in the video code from the boot
266# directory.
267.code16
268mode_set:
269 movw %ax, %bx
270 subb $VIDEO_FIRST_VESA>>8, %bh
271 cmpb $2, %bh
272 jb check_vesa
273
274setbad:
275 clc
276 ret
277
278check_vesa:
279 orw $0x4000, %bx # Use linear frame buffer
280 movw $0x4f02, %ax # VESA BIOS mode set call
281 int $0x10
282 cmpw $0x004f, %ax # AL=4f if implemented
283 jnz setbad # AH=0 if OK
284
285 stc
286 ret
287
288wakeup_stack_begin: # Stack grows down
289
290.org 0xff0
291wakeup_stack: # Just below end of page
292
293.org 0x1000
294ENTRY(wakeup_level4_pgt)
295 .quad level3_ident_pgt - __START_KERNEL_map + _KERNPG_TABLE
296 .fill 510,8,0
297 /* (2^48-(2*1024*1024*1024))/(2^39) = 511 */
298 .quad level3_kernel_pgt - __START_KERNEL_map + _KERNPG_TABLE
299
300ENTRY(wakeup_end)
301
302##
303# acpi_copy_wakeup_routine
304#
305# Copy the above routine to low memory.
306#
307# Parameters:
308# %rdi: place to copy wakeup routine to
309#
310# Returned address is location of code in low memory (past data and stack)
311#
312 .code64
313ENTRY(acpi_copy_wakeup_routine)
314 pushq %rax
315 pushq %rdx
316
317 movl saved_video_mode, %edx
318 movl %edx, video_mode - wakeup_start (,%rdi)
319 movl acpi_realmode_flags, %edx
320 movl %edx, realmode_flags - wakeup_start (,%rdi)
321 movq $0x12345678, real_magic - wakeup_start (,%rdi)
322 movq $0x123456789abcdef0, %rdx
323 movq %rdx, saved_magic
324
325 movq saved_magic, %rax
326 movq $0x123456789abcdef0, %rdx
327 cmpq %rdx, %rax
328 jne bogus_64_magic
329
330 # restore the regs we used
331 popq %rdx
332 popq %rax
333ENTRY(do_suspend_lowlevel_s4bios)
334 ret
335 40
336 .align 2 41 .align 2
337 .p2align 4,,15 42 .p2align 4,,15
@@ -414,7 +119,7 @@ do_suspend_lowlevel:
414 jmp restore_processor_state 119 jmp restore_processor_state
415.LFE5: 120.LFE5:
416.Lfe5: 121.Lfe5:
417 .size do_suspend_lowlevel,.Lfe5-do_suspend_lowlevel 122 .size do_suspend_lowlevel, .Lfe5-do_suspend_lowlevel
418 123
419.data 124.data
420ALIGN 125ALIGN
diff --git a/arch/x86/kernel/acpi/wakeup_rm.S b/arch/x86/kernel/acpi/wakeup_rm.S
new file mode 100644
index 000000000000..6ff3b5730575
--- /dev/null
+++ b/arch/x86/kernel/acpi/wakeup_rm.S
@@ -0,0 +1,10 @@
1/*
2 * Wrapper script for the realmode binary as a transport object
3 * before copying to low memory.
4 */
5 .section ".rodata","a"
6 .globl wakeup_code_start, wakeup_code_end
7wakeup_code_start:
8 .incbin "arch/x86/kernel/acpi/realmode/wakeup.bin"
9wakeup_code_end:
10 .size wakeup_code_start, .-wakeup_code_start
diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
index 5fed98ca0e1f..df4099dc1c68 100644
--- a/arch/x86/kernel/alternative.c
+++ b/arch/x86/kernel/alternative.c
@@ -11,6 +11,8 @@
11#include <asm/mce.h> 11#include <asm/mce.h>
12#include <asm/nmi.h> 12#include <asm/nmi.h>
13#include <asm/vsyscall.h> 13#include <asm/vsyscall.h>
14#include <asm/cacheflush.h>
15#include <asm/io.h>
14 16
15#define MAX_PATCH_LEN (255-1) 17#define MAX_PATCH_LEN (255-1)
16 18
@@ -177,7 +179,7 @@ static const unsigned char*const * find_nop_table(void)
177#endif /* CONFIG_X86_64 */ 179#endif /* CONFIG_X86_64 */
178 180
179/* Use this to add nops to a buffer, then text_poke the whole buffer. */ 181/* Use this to add nops to a buffer, then text_poke the whole buffer. */
180static void add_nops(void *insns, unsigned int len) 182void add_nops(void *insns, unsigned int len)
181{ 183{
182 const unsigned char *const *noptable = find_nop_table(); 184 const unsigned char *const *noptable = find_nop_table();
183 185
@@ -190,6 +192,7 @@ static void add_nops(void *insns, unsigned int len)
190 len -= noplen; 192 len -= noplen;
191 } 193 }
192} 194}
195EXPORT_SYMBOL_GPL(add_nops);
193 196
194extern struct alt_instr __alt_instructions[], __alt_instructions_end[]; 197extern struct alt_instr __alt_instructions[], __alt_instructions_end[];
195extern u8 *__smp_locks[], *__smp_locks_end[]; 198extern u8 *__smp_locks[], *__smp_locks_end[];
@@ -205,7 +208,7 @@ void apply_alternatives(struct alt_instr *start, struct alt_instr *end)
205 struct alt_instr *a; 208 struct alt_instr *a;
206 char insnbuf[MAX_PATCH_LEN]; 209 char insnbuf[MAX_PATCH_LEN];
207 210
208 DPRINTK("%s: alt table %p -> %p\n", __FUNCTION__, start, end); 211 DPRINTK("%s: alt table %p -> %p\n", __func__, start, end);
209 for (a = start; a < end; a++) { 212 for (a = start; a < end; a++) {
210 u8 *instr = a->instr; 213 u8 *instr = a->instr;
211 BUG_ON(a->replacementlen > a->instrlen); 214 BUG_ON(a->replacementlen > a->instrlen);
@@ -217,13 +220,13 @@ void apply_alternatives(struct alt_instr *start, struct alt_instr *end)
217 if (instr >= (u8 *)VSYSCALL_START && instr < (u8*)VSYSCALL_END) { 220 if (instr >= (u8 *)VSYSCALL_START && instr < (u8*)VSYSCALL_END) {
218 instr = __va(instr - (u8*)VSYSCALL_START + (u8*)__pa_symbol(&__vsyscall_0)); 221 instr = __va(instr - (u8*)VSYSCALL_START + (u8*)__pa_symbol(&__vsyscall_0));
219 DPRINTK("%s: vsyscall fixup: %p => %p\n", 222 DPRINTK("%s: vsyscall fixup: %p => %p\n",
220 __FUNCTION__, a->instr, instr); 223 __func__, a->instr, instr);
221 } 224 }
222#endif 225#endif
223 memcpy(insnbuf, a->replacement, a->replacementlen); 226 memcpy(insnbuf, a->replacement, a->replacementlen);
224 add_nops(insnbuf + a->replacementlen, 227 add_nops(insnbuf + a->replacementlen,
225 a->instrlen - a->replacementlen); 228 a->instrlen - a->replacementlen);
226 text_poke(instr, insnbuf, a->instrlen); 229 text_poke_early(instr, insnbuf, a->instrlen);
227 } 230 }
228} 231}
229 232
@@ -284,7 +287,6 @@ void alternatives_smp_module_add(struct module *mod, char *name,
284 void *text, void *text_end) 287 void *text, void *text_end)
285{ 288{
286 struct smp_alt_module *smp; 289 struct smp_alt_module *smp;
287 unsigned long flags;
288 290
289 if (noreplace_smp) 291 if (noreplace_smp)
290 return; 292 return;
@@ -307,42 +309,40 @@ void alternatives_smp_module_add(struct module *mod, char *name,
307 smp->text = text; 309 smp->text = text;
308 smp->text_end = text_end; 310 smp->text_end = text_end;
309 DPRINTK("%s: locks %p -> %p, text %p -> %p, name %s\n", 311 DPRINTK("%s: locks %p -> %p, text %p -> %p, name %s\n",
310 __FUNCTION__, smp->locks, smp->locks_end, 312 __func__, smp->locks, smp->locks_end,
311 smp->text, smp->text_end, smp->name); 313 smp->text, smp->text_end, smp->name);
312 314
313 spin_lock_irqsave(&smp_alt, flags); 315 spin_lock(&smp_alt);
314 list_add_tail(&smp->next, &smp_alt_modules); 316 list_add_tail(&smp->next, &smp_alt_modules);
315 if (boot_cpu_has(X86_FEATURE_UP)) 317 if (boot_cpu_has(X86_FEATURE_UP))
316 alternatives_smp_unlock(smp->locks, smp->locks_end, 318 alternatives_smp_unlock(smp->locks, smp->locks_end,
317 smp->text, smp->text_end); 319 smp->text, smp->text_end);
318 spin_unlock_irqrestore(&smp_alt, flags); 320 spin_unlock(&smp_alt);
319} 321}
320 322
321void alternatives_smp_module_del(struct module *mod) 323void alternatives_smp_module_del(struct module *mod)
322{ 324{
323 struct smp_alt_module *item; 325 struct smp_alt_module *item;
324 unsigned long flags;
325 326
326 if (smp_alt_once || noreplace_smp) 327 if (smp_alt_once || noreplace_smp)
327 return; 328 return;
328 329
329 spin_lock_irqsave(&smp_alt, flags); 330 spin_lock(&smp_alt);
330 list_for_each_entry(item, &smp_alt_modules, next) { 331 list_for_each_entry(item, &smp_alt_modules, next) {
331 if (mod != item->mod) 332 if (mod != item->mod)
332 continue; 333 continue;
333 list_del(&item->next); 334 list_del(&item->next);
334 spin_unlock_irqrestore(&smp_alt, flags); 335 spin_unlock(&smp_alt);
335 DPRINTK("%s: %s\n", __FUNCTION__, item->name); 336 DPRINTK("%s: %s\n", __func__, item->name);
336 kfree(item); 337 kfree(item);
337 return; 338 return;
338 } 339 }
339 spin_unlock_irqrestore(&smp_alt, flags); 340 spin_unlock(&smp_alt);
340} 341}
341 342
342void alternatives_smp_switch(int smp) 343void alternatives_smp_switch(int smp)
343{ 344{
344 struct smp_alt_module *mod; 345 struct smp_alt_module *mod;
345 unsigned long flags;
346 346
347#ifdef CONFIG_LOCKDEP 347#ifdef CONFIG_LOCKDEP
348 /* 348 /*
@@ -359,7 +359,7 @@ void alternatives_smp_switch(int smp)
359 return; 359 return;
360 BUG_ON(!smp && (num_online_cpus() > 1)); 360 BUG_ON(!smp && (num_online_cpus() > 1));
361 361
362 spin_lock_irqsave(&smp_alt, flags); 362 spin_lock(&smp_alt);
363 363
364 /* 364 /*
365 * Avoid unnecessary switches because it forces JIT based VMs to 365 * Avoid unnecessary switches because it forces JIT based VMs to
@@ -383,7 +383,7 @@ void alternatives_smp_switch(int smp)
383 mod->text, mod->text_end); 383 mod->text, mod->text_end);
384 } 384 }
385 smp_mode = smp; 385 smp_mode = smp;
386 spin_unlock_irqrestore(&smp_alt, flags); 386 spin_unlock(&smp_alt);
387} 387}
388 388
389#endif 389#endif
@@ -411,7 +411,7 @@ void apply_paravirt(struct paravirt_patch_site *start,
411 411
412 /* Pad the rest with nops */ 412 /* Pad the rest with nops */
413 add_nops(insnbuf + used, p->len - used); 413 add_nops(insnbuf + used, p->len - used);
414 text_poke(p->instr, insnbuf, p->len); 414 text_poke_early(p->instr, insnbuf, p->len);
415 } 415 }
416} 416}
417extern struct paravirt_patch_site __start_parainstructions[], 417extern struct paravirt_patch_site __start_parainstructions[],
@@ -420,8 +420,6 @@ extern struct paravirt_patch_site __start_parainstructions[],
420 420
421void __init alternative_instructions(void) 421void __init alternative_instructions(void)
422{ 422{
423 unsigned long flags;
424
425 /* The patching is not fully atomic, so try to avoid local interruptions 423 /* The patching is not fully atomic, so try to avoid local interruptions
426 that might execute the to be patched code. 424 that might execute the to be patched code.
427 Other CPUs are not running. */ 425 Other CPUs are not running. */
@@ -430,7 +428,6 @@ void __init alternative_instructions(void)
430 stop_mce(); 428 stop_mce();
431#endif 429#endif
432 430
433 local_irq_save(flags);
434 apply_alternatives(__alt_instructions, __alt_instructions_end); 431 apply_alternatives(__alt_instructions, __alt_instructions_end);
435 432
436 /* switch to patch-once-at-boottime-only mode and free the 433 /* switch to patch-once-at-boottime-only mode and free the
@@ -462,7 +459,6 @@ void __init alternative_instructions(void)
462 } 459 }
463#endif 460#endif
464 apply_paravirt(__parainstructions, __parainstructions_end); 461 apply_paravirt(__parainstructions, __parainstructions_end);
465 local_irq_restore(flags);
466 462
467 if (smp_alt_once) 463 if (smp_alt_once)
468 free_init_pages("SMP alternatives", 464 free_init_pages("SMP alternatives",
@@ -475,18 +471,71 @@ void __init alternative_instructions(void)
475#endif 471#endif
476} 472}
477 473
478/* 474/**
479 * Warning: 475 * text_poke_early - Update instructions on a live kernel at boot time
476 * @addr: address to modify
477 * @opcode: source of the copy
478 * @len: length to copy
479 *
480 * When you use this code to patch more than one byte of an instruction 480 * When you use this code to patch more than one byte of an instruction
481 * you need to make sure that other CPUs cannot execute this code in parallel. 481 * you need to make sure that other CPUs cannot execute this code in parallel.
482 * Also no thread must be currently preempted in the middle of these instructions. 482 * Also no thread must be currently preempted in the middle of these
483 * And on the local CPU you need to be protected again NMI or MCE handlers 483 * instructions. And on the local CPU you need to be protected again NMI or MCE
484 * seeing an inconsistent instruction while you patch. 484 * handlers seeing an inconsistent instruction while you patch.
485 */ 485 */
486void __kprobes text_poke(void *addr, unsigned char *opcode, int len) 486void *text_poke_early(void *addr, const void *opcode, size_t len)
487{ 487{
488 unsigned long flags;
489 local_irq_save(flags);
488 memcpy(addr, opcode, len); 490 memcpy(addr, opcode, len);
491 local_irq_restore(flags);
492 sync_core();
493 /* Could also do a CLFLUSH here to speed up CPU recovery; but
494 that causes hangs on some VIA CPUs. */
495 return addr;
496}
497
498/**
499 * text_poke - Update instructions on a live kernel
500 * @addr: address to modify
501 * @opcode: source of the copy
502 * @len: length to copy
503 *
504 * Only atomic text poke/set should be allowed when not doing early patching.
505 * It means the size must be writable atomically and the address must be aligned
506 * in a way that permits an atomic write. It also makes sure we fit on a single
507 * page.
508 */
509void *__kprobes text_poke(void *addr, const void *opcode, size_t len)
510{
511 unsigned long flags;
512 char *vaddr;
513 int nr_pages = 2;
514
515 BUG_ON(len > sizeof(long));
516 BUG_ON((((long)addr + len - 1) & ~(sizeof(long) - 1))
517 - ((long)addr & ~(sizeof(long) - 1)));
518 if (kernel_text_address((unsigned long)addr)) {
519 struct page *pages[2] = { virt_to_page(addr),
520 virt_to_page(addr + PAGE_SIZE) };
521 if (!pages[1])
522 nr_pages = 1;
523 vaddr = vmap(pages, nr_pages, VM_MAP, PAGE_KERNEL);
524 BUG_ON(!vaddr);
525 local_irq_save(flags);
526 memcpy(&vaddr[(unsigned long)addr & ~PAGE_MASK], opcode, len);
527 local_irq_restore(flags);
528 vunmap(vaddr);
529 } else {
530 /*
531 * modules are in vmalloc'ed memory, always writable.
532 */
533 local_irq_save(flags);
534 memcpy(addr, opcode, len);
535 local_irq_restore(flags);
536 }
489 sync_core(); 537 sync_core();
490 /* Could also do a CLFLUSH here to speed up CPU recovery; but 538 /* Could also do a CLFLUSH here to speed up CPU recovery; but
491 that causes hangs on some VIA CPUs. */ 539 that causes hangs on some VIA CPUs. */
540 return addr;
492} 541}
diff --git a/arch/x86/kernel/aperture_64.c b/arch/x86/kernel/aperture_64.c
index 00df126169b4..479926d9e004 100644
--- a/arch/x86/kernel/aperture_64.c
+++ b/arch/x86/kernel/aperture_64.c
@@ -27,11 +27,11 @@
27#include <asm/k8.h> 27#include <asm/k8.h>
28 28
29int gart_iommu_aperture; 29int gart_iommu_aperture;
30int gart_iommu_aperture_disabled __initdata = 0; 30int gart_iommu_aperture_disabled __initdata;
31int gart_iommu_aperture_allowed __initdata = 0; 31int gart_iommu_aperture_allowed __initdata;
32 32
33int fallback_aper_order __initdata = 1; /* 64MB */ 33int fallback_aper_order __initdata = 1; /* 64MB */
34int fallback_aper_force __initdata = 0; 34int fallback_aper_force __initdata;
35 35
36int fix_aperture __initdata = 1; 36int fix_aperture __initdata = 1;
37 37
diff --git a/arch/x86/kernel/apic_32.c b/arch/x86/kernel/apic_32.c
index 35a568ea8400..687208190b06 100644
--- a/arch/x86/kernel/apic_32.c
+++ b/arch/x86/kernel/apic_32.c
@@ -50,6 +50,11 @@
50# error SPURIOUS_APIC_VECTOR definition error 50# error SPURIOUS_APIC_VECTOR definition error
51#endif 51#endif
52 52
53unsigned long mp_lapic_addr;
54
55DEFINE_PER_CPU(u16, x86_bios_cpu_apicid) = BAD_APICID;
56EXPORT_PER_CPU_SYMBOL(x86_bios_cpu_apicid);
57
53/* 58/*
54 * Knob to control our willingness to enable the local APIC. 59 * Knob to control our willingness to enable the local APIC.
55 * 60 *
@@ -621,6 +626,35 @@ int setup_profiling_timer(unsigned int multiplier)
621} 626}
622 627
623/* 628/*
629 * Setup extended LVT, AMD specific (K8, family 10h)
630 *
631 * Vector mappings are hard coded. On K8 only offset 0 (APIC500) and
632 * MCE interrupts are supported. Thus MCE offset must be set to 0.
633 */
634
635#define APIC_EILVT_LVTOFF_MCE 0
636#define APIC_EILVT_LVTOFF_IBS 1
637
638static void setup_APIC_eilvt(u8 lvt_off, u8 vector, u8 msg_type, u8 mask)
639{
640 unsigned long reg = (lvt_off << 4) + APIC_EILVT0;
641 unsigned int v = (mask << 16) | (msg_type << 8) | vector;
642 apic_write(reg, v);
643}
644
645u8 setup_APIC_eilvt_mce(u8 vector, u8 msg_type, u8 mask)
646{
647 setup_APIC_eilvt(APIC_EILVT_LVTOFF_MCE, vector, msg_type, mask);
648 return APIC_EILVT_LVTOFF_MCE;
649}
650
651u8 setup_APIC_eilvt_ibs(u8 vector, u8 msg_type, u8 mask)
652{
653 setup_APIC_eilvt(APIC_EILVT_LVTOFF_IBS, vector, msg_type, mask);
654 return APIC_EILVT_LVTOFF_IBS;
655}
656
657/*
624 * Local APIC start and shutdown 658 * Local APIC start and shutdown
625 */ 659 */
626 660
@@ -868,12 +902,50 @@ void __init init_bsp_APIC(void)
868 apic_write_around(APIC_LVT1, value); 902 apic_write_around(APIC_LVT1, value);
869} 903}
870 904
905void __cpuinit lapic_setup_esr(void)
906{
907 unsigned long oldvalue, value, maxlvt;
908 if (lapic_is_integrated() && !esr_disable) {
909 /* !82489DX */
910 maxlvt = lapic_get_maxlvt();
911 if (maxlvt > 3) /* Due to the Pentium erratum 3AP. */
912 apic_write(APIC_ESR, 0);
913 oldvalue = apic_read(APIC_ESR);
914
915 /* enables sending errors */
916 value = ERROR_APIC_VECTOR;
917 apic_write_around(APIC_LVTERR, value);
918 /*
919 * spec says clear errors after enabling vector.
920 */
921 if (maxlvt > 3)
922 apic_write(APIC_ESR, 0);
923 value = apic_read(APIC_ESR);
924 if (value != oldvalue)
925 apic_printk(APIC_VERBOSE, "ESR value before enabling "
926 "vector: 0x%08lx after: 0x%08lx\n",
927 oldvalue, value);
928 } else {
929 if (esr_disable)
930 /*
931 * Something untraceable is creating bad interrupts on
932 * secondary quads ... for the moment, just leave the
933 * ESR disabled - we can't do anything useful with the
934 * errors anyway - mbligh
935 */
936 printk(KERN_INFO "Leaving ESR disabled.\n");
937 else
938 printk(KERN_INFO "No ESR for 82489DX.\n");
939 }
940}
941
942
871/** 943/**
872 * setup_local_APIC - setup the local APIC 944 * setup_local_APIC - setup the local APIC
873 */ 945 */
874void __cpuinit setup_local_APIC(void) 946void __cpuinit setup_local_APIC(void)
875{ 947{
876 unsigned long oldvalue, value, maxlvt, integrated; 948 unsigned long value, integrated;
877 int i, j; 949 int i, j;
878 950
879 /* Pound the ESR really hard over the head with a big hammer - mbligh */ 951 /* Pound the ESR really hard over the head with a big hammer - mbligh */
@@ -997,40 +1069,13 @@ void __cpuinit setup_local_APIC(void)
997 if (!integrated) /* 82489DX */ 1069 if (!integrated) /* 82489DX */
998 value |= APIC_LVT_LEVEL_TRIGGER; 1070 value |= APIC_LVT_LEVEL_TRIGGER;
999 apic_write_around(APIC_LVT1, value); 1071 apic_write_around(APIC_LVT1, value);
1072}
1000 1073
1001 if (integrated && !esr_disable) { 1074void __cpuinit end_local_APIC_setup(void)
1002 /* !82489DX */ 1075{
1003 maxlvt = lapic_get_maxlvt(); 1076 unsigned long value;
1004 if (maxlvt > 3) /* Due to the Pentium erratum 3AP. */
1005 apic_write(APIC_ESR, 0);
1006 oldvalue = apic_read(APIC_ESR);
1007
1008 /* enables sending errors */
1009 value = ERROR_APIC_VECTOR;
1010 apic_write_around(APIC_LVTERR, value);
1011 /*
1012 * spec says clear errors after enabling vector.
1013 */
1014 if (maxlvt > 3)
1015 apic_write(APIC_ESR, 0);
1016 value = apic_read(APIC_ESR);
1017 if (value != oldvalue)
1018 apic_printk(APIC_VERBOSE, "ESR value before enabling "
1019 "vector: 0x%08lx after: 0x%08lx\n",
1020 oldvalue, value);
1021 } else {
1022 if (esr_disable)
1023 /*
1024 * Something untraceable is creating bad interrupts on
1025 * secondary quads ... for the moment, just leave the
1026 * ESR disabled - we can't do anything useful with the
1027 * errors anyway - mbligh
1028 */
1029 printk(KERN_INFO "Leaving ESR disabled.\n");
1030 else
1031 printk(KERN_INFO "No ESR for 82489DX.\n");
1032 }
1033 1077
1078 lapic_setup_esr();
1034 /* Disable the local apic timer */ 1079 /* Disable the local apic timer */
1035 value = apic_read(APIC_LVTT); 1080 value = apic_read(APIC_LVTT);
1036 value |= (APIC_LVT_MASKED | LOCAL_TIMER_VECTOR); 1081 value |= (APIC_LVT_MASKED | LOCAL_TIMER_VECTOR);
@@ -1147,7 +1192,7 @@ void __init init_apic_mappings(void)
1147 * default configuration (or the MP table is broken). 1192 * default configuration (or the MP table is broken).
1148 */ 1193 */
1149 if (boot_cpu_physical_apicid == -1U) 1194 if (boot_cpu_physical_apicid == -1U)
1150 boot_cpu_physical_apicid = GET_APIC_ID(apic_read(APIC_ID)); 1195 boot_cpu_physical_apicid = GET_APIC_ID(read_apic_id());
1151 1196
1152#ifdef CONFIG_X86_IO_APIC 1197#ifdef CONFIG_X86_IO_APIC
1153 { 1198 {
@@ -1185,6 +1230,9 @@ fake_ioapic_page:
1185 * This initializes the IO-APIC and APIC hardware if this is 1230 * This initializes the IO-APIC and APIC hardware if this is
1186 * a UP kernel. 1231 * a UP kernel.
1187 */ 1232 */
1233
1234int apic_version[MAX_APICS];
1235
1188int __init APIC_init_uniprocessor(void) 1236int __init APIC_init_uniprocessor(void)
1189{ 1237{
1190 if (enable_local_apic < 0) 1238 if (enable_local_apic < 0)
@@ -1214,12 +1262,13 @@ int __init APIC_init_uniprocessor(void)
1214 * might be zero if read from MP tables. Get it from LAPIC. 1262 * might be zero if read from MP tables. Get it from LAPIC.
1215 */ 1263 */
1216#ifdef CONFIG_CRASH_DUMP 1264#ifdef CONFIG_CRASH_DUMP
1217 boot_cpu_physical_apicid = GET_APIC_ID(apic_read(APIC_ID)); 1265 boot_cpu_physical_apicid = GET_APIC_ID(read_apic_id());
1218#endif 1266#endif
1219 phys_cpu_present_map = physid_mask_of_physid(boot_cpu_physical_apicid); 1267 phys_cpu_present_map = physid_mask_of_physid(boot_cpu_physical_apicid);
1220 1268
1221 setup_local_APIC(); 1269 setup_local_APIC();
1222 1270
1271 end_local_APIC_setup();
1223#ifdef CONFIG_X86_IO_APIC 1272#ifdef CONFIG_X86_IO_APIC
1224 if (smp_found_config) 1273 if (smp_found_config)
1225 if (!skip_ioapic_setup && nr_ioapics) 1274 if (!skip_ioapic_setup && nr_ioapics)
@@ -1288,6 +1337,29 @@ void smp_error_interrupt(struct pt_regs *regs)
1288 irq_exit(); 1337 irq_exit();
1289} 1338}
1290 1339
1340#ifdef CONFIG_SMP
1341void __init smp_intr_init(void)
1342{
1343 /*
1344 * IRQ0 must be given a fixed assignment and initialized,
1345 * because it's used before the IO-APIC is set up.
1346 */
1347 set_intr_gate(FIRST_DEVICE_VECTOR, interrupt[0]);
1348
1349 /*
1350 * The reschedule interrupt is a CPU-to-CPU reschedule-helper
1351 * IPI, driven by wakeup.
1352 */
1353 set_intr_gate(RESCHEDULE_VECTOR, reschedule_interrupt);
1354
1355 /* IPI for invalidation */
1356 set_intr_gate(INVALIDATE_TLB_VECTOR, invalidate_interrupt);
1357
1358 /* IPI for generic function call */
1359 set_intr_gate(CALL_FUNCTION_VECTOR, call_function_interrupt);
1360}
1361#endif
1362
1291/* 1363/*
1292 * Initialize APIC interrupts 1364 * Initialize APIC interrupts
1293 */ 1365 */
@@ -1394,6 +1466,88 @@ void disconnect_bsp_APIC(int virt_wire_setup)
1394 } 1466 }
1395} 1467}
1396 1468
1469unsigned int __cpuinitdata maxcpus = NR_CPUS;
1470
1471void __cpuinit generic_processor_info(int apicid, int version)
1472{
1473 int cpu;
1474 cpumask_t tmp_map;
1475 physid_mask_t phys_cpu;
1476
1477 /*
1478 * Validate version
1479 */
1480 if (version == 0x0) {
1481 printk(KERN_WARNING "BIOS bug, APIC version is 0 for CPU#%d! "
1482 "fixing up to 0x10. (tell your hw vendor)\n",
1483 version);
1484 version = 0x10;
1485 }
1486 apic_version[apicid] = version;
1487
1488 phys_cpu = apicid_to_cpu_present(apicid);
1489 physids_or(phys_cpu_present_map, phys_cpu_present_map, phys_cpu);
1490
1491 if (num_processors >= NR_CPUS) {
1492 printk(KERN_WARNING "WARNING: NR_CPUS limit of %i reached."
1493 " Processor ignored.\n", NR_CPUS);
1494 return;
1495 }
1496
1497 if (num_processors >= maxcpus) {
1498 printk(KERN_WARNING "WARNING: maxcpus limit of %i reached."
1499 " Processor ignored.\n", maxcpus);
1500 return;
1501 }
1502
1503 num_processors++;
1504 cpus_complement(tmp_map, cpu_present_map);
1505 cpu = first_cpu(tmp_map);
1506
1507 if (apicid == boot_cpu_physical_apicid)
1508 /*
1509 * x86_bios_cpu_apicid is required to have processors listed
1510 * in same order as logical cpu numbers. Hence the first
1511 * entry is BSP, and so on.
1512 */
1513 cpu = 0;
1514
1515 /*
1516 * Would be preferable to switch to bigsmp when CONFIG_HOTPLUG_CPU=y
1517 * but we need to work other dependencies like SMP_SUSPEND etc
1518 * before this can be done without some confusion.
1519 * if (CPU_HOTPLUG_ENABLED || num_processors > 8)
1520 * - Ashok Raj <ashok.raj@intel.com>
1521 */
1522 if (num_processors > 8) {
1523 switch (boot_cpu_data.x86_vendor) {
1524 case X86_VENDOR_INTEL:
1525 if (!APIC_XAPIC(version)) {
1526 def_to_bigsmp = 0;
1527 break;
1528 }
1529 /* If P4 and above fall through */
1530 case X86_VENDOR_AMD:
1531 def_to_bigsmp = 1;
1532 }
1533 }
1534#ifdef CONFIG_SMP
1535 /* are we being called early in kernel startup? */
1536 if (x86_cpu_to_apicid_early_ptr) {
1537 u16 *cpu_to_apicid = x86_cpu_to_apicid_early_ptr;
1538 u16 *bios_cpu_apicid = x86_bios_cpu_apicid_early_ptr;
1539
1540 cpu_to_apicid[cpu] = apicid;
1541 bios_cpu_apicid[cpu] = apicid;
1542 } else {
1543 per_cpu(x86_cpu_to_apicid, cpu) = apicid;
1544 per_cpu(x86_bios_cpu_apicid, cpu) = apicid;
1545 }
1546#endif
1547 cpu_set(cpu, cpu_possible_map);
1548 cpu_set(cpu, cpu_present_map);
1549}
1550
1397/* 1551/*
1398 * Power management 1552 * Power management
1399 */ 1553 */
diff --git a/arch/x86/kernel/apic_64.c b/arch/x86/kernel/apic_64.c
index d8d03e09dea2..9e8e5c050c55 100644
--- a/arch/x86/kernel/apic_64.c
+++ b/arch/x86/kernel/apic_64.c
@@ -34,13 +34,15 @@
34#include <asm/mpspec.h> 34#include <asm/mpspec.h>
35#include <asm/hpet.h> 35#include <asm/hpet.h>
36#include <asm/pgalloc.h> 36#include <asm/pgalloc.h>
37#include <asm/mach_apic.h>
38#include <asm/nmi.h> 37#include <asm/nmi.h>
39#include <asm/idle.h> 38#include <asm/idle.h>
40#include <asm/proto.h> 39#include <asm/proto.h>
41#include <asm/timex.h> 40#include <asm/timex.h>
42#include <asm/apic.h> 41#include <asm/apic.h>
43 42
43#include <mach_ipi.h>
44#include <mach_apic.h>
45
44int disable_apic_timer __cpuinitdata; 46int disable_apic_timer __cpuinitdata;
45static int apic_calibrate_pmtmr __initdata; 47static int apic_calibrate_pmtmr __initdata;
46int disable_apic; 48int disable_apic;
@@ -83,6 +85,12 @@ static DEFINE_PER_CPU(struct clock_event_device, lapic_events);
83 85
84static unsigned long apic_phys; 86static unsigned long apic_phys;
85 87
88unsigned long mp_lapic_addr;
89
90DEFINE_PER_CPU(u16, x86_bios_cpu_apicid) = BAD_APICID;
91EXPORT_PER_CPU_SYMBOL(x86_bios_cpu_apicid);
92
93unsigned int __cpuinitdata maxcpus = NR_CPUS;
86/* 94/*
87 * Get the LAPIC version 95 * Get the LAPIC version
88 */ 96 */
@@ -431,7 +439,8 @@ void __cpuinit check_boot_apic_timer_broadcast(void)
431 lapic_clockevent.features |= CLOCK_EVT_FEAT_DUMMY; 439 lapic_clockevent.features |= CLOCK_EVT_FEAT_DUMMY;
432 440
433 local_irq_enable(); 441 local_irq_enable();
434 clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_FORCE, &boot_cpu_id); 442 clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_FORCE,
443 &boot_cpu_physical_apicid);
435 local_irq_disable(); 444 local_irq_disable();
436} 445}
437 446
@@ -640,10 +649,10 @@ int __init verify_local_APIC(void)
640 /* 649 /*
641 * The ID register is read/write in a real APIC. 650 * The ID register is read/write in a real APIC.
642 */ 651 */
643 reg0 = apic_read(APIC_ID); 652 reg0 = read_apic_id();
644 apic_printk(APIC_DEBUG, "Getting ID: %x\n", reg0); 653 apic_printk(APIC_DEBUG, "Getting ID: %x\n", reg0);
645 apic_write(APIC_ID, reg0 ^ APIC_ID_MASK); 654 apic_write(APIC_ID, reg0 ^ APIC_ID_MASK);
646 reg1 = apic_read(APIC_ID); 655 reg1 = read_apic_id();
647 apic_printk(APIC_DEBUG, "Getting ID: %x\n", reg1); 656 apic_printk(APIC_DEBUG, "Getting ID: %x\n", reg1);
648 apic_write(APIC_ID, reg0); 657 apic_write(APIC_ID, reg0);
649 if (reg1 != (reg0 ^ APIC_ID_MASK)) 658 if (reg1 != (reg0 ^ APIC_ID_MASK))
@@ -728,6 +737,7 @@ void __cpuinit setup_local_APIC(void)
728 unsigned int value; 737 unsigned int value;
729 int i, j; 738 int i, j;
730 739
740 preempt_disable();
731 value = apic_read(APIC_LVR); 741 value = apic_read(APIC_LVR);
732 742
733 BUILD_BUG_ON((SPURIOUS_APIC_VECTOR & 0x0f) != 0x0f); 743 BUILD_BUG_ON((SPURIOUS_APIC_VECTOR & 0x0f) != 0x0f);
@@ -821,6 +831,7 @@ void __cpuinit setup_local_APIC(void)
821 else 831 else
822 value = APIC_DM_NMI | APIC_LVT_MASKED; 832 value = APIC_DM_NMI | APIC_LVT_MASKED;
823 apic_write(APIC_LVT1, value); 833 apic_write(APIC_LVT1, value);
834 preempt_enable();
824} 835}
825 836
826void __cpuinit lapic_setup_esr(void) 837void __cpuinit lapic_setup_esr(void)
@@ -857,10 +868,34 @@ static int __init detect_init_APIC(void)
857 } 868 }
858 869
859 mp_lapic_addr = APIC_DEFAULT_PHYS_BASE; 870 mp_lapic_addr = APIC_DEFAULT_PHYS_BASE;
860 boot_cpu_id = 0; 871 boot_cpu_physical_apicid = 0;
861 return 0; 872 return 0;
862} 873}
863 874
875void __init early_init_lapic_mapping(void)
876{
877 unsigned long apic_phys;
878
879 /*
880 * If no local APIC can be found then go out
881 * : it means there is no mpatable and MADT
882 */
883 if (!smp_found_config)
884 return;
885
886 apic_phys = mp_lapic_addr;
887
888 set_fixmap_nocache(FIX_APIC_BASE, apic_phys);
889 apic_printk(APIC_VERBOSE, "mapped APIC to %16lx (%16lx)\n",
890 APIC_BASE, apic_phys);
891
892 /*
893 * Fetch the APIC ID of the BSP in case we have a
894 * default configuration (or the MP table is broken).
895 */
896 boot_cpu_physical_apicid = GET_APIC_ID(read_apic_id());
897}
898
864/** 899/**
865 * init_apic_mappings - initialize APIC mappings 900 * init_apic_mappings - initialize APIC mappings
866 */ 901 */
@@ -881,16 +916,11 @@ void __init init_apic_mappings(void)
881 apic_printk(APIC_VERBOSE, "mapped APIC to %16lx (%16lx)\n", 916 apic_printk(APIC_VERBOSE, "mapped APIC to %16lx (%16lx)\n",
882 APIC_BASE, apic_phys); 917 APIC_BASE, apic_phys);
883 918
884 /* Put local APIC into the resource map. */
885 lapic_resource.start = apic_phys;
886 lapic_resource.end = lapic_resource.start + PAGE_SIZE - 1;
887 insert_resource(&iomem_resource, &lapic_resource);
888
889 /* 919 /*
890 * Fetch the APIC ID of the BSP in case we have a 920 * Fetch the APIC ID of the BSP in case we have a
891 * default configuration (or the MP table is broken). 921 * default configuration (or the MP table is broken).
892 */ 922 */
893 boot_cpu_id = GET_APIC_ID(apic_read(APIC_ID)); 923 boot_cpu_physical_apicid = GET_APIC_ID(read_apic_id());
894} 924}
895 925
896/* 926/*
@@ -911,8 +941,8 @@ int __init APIC_init_uniprocessor(void)
911 941
912 verify_local_APIC(); 942 verify_local_APIC();
913 943
914 phys_cpu_present_map = physid_mask_of_physid(boot_cpu_id); 944 phys_cpu_present_map = physid_mask_of_physid(boot_cpu_physical_apicid);
915 apic_write(APIC_ID, SET_APIC_ID(boot_cpu_id)); 945 apic_write(APIC_ID, SET_APIC_ID(boot_cpu_physical_apicid));
916 946
917 setup_local_APIC(); 947 setup_local_APIC();
918 948
@@ -1029,6 +1059,52 @@ void disconnect_bsp_APIC(int virt_wire_setup)
1029 apic_write(APIC_LVT1, value); 1059 apic_write(APIC_LVT1, value);
1030} 1060}
1031 1061
1062void __cpuinit generic_processor_info(int apicid, int version)
1063{
1064 int cpu;
1065 cpumask_t tmp_map;
1066
1067 if (num_processors >= NR_CPUS) {
1068 printk(KERN_WARNING "WARNING: NR_CPUS limit of %i reached."
1069 " Processor ignored.\n", NR_CPUS);
1070 return;
1071 }
1072
1073 if (num_processors >= maxcpus) {
1074 printk(KERN_WARNING "WARNING: maxcpus limit of %i reached."
1075 " Processor ignored.\n", maxcpus);
1076 return;
1077 }
1078
1079 num_processors++;
1080 cpus_complement(tmp_map, cpu_present_map);
1081 cpu = first_cpu(tmp_map);
1082
1083 physid_set(apicid, phys_cpu_present_map);
1084 if (apicid == boot_cpu_physical_apicid) {
1085 /*
1086 * x86_bios_cpu_apicid is required to have processors listed
1087 * in same order as logical cpu numbers. Hence the first
1088 * entry is BSP, and so on.
1089 */
1090 cpu = 0;
1091 }
1092 /* are we being called early in kernel startup? */
1093 if (x86_cpu_to_apicid_early_ptr) {
1094 u16 *cpu_to_apicid = x86_cpu_to_apicid_early_ptr;
1095 u16 *bios_cpu_apicid = x86_bios_cpu_apicid_early_ptr;
1096
1097 cpu_to_apicid[cpu] = apicid;
1098 bios_cpu_apicid[cpu] = apicid;
1099 } else {
1100 per_cpu(x86_cpu_to_apicid, cpu) = apicid;
1101 per_cpu(x86_bios_cpu_apicid, cpu) = apicid;
1102 }
1103
1104 cpu_set(cpu, cpu_possible_map);
1105 cpu_set(cpu, cpu_present_map);
1106}
1107
1032/* 1108/*
1033 * Power management 1109 * Power management
1034 */ 1110 */
@@ -1065,7 +1141,7 @@ static int lapic_suspend(struct sys_device *dev, pm_message_t state)
1065 1141
1066 maxlvt = lapic_get_maxlvt(); 1142 maxlvt = lapic_get_maxlvt();
1067 1143
1068 apic_pm_state.apic_id = apic_read(APIC_ID); 1144 apic_pm_state.apic_id = read_apic_id();
1069 apic_pm_state.apic_taskpri = apic_read(APIC_TASKPRI); 1145 apic_pm_state.apic_taskpri = apic_read(APIC_TASKPRI);
1070 apic_pm_state.apic_ldr = apic_read(APIC_LDR); 1146 apic_pm_state.apic_ldr = apic_read(APIC_LDR);
1071 apic_pm_state.apic_dfr = apic_read(APIC_DFR); 1147 apic_pm_state.apic_dfr = apic_read(APIC_DFR);
@@ -1180,9 +1256,19 @@ __cpuinit int apic_is_clustered_box(void)
1180{ 1256{
1181 int i, clusters, zeros; 1257 int i, clusters, zeros;
1182 unsigned id; 1258 unsigned id;
1183 u16 *bios_cpu_apicid = x86_bios_cpu_apicid_early_ptr; 1259 u16 *bios_cpu_apicid;
1184 DECLARE_BITMAP(clustermap, NUM_APIC_CLUSTERS); 1260 DECLARE_BITMAP(clustermap, NUM_APIC_CLUSTERS);
1185 1261
1262 /*
1263 * there is not this kind of box with AMD CPU yet.
1264 * Some AMD box with quadcore cpu and 8 sockets apicid
1265 * will be [4, 0x23] or [8, 0x27] could be thought to
1266 * vsmp box still need checking...
1267 */
1268 if ((boot_cpu_data.x86_vendor == X86_VENDOR_AMD) && !is_vsmp_box())
1269 return 0;
1270
1271 bios_cpu_apicid = x86_bios_cpu_apicid_early_ptr;
1186 bitmap_zero(clustermap, NUM_APIC_CLUSTERS); 1272 bitmap_zero(clustermap, NUM_APIC_CLUSTERS);
1187 1273
1188 for (i = 0; i < NR_CPUS; i++) { 1274 for (i = 0; i < NR_CPUS; i++) {
@@ -1219,6 +1305,12 @@ __cpuinit int apic_is_clustered_box(void)
1219 ++zeros; 1305 ++zeros;
1220 } 1306 }
1221 1307
1308 /* ScaleMP vSMPowered boxes have one cluster per board and TSCs are
1309 * not guaranteed to be synced between boards
1310 */
1311 if (is_vsmp_box() && clusters > 1)
1312 return 1;
1313
1222 /* 1314 /*
1223 * If clusters > 2, then should be multi-chassis. 1315 * If clusters > 2, then should be multi-chassis.
1224 * May have to revisit this when multi-core + hyperthreaded CPUs come 1316 * May have to revisit this when multi-core + hyperthreaded CPUs come
@@ -1290,3 +1382,21 @@ static __init int setup_apicpmtimer(char *s)
1290} 1382}
1291__setup("apicpmtimer", setup_apicpmtimer); 1383__setup("apicpmtimer", setup_apicpmtimer);
1292 1384
1385static int __init lapic_insert_resource(void)
1386{
1387 if (!apic_phys)
1388 return -1;
1389
1390 /* Put local APIC into the resource map. */
1391 lapic_resource.start = apic_phys;
1392 lapic_resource.end = lapic_resource.start + PAGE_SIZE - 1;
1393 insert_resource(&iomem_resource, &lapic_resource);
1394
1395 return 0;
1396}
1397
1398/*
1399 * need call insert after e820_reserve_resources()
1400 * that is using request_resource
1401 */
1402late_initcall(lapic_insert_resource);
diff --git a/arch/x86/kernel/apm_32.c b/arch/x86/kernel/apm_32.c
index d4438ef296d8..f0030a0999c7 100644
--- a/arch/x86/kernel/apm_32.c
+++ b/arch/x86/kernel/apm_32.c
@@ -2217,7 +2217,6 @@ static struct dmi_system_id __initdata apm_dmi_table[] = {
2217 */ 2217 */
2218static int __init apm_init(void) 2218static int __init apm_init(void)
2219{ 2219{
2220 struct proc_dir_entry *apm_proc;
2221 struct desc_struct *gdt; 2220 struct desc_struct *gdt;
2222 int err; 2221 int err;
2223 2222
@@ -2322,9 +2321,7 @@ static int __init apm_init(void)
2322 set_base(gdt[APM_DS >> 3], 2321 set_base(gdt[APM_DS >> 3],
2323 __va((unsigned long)apm_info.bios.dseg << 4)); 2322 __va((unsigned long)apm_info.bios.dseg << 4));
2324 2323
2325 apm_proc = create_proc_entry("apm", 0, NULL); 2324 proc_create("apm", 0, NULL, &apm_file_ops);
2326 if (apm_proc)
2327 apm_proc->proc_fops = &apm_file_ops;
2328 2325
2329 kapmd_task = kthread_create(apm, NULL, "kapmd"); 2326 kapmd_task = kthread_create(apm, NULL, "kapmd");
2330 if (IS_ERR(kapmd_task)) { 2327 if (IS_ERR(kapmd_task)) {
diff --git a/arch/x86/kernel/asm-offsets_32.c b/arch/x86/kernel/asm-offsets_32.c
index 8ea040124f7d..670c3c311289 100644
--- a/arch/x86/kernel/asm-offsets_32.c
+++ b/arch/x86/kernel/asm-offsets_32.c
@@ -10,7 +10,7 @@
10#include <linux/personality.h> 10#include <linux/personality.h>
11#include <linux/suspend.h> 11#include <linux/suspend.h>
12#include <asm/ucontext.h> 12#include <asm/ucontext.h>
13#include "sigframe_32.h" 13#include "sigframe.h"
14#include <asm/pgtable.h> 14#include <asm/pgtable.h>
15#include <asm/fixmap.h> 15#include <asm/fixmap.h>
16#include <asm/processor.h> 16#include <asm/processor.h>
diff --git a/arch/x86/kernel/bugs_64.c b/arch/x86/kernel/bugs_64.c
index 8f520f93ffd4..9a3ed0649d4e 100644
--- a/arch/x86/kernel/bugs_64.c
+++ b/arch/x86/kernel/bugs_64.c
@@ -9,13 +9,25 @@
9#include <asm/bugs.h> 9#include <asm/bugs.h>
10#include <asm/processor.h> 10#include <asm/processor.h>
11#include <asm/mtrr.h> 11#include <asm/mtrr.h>
12#include <asm/cacheflush.h>
12 13
13void __init check_bugs(void) 14void __init check_bugs(void)
14{ 15{
15 identify_cpu(&boot_cpu_data); 16 identify_boot_cpu();
16#if !defined(CONFIG_SMP) 17#if !defined(CONFIG_SMP)
17 printk("CPU: "); 18 printk("CPU: ");
18 print_cpu_info(&boot_cpu_data); 19 print_cpu_info(&boot_cpu_data);
19#endif 20#endif
20 alternative_instructions(); 21 alternative_instructions();
22
23 /*
24 * Make sure the first 2MB area is not mapped by huge pages
25 * There are typically fixed size MTRRs in there and overlapping
26 * MTRRs into large pages causes slow downs.
27 *
28 * Right now we don't do that with gbpages because there seems
29 * very little benefit for that case.
30 */
31 if (!direct_gbpages)
32 set_memory_4k((unsigned long)__va(0), 1);
21} 33}
diff --git a/arch/x86/kernel/cpu/Makefile b/arch/x86/kernel/cpu/Makefile
index a0c4d7c5dbd7..ee7c45235e54 100644
--- a/arch/x86/kernel/cpu/Makefile
+++ b/arch/x86/kernel/cpu/Makefile
@@ -3,9 +3,9 @@
3# 3#
4 4
5obj-y := intel_cacheinfo.o addon_cpuid_features.o 5obj-y := intel_cacheinfo.o addon_cpuid_features.o
6obj-y += feature_names.o 6obj-y += proc.o feature_names.o
7 7
8obj-$(CONFIG_X86_32) += common.o proc.o bugs.o 8obj-$(CONFIG_X86_32) += common.o bugs.o
9obj-$(CONFIG_X86_32) += amd.o 9obj-$(CONFIG_X86_32) += amd.o
10obj-$(CONFIG_X86_32) += cyrix.o 10obj-$(CONFIG_X86_32) += cyrix.o
11obj-$(CONFIG_X86_32) += centaur.o 11obj-$(CONFIG_X86_32) += centaur.o
diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c
index 693e353999cd..0173065dc3b7 100644
--- a/arch/x86/kernel/cpu/amd.c
+++ b/arch/x86/kernel/cpu/amd.c
@@ -4,8 +4,8 @@
4#include <asm/io.h> 4#include <asm/io.h>
5#include <asm/processor.h> 5#include <asm/processor.h>
6#include <asm/apic.h> 6#include <asm/apic.h>
7#include <asm/mach_apic.h>
8 7
8#include <mach_apic.h>
9#include "cpu.h" 9#include "cpu.h"
10 10
11/* 11/*
@@ -20,7 +20,7 @@
20 * the chip setting when fixing the bug but they also tweaked some 20 * the chip setting when fixing the bug but they also tweaked some
21 * performance at the same time.. 21 * performance at the same time..
22 */ 22 */
23 23
24extern void vide(void); 24extern void vide(void);
25__asm__(".align 4\nvide: ret"); 25__asm__(".align 4\nvide: ret");
26 26
@@ -63,12 +63,12 @@ static __cpuinit int amd_apic_timer_broken(void)
63 63
64int force_mwait __cpuinitdata; 64int force_mwait __cpuinitdata;
65 65
66void __cpuinit early_init_amd(struct cpuinfo_x86 *c) 66static void __cpuinit early_init_amd(struct cpuinfo_x86 *c)
67{ 67{
68 if (cpuid_eax(0x80000000) >= 0x80000007) { 68 if (cpuid_eax(0x80000000) >= 0x80000007) {
69 c->x86_power = cpuid_edx(0x80000007); 69 c->x86_power = cpuid_edx(0x80000007);
70 if (c->x86_power & (1<<8)) 70 if (c->x86_power & (1<<8))
71 set_bit(X86_FEATURE_CONSTANT_TSC, c->x86_capability); 71 set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
72 } 72 }
73} 73}
74 74
@@ -81,7 +81,8 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c)
81#ifdef CONFIG_SMP 81#ifdef CONFIG_SMP
82 unsigned long long value; 82 unsigned long long value;
83 83
84 /* Disable TLB flush filter by setting HWCR.FFDIS on K8 84 /*
85 * Disable TLB flush filter by setting HWCR.FFDIS on K8
85 * bit 6 of msr C001_0015 86 * bit 6 of msr C001_0015
86 * 87 *
87 * Errata 63 for SH-B3 steppings 88 * Errata 63 for SH-B3 steppings
@@ -102,15 +103,16 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c)
102 * no bus pipeline) 103 * no bus pipeline)
103 */ 104 */
104 105
105 /* Bit 31 in normal CPUID used for nonstandard 3DNow ID; 106 /*
106 3DNow is IDd by bit 31 in extended CPUID (1*32+31) anyway */ 107 * Bit 31 in normal CPUID used for nonstandard 3DNow ID;
107 clear_bit(0*32+31, c->x86_capability); 108 * 3DNow is IDd by bit 31 in extended CPUID (1*32+31) anyway
108 109 */
110 clear_cpu_cap(c, 0*32+31);
111
109 r = get_model_name(c); 112 r = get_model_name(c);
110 113
111 switch(c->x86) 114 switch (c->x86) {
112 { 115 case 4:
113 case 4:
114 /* 116 /*
115 * General Systems BIOSen alias the cpu frequency registers 117 * General Systems BIOSen alias the cpu frequency registers
116 * of the Elan at 0x000df000. Unfortuantly, one of the Linux 118 * of the Elan at 0x000df000. Unfortuantly, one of the Linux
@@ -120,61 +122,60 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c)
120#define CBAR (0xfffc) /* Configuration Base Address (32-bit) */ 122#define CBAR (0xfffc) /* Configuration Base Address (32-bit) */
121#define CBAR_ENB (0x80000000) 123#define CBAR_ENB (0x80000000)
122#define CBAR_KEY (0X000000CB) 124#define CBAR_KEY (0X000000CB)
123 if (c->x86_model==9 || c->x86_model == 10) { 125 if (c->x86_model == 9 || c->x86_model == 10) {
124 if (inl (CBAR) & CBAR_ENB) 126 if (inl (CBAR) & CBAR_ENB)
125 outl (0 | CBAR_KEY, CBAR); 127 outl (0 | CBAR_KEY, CBAR);
126 } 128 }
127 break; 129 break;
128 case 5: 130 case 5:
129 if( c->x86_model < 6 ) 131 if (c->x86_model < 6) {
130 {
131 /* Based on AMD doc 20734R - June 2000 */ 132 /* Based on AMD doc 20734R - June 2000 */
132 if ( c->x86_model == 0 ) { 133 if (c->x86_model == 0) {
133 clear_bit(X86_FEATURE_APIC, c->x86_capability); 134 clear_cpu_cap(c, X86_FEATURE_APIC);
134 set_bit(X86_FEATURE_PGE, c->x86_capability); 135 set_cpu_cap(c, X86_FEATURE_PGE);
135 } 136 }
136 break; 137 break;
137 } 138 }
138 139
139 if ( c->x86_model == 6 && c->x86_mask == 1 ) { 140 if (c->x86_model == 6 && c->x86_mask == 1) {
140 const int K6_BUG_LOOP = 1000000; 141 const int K6_BUG_LOOP = 1000000;
141 int n; 142 int n;
142 void (*f_vide)(void); 143 void (*f_vide)(void);
143 unsigned long d, d2; 144 unsigned long d, d2;
144 145
145 printk(KERN_INFO "AMD K6 stepping B detected - "); 146 printk(KERN_INFO "AMD K6 stepping B detected - ");
146 147
147 /* 148 /*
148 * It looks like AMD fixed the 2.6.2 bug and improved indirect 149 * It looks like AMD fixed the 2.6.2 bug and improved indirect
149 * calls at the same time. 150 * calls at the same time.
150 */ 151 */
151 152
152 n = K6_BUG_LOOP; 153 n = K6_BUG_LOOP;
153 f_vide = vide; 154 f_vide = vide;
154 rdtscl(d); 155 rdtscl(d);
155 while (n--) 156 while (n--)
156 f_vide(); 157 f_vide();
157 rdtscl(d2); 158 rdtscl(d2);
158 d = d2-d; 159 d = d2-d;
159 160
160 if (d > 20*K6_BUG_LOOP) 161 if (d > 20*K6_BUG_LOOP)
161 printk("system stability may be impaired when more than 32 MB are used.\n"); 162 printk("system stability may be impaired when more than 32 MB are used.\n");
162 else 163 else
163 printk("probably OK (after B9730xxxx).\n"); 164 printk("probably OK (after B9730xxxx).\n");
164 printk(KERN_INFO "Please see http://membres.lycos.fr/poulot/k6bug.html\n"); 165 printk(KERN_INFO "Please see http://membres.lycos.fr/poulot/k6bug.html\n");
165 } 166 }
166 167
167 /* K6 with old style WHCR */ 168 /* K6 with old style WHCR */
168 if (c->x86_model < 8 || 169 if (c->x86_model < 8 ||
169 (c->x86_model== 8 && c->x86_mask < 8)) { 170 (c->x86_model == 8 && c->x86_mask < 8)) {
170 /* We can only write allocate on the low 508Mb */ 171 /* We can only write allocate on the low 508Mb */
171 if(mbytes>508) 172 if (mbytes > 508)
172 mbytes=508; 173 mbytes = 508;
173 174
174 rdmsr(MSR_K6_WHCR, l, h); 175 rdmsr(MSR_K6_WHCR, l, h);
175 if ((l&0x0000FFFF)==0) { 176 if ((l&0x0000FFFF) == 0) {
176 unsigned long flags; 177 unsigned long flags;
177 l=(1<<0)|((mbytes/4)<<1); 178 l = (1<<0)|((mbytes/4)<<1);
178 local_irq_save(flags); 179 local_irq_save(flags);
179 wbinvd(); 180 wbinvd();
180 wrmsr(MSR_K6_WHCR, l, h); 181 wrmsr(MSR_K6_WHCR, l, h);
@@ -185,17 +186,17 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c)
185 break; 186 break;
186 } 187 }
187 188
188 if ((c->x86_model == 8 && c->x86_mask >7) || 189 if ((c->x86_model == 8 && c->x86_mask > 7) ||
189 c->x86_model == 9 || c->x86_model == 13) { 190 c->x86_model == 9 || c->x86_model == 13) {
190 /* The more serious chips .. */ 191 /* The more serious chips .. */
191 192
192 if(mbytes>4092) 193 if (mbytes > 4092)
193 mbytes=4092; 194 mbytes = 4092;
194 195
195 rdmsr(MSR_K6_WHCR, l, h); 196 rdmsr(MSR_K6_WHCR, l, h);
196 if ((l&0xFFFF0000)==0) { 197 if ((l&0xFFFF0000) == 0) {
197 unsigned long flags; 198 unsigned long flags;
198 l=((mbytes>>2)<<22)|(1<<16); 199 l = ((mbytes>>2)<<22)|(1<<16);
199 local_irq_save(flags); 200 local_irq_save(flags);
200 wbinvd(); 201 wbinvd();
201 wrmsr(MSR_K6_WHCR, l, h); 202 wrmsr(MSR_K6_WHCR, l, h);
@@ -207,7 +208,7 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c)
207 /* Set MTRR capability flag if appropriate */ 208 /* Set MTRR capability flag if appropriate */
208 if (c->x86_model == 13 || c->x86_model == 9 || 209 if (c->x86_model == 13 || c->x86_model == 9 ||
209 (c->x86_model == 8 && c->x86_mask >= 8)) 210 (c->x86_model == 8 && c->x86_mask >= 8))
210 set_bit(X86_FEATURE_K6_MTRR, c->x86_capability); 211 set_cpu_cap(c, X86_FEATURE_K6_MTRR);
211 break; 212 break;
212 } 213 }
213 214
@@ -217,10 +218,11 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c)
217 break; 218 break;
218 } 219 }
219 break; 220 break;
220 case 6: /* An Athlon/Duron */ 221 case 6: /* An Athlon/Duron */
221 222
222 /* Bit 15 of Athlon specific MSR 15, needs to be 0 223 /*
223 * to enable SSE on Palomino/Morgan/Barton CPU's. 224 * Bit 15 of Athlon specific MSR 15, needs to be 0
225 * to enable SSE on Palomino/Morgan/Barton CPU's.
224 * If the BIOS didn't enable it already, enable it here. 226 * If the BIOS didn't enable it already, enable it here.
225 */ 227 */
226 if (c->x86_model >= 6 && c->x86_model <= 10) { 228 if (c->x86_model >= 6 && c->x86_model <= 10) {
@@ -229,15 +231,16 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c)
229 rdmsr(MSR_K7_HWCR, l, h); 231 rdmsr(MSR_K7_HWCR, l, h);
230 l &= ~0x00008000; 232 l &= ~0x00008000;
231 wrmsr(MSR_K7_HWCR, l, h); 233 wrmsr(MSR_K7_HWCR, l, h);
232 set_bit(X86_FEATURE_XMM, c->x86_capability); 234 set_cpu_cap(c, X86_FEATURE_XMM);
233 } 235 }
234 } 236 }
235 237
236 /* It's been determined by AMD that Athlons since model 8 stepping 1 238 /*
239 * It's been determined by AMD that Athlons since model 8 stepping 1
237 * are more robust with CLK_CTL set to 200xxxxx instead of 600xxxxx 240 * are more robust with CLK_CTL set to 200xxxxx instead of 600xxxxx
238 * As per AMD technical note 27212 0.2 241 * As per AMD technical note 27212 0.2
239 */ 242 */
240 if ((c->x86_model == 8 && c->x86_mask>=1) || (c->x86_model > 8)) { 243 if ((c->x86_model == 8 && c->x86_mask >= 1) || (c->x86_model > 8)) {
241 rdmsr(MSR_K7_CLK_CTL, l, h); 244 rdmsr(MSR_K7_CLK_CTL, l, h);
242 if ((l & 0xfff00000) != 0x20000000) { 245 if ((l & 0xfff00000) != 0x20000000) {
243 printk ("CPU: CLK_CTL MSR was %x. Reprogramming to %x\n", l, 246 printk ("CPU: CLK_CTL MSR was %x. Reprogramming to %x\n", l,
@@ -253,20 +256,19 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c)
253 /* Use K8 tuning for Fam10h and Fam11h */ 256 /* Use K8 tuning for Fam10h and Fam11h */
254 case 0x10: 257 case 0x10:
255 case 0x11: 258 case 0x11:
256 set_bit(X86_FEATURE_K8, c->x86_capability); 259 set_cpu_cap(c, X86_FEATURE_K8);
257 break; 260 break;
258 case 6: 261 case 6:
259 set_bit(X86_FEATURE_K7, c->x86_capability); 262 set_cpu_cap(c, X86_FEATURE_K7);
260 break; 263 break;
261 } 264 }
262 if (c->x86 >= 6) 265 if (c->x86 >= 6)
263 set_bit(X86_FEATURE_FXSAVE_LEAK, c->x86_capability); 266 set_cpu_cap(c, X86_FEATURE_FXSAVE_LEAK);
264 267
265 display_cacheinfo(c); 268 display_cacheinfo(c);
266 269
267 if (cpuid_eax(0x80000000) >= 0x80000008) { 270 if (cpuid_eax(0x80000000) >= 0x80000008)
268 c->x86_max_cores = (cpuid_ecx(0x80000008) & 0xff) + 1; 271 c->x86_max_cores = (cpuid_ecx(0x80000008) & 0xff) + 1;
269 }
270 272
271#ifdef CONFIG_X86_HT 273#ifdef CONFIG_X86_HT
272 /* 274 /*
@@ -302,20 +304,20 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c)
302 304
303 /* K6s reports MCEs but don't actually have all the MSRs */ 305 /* K6s reports MCEs but don't actually have all the MSRs */
304 if (c->x86 < 6) 306 if (c->x86 < 6)
305 clear_bit(X86_FEATURE_MCE, c->x86_capability); 307 clear_cpu_cap(c, X86_FEATURE_MCE);
306 308
307 if (cpu_has_xmm2) 309 if (cpu_has_xmm2)
308 set_bit(X86_FEATURE_MFENCE_RDTSC, c->x86_capability); 310 set_cpu_cap(c, X86_FEATURE_MFENCE_RDTSC);
309} 311}
310 312
311static unsigned int __cpuinit amd_size_cache(struct cpuinfo_x86 * c, unsigned int size) 313static unsigned int __cpuinit amd_size_cache(struct cpuinfo_x86 *c, unsigned int size)
312{ 314{
313 /* AMD errata T13 (order #21922) */ 315 /* AMD errata T13 (order #21922) */
314 if ((c->x86 == 6)) { 316 if ((c->x86 == 6)) {
315 if (c->x86_model == 3 && c->x86_mask == 0) /* Duron Rev A0 */ 317 if (c->x86_model == 3 && c->x86_mask == 0) /* Duron Rev A0 */
316 size = 64; 318 size = 64;
317 if (c->x86_model == 4 && 319 if (c->x86_model == 4 &&
318 (c->x86_mask==0 || c->x86_mask==1)) /* Tbird rev A1/A2 */ 320 (c->x86_mask == 0 || c->x86_mask == 1)) /* Tbird rev A1/A2 */
319 size = 256; 321 size = 256;
320 } 322 }
321 return size; 323 return size;
@@ -323,19 +325,20 @@ static unsigned int __cpuinit amd_size_cache(struct cpuinfo_x86 * c, unsigned in
323 325
324static struct cpu_dev amd_cpu_dev __cpuinitdata = { 326static struct cpu_dev amd_cpu_dev __cpuinitdata = {
325 .c_vendor = "AMD", 327 .c_vendor = "AMD",
326 .c_ident = { "AuthenticAMD" }, 328 .c_ident = { "AuthenticAMD" },
327 .c_models = { 329 .c_models = {
328 { .vendor = X86_VENDOR_AMD, .family = 4, .model_names = 330 { .vendor = X86_VENDOR_AMD, .family = 4, .model_names =
329 { 331 {
330 [3] = "486 DX/2", 332 [3] = "486 DX/2",
331 [7] = "486 DX/2-WB", 333 [7] = "486 DX/2-WB",
332 [8] = "486 DX/4", 334 [8] = "486 DX/4",
333 [9] = "486 DX/4-WB", 335 [9] = "486 DX/4-WB",
334 [14] = "Am5x86-WT", 336 [14] = "Am5x86-WT",
335 [15] = "Am5x86-WB" 337 [15] = "Am5x86-WB"
336 } 338 }
337 }, 339 },
338 }, 340 },
341 .c_early_init = early_init_amd,
339 .c_init = init_amd, 342 .c_init = init_amd,
340 .c_size_cache = amd_size_cache, 343 .c_size_cache = amd_size_cache,
341}; 344};
@@ -345,3 +348,5 @@ int __init amd_init_cpu(void)
345 cpu_devs[X86_VENDOR_AMD] = &amd_cpu_dev; 348 cpu_devs[X86_VENDOR_AMD] = &amd_cpu_dev;
346 return 0; 349 return 0;
347} 350}
351
352cpu_vendor_dev_register(X86_VENDOR_AMD, &amd_cpu_dev);
diff --git a/arch/x86/kernel/cpu/centaur.c b/arch/x86/kernel/cpu/centaur.c
index 9681fa15ddf0..e0f45edd6a55 100644
--- a/arch/x86/kernel/cpu/centaur.c
+++ b/arch/x86/kernel/cpu/centaur.c
@@ -1,31 +1,34 @@
1#include <linux/kernel.h> 1#include <linux/kernel.h>
2#include <linux/init.h> 2#include <linux/init.h>
3#include <linux/bitops.h> 3#include <linux/bitops.h>
4
4#include <asm/processor.h> 5#include <asm/processor.h>
5#include <asm/msr.h> 6#include <asm/msr.h>
6#include <asm/e820.h> 7#include <asm/e820.h>
7#include <asm/mtrr.h> 8#include <asm/mtrr.h>
9
8#include "cpu.h" 10#include "cpu.h"
9 11
10#ifdef CONFIG_X86_OOSTORE 12#ifdef CONFIG_X86_OOSTORE
11 13
12static u32 __cpuinit power2(u32 x) 14static u32 __cpuinit power2(u32 x)
13{ 15{
14 u32 s=1; 16 u32 s = 1;
15 while(s<=x) 17
16 s<<=1; 18 while (s <= x)
17 return s>>=1; 19 s <<= 1;
20
21 return s >>= 1;
18} 22}
19 23
20 24
21/* 25/*
22 * Set up an actual MCR 26 * Set up an actual MCR
23 */ 27 */
24
25static void __cpuinit centaur_mcr_insert(int reg, u32 base, u32 size, int key) 28static void __cpuinit centaur_mcr_insert(int reg, u32 base, u32 size, int key)
26{ 29{
27 u32 lo, hi; 30 u32 lo, hi;
28 31
29 hi = base & ~0xFFF; 32 hi = base & ~0xFFF;
30 lo = ~(size-1); /* Size is a power of 2 so this makes a mask */ 33 lo = ~(size-1); /* Size is a power of 2 so this makes a mask */
31 lo &= ~0xFFF; /* Remove the ctrl value bits */ 34 lo &= ~0xFFF; /* Remove the ctrl value bits */
@@ -35,30 +38,28 @@ static void __cpuinit centaur_mcr_insert(int reg, u32 base, u32 size, int key)
35} 38}
36 39
37/* 40/*
38 * Figure what we can cover with MCR's 41 * Figure what we can cover with MCR's
39 * 42 *
40 * Shortcut: We know you can't put 4Gig of RAM on a winchip 43 * Shortcut: We know you can't put 4Gig of RAM on a winchip
41 */ 44 */
42 45static u32 __cpuinit ramtop(void)
43static u32 __cpuinit ramtop(void) /* 16388 */
44{ 46{
45 int i;
46 u32 top = 0;
47 u32 clip = 0xFFFFFFFFUL; 47 u32 clip = 0xFFFFFFFFUL;
48 48 u32 top = 0;
49 int i;
50
49 for (i = 0; i < e820.nr_map; i++) { 51 for (i = 0; i < e820.nr_map; i++) {
50 unsigned long start, end; 52 unsigned long start, end;
51 53
52 if (e820.map[i].addr > 0xFFFFFFFFUL) 54 if (e820.map[i].addr > 0xFFFFFFFFUL)
53 continue; 55 continue;
54 /* 56 /*
55 * Don't MCR over reserved space. Ignore the ISA hole 57 * Don't MCR over reserved space. Ignore the ISA hole
56 * we frob around that catastrophe already 58 * we frob around that catastrophe already
57 */ 59 */
58 60 if (e820.map[i].type == E820_RESERVED) {
59 if (e820.map[i].type == E820_RESERVED) 61 if (e820.map[i].addr >= 0x100000UL &&
60 { 62 e820.map[i].addr < clip)
61 if(e820.map[i].addr >= 0x100000UL && e820.map[i].addr < clip)
62 clip = e820.map[i].addr; 63 clip = e820.map[i].addr;
63 continue; 64 continue;
64 } 65 }
@@ -69,28 +70,27 @@ static u32 __cpuinit ramtop(void) /* 16388 */
69 if (end > top) 70 if (end > top)
70 top = end; 71 top = end;
71 } 72 }
72 /* Everything below 'top' should be RAM except for the ISA hole. 73 /*
73 Because of the limited MCR's we want to map NV/ACPI into our 74 * Everything below 'top' should be RAM except for the ISA hole.
74 MCR range for gunk in RAM 75 * Because of the limited MCR's we want to map NV/ACPI into our
75 76 * MCR range for gunk in RAM
76 Clip might cause us to MCR insufficient RAM but that is an 77 *
77 acceptable failure mode and should only bite obscure boxes with 78 * Clip might cause us to MCR insufficient RAM but that is an
78 a VESA hole at 15Mb 79 * acceptable failure mode and should only bite obscure boxes with
79 80 * a VESA hole at 15Mb
80 The second case Clip sometimes kicks in is when the EBDA is marked 81 *
81 as reserved. Again we fail safe with reasonable results 82 * The second case Clip sometimes kicks in is when the EBDA is marked
82 */ 83 * as reserved. Again we fail safe with reasonable results
83 84 */
84 if(top>clip) 85 if (top > clip)
85 top=clip; 86 top = clip;
86 87
87 return top; 88 return top;
88} 89}
89 90
90/* 91/*
91 * Compute a set of MCR's to give maximum coverage 92 * Compute a set of MCR's to give maximum coverage
92 */ 93 */
93
94static int __cpuinit centaur_mcr_compute(int nr, int key) 94static int __cpuinit centaur_mcr_compute(int nr, int key)
95{ 95{
96 u32 mem = ramtop(); 96 u32 mem = ramtop();
@@ -99,141 +99,131 @@ static int __cpuinit centaur_mcr_compute(int nr, int key)
99 u32 top = root; 99 u32 top = root;
100 u32 floor = 0; 100 u32 floor = 0;
101 int ct = 0; 101 int ct = 0;
102 102
103 while(ct<nr) 103 while (ct < nr) {
104 {
105 u32 fspace = 0; 104 u32 fspace = 0;
105 u32 high;
106 u32 low;
106 107
107 /* 108 /*
108 * Find the largest block we will fill going upwards 109 * Find the largest block we will fill going upwards
109 */ 110 */
110 111 high = power2(mem-top);
111 u32 high = power2(mem-top);
112 112
113 /* 113 /*
114 * Find the largest block we will fill going downwards 114 * Find the largest block we will fill going downwards
115 */ 115 */
116 116 low = base/2;
117 u32 low = base/2;
118 117
119 /* 118 /*
120 * Don't fill below 1Mb going downwards as there 119 * Don't fill below 1Mb going downwards as there
121 * is an ISA hole in the way. 120 * is an ISA hole in the way.
122 */ 121 */
123 122 if (base <= 1024*1024)
124 if(base <= 1024*1024)
125 low = 0; 123 low = 0;
126 124
127 /* 125 /*
128 * See how much space we could cover by filling below 126 * See how much space we could cover by filling below
129 * the ISA hole 127 * the ISA hole
130 */ 128 */
131 129
132 if(floor == 0) 130 if (floor == 0)
133 fspace = 512*1024; 131 fspace = 512*1024;
134 else if(floor ==512*1024) 132 else if (floor == 512*1024)
135 fspace = 128*1024; 133 fspace = 128*1024;
136 134
137 /* And forget ROM space */ 135 /* And forget ROM space */
138 136
139 /* 137 /*
140 * Now install the largest coverage we get 138 * Now install the largest coverage we get
141 */ 139 */
142 140 if (fspace > high && fspace > low) {
143 if(fspace > high && fspace > low)
144 {
145 centaur_mcr_insert(ct, floor, fspace, key); 141 centaur_mcr_insert(ct, floor, fspace, key);
146 floor += fspace; 142 floor += fspace;
147 } 143 } else if (high > low) {
148 else if(high > low)
149 {
150 centaur_mcr_insert(ct, top, high, key); 144 centaur_mcr_insert(ct, top, high, key);
151 top += high; 145 top += high;
152 } 146 } else if (low > 0) {
153 else if(low > 0)
154 {
155 base -= low; 147 base -= low;
156 centaur_mcr_insert(ct, base, low, key); 148 centaur_mcr_insert(ct, base, low, key);
157 } 149 } else
158 else break; 150 break;
159 ct++; 151 ct++;
160 } 152 }
161 /* 153 /*
162 * We loaded ct values. We now need to set the mask. The caller 154 * We loaded ct values. We now need to set the mask. The caller
163 * must do this bit. 155 * must do this bit.
164 */ 156 */
165
166 return ct; 157 return ct;
167} 158}
168 159
169static void __cpuinit centaur_create_optimal_mcr(void) 160static void __cpuinit centaur_create_optimal_mcr(void)
170{ 161{
162 int used;
171 int i; 163 int i;
164
172 /* 165 /*
173 * Allocate up to 6 mcrs to mark as much of ram as possible 166 * Allocate up to 6 mcrs to mark as much of ram as possible
174 * as write combining and weak write ordered. 167 * as write combining and weak write ordered.
175 * 168 *
176 * To experiment with: Linux never uses stack operations for 169 * To experiment with: Linux never uses stack operations for
177 * mmio spaces so we could globally enable stack operation wc 170 * mmio spaces so we could globally enable stack operation wc
178 * 171 *
179 * Load the registers with type 31 - full write combining, all 172 * Load the registers with type 31 - full write combining, all
180 * writes weakly ordered. 173 * writes weakly ordered.
181 */ 174 */
182 int used = centaur_mcr_compute(6, 31); 175 used = centaur_mcr_compute(6, 31);
183 176
184 /* 177 /*
185 * Wipe unused MCRs 178 * Wipe unused MCRs
186 */ 179 */
187 180 for (i = used; i < 8; i++)
188 for(i=used;i<8;i++)
189 wrmsr(MSR_IDT_MCR0+i, 0, 0); 181 wrmsr(MSR_IDT_MCR0+i, 0, 0);
190} 182}
191 183
192static void __cpuinit winchip2_create_optimal_mcr(void) 184static void __cpuinit winchip2_create_optimal_mcr(void)
193{ 185{
194 u32 lo, hi; 186 u32 lo, hi;
187 int used;
195 int i; 188 int i;
196 189
197 /* 190 /*
198 * Allocate up to 6 mcrs to mark as much of ram as possible 191 * Allocate up to 6 mcrs to mark as much of ram as possible
199 * as write combining, weak store ordered. 192 * as write combining, weak store ordered.
200 * 193 *
201 * Load the registers with type 25 194 * Load the registers with type 25
202 * 8 - weak write ordering 195 * 8 - weak write ordering
203 * 16 - weak read ordering 196 * 16 - weak read ordering
204 * 1 - write combining 197 * 1 - write combining
205 */ 198 */
199 used = centaur_mcr_compute(6, 25);
206 200
207 int used = centaur_mcr_compute(6, 25);
208
209 /* 201 /*
210 * Mark the registers we are using. 202 * Mark the registers we are using.
211 */ 203 */
212
213 rdmsr(MSR_IDT_MCR_CTRL, lo, hi); 204 rdmsr(MSR_IDT_MCR_CTRL, lo, hi);
214 for(i=0;i<used;i++) 205 for (i = 0; i < used; i++)
215 lo|=1<<(9+i); 206 lo |= 1<<(9+i);
216 wrmsr(MSR_IDT_MCR_CTRL, lo, hi); 207 wrmsr(MSR_IDT_MCR_CTRL, lo, hi);
217 208
218 /* 209 /*
219 * Wipe unused MCRs 210 * Wipe unused MCRs
220 */ 211 */
221 212
222 for(i=used;i<8;i++) 213 for (i = used; i < 8; i++)
223 wrmsr(MSR_IDT_MCR0+i, 0, 0); 214 wrmsr(MSR_IDT_MCR0+i, 0, 0);
224} 215}
225 216
226/* 217/*
227 * Handle the MCR key on the Winchip 2. 218 * Handle the MCR key on the Winchip 2.
228 */ 219 */
229
230static void __cpuinit winchip2_unprotect_mcr(void) 220static void __cpuinit winchip2_unprotect_mcr(void)
231{ 221{
232 u32 lo, hi; 222 u32 lo, hi;
233 u32 key; 223 u32 key;
234 224
235 rdmsr(MSR_IDT_MCR_CTRL, lo, hi); 225 rdmsr(MSR_IDT_MCR_CTRL, lo, hi);
236 lo&=~0x1C0; /* blank bits 8-6 */ 226 lo &= ~0x1C0; /* blank bits 8-6 */
237 key = (lo>>17) & 7; 227 key = (lo>>17) & 7;
238 lo |= key<<6; /* replace with unlock key */ 228 lo |= key<<6; /* replace with unlock key */
239 wrmsr(MSR_IDT_MCR_CTRL, lo, hi); 229 wrmsr(MSR_IDT_MCR_CTRL, lo, hi);
@@ -242,9 +232,9 @@ static void __cpuinit winchip2_unprotect_mcr(void)
242static void __cpuinit winchip2_protect_mcr(void) 232static void __cpuinit winchip2_protect_mcr(void)
243{ 233{
244 u32 lo, hi; 234 u32 lo, hi;
245 235
246 rdmsr(MSR_IDT_MCR_CTRL, lo, hi); 236 rdmsr(MSR_IDT_MCR_CTRL, lo, hi);
247 lo&=~0x1C0; /* blank bits 8-6 */ 237 lo &= ~0x1C0; /* blank bits 8-6 */
248 wrmsr(MSR_IDT_MCR_CTRL, lo, hi); 238 wrmsr(MSR_IDT_MCR_CTRL, lo, hi);
249} 239}
250#endif /* CONFIG_X86_OOSTORE */ 240#endif /* CONFIG_X86_OOSTORE */
@@ -267,17 +257,17 @@ static void __cpuinit init_c3(struct cpuinfo_x86 *c)
267 257
268 /* enable ACE unit, if present and disabled */ 258 /* enable ACE unit, if present and disabled */
269 if ((tmp & (ACE_PRESENT | ACE_ENABLED)) == ACE_PRESENT) { 259 if ((tmp & (ACE_PRESENT | ACE_ENABLED)) == ACE_PRESENT) {
270 rdmsr (MSR_VIA_FCR, lo, hi); 260 rdmsr(MSR_VIA_FCR, lo, hi);
271 lo |= ACE_FCR; /* enable ACE unit */ 261 lo |= ACE_FCR; /* enable ACE unit */
272 wrmsr (MSR_VIA_FCR, lo, hi); 262 wrmsr(MSR_VIA_FCR, lo, hi);
273 printk(KERN_INFO "CPU: Enabled ACE h/w crypto\n"); 263 printk(KERN_INFO "CPU: Enabled ACE h/w crypto\n");
274 } 264 }
275 265
276 /* enable RNG unit, if present and disabled */ 266 /* enable RNG unit, if present and disabled */
277 if ((tmp & (RNG_PRESENT | RNG_ENABLED)) == RNG_PRESENT) { 267 if ((tmp & (RNG_PRESENT | RNG_ENABLED)) == RNG_PRESENT) {
278 rdmsr (MSR_VIA_RNG, lo, hi); 268 rdmsr(MSR_VIA_RNG, lo, hi);
279 lo |= RNG_ENABLE; /* enable RNG unit */ 269 lo |= RNG_ENABLE; /* enable RNG unit */
280 wrmsr (MSR_VIA_RNG, lo, hi); 270 wrmsr(MSR_VIA_RNG, lo, hi);
281 printk(KERN_INFO "CPU: Enabled h/w RNG\n"); 271 printk(KERN_INFO "CPU: Enabled h/w RNG\n");
282 } 272 }
283 273
@@ -288,171 +278,183 @@ static void __cpuinit init_c3(struct cpuinfo_x86 *c)
288 } 278 }
289 279
290 /* Cyrix III family needs CX8 & PGE explicitly enabled. */ 280 /* Cyrix III family needs CX8 & PGE explicitly enabled. */
291 if (c->x86_model >=6 && c->x86_model <= 9) { 281 if (c->x86_model >= 6 && c->x86_model <= 9) {
292 rdmsr (MSR_VIA_FCR, lo, hi); 282 rdmsr(MSR_VIA_FCR, lo, hi);
293 lo |= (1<<1 | 1<<7); 283 lo |= (1<<1 | 1<<7);
294 wrmsr (MSR_VIA_FCR, lo, hi); 284 wrmsr(MSR_VIA_FCR, lo, hi);
295 set_bit(X86_FEATURE_CX8, c->x86_capability); 285 set_cpu_cap(c, X86_FEATURE_CX8);
296 } 286 }
297 287
298 /* Before Nehemiah, the C3's had 3dNOW! */ 288 /* Before Nehemiah, the C3's had 3dNOW! */
299 if (c->x86_model >=6 && c->x86_model <9) 289 if (c->x86_model >= 6 && c->x86_model < 9)
300 set_bit(X86_FEATURE_3DNOW, c->x86_capability); 290 set_cpu_cap(c, X86_FEATURE_3DNOW);
301 291
302 get_model_name(c); 292 get_model_name(c);
303 display_cacheinfo(c); 293 display_cacheinfo(c);
304} 294}
305 295
296enum {
297 ECX8 = 1<<1,
298 EIERRINT = 1<<2,
299 DPM = 1<<3,
300 DMCE = 1<<4,
301 DSTPCLK = 1<<5,
302 ELINEAR = 1<<6,
303 DSMC = 1<<7,
304 DTLOCK = 1<<8,
305 EDCTLB = 1<<8,
306 EMMX = 1<<9,
307 DPDC = 1<<11,
308 EBRPRED = 1<<12,
309 DIC = 1<<13,
310 DDC = 1<<14,
311 DNA = 1<<15,
312 ERETSTK = 1<<16,
313 E2MMX = 1<<19,
314 EAMD3D = 1<<20,
315};
316
306static void __cpuinit init_centaur(struct cpuinfo_x86 *c) 317static void __cpuinit init_centaur(struct cpuinfo_x86 *c)
307{ 318{
308 enum {
309 ECX8=1<<1,
310 EIERRINT=1<<2,
311 DPM=1<<3,
312 DMCE=1<<4,
313 DSTPCLK=1<<5,
314 ELINEAR=1<<6,
315 DSMC=1<<7,
316 DTLOCK=1<<8,
317 EDCTLB=1<<8,
318 EMMX=1<<9,
319 DPDC=1<<11,
320 EBRPRED=1<<12,
321 DIC=1<<13,
322 DDC=1<<14,
323 DNA=1<<15,
324 ERETSTK=1<<16,
325 E2MMX=1<<19,
326 EAMD3D=1<<20,
327 };
328 319
329 char *name; 320 char *name;
330 u32 fcr_set=0; 321 u32 fcr_set = 0;
331 u32 fcr_clr=0; 322 u32 fcr_clr = 0;
332 u32 lo,hi,newlo; 323 u32 lo, hi, newlo;
333 u32 aa,bb,cc,dd; 324 u32 aa, bb, cc, dd;
334 325
335 /* Bit 31 in normal CPUID used for nonstandard 3DNow ID; 326 /*
336 3DNow is IDd by bit 31 in extended CPUID (1*32+31) anyway */ 327 * Bit 31 in normal CPUID used for nonstandard 3DNow ID;
337 clear_bit(0*32+31, c->x86_capability); 328 * 3DNow is IDd by bit 31 in extended CPUID (1*32+31) anyway
329 */
330 clear_cpu_cap(c, 0*32+31);
338 331
339 switch (c->x86) { 332 switch (c->x86) {
340 333 case 5:
341 case 5: 334 switch (c->x86_model) {
342 switch(c->x86_model) { 335 case 4:
343 case 4: 336 name = "C6";
344 name="C6"; 337 fcr_set = ECX8|DSMC|EDCTLB|EMMX|ERETSTK;
345 fcr_set=ECX8|DSMC|EDCTLB|EMMX|ERETSTK; 338 fcr_clr = DPDC;
346 fcr_clr=DPDC; 339 printk(KERN_NOTICE "Disabling bugged TSC.\n");
347 printk(KERN_NOTICE "Disabling bugged TSC.\n"); 340 clear_cpu_cap(c, X86_FEATURE_TSC);
348 clear_bit(X86_FEATURE_TSC, c->x86_capability);
349#ifdef CONFIG_X86_OOSTORE 341#ifdef CONFIG_X86_OOSTORE
350 centaur_create_optimal_mcr(); 342 centaur_create_optimal_mcr();
351 /* Enable 343 /*
352 write combining on non-stack, non-string 344 * Enable:
353 write combining on string, all types 345 * write combining on non-stack, non-string
354 weak write ordering 346 * write combining on string, all types
355 347 * weak write ordering
356 The C6 original lacks weak read order 348 *
357 349 * The C6 original lacks weak read order
358 Note 0x120 is write only on Winchip 1 */ 350 *
359 351 * Note 0x120 is write only on Winchip 1
360 wrmsr(MSR_IDT_MCR_CTRL, 0x01F0001F, 0); 352 */
361#endif 353 wrmsr(MSR_IDT_MCR_CTRL, 0x01F0001F, 0);
354#endif
355 break;
356 case 8:
357 switch (c->x86_mask) {
358 default:
359 name = "2";
360 break;
361 case 7 ... 9:
362 name = "2A";
362 break; 363 break;
363 case 8: 364 case 10 ... 15:
364 switch(c->x86_mask) { 365 name = "2B";
365 default: 366 break;
366 name="2"; 367 }
367 break; 368 fcr_set = ECX8|DSMC|DTLOCK|EMMX|EBRPRED|ERETSTK|
368 case 7 ... 9: 369 E2MMX|EAMD3D;
369 name="2A"; 370 fcr_clr = DPDC;
370 break;
371 case 10 ... 15:
372 name="2B";
373 break;
374 }
375 fcr_set=ECX8|DSMC|DTLOCK|EMMX|EBRPRED|ERETSTK|E2MMX|EAMD3D;
376 fcr_clr=DPDC;
377#ifdef CONFIG_X86_OOSTORE 371#ifdef CONFIG_X86_OOSTORE
378 winchip2_unprotect_mcr(); 372 winchip2_unprotect_mcr();
379 winchip2_create_optimal_mcr(); 373 winchip2_create_optimal_mcr();
380 rdmsr(MSR_IDT_MCR_CTRL, lo, hi); 374 rdmsr(MSR_IDT_MCR_CTRL, lo, hi);
381 /* Enable 375 /*
382 write combining on non-stack, non-string 376 * Enable:
383 write combining on string, all types 377 * write combining on non-stack, non-string
384 weak write ordering 378 * write combining on string, all types
385 */ 379 * weak write ordering
386 lo|=31; 380 */
387 wrmsr(MSR_IDT_MCR_CTRL, lo, hi); 381 lo |= 31;
388 winchip2_protect_mcr(); 382 wrmsr(MSR_IDT_MCR_CTRL, lo, hi);
383 winchip2_protect_mcr();
389#endif 384#endif
390 break; 385 break;
391 case 9: 386 case 9:
392 name="3"; 387 name = "3";
393 fcr_set=ECX8|DSMC|DTLOCK|EMMX|EBRPRED|ERETSTK|E2MMX|EAMD3D; 388 fcr_set = ECX8|DSMC|DTLOCK|EMMX|EBRPRED|ERETSTK|
394 fcr_clr=DPDC; 389 E2MMX|EAMD3D;
390 fcr_clr = DPDC;
395#ifdef CONFIG_X86_OOSTORE 391#ifdef CONFIG_X86_OOSTORE
396 winchip2_unprotect_mcr(); 392 winchip2_unprotect_mcr();
397 winchip2_create_optimal_mcr(); 393 winchip2_create_optimal_mcr();
398 rdmsr(MSR_IDT_MCR_CTRL, lo, hi); 394 rdmsr(MSR_IDT_MCR_CTRL, lo, hi);
399 /* Enable 395 /*
400 write combining on non-stack, non-string 396 * Enable:
401 write combining on string, all types 397 * write combining on non-stack, non-string
402 weak write ordering 398 * write combining on string, all types
403 */ 399 * weak write ordering
404 lo|=31; 400 */
405 wrmsr(MSR_IDT_MCR_CTRL, lo, hi); 401 lo |= 31;
406 winchip2_protect_mcr(); 402 wrmsr(MSR_IDT_MCR_CTRL, lo, hi);
403 winchip2_protect_mcr();
407#endif 404#endif
408 break; 405 break;
409 default: 406 default:
410 name="??"; 407 name = "??";
411 } 408 }
412 409
413 rdmsr(MSR_IDT_FCR1, lo, hi); 410 rdmsr(MSR_IDT_FCR1, lo, hi);
414 newlo=(lo|fcr_set) & (~fcr_clr); 411 newlo = (lo|fcr_set) & (~fcr_clr);
415 412
416 if (newlo!=lo) { 413 if (newlo != lo) {
417 printk(KERN_INFO "Centaur FCR was 0x%X now 0x%X\n", lo, newlo ); 414 printk(KERN_INFO "Centaur FCR was 0x%X now 0x%X\n",
418 wrmsr(MSR_IDT_FCR1, newlo, hi ); 415 lo, newlo);
419 } else { 416 wrmsr(MSR_IDT_FCR1, newlo, hi);
420 printk(KERN_INFO "Centaur FCR is 0x%X\n",lo); 417 } else {
421 } 418 printk(KERN_INFO "Centaur FCR is 0x%X\n", lo);
422 /* Emulate MTRRs using Centaur's MCR. */ 419 }
423 set_bit(X86_FEATURE_CENTAUR_MCR, c->x86_capability); 420 /* Emulate MTRRs using Centaur's MCR. */
424 /* Report CX8 */ 421 set_cpu_cap(c, X86_FEATURE_CENTAUR_MCR);
425 set_bit(X86_FEATURE_CX8, c->x86_capability); 422 /* Report CX8 */
426 /* Set 3DNow! on Winchip 2 and above. */ 423 set_cpu_cap(c, X86_FEATURE_CX8);
427 if (c->x86_model >=8) 424 /* Set 3DNow! on Winchip 2 and above. */
428 set_bit(X86_FEATURE_3DNOW, c->x86_capability); 425 if (c->x86_model >= 8)
429 /* See if we can find out some more. */ 426 set_cpu_cap(c, X86_FEATURE_3DNOW);
430 if ( cpuid_eax(0x80000000) >= 0x80000005 ) { 427 /* See if we can find out some more. */
431 /* Yes, we can. */ 428 if (cpuid_eax(0x80000000) >= 0x80000005) {
432 cpuid(0x80000005,&aa,&bb,&cc,&dd); 429 /* Yes, we can. */
433 /* Add L1 data and code cache sizes. */ 430 cpuid(0x80000005, &aa, &bb, &cc, &dd);
434 c->x86_cache_size = (cc>>24)+(dd>>24); 431 /* Add L1 data and code cache sizes. */
435 } 432 c->x86_cache_size = (cc>>24)+(dd>>24);
436 sprintf( c->x86_model_id, "WinChip %s", name ); 433 }
437 break; 434 sprintf(c->x86_model_id, "WinChip %s", name);
435 break;
438 436
439 case 6: 437 case 6:
440 init_c3(c); 438 init_c3(c);
441 break; 439 break;
442 } 440 }
443} 441}
444 442
445static unsigned int __cpuinit centaur_size_cache(struct cpuinfo_x86 * c, unsigned int size) 443static unsigned int __cpuinit
444centaur_size_cache(struct cpuinfo_x86 *c, unsigned int size)
446{ 445{
447 /* VIA C3 CPUs (670-68F) need further shifting. */ 446 /* VIA C3 CPUs (670-68F) need further shifting. */
448 if ((c->x86 == 6) && ((c->x86_model == 7) || (c->x86_model == 8))) 447 if ((c->x86 == 6) && ((c->x86_model == 7) || (c->x86_model == 8)))
449 size >>= 8; 448 size >>= 8;
450 449
451 /* VIA also screwed up Nehemiah stepping 1, and made 450 /*
452 it return '65KB' instead of '64KB' 451 * There's also an erratum in Nehemiah stepping 1, which
453 - Note, it seems this may only be in engineering samples. */ 452 * returns '65KB' instead of '64KB'
454 if ((c->x86==6) && (c->x86_model==9) && (c->x86_mask==1) && (size==65)) 453 * - Note, it seems this may only be in engineering samples.
455 size -=1; 454 */
455 if ((c->x86 == 6) && (c->x86_model == 9) &&
456 (c->x86_mask == 1) && (size == 65))
457 size -= 1;
456 458
457 return size; 459 return size;
458} 460}
@@ -464,8 +466,4 @@ static struct cpu_dev centaur_cpu_dev __cpuinitdata = {
464 .c_size_cache = centaur_size_cache, 466 .c_size_cache = centaur_size_cache,
465}; 467};
466 468
467int __init centaur_init_cpu(void) 469cpu_vendor_dev_register(X86_VENDOR_CENTAUR, &centaur_cpu_dev);
468{
469 cpu_devs[X86_VENDOR_CENTAUR] = &centaur_cpu_dev;
470 return 0;
471}
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index a38aafaefc23..d999d7833bc2 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -62,9 +62,9 @@ __u32 cleared_cpu_caps[NCAPINTS] __cpuinitdata;
62static int cachesize_override __cpuinitdata = -1; 62static int cachesize_override __cpuinitdata = -1;
63static int disable_x86_serial_nr __cpuinitdata = 1; 63static int disable_x86_serial_nr __cpuinitdata = 1;
64 64
65struct cpu_dev * cpu_devs[X86_VENDOR_NUM] = {}; 65struct cpu_dev *cpu_devs[X86_VENDOR_NUM] = {};
66 66
67static void __cpuinit default_init(struct cpuinfo_x86 * c) 67static void __cpuinit default_init(struct cpuinfo_x86 *c)
68{ 68{
69 /* Not much we can do here... */ 69 /* Not much we can do here... */
70 /* Check if at least it has cpuid */ 70 /* Check if at least it has cpuid */
@@ -81,11 +81,11 @@ static struct cpu_dev __cpuinitdata default_cpu = {
81 .c_init = default_init, 81 .c_init = default_init,
82 .c_vendor = "Unknown", 82 .c_vendor = "Unknown",
83}; 83};
84static struct cpu_dev * this_cpu __cpuinitdata = &default_cpu; 84static struct cpu_dev *this_cpu __cpuinitdata = &default_cpu;
85 85
86static int __init cachesize_setup(char *str) 86static int __init cachesize_setup(char *str)
87{ 87{
88 get_option (&str, &cachesize_override); 88 get_option(&str, &cachesize_override);
89 return 1; 89 return 1;
90} 90}
91__setup("cachesize=", cachesize_setup); 91__setup("cachesize=", cachesize_setup);
@@ -107,12 +107,12 @@ int __cpuinit get_model_name(struct cpuinfo_x86 *c)
107 /* Intel chips right-justify this string for some dumb reason; 107 /* Intel chips right-justify this string for some dumb reason;
108 undo that brain damage */ 108 undo that brain damage */
109 p = q = &c->x86_model_id[0]; 109 p = q = &c->x86_model_id[0];
110 while ( *p == ' ' ) 110 while (*p == ' ')
111 p++; 111 p++;
112 if ( p != q ) { 112 if (p != q) {
113 while ( *p ) 113 while (*p)
114 *q++ = *p++; 114 *q++ = *p++;
115 while ( q <= &c->x86_model_id[48] ) 115 while (q <= &c->x86_model_id[48])
116 *q++ = '\0'; /* Zero-pad the rest */ 116 *q++ = '\0'; /* Zero-pad the rest */
117 } 117 }
118 118
@@ -130,7 +130,7 @@ void __cpuinit display_cacheinfo(struct cpuinfo_x86 *c)
130 cpuid(0x80000005, &dummy, &dummy, &ecx, &edx); 130 cpuid(0x80000005, &dummy, &dummy, &ecx, &edx);
131 printk(KERN_INFO "CPU: L1 I Cache: %dK (%d bytes/line), D cache %dK (%d bytes/line)\n", 131 printk(KERN_INFO "CPU: L1 I Cache: %dK (%d bytes/line), D cache %dK (%d bytes/line)\n",
132 edx>>24, edx&0xFF, ecx>>24, ecx&0xFF); 132 edx>>24, edx&0xFF, ecx>>24, ecx&0xFF);
133 c->x86_cache_size=(ecx>>24)+(edx>>24); 133 c->x86_cache_size = (ecx>>24)+(edx>>24);
134 } 134 }
135 135
136 if (n < 0x80000006) /* Some chips just has a large L1. */ 136 if (n < 0x80000006) /* Some chips just has a large L1. */
@@ -138,16 +138,16 @@ void __cpuinit display_cacheinfo(struct cpuinfo_x86 *c)
138 138
139 ecx = cpuid_ecx(0x80000006); 139 ecx = cpuid_ecx(0x80000006);
140 l2size = ecx >> 16; 140 l2size = ecx >> 16;
141 141
142 /* do processor-specific cache resizing */ 142 /* do processor-specific cache resizing */
143 if (this_cpu->c_size_cache) 143 if (this_cpu->c_size_cache)
144 l2size = this_cpu->c_size_cache(c,l2size); 144 l2size = this_cpu->c_size_cache(c, l2size);
145 145
146 /* Allow user to override all this if necessary. */ 146 /* Allow user to override all this if necessary. */
147 if (cachesize_override != -1) 147 if (cachesize_override != -1)
148 l2size = cachesize_override; 148 l2size = cachesize_override;
149 149
150 if ( l2size == 0 ) 150 if (l2size == 0)
151 return; /* Again, no L2 cache is possible */ 151 return; /* Again, no L2 cache is possible */
152 152
153 c->x86_cache_size = l2size; 153 c->x86_cache_size = l2size;
@@ -156,16 +156,19 @@ void __cpuinit display_cacheinfo(struct cpuinfo_x86 *c)
156 l2size, ecx & 0xFF); 156 l2size, ecx & 0xFF);
157} 157}
158 158
159/* Naming convention should be: <Name> [(<Codename>)] */ 159/*
160/* This table only is used unless init_<vendor>() below doesn't set it; */ 160 * Naming convention should be: <Name> [(<Codename>)]
161/* in particular, if CPUID levels 0x80000002..4 are supported, this isn't used */ 161 * This table only is used unless init_<vendor>() below doesn't set it;
162 * in particular, if CPUID levels 0x80000002..4 are supported, this isn't used
163 *
164 */
162 165
163/* Look up CPU names by table lookup. */ 166/* Look up CPU names by table lookup. */
164static char __cpuinit *table_lookup_model(struct cpuinfo_x86 *c) 167static char __cpuinit *table_lookup_model(struct cpuinfo_x86 *c)
165{ 168{
166 struct cpu_model_info *info; 169 struct cpu_model_info *info;
167 170
168 if ( c->x86_model >= 16 ) 171 if (c->x86_model >= 16)
169 return NULL; /* Range check */ 172 return NULL; /* Range check */
170 173
171 if (!this_cpu) 174 if (!this_cpu)
@@ -190,9 +193,9 @@ static void __cpuinit get_cpu_vendor(struct cpuinfo_x86 *c, int early)
190 193
191 for (i = 0; i < X86_VENDOR_NUM; i++) { 194 for (i = 0; i < X86_VENDOR_NUM; i++) {
192 if (cpu_devs[i]) { 195 if (cpu_devs[i]) {
193 if (!strcmp(v,cpu_devs[i]->c_ident[0]) || 196 if (!strcmp(v, cpu_devs[i]->c_ident[0]) ||
194 (cpu_devs[i]->c_ident[1] && 197 (cpu_devs[i]->c_ident[1] &&
195 !strcmp(v,cpu_devs[i]->c_ident[1]))) { 198 !strcmp(v, cpu_devs[i]->c_ident[1]))) {
196 c->x86_vendor = i; 199 c->x86_vendor = i;
197 if (!early) 200 if (!early)
198 this_cpu = cpu_devs[i]; 201 this_cpu = cpu_devs[i];
@@ -210,7 +213,7 @@ static void __cpuinit get_cpu_vendor(struct cpuinfo_x86 *c, int early)
210} 213}
211 214
212 215
213static int __init x86_fxsr_setup(char * s) 216static int __init x86_fxsr_setup(char *s)
214{ 217{
215 setup_clear_cpu_cap(X86_FEATURE_FXSR); 218 setup_clear_cpu_cap(X86_FEATURE_FXSR);
216 setup_clear_cpu_cap(X86_FEATURE_XMM); 219 setup_clear_cpu_cap(X86_FEATURE_XMM);
@@ -219,7 +222,7 @@ static int __init x86_fxsr_setup(char * s)
219__setup("nofxsr", x86_fxsr_setup); 222__setup("nofxsr", x86_fxsr_setup);
220 223
221 224
222static int __init x86_sep_setup(char * s) 225static int __init x86_sep_setup(char *s)
223{ 226{
224 setup_clear_cpu_cap(X86_FEATURE_SEP); 227 setup_clear_cpu_cap(X86_FEATURE_SEP);
225 return 1; 228 return 1;
@@ -306,14 +309,30 @@ static void __cpuinit early_get_cap(struct cpuinfo_x86 *c)
306 309
307 } 310 }
308 311
309} 312 clear_cpu_cap(c, X86_FEATURE_PAT);
313
314 switch (c->x86_vendor) {
315 case X86_VENDOR_AMD:
316 if (c->x86 >= 0xf && c->x86 <= 0x11)
317 set_cpu_cap(c, X86_FEATURE_PAT);
318 break;
319 case X86_VENDOR_INTEL:
320 if (c->x86 == 0xF || (c->x86 == 6 && c->x86_model >= 15))
321 set_cpu_cap(c, X86_FEATURE_PAT);
322 break;
323 }
310 324
311/* Do minimum CPU detection early. 325}
312 Fields really needed: vendor, cpuid_level, family, model, mask, cache alignment.
313 The others are not touched to avoid unwanted side effects.
314 326
315 WARNING: this function is only called on the BP. Don't add code here 327/*
316 that is supposed to run on all CPUs. */ 328 * Do minimum CPU detection early.
329 * Fields really needed: vendor, cpuid_level, family, model, mask,
330 * cache alignment.
331 * The others are not touched to avoid unwanted side effects.
332 *
333 * WARNING: this function is only called on the BP. Don't add code here
334 * that is supposed to run on all CPUs.
335 */
317static void __init early_cpu_detect(void) 336static void __init early_cpu_detect(void)
318{ 337{
319 struct cpuinfo_x86 *c = &boot_cpu_data; 338 struct cpuinfo_x86 *c = &boot_cpu_data;
@@ -328,19 +347,14 @@ static void __init early_cpu_detect(void)
328 347
329 get_cpu_vendor(c, 1); 348 get_cpu_vendor(c, 1);
330 349
331 switch (c->x86_vendor) { 350 if (c->x86_vendor != X86_VENDOR_UNKNOWN &&
332 case X86_VENDOR_AMD: 351 cpu_devs[c->x86_vendor]->c_early_init)
333 early_init_amd(c); 352 cpu_devs[c->x86_vendor]->c_early_init(c);
334 break;
335 case X86_VENDOR_INTEL:
336 early_init_intel(c);
337 break;
338 }
339 353
340 early_get_cap(c); 354 early_get_cap(c);
341} 355}
342 356
343static void __cpuinit generic_identify(struct cpuinfo_x86 * c) 357static void __cpuinit generic_identify(struct cpuinfo_x86 *c)
344{ 358{
345 u32 tfms, xlvl; 359 u32 tfms, xlvl;
346 unsigned int ebx; 360 unsigned int ebx;
@@ -351,13 +365,12 @@ static void __cpuinit generic_identify(struct cpuinfo_x86 * c)
351 (unsigned int *)&c->x86_vendor_id[0], 365 (unsigned int *)&c->x86_vendor_id[0],
352 (unsigned int *)&c->x86_vendor_id[8], 366 (unsigned int *)&c->x86_vendor_id[8],
353 (unsigned int *)&c->x86_vendor_id[4]); 367 (unsigned int *)&c->x86_vendor_id[4]);
354 368
355 get_cpu_vendor(c, 0); 369 get_cpu_vendor(c, 0);
356 /* Initialize the standard set of capabilities */ 370 /* Initialize the standard set of capabilities */
357 /* Note that the vendor-specific code below might override */ 371 /* Note that the vendor-specific code below might override */
358
359 /* Intel-defined flags: level 0x00000001 */ 372 /* Intel-defined flags: level 0x00000001 */
360 if ( c->cpuid_level >= 0x00000001 ) { 373 if (c->cpuid_level >= 0x00000001) {
361 u32 capability, excap; 374 u32 capability, excap;
362 cpuid(0x00000001, &tfms, &ebx, &excap, &capability); 375 cpuid(0x00000001, &tfms, &ebx, &excap, &capability);
363 c->x86_capability[0] = capability; 376 c->x86_capability[0] = capability;
@@ -369,12 +382,14 @@ static void __cpuinit generic_identify(struct cpuinfo_x86 * c)
369 if (c->x86 >= 0x6) 382 if (c->x86 >= 0x6)
370 c->x86_model += ((tfms >> 16) & 0xF) << 4; 383 c->x86_model += ((tfms >> 16) & 0xF) << 4;
371 c->x86_mask = tfms & 15; 384 c->x86_mask = tfms & 15;
385 c->initial_apicid = (ebx >> 24) & 0xFF;
372#ifdef CONFIG_X86_HT 386#ifdef CONFIG_X86_HT
373 c->apicid = phys_pkg_id((ebx >> 24) & 0xFF, 0); 387 c->apicid = phys_pkg_id(c->initial_apicid, 0);
388 c->phys_proc_id = c->initial_apicid;
374#else 389#else
375 c->apicid = (ebx >> 24) & 0xFF; 390 c->apicid = c->initial_apicid;
376#endif 391#endif
377 if (c->x86_capability[0] & (1<<19)) 392 if (test_cpu_cap(c, X86_FEATURE_CLFLSH))
378 c->x86_clflush_size = ((ebx >> 8) & 0xff) * 8; 393 c->x86_clflush_size = ((ebx >> 8) & 0xff) * 8;
379 } else { 394 } else {
380 /* Have CPUID level 0 only - unheard of */ 395 /* Have CPUID level 0 only - unheard of */
@@ -383,33 +398,42 @@ static void __cpuinit generic_identify(struct cpuinfo_x86 * c)
383 398
384 /* AMD-defined flags: level 0x80000001 */ 399 /* AMD-defined flags: level 0x80000001 */
385 xlvl = cpuid_eax(0x80000000); 400 xlvl = cpuid_eax(0x80000000);
386 if ( (xlvl & 0xffff0000) == 0x80000000 ) { 401 if ((xlvl & 0xffff0000) == 0x80000000) {
387 if ( xlvl >= 0x80000001 ) { 402 if (xlvl >= 0x80000001) {
388 c->x86_capability[1] = cpuid_edx(0x80000001); 403 c->x86_capability[1] = cpuid_edx(0x80000001);
389 c->x86_capability[6] = cpuid_ecx(0x80000001); 404 c->x86_capability[6] = cpuid_ecx(0x80000001);
390 } 405 }
391 if ( xlvl >= 0x80000004 ) 406 if (xlvl >= 0x80000004)
392 get_model_name(c); /* Default name */ 407 get_model_name(c); /* Default name */
393 } 408 }
394 409
395 init_scattered_cpuid_features(c); 410 init_scattered_cpuid_features(c);
396 } 411 }
397 412
398#ifdef CONFIG_X86_HT 413 clear_cpu_cap(c, X86_FEATURE_PAT);
399 c->phys_proc_id = (cpuid_ebx(1) >> 24) & 0xff; 414
400#endif 415 switch (c->x86_vendor) {
416 case X86_VENDOR_AMD:
417 if (c->x86 >= 0xf && c->x86 <= 0x11)
418 set_cpu_cap(c, X86_FEATURE_PAT);
419 break;
420 case X86_VENDOR_INTEL:
421 if (c->x86 == 0xF || (c->x86 == 6 && c->x86_model >= 15))
422 set_cpu_cap(c, X86_FEATURE_PAT);
423 break;
424 }
401} 425}
402 426
403static void __cpuinit squash_the_stupid_serial_number(struct cpuinfo_x86 *c) 427static void __cpuinit squash_the_stupid_serial_number(struct cpuinfo_x86 *c)
404{ 428{
405 if (cpu_has(c, X86_FEATURE_PN) && disable_x86_serial_nr ) { 429 if (cpu_has(c, X86_FEATURE_PN) && disable_x86_serial_nr) {
406 /* Disable processor serial number */ 430 /* Disable processor serial number */
407 unsigned long lo,hi; 431 unsigned long lo, hi;
408 rdmsr(MSR_IA32_BBL_CR_CTL,lo,hi); 432 rdmsr(MSR_IA32_BBL_CR_CTL, lo, hi);
409 lo |= 0x200000; 433 lo |= 0x200000;
410 wrmsr(MSR_IA32_BBL_CR_CTL,lo,hi); 434 wrmsr(MSR_IA32_BBL_CR_CTL, lo, hi);
411 printk(KERN_NOTICE "CPU serial number disabled.\n"); 435 printk(KERN_NOTICE "CPU serial number disabled.\n");
412 clear_bit(X86_FEATURE_PN, c->x86_capability); 436 clear_cpu_cap(c, X86_FEATURE_PN);
413 437
414 /* Disabling the serial number may affect the cpuid level */ 438 /* Disabling the serial number may affect the cpuid level */
415 c->cpuid_level = cpuid_eax(0); 439 c->cpuid_level = cpuid_eax(0);
@@ -444,9 +468,11 @@ void __cpuinit identify_cpu(struct cpuinfo_x86 *c)
444 memset(&c->x86_capability, 0, sizeof c->x86_capability); 468 memset(&c->x86_capability, 0, sizeof c->x86_capability);
445 469
446 if (!have_cpuid_p()) { 470 if (!have_cpuid_p()) {
447 /* First of all, decide if this is a 486 or higher */ 471 /*
448 /* It's a 486 if we can modify the AC flag */ 472 * First of all, decide if this is a 486 or higher
449 if ( flag_is_changeable_p(X86_EFLAGS_AC) ) 473 * It's a 486 if we can modify the AC flag
474 */
475 if (flag_is_changeable_p(X86_EFLAGS_AC))
450 c->x86 = 4; 476 c->x86 = 4;
451 else 477 else
452 c->x86 = 3; 478 c->x86 = 3;
@@ -479,10 +505,10 @@ void __cpuinit identify_cpu(struct cpuinfo_x86 *c)
479 */ 505 */
480 506
481 /* If the model name is still unset, do table lookup. */ 507 /* If the model name is still unset, do table lookup. */
482 if ( !c->x86_model_id[0] ) { 508 if (!c->x86_model_id[0]) {
483 char *p; 509 char *p;
484 p = table_lookup_model(c); 510 p = table_lookup_model(c);
485 if ( p ) 511 if (p)
486 strcpy(c->x86_model_id, p); 512 strcpy(c->x86_model_id, p);
487 else 513 else
488 /* Last resort... */ 514 /* Last resort... */
@@ -496,9 +522,9 @@ void __cpuinit identify_cpu(struct cpuinfo_x86 *c)
496 * common between the CPUs. The first time this routine gets 522 * common between the CPUs. The first time this routine gets
497 * executed, c == &boot_cpu_data. 523 * executed, c == &boot_cpu_data.
498 */ 524 */
499 if ( c != &boot_cpu_data ) { 525 if (c != &boot_cpu_data) {
500 /* AND the already accumulated flags with these */ 526 /* AND the already accumulated flags with these */
501 for ( i = 0 ; i < NCAPINTS ; i++ ) 527 for (i = 0 ; i < NCAPINTS ; i++)
502 boot_cpu_data.x86_capability[i] &= c->x86_capability[i]; 528 boot_cpu_data.x86_capability[i] &= c->x86_capability[i];
503 } 529 }
504 530
@@ -542,7 +568,7 @@ void __cpuinit detect_ht(struct cpuinfo_x86 *c)
542 568
543 if (smp_num_siblings == 1) { 569 if (smp_num_siblings == 1) {
544 printk(KERN_INFO "CPU: Hyper-Threading is disabled\n"); 570 printk(KERN_INFO "CPU: Hyper-Threading is disabled\n");
545 } else if (smp_num_siblings > 1 ) { 571 } else if (smp_num_siblings > 1) {
546 572
547 if (smp_num_siblings > NR_CPUS) { 573 if (smp_num_siblings > NR_CPUS) {
548 printk(KERN_WARNING "CPU: Unsupported number of the " 574 printk(KERN_WARNING "CPU: Unsupported number of the "
@@ -552,7 +578,7 @@ void __cpuinit detect_ht(struct cpuinfo_x86 *c)
552 } 578 }
553 579
554 index_msb = get_count_order(smp_num_siblings); 580 index_msb = get_count_order(smp_num_siblings);
555 c->phys_proc_id = phys_pkg_id((ebx >> 24) & 0xFF, index_msb); 581 c->phys_proc_id = phys_pkg_id(c->initial_apicid, index_msb);
556 582
557 printk(KERN_INFO "CPU: Physical Processor ID: %d\n", 583 printk(KERN_INFO "CPU: Physical Processor ID: %d\n",
558 c->phys_proc_id); 584 c->phys_proc_id);
@@ -563,7 +589,7 @@ void __cpuinit detect_ht(struct cpuinfo_x86 *c)
563 589
564 core_bits = get_count_order(c->x86_max_cores); 590 core_bits = get_count_order(c->x86_max_cores);
565 591
566 c->cpu_core_id = phys_pkg_id((ebx >> 24) & 0xFF, index_msb) & 592 c->cpu_core_id = phys_pkg_id(c->initial_apicid, index_msb) &
567 ((1 << core_bits) - 1); 593 ((1 << core_bits) - 1);
568 594
569 if (c->x86_max_cores > 1) 595 if (c->x86_max_cores > 1)
@@ -597,7 +623,7 @@ void __cpuinit print_cpu_info(struct cpuinfo_x86 *c)
597 else 623 else
598 printk("%s", c->x86_model_id); 624 printk("%s", c->x86_model_id);
599 625
600 if (c->x86_mask || c->cpuid_level >= 0) 626 if (c->x86_mask || c->cpuid_level >= 0)
601 printk(" stepping %02x\n", c->x86_mask); 627 printk(" stepping %02x\n", c->x86_mask);
602 else 628 else
603 printk("\n"); 629 printk("\n");
@@ -616,23 +642,15 @@ __setup("clearcpuid=", setup_disablecpuid);
616 642
617cpumask_t cpu_initialized __cpuinitdata = CPU_MASK_NONE; 643cpumask_t cpu_initialized __cpuinitdata = CPU_MASK_NONE;
618 644
619/* This is hacky. :)
620 * We're emulating future behavior.
621 * In the future, the cpu-specific init functions will be called implicitly
622 * via the magic of initcalls.
623 * They will insert themselves into the cpu_devs structure.
624 * Then, when cpu_init() is called, we can just iterate over that array.
625 */
626void __init early_cpu_init(void) 645void __init early_cpu_init(void)
627{ 646{
628 intel_cpu_init(); 647 struct cpu_vendor_dev *cvdev;
629 cyrix_init_cpu(); 648
630 nsc_init_cpu(); 649 for (cvdev = __x86cpuvendor_start ;
631 amd_init_cpu(); 650 cvdev < __x86cpuvendor_end ;
632 centaur_init_cpu(); 651 cvdev++)
633 transmeta_init_cpu(); 652 cpu_devs[cvdev->vendor] = cvdev->cpu_dev;
634 nexgen_init_cpu(); 653
635 umc_init_cpu();
636 early_cpu_detect(); 654 early_cpu_detect();
637} 655}
638 656
@@ -666,7 +684,7 @@ void __cpuinit cpu_init(void)
666{ 684{
667 int cpu = smp_processor_id(); 685 int cpu = smp_processor_id();
668 struct task_struct *curr = current; 686 struct task_struct *curr = current;
669 struct tss_struct * t = &per_cpu(init_tss, cpu); 687 struct tss_struct *t = &per_cpu(init_tss, cpu);
670 struct thread_struct *thread = &curr->thread; 688 struct thread_struct *thread = &curr->thread;
671 689
672 if (cpu_test_and_set(cpu, cpu_initialized)) { 690 if (cpu_test_and_set(cpu, cpu_initialized)) {
@@ -692,7 +710,7 @@ void __cpuinit cpu_init(void)
692 enter_lazy_tlb(&init_mm, curr); 710 enter_lazy_tlb(&init_mm, curr);
693 711
694 load_sp0(t, thread); 712 load_sp0(t, thread);
695 set_tss_desc(cpu,t); 713 set_tss_desc(cpu, t);
696 load_TR_desc(); 714 load_TR_desc();
697 load_LDT(&init_mm.context); 715 load_LDT(&init_mm.context);
698 716
diff --git a/arch/x86/kernel/cpu/cpu.h b/arch/x86/kernel/cpu/cpu.h
index e0b38c33d842..783691b2a738 100644
--- a/arch/x86/kernel/cpu/cpu.h
+++ b/arch/x86/kernel/cpu/cpu.h
@@ -14,6 +14,7 @@ struct cpu_dev {
14 14
15 struct cpu_model_info c_models[4]; 15 struct cpu_model_info c_models[4];
16 16
17 void (*c_early_init)(struct cpuinfo_x86 *c);
17 void (*c_init)(struct cpuinfo_x86 * c); 18 void (*c_init)(struct cpuinfo_x86 * c);
18 void (*c_identify)(struct cpuinfo_x86 * c); 19 void (*c_identify)(struct cpuinfo_x86 * c);
19 unsigned int (*c_size_cache)(struct cpuinfo_x86 * c, unsigned int size); 20 unsigned int (*c_size_cache)(struct cpuinfo_x86 * c, unsigned int size);
@@ -21,18 +22,17 @@ struct cpu_dev {
21 22
22extern struct cpu_dev * cpu_devs [X86_VENDOR_NUM]; 23extern struct cpu_dev * cpu_devs [X86_VENDOR_NUM];
23 24
25struct cpu_vendor_dev {
26 int vendor;
27 struct cpu_dev *cpu_dev;
28};
29
30#define cpu_vendor_dev_register(cpu_vendor_id, cpu_dev) \
31 static struct cpu_vendor_dev __cpu_vendor_dev_##cpu_vendor_id __used \
32 __attribute__((__section__(".x86cpuvendor.init"))) = \
33 { cpu_vendor_id, cpu_dev }
34
35extern struct cpu_vendor_dev __x86cpuvendor_start[], __x86cpuvendor_end[];
36
24extern int get_model_name(struct cpuinfo_x86 *c); 37extern int get_model_name(struct cpuinfo_x86 *c);
25extern void display_cacheinfo(struct cpuinfo_x86 *c); 38extern void display_cacheinfo(struct cpuinfo_x86 *c);
26
27extern void early_init_intel(struct cpuinfo_x86 *c);
28extern void early_init_amd(struct cpuinfo_x86 *c);
29
30/* Specific CPU type init functions */
31int intel_cpu_init(void);
32int amd_init_cpu(void);
33int cyrix_init_cpu(void);
34int nsc_init_cpu(void);
35int centaur_init_cpu(void);
36int transmeta_init_cpu(void);
37int nexgen_init_cpu(void);
38int umc_init_cpu(void);
diff --git a/arch/x86/kernel/cpu/cyrix.c b/arch/x86/kernel/cpu/cyrix.c
index 7139b0262703..3fd7a67bb06a 100644
--- a/arch/x86/kernel/cpu/cyrix.c
+++ b/arch/x86/kernel/cpu/cyrix.c
@@ -19,7 +19,7 @@ static void __cpuinit do_cyrix_devid(unsigned char *dir0, unsigned char *dir1)
19{ 19{
20 unsigned char ccr2, ccr3; 20 unsigned char ccr2, ccr3;
21 unsigned long flags; 21 unsigned long flags;
22 22
23 /* we test for DEVID by checking whether CCR3 is writable */ 23 /* we test for DEVID by checking whether CCR3 is writable */
24 local_irq_save(flags); 24 local_irq_save(flags);
25 ccr3 = getCx86(CX86_CCR3); 25 ccr3 = getCx86(CX86_CCR3);
@@ -37,8 +37,7 @@ static void __cpuinit do_cyrix_devid(unsigned char *dir0, unsigned char *dir1)
37 setCx86(CX86_CCR2, ccr2); 37 setCx86(CX86_CCR2, ccr2);
38 *dir0 = 0xfe; 38 *dir0 = 0xfe;
39 } 39 }
40 } 40 } else {
41 else {
42 setCx86(CX86_CCR3, ccr3); /* restore CCR3 */ 41 setCx86(CX86_CCR3, ccr3); /* restore CCR3 */
43 42
44 /* read DIR0 and DIR1 CPU registers */ 43 /* read DIR0 and DIR1 CPU registers */
@@ -86,7 +85,7 @@ static char cyrix_model_mult2[] __cpuinitdata = "12233445";
86static void __cpuinit check_cx686_slop(struct cpuinfo_x86 *c) 85static void __cpuinit check_cx686_slop(struct cpuinfo_x86 *c)
87{ 86{
88 unsigned long flags; 87 unsigned long flags;
89 88
90 if (Cx86_dir0_msb == 3) { 89 if (Cx86_dir0_msb == 3) {
91 unsigned char ccr3, ccr5; 90 unsigned char ccr3, ccr5;
92 91
@@ -132,7 +131,7 @@ static void __cpuinit set_cx86_memwb(void)
132 /* set 'Not Write-through' */ 131 /* set 'Not Write-through' */
133 write_cr0(read_cr0() | X86_CR0_NW); 132 write_cr0(read_cr0() | X86_CR0_NW);
134 /* CCR2 bit 2: lock NW bit and set WT1 */ 133 /* CCR2 bit 2: lock NW bit and set WT1 */
135 setCx86(CX86_CCR2, getCx86(CX86_CCR2) | 0x14 ); 134 setCx86(CX86_CCR2, getCx86(CX86_CCR2) | 0x14);
136} 135}
137 136
138static void __cpuinit set_cx86_inc(void) 137static void __cpuinit set_cx86_inc(void)
@@ -148,7 +147,7 @@ static void __cpuinit set_cx86_inc(void)
148 setCx86(CX86_PCR1, getCx86(CX86_PCR1) | 0x02); 147 setCx86(CX86_PCR1, getCx86(CX86_PCR1) | 0x02);
149 /* PCR0 -- Performance Control */ 148 /* PCR0 -- Performance Control */
150 /* Incrementor Margin 10 */ 149 /* Incrementor Margin 10 */
151 setCx86(CX86_PCR0, getCx86(CX86_PCR0) | 0x04); 150 setCx86(CX86_PCR0, getCx86(CX86_PCR0) | 0x04);
152 setCx86(CX86_CCR3, ccr3); /* disable MAPEN */ 151 setCx86(CX86_CCR3, ccr3); /* disable MAPEN */
153} 152}
154 153
@@ -167,16 +166,16 @@ static void __cpuinit geode_configure(void)
167 166
168 ccr3 = getCx86(CX86_CCR3); 167 ccr3 = getCx86(CX86_CCR3);
169 setCx86(CX86_CCR3, (ccr3 & 0x0f) | 0x10); /* enable MAPEN */ 168 setCx86(CX86_CCR3, (ccr3 & 0x0f) | 0x10); /* enable MAPEN */
170 169
171 170
172 /* FPU fast, DTE cache, Mem bypass */ 171 /* FPU fast, DTE cache, Mem bypass */
173 setCx86(CX86_CCR4, getCx86(CX86_CCR4) | 0x38); 172 setCx86(CX86_CCR4, getCx86(CX86_CCR4) | 0x38);
174 setCx86(CX86_CCR3, ccr3); /* disable MAPEN */ 173 setCx86(CX86_CCR3, ccr3); /* disable MAPEN */
175 174
176 set_cx86_memwb(); 175 set_cx86_memwb();
177 set_cx86_reorder(); 176 set_cx86_reorder();
178 set_cx86_inc(); 177 set_cx86_inc();
179 178
180 local_irq_restore(flags); 179 local_irq_restore(flags);
181} 180}
182 181
@@ -187,14 +186,16 @@ static void __cpuinit init_cyrix(struct cpuinfo_x86 *c)
187 char *buf = c->x86_model_id; 186 char *buf = c->x86_model_id;
188 const char *p = NULL; 187 const char *p = NULL;
189 188
190 /* Bit 31 in normal CPUID used for nonstandard 3DNow ID; 189 /*
191 3DNow is IDd by bit 31 in extended CPUID (1*32+31) anyway */ 190 * Bit 31 in normal CPUID used for nonstandard 3DNow ID;
192 clear_bit(0*32+31, c->x86_capability); 191 * 3DNow is IDd by bit 31 in extended CPUID (1*32+31) anyway
192 */
193 clear_cpu_cap(c, 0*32+31);
193 194
194 /* Cyrix used bit 24 in extended (AMD) CPUID for Cyrix MMX extensions */ 195 /* Cyrix used bit 24 in extended (AMD) CPUID for Cyrix MMX extensions */
195 if ( test_bit(1*32+24, c->x86_capability) ) { 196 if (test_cpu_cap(c, 1*32+24)) {
196 clear_bit(1*32+24, c->x86_capability); 197 clear_cpu_cap(c, 1*32+24);
197 set_bit(X86_FEATURE_CXMMX, c->x86_capability); 198 set_cpu_cap(c, X86_FEATURE_CXMMX);
198 } 199 }
199 200
200 do_cyrix_devid(&dir0, &dir1); 201 do_cyrix_devid(&dir0, &dir1);
@@ -213,7 +214,7 @@ static void __cpuinit init_cyrix(struct cpuinfo_x86 *c)
213 * the model, multiplier and stepping. Black magic included, 214 * the model, multiplier and stepping. Black magic included,
214 * to make the silicon step/rev numbers match the printed ones. 215 * to make the silicon step/rev numbers match the printed ones.
215 */ 216 */
216 217
217 switch (dir0_msn) { 218 switch (dir0_msn) {
218 unsigned char tmp; 219 unsigned char tmp;
219 220
@@ -241,7 +242,7 @@ static void __cpuinit init_cyrix(struct cpuinfo_x86 *c)
241 } else /* 686 */ 242 } else /* 686 */
242 p = Cx86_cb+1; 243 p = Cx86_cb+1;
243 /* Emulate MTRRs using Cyrix's ARRs. */ 244 /* Emulate MTRRs using Cyrix's ARRs. */
244 set_bit(X86_FEATURE_CYRIX_ARR, c->x86_capability); 245 set_cpu_cap(c, X86_FEATURE_CYRIX_ARR);
245 /* 6x86's contain this bug */ 246 /* 6x86's contain this bug */
246 c->coma_bug = 1; 247 c->coma_bug = 1;
247 break; 248 break;
@@ -250,17 +251,18 @@ static void __cpuinit init_cyrix(struct cpuinfo_x86 *c)
250#ifdef CONFIG_PCI 251#ifdef CONFIG_PCI
251 { 252 {
252 u32 vendor, device; 253 u32 vendor, device;
253 /* It isn't really a PCI quirk directly, but the cure is the 254 /*
254 same. The MediaGX has deep magic SMM stuff that handles the 255 * It isn't really a PCI quirk directly, but the cure is the
255 SB emulation. It throws away the fifo on disable_dma() which 256 * same. The MediaGX has deep magic SMM stuff that handles the
256 is wrong and ruins the audio. 257 * SB emulation. It throws away the fifo on disable_dma() which
257 258 * is wrong and ruins the audio.
258 Bug2: VSA1 has a wrap bug so that using maximum sized DMA 259 *
259 causes bad things. According to NatSemi VSA2 has another 260 * Bug2: VSA1 has a wrap bug so that using maximum sized DMA
260 bug to do with 'hlt'. I've not seen any boards using VSA2 261 * causes bad things. According to NatSemi VSA2 has another
261 and X doesn't seem to support it either so who cares 8). 262 * bug to do with 'hlt'. I've not seen any boards using VSA2
262 VSA1 we work around however. 263 * and X doesn't seem to support it either so who cares 8).
263 */ 264 * VSA1 we work around however.
265 */
264 266
265 printk(KERN_INFO "Working around Cyrix MediaGX virtual DMA bugs.\n"); 267 printk(KERN_INFO "Working around Cyrix MediaGX virtual DMA bugs.\n");
266 isa_dma_bridge_buggy = 2; 268 isa_dma_bridge_buggy = 2;
@@ -273,55 +275,51 @@ static void __cpuinit init_cyrix(struct cpuinfo_x86 *c)
273 275
274 /* 276 /*
275 * The 5510/5520 companion chips have a funky PIT. 277 * The 5510/5520 companion chips have a funky PIT.
276 */ 278 */
277 if (vendor == PCI_VENDOR_ID_CYRIX && 279 if (vendor == PCI_VENDOR_ID_CYRIX &&
278 (device == PCI_DEVICE_ID_CYRIX_5510 || device == PCI_DEVICE_ID_CYRIX_5520)) 280 (device == PCI_DEVICE_ID_CYRIX_5510 || device == PCI_DEVICE_ID_CYRIX_5520))
279 mark_tsc_unstable("cyrix 5510/5520 detected"); 281 mark_tsc_unstable("cyrix 5510/5520 detected");
280 } 282 }
281#endif 283#endif
282 c->x86_cache_size=16; /* Yep 16K integrated cache thats it */ 284 c->x86_cache_size = 16; /* Yep 16K integrated cache thats it */
283 285
284 /* GXm supports extended cpuid levels 'ala' AMD */ 286 /* GXm supports extended cpuid levels 'ala' AMD */
285 if (c->cpuid_level == 2) { 287 if (c->cpuid_level == 2) {
286 /* Enable cxMMX extensions (GX1 Datasheet 54) */ 288 /* Enable cxMMX extensions (GX1 Datasheet 54) */
287 setCx86(CX86_CCR7, getCx86(CX86_CCR7) | 1); 289 setCx86(CX86_CCR7, getCx86(CX86_CCR7) | 1);
288 290
289 /* 291 /*
290 * GXm : 0x30 ... 0x5f GXm datasheet 51 292 * GXm : 0x30 ... 0x5f GXm datasheet 51
291 * GXlv: 0x6x GXlv datasheet 54 293 * GXlv: 0x6x GXlv datasheet 54
292 * ? : 0x7x 294 * ? : 0x7x
293 * GX1 : 0x8x GX1 datasheet 56 295 * GX1 : 0x8x GX1 datasheet 56
294 */ 296 */
295 if((0x30 <= dir1 && dir1 <= 0x6f) || (0x80 <=dir1 && dir1 <= 0x8f)) 297 if ((0x30 <= dir1 && dir1 <= 0x6f) || (0x80 <= dir1 && dir1 <= 0x8f))
296 geode_configure(); 298 geode_configure();
297 get_model_name(c); /* get CPU marketing name */ 299 get_model_name(c); /* get CPU marketing name */
298 return; 300 return;
299 } 301 } else { /* MediaGX */
300 else { /* MediaGX */
301 Cx86_cb[2] = (dir0_lsn & 1) ? '3' : '4'; 302 Cx86_cb[2] = (dir0_lsn & 1) ? '3' : '4';
302 p = Cx86_cb+2; 303 p = Cx86_cb+2;
303 c->x86_model = (dir1 & 0x20) ? 1 : 2; 304 c->x86_model = (dir1 & 0x20) ? 1 : 2;
304 } 305 }
305 break; 306 break;
306 307
307 case 5: /* 6x86MX/M II */ 308 case 5: /* 6x86MX/M II */
308 if (dir1 > 7) 309 if (dir1 > 7) {
309 {
310 dir0_msn++; /* M II */ 310 dir0_msn++; /* M II */
311 /* Enable MMX extensions (App note 108) */ 311 /* Enable MMX extensions (App note 108) */
312 setCx86(CX86_CCR7, getCx86(CX86_CCR7)|1); 312 setCx86(CX86_CCR7, getCx86(CX86_CCR7)|1);
313 } 313 } else {
314 else
315 {
316 c->coma_bug = 1; /* 6x86MX, it has the bug. */ 314 c->coma_bug = 1; /* 6x86MX, it has the bug. */
317 } 315 }
318 tmp = (!(dir0_lsn & 7) || dir0_lsn & 1) ? 2 : 0; 316 tmp = (!(dir0_lsn & 7) || dir0_lsn & 1) ? 2 : 0;
319 Cx86_cb[tmp] = cyrix_model_mult2[dir0_lsn & 7]; 317 Cx86_cb[tmp] = cyrix_model_mult2[dir0_lsn & 7];
320 p = Cx86_cb+tmp; 318 p = Cx86_cb+tmp;
321 if (((dir1 & 0x0f) > 4) || ((dir1 & 0xf0) == 0x20)) 319 if (((dir1 & 0x0f) > 4) || ((dir1 & 0xf0) == 0x20))
322 (c->x86_model)++; 320 (c->x86_model)++;
323 /* Emulate MTRRs using Cyrix's ARRs. */ 321 /* Emulate MTRRs using Cyrix's ARRs. */
324 set_bit(X86_FEATURE_CYRIX_ARR, c->x86_capability); 322 set_cpu_cap(c, X86_FEATURE_CYRIX_ARR);
325 break; 323 break;
326 324
327 case 0xf: /* Cyrix 486 without DEVID registers */ 325 case 0xf: /* Cyrix 486 without DEVID registers */
@@ -343,7 +341,8 @@ static void __cpuinit init_cyrix(struct cpuinfo_x86 *c)
343 break; 341 break;
344 } 342 }
345 strcpy(buf, Cx86_model[dir0_msn & 7]); 343 strcpy(buf, Cx86_model[dir0_msn & 7]);
346 if (p) strcat(buf, p); 344 if (p)
345 strcat(buf, p);
347 return; 346 return;
348} 347}
349 348
@@ -352,7 +351,8 @@ static void __cpuinit init_cyrix(struct cpuinfo_x86 *c)
352 */ 351 */
353static void __cpuinit init_nsc(struct cpuinfo_x86 *c) 352static void __cpuinit init_nsc(struct cpuinfo_x86 *c)
354{ 353{
355 /* There may be GX1 processors in the wild that are branded 354 /*
355 * There may be GX1 processors in the wild that are branded
356 * NSC and not Cyrix. 356 * NSC and not Cyrix.
357 * 357 *
358 * This function only handles the GX processor, and kicks every 358 * This function only handles the GX processor, and kicks every
@@ -377,7 +377,7 @@ static void __cpuinit init_nsc(struct cpuinfo_x86 *c)
377 * by the fact that they preserve the flags across the division of 5/2. 377 * by the fact that they preserve the flags across the division of 5/2.
378 * PII and PPro exhibit this behavior too, but they have cpuid available. 378 * PII and PPro exhibit this behavior too, but they have cpuid available.
379 */ 379 */
380 380
381/* 381/*
382 * Perform the Cyrix 5/2 test. A Cyrix won't change 382 * Perform the Cyrix 5/2 test. A Cyrix won't change
383 * the flags, while other 486 chips will. 383 * the flags, while other 486 chips will.
@@ -398,27 +398,26 @@ static inline int test_cyrix_52div(void)
398 return (unsigned char) (test >> 8) == 0x02; 398 return (unsigned char) (test >> 8) == 0x02;
399} 399}
400 400
401static void __cpuinit cyrix_identify(struct cpuinfo_x86 * c) 401static void __cpuinit cyrix_identify(struct cpuinfo_x86 *c)
402{ 402{
403 /* Detect Cyrix with disabled CPUID */ 403 /* Detect Cyrix with disabled CPUID */
404 if ( c->x86 == 4 && test_cyrix_52div() ) { 404 if (c->x86 == 4 && test_cyrix_52div()) {
405 unsigned char dir0, dir1; 405 unsigned char dir0, dir1;
406 406
407 strcpy(c->x86_vendor_id, "CyrixInstead"); 407 strcpy(c->x86_vendor_id, "CyrixInstead");
408 c->x86_vendor = X86_VENDOR_CYRIX; 408 c->x86_vendor = X86_VENDOR_CYRIX;
409 409
410 /* Actually enable cpuid on the older cyrix */ 410 /* Actually enable cpuid on the older cyrix */
411 411
412 /* Retrieve CPU revisions */ 412 /* Retrieve CPU revisions */
413 413
414 do_cyrix_devid(&dir0, &dir1); 414 do_cyrix_devid(&dir0, &dir1);
415 415
416 dir0>>=4; 416 dir0 >>= 4;
417 417
418 /* Check it is an affected model */ 418 /* Check it is an affected model */
419 419
420 if (dir0 == 5 || dir0 == 3) 420 if (dir0 == 5 || dir0 == 3) {
421 {
422 unsigned char ccr3; 421 unsigned char ccr3;
423 unsigned long flags; 422 unsigned long flags;
424 printk(KERN_INFO "Enabling CPUID on Cyrix processor.\n"); 423 printk(KERN_INFO "Enabling CPUID on Cyrix processor.\n");
@@ -434,26 +433,17 @@ static void __cpuinit cyrix_identify(struct cpuinfo_x86 * c)
434 433
435static struct cpu_dev cyrix_cpu_dev __cpuinitdata = { 434static struct cpu_dev cyrix_cpu_dev __cpuinitdata = {
436 .c_vendor = "Cyrix", 435 .c_vendor = "Cyrix",
437 .c_ident = { "CyrixInstead" }, 436 .c_ident = { "CyrixInstead" },
438 .c_init = init_cyrix, 437 .c_init = init_cyrix,
439 .c_identify = cyrix_identify, 438 .c_identify = cyrix_identify,
440}; 439};
441 440
442int __init cyrix_init_cpu(void) 441cpu_vendor_dev_register(X86_VENDOR_CYRIX, &cyrix_cpu_dev);
443{
444 cpu_devs[X86_VENDOR_CYRIX] = &cyrix_cpu_dev;
445 return 0;
446}
447 442
448static struct cpu_dev nsc_cpu_dev __cpuinitdata = { 443static struct cpu_dev nsc_cpu_dev __cpuinitdata = {
449 .c_vendor = "NSC", 444 .c_vendor = "NSC",
450 .c_ident = { "Geode by NSC" }, 445 .c_ident = { "Geode by NSC" },
451 .c_init = init_nsc, 446 .c_init = init_nsc,
452}; 447};
453 448
454int __init nsc_init_cpu(void) 449cpu_vendor_dev_register(X86_VENDOR_NSC, &nsc_cpu_dev);
455{
456 cpu_devs[X86_VENDOR_NSC] = &nsc_cpu_dev;
457 return 0;
458}
459
diff --git a/arch/x86/kernel/cpu/feature_names.c b/arch/x86/kernel/cpu/feature_names.c
index ee975ac6bbcb..e43ad4ad4cba 100644
--- a/arch/x86/kernel/cpu/feature_names.c
+++ b/arch/x86/kernel/cpu/feature_names.c
@@ -4,7 +4,7 @@
4 * This file must not contain any executable code. 4 * This file must not contain any executable code.
5 */ 5 */
6 6
7#include "asm/cpufeature.h" 7#include <asm/cpufeature.h>
8 8
9/* 9/*
10 * These flag bits must match the definitions in <asm/cpufeature.h>. 10 * These flag bits must match the definitions in <asm/cpufeature.h>.
diff --git a/arch/x86/kernel/cpu/intel.c b/arch/x86/kernel/cpu/intel.c
index fae31ce747bd..fe9224c51d37 100644
--- a/arch/x86/kernel/cpu/intel.c
+++ b/arch/x86/kernel/cpu/intel.c
@@ -30,7 +30,7 @@
30struct movsl_mask movsl_mask __read_mostly; 30struct movsl_mask movsl_mask __read_mostly;
31#endif 31#endif
32 32
33void __cpuinit early_init_intel(struct cpuinfo_x86 *c) 33static void __cpuinit early_init_intel(struct cpuinfo_x86 *c)
34{ 34{
35 /* Netburst reports 64 bytes clflush size, but does IO in 128 bytes */ 35 /* Netburst reports 64 bytes clflush size, but does IO in 128 bytes */
36 if (c->x86 == 15 && c->x86_cache_alignment == 64) 36 if (c->x86 == 15 && c->x86_cache_alignment == 64)
@@ -45,7 +45,7 @@ void __cpuinit early_init_intel(struct cpuinfo_x86 *c)
45 * 45 *
46 * This is called before we do cpu ident work 46 * This is called before we do cpu ident work
47 */ 47 */
48 48
49int __cpuinit ppro_with_ram_bug(void) 49int __cpuinit ppro_with_ram_bug(void)
50{ 50{
51 /* Uses data from early_cpu_detect now */ 51 /* Uses data from early_cpu_detect now */
@@ -58,7 +58,7 @@ int __cpuinit ppro_with_ram_bug(void)
58 } 58 }
59 return 0; 59 return 0;
60} 60}
61 61
62 62
63/* 63/*
64 * P4 Xeon errata 037 workaround. 64 * P4 Xeon errata 037 workaround.
@@ -69,7 +69,7 @@ static void __cpuinit Intel_errata_workarounds(struct cpuinfo_x86 *c)
69 unsigned long lo, hi; 69 unsigned long lo, hi;
70 70
71 if ((c->x86 == 15) && (c->x86_model == 1) && (c->x86_mask == 1)) { 71 if ((c->x86 == 15) && (c->x86_model == 1) && (c->x86_mask == 1)) {
72 rdmsr (MSR_IA32_MISC_ENABLE, lo, hi); 72 rdmsr(MSR_IA32_MISC_ENABLE, lo, hi);
73 if ((lo & (1<<9)) == 0) { 73 if ((lo & (1<<9)) == 0) {
74 printk (KERN_INFO "CPU: C0 stepping P4 Xeon detected.\n"); 74 printk (KERN_INFO "CPU: C0 stepping P4 Xeon detected.\n");
75 printk (KERN_INFO "CPU: Disabling hardware prefetching (Errata 037)\n"); 75 printk (KERN_INFO "CPU: Disabling hardware prefetching (Errata 037)\n");
@@ -127,10 +127,10 @@ static void __cpuinit init_intel(struct cpuinfo_x86 *c)
127 */ 127 */
128 c->f00f_bug = 0; 128 c->f00f_bug = 0;
129 if (!paravirt_enabled() && c->x86 == 5) { 129 if (!paravirt_enabled() && c->x86 == 5) {
130 static int f00f_workaround_enabled = 0; 130 static int f00f_workaround_enabled;
131 131
132 c->f00f_bug = 1; 132 c->f00f_bug = 1;
133 if ( !f00f_workaround_enabled ) { 133 if (!f00f_workaround_enabled) {
134 trap_init_f00f_bug(); 134 trap_init_f00f_bug();
135 printk(KERN_NOTICE "Intel Pentium with F0 0F bug - workaround enabled.\n"); 135 printk(KERN_NOTICE "Intel Pentium with F0 0F bug - workaround enabled.\n");
136 f00f_workaround_enabled = 1; 136 f00f_workaround_enabled = 1;
@@ -139,20 +139,22 @@ static void __cpuinit init_intel(struct cpuinfo_x86 *c)
139#endif 139#endif
140 140
141 l2 = init_intel_cacheinfo(c); 141 l2 = init_intel_cacheinfo(c);
142 if (c->cpuid_level > 9 ) { 142 if (c->cpuid_level > 9) {
143 unsigned eax = cpuid_eax(10); 143 unsigned eax = cpuid_eax(10);
144 /* Check for version and the number of counters */ 144 /* Check for version and the number of counters */
145 if ((eax & 0xff) && (((eax>>8) & 0xff) > 1)) 145 if ((eax & 0xff) && (((eax>>8) & 0xff) > 1))
146 set_bit(X86_FEATURE_ARCH_PERFMON, c->x86_capability); 146 set_cpu_cap(c, X86_FEATURE_ARCH_PERFMON);
147 } 147 }
148 148
149 /* SEP CPUID bug: Pentium Pro reports SEP but doesn't have it until model 3 mask 3 */ 149 /* SEP CPUID bug: Pentium Pro reports SEP but doesn't have it until model 3 mask 3 */
150 if ((c->x86<<8 | c->x86_model<<4 | c->x86_mask) < 0x633) 150 if ((c->x86<<8 | c->x86_model<<4 | c->x86_mask) < 0x633)
151 clear_bit(X86_FEATURE_SEP, c->x86_capability); 151 clear_cpu_cap(c, X86_FEATURE_SEP);
152 152
153 /* Names for the Pentium II/Celeron processors 153 /*
154 detectable only by also checking the cache size. 154 * Names for the Pentium II/Celeron processors
155 Dixon is NOT a Celeron. */ 155 * detectable only by also checking the cache size.
156 * Dixon is NOT a Celeron.
157 */
156 if (c->x86 == 6) { 158 if (c->x86 == 6) {
157 switch (c->x86_model) { 159 switch (c->x86_model) {
158 case 5: 160 case 5:
@@ -163,14 +165,14 @@ static void __cpuinit init_intel(struct cpuinfo_x86 *c)
163 p = "Mobile Pentium II (Dixon)"; 165 p = "Mobile Pentium II (Dixon)";
164 } 166 }
165 break; 167 break;
166 168
167 case 6: 169 case 6:
168 if (l2 == 128) 170 if (l2 == 128)
169 p = "Celeron (Mendocino)"; 171 p = "Celeron (Mendocino)";
170 else if (c->x86_mask == 0 || c->x86_mask == 5) 172 else if (c->x86_mask == 0 || c->x86_mask == 5)
171 p = "Celeron-A"; 173 p = "Celeron-A";
172 break; 174 break;
173 175
174 case 8: 176 case 8:
175 if (l2 == 128) 177 if (l2 == 128)
176 p = "Celeron (Coppermine)"; 178 p = "Celeron (Coppermine)";
@@ -178,9 +180,9 @@ static void __cpuinit init_intel(struct cpuinfo_x86 *c)
178 } 180 }
179 } 181 }
180 182
181 if ( p ) 183 if (p)
182 strcpy(c->x86_model_id, p); 184 strcpy(c->x86_model_id, p);
183 185
184 c->x86_max_cores = num_cpu_cores(c); 186 c->x86_max_cores = num_cpu_cores(c);
185 187
186 detect_ht(c); 188 detect_ht(c);
@@ -207,28 +209,29 @@ static void __cpuinit init_intel(struct cpuinfo_x86 *c)
207#endif 209#endif
208 210
209 if (cpu_has_xmm2) 211 if (cpu_has_xmm2)
210 set_bit(X86_FEATURE_LFENCE_RDTSC, c->x86_capability); 212 set_cpu_cap(c, X86_FEATURE_LFENCE_RDTSC);
211 if (c->x86 == 15) { 213 if (c->x86 == 15) {
212 set_bit(X86_FEATURE_P4, c->x86_capability); 214 set_cpu_cap(c, X86_FEATURE_P4);
213 } 215 }
214 if (c->x86 == 6) 216 if (c->x86 == 6)
215 set_bit(X86_FEATURE_P3, c->x86_capability); 217 set_cpu_cap(c, X86_FEATURE_P3);
216 if (cpu_has_ds) { 218 if (cpu_has_ds) {
217 unsigned int l1; 219 unsigned int l1;
218 rdmsr(MSR_IA32_MISC_ENABLE, l1, l2); 220 rdmsr(MSR_IA32_MISC_ENABLE, l1, l2);
219 if (!(l1 & (1<<11))) 221 if (!(l1 & (1<<11)))
220 set_bit(X86_FEATURE_BTS, c->x86_capability); 222 set_cpu_cap(c, X86_FEATURE_BTS);
221 if (!(l1 & (1<<12))) 223 if (!(l1 & (1<<12)))
222 set_bit(X86_FEATURE_PEBS, c->x86_capability); 224 set_cpu_cap(c, X86_FEATURE_PEBS);
223 } 225 }
224 226
225 if (cpu_has_bts) 227 if (cpu_has_bts)
226 ds_init_intel(c); 228 ds_init_intel(c);
227} 229}
228 230
229static unsigned int __cpuinit intel_size_cache(struct cpuinfo_x86 * c, unsigned int size) 231static unsigned int __cpuinit intel_size_cache(struct cpuinfo_x86 *c, unsigned int size)
230{ 232{
231 /* Intel PIII Tualatin. This comes in two flavours. 233 /*
234 * Intel PIII Tualatin. This comes in two flavours.
232 * One has 256kb of cache, the other 512. We have no way 235 * One has 256kb of cache, the other 512. We have no way
233 * to determine which, so we use a boottime override 236 * to determine which, so we use a boottime override
234 * for the 512kb model, and assume 256 otherwise. 237 * for the 512kb model, and assume 256 otherwise.
@@ -240,42 +243,42 @@ static unsigned int __cpuinit intel_size_cache(struct cpuinfo_x86 * c, unsigned
240 243
241static struct cpu_dev intel_cpu_dev __cpuinitdata = { 244static struct cpu_dev intel_cpu_dev __cpuinitdata = {
242 .c_vendor = "Intel", 245 .c_vendor = "Intel",
243 .c_ident = { "GenuineIntel" }, 246 .c_ident = { "GenuineIntel" },
244 .c_models = { 247 .c_models = {
245 { .vendor = X86_VENDOR_INTEL, .family = 4, .model_names = 248 { .vendor = X86_VENDOR_INTEL, .family = 4, .model_names =
246 { 249 {
247 [0] = "486 DX-25/33", 250 [0] = "486 DX-25/33",
248 [1] = "486 DX-50", 251 [1] = "486 DX-50",
249 [2] = "486 SX", 252 [2] = "486 SX",
250 [3] = "486 DX/2", 253 [3] = "486 DX/2",
251 [4] = "486 SL", 254 [4] = "486 SL",
252 [5] = "486 SX/2", 255 [5] = "486 SX/2",
253 [7] = "486 DX/2-WB", 256 [7] = "486 DX/2-WB",
254 [8] = "486 DX/4", 257 [8] = "486 DX/4",
255 [9] = "486 DX/4-WB" 258 [9] = "486 DX/4-WB"
256 } 259 }
257 }, 260 },
258 { .vendor = X86_VENDOR_INTEL, .family = 5, .model_names = 261 { .vendor = X86_VENDOR_INTEL, .family = 5, .model_names =
259 { 262 {
260 [0] = "Pentium 60/66 A-step", 263 [0] = "Pentium 60/66 A-step",
261 [1] = "Pentium 60/66", 264 [1] = "Pentium 60/66",
262 [2] = "Pentium 75 - 200", 265 [2] = "Pentium 75 - 200",
263 [3] = "OverDrive PODP5V83", 266 [3] = "OverDrive PODP5V83",
264 [4] = "Pentium MMX", 267 [4] = "Pentium MMX",
265 [7] = "Mobile Pentium 75 - 200", 268 [7] = "Mobile Pentium 75 - 200",
266 [8] = "Mobile Pentium MMX" 269 [8] = "Mobile Pentium MMX"
267 } 270 }
268 }, 271 },
269 { .vendor = X86_VENDOR_INTEL, .family = 6, .model_names = 272 { .vendor = X86_VENDOR_INTEL, .family = 6, .model_names =
270 { 273 {
271 [0] = "Pentium Pro A-step", 274 [0] = "Pentium Pro A-step",
272 [1] = "Pentium Pro", 275 [1] = "Pentium Pro",
273 [3] = "Pentium II (Klamath)", 276 [3] = "Pentium II (Klamath)",
274 [4] = "Pentium II (Deschutes)", 277 [4] = "Pentium II (Deschutes)",
275 [5] = "Pentium II (Deschutes)", 278 [5] = "Pentium II (Deschutes)",
276 [6] = "Mobile Pentium II", 279 [6] = "Mobile Pentium II",
277 [7] = "Pentium III (Katmai)", 280 [7] = "Pentium III (Katmai)",
278 [8] = "Pentium III (Coppermine)", 281 [8] = "Pentium III (Coppermine)",
279 [10] = "Pentium III (Cascades)", 282 [10] = "Pentium III (Cascades)",
280 [11] = "Pentium III (Tualatin)", 283 [11] = "Pentium III (Tualatin)",
281 } 284 }
@@ -290,15 +293,12 @@ static struct cpu_dev intel_cpu_dev __cpuinitdata = {
290 } 293 }
291 }, 294 },
292 }, 295 },
296 .c_early_init = early_init_intel,
293 .c_init = init_intel, 297 .c_init = init_intel,
294 .c_size_cache = intel_size_cache, 298 .c_size_cache = intel_size_cache,
295}; 299};
296 300
297__init int intel_cpu_init(void) 301cpu_vendor_dev_register(X86_VENDOR_INTEL, &intel_cpu_dev);
298{
299 cpu_devs[X86_VENDOR_INTEL] = &intel_cpu_dev;
300 return 0;
301}
302 302
303#ifndef CONFIG_X86_CMPXCHG 303#ifndef CONFIG_X86_CMPXCHG
304unsigned long cmpxchg_386_u8(volatile void *ptr, u8 old, u8 new) 304unsigned long cmpxchg_386_u8(volatile void *ptr, u8 old, u8 new)
@@ -364,5 +364,5 @@ unsigned long long cmpxchg_486_u64(volatile void *ptr, u64 old, u64 new)
364EXPORT_SYMBOL(cmpxchg_486_u64); 364EXPORT_SYMBOL(cmpxchg_486_u64);
365#endif 365#endif
366 366
367// arch_initcall(intel_cpu_init); 367/* arch_initcall(intel_cpu_init); */
368 368
diff --git a/arch/x86/kernel/cpu/mcheck/mce_32.c b/arch/x86/kernel/cpu/mcheck/mce_32.c
index a5182dcd94ae..774d87cfd8cd 100644
--- a/arch/x86/kernel/cpu/mcheck/mce_32.c
+++ b/arch/x86/kernel/cpu/mcheck/mce_32.c
@@ -10,20 +10,20 @@
10#include <linux/smp.h> 10#include <linux/smp.h>
11#include <linux/thread_info.h> 11#include <linux/thread_info.h>
12 12
13#include <asm/processor.h> 13#include <asm/processor.h>
14#include <asm/system.h> 14#include <asm/system.h>
15#include <asm/mce.h> 15#include <asm/mce.h>
16 16
17#include "mce.h" 17#include "mce.h"
18 18
19int mce_disabled = 0; 19int mce_disabled;
20int nr_mce_banks; 20int nr_mce_banks;
21 21
22EXPORT_SYMBOL_GPL(nr_mce_banks); /* non-fatal.o */ 22EXPORT_SYMBOL_GPL(nr_mce_banks); /* non-fatal.o */
23 23
24/* Handle unconfigured int18 (should never happen) */ 24/* Handle unconfigured int18 (should never happen) */
25static void unexpected_machine_check(struct pt_regs * regs, long error_code) 25static void unexpected_machine_check(struct pt_regs *regs, long error_code)
26{ 26{
27 printk(KERN_ERR "CPU#%d: Unexpected int18 (Machine Check).\n", smp_processor_id()); 27 printk(KERN_ERR "CPU#%d: Unexpected int18 (Machine Check).\n", smp_processor_id());
28} 28}
29 29
@@ -33,30 +33,30 @@ void (*machine_check_vector)(struct pt_regs *, long error_code) = unexpected_mac
33/* This has to be run for each processor */ 33/* This has to be run for each processor */
34void mcheck_init(struct cpuinfo_x86 *c) 34void mcheck_init(struct cpuinfo_x86 *c)
35{ 35{
36 if (mce_disabled==1) 36 if (mce_disabled == 1)
37 return; 37 return;
38 38
39 switch (c->x86_vendor) { 39 switch (c->x86_vendor) {
40 case X86_VENDOR_AMD: 40 case X86_VENDOR_AMD:
41 amd_mcheck_init(c); 41 amd_mcheck_init(c);
42 break; 42 break;
43 43
44 case X86_VENDOR_INTEL: 44 case X86_VENDOR_INTEL:
45 if (c->x86==5) 45 if (c->x86 == 5)
46 intel_p5_mcheck_init(c); 46 intel_p5_mcheck_init(c);
47 if (c->x86==6) 47 if (c->x86 == 6)
48 intel_p6_mcheck_init(c); 48 intel_p6_mcheck_init(c);
49 if (c->x86==15) 49 if (c->x86 == 15)
50 intel_p4_mcheck_init(c); 50 intel_p4_mcheck_init(c);
51 break; 51 break;
52 52
53 case X86_VENDOR_CENTAUR: 53 case X86_VENDOR_CENTAUR:
54 if (c->x86==5) 54 if (c->x86 == 5)
55 winchip_mcheck_init(c); 55 winchip_mcheck_init(c);
56 break; 56 break;
57 57
58 default: 58 default:
59 break; 59 break;
60 } 60 }
61} 61}
62 62
diff --git a/arch/x86/kernel/cpu/mcheck/non-fatal.c b/arch/x86/kernel/cpu/mcheck/non-fatal.c
index bf39409b3838..00ccb6c14ec2 100644
--- a/arch/x86/kernel/cpu/mcheck/non-fatal.c
+++ b/arch/x86/kernel/cpu/mcheck/non-fatal.c
@@ -16,7 +16,7 @@
16#include <linux/smp.h> 16#include <linux/smp.h>
17#include <linux/module.h> 17#include <linux/module.h>
18 18
19#include <asm/processor.h> 19#include <asm/processor.h>
20#include <asm/system.h> 20#include <asm/system.h>
21#include <asm/msr.h> 21#include <asm/msr.h>
22 22
@@ -26,23 +26,26 @@ static int firstbank;
26 26
27#define MCE_RATE 15*HZ /* timer rate is 15s */ 27#define MCE_RATE 15*HZ /* timer rate is 15s */
28 28
29static void mce_checkregs (void *info) 29static void mce_checkregs(void *info)
30{ 30{
31 u32 low, high; 31 u32 low, high;
32 int i; 32 int i;
33 33
34 for (i=firstbank; i<nr_mce_banks; i++) { 34 for (i = firstbank; i < nr_mce_banks; i++) {
35 rdmsr (MSR_IA32_MC0_STATUS+i*4, low, high); 35 rdmsr(MSR_IA32_MC0_STATUS+i*4, low, high);
36 36
37 if (high & (1<<31)) { 37 if (high & (1<<31)) {
38 printk(KERN_INFO "MCE: The hardware reports a non " 38 printk(KERN_INFO "MCE: The hardware reports a non "
39 "fatal, correctable incident occurred on " 39 "fatal, correctable incident occurred on "
40 "CPU %d.\n", 40 "CPU %d.\n",
41 smp_processor_id()); 41 smp_processor_id());
42 printk (KERN_INFO "Bank %d: %08x%08x\n", i, high, low); 42 printk(KERN_INFO "Bank %d: %08x%08x\n", i, high, low);
43 43
44 /* Scrub the error so we don't pick it up in MCE_RATE seconds time. */ 44 /*
45 wrmsr (MSR_IA32_MC0_STATUS+i*4, 0UL, 0UL); 45 * Scrub the error so we don't pick it up in MCE_RATE
46 * seconds time.
47 */
48 wrmsr(MSR_IA32_MC0_STATUS+i*4, 0UL, 0UL);
46 49
47 /* Serialize */ 50 /* Serialize */
48 wmb(); 51 wmb();
@@ -55,10 +58,10 @@ static void mce_work_fn(struct work_struct *work);
55static DECLARE_DELAYED_WORK(mce_work, mce_work_fn); 58static DECLARE_DELAYED_WORK(mce_work, mce_work_fn);
56 59
57static void mce_work_fn(struct work_struct *work) 60static void mce_work_fn(struct work_struct *work)
58{ 61{
59 on_each_cpu(mce_checkregs, NULL, 1, 1); 62 on_each_cpu(mce_checkregs, NULL, 1, 1);
60 schedule_delayed_work(&mce_work, round_jiffies_relative(MCE_RATE)); 63 schedule_delayed_work(&mce_work, round_jiffies_relative(MCE_RATE));
61} 64}
62 65
63static int __init init_nonfatal_mce_checker(void) 66static int __init init_nonfatal_mce_checker(void)
64{ 67{
diff --git a/arch/x86/kernel/cpu/mcheck/p5.c b/arch/x86/kernel/cpu/mcheck/p5.c
index a18310aaae0c..bfa5817afdda 100644
--- a/arch/x86/kernel/cpu/mcheck/p5.c
+++ b/arch/x86/kernel/cpu/mcheck/p5.c
@@ -9,20 +9,20 @@
9#include <linux/interrupt.h> 9#include <linux/interrupt.h>
10#include <linux/smp.h> 10#include <linux/smp.h>
11 11
12#include <asm/processor.h> 12#include <asm/processor.h>
13#include <asm/system.h> 13#include <asm/system.h>
14#include <asm/msr.h> 14#include <asm/msr.h>
15 15
16#include "mce.h" 16#include "mce.h"
17 17
18/* Machine check handler for Pentium class Intel */ 18/* Machine check handler for Pentium class Intel */
19static void pentium_machine_check(struct pt_regs * regs, long error_code) 19static void pentium_machine_check(struct pt_regs *regs, long error_code)
20{ 20{
21 u32 loaddr, hi, lotype; 21 u32 loaddr, hi, lotype;
22 rdmsr(MSR_IA32_P5_MC_ADDR, loaddr, hi); 22 rdmsr(MSR_IA32_P5_MC_ADDR, loaddr, hi);
23 rdmsr(MSR_IA32_P5_MC_TYPE, lotype, hi); 23 rdmsr(MSR_IA32_P5_MC_TYPE, lotype, hi);
24 printk(KERN_EMERG "CPU#%d: Machine Check Exception: 0x%8X (type 0x%8X).\n", smp_processor_id(), loaddr, lotype); 24 printk(KERN_EMERG "CPU#%d: Machine Check Exception: 0x%8X (type 0x%8X).\n", smp_processor_id(), loaddr, lotype);
25 if(lotype&(1<<5)) 25 if (lotype&(1<<5))
26 printk(KERN_EMERG "CPU#%d: Possible thermal failure (CPU on fire ?).\n", smp_processor_id()); 26 printk(KERN_EMERG "CPU#%d: Possible thermal failure (CPU on fire ?).\n", smp_processor_id());
27 add_taint(TAINT_MACHINE_CHECK); 27 add_taint(TAINT_MACHINE_CHECK);
28} 28}
@@ -31,13 +31,13 @@ static void pentium_machine_check(struct pt_regs * regs, long error_code)
31void intel_p5_mcheck_init(struct cpuinfo_x86 *c) 31void intel_p5_mcheck_init(struct cpuinfo_x86 *c)
32{ 32{
33 u32 l, h; 33 u32 l, h;
34 34
35 /*Check for MCE support */ 35 /*Check for MCE support */
36 if( !cpu_has(c, X86_FEATURE_MCE) ) 36 if (!cpu_has(c, X86_FEATURE_MCE))
37 return; 37 return;
38 38
39 /* Default P5 to off as its often misconnected */ 39 /* Default P5 to off as its often misconnected */
40 if(mce_disabled != -1) 40 if (mce_disabled != -1)
41 return; 41 return;
42 machine_check_vector = pentium_machine_check; 42 machine_check_vector = pentium_machine_check;
43 wmb(); 43 wmb();
@@ -47,7 +47,7 @@ void intel_p5_mcheck_init(struct cpuinfo_x86 *c)
47 rdmsr(MSR_IA32_P5_MC_TYPE, l, h); 47 rdmsr(MSR_IA32_P5_MC_TYPE, l, h);
48 printk(KERN_INFO "Intel old style machine check architecture supported.\n"); 48 printk(KERN_INFO "Intel old style machine check architecture supported.\n");
49 49
50 /* Enable MCE */ 50 /* Enable MCE */
51 set_in_cr4(X86_CR4_MCE); 51 set_in_cr4(X86_CR4_MCE);
52 printk(KERN_INFO "Intel old style machine check reporting enabled on CPU#%d.\n", smp_processor_id()); 52 printk(KERN_INFO "Intel old style machine check reporting enabled on CPU#%d.\n", smp_processor_id());
53} 53}
diff --git a/arch/x86/kernel/cpu/mcheck/p6.c b/arch/x86/kernel/cpu/mcheck/p6.c
index 74342604d30e..62efc9c2b3af 100644
--- a/arch/x86/kernel/cpu/mcheck/p6.c
+++ b/arch/x86/kernel/cpu/mcheck/p6.c
@@ -9,23 +9,23 @@
9#include <linux/interrupt.h> 9#include <linux/interrupt.h>
10#include <linux/smp.h> 10#include <linux/smp.h>
11 11
12#include <asm/processor.h> 12#include <asm/processor.h>
13#include <asm/system.h> 13#include <asm/system.h>
14#include <asm/msr.h> 14#include <asm/msr.h>
15 15
16#include "mce.h" 16#include "mce.h"
17 17
18/* Machine Check Handler For PII/PIII */ 18/* Machine Check Handler For PII/PIII */
19static void intel_machine_check(struct pt_regs * regs, long error_code) 19static void intel_machine_check(struct pt_regs *regs, long error_code)
20{ 20{
21 int recover=1; 21 int recover = 1;
22 u32 alow, ahigh, high, low; 22 u32 alow, ahigh, high, low;
23 u32 mcgstl, mcgsth; 23 u32 mcgstl, mcgsth;
24 int i; 24 int i;
25 25
26 rdmsr (MSR_IA32_MCG_STATUS, mcgstl, mcgsth); 26 rdmsr(MSR_IA32_MCG_STATUS, mcgstl, mcgsth);
27 if (mcgstl & (1<<0)) /* Recoverable ? */ 27 if (mcgstl & (1<<0)) /* Recoverable ? */
28 recover=0; 28 recover = 0;
29 29
30 printk(KERN_EMERG "CPU %d: Machine Check Exception: %08x%08x\n", 30 printk(KERN_EMERG "CPU %d: Machine Check Exception: %08x%08x\n",
31 smp_processor_id(), mcgsth, mcgstl); 31 smp_processor_id(), mcgsth, mcgstl);
@@ -55,30 +55,30 @@ static void intel_machine_check(struct pt_regs * regs, long error_code)
55 } 55 }
56 56
57 if (recover & 2) 57 if (recover & 2)
58 panic ("CPU context corrupt"); 58 panic("CPU context corrupt");
59 if (recover & 1) 59 if (recover & 1)
60 panic ("Unable to continue"); 60 panic("Unable to continue");
61 61
62 printk (KERN_EMERG "Attempting to continue.\n"); 62 printk(KERN_EMERG "Attempting to continue.\n");
63 /* 63 /*
64 * Do not clear the MSR_IA32_MCi_STATUS if the error is not 64 * Do not clear the MSR_IA32_MCi_STATUS if the error is not
65 * recoverable/continuable.This will allow BIOS to look at the MSRs 65 * recoverable/continuable.This will allow BIOS to look at the MSRs
66 * for errors if the OS could not log the error. 66 * for errors if the OS could not log the error.
67 */ 67 */
68 for (i=0; i<nr_mce_banks; i++) { 68 for (i = 0; i < nr_mce_banks; i++) {
69 unsigned int msr; 69 unsigned int msr;
70 msr = MSR_IA32_MC0_STATUS+i*4; 70 msr = MSR_IA32_MC0_STATUS+i*4;
71 rdmsr (msr,low, high); 71 rdmsr(msr, low, high);
72 if (high & (1<<31)) { 72 if (high & (1<<31)) {
73 /* Clear it */ 73 /* Clear it */
74 wrmsr (msr, 0UL, 0UL); 74 wrmsr(msr, 0UL, 0UL);
75 /* Serialize */ 75 /* Serialize */
76 wmb(); 76 wmb();
77 add_taint(TAINT_MACHINE_CHECK); 77 add_taint(TAINT_MACHINE_CHECK);
78 } 78 }
79 } 79 }
80 mcgstl &= ~(1<<2); 80 mcgstl &= ~(1<<2);
81 wrmsr (MSR_IA32_MCG_STATUS,mcgstl, mcgsth); 81 wrmsr(MSR_IA32_MCG_STATUS, mcgstl, mcgsth);
82} 82}
83 83
84/* Set up machine check reporting for processors with Intel style MCE */ 84/* Set up machine check reporting for processors with Intel style MCE */
@@ -86,21 +86,21 @@ void intel_p6_mcheck_init(struct cpuinfo_x86 *c)
86{ 86{
87 u32 l, h; 87 u32 l, h;
88 int i; 88 int i;
89 89
90 /* Check for MCE support */ 90 /* Check for MCE support */
91 if (!cpu_has(c, X86_FEATURE_MCE)) 91 if (!cpu_has(c, X86_FEATURE_MCE))
92 return; 92 return;
93 93
94 /* Check for PPro style MCA */ 94 /* Check for PPro style MCA */
95 if (!cpu_has(c, X86_FEATURE_MCA)) 95 if (!cpu_has(c, X86_FEATURE_MCA))
96 return; 96 return;
97 97
98 /* Ok machine check is available */ 98 /* Ok machine check is available */
99 machine_check_vector = intel_machine_check; 99 machine_check_vector = intel_machine_check;
100 wmb(); 100 wmb();
101 101
102 printk (KERN_INFO "Intel machine check architecture supported.\n"); 102 printk(KERN_INFO "Intel machine check architecture supported.\n");
103 rdmsr (MSR_IA32_MCG_CAP, l, h); 103 rdmsr(MSR_IA32_MCG_CAP, l, h);
104 if (l & (1<<8)) /* Control register present ? */ 104 if (l & (1<<8)) /* Control register present ? */
105 wrmsr(MSR_IA32_MCG_CTL, 0xffffffff, 0xffffffff); 105 wrmsr(MSR_IA32_MCG_CTL, 0xffffffff, 0xffffffff);
106 nr_mce_banks = l & 0xff; 106 nr_mce_banks = l & 0xff;
@@ -110,13 +110,13 @@ void intel_p6_mcheck_init(struct cpuinfo_x86 *c)
110 * - MC0_CTL should not be written 110 * - MC0_CTL should not be written
111 * - Status registers on all banks should be cleared on reset 111 * - Status registers on all banks should be cleared on reset
112 */ 112 */
113 for (i=1; i<nr_mce_banks; i++) 113 for (i = 1; i < nr_mce_banks; i++)
114 wrmsr (MSR_IA32_MC0_CTL+4*i, 0xffffffff, 0xffffffff); 114 wrmsr(MSR_IA32_MC0_CTL+4*i, 0xffffffff, 0xffffffff);
115 115
116 for (i=0; i<nr_mce_banks; i++) 116 for (i = 0; i < nr_mce_banks; i++)
117 wrmsr (MSR_IA32_MC0_STATUS+4*i, 0x0, 0x0); 117 wrmsr(MSR_IA32_MC0_STATUS+4*i, 0x0, 0x0);
118 118
119 set_in_cr4 (X86_CR4_MCE); 119 set_in_cr4(X86_CR4_MCE);
120 printk (KERN_INFO "Intel machine check reporting enabled on CPU#%d.\n", 120 printk(KERN_INFO "Intel machine check reporting enabled on CPU#%d.\n",
121 smp_processor_id()); 121 smp_processor_id());
122} 122}
diff --git a/arch/x86/kernel/cpu/mcheck/winchip.c b/arch/x86/kernel/cpu/mcheck/winchip.c
index 3d428d5afc52..f2be3e190c6b 100644
--- a/arch/x86/kernel/cpu/mcheck/winchip.c
+++ b/arch/x86/kernel/cpu/mcheck/winchip.c
@@ -8,14 +8,14 @@
8#include <linux/kernel.h> 8#include <linux/kernel.h>
9#include <linux/interrupt.h> 9#include <linux/interrupt.h>
10 10
11#include <asm/processor.h> 11#include <asm/processor.h>
12#include <asm/system.h> 12#include <asm/system.h>
13#include <asm/msr.h> 13#include <asm/msr.h>
14 14
15#include "mce.h" 15#include "mce.h"
16 16
17/* Machine check handler for WinChip C6 */ 17/* Machine check handler for WinChip C6 */
18static void winchip_machine_check(struct pt_regs * regs, long error_code) 18static void winchip_machine_check(struct pt_regs *regs, long error_code)
19{ 19{
20 printk(KERN_EMERG "CPU0: Machine Check Exception.\n"); 20 printk(KERN_EMERG "CPU0: Machine Check Exception.\n");
21 add_taint(TAINT_MACHINE_CHECK); 21 add_taint(TAINT_MACHINE_CHECK);
@@ -28,8 +28,8 @@ void winchip_mcheck_init(struct cpuinfo_x86 *c)
28 machine_check_vector = winchip_machine_check; 28 machine_check_vector = winchip_machine_check;
29 wmb(); 29 wmb();
30 rdmsr(MSR_IDT_FCR1, lo, hi); 30 rdmsr(MSR_IDT_FCR1, lo, hi);
31 lo|= (1<<2); /* Enable EIERRINT (int 18 MCE) */ 31 lo |= (1<<2); /* Enable EIERRINT (int 18 MCE) */
32 lo&= ~(1<<4); /* Enable MCE */ 32 lo &= ~(1<<4); /* Enable MCE */
33 wrmsr(MSR_IDT_FCR1, lo, hi); 33 wrmsr(MSR_IDT_FCR1, lo, hi);
34 set_in_cr4(X86_CR4_MCE); 34 set_in_cr4(X86_CR4_MCE);
35 printk(KERN_INFO "Winchip machine check reporting enabled on CPU#0.\n"); 35 printk(KERN_INFO "Winchip machine check reporting enabled on CPU#0.\n");
diff --git a/arch/x86/kernel/cpu/mtrr/generic.c b/arch/x86/kernel/cpu/mtrr/generic.c
index 3e18db4cefee..353efe4f5017 100644
--- a/arch/x86/kernel/cpu/mtrr/generic.c
+++ b/arch/x86/kernel/cpu/mtrr/generic.c
@@ -11,6 +11,7 @@
11#include <asm/cpufeature.h> 11#include <asm/cpufeature.h>
12#include <asm/processor-flags.h> 12#include <asm/processor-flags.h>
13#include <asm/tlbflush.h> 13#include <asm/tlbflush.h>
14#include <asm/pat.h>
14#include "mtrr.h" 15#include "mtrr.h"
15 16
16struct mtrr_state { 17struct mtrr_state {
@@ -35,6 +36,8 @@ static struct fixed_range_block fixed_range_blocks[] = {
35 36
36static unsigned long smp_changes_mask; 37static unsigned long smp_changes_mask;
37static struct mtrr_state mtrr_state = {}; 38static struct mtrr_state mtrr_state = {};
39static int mtrr_state_set;
40static u64 tom2;
38 41
39#undef MODULE_PARAM_PREFIX 42#undef MODULE_PARAM_PREFIX
40#define MODULE_PARAM_PREFIX "mtrr." 43#define MODULE_PARAM_PREFIX "mtrr."
@@ -42,6 +45,111 @@ static struct mtrr_state mtrr_state = {};
42static int mtrr_show; 45static int mtrr_show;
43module_param_named(show, mtrr_show, bool, 0); 46module_param_named(show, mtrr_show, bool, 0);
44 47
48/*
49 * Returns the effective MTRR type for the region
50 * Error returns:
51 * - 0xFE - when the range is "not entirely covered" by _any_ var range MTRR
52 * - 0xFF - when MTRR is not enabled
53 */
54u8 mtrr_type_lookup(u64 start, u64 end)
55{
56 int i;
57 u64 base, mask;
58 u8 prev_match, curr_match;
59
60 if (!mtrr_state_set)
61 return 0xFF;
62
63 if (!mtrr_state.enabled)
64 return 0xFF;
65
66 /* Make end inclusive end, instead of exclusive */
67 end--;
68
69 /* Look in fixed ranges. Just return the type as per start */
70 if (mtrr_state.have_fixed && (start < 0x100000)) {
71 int idx;
72
73 if (start < 0x80000) {
74 idx = 0;
75 idx += (start >> 16);
76 return mtrr_state.fixed_ranges[idx];
77 } else if (start < 0xC0000) {
78 idx = 1 * 8;
79 idx += ((start - 0x80000) >> 14);
80 return mtrr_state.fixed_ranges[idx];
81 } else if (start < 0x1000000) {
82 idx = 3 * 8;
83 idx += ((start - 0xC0000) >> 12);
84 return mtrr_state.fixed_ranges[idx];
85 }
86 }
87
88 /*
89 * Look in variable ranges
90 * Look of multiple ranges matching this address and pick type
91 * as per MTRR precedence
92 */
93 if (!mtrr_state.enabled & 2) {
94 return mtrr_state.def_type;
95 }
96
97 prev_match = 0xFF;
98 for (i = 0; i < num_var_ranges; ++i) {
99 unsigned short start_state, end_state;
100
101 if (!(mtrr_state.var_ranges[i].mask_lo & (1 << 11)))
102 continue;
103
104 base = (((u64)mtrr_state.var_ranges[i].base_hi) << 32) +
105 (mtrr_state.var_ranges[i].base_lo & PAGE_MASK);
106 mask = (((u64)mtrr_state.var_ranges[i].mask_hi) << 32) +
107 (mtrr_state.var_ranges[i].mask_lo & PAGE_MASK);
108
109 start_state = ((start & mask) == (base & mask));
110 end_state = ((end & mask) == (base & mask));
111 if (start_state != end_state)
112 return 0xFE;
113
114 if ((start & mask) != (base & mask)) {
115 continue;
116 }
117
118 curr_match = mtrr_state.var_ranges[i].base_lo & 0xff;
119 if (prev_match == 0xFF) {
120 prev_match = curr_match;
121 continue;
122 }
123
124 if (prev_match == MTRR_TYPE_UNCACHABLE ||
125 curr_match == MTRR_TYPE_UNCACHABLE) {
126 return MTRR_TYPE_UNCACHABLE;
127 }
128
129 if ((prev_match == MTRR_TYPE_WRBACK &&
130 curr_match == MTRR_TYPE_WRTHROUGH) ||
131 (prev_match == MTRR_TYPE_WRTHROUGH &&
132 curr_match == MTRR_TYPE_WRBACK)) {
133 prev_match = MTRR_TYPE_WRTHROUGH;
134 curr_match = MTRR_TYPE_WRTHROUGH;
135 }
136
137 if (prev_match != curr_match) {
138 return MTRR_TYPE_UNCACHABLE;
139 }
140 }
141
142 if (tom2) {
143 if (start >= (1ULL<<32) && (end < tom2))
144 return MTRR_TYPE_WRBACK;
145 }
146
147 if (prev_match != 0xFF)
148 return prev_match;
149
150 return mtrr_state.def_type;
151}
152
45/* Get the MSR pair relating to a var range */ 153/* Get the MSR pair relating to a var range */
46static void 154static void
47get_mtrr_var_range(unsigned int index, struct mtrr_var_range *vr) 155get_mtrr_var_range(unsigned int index, struct mtrr_var_range *vr)
@@ -79,12 +187,16 @@ static void print_fixed(unsigned base, unsigned step, const mtrr_type*types)
79 base, base + step - 1, mtrr_attrib_to_str(*types)); 187 base, base + step - 1, mtrr_attrib_to_str(*types));
80} 188}
81 189
190static void prepare_set(void);
191static void post_set(void);
192
82/* Grab all of the MTRR state for this CPU into *state */ 193/* Grab all of the MTRR state for this CPU into *state */
83void __init get_mtrr_state(void) 194void __init get_mtrr_state(void)
84{ 195{
85 unsigned int i; 196 unsigned int i;
86 struct mtrr_var_range *vrs; 197 struct mtrr_var_range *vrs;
87 unsigned lo, dummy; 198 unsigned lo, dummy;
199 unsigned long flags;
88 200
89 vrs = mtrr_state.var_ranges; 201 vrs = mtrr_state.var_ranges;
90 202
@@ -100,6 +212,15 @@ void __init get_mtrr_state(void)
100 mtrr_state.def_type = (lo & 0xff); 212 mtrr_state.def_type = (lo & 0xff);
101 mtrr_state.enabled = (lo & 0xc00) >> 10; 213 mtrr_state.enabled = (lo & 0xc00) >> 10;
102 214
215 if (amd_special_default_mtrr()) {
216 unsigned lo, hi;
217 /* TOP_MEM2 */
218 rdmsr(MSR_K8_TOP_MEM2, lo, hi);
219 tom2 = hi;
220 tom2 <<= 32;
221 tom2 |= lo;
222 tom2 &= 0xffffff8000000ULL;
223 }
103 if (mtrr_show) { 224 if (mtrr_show) {
104 int high_width; 225 int high_width;
105 226
@@ -130,7 +251,22 @@ void __init get_mtrr_state(void)
130 else 251 else
131 printk(KERN_INFO "MTRR %u disabled\n", i); 252 printk(KERN_INFO "MTRR %u disabled\n", i);
132 } 253 }
254 if (tom2) {
255 printk(KERN_INFO "TOM2: %016llx aka %lldM\n",
256 tom2, tom2>>20);
257 }
133 } 258 }
259 mtrr_state_set = 1;
260
261 /* PAT setup for BP. We need to go through sync steps here */
262 local_irq_save(flags);
263 prepare_set();
264
265 pat_init();
266
267 post_set();
268 local_irq_restore(flags);
269
134} 270}
135 271
136/* Some BIOS's are fucked and don't set all MTRRs the same! */ 272/* Some BIOS's are fucked and don't set all MTRRs the same! */
@@ -397,6 +533,9 @@ static void generic_set_all(void)
397 /* Actually set the state */ 533 /* Actually set the state */
398 mask = set_mtrr_state(); 534 mask = set_mtrr_state();
399 535
536 /* also set PAT */
537 pat_init();
538
400 post_set(); 539 post_set();
401 local_irq_restore(flags); 540 local_irq_restore(flags);
402 541
diff --git a/arch/x86/kernel/cpu/mtrr/if.c b/arch/x86/kernel/cpu/mtrr/if.c
index 91e150acb46c..1960f1985e5e 100644
--- a/arch/x86/kernel/cpu/mtrr/if.c
+++ b/arch/x86/kernel/cpu/mtrr/if.c
@@ -424,11 +424,10 @@ static int __init mtrr_if_init(void)
424 return -ENODEV; 424 return -ENODEV;
425 425
426 proc_root_mtrr = 426 proc_root_mtrr =
427 create_proc_entry("mtrr", S_IWUSR | S_IRUGO, &proc_root); 427 proc_create("mtrr", S_IWUSR | S_IRUGO, &proc_root, &mtrr_fops);
428 if (proc_root_mtrr) { 428
429 if (proc_root_mtrr)
429 proc_root_mtrr->owner = THIS_MODULE; 430 proc_root_mtrr->owner = THIS_MODULE;
430 proc_root_mtrr->proc_fops = &mtrr_fops;
431 }
432 return 0; 431 return 0;
433} 432}
434 433
diff --git a/arch/x86/kernel/cpu/mtrr/main.c b/arch/x86/kernel/cpu/mtrr/main.c
index a6450b3ae759..6a1e278d9323 100644
--- a/arch/x86/kernel/cpu/mtrr/main.c
+++ b/arch/x86/kernel/cpu/mtrr/main.c
@@ -627,7 +627,7 @@ early_param("disable_mtrr_trim", disable_mtrr_trim_setup);
627#define Tom2Enabled (1U << 21) 627#define Tom2Enabled (1U << 21)
628#define Tom2ForceMemTypeWB (1U << 22) 628#define Tom2ForceMemTypeWB (1U << 22)
629 629
630static __init int amd_special_default_mtrr(void) 630int __init amd_special_default_mtrr(void)
631{ 631{
632 u32 l, h; 632 u32 l, h;
633 633
diff --git a/arch/x86/kernel/cpu/mtrr/state.c b/arch/x86/kernel/cpu/mtrr/state.c
index 9f8ba923d1c9..7f7e2753685b 100644
--- a/arch/x86/kernel/cpu/mtrr/state.c
+++ b/arch/x86/kernel/cpu/mtrr/state.c
@@ -19,13 +19,15 @@ void set_mtrr_prepare_save(struct set_mtrr_context *ctxt)
19 if (use_intel() || is_cpu(CYRIX)) { 19 if (use_intel() || is_cpu(CYRIX)) {
20 20
21 /* Save value of CR4 and clear Page Global Enable (bit 7) */ 21 /* Save value of CR4 and clear Page Global Enable (bit 7) */
22 if ( cpu_has_pge ) { 22 if (cpu_has_pge) {
23 ctxt->cr4val = read_cr4(); 23 ctxt->cr4val = read_cr4();
24 write_cr4(ctxt->cr4val & ~X86_CR4_PGE); 24 write_cr4(ctxt->cr4val & ~X86_CR4_PGE);
25 } 25 }
26 26
27 /* Disable and flush caches. Note that wbinvd flushes the TLBs as 27 /*
28 a side-effect */ 28 * Disable and flush caches. Note that wbinvd flushes the TLBs
29 * as a side-effect
30 */
29 cr0 = read_cr0() | X86_CR0_CD; 31 cr0 = read_cr0() | X86_CR0_CD;
30 wbinvd(); 32 wbinvd();
31 write_cr0(cr0); 33 write_cr0(cr0);
@@ -42,7 +44,7 @@ void set_mtrr_prepare_save(struct set_mtrr_context *ctxt)
42 44
43void set_mtrr_cache_disable(struct set_mtrr_context *ctxt) 45void set_mtrr_cache_disable(struct set_mtrr_context *ctxt)
44{ 46{
45 if (use_intel()) 47 if (use_intel())
46 /* Disable MTRRs, and set the default type to uncached */ 48 /* Disable MTRRs, and set the default type to uncached */
47 mtrr_wrmsr(MTRRdefType_MSR, ctxt->deftype_lo & 0xf300UL, 49 mtrr_wrmsr(MTRRdefType_MSR, ctxt->deftype_lo & 0xf300UL,
48 ctxt->deftype_hi); 50 ctxt->deftype_hi);
@@ -66,12 +68,12 @@ void set_mtrr_done(struct set_mtrr_context *ctxt)
66 else 68 else
67 /* Cyrix ARRs - everything else was excluded at the top */ 69 /* Cyrix ARRs - everything else was excluded at the top */
68 setCx86(CX86_CCR3, ctxt->ccr3); 70 setCx86(CX86_CCR3, ctxt->ccr3);
69 71
70 /* Enable caches */ 72 /* Enable caches */
71 write_cr0(read_cr0() & 0xbfffffff); 73 write_cr0(read_cr0() & 0xbfffffff);
72 74
73 /* Restore value of CR4 */ 75 /* Restore value of CR4 */
74 if ( cpu_has_pge ) 76 if (cpu_has_pge)
75 write_cr4(ctxt->cr4val); 77 write_cr4(ctxt->cr4val);
76 } 78 }
77 /* Re-enable interrupts locally (if enabled previously) */ 79 /* Re-enable interrupts locally (if enabled previously) */
diff --git a/arch/x86/kernel/cpu/nexgen.c b/arch/x86/kernel/cpu/nexgen.c
index 961fbe1a748f..5d5e1c134123 100644
--- a/arch/x86/kernel/cpu/nexgen.c
+++ b/arch/x86/kernel/cpu/nexgen.c
@@ -9,11 +9,11 @@
9 * Detect a NexGen CPU running without BIOS hypercode new enough 9 * Detect a NexGen CPU running without BIOS hypercode new enough
10 * to have CPUID. (Thanks to Herbert Oppmann) 10 * to have CPUID. (Thanks to Herbert Oppmann)
11 */ 11 */
12 12
13static int __cpuinit deep_magic_nexgen_probe(void) 13static int __cpuinit deep_magic_nexgen_probe(void)
14{ 14{
15 int ret; 15 int ret;
16 16
17 __asm__ __volatile__ ( 17 __asm__ __volatile__ (
18 " movw $0x5555, %%ax\n" 18 " movw $0x5555, %%ax\n"
19 " xorw %%dx,%%dx\n" 19 " xorw %%dx,%%dx\n"
@@ -22,22 +22,21 @@ static int __cpuinit deep_magic_nexgen_probe(void)
22 " movl $0, %%eax\n" 22 " movl $0, %%eax\n"
23 " jnz 1f\n" 23 " jnz 1f\n"
24 " movl $1, %%eax\n" 24 " movl $1, %%eax\n"
25 "1:\n" 25 "1:\n"
26 : "=a" (ret) : : "cx", "dx" ); 26 : "=a" (ret) : : "cx", "dx");
27 return ret; 27 return ret;
28} 28}
29 29
30static void __cpuinit init_nexgen(struct cpuinfo_x86 * c) 30static void __cpuinit init_nexgen(struct cpuinfo_x86 *c)
31{ 31{
32 c->x86_cache_size = 256; /* A few had 1 MB... */ 32 c->x86_cache_size = 256; /* A few had 1 MB... */
33} 33}
34 34
35static void __cpuinit nexgen_identify(struct cpuinfo_x86 * c) 35static void __cpuinit nexgen_identify(struct cpuinfo_x86 *c)
36{ 36{
37 /* Detect NexGen with old hypercode */ 37 /* Detect NexGen with old hypercode */
38 if ( deep_magic_nexgen_probe() ) { 38 if (deep_magic_nexgen_probe())
39 strcpy(c->x86_vendor_id, "NexGenDriven"); 39 strcpy(c->x86_vendor_id, "NexGenDriven");
40 }
41} 40}
42 41
43static struct cpu_dev nexgen_cpu_dev __cpuinitdata = { 42static struct cpu_dev nexgen_cpu_dev __cpuinitdata = {
diff --git a/arch/x86/kernel/cpu/proc.c b/arch/x86/kernel/cpu/proc.c
index af11d31dce0a..0978a4a39418 100644
--- a/arch/x86/kernel/cpu/proc.c
+++ b/arch/x86/kernel/cpu/proc.c
@@ -8,78 +8,139 @@
8/* 8/*
9 * Get CPU information for use by the procfs. 9 * Get CPU information for use by the procfs.
10 */ 10 */
11#ifdef CONFIG_X86_32
12static void show_cpuinfo_core(struct seq_file *m, struct cpuinfo_x86 *c,
13 unsigned int cpu)
14{
15#ifdef CONFIG_X86_HT
16 if (c->x86_max_cores * smp_num_siblings > 1) {
17 seq_printf(m, "physical id\t: %d\n", c->phys_proc_id);
18 seq_printf(m, "siblings\t: %d\n",
19 cpus_weight(per_cpu(cpu_core_map, cpu)));
20 seq_printf(m, "core id\t\t: %d\n", c->cpu_core_id);
21 seq_printf(m, "cpu cores\t: %d\n", c->booted_cores);
22 seq_printf(m, "apicid\t\t: %d\n", c->apicid);
23 seq_printf(m, "initial apicid\t: %d\n", c->initial_apicid);
24 }
25#endif
26}
27
28static void show_cpuinfo_misc(struct seq_file *m, struct cpuinfo_x86 *c)
29{
30 /*
31 * We use exception 16 if we have hardware math and we've either seen
32 * it or the CPU claims it is internal
33 */
34 int fpu_exception = c->hard_math && (ignore_fpu_irq || cpu_has_fpu);
35 seq_printf(m,
36 "fdiv_bug\t: %s\n"
37 "hlt_bug\t\t: %s\n"
38 "f00f_bug\t: %s\n"
39 "coma_bug\t: %s\n"
40 "fpu\t\t: %s\n"
41 "fpu_exception\t: %s\n"
42 "cpuid level\t: %d\n"
43 "wp\t\t: %s\n",
44 c->fdiv_bug ? "yes" : "no",
45 c->hlt_works_ok ? "no" : "yes",
46 c->f00f_bug ? "yes" : "no",
47 c->coma_bug ? "yes" : "no",
48 c->hard_math ? "yes" : "no",
49 fpu_exception ? "yes" : "no",
50 c->cpuid_level,
51 c->wp_works_ok ? "yes" : "no");
52}
53#else
54static void show_cpuinfo_core(struct seq_file *m, struct cpuinfo_x86 *c,
55 unsigned int cpu)
56{
57#ifdef CONFIG_SMP
58 if (c->x86_max_cores * smp_num_siblings > 1) {
59 seq_printf(m, "physical id\t: %d\n", c->phys_proc_id);
60 seq_printf(m, "siblings\t: %d\n",
61 cpus_weight(per_cpu(cpu_core_map, cpu)));
62 seq_printf(m, "core id\t\t: %d\n", c->cpu_core_id);
63 seq_printf(m, "cpu cores\t: %d\n", c->booted_cores);
64 seq_printf(m, "apicid\t\t: %d\n", c->apicid);
65 seq_printf(m, "initial apicid\t: %d\n", c->initial_apicid);
66 }
67#endif
68}
69
70static void show_cpuinfo_misc(struct seq_file *m, struct cpuinfo_x86 *c)
71{
72 seq_printf(m,
73 "fpu\t\t: yes\n"
74 "fpu_exception\t: yes\n"
75 "cpuid level\t: %d\n"
76 "wp\t\t: yes\n",
77 c->cpuid_level);
78}
79#endif
80
11static int show_cpuinfo(struct seq_file *m, void *v) 81static int show_cpuinfo(struct seq_file *m, void *v)
12{ 82{
13 struct cpuinfo_x86 *c = v; 83 struct cpuinfo_x86 *c = v;
14 int i, n = 0; 84 unsigned int cpu = 0;
15 int fpu_exception; 85 int i;
16 86
17#ifdef CONFIG_SMP 87#ifdef CONFIG_SMP
18 n = c->cpu_index; 88 cpu = c->cpu_index;
19#endif 89#endif
20 seq_printf(m, "processor\t: %d\n" 90 seq_printf(m, "processor\t: %u\n"
21 "vendor_id\t: %s\n" 91 "vendor_id\t: %s\n"
22 "cpu family\t: %d\n" 92 "cpu family\t: %d\n"
23 "model\t\t: %d\n" 93 "model\t\t: %u\n"
24 "model name\t: %s\n", 94 "model name\t: %s\n",
25 n, 95 cpu,
26 c->x86_vendor_id[0] ? c->x86_vendor_id : "unknown", 96 c->x86_vendor_id[0] ? c->x86_vendor_id : "unknown",
27 c->x86, 97 c->x86,
28 c->x86_model, 98 c->x86_model,
29 c->x86_model_id[0] ? c->x86_model_id : "unknown"); 99 c->x86_model_id[0] ? c->x86_model_id : "unknown");
30 100
31 if (c->x86_mask || c->cpuid_level >= 0) 101 if (c->x86_mask || c->cpuid_level >= 0)
32 seq_printf(m, "stepping\t: %d\n", c->x86_mask); 102 seq_printf(m, "stepping\t: %d\n", c->x86_mask);
33 else 103 else
34 seq_printf(m, "stepping\t: unknown\n"); 104 seq_printf(m, "stepping\t: unknown\n");
35 105
36 if ( cpu_has(c, X86_FEATURE_TSC) ) { 106 if (cpu_has(c, X86_FEATURE_TSC)) {
37 unsigned int freq = cpufreq_quick_get(n); 107 unsigned int freq = cpufreq_quick_get(cpu);
108
38 if (!freq) 109 if (!freq)
39 freq = cpu_khz; 110 freq = cpu_khz;
40 seq_printf(m, "cpu MHz\t\t: %u.%03u\n", 111 seq_printf(m, "cpu MHz\t\t: %u.%03u\n",
41 freq / 1000, (freq % 1000)); 112 freq / 1000, (freq % 1000));
42 } 113 }
43 114
44 /* Cache size */ 115 /* Cache size */
45 if (c->x86_cache_size >= 0) 116 if (c->x86_cache_size >= 0)
46 seq_printf(m, "cache size\t: %d KB\n", c->x86_cache_size); 117 seq_printf(m, "cache size\t: %d KB\n", c->x86_cache_size);
47#ifdef CONFIG_X86_HT 118
48 if (c->x86_max_cores * smp_num_siblings > 1) { 119 show_cpuinfo_core(m, c, cpu);
49 seq_printf(m, "physical id\t: %d\n", c->phys_proc_id); 120 show_cpuinfo_misc(m, c);
50 seq_printf(m, "siblings\t: %d\n", 121
51 cpus_weight(per_cpu(cpu_core_map, n))); 122 seq_printf(m, "flags\t\t:");
52 seq_printf(m, "core id\t\t: %d\n", c->cpu_core_id); 123 for (i = 0; i < 32*NCAPINTS; i++)
53 seq_printf(m, "cpu cores\t: %d\n", c->booted_cores); 124 if (cpu_has(c, i) && x86_cap_flags[i] != NULL)
54 }
55#endif
56
57 /* We use exception 16 if we have hardware math and we've either seen it or the CPU claims it is internal */
58 fpu_exception = c->hard_math && (ignore_fpu_irq || cpu_has_fpu);
59 seq_printf(m, "fdiv_bug\t: %s\n"
60 "hlt_bug\t\t: %s\n"
61 "f00f_bug\t: %s\n"
62 "coma_bug\t: %s\n"
63 "fpu\t\t: %s\n"
64 "fpu_exception\t: %s\n"
65 "cpuid level\t: %d\n"
66 "wp\t\t: %s\n"
67 "flags\t\t:",
68 c->fdiv_bug ? "yes" : "no",
69 c->hlt_works_ok ? "no" : "yes",
70 c->f00f_bug ? "yes" : "no",
71 c->coma_bug ? "yes" : "no",
72 c->hard_math ? "yes" : "no",
73 fpu_exception ? "yes" : "no",
74 c->cpuid_level,
75 c->wp_works_ok ? "yes" : "no");
76
77 for ( i = 0 ; i < 32*NCAPINTS ; i++ )
78 if ( test_bit(i, c->x86_capability) &&
79 x86_cap_flags[i] != NULL )
80 seq_printf(m, " %s", x86_cap_flags[i]); 125 seq_printf(m, " %s", x86_cap_flags[i]);
81 126
82 for (i = 0; i < 32; i++) 127 seq_printf(m, "\nbogomips\t: %lu.%02lu\n",
128 c->loops_per_jiffy/(500000/HZ),
129 (c->loops_per_jiffy/(5000/HZ)) % 100);
130
131#ifdef CONFIG_X86_64
132 if (c->x86_tlbsize > 0)
133 seq_printf(m, "TLB size\t: %d 4K pages\n", c->x86_tlbsize);
134#endif
135 seq_printf(m, "clflush size\t: %u\n", c->x86_clflush_size);
136#ifdef CONFIG_X86_64
137 seq_printf(m, "cache_alignment\t: %d\n", c->x86_cache_alignment);
138 seq_printf(m, "address sizes\t: %u bits physical, %u bits virtual\n",
139 c->x86_phys_bits, c->x86_virt_bits);
140#endif
141
142 seq_printf(m, "power management:");
143 for (i = 0; i < 32; i++) {
83 if (c->x86_power & (1 << i)) { 144 if (c->x86_power & (1 << i)) {
84 if (i < ARRAY_SIZE(x86_power_flags) && 145 if (i < ARRAY_SIZE(x86_power_flags) &&
85 x86_power_flags[i]) 146 x86_power_flags[i])
@@ -89,11 +150,9 @@ static int show_cpuinfo(struct seq_file *m, void *v)
89 else 150 else
90 seq_printf(m, " [%d]", i); 151 seq_printf(m, " [%d]", i);
91 } 152 }
153 }
92 154
93 seq_printf(m, "\nbogomips\t: %lu.%02lu\n", 155 seq_printf(m, "\n\n");
94 c->loops_per_jiffy/(500000/HZ),
95 (c->loops_per_jiffy/(5000/HZ)) % 100);
96 seq_printf(m, "clflush size\t: %u\n\n", c->x86_clflush_size);
97 156
98 return 0; 157 return 0;
99} 158}
@@ -106,14 +165,17 @@ static void *c_start(struct seq_file *m, loff_t *pos)
106 return &cpu_data(*pos); 165 return &cpu_data(*pos);
107 return NULL; 166 return NULL;
108} 167}
168
109static void *c_next(struct seq_file *m, void *v, loff_t *pos) 169static void *c_next(struct seq_file *m, void *v, loff_t *pos)
110{ 170{
111 *pos = next_cpu(*pos, cpu_online_map); 171 *pos = next_cpu(*pos, cpu_online_map);
112 return c_start(m, pos); 172 return c_start(m, pos);
113} 173}
174
114static void c_stop(struct seq_file *m, void *v) 175static void c_stop(struct seq_file *m, void *v)
115{ 176{
116} 177}
178
117const struct seq_operations cpuinfo_op = { 179const struct seq_operations cpuinfo_op = {
118 .start = c_start, 180 .start = c_start,
119 .next = c_next, 181 .next = c_next,
diff --git a/arch/x86/kernel/cpu/transmeta.c b/arch/x86/kernel/cpu/transmeta.c
index e8b422c1c512..b911a2c61b8f 100644
--- a/arch/x86/kernel/cpu/transmeta.c
+++ b/arch/x86/kernel/cpu/transmeta.c
@@ -18,8 +18,8 @@ static void __cpuinit init_transmeta(struct cpuinfo_x86 *c)
18 /* Print CMS and CPU revision */ 18 /* Print CMS and CPU revision */
19 max = cpuid_eax(0x80860000); 19 max = cpuid_eax(0x80860000);
20 cpu_rev = 0; 20 cpu_rev = 0;
21 if ( max >= 0x80860001 ) { 21 if (max >= 0x80860001) {
22 cpuid(0x80860001, &dummy, &cpu_rev, &cpu_freq, &cpu_flags); 22 cpuid(0x80860001, &dummy, &cpu_rev, &cpu_freq, &cpu_flags);
23 if (cpu_rev != 0x02000000) { 23 if (cpu_rev != 0x02000000) {
24 printk(KERN_INFO "CPU: Processor revision %u.%u.%u.%u, %u MHz\n", 24 printk(KERN_INFO "CPU: Processor revision %u.%u.%u.%u, %u MHz\n",
25 (cpu_rev >> 24) & 0xff, 25 (cpu_rev >> 24) & 0xff,
@@ -29,7 +29,7 @@ static void __cpuinit init_transmeta(struct cpuinfo_x86 *c)
29 cpu_freq); 29 cpu_freq);
30 } 30 }
31 } 31 }
32 if ( max >= 0x80860002 ) { 32 if (max >= 0x80860002) {
33 cpuid(0x80860002, &new_cpu_rev, &cms_rev1, &cms_rev2, &dummy); 33 cpuid(0x80860002, &new_cpu_rev, &cms_rev1, &cms_rev2, &dummy);
34 if (cpu_rev == 0x02000000) { 34 if (cpu_rev == 0x02000000) {
35 printk(KERN_INFO "CPU: Processor revision %08X, %u MHz\n", 35 printk(KERN_INFO "CPU: Processor revision %08X, %u MHz\n",
@@ -42,7 +42,7 @@ static void __cpuinit init_transmeta(struct cpuinfo_x86 *c)
42 cms_rev1 & 0xff, 42 cms_rev1 & 0xff,
43 cms_rev2); 43 cms_rev2);
44 } 44 }
45 if ( max >= 0x80860006 ) { 45 if (max >= 0x80860006) {
46 cpuid(0x80860003, 46 cpuid(0x80860003,
47 (void *)&cpu_info[0], 47 (void *)&cpu_info[0],
48 (void *)&cpu_info[4], 48 (void *)&cpu_info[4],
@@ -74,23 +74,25 @@ static void __cpuinit init_transmeta(struct cpuinfo_x86 *c)
74 wrmsr(0x80860004, cap_mask, uk); 74 wrmsr(0x80860004, cap_mask, uk);
75 75
76 /* All Transmeta CPUs have a constant TSC */ 76 /* All Transmeta CPUs have a constant TSC */
77 set_bit(X86_FEATURE_CONSTANT_TSC, c->x86_capability); 77 set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
78 78
79#ifdef CONFIG_SYSCTL 79#ifdef CONFIG_SYSCTL
80 /* randomize_va_space slows us down enormously; 80 /*
81 it probably triggers retranslation of x86->native bytecode */ 81 * randomize_va_space slows us down enormously;
82 * it probably triggers retranslation of x86->native bytecode
83 */
82 randomize_va_space = 0; 84 randomize_va_space = 0;
83#endif 85#endif
84} 86}
85 87
86static void __cpuinit transmeta_identify(struct cpuinfo_x86 * c) 88static void __cpuinit transmeta_identify(struct cpuinfo_x86 *c)
87{ 89{
88 u32 xlvl; 90 u32 xlvl;
89 91
90 /* Transmeta-defined flags: level 0x80860001 */ 92 /* Transmeta-defined flags: level 0x80860001 */
91 xlvl = cpuid_eax(0x80860000); 93 xlvl = cpuid_eax(0x80860000);
92 if ( (xlvl & 0xffff0000) == 0x80860000 ) { 94 if ((xlvl & 0xffff0000) == 0x80860000) {
93 if ( xlvl >= 0x80860001 ) 95 if (xlvl >= 0x80860001)
94 c->x86_capability[2] = cpuid_edx(0x80860001); 96 c->x86_capability[2] = cpuid_edx(0x80860001);
95 } 97 }
96} 98}
@@ -102,8 +104,4 @@ static struct cpu_dev transmeta_cpu_dev __cpuinitdata = {
102 .c_identify = transmeta_identify, 104 .c_identify = transmeta_identify,
103}; 105};
104 106
105int __init transmeta_init_cpu(void) 107cpu_vendor_dev_register(X86_VENDOR_TRANSMETA, &transmeta_cpu_dev);
106{
107 cpu_devs[X86_VENDOR_TRANSMETA] = &transmeta_cpu_dev;
108 return 0;
109}
diff --git a/arch/x86/kernel/cpu/umc.c b/arch/x86/kernel/cpu/umc.c
index a7a4e75bdcd7..b1fc90989d75 100644
--- a/arch/x86/kernel/cpu/umc.c
+++ b/arch/x86/kernel/cpu/umc.c
@@ -3,24 +3,23 @@
3#include <asm/processor.h> 3#include <asm/processor.h>
4#include "cpu.h" 4#include "cpu.h"
5 5
6/* UMC chips appear to be only either 386 or 486, so no special init takes place. 6/*
7 * UMC chips appear to be only either 386 or 486,
8 * so no special init takes place.
7 */ 9 */
8 10
9static struct cpu_dev umc_cpu_dev __cpuinitdata = { 11static struct cpu_dev umc_cpu_dev __cpuinitdata = {
10 .c_vendor = "UMC", 12 .c_vendor = "UMC",
11 .c_ident = { "UMC UMC UMC" }, 13 .c_ident = { "UMC UMC UMC" },
12 .c_models = { 14 .c_models = {
13 { .vendor = X86_VENDOR_UMC, .family = 4, .model_names = 15 { .vendor = X86_VENDOR_UMC, .family = 4, .model_names =
14 { 16 {
15 [1] = "U5D", 17 [1] = "U5D",
16 [2] = "U5S", 18 [2] = "U5S",
17 } 19 }
18 }, 20 },
19 }, 21 },
20}; 22};
21 23
22int __init umc_init_cpu(void) 24cpu_vendor_dev_register(X86_VENDOR_UMC, &umc_cpu_dev);
23{ 25
24 cpu_devs[X86_VENDOR_UMC] = &umc_cpu_dev;
25 return 0;
26}
diff --git a/arch/x86/kernel/crash.c b/arch/x86/kernel/crash.c
index 9a5fa0abfcc7..2251d0ae9570 100644
--- a/arch/x86/kernel/crash.c
+++ b/arch/x86/kernel/crash.c
@@ -26,11 +26,7 @@
26#include <linux/kdebug.h> 26#include <linux/kdebug.h>
27#include <asm/smp.h> 27#include <asm/smp.h>
28 28
29#ifdef CONFIG_X86_32
30#include <mach_ipi.h> 29#include <mach_ipi.h>
31#else
32#include <asm/mach_apic.h>
33#endif
34 30
35/* This keeps a track of which one is crashing cpu. */ 31/* This keeps a track of which one is crashing cpu. */
36static int crashing_cpu; 32static int crashing_cpu;
diff --git a/arch/x86/kernel/ds.c b/arch/x86/kernel/ds.c
index dcd918c1580d..11c11b8ec48d 100644
--- a/arch/x86/kernel/ds.c
+++ b/arch/x86/kernel/ds.c
@@ -220,11 +220,11 @@ int ds_allocate(void **dsp, size_t bts_size_in_bytes)
220 220
221int ds_free(void **dsp) 221int ds_free(void **dsp)
222{ 222{
223 if (*dsp) 223 if (*dsp) {
224 kfree((void *)get_bts_buffer_base(*dsp)); 224 kfree((void *)get_bts_buffer_base(*dsp));
225 kfree(*dsp); 225 kfree(*dsp);
226 *dsp = NULL; 226 *dsp = NULL;
227 227 }
228 return 0; 228 return 0;
229} 229}
230 230
diff --git a/arch/x86/kernel/e820_32.c b/arch/x86/kernel/e820_32.c
index 80444c5c9b14..0240cd778365 100644
--- a/arch/x86/kernel/e820_32.c
+++ b/arch/x86/kernel/e820_32.c
@@ -450,38 +450,25 @@ int __init sanitize_e820_map(struct e820entry * biosmap, char * pnr_map)
450 * thinkpad 560x, for example, does not cooperate with the memory 450 * thinkpad 560x, for example, does not cooperate with the memory
451 * detection code.) 451 * detection code.)
452 */ 452 */
453int __init copy_e820_map(struct e820entry * biosmap, int nr_map) 453int __init copy_e820_map(struct e820entry *biosmap, int nr_map)
454{ 454{
455 /* Only one memory region (or negative)? Ignore it */ 455 /* Only one memory region (or negative)? Ignore it */
456 if (nr_map < 2) 456 if (nr_map < 2)
457 return -1; 457 return -1;
458 458
459 do { 459 do {
460 unsigned long long start = biosmap->addr; 460 u64 start = biosmap->addr;
461 unsigned long long size = biosmap->size; 461 u64 size = biosmap->size;
462 unsigned long long end = start + size; 462 u64 end = start + size;
463 unsigned long type = biosmap->type; 463 u32 type = biosmap->type;
464 464
465 /* Overflow in 64 bits? Ignore the memory map. */ 465 /* Overflow in 64 bits? Ignore the memory map. */
466 if (start > end) 466 if (start > end)
467 return -1; 467 return -1;
468 468
469 /*
470 * Some BIOSes claim RAM in the 640k - 1M region.
471 * Not right. Fix it up.
472 */
473 if (type == E820_RAM) {
474 if (start < 0x100000ULL && end > 0xA0000ULL) {
475 if (start < 0xA0000ULL)
476 add_memory_region(start, 0xA0000ULL-start, type);
477 if (end <= 0x100000ULL)
478 continue;
479 start = 0x100000ULL;
480 size = end - start;
481 }
482 }
483 add_memory_region(start, size, type); 469 add_memory_region(start, size, type);
484 } while (biosmap++,--nr_map); 470 } while (biosmap++, --nr_map);
471
485 return 0; 472 return 0;
486} 473}
487 474
diff --git a/arch/x86/kernel/e820_64.c b/arch/x86/kernel/e820_64.c
index 9be697126013..7f6c0c85c8f6 100644
--- a/arch/x86/kernel/e820_64.c
+++ b/arch/x86/kernel/e820_64.c
@@ -27,6 +27,7 @@
27#include <asm/setup.h> 27#include <asm/setup.h>
28#include <asm/sections.h> 28#include <asm/sections.h>
29#include <asm/kdebug.h> 29#include <asm/kdebug.h>
30#include <asm/trampoline.h>
30 31
31struct e820map e820; 32struct e820map e820;
32 33
@@ -36,11 +37,11 @@ struct e820map e820;
36unsigned long end_pfn; 37unsigned long end_pfn;
37 38
38/* 39/*
39 * end_pfn only includes RAM, while end_pfn_map includes all e820 entries. 40 * end_pfn only includes RAM, while max_pfn_mapped includes all e820 entries.
40 * The direct mapping extends to end_pfn_map, so that we can directly access 41 * The direct mapping extends to max_pfn_mapped, so that we can directly access
41 * apertures, ACPI and other tables without having to play with fixmaps. 42 * apertures, ACPI and other tables without having to play with fixmaps.
42 */ 43 */
43unsigned long end_pfn_map; 44unsigned long max_pfn_mapped;
44 45
45/* 46/*
46 * Last pfn which the user wants to use. 47 * Last pfn which the user wants to use.
@@ -58,8 +59,8 @@ struct early_res {
58}; 59};
59static struct early_res early_res[MAX_EARLY_RES] __initdata = { 60static struct early_res early_res[MAX_EARLY_RES] __initdata = {
60 { 0, PAGE_SIZE, "BIOS data page" }, /* BIOS data page */ 61 { 0, PAGE_SIZE, "BIOS data page" }, /* BIOS data page */
61#ifdef CONFIG_SMP 62#ifdef CONFIG_X86_TRAMPOLINE
62 { SMP_TRAMPOLINE_BASE, SMP_TRAMPOLINE_BASE + 2*PAGE_SIZE, "SMP_TRAMPOLINE" }, 63 { TRAMPOLINE_BASE, TRAMPOLINE_BASE + 2 * PAGE_SIZE, "TRAMPOLINE" },
63#endif 64#endif
64 {} 65 {}
65}; 66};
@@ -95,7 +96,8 @@ void __init early_res_to_bootmem(void)
95} 96}
96 97
97/* Check for already reserved areas */ 98/* Check for already reserved areas */
98static inline int bad_addr(unsigned long *addrp, unsigned long size) 99static inline int
100bad_addr(unsigned long *addrp, unsigned long size, unsigned long align)
99{ 101{
100 int i; 102 int i;
101 unsigned long addr = *addrp, last; 103 unsigned long addr = *addrp, last;
@@ -105,7 +107,7 @@ again:
105 for (i = 0; i < MAX_EARLY_RES && early_res[i].end; i++) { 107 for (i = 0; i < MAX_EARLY_RES && early_res[i].end; i++) {
106 struct early_res *r = &early_res[i]; 108 struct early_res *r = &early_res[i];
107 if (last >= r->start && addr < r->end) { 109 if (last >= r->start && addr < r->end) {
108 *addrp = addr = r->end; 110 *addrp = addr = round_up(r->end, align);
109 changed = 1; 111 changed = 1;
110 goto again; 112 goto again;
111 } 113 }
@@ -113,6 +115,40 @@ again:
113 return changed; 115 return changed;
114} 116}
115 117
118/* Check for already reserved areas */
119static inline int
120bad_addr_size(unsigned long *addrp, unsigned long *sizep, unsigned long align)
121{
122 int i;
123 unsigned long addr = *addrp, last;
124 unsigned long size = *sizep;
125 int changed = 0;
126again:
127 last = addr + size;
128 for (i = 0; i < MAX_EARLY_RES && early_res[i].end; i++) {
129 struct early_res *r = &early_res[i];
130 if (last > r->start && addr < r->start) {
131 size = r->start - addr;
132 changed = 1;
133 goto again;
134 }
135 if (last > r->end && addr < r->end) {
136 addr = round_up(r->end, align);
137 size = last - addr;
138 changed = 1;
139 goto again;
140 }
141 if (last <= r->end && addr >= r->start) {
142 (*sizep)++;
143 return 0;
144 }
145 }
146 if (changed) {
147 *addrp = addr;
148 *sizep = size;
149 }
150 return changed;
151}
116/* 152/*
117 * This function checks if any part of the range <start,end> is mapped 153 * This function checks if any part of the range <start,end> is mapped
118 * with type. 154 * with type.
@@ -174,26 +210,27 @@ int __init e820_all_mapped(unsigned long start, unsigned long end,
174 * Find a free area with specified alignment in a specific range. 210 * Find a free area with specified alignment in a specific range.
175 */ 211 */
176unsigned long __init find_e820_area(unsigned long start, unsigned long end, 212unsigned long __init find_e820_area(unsigned long start, unsigned long end,
177 unsigned size, unsigned long align) 213 unsigned long size, unsigned long align)
178{ 214{
179 int i; 215 int i;
180 unsigned long mask = ~(align - 1);
181 216
182 for (i = 0; i < e820.nr_map; i++) { 217 for (i = 0; i < e820.nr_map; i++) {
183 struct e820entry *ei = &e820.map[i]; 218 struct e820entry *ei = &e820.map[i];
184 unsigned long addr = ei->addr, last; 219 unsigned long addr, last;
220 unsigned long ei_last;
185 221
186 if (ei->type != E820_RAM) 222 if (ei->type != E820_RAM)
187 continue; 223 continue;
224 addr = round_up(ei->addr, align);
225 ei_last = ei->addr + ei->size;
188 if (addr < start) 226 if (addr < start)
189 addr = start; 227 addr = round_up(start, align);
190 if (addr > ei->addr + ei->size) 228 if (addr >= ei_last)
191 continue; 229 continue;
192 while (bad_addr(&addr, size) && addr+size <= ei->addr+ei->size) 230 while (bad_addr(&addr, size, align) && addr+size <= ei_last)
193 ; 231 ;
194 addr = (addr + align - 1) & mask;
195 last = addr + size; 232 last = addr + size;
196 if (last > ei->addr + ei->size) 233 if (last > ei_last)
197 continue; 234 continue;
198 if (last > end) 235 if (last > end)
199 continue; 236 continue;
@@ -203,6 +240,40 @@ unsigned long __init find_e820_area(unsigned long start, unsigned long end,
203} 240}
204 241
205/* 242/*
243 * Find next free range after *start
244 */
245unsigned long __init find_e820_area_size(unsigned long start,
246 unsigned long *sizep,
247 unsigned long align)
248{
249 int i;
250
251 for (i = 0; i < e820.nr_map; i++) {
252 struct e820entry *ei = &e820.map[i];
253 unsigned long addr, last;
254 unsigned long ei_last;
255
256 if (ei->type != E820_RAM)
257 continue;
258 addr = round_up(ei->addr, align);
259 ei_last = ei->addr + ei->size;
260 if (addr < start)
261 addr = round_up(start, align);
262 if (addr >= ei_last)
263 continue;
264 *sizep = ei_last - addr;
265 while (bad_addr_size(&addr, sizep, align) &&
266 addr + *sizep <= ei_last)
267 ;
268 last = addr + *sizep;
269 if (last > ei_last)
270 continue;
271 return addr;
272 }
273 return -1UL;
274
275}
276/*
206 * Find the highest page frame number we have available 277 * Find the highest page frame number we have available
207 */ 278 */
208unsigned long __init e820_end_of_ram(void) 279unsigned long __init e820_end_of_ram(void)
@@ -211,29 +282,29 @@ unsigned long __init e820_end_of_ram(void)
211 282
212 end_pfn = find_max_pfn_with_active_regions(); 283 end_pfn = find_max_pfn_with_active_regions();
213 284
214 if (end_pfn > end_pfn_map) 285 if (end_pfn > max_pfn_mapped)
215 end_pfn_map = end_pfn; 286 max_pfn_mapped = end_pfn;
216 if (end_pfn_map > MAXMEM>>PAGE_SHIFT) 287 if (max_pfn_mapped > MAXMEM>>PAGE_SHIFT)
217 end_pfn_map = MAXMEM>>PAGE_SHIFT; 288 max_pfn_mapped = MAXMEM>>PAGE_SHIFT;
218 if (end_pfn > end_user_pfn) 289 if (end_pfn > end_user_pfn)
219 end_pfn = end_user_pfn; 290 end_pfn = end_user_pfn;
220 if (end_pfn > end_pfn_map) 291 if (end_pfn > max_pfn_mapped)
221 end_pfn = end_pfn_map; 292 end_pfn = max_pfn_mapped;
222 293
223 printk(KERN_INFO "end_pfn_map = %lu\n", end_pfn_map); 294 printk(KERN_INFO "max_pfn_mapped = %lu\n", max_pfn_mapped);
224 return end_pfn; 295 return end_pfn;
225} 296}
226 297
227/* 298/*
228 * Mark e820 reserved areas as busy for the resource manager. 299 * Mark e820 reserved areas as busy for the resource manager.
229 */ 300 */
230void __init e820_reserve_resources(struct resource *code_resource, 301void __init e820_reserve_resources(void)
231 struct resource *data_resource, struct resource *bss_resource)
232{ 302{
233 int i; 303 int i;
304 struct resource *res;
305
306 res = alloc_bootmem_low(sizeof(struct resource) * e820.nr_map);
234 for (i = 0; i < e820.nr_map; i++) { 307 for (i = 0; i < e820.nr_map; i++) {
235 struct resource *res;
236 res = alloc_bootmem_low(sizeof(struct resource));
237 switch (e820.map[i].type) { 308 switch (e820.map[i].type) {
238 case E820_RAM: res->name = "System RAM"; break; 309 case E820_RAM: res->name = "System RAM"; break;
239 case E820_ACPI: res->name = "ACPI Tables"; break; 310 case E820_ACPI: res->name = "ACPI Tables"; break;
@@ -243,21 +314,8 @@ void __init e820_reserve_resources(struct resource *code_resource,
243 res->start = e820.map[i].addr; 314 res->start = e820.map[i].addr;
244 res->end = res->start + e820.map[i].size - 1; 315 res->end = res->start + e820.map[i].size - 1;
245 res->flags = IORESOURCE_MEM | IORESOURCE_BUSY; 316 res->flags = IORESOURCE_MEM | IORESOURCE_BUSY;
246 request_resource(&iomem_resource, res); 317 insert_resource(&iomem_resource, res);
247 if (e820.map[i].type == E820_RAM) { 318 res++;
248 /*
249 * We don't know which RAM region contains kernel data,
250 * so we try it repeatedly and let the resource manager
251 * test it.
252 */
253 request_resource(res, code_resource);
254 request_resource(res, data_resource);
255 request_resource(res, bss_resource);
256#ifdef CONFIG_KEXEC
257 if (crashk_res.start != crashk_res.end)
258 request_resource(res, &crashk_res);
259#endif
260 }
261 } 319 }
262} 320}
263 321
@@ -309,9 +367,9 @@ static int __init e820_find_active_region(const struct e820entry *ei,
309 if (*ei_startpfn >= *ei_endpfn) 367 if (*ei_startpfn >= *ei_endpfn)
310 return 0; 368 return 0;
311 369
312 /* Check if end_pfn_map should be updated */ 370 /* Check if max_pfn_mapped should be updated */
313 if (ei->type != E820_RAM && *ei_endpfn > end_pfn_map) 371 if (ei->type != E820_RAM && *ei_endpfn > max_pfn_mapped)
314 end_pfn_map = *ei_endpfn; 372 max_pfn_mapped = *ei_endpfn;
315 373
316 /* Skip if map is outside the node */ 374 /* Skip if map is outside the node */
317 if (ei->type != E820_RAM || *ei_endpfn <= start_pfn || 375 if (ei->type != E820_RAM || *ei_endpfn <= start_pfn ||
@@ -634,10 +692,10 @@ static int __init copy_e820_map(struct e820entry *biosmap, int nr_map)
634 return -1; 692 return -1;
635 693
636 do { 694 do {
637 unsigned long start = biosmap->addr; 695 u64 start = biosmap->addr;
638 unsigned long size = biosmap->size; 696 u64 size = biosmap->size;
639 unsigned long end = start + size; 697 u64 end = start + size;
640 unsigned long type = biosmap->type; 698 u32 type = biosmap->type;
641 699
642 /* Overflow in 64 bits? Ignore the memory map. */ 700 /* Overflow in 64 bits? Ignore the memory map. */
643 if (start > end) 701 if (start > end)
@@ -702,7 +760,7 @@ static int __init parse_memmap_opt(char *p)
702 saved_max_pfn = e820_end_of_ram(); 760 saved_max_pfn = e820_end_of_ram();
703 remove_all_active_ranges(); 761 remove_all_active_ranges();
704#endif 762#endif
705 end_pfn_map = 0; 763 max_pfn_mapped = 0;
706 e820.nr_map = 0; 764 e820.nr_map = 0;
707 userdef = 1; 765 userdef = 1;
708 return 0; 766 return 0;
diff --git a/arch/x86/kernel/early_printk.c b/arch/x86/kernel/early_printk.c
index cff84cd9987f..643fd861b724 100644
--- a/arch/x86/kernel/early_printk.c
+++ b/arch/x86/kernel/early_printk.c
@@ -13,7 +13,7 @@
13#define VGABASE (__ISA_IO_base + 0xb8000) 13#define VGABASE (__ISA_IO_base + 0xb8000)
14 14
15static int max_ypos = 25, max_xpos = 80; 15static int max_ypos = 25, max_xpos = 80;
16static int current_ypos = 25, current_xpos = 0; 16static int current_ypos = 25, current_xpos;
17 17
18static void early_vga_write(struct console *con, const char *str, unsigned n) 18static void early_vga_write(struct console *con, const char *str, unsigned n)
19{ 19{
@@ -108,12 +108,12 @@ static __init void early_serial_init(char *s)
108 108
109 if (*s) { 109 if (*s) {
110 unsigned port; 110 unsigned port;
111 if (!strncmp(s,"0x",2)) { 111 if (!strncmp(s, "0x", 2)) {
112 early_serial_base = simple_strtoul(s, &e, 16); 112 early_serial_base = simple_strtoul(s, &e, 16);
113 } else { 113 } else {
114 static int bases[] = { 0x3f8, 0x2f8 }; 114 static int bases[] = { 0x3f8, 0x2f8 };
115 115
116 if (!strncmp(s,"ttyS",4)) 116 if (!strncmp(s, "ttyS", 4))
117 s += 4; 117 s += 4;
118 port = simple_strtoul(s, &e, 10); 118 port = simple_strtoul(s, &e, 10);
119 if (port > 1 || s == e) 119 if (port > 1 || s == e)
@@ -194,7 +194,7 @@ static struct console simnow_console = {
194 194
195/* Direct interface for emergencies */ 195/* Direct interface for emergencies */
196static struct console *early_console = &early_vga_console; 196static struct console *early_console = &early_vga_console;
197static int early_console_initialized = 0; 197static int early_console_initialized;
198 198
199void early_printk(const char *fmt, ...) 199void early_printk(const char *fmt, ...)
200{ 200{
@@ -202,9 +202,9 @@ void early_printk(const char *fmt, ...)
202 int n; 202 int n;
203 va_list ap; 203 va_list ap;
204 204
205 va_start(ap,fmt); 205 va_start(ap, fmt);
206 n = vscnprintf(buf,512,fmt,ap); 206 n = vscnprintf(buf, 512, fmt, ap);
207 early_console->write(early_console,buf,n); 207 early_console->write(early_console, buf, n);
208 va_end(ap); 208 va_end(ap);
209} 209}
210 210
@@ -229,15 +229,15 @@ static int __init setup_early_printk(char *buf)
229 early_serial_init(buf); 229 early_serial_init(buf);
230 early_console = &early_serial_console; 230 early_console = &early_serial_console;
231 } else if (!strncmp(buf, "vga", 3) 231 } else if (!strncmp(buf, "vga", 3)
232 && boot_params.screen_info.orig_video_isVGA == 1) { 232 && boot_params.screen_info.orig_video_isVGA == 1) {
233 max_xpos = boot_params.screen_info.orig_video_cols; 233 max_xpos = boot_params.screen_info.orig_video_cols;
234 max_ypos = boot_params.screen_info.orig_video_lines; 234 max_ypos = boot_params.screen_info.orig_video_lines;
235 current_ypos = boot_params.screen_info.orig_y; 235 current_ypos = boot_params.screen_info.orig_y;
236 early_console = &early_vga_console; 236 early_console = &early_vga_console;
237 } else if (!strncmp(buf, "simnow", 6)) { 237 } else if (!strncmp(buf, "simnow", 6)) {
238 simnow_init(buf + 6); 238 simnow_init(buf + 6);
239 early_console = &simnow_console; 239 early_console = &simnow_console;
240 keep_early = 1; 240 keep_early = 1;
241#ifdef CONFIG_HVC_XEN 241#ifdef CONFIG_HVC_XEN
242 } else if (!strncmp(buf, "xen", 3)) { 242 } else if (!strncmp(buf, "xen", 3)) {
243 early_console = &xenboot_console; 243 early_console = &xenboot_console;
diff --git a/arch/x86/kernel/entry_32.S b/arch/x86/kernel/entry_32.S
index 4b87c32b639f..9ba49a26dff8 100644
--- a/arch/x86/kernel/entry_32.S
+++ b/arch/x86/kernel/entry_32.S
@@ -51,6 +51,7 @@
51#include <asm/desc.h> 51#include <asm/desc.h>
52#include <asm/percpu.h> 52#include <asm/percpu.h>
53#include <asm/dwarf2.h> 53#include <asm/dwarf2.h>
54#include <asm/processor-flags.h>
54#include "irq_vectors.h" 55#include "irq_vectors.h"
55 56
56/* 57/*
@@ -68,13 +69,6 @@
68 69
69#define nr_syscalls ((syscall_table_size)/4) 70#define nr_syscalls ((syscall_table_size)/4)
70 71
71CF_MASK = 0x00000001
72TF_MASK = 0x00000100
73IF_MASK = 0x00000200
74DF_MASK = 0x00000400
75NT_MASK = 0x00004000
76VM_MASK = 0x00020000
77
78#ifdef CONFIG_PREEMPT 72#ifdef CONFIG_PREEMPT
79#define preempt_stop(clobbers) DISABLE_INTERRUPTS(clobbers); TRACE_IRQS_OFF 73#define preempt_stop(clobbers) DISABLE_INTERRUPTS(clobbers); TRACE_IRQS_OFF
80#else 74#else
@@ -84,7 +78,7 @@ VM_MASK = 0x00020000
84 78
85.macro TRACE_IRQS_IRET 79.macro TRACE_IRQS_IRET
86#ifdef CONFIG_TRACE_IRQFLAGS 80#ifdef CONFIG_TRACE_IRQFLAGS
87 testl $IF_MASK,PT_EFLAGS(%esp) # interrupts off? 81 testl $X86_EFLAGS_IF,PT_EFLAGS(%esp) # interrupts off?
88 jz 1f 82 jz 1f
89 TRACE_IRQS_ON 83 TRACE_IRQS_ON
901: 841:
@@ -246,7 +240,7 @@ ret_from_intr:
246check_userspace: 240check_userspace:
247 movl PT_EFLAGS(%esp), %eax # mix EFLAGS and CS 241 movl PT_EFLAGS(%esp), %eax # mix EFLAGS and CS
248 movb PT_CS(%esp), %al 242 movb PT_CS(%esp), %al
249 andl $(VM_MASK | SEGMENT_RPL_MASK), %eax 243 andl $(X86_EFLAGS_VM | SEGMENT_RPL_MASK), %eax
250 cmpl $USER_RPL, %eax 244 cmpl $USER_RPL, %eax
251 jb resume_kernel # not returning to v8086 or userspace 245 jb resume_kernel # not returning to v8086 or userspace
252 246
@@ -271,7 +265,7 @@ need_resched:
271 movl TI_flags(%ebp), %ecx # need_resched set ? 265 movl TI_flags(%ebp), %ecx # need_resched set ?
272 testb $_TIF_NEED_RESCHED, %cl 266 testb $_TIF_NEED_RESCHED, %cl
273 jz restore_all 267 jz restore_all
274 testl $IF_MASK,PT_EFLAGS(%esp) # interrupts off (exception path) ? 268 testl $X86_EFLAGS_IF,PT_EFLAGS(%esp) # interrupts off (exception path) ?
275 jz restore_all 269 jz restore_all
276 call preempt_schedule_irq 270 call preempt_schedule_irq
277 jmp need_resched 271 jmp need_resched
@@ -291,10 +285,10 @@ ENTRY(ia32_sysenter_target)
291 movl TSS_sysenter_sp0(%esp),%esp 285 movl TSS_sysenter_sp0(%esp),%esp
292sysenter_past_esp: 286sysenter_past_esp:
293 /* 287 /*
294 * No need to follow this irqs on/off section: the syscall 288 * Interrupts are disabled here, but we can't trace it until
295 * disabled irqs and here we enable it straight after entry: 289 * enough kernel state to call TRACE_IRQS_OFF can be called - but
290 * we immediately enable interrupts at that point anyway.
296 */ 291 */
297 ENABLE_INTERRUPTS(CLBR_NONE)
298 pushl $(__USER_DS) 292 pushl $(__USER_DS)
299 CFI_ADJUST_CFA_OFFSET 4 293 CFI_ADJUST_CFA_OFFSET 4
300 /*CFI_REL_OFFSET ss, 0*/ 294 /*CFI_REL_OFFSET ss, 0*/
@@ -302,6 +296,7 @@ sysenter_past_esp:
302 CFI_ADJUST_CFA_OFFSET 4 296 CFI_ADJUST_CFA_OFFSET 4
303 CFI_REL_OFFSET esp, 0 297 CFI_REL_OFFSET esp, 0
304 pushfl 298 pushfl
299 orl $X86_EFLAGS_IF, (%esp)
305 CFI_ADJUST_CFA_OFFSET 4 300 CFI_ADJUST_CFA_OFFSET 4
306 pushl $(__USER_CS) 301 pushl $(__USER_CS)
307 CFI_ADJUST_CFA_OFFSET 4 302 CFI_ADJUST_CFA_OFFSET 4
@@ -315,6 +310,11 @@ sysenter_past_esp:
315 CFI_ADJUST_CFA_OFFSET 4 310 CFI_ADJUST_CFA_OFFSET 4
316 CFI_REL_OFFSET eip, 0 311 CFI_REL_OFFSET eip, 0
317 312
313 pushl %eax
314 CFI_ADJUST_CFA_OFFSET 4
315 SAVE_ALL
316 ENABLE_INTERRUPTS(CLBR_NONE)
317
318/* 318/*
319 * Load the potential sixth argument from user stack. 319 * Load the potential sixth argument from user stack.
320 * Careful about security. 320 * Careful about security.
@@ -322,14 +322,12 @@ sysenter_past_esp:
322 cmpl $__PAGE_OFFSET-3,%ebp 322 cmpl $__PAGE_OFFSET-3,%ebp
323 jae syscall_fault 323 jae syscall_fault
3241: movl (%ebp),%ebp 3241: movl (%ebp),%ebp
325 movl %ebp,PT_EBP(%esp)
325.section __ex_table,"a" 326.section __ex_table,"a"
326 .align 4 327 .align 4
327 .long 1b,syscall_fault 328 .long 1b,syscall_fault
328.previous 329.previous
329 330
330 pushl %eax
331 CFI_ADJUST_CFA_OFFSET 4
332 SAVE_ALL
333 GET_THREAD_INFO(%ebp) 331 GET_THREAD_INFO(%ebp)
334 332
335 /* Note, _TIF_SECCOMP is bit number 8, and so it needs testw and not testb */ 333 /* Note, _TIF_SECCOMP is bit number 8, and so it needs testw and not testb */
@@ -384,7 +382,7 @@ syscall_exit:
384 # setting need_resched or sigpending 382 # setting need_resched or sigpending
385 # between sampling and the iret 383 # between sampling and the iret
386 TRACE_IRQS_OFF 384 TRACE_IRQS_OFF
387 testl $TF_MASK,PT_EFLAGS(%esp) # If tracing set singlestep flag on exit 385 testl $X86_EFLAGS_TF,PT_EFLAGS(%esp) # If tracing set singlestep flag on exit
388 jz no_singlestep 386 jz no_singlestep
389 orl $_TIF_SINGLESTEP,TI_flags(%ebp) 387 orl $_TIF_SINGLESTEP,TI_flags(%ebp)
390no_singlestep: 388no_singlestep:
@@ -399,7 +397,7 @@ restore_all:
399 # See comments in process.c:copy_thread() for details. 397 # See comments in process.c:copy_thread() for details.
400 movb PT_OLDSS(%esp), %ah 398 movb PT_OLDSS(%esp), %ah
401 movb PT_CS(%esp), %al 399 movb PT_CS(%esp), %al
402 andl $(VM_MASK | (SEGMENT_TI_MASK << 8) | SEGMENT_RPL_MASK), %eax 400 andl $(X86_EFLAGS_VM | (SEGMENT_TI_MASK << 8) | SEGMENT_RPL_MASK), %eax
403 cmpl $((SEGMENT_LDT << 8) | USER_RPL), %eax 401 cmpl $((SEGMENT_LDT << 8) | USER_RPL), %eax
404 CFI_REMEMBER_STATE 402 CFI_REMEMBER_STATE
405 je ldt_ss # returning to user-space with LDT SS 403 je ldt_ss # returning to user-space with LDT SS
@@ -486,7 +484,7 @@ work_resched:
486work_notifysig: # deal with pending signals and 484work_notifysig: # deal with pending signals and
487 # notify-resume requests 485 # notify-resume requests
488#ifdef CONFIG_VM86 486#ifdef CONFIG_VM86
489 testl $VM_MASK, PT_EFLAGS(%esp) 487 testl $X86_EFLAGS_VM, PT_EFLAGS(%esp)
490 movl %esp, %eax 488 movl %esp, %eax
491 jne work_notifysig_v86 # returning to kernel-space or 489 jne work_notifysig_v86 # returning to kernel-space or
492 # vm86-space 490 # vm86-space
@@ -543,9 +541,6 @@ END(syscall_exit_work)
543 541
544 RING0_INT_FRAME # can't unwind into user space anyway 542 RING0_INT_FRAME # can't unwind into user space anyway
545syscall_fault: 543syscall_fault:
546 pushl %eax # save orig_eax
547 CFI_ADJUST_CFA_OFFSET 4
548 SAVE_ALL
549 GET_THREAD_INFO(%ebp) 544 GET_THREAD_INFO(%ebp)
550 movl $-EFAULT,PT_EAX(%esp) 545 movl $-EFAULT,PT_EAX(%esp)
551 jmp resume_userspace 546 jmp resume_userspace
diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S
index c20c9e7e08dd..556a8df522a7 100644
--- a/arch/x86/kernel/entry_64.S
+++ b/arch/x86/kernel/entry_64.S
@@ -319,19 +319,17 @@ badsys:
319 /* Do syscall tracing */ 319 /* Do syscall tracing */
320tracesys: 320tracesys:
321 SAVE_REST 321 SAVE_REST
322 movq $-ENOSYS,RAX(%rsp) 322 movq $-ENOSYS,RAX(%rsp) /* ptrace can change this for a bad syscall */
323 FIXUP_TOP_OF_STACK %rdi 323 FIXUP_TOP_OF_STACK %rdi
324 movq %rsp,%rdi 324 movq %rsp,%rdi
325 call syscall_trace_enter 325 call syscall_trace_enter
326 LOAD_ARGS ARGOFFSET /* reload args from stack in case ptrace changed it */ 326 LOAD_ARGS ARGOFFSET /* reload args from stack in case ptrace changed it */
327 RESTORE_REST 327 RESTORE_REST
328 cmpq $__NR_syscall_max,%rax 328 cmpq $__NR_syscall_max,%rax
329 movq $-ENOSYS,%rcx 329 ja int_ret_from_sys_call /* RAX(%rsp) set to -ENOSYS above */
330 cmova %rcx,%rax
331 ja 1f
332 movq %r10,%rcx /* fixup for C */ 330 movq %r10,%rcx /* fixup for C */
333 call *sys_call_table(,%rax,8) 331 call *sys_call_table(,%rax,8)
3341: movq %rax,RAX-ARGOFFSET(%rsp) 332 movq %rax,RAX-ARGOFFSET(%rsp)
335 /* Use IRET because user could have changed frame */ 333 /* Use IRET because user could have changed frame */
336 334
337/* 335/*
diff --git a/arch/x86/kernel/genapic_64.c b/arch/x86/kernel/genapic_64.c
index 4ae7b6440260..9546ef408b92 100644
--- a/arch/x86/kernel/genapic_64.c
+++ b/arch/x86/kernel/genapic_64.c
@@ -15,6 +15,7 @@
15#include <linux/kernel.h> 15#include <linux/kernel.h>
16#include <linux/ctype.h> 16#include <linux/ctype.h>
17#include <linux/init.h> 17#include <linux/init.h>
18#include <linux/hardirq.h>
18 19
19#include <asm/smp.h> 20#include <asm/smp.h>
20#include <asm/ipi.h> 21#include <asm/ipi.h>
@@ -24,20 +25,20 @@
24#include <acpi/acpi_bus.h> 25#include <acpi/acpi_bus.h>
25#endif 26#endif
26 27
27/* which logical CPU number maps to which CPU (physical APIC ID) */ 28DEFINE_PER_CPU(int, x2apic_extra_bits);
28u16 x86_cpu_to_apicid_init[NR_CPUS] __initdata
29 = { [0 ... NR_CPUS-1] = BAD_APICID };
30void *x86_cpu_to_apicid_early_ptr;
31DEFINE_PER_CPU(u16, x86_cpu_to_apicid) = BAD_APICID;
32EXPORT_PER_CPU_SYMBOL(x86_cpu_to_apicid);
33 29
34struct genapic __read_mostly *genapic = &apic_flat; 30struct genapic __read_mostly *genapic = &apic_flat;
35 31
32static enum uv_system_type uv_system_type;
33
36/* 34/*
37 * Check the APIC IDs in bios_cpu_apicid and choose the APIC mode. 35 * Check the APIC IDs in bios_cpu_apicid and choose the APIC mode.
38 */ 36 */
39void __init setup_apic_routing(void) 37void __init setup_apic_routing(void)
40{ 38{
39 if (uv_system_type == UV_NON_UNIQUE_APIC)
40 genapic = &apic_x2apic_uv_x;
41 else
41#ifdef CONFIG_ACPI 42#ifdef CONFIG_ACPI
42 /* 43 /*
43 * Quirk: some x86_64 machines can only use physical APIC mode 44 * Quirk: some x86_64 machines can only use physical APIC mode
@@ -64,3 +65,37 @@ void send_IPI_self(int vector)
64{ 65{
65 __send_IPI_shortcut(APIC_DEST_SELF, vector, APIC_DEST_PHYSICAL); 66 __send_IPI_shortcut(APIC_DEST_SELF, vector, APIC_DEST_PHYSICAL);
66} 67}
68
69int __init acpi_madt_oem_check(char *oem_id, char *oem_table_id)
70{
71 if (!strcmp(oem_id, "SGI")) {
72 if (!strcmp(oem_table_id, "UVL"))
73 uv_system_type = UV_LEGACY_APIC;
74 else if (!strcmp(oem_table_id, "UVX"))
75 uv_system_type = UV_X2APIC;
76 else if (!strcmp(oem_table_id, "UVH"))
77 uv_system_type = UV_NON_UNIQUE_APIC;
78 }
79 return 0;
80}
81
82unsigned int read_apic_id(void)
83{
84 unsigned int id;
85
86 WARN_ON(preemptible());
87 id = apic_read(APIC_ID);
88 if (uv_system_type >= UV_X2APIC)
89 id |= __get_cpu_var(x2apic_extra_bits);
90 return id;
91}
92
93enum uv_system_type get_uv_system_type(void)
94{
95 return uv_system_type;
96}
97
98int is_uv_system(void)
99{
100 return uv_system_type != UV_NONE;
101}
diff --git a/arch/x86/kernel/genapic_flat_64.c b/arch/x86/kernel/genapic_flat_64.c
index 07352b74bda6..1a9c68845ee8 100644
--- a/arch/x86/kernel/genapic_flat_64.c
+++ b/arch/x86/kernel/genapic_flat_64.c
@@ -97,7 +97,7 @@ static void flat_send_IPI_all(int vector)
97 97
98static int flat_apic_id_registered(void) 98static int flat_apic_id_registered(void)
99{ 99{
100 return physid_isset(GET_APIC_ID(apic_read(APIC_ID)), phys_cpu_present_map); 100 return physid_isset(GET_APIC_ID(read_apic_id()), phys_cpu_present_map);
101} 101}
102 102
103static unsigned int flat_cpu_mask_to_apicid(cpumask_t cpumask) 103static unsigned int flat_cpu_mask_to_apicid(cpumask_t cpumask)
@@ -138,12 +138,9 @@ static cpumask_t physflat_target_cpus(void)
138 138
139static cpumask_t physflat_vector_allocation_domain(int cpu) 139static cpumask_t physflat_vector_allocation_domain(int cpu)
140{ 140{
141 cpumask_t domain = CPU_MASK_NONE; 141 return cpumask_of_cpu(cpu);
142 cpu_set(cpu, domain);
143 return domain;
144} 142}
145 143
146
147static void physflat_send_IPI_mask(cpumask_t cpumask, int vector) 144static void physflat_send_IPI_mask(cpumask_t cpumask, int vector)
148{ 145{
149 send_IPI_mask_sequence(cpumask, vector); 146 send_IPI_mask_sequence(cpumask, vector);
diff --git a/arch/x86/kernel/genx2apic_uv_x.c b/arch/x86/kernel/genx2apic_uv_x.c
new file mode 100644
index 000000000000..5d77c9cd8e15
--- /dev/null
+++ b/arch/x86/kernel/genx2apic_uv_x.c
@@ -0,0 +1,245 @@
1/*
2 * This file is subject to the terms and conditions of the GNU General Public
3 * License. See the file "COPYING" in the main directory of this archive
4 * for more details.
5 *
6 * SGI UV APIC functions (note: not an Intel compatible APIC)
7 *
8 * Copyright (C) 2007 Silicon Graphics, Inc. All rights reserved.
9 */
10
11#include <linux/threads.h>
12#include <linux/cpumask.h>
13#include <linux/string.h>
14#include <linux/kernel.h>
15#include <linux/ctype.h>
16#include <linux/init.h>
17#include <linux/sched.h>
18#include <linux/bootmem.h>
19#include <linux/module.h>
20#include <asm/smp.h>
21#include <asm/ipi.h>
22#include <asm/genapic.h>
23#include <asm/uv/uv_mmrs.h>
24#include <asm/uv/uv_hub.h>
25
26DEFINE_PER_CPU(struct uv_hub_info_s, __uv_hub_info);
27EXPORT_PER_CPU_SYMBOL_GPL(__uv_hub_info);
28
29struct uv_blade_info *uv_blade_info;
30EXPORT_SYMBOL_GPL(uv_blade_info);
31
32short *uv_node_to_blade;
33EXPORT_SYMBOL_GPL(uv_node_to_blade);
34
35short *uv_cpu_to_blade;
36EXPORT_SYMBOL_GPL(uv_cpu_to_blade);
37
38short uv_possible_blades;
39EXPORT_SYMBOL_GPL(uv_possible_blades);
40
41/* Start with all IRQs pointing to boot CPU. IRQ balancing will shift them. */
42
43static cpumask_t uv_target_cpus(void)
44{
45 return cpumask_of_cpu(0);
46}
47
48static cpumask_t uv_vector_allocation_domain(int cpu)
49{
50 cpumask_t domain = CPU_MASK_NONE;
51 cpu_set(cpu, domain);
52 return domain;
53}
54
55int uv_wakeup_secondary(int phys_apicid, unsigned int start_rip)
56{
57 unsigned long val;
58 int nasid;
59
60 nasid = uv_apicid_to_nasid(phys_apicid);
61 val = (1UL << UVH_IPI_INT_SEND_SHFT) |
62 (phys_apicid << UVH_IPI_INT_APIC_ID_SHFT) |
63 (((long)start_rip << UVH_IPI_INT_VECTOR_SHFT) >> 12) |
64 (6 << UVH_IPI_INT_DELIVERY_MODE_SHFT);
65 uv_write_global_mmr64(nasid, UVH_IPI_INT, val);
66 return 0;
67}
68
69static void uv_send_IPI_one(int cpu, int vector)
70{
71 unsigned long val, apicid;
72 int nasid;
73
74 apicid = per_cpu(x86_cpu_to_apicid, cpu); /* ZZZ - cache node-local ? */
75 nasid = uv_apicid_to_nasid(apicid);
76 val =
77 (1UL << UVH_IPI_INT_SEND_SHFT) | (apicid <<
78 UVH_IPI_INT_APIC_ID_SHFT) |
79 (vector << UVH_IPI_INT_VECTOR_SHFT);
80 uv_write_global_mmr64(nasid, UVH_IPI_INT, val);
81 printk(KERN_DEBUG
82 "UV: IPI to cpu %d, apicid 0x%lx, vec %d, nasid%d, val 0x%lx\n",
83 cpu, apicid, vector, nasid, val);
84}
85
86static void uv_send_IPI_mask(cpumask_t mask, int vector)
87{
88 unsigned int cpu;
89
90 for (cpu = 0; cpu < NR_CPUS; ++cpu)
91 if (cpu_isset(cpu, mask))
92 uv_send_IPI_one(cpu, vector);
93}
94
95static void uv_send_IPI_allbutself(int vector)
96{
97 cpumask_t mask = cpu_online_map;
98
99 cpu_clear(smp_processor_id(), mask);
100
101 if (!cpus_empty(mask))
102 uv_send_IPI_mask(mask, vector);
103}
104
105static void uv_send_IPI_all(int vector)
106{
107 uv_send_IPI_mask(cpu_online_map, vector);
108}
109
110static int uv_apic_id_registered(void)
111{
112 return 1;
113}
114
115static unsigned int uv_cpu_mask_to_apicid(cpumask_t cpumask)
116{
117 int cpu;
118
119 /*
120 * We're using fixed IRQ delivery, can only return one phys APIC ID.
121 * May as well be the first.
122 */
123 cpu = first_cpu(cpumask);
124 if ((unsigned)cpu < NR_CPUS)
125 return per_cpu(x86_cpu_to_apicid, cpu);
126 else
127 return BAD_APICID;
128}
129
130static unsigned int phys_pkg_id(int index_msb)
131{
132 return GET_APIC_ID(read_apic_id()) >> index_msb;
133}
134
135#ifdef ZZZ /* Needs x2apic patch */
136static void uv_send_IPI_self(int vector)
137{
138 apic_write(APIC_SELF_IPI, vector);
139}
140#endif
141
142struct genapic apic_x2apic_uv_x = {
143 .name = "UV large system",
144 .int_delivery_mode = dest_Fixed,
145 .int_dest_mode = (APIC_DEST_PHYSICAL != 0),
146 .target_cpus = uv_target_cpus,
147 .vector_allocation_domain = uv_vector_allocation_domain,/* Fixme ZZZ */
148 .apic_id_registered = uv_apic_id_registered,
149 .send_IPI_all = uv_send_IPI_all,
150 .send_IPI_allbutself = uv_send_IPI_allbutself,
151 .send_IPI_mask = uv_send_IPI_mask,
152 /* ZZZ.send_IPI_self = uv_send_IPI_self, */
153 .cpu_mask_to_apicid = uv_cpu_mask_to_apicid,
154 .phys_pkg_id = phys_pkg_id, /* Fixme ZZZ */
155};
156
157static __cpuinit void set_x2apic_extra_bits(int nasid)
158{
159 __get_cpu_var(x2apic_extra_bits) = ((nasid >> 1) << 6);
160}
161
162/*
163 * Called on boot cpu.
164 */
165static __init void uv_system_init(void)
166{
167 union uvh_si_addr_map_config_u m_n_config;
168 int bytes, nid, cpu, lcpu, nasid, last_nasid, blade;
169 unsigned long mmr_base;
170
171 m_n_config.v = uv_read_local_mmr(UVH_SI_ADDR_MAP_CONFIG);
172 mmr_base =
173 uv_read_local_mmr(UVH_RH_GAM_MMR_OVERLAY_CONFIG_MMR) &
174 ~UV_MMR_ENABLE;
175 printk(KERN_DEBUG "UV: global MMR base 0x%lx\n", mmr_base);
176
177 last_nasid = -1;
178 for_each_possible_cpu(cpu) {
179 nid = cpu_to_node(cpu);
180 nasid = uv_apicid_to_nasid(per_cpu(x86_cpu_to_apicid, cpu));
181 if (nasid != last_nasid)
182 uv_possible_blades++;
183 last_nasid = nasid;
184 }
185 printk(KERN_DEBUG "UV: Found %d blades\n", uv_num_possible_blades());
186
187 bytes = sizeof(struct uv_blade_info) * uv_num_possible_blades();
188 uv_blade_info = alloc_bootmem_pages(bytes);
189
190 bytes = sizeof(uv_node_to_blade[0]) * num_possible_nodes();
191 uv_node_to_blade = alloc_bootmem_pages(bytes);
192 memset(uv_node_to_blade, 255, bytes);
193
194 bytes = sizeof(uv_cpu_to_blade[0]) * num_possible_cpus();
195 uv_cpu_to_blade = alloc_bootmem_pages(bytes);
196 memset(uv_cpu_to_blade, 255, bytes);
197
198 last_nasid = -1;
199 blade = -1;
200 lcpu = -1;
201 for_each_possible_cpu(cpu) {
202 nid = cpu_to_node(cpu);
203 nasid = uv_apicid_to_nasid(per_cpu(x86_cpu_to_apicid, cpu));
204 if (nasid != last_nasid) {
205 blade++;
206 lcpu = -1;
207 uv_blade_info[blade].nr_posible_cpus = 0;
208 uv_blade_info[blade].nr_online_cpus = 0;
209 }
210 last_nasid = nasid;
211 lcpu++;
212
213 uv_cpu_hub_info(cpu)->m_val = m_n_config.s.m_skt;
214 uv_cpu_hub_info(cpu)->n_val = m_n_config.s.n_skt;
215 uv_cpu_hub_info(cpu)->numa_blade_id = blade;
216 uv_cpu_hub_info(cpu)->blade_processor_id = lcpu;
217 uv_cpu_hub_info(cpu)->local_nasid = nasid;
218 uv_cpu_hub_info(cpu)->gnode_upper =
219 nasid & ~((1 << uv_hub_info->n_val) - 1);
220 uv_cpu_hub_info(cpu)->global_mmr_base = mmr_base;
221 uv_cpu_hub_info(cpu)->coherency_domain_number = 0;/* ZZZ */
222 uv_blade_info[blade].nasid = nasid;
223 uv_blade_info[blade].nr_posible_cpus++;
224 uv_node_to_blade[nid] = blade;
225 uv_cpu_to_blade[cpu] = blade;
226
227 printk(KERN_DEBUG "UV cpu %d, apicid 0x%x, nasid %d, nid %d\n",
228 cpu, per_cpu(x86_cpu_to_apicid, cpu), nasid, nid);
229 printk(KERN_DEBUG "UV lcpu %d, blade %d\n", lcpu, blade);
230 }
231}
232
233/*
234 * Called on each cpu to initialize the per_cpu UV data area.
235 */
236void __cpuinit uv_cpu_init(void)
237{
238 if (!uv_node_to_blade)
239 uv_system_init();
240
241 uv_blade_info[uv_numa_blade_id()].nr_online_cpus++;
242
243 if (get_uv_system_type() == UV_NON_UNIQUE_APIC)
244 set_x2apic_extra_bits(uv_hub_info->local_nasid);
245}
diff --git a/arch/x86/kernel/head32.c b/arch/x86/kernel/head32.c
new file mode 100644
index 000000000000..3db059058927
--- /dev/null
+++ b/arch/x86/kernel/head32.c
@@ -0,0 +1,14 @@
1/*
2 * linux/arch/i386/kernel/head32.c -- prepare to run common code
3 *
4 * Copyright (C) 2000 Andrea Arcangeli <andrea@suse.de> SuSE
5 * Copyright (C) 2007 Eric Biederman <ebiederm@xmission.com>
6 */
7
8#include <linux/init.h>
9#include <linux/start_kernel.h>
10
11void __init i386_start_kernel(void)
12{
13 start_kernel();
14}
diff --git a/arch/x86/kernel/head64.c b/arch/x86/kernel/head64.c
index ad2440832de0..d6d54faa84df 100644
--- a/arch/x86/kernel/head64.c
+++ b/arch/x86/kernel/head64.c
@@ -49,39 +49,75 @@ static void __init copy_bootdata(char *real_mode_data)
49 } 49 }
50} 50}
51 51
52#define EBDA_ADDR_POINTER 0x40E 52#define BIOS_EBDA_SEGMENT 0x40E
53#define BIOS_LOWMEM_KILOBYTES 0x413
53 54
54static __init void reserve_ebda(void) 55/*
56 * The BIOS places the EBDA/XBDA at the top of conventional
57 * memory, and usually decreases the reported amount of
58 * conventional memory (int 0x12) too. This also contains a
59 * workaround for Dell systems that neglect to reserve EBDA.
60 * The same workaround also avoids a problem with the AMD768MPX
61 * chipset: reserve a page before VGA to prevent PCI prefetch
62 * into it (errata #56). Usually the page is reserved anyways,
63 * unless you have no PS/2 mouse plugged in.
64 */
65static void __init reserve_ebda_region(void)
55{ 66{
56 unsigned ebda_addr, ebda_size; 67 unsigned int lowmem, ebda_addr;
68
69 /* To determine the position of the EBDA and the */
70 /* end of conventional memory, we need to look at */
71 /* the BIOS data area. In a paravirtual environment */
72 /* that area is absent. We'll just have to assume */
73 /* that the paravirt case can handle memory setup */
74 /* correctly, without our help. */
75 if (paravirt_enabled())
76 return;
57 77
58 /* 78 /* end of low (conventional) memory */
59 * there is a real-mode segmented pointer pointing to the 79 lowmem = *(unsigned short *)__va(BIOS_LOWMEM_KILOBYTES);
60 * 4K EBDA area at 0x40E 80 lowmem <<= 10;
61 */ 81
62 ebda_addr = *(unsigned short *)__va(EBDA_ADDR_POINTER); 82 /* start of EBDA area */
83 ebda_addr = *(unsigned short *)__va(BIOS_EBDA_SEGMENT);
63 ebda_addr <<= 4; 84 ebda_addr <<= 4;
64 85
65 if (!ebda_addr) 86 /* Fixup: bios puts an EBDA in the top 64K segment */
66 return; 87 /* of conventional memory, but does not adjust lowmem. */
88 if ((lowmem - ebda_addr) <= 0x10000)
89 lowmem = ebda_addr;
67 90
68 ebda_size = *(unsigned short *)__va(ebda_addr); 91 /* Fixup: bios does not report an EBDA at all. */
92 /* Some old Dells seem to need 4k anyhow (bugzilla 2990) */
93 if ((ebda_addr == 0) && (lowmem >= 0x9f000))
94 lowmem = 0x9f000;
69 95
70 /* Round EBDA up to pages */ 96 /* Paranoia: should never happen, but... */
71 if (ebda_size == 0) 97 if ((lowmem == 0) || (lowmem >= 0x100000))
72 ebda_size = 1; 98 lowmem = 0x9f000;
73 ebda_size <<= 10;
74 ebda_size = round_up(ebda_size + (ebda_addr & ~PAGE_MASK), PAGE_SIZE);
75 if (ebda_size > 64*1024)
76 ebda_size = 64*1024;
77 99
78 reserve_early(ebda_addr, ebda_addr + ebda_size, "EBDA"); 100 /* reserve all memory between lowmem and the 1MB mark */
101 reserve_early(lowmem, 0x100000, "BIOS reserved");
79} 102}
80 103
81void __init x86_64_start_kernel(char * real_mode_data) 104void __init x86_64_start_kernel(char * real_mode_data)
82{ 105{
83 int i; 106 int i;
84 107
108 /*
109 * Build-time sanity checks on the kernel image and module
110 * area mappings. (these are purely build-time and produce no code)
111 */
112 BUILD_BUG_ON(MODULES_VADDR < KERNEL_IMAGE_START);
113 BUILD_BUG_ON(MODULES_VADDR-KERNEL_IMAGE_START < KERNEL_IMAGE_SIZE);
114 BUILD_BUG_ON(MODULES_LEN + KERNEL_IMAGE_SIZE > 2*PUD_SIZE);
115 BUILD_BUG_ON((KERNEL_IMAGE_START & ~PMD_MASK) != 0);
116 BUILD_BUG_ON((MODULES_VADDR & ~PMD_MASK) != 0);
117 BUILD_BUG_ON(!(MODULES_VADDR > __START_KERNEL));
118 BUILD_BUG_ON(!(((MODULES_END - 1) & PGDIR_MASK) ==
119 (__START_KERNEL & PGDIR_MASK)));
120
85 /* clear bss before set_intr_gate with early_idt_handler */ 121 /* clear bss before set_intr_gate with early_idt_handler */
86 clear_bss(); 122 clear_bss();
87 123
@@ -91,7 +127,7 @@ void __init x86_64_start_kernel(char * real_mode_data)
91 /* Cleanup the over mapped high alias */ 127 /* Cleanup the over mapped high alias */
92 cleanup_highmap(); 128 cleanup_highmap();
93 129
94 for (i = 0; i < IDT_ENTRIES; i++) { 130 for (i = 0; i < NUM_EXCEPTION_VECTORS; i++) {
95#ifdef CONFIG_EARLY_PRINTK 131#ifdef CONFIG_EARLY_PRINTK
96 set_intr_gate(i, &early_idt_handlers[i]); 132 set_intr_gate(i, &early_idt_handlers[i]);
97#else 133#else
@@ -118,7 +154,7 @@ void __init x86_64_start_kernel(char * real_mode_data)
118 reserve_early(ramdisk_image, ramdisk_end, "RAMDISK"); 154 reserve_early(ramdisk_image, ramdisk_end, "RAMDISK");
119 } 155 }
120 156
121 reserve_ebda(); 157 reserve_ebda_region();
122 158
123 /* 159 /*
124 * At this point everything still needed from the boot loader 160 * At this point everything still needed from the boot loader
diff --git a/arch/x86/kernel/head_32.S b/arch/x86/kernel/head_32.S
index 74d87ea85b5c..826988a6e964 100644
--- a/arch/x86/kernel/head_32.S
+++ b/arch/x86/kernel/head_32.S
@@ -450,7 +450,7 @@ is386: movl $2,%ecx # set MP
450 jmp initialize_secondary # all other CPUs call initialize_secondary 450 jmp initialize_secondary # all other CPUs call initialize_secondary
4511: 4511:
452#endif /* CONFIG_SMP */ 452#endif /* CONFIG_SMP */
453 jmp start_kernel 453 jmp i386_start_kernel
454 454
455/* 455/*
456 * We depend on ET to be correct. This checks for 287/387. 456 * We depend on ET to be correct. This checks for 287/387.
diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S
index a007454133a3..10a1955bb1d1 100644
--- a/arch/x86/kernel/head_64.S
+++ b/arch/x86/kernel/head_64.S
@@ -132,10 +132,6 @@ ident_complete:
132 addq %rbp, trampoline_level4_pgt + 0(%rip) 132 addq %rbp, trampoline_level4_pgt + 0(%rip)
133 addq %rbp, trampoline_level4_pgt + (511*8)(%rip) 133 addq %rbp, trampoline_level4_pgt + (511*8)(%rip)
134#endif 134#endif
135#ifdef CONFIG_ACPI_SLEEP
136 addq %rbp, wakeup_level4_pgt + 0(%rip)
137 addq %rbp, wakeup_level4_pgt + (511*8)(%rip)
138#endif
139 135
140 /* Due to ENTRY(), sometimes the empty space gets filled with 136 /* Due to ENTRY(), sometimes the empty space gets filled with
141 * zeros. Better take a jmp than relying on empty space being 137 * zeros. Better take a jmp than relying on empty space being
@@ -267,21 +263,16 @@ ENTRY(secondary_startup_64)
267bad_address: 263bad_address:
268 jmp bad_address 264 jmp bad_address
269 265
266 .section ".init.text","ax"
270#ifdef CONFIG_EARLY_PRINTK 267#ifdef CONFIG_EARLY_PRINTK
271.macro early_idt_tramp first, last
272 .ifgt \last-\first
273 early_idt_tramp \first, \last-1
274 .endif
275 movl $\last,%esi
276 jmp early_idt_handler
277.endm
278
279 .globl early_idt_handlers 268 .globl early_idt_handlers
280early_idt_handlers: 269early_idt_handlers:
281 early_idt_tramp 0, 63 270 i = 0
282 early_idt_tramp 64, 127 271 .rept NUM_EXCEPTION_VECTORS
283 early_idt_tramp 128, 191 272 movl $i, %esi
284 early_idt_tramp 192, 255 273 jmp early_idt_handler
274 i = i + 1
275 .endr
285#endif 276#endif
286 277
287ENTRY(early_idt_handler) 278ENTRY(early_idt_handler)
@@ -327,6 +318,7 @@ early_idt_msg:
327early_idt_ripmsg: 318early_idt_ripmsg:
328 .asciz "RIP %s\n" 319 .asciz "RIP %s\n"
329#endif /* CONFIG_EARLY_PRINTK */ 320#endif /* CONFIG_EARLY_PRINTK */
321 .previous
330 322
331.balign PAGE_SIZE 323.balign PAGE_SIZE
332 324
@@ -383,12 +375,12 @@ NEXT_PAGE(level2_ident_pgt)
383 375
384NEXT_PAGE(level2_kernel_pgt) 376NEXT_PAGE(level2_kernel_pgt)
385 /* 377 /*
386 * 128 MB kernel mapping. We spend a full page on this pagetable 378 * 512 MB kernel mapping. We spend a full page on this pagetable
387 * anyway. 379 * anyway.
388 * 380 *
389 * The kernel code+data+bss must not be bigger than that. 381 * The kernel code+data+bss must not be bigger than that.
390 * 382 *
391 * (NOTE: at +128MB starts the module area, see MODULES_VADDR. 383 * (NOTE: at +512MB starts the module area, see MODULES_VADDR.
392 * If you want to increase this then increase MODULES_VADDR 384 * If you want to increase this then increase MODULES_VADDR
393 * too.) 385 * too.)
394 */ 386 */
diff --git a/arch/x86/kernel/i386_ksyms_32.c b/arch/x86/kernel/i386_ksyms_32.c
index 061627806a2d..deb43785e923 100644
--- a/arch/x86/kernel/i386_ksyms_32.c
+++ b/arch/x86/kernel/i386_ksyms_32.c
@@ -1,13 +1,8 @@
1#include <linux/module.h> 1#include <linux/module.h>
2#include <asm/semaphore.h>
3#include <asm/checksum.h> 2#include <asm/checksum.h>
4#include <asm/desc.h> 3#include <asm/desc.h>
5#include <asm/pgtable.h> 4#include <asm/pgtable.h>
6 5
7EXPORT_SYMBOL(__down_failed);
8EXPORT_SYMBOL(__down_failed_interruptible);
9EXPORT_SYMBOL(__down_failed_trylock);
10EXPORT_SYMBOL(__up_wakeup);
11/* Networking helper routines. */ 6/* Networking helper routines. */
12EXPORT_SYMBOL(csum_partial_copy_generic); 7EXPORT_SYMBOL(csum_partial_copy_generic);
13 8
diff --git a/arch/x86/kernel/i387.c b/arch/x86/kernel/i387.c
index d2e39e69aaf8..8f8102d967b3 100644
--- a/arch/x86/kernel/i387.c
+++ b/arch/x86/kernel/i387.c
@@ -5,45 +5,41 @@
5 * General FPU state handling cleanups 5 * General FPU state handling cleanups
6 * Gareth Hughes <gareth@valinux.com>, May 2000 6 * Gareth Hughes <gareth@valinux.com>, May 2000
7 */ 7 */
8
9#include <linux/sched.h>
10#include <linux/module.h> 8#include <linux/module.h>
11#include <linux/regset.h> 9#include <linux/regset.h>
10#include <linux/sched.h>
11
12#include <asm/sigcontext.h>
12#include <asm/processor.h> 13#include <asm/processor.h>
13#include <asm/i387.h>
14#include <asm/math_emu.h> 14#include <asm/math_emu.h>
15#include <asm/sigcontext.h>
16#include <asm/user.h>
17#include <asm/ptrace.h>
18#include <asm/uaccess.h> 15#include <asm/uaccess.h>
16#include <asm/ptrace.h>
17#include <asm/i387.h>
18#include <asm/user.h>
19 19
20#ifdef CONFIG_X86_64 20#ifdef CONFIG_X86_64
21 21# include <asm/sigcontext32.h>
22#include <asm/sigcontext32.h> 22# include <asm/user32.h>
23#include <asm/user32.h>
24
25#else 23#else
26 24# define save_i387_ia32 save_i387
27#define save_i387_ia32 save_i387 25# define restore_i387_ia32 restore_i387
28#define restore_i387_ia32 restore_i387 26# define _fpstate_ia32 _fpstate
29 27# define user_i387_ia32_struct user_i387_struct
30#define _fpstate_ia32 _fpstate 28# define user32_fxsr_struct user_fxsr_struct
31#define user_i387_ia32_struct user_i387_struct
32#define user32_fxsr_struct user_fxsr_struct
33
34#endif 29#endif
35 30
36#ifdef CONFIG_MATH_EMULATION 31#ifdef CONFIG_MATH_EMULATION
37#define HAVE_HWFP (boot_cpu_data.hard_math) 32# define HAVE_HWFP (boot_cpu_data.hard_math)
38#else 33#else
39#define HAVE_HWFP 1 34# define HAVE_HWFP 1
40#endif 35#endif
41 36
42static unsigned int mxcsr_feature_mask __read_mostly = 0xffffffffu; 37static unsigned int mxcsr_feature_mask __read_mostly = 0xffffffffu;
43 38
44void mxcsr_feature_mask_init(void) 39void mxcsr_feature_mask_init(void)
45{ 40{
46 unsigned long mask = 0; 41 unsigned long mask = 0;
42
47 clts(); 43 clts();
48 if (cpu_has_fxsr) { 44 if (cpu_has_fxsr) {
49 memset(&current->thread.i387.fxsave, 0, 45 memset(&current->thread.i387.fxsave, 0,
@@ -69,10 +65,11 @@ void __cpuinit fpu_init(void)
69 65
70 if (offsetof(struct task_struct, thread.i387.fxsave) & 15) 66 if (offsetof(struct task_struct, thread.i387.fxsave) & 15)
71 __bad_fxsave_alignment(); 67 __bad_fxsave_alignment();
68
72 set_in_cr4(X86_CR4_OSFXSR); 69 set_in_cr4(X86_CR4_OSFXSR);
73 set_in_cr4(X86_CR4_OSXMMEXCPT); 70 set_in_cr4(X86_CR4_OSXMMEXCPT);
74 71
75 write_cr0(oldcr0 & ~((1UL<<3)|(1UL<<2))); /* clear TS and EM */ 72 write_cr0(oldcr0 & ~(X86_CR0_TS|X86_CR0_EM)); /* clear TS and EM */
76 73
77 mxcsr_feature_mask_init(); 74 mxcsr_feature_mask_init();
78 /* clean state in init */ 75 /* clean state in init */
@@ -178,6 +175,7 @@ static inline unsigned short twd_i387_to_fxsr(unsigned short twd)
178 tmp = (tmp | (tmp >> 1)) & 0x3333; /* 00VV00VV00VV00VV */ 175 tmp = (tmp | (tmp >> 1)) & 0x3333; /* 00VV00VV00VV00VV */
179 tmp = (tmp | (tmp >> 2)) & 0x0f0f; /* 0000VVVV0000VVVV */ 176 tmp = (tmp | (tmp >> 2)) & 0x0f0f; /* 0000VVVV0000VVVV */
180 tmp = (tmp | (tmp >> 4)) & 0x00ff; /* 00000000VVVVVVVV */ 177 tmp = (tmp | (tmp >> 4)) & 0x00ff; /* 00000000VVVVVVVV */
178
181 return tmp; 179 return tmp;
182} 180}
183 181
@@ -232,8 +230,8 @@ static inline u32 twd_fxsr_to_i387(struct i387_fxsave_struct *fxsave)
232 * FXSR floating point environment conversions. 230 * FXSR floating point environment conversions.
233 */ 231 */
234 232
235static void convert_from_fxsr(struct user_i387_ia32_struct *env, 233static void
236 struct task_struct *tsk) 234convert_from_fxsr(struct user_i387_ia32_struct *env, struct task_struct *tsk)
237{ 235{
238 struct i387_fxsave_struct *fxsave = &tsk->thread.i387.fxsave; 236 struct i387_fxsave_struct *fxsave = &tsk->thread.i387.fxsave;
239 struct _fpreg *to = (struct _fpreg *) &env->st_space[0]; 237 struct _fpreg *to = (struct _fpreg *) &env->st_space[0];
@@ -252,10 +250,11 @@ static void convert_from_fxsr(struct user_i387_ia32_struct *env,
252 * should be actually ds/cs at fpu exception time, but 250 * should be actually ds/cs at fpu exception time, but
253 * that information is not available in 64bit mode. 251 * that information is not available in 64bit mode.
254 */ 252 */
255 asm("mov %%ds,%0" : "=r" (env->fos)); 253 asm("mov %%ds, %[fos]" : [fos] "=r" (env->fos));
256 asm("mov %%cs,%0" : "=r" (env->fcs)); 254 asm("mov %%cs, %[fcs]" : [fcs] "=r" (env->fcs));
257 } else { 255 } else {
258 struct pt_regs *regs = task_pt_regs(tsk); 256 struct pt_regs *regs = task_pt_regs(tsk);
257
259 env->fos = 0xffff0000 | tsk->thread.ds; 258 env->fos = 0xffff0000 | tsk->thread.ds;
260 env->fcs = regs->cs; 259 env->fcs = regs->cs;
261 } 260 }
@@ -309,9 +308,10 @@ int fpregs_get(struct task_struct *target, const struct user_regset *regset,
309 308
310 init_fpu(target); 309 init_fpu(target);
311 310
312 if (!cpu_has_fxsr) 311 if (!cpu_has_fxsr) {
313 return user_regset_copyout(&pos, &count, &kbuf, &ubuf, 312 return user_regset_copyout(&pos, &count, &kbuf, &ubuf,
314 &target->thread.i387.fsave, 0, -1); 313 &target->thread.i387.fsave, 0, -1);
314 }
315 315
316 if (kbuf && pos == 0 && count == sizeof(env)) { 316 if (kbuf && pos == 0 && count == sizeof(env)) {
317 convert_from_fxsr(kbuf, target); 317 convert_from_fxsr(kbuf, target);
@@ -319,6 +319,7 @@ int fpregs_get(struct task_struct *target, const struct user_regset *regset,
319 } 319 }
320 320
321 convert_from_fxsr(&env, target); 321 convert_from_fxsr(&env, target);
322
322 return user_regset_copyout(&pos, &count, &kbuf, &ubuf, &env, 0, -1); 323 return user_regset_copyout(&pos, &count, &kbuf, &ubuf, &env, 0, -1);
323} 324}
324 325
@@ -335,9 +336,10 @@ int fpregs_set(struct task_struct *target, const struct user_regset *regset,
335 init_fpu(target); 336 init_fpu(target);
336 set_stopped_child_used_math(target); 337 set_stopped_child_used_math(target);
337 338
338 if (!cpu_has_fxsr) 339 if (!cpu_has_fxsr) {
339 return user_regset_copyin(&pos, &count, &kbuf, &ubuf, 340 return user_regset_copyin(&pos, &count, &kbuf, &ubuf,
340 &target->thread.i387.fsave, 0, -1); 341 &target->thread.i387.fsave, 0, -1);
342 }
341 343
342 if (pos > 0 || count < sizeof(env)) 344 if (pos > 0 || count < sizeof(env))
343 convert_from_fxsr(&env, target); 345 convert_from_fxsr(&env, target);
@@ -392,28 +394,28 @@ int save_i387_ia32(struct _fpstate_ia32 __user *buf)
392{ 394{
393 if (!used_math()) 395 if (!used_math())
394 return 0; 396 return 0;
395 397 /*
396 /* This will cause a "finit" to be triggered by the next 398 * This will cause a "finit" to be triggered by the next
397 * attempted FPU operation by the 'current' process. 399 * attempted FPU operation by the 'current' process.
398 */ 400 */
399 clear_used_math(); 401 clear_used_math();
400 402
401 if (HAVE_HWFP) { 403 if (!HAVE_HWFP) {
402 if (cpu_has_fxsr) {
403 return save_i387_fxsave(buf);
404 } else {
405 return save_i387_fsave(buf);
406 }
407 } else {
408 return fpregs_soft_get(current, NULL, 404 return fpregs_soft_get(current, NULL,
409 0, sizeof(struct user_i387_ia32_struct), 405 0, sizeof(struct user_i387_ia32_struct),
410 NULL, buf) ? -1 : 1; 406 NULL, buf) ? -1 : 1;
411 } 407 }
408
409 if (cpu_has_fxsr)
410 return save_i387_fxsave(buf);
411 else
412 return save_i387_fsave(buf);
412} 413}
413 414
414static inline int restore_i387_fsave(struct _fpstate_ia32 __user *buf) 415static inline int restore_i387_fsave(struct _fpstate_ia32 __user *buf)
415{ 416{
416 struct task_struct *tsk = current; 417 struct task_struct *tsk = current;
418
417 clear_fpu(tsk); 419 clear_fpu(tsk);
418 return __copy_from_user(&tsk->thread.i387.fsave, buf, 420 return __copy_from_user(&tsk->thread.i387.fsave, buf,
419 sizeof(struct i387_fsave_struct)); 421 sizeof(struct i387_fsave_struct));
@@ -421,9 +423,10 @@ static inline int restore_i387_fsave(struct _fpstate_ia32 __user *buf)
421 423
422static int restore_i387_fxsave(struct _fpstate_ia32 __user *buf) 424static int restore_i387_fxsave(struct _fpstate_ia32 __user *buf)
423{ 425{
424 int err;
425 struct task_struct *tsk = current; 426 struct task_struct *tsk = current;
426 struct user_i387_ia32_struct env; 427 struct user_i387_ia32_struct env;
428 int err;
429
427 clear_fpu(tsk); 430 clear_fpu(tsk);
428 err = __copy_from_user(&tsk->thread.i387.fxsave, &buf->_fxsr_env[0], 431 err = __copy_from_user(&tsk->thread.i387.fxsave, &buf->_fxsr_env[0],
429 sizeof(struct i387_fxsave_struct)); 432 sizeof(struct i387_fxsave_struct));
@@ -432,6 +435,7 @@ static int restore_i387_fxsave(struct _fpstate_ia32 __user *buf)
432 if (err || __copy_from_user(&env, buf, sizeof(env))) 435 if (err || __copy_from_user(&env, buf, sizeof(env)))
433 return 1; 436 return 1;
434 convert_to_fxsr(tsk, &env); 437 convert_to_fxsr(tsk, &env);
438
435 return 0; 439 return 0;
436} 440}
437 441
@@ -440,17 +444,17 @@ int restore_i387_ia32(struct _fpstate_ia32 __user *buf)
440 int err; 444 int err;
441 445
442 if (HAVE_HWFP) { 446 if (HAVE_HWFP) {
443 if (cpu_has_fxsr) { 447 if (cpu_has_fxsr)
444 err = restore_i387_fxsave(buf); 448 err = restore_i387_fxsave(buf);
445 } else { 449 else
446 err = restore_i387_fsave(buf); 450 err = restore_i387_fsave(buf);
447 }
448 } else { 451 } else {
449 err = fpregs_soft_set(current, NULL, 452 err = fpregs_soft_set(current, NULL,
450 0, sizeof(struct user_i387_ia32_struct), 453 0, sizeof(struct user_i387_ia32_struct),
451 NULL, buf) != 0; 454 NULL, buf) != 0;
452 } 455 }
453 set_used_math(); 456 set_used_math();
457
454 return err; 458 return err;
455} 459}
456 460
@@ -463,8 +467,8 @@ int restore_i387_ia32(struct _fpstate_ia32 __user *buf)
463 */ 467 */
464int dump_fpu(struct pt_regs *regs, struct user_i387_struct *fpu) 468int dump_fpu(struct pt_regs *regs, struct user_i387_struct *fpu)
465{ 469{
466 int fpvalid;
467 struct task_struct *tsk = current; 470 struct task_struct *tsk = current;
471 int fpvalid;
468 472
469 fpvalid = !!used_math(); 473 fpvalid = !!used_math();
470 if (fpvalid) 474 if (fpvalid)
diff --git a/arch/x86/kernel/io_apic_32.c b/arch/x86/kernel/io_apic_32.c
index 4ca548632c8d..2e2f42074e18 100644
--- a/arch/x86/kernel/io_apic_32.c
+++ b/arch/x86/kernel/io_apic_32.c
@@ -71,6 +71,16 @@ int sis_apic_bug = -1;
71 */ 71 */
72int nr_ioapic_registers[MAX_IO_APICS]; 72int nr_ioapic_registers[MAX_IO_APICS];
73 73
74/* I/O APIC entries */
75struct mpc_config_ioapic mp_ioapics[MAX_IO_APICS];
76int nr_ioapics;
77
78/* MP IRQ source entries */
79struct mpc_config_intsrc mp_irqs[MAX_IRQ_SOURCES];
80
81/* # of MP IRQ source entries */
82int mp_irq_entries;
83
74static int disable_timer_pin_1 __initdata; 84static int disable_timer_pin_1 __initdata;
75 85
76/* 86/*
@@ -810,10 +820,7 @@ static int __init find_isa_irq_pin(int irq, int type)
810 for (i = 0; i < mp_irq_entries; i++) { 820 for (i = 0; i < mp_irq_entries; i++) {
811 int lbus = mp_irqs[i].mpc_srcbus; 821 int lbus = mp_irqs[i].mpc_srcbus;
812 822
813 if ((mp_bus_id_to_type[lbus] == MP_BUS_ISA || 823 if (test_bit(lbus, mp_bus_not_pci) &&
814 mp_bus_id_to_type[lbus] == MP_BUS_EISA ||
815 mp_bus_id_to_type[lbus] == MP_BUS_MCA
816 ) &&
817 (mp_irqs[i].mpc_irqtype == type) && 824 (mp_irqs[i].mpc_irqtype == type) &&
818 (mp_irqs[i].mpc_srcbusirq == irq)) 825 (mp_irqs[i].mpc_srcbusirq == irq))
819 826
@@ -829,10 +836,7 @@ static int __init find_isa_irq_apic(int irq, int type)
829 for (i = 0; i < mp_irq_entries; i++) { 836 for (i = 0; i < mp_irq_entries; i++) {
830 int lbus = mp_irqs[i].mpc_srcbus; 837 int lbus = mp_irqs[i].mpc_srcbus;
831 838
832 if ((mp_bus_id_to_type[lbus] == MP_BUS_ISA || 839 if (test_bit(lbus, mp_bus_not_pci) &&
833 mp_bus_id_to_type[lbus] == MP_BUS_EISA ||
834 mp_bus_id_to_type[lbus] == MP_BUS_MCA
835 ) &&
836 (mp_irqs[i].mpc_irqtype == type) && 840 (mp_irqs[i].mpc_irqtype == type) &&
837 (mp_irqs[i].mpc_srcbusirq == irq)) 841 (mp_irqs[i].mpc_srcbusirq == irq))
838 break; 842 break;
@@ -872,7 +876,7 @@ int IO_APIC_get_PCI_irq_vector(int bus, int slot, int pin)
872 mp_irqs[i].mpc_dstapic == MP_APIC_ALL) 876 mp_irqs[i].mpc_dstapic == MP_APIC_ALL)
873 break; 877 break;
874 878
875 if ((mp_bus_id_to_type[lbus] == MP_BUS_PCI) && 879 if (!test_bit(lbus, mp_bus_not_pci) &&
876 !mp_irqs[i].mpc_irqtype && 880 !mp_irqs[i].mpc_irqtype &&
877 (bus == lbus) && 881 (bus == lbus) &&
878 (slot == ((mp_irqs[i].mpc_srcbusirq >> 2) & 0x1f))) { 882 (slot == ((mp_irqs[i].mpc_srcbusirq >> 2) & 0x1f))) {
@@ -921,6 +925,7 @@ void __init setup_ioapic_dest(void)
921} 925}
922#endif 926#endif
923 927
928#if defined(CONFIG_EISA) || defined(CONFIG_MCA)
924/* 929/*
925 * EISA Edge/Level control register, ELCR 930 * EISA Edge/Level control register, ELCR
926 */ 931 */
@@ -934,6 +939,13 @@ static int EISA_ELCR(unsigned int irq)
934 "Broken MPtable reports ISA irq %d\n", irq); 939 "Broken MPtable reports ISA irq %d\n", irq);
935 return 0; 940 return 0;
936} 941}
942#endif
943
944/* ISA interrupts are always polarity zero edge triggered,
945 * when listed as conforming in the MP table. */
946
947#define default_ISA_trigger(idx) (0)
948#define default_ISA_polarity(idx) (0)
937 949
938/* EISA interrupts are always polarity zero and can be edge or level 950/* EISA interrupts are always polarity zero and can be edge or level
939 * trigger depending on the ELCR value. If an interrupt is listed as 951 * trigger depending on the ELCR value. If an interrupt is listed as
@@ -941,13 +953,7 @@ static int EISA_ELCR(unsigned int irq)
941 * be read in from the ELCR */ 953 * be read in from the ELCR */
942 954
943#define default_EISA_trigger(idx) (EISA_ELCR(mp_irqs[idx].mpc_srcbusirq)) 955#define default_EISA_trigger(idx) (EISA_ELCR(mp_irqs[idx].mpc_srcbusirq))
944#define default_EISA_polarity(idx) (0) 956#define default_EISA_polarity(idx) default_ISA_polarity(idx)
945
946/* ISA interrupts are always polarity zero edge triggered,
947 * when listed as conforming in the MP table. */
948
949#define default_ISA_trigger(idx) (0)
950#define default_ISA_polarity(idx) (0)
951 957
952/* PCI interrupts are always polarity one level triggered, 958/* PCI interrupts are always polarity one level triggered,
953 * when listed as conforming in the MP table. */ 959 * when listed as conforming in the MP table. */
@@ -959,7 +965,7 @@ static int EISA_ELCR(unsigned int irq)
959 * when listed as conforming in the MP table. */ 965 * when listed as conforming in the MP table. */
960 966
961#define default_MCA_trigger(idx) (1) 967#define default_MCA_trigger(idx) (1)
962#define default_MCA_polarity(idx) (0) 968#define default_MCA_polarity(idx) default_ISA_polarity(idx)
963 969
964static int MPBIOS_polarity(int idx) 970static int MPBIOS_polarity(int idx)
965{ 971{
@@ -973,35 +979,9 @@ static int MPBIOS_polarity(int idx)
973 { 979 {
974 case 0: /* conforms, ie. bus-type dependent polarity */ 980 case 0: /* conforms, ie. bus-type dependent polarity */
975 { 981 {
976 switch (mp_bus_id_to_type[bus]) 982 polarity = test_bit(bus, mp_bus_not_pci)?
977 { 983 default_ISA_polarity(idx):
978 case MP_BUS_ISA: /* ISA pin */ 984 default_PCI_polarity(idx);
979 {
980 polarity = default_ISA_polarity(idx);
981 break;
982 }
983 case MP_BUS_EISA: /* EISA pin */
984 {
985 polarity = default_EISA_polarity(idx);
986 break;
987 }
988 case MP_BUS_PCI: /* PCI pin */
989 {
990 polarity = default_PCI_polarity(idx);
991 break;
992 }
993 case MP_BUS_MCA: /* MCA pin */
994 {
995 polarity = default_MCA_polarity(idx);
996 break;
997 }
998 default:
999 {
1000 printk(KERN_WARNING "broken BIOS!!\n");
1001 polarity = 1;
1002 break;
1003 }
1004 }
1005 break; 985 break;
1006 } 986 }
1007 case 1: /* high active */ 987 case 1: /* high active */
@@ -1042,11 +1022,15 @@ static int MPBIOS_trigger(int idx)
1042 { 1022 {
1043 case 0: /* conforms, ie. bus-type dependent */ 1023 case 0: /* conforms, ie. bus-type dependent */
1044 { 1024 {
1025 trigger = test_bit(bus, mp_bus_not_pci)?
1026 default_ISA_trigger(idx):
1027 default_PCI_trigger(idx);
1028#if defined(CONFIG_EISA) || defined(CONFIG_MCA)
1045 switch (mp_bus_id_to_type[bus]) 1029 switch (mp_bus_id_to_type[bus])
1046 { 1030 {
1047 case MP_BUS_ISA: /* ISA pin */ 1031 case MP_BUS_ISA: /* ISA pin */
1048 { 1032 {
1049 trigger = default_ISA_trigger(idx); 1033 /* set before the switch */
1050 break; 1034 break;
1051 } 1035 }
1052 case MP_BUS_EISA: /* EISA pin */ 1036 case MP_BUS_EISA: /* EISA pin */
@@ -1056,7 +1040,7 @@ static int MPBIOS_trigger(int idx)
1056 } 1040 }
1057 case MP_BUS_PCI: /* PCI pin */ 1041 case MP_BUS_PCI: /* PCI pin */
1058 { 1042 {
1059 trigger = default_PCI_trigger(idx); 1043 /* set before the switch */
1060 break; 1044 break;
1061 } 1045 }
1062 case MP_BUS_MCA: /* MCA pin */ 1046 case MP_BUS_MCA: /* MCA pin */
@@ -1071,6 +1055,7 @@ static int MPBIOS_trigger(int idx)
1071 break; 1055 break;
1072 } 1056 }
1073 } 1057 }
1058#endif
1074 break; 1059 break;
1075 } 1060 }
1076 case 1: /* edge */ 1061 case 1: /* edge */
@@ -1120,39 +1105,22 @@ static int pin_2_irq(int idx, int apic, int pin)
1120 if (mp_irqs[idx].mpc_dstirq != pin) 1105 if (mp_irqs[idx].mpc_dstirq != pin)
1121 printk(KERN_ERR "broken BIOS or MPTABLE parser, ayiee!!\n"); 1106 printk(KERN_ERR "broken BIOS or MPTABLE parser, ayiee!!\n");
1122 1107
1123 switch (mp_bus_id_to_type[bus]) 1108 if (test_bit(bus, mp_bus_not_pci))
1124 { 1109 irq = mp_irqs[idx].mpc_srcbusirq;
1125 case MP_BUS_ISA: /* ISA pin */ 1110 else {
1126 case MP_BUS_EISA: 1111 /*
1127 case MP_BUS_MCA: 1112 * PCI IRQs are mapped in order
1128 { 1113 */
1129 irq = mp_irqs[idx].mpc_srcbusirq; 1114 i = irq = 0;
1130 break; 1115 while (i < apic)
1131 } 1116 irq += nr_ioapic_registers[i++];
1132 case MP_BUS_PCI: /* PCI pin */ 1117 irq += pin;
1133 {
1134 /*
1135 * PCI IRQs are mapped in order
1136 */
1137 i = irq = 0;
1138 while (i < apic)
1139 irq += nr_ioapic_registers[i++];
1140 irq += pin;
1141
1142 /*
1143 * For MPS mode, so far only needed by ES7000 platform
1144 */
1145 if (ioapic_renumber_irq)
1146 irq = ioapic_renumber_irq(apic, irq);
1147 1118
1148 break; 1119 /*
1149 } 1120 * For MPS mode, so far only needed by ES7000 platform
1150 default: 1121 */
1151 { 1122 if (ioapic_renumber_irq)
1152 printk(KERN_ERR "unknown bus type %d.\n",bus); 1123 irq = ioapic_renumber_irq(apic, irq);
1153 irq = 0;
1154 break;
1155 }
1156 } 1124 }
1157 1125
1158 /* 1126 /*
@@ -1260,7 +1228,6 @@ static void __init setup_IO_APIC_irqs(void)
1260{ 1228{
1261 struct IO_APIC_route_entry entry; 1229 struct IO_APIC_route_entry entry;
1262 int apic, pin, idx, irq, first_notcon = 1, vector; 1230 int apic, pin, idx, irq, first_notcon = 1, vector;
1263 unsigned long flags;
1264 1231
1265 apic_printk(APIC_VERBOSE, KERN_DEBUG "init IO_APIC IRQs\n"); 1232 apic_printk(APIC_VERBOSE, KERN_DEBUG "init IO_APIC IRQs\n");
1266 1233
@@ -1326,9 +1293,7 @@ static void __init setup_IO_APIC_irqs(void)
1326 if (!apic && (irq < 16)) 1293 if (!apic && (irq < 16))
1327 disable_8259A_irq(irq); 1294 disable_8259A_irq(irq);
1328 } 1295 }
1329 spin_lock_irqsave(&ioapic_lock, flags); 1296 ioapic_write_entry(apic, pin, entry);
1330 __ioapic_write_entry(apic, pin, entry);
1331 spin_unlock_irqrestore(&ioapic_lock, flags);
1332 } 1297 }
1333 } 1298 }
1334 1299
@@ -1524,8 +1489,8 @@ void /*__init*/ print_local_APIC(void * dummy)
1524 1489
1525 printk("\n" KERN_DEBUG "printing local APIC contents on CPU#%d/%d:\n", 1490 printk("\n" KERN_DEBUG "printing local APIC contents on CPU#%d/%d:\n",
1526 smp_processor_id(), hard_smp_processor_id()); 1491 smp_processor_id(), hard_smp_processor_id());
1527 v = apic_read(APIC_ID); 1492 printk(KERN_INFO "... APIC ID: %08x (%01x)\n", v,
1528 printk(KERN_INFO "... APIC ID: %08x (%01x)\n", v, GET_APIC_ID(v)); 1493 GET_APIC_ID(read_apic_id()));
1529 v = apic_read(APIC_LVR); 1494 v = apic_read(APIC_LVR);
1530 printk(KERN_INFO "... APIC VERSION: %08x\n", v); 1495 printk(KERN_INFO "... APIC VERSION: %08x\n", v);
1531 ver = GET_APIC_VERSION(v); 1496 ver = GET_APIC_VERSION(v);
@@ -1734,7 +1699,7 @@ void disable_IO_APIC(void)
1734 entry.delivery_mode = dest_ExtINT; /* ExtInt */ 1699 entry.delivery_mode = dest_ExtINT; /* ExtInt */
1735 entry.vector = 0; 1700 entry.vector = 0;
1736 entry.dest.physical.physical_dest = 1701 entry.dest.physical.physical_dest =
1737 GET_APIC_ID(apic_read(APIC_ID)); 1702 GET_APIC_ID(read_apic_id());
1738 1703
1739 /* 1704 /*
1740 * Add it to the IO-APIC irq-routing table: 1705 * Add it to the IO-APIC irq-routing table:
@@ -2031,8 +1996,7 @@ static inline void init_IO_APIC_traps(void)
2031 * 0x80, because int 0x80 is hm, kind of importantish. ;) 1996 * 0x80, because int 0x80 is hm, kind of importantish. ;)
2032 */ 1997 */
2033 for (irq = 0; irq < NR_IRQS ; irq++) { 1998 for (irq = 0; irq < NR_IRQS ; irq++) {
2034 int tmp = irq; 1999 if (IO_APIC_IRQ(irq) && !irq_vector[irq]) {
2035 if (IO_APIC_IRQ(tmp) && !irq_vector[tmp]) {
2036 /* 2000 /*
2037 * Hmm.. We don't have an entry for this, 2001 * Hmm.. We don't have an entry for this,
2038 * so default to an old-fashioned 8259 2002 * so default to an old-fashioned 8259
@@ -2156,8 +2120,6 @@ static inline void unlock_ExtINT_logic(void)
2156 ioapic_write_entry(apic, pin, entry0); 2120 ioapic_write_entry(apic, pin, entry0);
2157} 2121}
2158 2122
2159int timer_uses_ioapic_pin_0;
2160
2161/* 2123/*
2162 * This code may look a bit paranoid, but it's supposed to cooperate with 2124 * This code may look a bit paranoid, but it's supposed to cooperate with
2163 * a wide range of boards and BIOS bugs. Fortunately only the timer IRQ 2125 * a wide range of boards and BIOS bugs. Fortunately only the timer IRQ
@@ -2168,10 +2130,14 @@ static inline void __init check_timer(void)
2168{ 2130{
2169 int apic1, pin1, apic2, pin2; 2131 int apic1, pin1, apic2, pin2;
2170 int vector; 2132 int vector;
2133 unsigned int ver;
2171 unsigned long flags; 2134 unsigned long flags;
2172 2135
2173 local_irq_save(flags); 2136 local_irq_save(flags);
2174 2137
2138 ver = apic_read(APIC_LVR);
2139 ver = GET_APIC_VERSION(ver);
2140
2175 /* 2141 /*
2176 * get/set the timer IRQ vector: 2142 * get/set the timer IRQ vector:
2177 */ 2143 */
@@ -2184,11 +2150,15 @@ static inline void __init check_timer(void)
2184 * mode for the 8259A whenever interrupts are routed 2150 * mode for the 8259A whenever interrupts are routed
2185 * through I/O APICs. Also IRQ0 has to be enabled in 2151 * through I/O APICs. Also IRQ0 has to be enabled in
2186 * the 8259A which implies the virtual wire has to be 2152 * the 8259A which implies the virtual wire has to be
2187 * disabled in the local APIC. 2153 * disabled in the local APIC. Finally timer interrupts
2154 * need to be acknowledged manually in the 8259A for
2155 * timer_interrupt() and for the i82489DX when using
2156 * the NMI watchdog.
2188 */ 2157 */
2189 apic_write_around(APIC_LVT0, APIC_LVT_MASKED | APIC_DM_EXTINT); 2158 apic_write_around(APIC_LVT0, APIC_LVT_MASKED | APIC_DM_EXTINT);
2190 init_8259A(1); 2159 init_8259A(1);
2191 timer_ack = 1; 2160 timer_ack = !cpu_has_tsc;
2161 timer_ack |= (nmi_watchdog == NMI_IO_APIC && !APIC_INTEGRATED(ver));
2192 if (timer_over_8254 > 0) 2162 if (timer_over_8254 > 0)
2193 enable_8259A_irq(0); 2163 enable_8259A_irq(0);
2194 2164
@@ -2197,9 +2167,6 @@ static inline void __init check_timer(void)
2197 pin2 = ioapic_i8259.pin; 2167 pin2 = ioapic_i8259.pin;
2198 apic2 = ioapic_i8259.apic; 2168 apic2 = ioapic_i8259.apic;
2199 2169
2200 if (pin1 == 0)
2201 timer_uses_ioapic_pin_0 = 1;
2202
2203 printk(KERN_INFO "..TIMER: vector=0x%02X apic1=%d pin1=%d apic2=%d pin2=%d\n", 2170 printk(KERN_INFO "..TIMER: vector=0x%02X apic1=%d pin1=%d apic2=%d pin2=%d\n",
2204 vector, apic1, pin1, apic2, pin2); 2171 vector, apic1, pin1, apic2, pin2);
2205 2172
@@ -2789,7 +2756,6 @@ int __init io_apic_get_redir_entries (int ioapic)
2789int io_apic_set_pci_routing (int ioapic, int pin, int irq, int edge_level, int active_high_low) 2756int io_apic_set_pci_routing (int ioapic, int pin, int irq, int edge_level, int active_high_low)
2790{ 2757{
2791 struct IO_APIC_route_entry entry; 2758 struct IO_APIC_route_entry entry;
2792 unsigned long flags;
2793 2759
2794 if (!IO_APIC_IRQ(irq)) { 2760 if (!IO_APIC_IRQ(irq)) {
2795 printk(KERN_ERR "IOAPIC[%d]: Invalid reference to IRQ 0\n", 2761 printk(KERN_ERR "IOAPIC[%d]: Invalid reference to IRQ 0\n",
@@ -2830,9 +2796,7 @@ int io_apic_set_pci_routing (int ioapic, int pin, int irq, int edge_level, int a
2830 if (!ioapic && (irq < 16)) 2796 if (!ioapic && (irq < 16))
2831 disable_8259A_irq(irq); 2797 disable_8259A_irq(irq);
2832 2798
2833 spin_lock_irqsave(&ioapic_lock, flags); 2799 ioapic_write_entry(ioapic, pin, entry);
2834 __ioapic_write_entry(ioapic, pin, entry);
2835 spin_unlock_irqrestore(&ioapic_lock, flags);
2836 2800
2837 return 0; 2801 return 0;
2838} 2802}
diff --git a/arch/x86/kernel/io_apic_64.c b/arch/x86/kernel/io_apic_64.c
index 1627c0d53e0b..b54464b26658 100644
--- a/arch/x86/kernel/io_apic_64.c
+++ b/arch/x86/kernel/io_apic_64.c
@@ -43,13 +43,15 @@
43#include <asm/smp.h> 43#include <asm/smp.h>
44#include <asm/desc.h> 44#include <asm/desc.h>
45#include <asm/proto.h> 45#include <asm/proto.h>
46#include <asm/mach_apic.h>
47#include <asm/acpi.h> 46#include <asm/acpi.h>
48#include <asm/dma.h> 47#include <asm/dma.h>
49#include <asm/nmi.h> 48#include <asm/nmi.h>
50#include <asm/msidef.h> 49#include <asm/msidef.h>
51#include <asm/hypertransport.h> 50#include <asm/hypertransport.h>
52 51
52#include <mach_ipi.h>
53#include <mach_apic.h>
54
53struct irq_cfg { 55struct irq_cfg {
54 cpumask_t domain; 56 cpumask_t domain;
55 cpumask_t old_domain; 57 cpumask_t old_domain;
@@ -101,6 +103,16 @@ DEFINE_SPINLOCK(vector_lock);
101 */ 103 */
102int nr_ioapic_registers[MAX_IO_APICS]; 104int nr_ioapic_registers[MAX_IO_APICS];
103 105
106/* I/O APIC entries */
107struct mpc_config_ioapic mp_ioapics[MAX_IO_APICS];
108int nr_ioapics;
109
110/* MP IRQ source entries */
111struct mpc_config_intsrc mp_irqs[MAX_IRQ_SOURCES];
112
113/* # of MP IRQ source entries */
114int mp_irq_entries;
115
104/* 116/*
105 * Rough estimation of how many shared IRQs there are, can 117 * Rough estimation of how many shared IRQs there are, can
106 * be changed anytime. 118 * be changed anytime.
@@ -155,11 +167,10 @@ static inline void io_apic_modify(unsigned int apic, unsigned int value)
155 writel(value, &io_apic->data); 167 writel(value, &io_apic->data);
156} 168}
157 169
158static int io_apic_level_ack_pending(unsigned int irq) 170static bool io_apic_level_ack_pending(unsigned int irq)
159{ 171{
160 struct irq_pin_list *entry; 172 struct irq_pin_list *entry;
161 unsigned long flags; 173 unsigned long flags;
162 int pending = 0;
163 174
164 spin_lock_irqsave(&ioapic_lock, flags); 175 spin_lock_irqsave(&ioapic_lock, flags);
165 entry = irq_2_pin + irq; 176 entry = irq_2_pin + irq;
@@ -172,13 +183,17 @@ static int io_apic_level_ack_pending(unsigned int irq)
172 break; 183 break;
173 reg = io_apic_read(entry->apic, 0x10 + pin*2); 184 reg = io_apic_read(entry->apic, 0x10 + pin*2);
174 /* Is the remote IRR bit set? */ 185 /* Is the remote IRR bit set? */
175 pending |= (reg >> 14) & 1; 186 if ((reg >> 14) & 1) {
187 spin_unlock_irqrestore(&ioapic_lock, flags);
188 return true;
189 }
176 if (!entry->next) 190 if (!entry->next)
177 break; 191 break;
178 entry = irq_2_pin + entry->next; 192 entry = irq_2_pin + entry->next;
179 } 193 }
180 spin_unlock_irqrestore(&ioapic_lock, flags); 194 spin_unlock_irqrestore(&ioapic_lock, flags);
181 return pending; 195
196 return false;
182} 197}
183 198
184/* 199/*
@@ -902,9 +917,8 @@ static void __init setup_IO_APIC_irqs(void)
902static void __init setup_ExtINT_IRQ0_pin(unsigned int apic, unsigned int pin, int vector) 917static void __init setup_ExtINT_IRQ0_pin(unsigned int apic, unsigned int pin, int vector)
903{ 918{
904 struct IO_APIC_route_entry entry; 919 struct IO_APIC_route_entry entry;
905 unsigned long flags;
906 920
907 memset(&entry,0,sizeof(entry)); 921 memset(&entry, 0, sizeof(entry));
908 922
909 disable_8259A_irq(0); 923 disable_8259A_irq(0);
910 924
@@ -932,10 +946,7 @@ static void __init setup_ExtINT_IRQ0_pin(unsigned int apic, unsigned int pin, in
932 /* 946 /*
933 * Add it to the IO-APIC irq-routing table: 947 * Add it to the IO-APIC irq-routing table:
934 */ 948 */
935 spin_lock_irqsave(&ioapic_lock, flags); 949 ioapic_write_entry(apic, pin, entry);
936 io_apic_write(apic, 0x11+2*pin, *(((int *)&entry)+1));
937 io_apic_write(apic, 0x10+2*pin, *(((int *)&entry)+0));
938 spin_unlock_irqrestore(&ioapic_lock, flags);
939 950
940 enable_8259A_irq(0); 951 enable_8259A_irq(0);
941} 952}
@@ -1066,8 +1077,7 @@ void __apicdebuginit print_local_APIC(void * dummy)
1066 1077
1067 printk("\n" KERN_DEBUG "printing local APIC contents on CPU#%d/%d:\n", 1078 printk("\n" KERN_DEBUG "printing local APIC contents on CPU#%d/%d:\n",
1068 smp_processor_id(), hard_smp_processor_id()); 1079 smp_processor_id(), hard_smp_processor_id());
1069 v = apic_read(APIC_ID); 1080 printk(KERN_INFO "... APIC ID: %08x (%01x)\n", v, GET_APIC_ID(read_apic_id()));
1070 printk(KERN_INFO "... APIC ID: %08x (%01x)\n", v, GET_APIC_ID(v));
1071 v = apic_read(APIC_LVR); 1081 v = apic_read(APIC_LVR);
1072 printk(KERN_INFO "... APIC VERSION: %08x\n", v); 1082 printk(KERN_INFO "... APIC VERSION: %08x\n", v);
1073 ver = GET_APIC_VERSION(v); 1083 ver = GET_APIC_VERSION(v);
@@ -1261,7 +1271,7 @@ void disable_IO_APIC(void)
1261 entry.dest_mode = 0; /* Physical */ 1271 entry.dest_mode = 0; /* Physical */
1262 entry.delivery_mode = dest_ExtINT; /* ExtInt */ 1272 entry.delivery_mode = dest_ExtINT; /* ExtInt */
1263 entry.vector = 0; 1273 entry.vector = 0;
1264 entry.dest = GET_APIC_ID(apic_read(APIC_ID)); 1274 entry.dest = GET_APIC_ID(read_apic_id());
1265 1275
1266 /* 1276 /*
1267 * Add it to the IO-APIC irq-routing table: 1277 * Add it to the IO-APIC irq-routing table:
@@ -1352,9 +1362,7 @@ static int ioapic_retrigger_irq(unsigned int irq)
1352 unsigned long flags; 1362 unsigned long flags;
1353 1363
1354 spin_lock_irqsave(&vector_lock, flags); 1364 spin_lock_irqsave(&vector_lock, flags);
1355 cpus_clear(mask); 1365 mask = cpumask_of_cpu(first_cpu(cfg->domain));
1356 cpu_set(first_cpu(cfg->domain), mask);
1357
1358 send_IPI_mask(mask, cfg->vector); 1366 send_IPI_mask(mask, cfg->vector);
1359 spin_unlock_irqrestore(&vector_lock, flags); 1367 spin_unlock_irqrestore(&vector_lock, flags);
1360 1368
@@ -1517,8 +1525,7 @@ static inline void init_IO_APIC_traps(void)
1517 * 0x80, because int 0x80 is hm, kind of importantish. ;) 1525 * 0x80, because int 0x80 is hm, kind of importantish. ;)
1518 */ 1526 */
1519 for (irq = 0; irq < NR_IRQS ; irq++) { 1527 for (irq = 0; irq < NR_IRQS ; irq++) {
1520 int tmp = irq; 1528 if (IO_APIC_IRQ(irq) && !irq_cfg[irq].vector) {
1521 if (IO_APIC_IRQ(tmp) && !irq_cfg[tmp].vector) {
1522 /* 1529 /*
1523 * Hmm.. We don't have an entry for this, 1530 * Hmm.. We don't have an entry for this,
1524 * so default to an old-fashioned 8259 1531 * so default to an old-fashioned 8259
@@ -1597,17 +1604,14 @@ static inline void unlock_ExtINT_logic(void)
1597 int apic, pin, i; 1604 int apic, pin, i;
1598 struct IO_APIC_route_entry entry0, entry1; 1605 struct IO_APIC_route_entry entry0, entry1;
1599 unsigned char save_control, save_freq_select; 1606 unsigned char save_control, save_freq_select;
1600 unsigned long flags;
1601 1607
1602 pin = find_isa_irq_pin(8, mp_INT); 1608 pin = find_isa_irq_pin(8, mp_INT);
1603 apic = find_isa_irq_apic(8, mp_INT); 1609 apic = find_isa_irq_apic(8, mp_INT);
1604 if (pin == -1) 1610 if (pin == -1)
1605 return; 1611 return;
1606 1612
1607 spin_lock_irqsave(&ioapic_lock, flags); 1613 entry0 = ioapic_read_entry(apic, pin);
1608 *(((int *)&entry0) + 1) = io_apic_read(apic, 0x11 + 2 * pin); 1614
1609 *(((int *)&entry0) + 0) = io_apic_read(apic, 0x10 + 2 * pin);
1610 spin_unlock_irqrestore(&ioapic_lock, flags);
1611 clear_IO_APIC_pin(apic, pin); 1615 clear_IO_APIC_pin(apic, pin);
1612 1616
1613 memset(&entry1, 0, sizeof(entry1)); 1617 memset(&entry1, 0, sizeof(entry1));
@@ -1620,10 +1624,7 @@ static inline void unlock_ExtINT_logic(void)
1620 entry1.trigger = 0; 1624 entry1.trigger = 0;
1621 entry1.vector = 0; 1625 entry1.vector = 0;
1622 1626
1623 spin_lock_irqsave(&ioapic_lock, flags); 1627 ioapic_write_entry(apic, pin, entry1);
1624 io_apic_write(apic, 0x11 + 2 * pin, *(((int *)&entry1) + 1));
1625 io_apic_write(apic, 0x10 + 2 * pin, *(((int *)&entry1) + 0));
1626 spin_unlock_irqrestore(&ioapic_lock, flags);
1627 1628
1628 save_control = CMOS_READ(RTC_CONTROL); 1629 save_control = CMOS_READ(RTC_CONTROL);
1629 save_freq_select = CMOS_READ(RTC_FREQ_SELECT); 1630 save_freq_select = CMOS_READ(RTC_FREQ_SELECT);
@@ -1642,10 +1643,7 @@ static inline void unlock_ExtINT_logic(void)
1642 CMOS_WRITE(save_freq_select, RTC_FREQ_SELECT); 1643 CMOS_WRITE(save_freq_select, RTC_FREQ_SELECT);
1643 clear_IO_APIC_pin(apic, pin); 1644 clear_IO_APIC_pin(apic, pin);
1644 1645
1645 spin_lock_irqsave(&ioapic_lock, flags); 1646 ioapic_write_entry(apic, pin, entry0);
1646 io_apic_write(apic, 0x11 + 2 * pin, *(((int *)&entry0) + 1));
1647 io_apic_write(apic, 0x10 + 2 * pin, *(((int *)&entry0) + 0));
1648 spin_unlock_irqrestore(&ioapic_lock, flags);
1649} 1647}
1650 1648
1651/* 1649/*
@@ -2314,7 +2312,6 @@ static struct resource * __init ioapic_setup_resources(void)
2314 res = (void *)mem; 2312 res = (void *)mem;
2315 2313
2316 if (mem != NULL) { 2314 if (mem != NULL) {
2317 memset(mem, 0, n);
2318 mem += sizeof(struct resource) * nr_ioapics; 2315 mem += sizeof(struct resource) * nr_ioapics;
2319 2316
2320 for (i = 0; i < nr_ioapics; i++) { 2317 for (i = 0; i < nr_ioapics; i++) {
diff --git a/arch/x86/kernel/ipi.c b/arch/x86/kernel/ipi.c
new file mode 100644
index 000000000000..c0df7b89ca23
--- /dev/null
+++ b/arch/x86/kernel/ipi.c
@@ -0,0 +1,178 @@
1#include <linux/cpumask.h>
2#include <linux/interrupt.h>
3#include <linux/init.h>
4
5#include <linux/mm.h>
6#include <linux/delay.h>
7#include <linux/spinlock.h>
8#include <linux/kernel_stat.h>
9#include <linux/mc146818rtc.h>
10#include <linux/cache.h>
11#include <linux/interrupt.h>
12#include <linux/cpu.h>
13#include <linux/module.h>
14
15#include <asm/smp.h>
16#include <asm/mtrr.h>
17#include <asm/tlbflush.h>
18#include <asm/mmu_context.h>
19#include <asm/apic.h>
20#include <asm/proto.h>
21
22#ifdef CONFIG_X86_32
23#include <mach_apic.h>
24/*
25 * the following functions deal with sending IPIs between CPUs.
26 *
27 * We use 'broadcast', CPU->CPU IPIs and self-IPIs too.
28 */
29
30static inline int __prepare_ICR(unsigned int shortcut, int vector)
31{
32 unsigned int icr = shortcut | APIC_DEST_LOGICAL;
33
34 switch (vector) {
35 default:
36 icr |= APIC_DM_FIXED | vector;
37 break;
38 case NMI_VECTOR:
39 icr |= APIC_DM_NMI;
40 break;
41 }
42 return icr;
43}
44
45static inline int __prepare_ICR2(unsigned int mask)
46{
47 return SET_APIC_DEST_FIELD(mask);
48}
49
50void __send_IPI_shortcut(unsigned int shortcut, int vector)
51{
52 /*
53 * Subtle. In the case of the 'never do double writes' workaround
54 * we have to lock out interrupts to be safe. As we don't care
55 * of the value read we use an atomic rmw access to avoid costly
56 * cli/sti. Otherwise we use an even cheaper single atomic write
57 * to the APIC.
58 */
59 unsigned int cfg;
60
61 /*
62 * Wait for idle.
63 */
64 apic_wait_icr_idle();
65
66 /*
67 * No need to touch the target chip field
68 */
69 cfg = __prepare_ICR(shortcut, vector);
70
71 /*
72 * Send the IPI. The write to APIC_ICR fires this off.
73 */
74 apic_write_around(APIC_ICR, cfg);
75}
76
77void send_IPI_self(int vector)
78{
79 __send_IPI_shortcut(APIC_DEST_SELF, vector);
80}
81
82/*
83 * This is used to send an IPI with no shorthand notation (the destination is
84 * specified in bits 56 to 63 of the ICR).
85 */
86static inline void __send_IPI_dest_field(unsigned long mask, int vector)
87{
88 unsigned long cfg;
89
90 /*
91 * Wait for idle.
92 */
93 if (unlikely(vector == NMI_VECTOR))
94 safe_apic_wait_icr_idle();
95 else
96 apic_wait_icr_idle();
97
98 /*
99 * prepare target chip field
100 */
101 cfg = __prepare_ICR2(mask);
102 apic_write_around(APIC_ICR2, cfg);
103
104 /*
105 * program the ICR
106 */
107 cfg = __prepare_ICR(0, vector);
108
109 /*
110 * Send the IPI. The write to APIC_ICR fires this off.
111 */
112 apic_write_around(APIC_ICR, cfg);
113}
114
115/*
116 * This is only used on smaller machines.
117 */
118void send_IPI_mask_bitmask(cpumask_t cpumask, int vector)
119{
120 unsigned long mask = cpus_addr(cpumask)[0];
121 unsigned long flags;
122
123 local_irq_save(flags);
124 WARN_ON(mask & ~cpus_addr(cpu_online_map)[0]);
125 __send_IPI_dest_field(mask, vector);
126 local_irq_restore(flags);
127}
128
129void send_IPI_mask_sequence(cpumask_t mask, int vector)
130{
131 unsigned long flags;
132 unsigned int query_cpu;
133
134 /*
135 * Hack. The clustered APIC addressing mode doesn't allow us to send
136 * to an arbitrary mask, so I do a unicasts to each CPU instead. This
137 * should be modified to do 1 message per cluster ID - mbligh
138 */
139
140 local_irq_save(flags);
141 for_each_possible_cpu(query_cpu) {
142 if (cpu_isset(query_cpu, mask)) {
143 __send_IPI_dest_field(cpu_to_logical_apicid(query_cpu),
144 vector);
145 }
146 }
147 local_irq_restore(flags);
148}
149
150/* must come after the send_IPI functions above for inlining */
151#include <mach_ipi.h>
152static int convert_apicid_to_cpu(int apic_id)
153{
154 int i;
155
156 for_each_possible_cpu(i) {
157 if (per_cpu(x86_cpu_to_apicid, i) == apic_id)
158 return i;
159 }
160 return -1;
161}
162
163int safe_smp_processor_id(void)
164{
165 int apicid, cpuid;
166
167 if (!boot_cpu_has(X86_FEATURE_APIC))
168 return 0;
169
170 apicid = hard_smp_processor_id();
171 if (apicid == BAD_APICID)
172 return 0;
173
174 cpuid = convert_apicid_to_cpu(apicid);
175
176 return cpuid >= 0 ? cpuid : 0;
177}
178#endif
diff --git a/arch/x86/kernel/irq_32.c b/arch/x86/kernel/irq_32.c
index cef054b09d27..6ea67b76a214 100644
--- a/arch/x86/kernel/irq_32.c
+++ b/arch/x86/kernel/irq_32.c
@@ -79,7 +79,7 @@ unsigned int do_IRQ(struct pt_regs *regs)
79 79
80 if (unlikely((unsigned)irq >= NR_IRQS)) { 80 if (unlikely((unsigned)irq >= NR_IRQS)) {
81 printk(KERN_EMERG "%s: cannot handle IRQ %d\n", 81 printk(KERN_EMERG "%s: cannot handle IRQ %d\n",
82 __FUNCTION__, irq); 82 __func__, irq);
83 BUG(); 83 BUG();
84 } 84 }
85 85
diff --git a/arch/x86/kernel/kprobes.c b/arch/x86/kernel/kprobes.c
index 34a591283f5d..b8c6743a13da 100644
--- a/arch/x86/kernel/kprobes.c
+++ b/arch/x86/kernel/kprobes.c
@@ -410,13 +410,13 @@ static void __kprobes set_current_kprobe(struct kprobe *p, struct pt_regs *regs,
410static void __kprobes clear_btf(void) 410static void __kprobes clear_btf(void)
411{ 411{
412 if (test_thread_flag(TIF_DEBUGCTLMSR)) 412 if (test_thread_flag(TIF_DEBUGCTLMSR))
413 wrmsrl(MSR_IA32_DEBUGCTLMSR, 0); 413 update_debugctlmsr(0);
414} 414}
415 415
416static void __kprobes restore_btf(void) 416static void __kprobes restore_btf(void)
417{ 417{
418 if (test_thread_flag(TIF_DEBUGCTLMSR)) 418 if (test_thread_flag(TIF_DEBUGCTLMSR))
419 wrmsrl(MSR_IA32_DEBUGCTLMSR, current->thread.debugctlmsr); 419 update_debugctlmsr(current->thread.debugctlmsr);
420} 420}
421 421
422static void __kprobes prepare_singlestep(struct kprobe *p, struct pt_regs *regs) 422static void __kprobes prepare_singlestep(struct kprobe *p, struct pt_regs *regs)
@@ -489,7 +489,7 @@ static int __kprobes reenter_kprobe(struct kprobe *p, struct pt_regs *regs,
489 break; 489 break;
490 case KPROBE_HIT_SS: 490 case KPROBE_HIT_SS:
491 if (p == kprobe_running()) { 491 if (p == kprobe_running()) {
492 regs->flags &= ~TF_MASK; 492 regs->flags &= ~X86_EFLAGS_TF;
493 regs->flags |= kcb->kprobe_saved_flags; 493 regs->flags |= kcb->kprobe_saved_flags;
494 return 0; 494 return 0;
495 } else { 495 } else {
@@ -858,15 +858,15 @@ static int __kprobes post_kprobe_handler(struct pt_regs *regs)
858 if (!cur) 858 if (!cur)
859 return 0; 859 return 0;
860 860
861 resume_execution(cur, regs, kcb);
862 regs->flags |= kcb->kprobe_saved_flags;
863 trace_hardirqs_fixup_flags(regs->flags);
864
861 if ((kcb->kprobe_status != KPROBE_REENTER) && cur->post_handler) { 865 if ((kcb->kprobe_status != KPROBE_REENTER) && cur->post_handler) {
862 kcb->kprobe_status = KPROBE_HIT_SSDONE; 866 kcb->kprobe_status = KPROBE_HIT_SSDONE;
863 cur->post_handler(cur, regs, 0); 867 cur->post_handler(cur, regs, 0);
864 } 868 }
865 869
866 resume_execution(cur, regs, kcb);
867 regs->flags |= kcb->kprobe_saved_flags;
868 trace_hardirqs_fixup_flags(regs->flags);
869
870 /* Restore back the original saved kprobes variables and continue. */ 870 /* Restore back the original saved kprobes variables and continue. */
871 if (kcb->kprobe_status == KPROBE_REENTER) { 871 if (kcb->kprobe_status == KPROBE_REENTER) {
872 restore_previous_kprobe(kcb); 872 restore_previous_kprobe(kcb);
diff --git a/arch/x86/kernel/mca_32.c b/arch/x86/kernel/mca_32.c
index 9482033ed0fe..2dc183758be3 100644
--- a/arch/x86/kernel/mca_32.c
+++ b/arch/x86/kernel/mca_32.c
@@ -53,9 +53,9 @@
53#include <linux/init.h> 53#include <linux/init.h>
54#include <asm/arch_hooks.h> 54#include <asm/arch_hooks.h>
55 55
56static unsigned char which_scsi = 0; 56static unsigned char which_scsi;
57 57
58int MCA_bus = 0; 58int MCA_bus;
59EXPORT_SYMBOL(MCA_bus); 59EXPORT_SYMBOL(MCA_bus);
60 60
61/* 61/*
@@ -68,15 +68,17 @@ static DEFINE_SPINLOCK(mca_lock);
68 68
69/* Build the status info for the adapter */ 69/* Build the status info for the adapter */
70 70
71static void mca_configure_adapter_status(struct mca_device *mca_dev) { 71static void mca_configure_adapter_status(struct mca_device *mca_dev)
72{
72 mca_dev->status = MCA_ADAPTER_NONE; 73 mca_dev->status = MCA_ADAPTER_NONE;
73 74
74 mca_dev->pos_id = mca_dev->pos[0] 75 mca_dev->pos_id = mca_dev->pos[0]
75 + (mca_dev->pos[1] << 8); 76 + (mca_dev->pos[1] << 8);
76 77
77 if(!mca_dev->pos_id && mca_dev->slot < MCA_MAX_SLOT_NR) { 78 if (!mca_dev->pos_id && mca_dev->slot < MCA_MAX_SLOT_NR) {
78 79
79 /* id = 0x0000 usually indicates hardware failure, 80 /*
81 * id = 0x0000 usually indicates hardware failure,
80 * however, ZP Gu (zpg@castle.net> reports that his 9556 82 * however, ZP Gu (zpg@castle.net> reports that his 9556
81 * has 0x0000 as id and everything still works. There 83 * has 0x0000 as id and everything still works. There
82 * also seem to be an adapter with id = 0x0000; the 84 * also seem to be an adapter with id = 0x0000; the
@@ -87,9 +89,10 @@ static void mca_configure_adapter_status(struct mca_device *mca_dev) {
87 mca_dev->status = MCA_ADAPTER_ERROR; 89 mca_dev->status = MCA_ADAPTER_ERROR;
88 90
89 return; 91 return;
90 } else if(mca_dev->pos_id != 0xffff) { 92 } else if (mca_dev->pos_id != 0xffff) {
91 93
92 /* 0xffff usually indicates that there's no adapter, 94 /*
95 * 0xffff usually indicates that there's no adapter,
93 * however, some integrated adapters may have 0xffff as 96 * however, some integrated adapters may have 0xffff as
94 * their id and still be valid. Examples are on-board 97 * their id and still be valid. Examples are on-board
95 * VGA of the 55sx, the integrated SCSI of the 56 & 57, 98 * VGA of the 55sx, the integrated SCSI of the 56 & 57,
@@ -99,19 +102,19 @@ static void mca_configure_adapter_status(struct mca_device *mca_dev) {
99 mca_dev->status = MCA_ADAPTER_NORMAL; 102 mca_dev->status = MCA_ADAPTER_NORMAL;
100 } 103 }
101 104
102 if((mca_dev->pos_id == 0xffff || 105 if ((mca_dev->pos_id == 0xffff ||
103 mca_dev->pos_id == 0x0000) && mca_dev->slot >= MCA_MAX_SLOT_NR) { 106 mca_dev->pos_id == 0x0000) && mca_dev->slot >= MCA_MAX_SLOT_NR) {
104 int j; 107 int j;
105 108
106 for(j = 2; j < 8; j++) { 109 for (j = 2; j < 8; j++) {
107 if(mca_dev->pos[j] != 0xff) { 110 if (mca_dev->pos[j] != 0xff) {
108 mca_dev->status = MCA_ADAPTER_NORMAL; 111 mca_dev->status = MCA_ADAPTER_NORMAL;
109 break; 112 break;
110 } 113 }
111 } 114 }
112 } 115 }
113 116
114 if(!(mca_dev->pos[2] & MCA_ENABLED)) { 117 if (!(mca_dev->pos[2] & MCA_ENABLED)) {
115 118
116 /* enabled bit is in POS 2 */ 119 /* enabled bit is in POS 2 */
117 120
@@ -133,7 +136,7 @@ static struct resource mca_standard_resources[] = {
133 136
134#define MCA_STANDARD_RESOURCES ARRAY_SIZE(mca_standard_resources) 137#define MCA_STANDARD_RESOURCES ARRAY_SIZE(mca_standard_resources)
135 138
136/** 139/*
137 * mca_read_and_store_pos - read the POS registers into a memory buffer 140 * mca_read_and_store_pos - read the POS registers into a memory buffer
138 * @pos: a char pointer to 8 bytes, contains the POS register value on 141 * @pos: a char pointer to 8 bytes, contains the POS register value on
139 * successful return 142 * successful return
@@ -141,12 +144,14 @@ static struct resource mca_standard_resources[] = {
141 * Returns 1 if a card actually exists (i.e. the pos isn't 144 * Returns 1 if a card actually exists (i.e. the pos isn't
142 * all 0xff) or 0 otherwise 145 * all 0xff) or 0 otherwise
143 */ 146 */
144static int mca_read_and_store_pos(unsigned char *pos) { 147static int mca_read_and_store_pos(unsigned char *pos)
148{
145 int j; 149 int j;
146 int found = 0; 150 int found = 0;
147 151
148 for(j=0; j<8; j++) { 152 for (j = 0; j < 8; j++) {
149 if((pos[j] = inb_p(MCA_POS_REG(j))) != 0xff) { 153 pos[j] = inb_p(MCA_POS_REG(j));
154 if (pos[j] != 0xff) {
150 /* 0xff all across means no device. 0x00 means 155 /* 0xff all across means no device. 0x00 means
151 * something's broken, but a device is 156 * something's broken, but a device is
152 * probably there. However, if you get 0x00 157 * probably there. However, if you get 0x00
@@ -167,11 +172,11 @@ static unsigned char mca_pc_read_pos(struct mca_device *mca_dev, int reg)
167 unsigned char byte; 172 unsigned char byte;
168 unsigned long flags; 173 unsigned long flags;
169 174
170 if(reg < 0 || reg >= 8) 175 if (reg < 0 || reg >= 8)
171 return 0; 176 return 0;
172 177
173 spin_lock_irqsave(&mca_lock, flags); 178 spin_lock_irqsave(&mca_lock, flags);
174 if(mca_dev->pos_register) { 179 if (mca_dev->pos_register) {
175 /* Disable adapter setup, enable motherboard setup */ 180 /* Disable adapter setup, enable motherboard setup */
176 181
177 outb_p(0, MCA_ADAPTER_SETUP_REG); 182 outb_p(0, MCA_ADAPTER_SETUP_REG);
@@ -203,7 +208,7 @@ static void mca_pc_write_pos(struct mca_device *mca_dev, int reg,
203{ 208{
204 unsigned long flags; 209 unsigned long flags;
205 210
206 if(reg < 0 || reg >= 8) 211 if (reg < 0 || reg >= 8)
207 return; 212 return;
208 213
209 spin_lock_irqsave(&mca_lock, flags); 214 spin_lock_irqsave(&mca_lock, flags);
@@ -227,17 +232,17 @@ static void mca_pc_write_pos(struct mca_device *mca_dev, int reg,
227} 232}
228 233
229/* for the primary MCA bus, we have identity transforms */ 234/* for the primary MCA bus, we have identity transforms */
230static int mca_dummy_transform_irq(struct mca_device * mca_dev, int irq) 235static int mca_dummy_transform_irq(struct mca_device *mca_dev, int irq)
231{ 236{
232 return irq; 237 return irq;
233} 238}
234 239
235static int mca_dummy_transform_ioport(struct mca_device * mca_dev, int port) 240static int mca_dummy_transform_ioport(struct mca_device *mca_dev, int port)
236{ 241{
237 return port; 242 return port;
238} 243}
239 244
240static void *mca_dummy_transform_memory(struct mca_device * mca_dev, void *mem) 245static void *mca_dummy_transform_memory(struct mca_device *mca_dev, void *mem)
241{ 246{
242 return mem; 247 return mem;
243} 248}
@@ -251,7 +256,8 @@ static int __init mca_init(void)
251 short mca_builtin_scsi_ports[] = {0xf7, 0xfd, 0x00}; 256 short mca_builtin_scsi_ports[] = {0xf7, 0xfd, 0x00};
252 struct mca_bus *bus; 257 struct mca_bus *bus;
253 258
254 /* WARNING: Be careful when making changes here. Putting an adapter 259 /*
260 * WARNING: Be careful when making changes here. Putting an adapter
255 * and the motherboard simultaneously into setup mode may result in 261 * and the motherboard simultaneously into setup mode may result in
256 * damage to chips (according to The Indispensible PC Hardware Book 262 * damage to chips (according to The Indispensible PC Hardware Book
257 * by Hans-Peter Messmer). Also, we disable system interrupts (so 263 * by Hans-Peter Messmer). Also, we disable system interrupts (so
@@ -283,7 +289,7 @@ static int __init mca_init(void)
283 289
284 /* get the motherboard device */ 290 /* get the motherboard device */
285 mca_dev = kzalloc(sizeof(struct mca_device), GFP_KERNEL); 291 mca_dev = kzalloc(sizeof(struct mca_device), GFP_KERNEL);
286 if(unlikely(!mca_dev)) 292 if (unlikely(!mca_dev))
287 goto out_nomem; 293 goto out_nomem;
288 294
289 /* 295 /*
@@ -309,7 +315,7 @@ static int __init mca_init(void)
309 mca_register_device(MCA_PRIMARY_BUS, mca_dev); 315 mca_register_device(MCA_PRIMARY_BUS, mca_dev);
310 316
311 mca_dev = kzalloc(sizeof(struct mca_device), GFP_ATOMIC); 317 mca_dev = kzalloc(sizeof(struct mca_device), GFP_ATOMIC);
312 if(unlikely(!mca_dev)) 318 if (unlikely(!mca_dev))
313 goto out_unlock_nomem; 319 goto out_unlock_nomem;
314 320
315 /* Put motherboard into video setup mode, read integrated video 321 /* Put motherboard into video setup mode, read integrated video
@@ -326,7 +332,8 @@ static int __init mca_init(void)
326 mca_dev->slot = MCA_INTEGVIDEO; 332 mca_dev->slot = MCA_INTEGVIDEO;
327 mca_register_device(MCA_PRIMARY_BUS, mca_dev); 333 mca_register_device(MCA_PRIMARY_BUS, mca_dev);
328 334
329 /* Put motherboard into scsi setup mode, read integrated scsi 335 /*
336 * Put motherboard into scsi setup mode, read integrated scsi
330 * POS registers, and turn motherboard setup off. 337 * POS registers, and turn motherboard setup off.
331 * 338 *
332 * It seems there are two possible SCSI registers. Martin says that 339 * It seems there are two possible SCSI registers. Martin says that
@@ -338,18 +345,18 @@ static int __init mca_init(void)
338 * machine. 345 * machine.
339 */ 346 */
340 347
341 for(i = 0; (which_scsi = mca_builtin_scsi_ports[i]) != 0; i++) { 348 for (i = 0; (which_scsi = mca_builtin_scsi_ports[i]) != 0; i++) {
342 outb_p(which_scsi, MCA_MOTHERBOARD_SETUP_REG); 349 outb_p(which_scsi, MCA_MOTHERBOARD_SETUP_REG);
343 if(mca_read_and_store_pos(pos)) 350 if (mca_read_and_store_pos(pos))
344 break; 351 break;
345 } 352 }
346 if(which_scsi) { 353 if (which_scsi) {
347 /* found a scsi card */ 354 /* found a scsi card */
348 mca_dev = kzalloc(sizeof(struct mca_device), GFP_ATOMIC); 355 mca_dev = kzalloc(sizeof(struct mca_device), GFP_ATOMIC);
349 if(unlikely(!mca_dev)) 356 if (unlikely(!mca_dev))
350 goto out_unlock_nomem; 357 goto out_unlock_nomem;
351 358
352 for(j = 0; j < 8; j++) 359 for (j = 0; j < 8; j++)
353 mca_dev->pos[j] = pos[j]; 360 mca_dev->pos[j] = pos[j];
354 361
355 mca_configure_adapter_status(mca_dev); 362 mca_configure_adapter_status(mca_dev);
@@ -364,21 +371,22 @@ static int __init mca_init(void)
364 371
365 outb_p(0xff, MCA_MOTHERBOARD_SETUP_REG); 372 outb_p(0xff, MCA_MOTHERBOARD_SETUP_REG);
366 373
367 /* Now loop over MCA slots: put each adapter into setup mode, and 374 /*
375 * Now loop over MCA slots: put each adapter into setup mode, and
368 * read its POS registers. Then put adapter setup off. 376 * read its POS registers. Then put adapter setup off.
369 */ 377 */
370 378
371 for(i=0; i<MCA_MAX_SLOT_NR; i++) { 379 for (i = 0; i < MCA_MAX_SLOT_NR; i++) {
372 outb_p(0x8|(i&0xf), MCA_ADAPTER_SETUP_REG); 380 outb_p(0x8|(i&0xf), MCA_ADAPTER_SETUP_REG);
373 if(!mca_read_and_store_pos(pos)) 381 if (!mca_read_and_store_pos(pos))
374 continue; 382 continue;
375 383
376 mca_dev = kzalloc(sizeof(struct mca_device), GFP_ATOMIC); 384 mca_dev = kzalloc(sizeof(struct mca_device), GFP_ATOMIC);
377 if(unlikely(!mca_dev)) 385 if (unlikely(!mca_dev))
378 goto out_unlock_nomem; 386 goto out_unlock_nomem;
379 387
380 for(j=0; j<8; j++) 388 for (j = 0; j < 8; j++)
381 mca_dev->pos[j]=pos[j]; 389 mca_dev->pos[j] = pos[j];
382 390
383 mca_dev->driver_loaded = 0; 391 mca_dev->driver_loaded = 0;
384 mca_dev->slot = i; 392 mca_dev->slot = i;
@@ -414,20 +422,20 @@ mca_handle_nmi_device(struct mca_device *mca_dev, int check_flag)
414{ 422{
415 int slot = mca_dev->slot; 423 int slot = mca_dev->slot;
416 424
417 if(slot == MCA_INTEGSCSI) { 425 if (slot == MCA_INTEGSCSI) {
418 printk(KERN_CRIT "NMI: caused by MCA integrated SCSI adapter (%s)\n", 426 printk(KERN_CRIT "NMI: caused by MCA integrated SCSI adapter (%s)\n",
419 mca_dev->name); 427 mca_dev->name);
420 } else if(slot == MCA_INTEGVIDEO) { 428 } else if (slot == MCA_INTEGVIDEO) {
421 printk(KERN_CRIT "NMI: caused by MCA integrated video adapter (%s)\n", 429 printk(KERN_CRIT "NMI: caused by MCA integrated video adapter (%s)\n",
422 mca_dev->name); 430 mca_dev->name);
423 } else if(slot == MCA_MOTHERBOARD) { 431 } else if (slot == MCA_MOTHERBOARD) {
424 printk(KERN_CRIT "NMI: caused by motherboard (%s)\n", 432 printk(KERN_CRIT "NMI: caused by motherboard (%s)\n",
425 mca_dev->name); 433 mca_dev->name);
426 } 434 }
427 435
428 /* More info available in POS 6 and 7? */ 436 /* More info available in POS 6 and 7? */
429 437
430 if(check_flag) { 438 if (check_flag) {
431 unsigned char pos6, pos7; 439 unsigned char pos6, pos7;
432 440
433 pos6 = mca_device_read_pos(mca_dev, 6); 441 pos6 = mca_device_read_pos(mca_dev, 6);
@@ -447,8 +455,9 @@ static int __kprobes mca_handle_nmi_callback(struct device *dev, void *data)
447 455
448 pos5 = mca_device_read_pos(mca_dev, 5); 456 pos5 = mca_device_read_pos(mca_dev, 5);
449 457
450 if(!(pos5 & 0x80)) { 458 if (!(pos5 & 0x80)) {
451 /* Bit 7 of POS 5 is reset when this adapter has a hardware 459 /*
460 * Bit 7 of POS 5 is reset when this adapter has a hardware
452 * error. Bit 7 it reset if there's error information 461 * error. Bit 7 it reset if there's error information
453 * available in POS 6 and 7. 462 * available in POS 6 and 7.
454 */ 463 */
@@ -460,7 +469,8 @@ static int __kprobes mca_handle_nmi_callback(struct device *dev, void *data)
460 469
461void __kprobes mca_handle_nmi(void) 470void __kprobes mca_handle_nmi(void)
462{ 471{
463 /* First try - scan the various adapters and see if a specific 472 /*
473 * First try - scan the various adapters and see if a specific
464 * adapter was responsible for the error. 474 * adapter was responsible for the error.
465 */ 475 */
466 bus_for_each_dev(&mca_bus_type, NULL, NULL, mca_handle_nmi_callback); 476 bus_for_each_dev(&mca_bus_type, NULL, NULL, mca_handle_nmi_callback);
diff --git a/arch/x86/kernel/microcode.c b/arch/x86/kernel/microcode.c
index f2702d01b8a8..25cf6dee4e56 100644
--- a/arch/x86/kernel/microcode.c
+++ b/arch/x86/kernel/microcode.c
@@ -290,7 +290,7 @@ static int get_maching_microcode(void *mc, int cpu)
290 } 290 }
291 return 0; 291 return 0;
292find: 292find:
293 pr_debug("microcode: CPU %d found a matching microcode update with" 293 pr_debug("microcode: CPU%d found a matching microcode update with"
294 " version 0x%x (current=0x%x)\n", cpu, mc_header->rev,uci->rev); 294 " version 0x%x (current=0x%x)\n", cpu, mc_header->rev,uci->rev);
295 new_mc = vmalloc(total_size); 295 new_mc = vmalloc(total_size);
296 if (!new_mc) { 296 if (!new_mc) {
@@ -336,11 +336,11 @@ static void apply_microcode(int cpu)
336 336
337 spin_unlock_irqrestore(&microcode_update_lock, flags); 337 spin_unlock_irqrestore(&microcode_update_lock, flags);
338 if (val[1] != uci->mc->hdr.rev) { 338 if (val[1] != uci->mc->hdr.rev) {
339 printk(KERN_ERR "microcode: CPU%d updated from revision " 339 printk(KERN_ERR "microcode: CPU%d update from revision "
340 "0x%x to 0x%x failed\n", cpu_num, uci->rev, val[1]); 340 "0x%x to 0x%x failed\n", cpu_num, uci->rev, val[1]);
341 return; 341 return;
342 } 342 }
343 pr_debug("microcode: CPU%d updated from revision " 343 printk(KERN_INFO "microcode: CPU%d updated from revision "
344 "0x%x to 0x%x, date = %08x \n", 344 "0x%x to 0x%x, date = %08x \n",
345 cpu_num, uci->rev, val[1], uci->mc->hdr.date); 345 cpu_num, uci->rev, val[1], uci->mc->hdr.date);
346 uci->rev = val[1]; 346 uci->rev = val[1];
@@ -534,7 +534,7 @@ static int cpu_request_microcode(int cpu)
534 c->x86, c->x86_model, c->x86_mask); 534 c->x86, c->x86_model, c->x86_mask);
535 error = request_firmware(&firmware, name, &microcode_pdev->dev); 535 error = request_firmware(&firmware, name, &microcode_pdev->dev);
536 if (error) { 536 if (error) {
537 pr_debug("ucode data file %s load failed\n", name); 537 pr_debug("microcode: ucode data file %s load failed\n", name);
538 return error; 538 return error;
539 } 539 }
540 buf = firmware->data; 540 buf = firmware->data;
@@ -709,7 +709,7 @@ static int __mc_sysdev_add(struct sys_device *sys_dev, int resume)
709 if (!cpu_online(cpu)) 709 if (!cpu_online(cpu))
710 return 0; 710 return 0;
711 711
712 pr_debug("Microcode:CPU %d added\n", cpu); 712 pr_debug("microcode: CPU%d added\n", cpu);
713 memset(uci, 0, sizeof(*uci)); 713 memset(uci, 0, sizeof(*uci));
714 714
715 err = sysfs_create_group(&sys_dev->kobj, &mc_attr_group); 715 err = sysfs_create_group(&sys_dev->kobj, &mc_attr_group);
@@ -733,7 +733,7 @@ static int mc_sysdev_remove(struct sys_device *sys_dev)
733 if (!cpu_online(cpu)) 733 if (!cpu_online(cpu))
734 return 0; 734 return 0;
735 735
736 pr_debug("Microcode:CPU %d removed\n", cpu); 736 pr_debug("microcode: CPU%d removed\n", cpu);
737 microcode_fini_cpu(cpu); 737 microcode_fini_cpu(cpu);
738 sysfs_remove_group(&sys_dev->kobj, &mc_attr_group); 738 sysfs_remove_group(&sys_dev->kobj, &mc_attr_group);
739 return 0; 739 return 0;
@@ -745,7 +745,7 @@ static int mc_sysdev_resume(struct sys_device *dev)
745 745
746 if (!cpu_online(cpu)) 746 if (!cpu_online(cpu))
747 return 0; 747 return 0;
748 pr_debug("Microcode:CPU %d resumed\n", cpu); 748 pr_debug("microcode: CPU%d resumed\n", cpu);
749 /* only CPU 0 will apply ucode here */ 749 /* only CPU 0 will apply ucode here */
750 apply_microcode(0); 750 apply_microcode(0);
751 return 0; 751 return 0;
@@ -783,7 +783,7 @@ mc_cpu_callback(struct notifier_block *nb, unsigned long action, void *hcpu)
783 } 783 }
784 case CPU_DOWN_FAILED_FROZEN: 784 case CPU_DOWN_FAILED_FROZEN:
785 if (sysfs_create_group(&sys_dev->kobj, &mc_attr_group)) 785 if (sysfs_create_group(&sys_dev->kobj, &mc_attr_group))
786 printk(KERN_ERR "Microcode: Failed to create the sysfs " 786 printk(KERN_ERR "microcode: Failed to create the sysfs "
787 "group for CPU%d\n", cpu); 787 "group for CPU%d\n", cpu);
788 break; 788 break;
789 case CPU_DOWN_PREPARE: 789 case CPU_DOWN_PREPARE:
diff --git a/arch/x86/kernel/mpparse_32.c b/arch/x86/kernel/mpparse.c
index f349e68e45a0..70744e344fa1 100644
--- a/arch/x86/kernel/mpparse_32.c
+++ b/arch/x86/kernel/mpparse.c
@@ -4,82 +4,56 @@
4 * 4 *
5 * (c) 1995 Alan Cox, Building #3 <alan@redhat.com> 5 * (c) 1995 Alan Cox, Building #3 <alan@redhat.com>
6 * (c) 1998, 1999, 2000 Ingo Molnar <mingo@redhat.com> 6 * (c) 1998, 1999, 2000 Ingo Molnar <mingo@redhat.com>
7 * 7 * (c) 2008 Alexey Starikovskiy <astarikovskiy@suse.de>
8 * Fixes
9 * Erich Boleyn : MP v1.4 and additional changes.
10 * Alan Cox : Added EBDA scanning
11 * Ingo Molnar : various cleanups and rewrites
12 * Maciej W. Rozycki: Bits for default MP configurations
13 * Paul Diefenbaugh: Added full ACPI support
14 */ 8 */
15 9
16#include <linux/mm.h> 10#include <linux/mm.h>
17#include <linux/init.h> 11#include <linux/init.h>
18#include <linux/acpi.h>
19#include <linux/delay.h> 12#include <linux/delay.h>
20#include <linux/bootmem.h> 13#include <linux/bootmem.h>
21#include <linux/kernel_stat.h> 14#include <linux/kernel_stat.h>
22#include <linux/mc146818rtc.h> 15#include <linux/mc146818rtc.h>
23#include <linux/bitops.h> 16#include <linux/bitops.h>
17#include <linux/acpi.h>
18#include <linux/module.h>
24 19
25#include <asm/smp.h> 20#include <asm/smp.h>
26#include <asm/acpi.h>
27#include <asm/mtrr.h> 21#include <asm/mtrr.h>
28#include <asm/mpspec.h> 22#include <asm/mpspec.h>
23#include <asm/pgalloc.h>
29#include <asm/io_apic.h> 24#include <asm/io_apic.h>
25#include <asm/proto.h>
26#include <asm/acpi.h>
27#include <asm/bios_ebda.h>
30 28
31#include <mach_apic.h> 29#include <mach_apic.h>
30#ifdef CONFIG_X86_32
32#include <mach_apicdef.h> 31#include <mach_apicdef.h>
33#include <mach_mpparse.h> 32#include <mach_mpparse.h>
34#include <bios_ebda.h> 33#endif
35 34
36/* Have we found an MP table */ 35/* Have we found an MP table */
37int smp_found_config; 36int smp_found_config;
38unsigned int __cpuinitdata maxcpus = NR_CPUS;
39 37
40/* 38/*
41 * Various Linux-internal data structures created from the 39 * Various Linux-internal data structures created from the
42 * MP-table. 40 * MP-table.
43 */ 41 */
44int apic_version [MAX_APICS]; 42#if defined (CONFIG_MCA) || defined (CONFIG_EISA)
45int mp_bus_id_to_type [MAX_MP_BUSSES]; 43int mp_bus_id_to_type[MAX_MP_BUSSES];
46int mp_bus_id_to_node [MAX_MP_BUSSES]; 44#endif
47int mp_bus_id_to_local [MAX_MP_BUSSES];
48int quad_local_to_mp_bus_id [NR_CPUS/4][4];
49int mp_bus_id_to_pci_bus [MAX_MP_BUSSES] = { [0 ... MAX_MP_BUSSES-1] = -1 };
50static int mp_current_pci_id;
51
52/* I/O APIC entries */
53struct mpc_config_ioapic mp_ioapics[MAX_IO_APICS];
54
55/* # of MP IRQ source entries */
56struct mpc_config_intsrc mp_irqs[MAX_IRQ_SOURCES];
57 45
58/* MP IRQ source entries */ 46DECLARE_BITMAP(mp_bus_not_pci, MAX_MP_BUSSES);
59int mp_irq_entries; 47int mp_bus_id_to_pci_bus[MAX_MP_BUSSES] = {[0 ... MAX_MP_BUSSES - 1] = -1 };
60 48
61int nr_ioapics; 49static int mp_current_pci_id;
62 50
63int pic_mode; 51int pic_mode;
64unsigned long mp_lapic_addr;
65
66unsigned int def_to_bigsmp = 0;
67
68/* Processor that is doing the boot up */
69unsigned int boot_cpu_physical_apicid = -1U;
70/* Internal processor count */
71unsigned int num_processors;
72
73/* Bitmask of physically existing CPUs */
74physid_mask_t phys_cpu_present_map;
75
76u8 bios_cpu_apicid[NR_CPUS] = { [0 ... NR_CPUS-1] = BAD_APICID };
77 52
78/* 53/*
79 * Intel MP BIOS table parsing routines: 54 * Intel MP BIOS table parsing routines:
80 */ 55 */
81 56
82
83/* 57/*
84 * Checksum an MP configuration block. 58 * Checksum an MP configuration block.
85 */ 59 */
@@ -94,216 +68,153 @@ static int __init mpf_checksum(unsigned char *mp, int len)
94 return sum & 0xFF; 68 return sum & 0xFF;
95} 69}
96 70
71#ifdef CONFIG_X86_NUMAQ
97/* 72/*
98 * Have to match translation table entries to main table entries by counter 73 * Have to match translation table entries to main table entries by counter
99 * hence the mpc_record variable .... can't see a less disgusting way of 74 * hence the mpc_record variable .... can't see a less disgusting way of
100 * doing this .... 75 * doing this ....
101 */ 76 */
102 77
103static int mpc_record; 78static int mpc_record;
104static struct mpc_config_translation *translation_table[MAX_MPC_ENTRY] __cpuinitdata; 79static struct mpc_config_translation *translation_table[MAX_MPC_ENTRY]
80 __cpuinitdata;
81#endif
105 82
106static void __cpuinit MP_processor_info (struct mpc_config_processor *m) 83static void __cpuinit MP_processor_info(struct mpc_config_processor *m)
107{ 84{
108 int ver, apicid; 85 int apicid;
109 physid_mask_t phys_cpu; 86 char *bootup_cpu = "";
110
111 if (!(m->mpc_cpuflag & CPU_ENABLED))
112 return;
113 87
88 if (!(m->mpc_cpuflag & CPU_ENABLED)) {
89 disabled_cpus++;
90 return;
91 }
92#ifdef CONFIG_X86_NUMAQ
114 apicid = mpc_apic_id(m, translation_table[mpc_record]); 93 apicid = mpc_apic_id(m, translation_table[mpc_record]);
115 94#else
116 if (m->mpc_featureflag&(1<<0)) 95 apicid = m->mpc_apicid;
117 Dprintk(" Floating point unit present.\n"); 96#endif
118 if (m->mpc_featureflag&(1<<7))
119 Dprintk(" Machine Exception supported.\n");
120 if (m->mpc_featureflag&(1<<8))
121 Dprintk(" 64 bit compare & exchange supported.\n");
122 if (m->mpc_featureflag&(1<<9))
123 Dprintk(" Internal APIC present.\n");
124 if (m->mpc_featureflag&(1<<11))
125 Dprintk(" SEP present.\n");
126 if (m->mpc_featureflag&(1<<12))
127 Dprintk(" MTRR present.\n");
128 if (m->mpc_featureflag&(1<<13))
129 Dprintk(" PGE present.\n");
130 if (m->mpc_featureflag&(1<<14))
131 Dprintk(" MCA present.\n");
132 if (m->mpc_featureflag&(1<<15))
133 Dprintk(" CMOV present.\n");
134 if (m->mpc_featureflag&(1<<16))
135 Dprintk(" PAT present.\n");
136 if (m->mpc_featureflag&(1<<17))
137 Dprintk(" PSE present.\n");
138 if (m->mpc_featureflag&(1<<18))
139 Dprintk(" PSN present.\n");
140 if (m->mpc_featureflag&(1<<19))
141 Dprintk(" Cache Line Flush Instruction present.\n");
142 /* 20 Reserved */
143 if (m->mpc_featureflag&(1<<21))
144 Dprintk(" Debug Trace and EMON Store present.\n");
145 if (m->mpc_featureflag&(1<<22))
146 Dprintk(" ACPI Thermal Throttle Registers present.\n");
147 if (m->mpc_featureflag&(1<<23))
148 Dprintk(" MMX present.\n");
149 if (m->mpc_featureflag&(1<<24))
150 Dprintk(" FXSR present.\n");
151 if (m->mpc_featureflag&(1<<25))
152 Dprintk(" XMM present.\n");
153 if (m->mpc_featureflag&(1<<26))
154 Dprintk(" Willamette New Instructions present.\n");
155 if (m->mpc_featureflag&(1<<27))
156 Dprintk(" Self Snoop present.\n");
157 if (m->mpc_featureflag&(1<<28))
158 Dprintk(" HT present.\n");
159 if (m->mpc_featureflag&(1<<29))
160 Dprintk(" Thermal Monitor present.\n");
161 /* 30, 31 Reserved */
162
163
164 if (m->mpc_cpuflag & CPU_BOOTPROCESSOR) { 97 if (m->mpc_cpuflag & CPU_BOOTPROCESSOR) {
165 Dprintk(" Bootup CPU\n"); 98 bootup_cpu = " (Bootup-CPU)";
166 boot_cpu_physical_apicid = m->mpc_apicid; 99 boot_cpu_physical_apicid = m->mpc_apicid;
167 } 100 }
168 101
169 ver = m->mpc_apicver; 102 printk(KERN_INFO "Processor #%d%s\n", m->mpc_apicid, bootup_cpu);
170 103 generic_processor_info(apicid, m->mpc_apicver);
171 /*
172 * Validate version
173 */
174 if (ver == 0x0) {
175 printk(KERN_WARNING "BIOS bug, APIC version is 0 for CPU#%d! "
176 "fixing up to 0x10. (tell your hw vendor)\n",
177 m->mpc_apicid);
178 ver = 0x10;
179 }
180 apic_version[m->mpc_apicid] = ver;
181
182 phys_cpu = apicid_to_cpu_present(apicid);
183 physids_or(phys_cpu_present_map, phys_cpu_present_map, phys_cpu);
184
185 if (num_processors >= NR_CPUS) {
186 printk(KERN_WARNING "WARNING: NR_CPUS limit of %i reached."
187 " Processor ignored.\n", NR_CPUS);
188 return;
189 }
190
191 if (num_processors >= maxcpus) {
192 printk(KERN_WARNING "WARNING: maxcpus limit of %i reached."
193 " Processor ignored.\n", maxcpus);
194 return;
195 }
196
197 cpu_set(num_processors, cpu_possible_map);
198 num_processors++;
199
200 /*
201 * Would be preferable to switch to bigsmp when CONFIG_HOTPLUG_CPU=y
202 * but we need to work other dependencies like SMP_SUSPEND etc
203 * before this can be done without some confusion.
204 * if (CPU_HOTPLUG_ENABLED || num_processors > 8)
205 * - Ashok Raj <ashok.raj@intel.com>
206 */
207 if (num_processors > 8) {
208 switch (boot_cpu_data.x86_vendor) {
209 case X86_VENDOR_INTEL:
210 if (!APIC_XAPIC(ver)) {
211 def_to_bigsmp = 0;
212 break;
213 }
214 /* If P4 and above fall through */
215 case X86_VENDOR_AMD:
216 def_to_bigsmp = 1;
217 }
218 }
219 bios_cpu_apicid[num_processors - 1] = m->mpc_apicid;
220} 104}
221 105
222static void __init MP_bus_info (struct mpc_config_bus *m) 106static void __init MP_bus_info(struct mpc_config_bus *m)
223{ 107{
224 char str[7]; 108 char str[7];
225 109
226 memcpy(str, m->mpc_bustype, 6); 110 memcpy(str, m->mpc_bustype, 6);
227 str[6] = 0; 111 str[6] = 0;
228 112
113#ifdef CONFIG_X86_NUMAQ
229 mpc_oem_bus_info(m, str, translation_table[mpc_record]); 114 mpc_oem_bus_info(m, str, translation_table[mpc_record]);
115#else
116 Dprintk("Bus #%d is %s\n", m->mpc_busid, str);
117#endif
230 118
231#if MAX_MP_BUSSES < 256 119#if MAX_MP_BUSSES < 256
232 if (m->mpc_busid >= MAX_MP_BUSSES) { 120 if (m->mpc_busid >= MAX_MP_BUSSES) {
233 printk(KERN_WARNING "MP table busid value (%d) for bustype %s " 121 printk(KERN_WARNING "MP table busid value (%d) for bustype %s "
234 " is too large, max. supported is %d\n", 122 " is too large, max. supported is %d\n",
235 m->mpc_busid, str, MAX_MP_BUSSES - 1); 123 m->mpc_busid, str, MAX_MP_BUSSES - 1);
236 return; 124 return;
237 } 125 }
238#endif 126#endif
239 127
240 if (strncmp(str, BUSTYPE_ISA, sizeof(BUSTYPE_ISA)-1) == 0) { 128 if (strncmp(str, BUSTYPE_ISA, sizeof(BUSTYPE_ISA) - 1) == 0) {
129 set_bit(m->mpc_busid, mp_bus_not_pci);
130#if defined(CONFIG_EISA) || defined (CONFIG_MCA)
241 mp_bus_id_to_type[m->mpc_busid] = MP_BUS_ISA; 131 mp_bus_id_to_type[m->mpc_busid] = MP_BUS_ISA;
242 } else if (strncmp(str, BUSTYPE_EISA, sizeof(BUSTYPE_EISA)-1) == 0) { 132#endif
243 mp_bus_id_to_type[m->mpc_busid] = MP_BUS_EISA; 133 } else if (strncmp(str, BUSTYPE_PCI, sizeof(BUSTYPE_PCI) - 1) == 0) {
244 } else if (strncmp(str, BUSTYPE_PCI, sizeof(BUSTYPE_PCI)-1) == 0) { 134#ifdef CONFIG_X86_NUMAQ
245 mpc_oem_pci_bus(m, translation_table[mpc_record]); 135 mpc_oem_pci_bus(m, translation_table[mpc_record]);
246 mp_bus_id_to_type[m->mpc_busid] = MP_BUS_PCI; 136#endif
137 clear_bit(m->mpc_busid, mp_bus_not_pci);
247 mp_bus_id_to_pci_bus[m->mpc_busid] = mp_current_pci_id; 138 mp_bus_id_to_pci_bus[m->mpc_busid] = mp_current_pci_id;
248 mp_current_pci_id++; 139 mp_current_pci_id++;
249 } else if (strncmp(str, BUSTYPE_MCA, sizeof(BUSTYPE_MCA)-1) == 0) { 140#if defined(CONFIG_EISA) || defined (CONFIG_MCA)
141 mp_bus_id_to_type[m->mpc_busid] = MP_BUS_PCI;
142 } else if (strncmp(str, BUSTYPE_EISA, sizeof(BUSTYPE_EISA) - 1) == 0) {
143 mp_bus_id_to_type[m->mpc_busid] = MP_BUS_EISA;
144 } else if (strncmp(str, BUSTYPE_MCA, sizeof(BUSTYPE_MCA) - 1) == 0) {
250 mp_bus_id_to_type[m->mpc_busid] = MP_BUS_MCA; 145 mp_bus_id_to_type[m->mpc_busid] = MP_BUS_MCA;
251 } else { 146#endif
147 } else
252 printk(KERN_WARNING "Unknown bustype %s - ignoring\n", str); 148 printk(KERN_WARNING "Unknown bustype %s - ignoring\n", str);
149}
150
151#ifdef CONFIG_X86_IO_APIC
152
153static int bad_ioapic(unsigned long address)
154{
155 if (nr_ioapics >= MAX_IO_APICS) {
156 printk(KERN_ERR "ERROR: Max # of I/O APICs (%d) exceeded "
157 "(found %d)\n", MAX_IO_APICS, nr_ioapics);
158 panic("Recompile kernel with bigger MAX_IO_APICS!\n");
159 }
160 if (!address) {
161 printk(KERN_ERR "WARNING: Bogus (zero) I/O APIC address"
162 " found in table, skipping!\n");
163 return 1;
253 } 164 }
165 return 0;
254} 166}
255 167
256static void __init MP_ioapic_info (struct mpc_config_ioapic *m) 168static void __init MP_ioapic_info(struct mpc_config_ioapic *m)
257{ 169{
258 if (!(m->mpc_flags & MPC_APIC_USABLE)) 170 if (!(m->mpc_flags & MPC_APIC_USABLE))
259 return; 171 return;
260 172
261 printk(KERN_INFO "I/O APIC #%d Version %d at 0x%X.\n", 173 printk(KERN_INFO "I/O APIC #%d Version %d at 0x%X.\n",
262 m->mpc_apicid, m->mpc_apicver, m->mpc_apicaddr); 174 m->mpc_apicid, m->mpc_apicver, m->mpc_apicaddr);
263 if (nr_ioapics >= MAX_IO_APICS) { 175
264 printk(KERN_CRIT "Max # of I/O APICs (%d) exceeded (found %d).\n", 176 if (bad_ioapic(m->mpc_apicaddr))
265 MAX_IO_APICS, nr_ioapics);
266 panic("Recompile kernel with bigger MAX_IO_APICS!.\n");
267 }
268 if (!m->mpc_apicaddr) {
269 printk(KERN_ERR "WARNING: bogus zero I/O APIC address"
270 " found in MP table, skipping!\n");
271 return; 177 return;
272 } 178
273 mp_ioapics[nr_ioapics] = *m; 179 mp_ioapics[nr_ioapics] = *m;
274 nr_ioapics++; 180 nr_ioapics++;
275} 181}
276 182
277static void __init MP_intsrc_info (struct mpc_config_intsrc *m) 183static void __init MP_intsrc_info(struct mpc_config_intsrc *m)
278{ 184{
279 mp_irqs [mp_irq_entries] = *m; 185 mp_irqs[mp_irq_entries] = *m;
280 Dprintk("Int: type %d, pol %d, trig %d, bus %d," 186 Dprintk("Int: type %d, pol %d, trig %d, bus %d,"
281 " IRQ %02x, APIC ID %x, APIC INT %02x\n", 187 " IRQ %02x, APIC ID %x, APIC INT %02x\n",
282 m->mpc_irqtype, m->mpc_irqflag & 3, 188 m->mpc_irqtype, m->mpc_irqflag & 3,
283 (m->mpc_irqflag >> 2) & 3, m->mpc_srcbus, 189 (m->mpc_irqflag >> 2) & 3, m->mpc_srcbus,
284 m->mpc_srcbusirq, m->mpc_dstapic, m->mpc_dstirq); 190 m->mpc_srcbusirq, m->mpc_dstapic, m->mpc_dstirq);
285 if (++mp_irq_entries == MAX_IRQ_SOURCES) 191 if (++mp_irq_entries == MAX_IRQ_SOURCES)
286 panic("Max # of irq sources exceeded!!\n"); 192 panic("Max # of irq sources exceeded!!\n");
287} 193}
288 194
289static void __init MP_lintsrc_info (struct mpc_config_lintsrc *m) 195#endif
196
197static void __init MP_lintsrc_info(struct mpc_config_lintsrc *m)
290{ 198{
291 Dprintk("Lint: type %d, pol %d, trig %d, bus %d," 199 Dprintk("Lint: type %d, pol %d, trig %d, bus %d,"
292 " IRQ %02x, APIC ID %x, APIC LINT %02x\n", 200 " IRQ %02x, APIC ID %x, APIC LINT %02x\n",
293 m->mpc_irqtype, m->mpc_irqflag & 3, 201 m->mpc_irqtype, m->mpc_irqflag & 3,
294 (m->mpc_irqflag >> 2) &3, m->mpc_srcbusid, 202 (m->mpc_irqflag >> 2) & 3, m->mpc_srcbusid,
295 m->mpc_srcbusirq, m->mpc_destapic, m->mpc_destapiclint); 203 m->mpc_srcbusirq, m->mpc_destapic, m->mpc_destapiclint);
296} 204}
297 205
298#ifdef CONFIG_X86_NUMAQ 206#ifdef CONFIG_X86_NUMAQ
299static void __init MP_translation_info (struct mpc_config_translation *m) 207static void __init MP_translation_info(struct mpc_config_translation *m)
300{ 208{
301 printk(KERN_INFO "Translation: record %d, type %d, quad %d, global %d, local %d\n", mpc_record, m->trans_type, m->trans_quad, m->trans_global, m->trans_local); 209 printk(KERN_INFO
210 "Translation: record %d, type %d, quad %d, global %d, local %d\n",
211 mpc_record, m->trans_type, m->trans_quad, m->trans_global,
212 m->trans_local);
302 213
303 if (mpc_record >= MAX_MPC_ENTRY) 214 if (mpc_record >= MAX_MPC_ENTRY)
304 printk(KERN_ERR "MAX_MPC_ENTRY exceeded!\n"); 215 printk(KERN_ERR "MAX_MPC_ENTRY exceeded!\n");
305 else 216 else
306 translation_table[mpc_record] = m; /* stash this for later */ 217 translation_table[mpc_record] = m; /* stash this for later */
307 if (m->trans_quad < MAX_NUMNODES && !node_online(m->trans_quad)) 218 if (m->trans_quad < MAX_NUMNODES && !node_online(m->trans_quad))
308 node_set_online(m->trans_quad); 219 node_set_online(m->trans_quad);
309} 220}
@@ -312,118 +223,124 @@ static void __init MP_translation_info (struct mpc_config_translation *m)
312 * Read/parse the MPC oem tables 223 * Read/parse the MPC oem tables
313 */ 224 */
314 225
315static void __init smp_read_mpc_oem(struct mp_config_oemtable *oemtable, \ 226static void __init smp_read_mpc_oem(struct mp_config_oemtable *oemtable,
316 unsigned short oemsize) 227 unsigned short oemsize)
317{ 228{
318 int count = sizeof (*oemtable); /* the header size */ 229 int count = sizeof(*oemtable); /* the header size */
319 unsigned char *oemptr = ((unsigned char *)oemtable)+count; 230 unsigned char *oemptr = ((unsigned char *)oemtable) + count;
320 231
321 mpc_record = 0; 232 mpc_record = 0;
322 printk(KERN_INFO "Found an OEM MPC table at %8p - parsing it ... \n", oemtable); 233 printk(KERN_INFO "Found an OEM MPC table at %8p - parsing it ... \n",
323 if (memcmp(oemtable->oem_signature,MPC_OEM_SIGNATURE,4)) 234 oemtable);
324 { 235 if (memcmp(oemtable->oem_signature, MPC_OEM_SIGNATURE, 4)) {
325 printk(KERN_WARNING "SMP mpc oemtable: bad signature [%c%c%c%c]!\n", 236 printk(KERN_WARNING
326 oemtable->oem_signature[0], 237 "SMP mpc oemtable: bad signature [%c%c%c%c]!\n",
327 oemtable->oem_signature[1], 238 oemtable->oem_signature[0], oemtable->oem_signature[1],
328 oemtable->oem_signature[2], 239 oemtable->oem_signature[2], oemtable->oem_signature[3]);
329 oemtable->oem_signature[3]);
330 return; 240 return;
331 } 241 }
332 if (mpf_checksum((unsigned char *)oemtable,oemtable->oem_length)) 242 if (mpf_checksum((unsigned char *)oemtable, oemtable->oem_length)) {
333 {
334 printk(KERN_WARNING "SMP oem mptable: checksum error!\n"); 243 printk(KERN_WARNING "SMP oem mptable: checksum error!\n");
335 return; 244 return;
336 } 245 }
337 while (count < oemtable->oem_length) { 246 while (count < oemtable->oem_length) {
338 switch (*oemptr) { 247 switch (*oemptr) {
339 case MP_TRANSLATION: 248 case MP_TRANSLATION:
340 { 249 {
341 struct mpc_config_translation *m= 250 struct mpc_config_translation *m =
342 (struct mpc_config_translation *)oemptr; 251 (struct mpc_config_translation *)oemptr;
343 MP_translation_info(m); 252 MP_translation_info(m);
344 oemptr += sizeof(*m); 253 oemptr += sizeof(*m);
345 count += sizeof(*m); 254 count += sizeof(*m);
346 ++mpc_record; 255 ++mpc_record;
347 break; 256 break;
348 } 257 }
349 default: 258 default:
350 { 259 {
351 printk(KERN_WARNING "Unrecognised OEM table entry type! - %d\n", (int) *oemptr); 260 printk(KERN_WARNING
261 "Unrecognised OEM table entry type! - %d\n",
262 (int)*oemptr);
352 return; 263 return;
353 } 264 }
354 } 265 }
355 } 266 }
356} 267}
357 268
358static inline void mps_oem_check(struct mp_config_table *mpc, char *oem, 269static inline void mps_oem_check(struct mp_config_table *mpc, char *oem,
359 char *productid) 270 char *productid)
360{ 271{
361 if (strncmp(oem, "IBM NUMA", 8)) 272 if (strncmp(oem, "IBM NUMA", 8))
362 printk("Warning! May not be a NUMA-Q system!\n"); 273 printk("Warning! May not be a NUMA-Q system!\n");
363 if (mpc->mpc_oemptr) 274 if (mpc->mpc_oemptr)
364 smp_read_mpc_oem((struct mp_config_oemtable *) mpc->mpc_oemptr, 275 smp_read_mpc_oem((struct mp_config_oemtable *)mpc->mpc_oemptr,
365 mpc->mpc_oemsize); 276 mpc->mpc_oemsize);
366} 277}
367#endif /* CONFIG_X86_NUMAQ */ 278#endif /* CONFIG_X86_NUMAQ */
368 279
369/* 280/*
370 * Read/parse the MPC 281 * Read/parse the MPC
371 */ 282 */
372 283
373static int __init smp_read_mpc(struct mp_config_table *mpc) 284static int __init smp_read_mpc(struct mp_config_table *mpc, unsigned early)
374{ 285{
375 char str[16]; 286 char str[16];
376 char oem[10]; 287 char oem[10];
377 int count=sizeof(*mpc); 288 int count = sizeof(*mpc);
378 unsigned char *mpt=((unsigned char *)mpc)+count; 289 unsigned char *mpt = ((unsigned char *)mpc) + count;
379 290
380 if (memcmp(mpc->mpc_signature,MPC_SIGNATURE,4)) { 291 if (memcmp(mpc->mpc_signature, MPC_SIGNATURE, 4)) {
381 printk(KERN_ERR "SMP mptable: bad signature [0x%x]!\n", 292 printk(KERN_ERR "MPTABLE: bad signature [%c%c%c%c]!\n",
382 *(u32 *)mpc->mpc_signature); 293 mpc->mpc_signature[0], mpc->mpc_signature[1],
294 mpc->mpc_signature[2], mpc->mpc_signature[3]);
383 return 0; 295 return 0;
384 } 296 }
385 if (mpf_checksum((unsigned char *)mpc,mpc->mpc_length)) { 297 if (mpf_checksum((unsigned char *)mpc, mpc->mpc_length)) {
386 printk(KERN_ERR "SMP mptable: checksum error!\n"); 298 printk(KERN_ERR "MPTABLE: checksum error!\n");
387 return 0; 299 return 0;
388 } 300 }
389 if (mpc->mpc_spec!=0x01 && mpc->mpc_spec!=0x04) { 301 if (mpc->mpc_spec != 0x01 && mpc->mpc_spec != 0x04) {
390 printk(KERN_ERR "SMP mptable: bad table version (%d)!!\n", 302 printk(KERN_ERR "MPTABLE: bad table version (%d)!!\n",
391 mpc->mpc_spec); 303 mpc->mpc_spec);
392 return 0; 304 return 0;
393 } 305 }
394 if (!mpc->mpc_lapic) { 306 if (!mpc->mpc_lapic) {
395 printk(KERN_ERR "SMP mptable: null local APIC address!\n"); 307 printk(KERN_ERR "MPTABLE: null local APIC address!\n");
396 return 0; 308 return 0;
397 } 309 }
398 memcpy(oem,mpc->mpc_oem,8); 310 memcpy(oem, mpc->mpc_oem, 8);
399 oem[8]=0; 311 oem[8] = 0;
400 printk(KERN_INFO "OEM ID: %s ",oem); 312 printk(KERN_INFO "MPTABLE: OEM ID: %s ", oem);
401 313
402 memcpy(str,mpc->mpc_productid,12); 314 memcpy(str, mpc->mpc_productid, 12);
403 str[12]=0; 315 str[12] = 0;
404 printk("Product ID: %s ",str); 316 printk("Product ID: %s ", str);
405 317
318#ifdef CONFIG_X86_32
406 mps_oem_check(mpc, oem, str); 319 mps_oem_check(mpc, oem, str);
320#endif
321 printk(KERN_INFO "MPTABLE: Product ID: %s ", str);
407 322
408 printk("APIC at: 0x%X\n", mpc->mpc_lapic); 323 printk(KERN_INFO "MPTABLE: APIC at: 0x%X\n", mpc->mpc_lapic);
409 324
410 /* 325 /* save the local APIC address, it might be non-default */
411 * Save the local APIC address (it might be non-default) -- but only
412 * if we're not using ACPI.
413 */
414 if (!acpi_lapic) 326 if (!acpi_lapic)
415 mp_lapic_addr = mpc->mpc_lapic; 327 mp_lapic_addr = mpc->mpc_lapic;
416 328
329 if (early)
330 return 1;
331
417 /* 332 /*
418 * Now process the configuration blocks. 333 * Now process the configuration blocks.
419 */ 334 */
335#ifdef CONFIG_X86_NUMAQ
420 mpc_record = 0; 336 mpc_record = 0;
337#endif
421 while (count < mpc->mpc_length) { 338 while (count < mpc->mpc_length) {
422 switch(*mpt) { 339 switch (*mpt) {
423 case MP_PROCESSOR: 340 case MP_PROCESSOR:
424 { 341 {
425 struct mpc_config_processor *m= 342 struct mpc_config_processor *m =
426 (struct mpc_config_processor *)mpt; 343 (struct mpc_config_processor *)mpt;
427 /* ACPI may have already provided this data */ 344 /* ACPI may have already provided this data */
428 if (!acpi_lapic) 345 if (!acpi_lapic)
429 MP_processor_info(m); 346 MP_processor_info(m);
@@ -431,57 +348,68 @@ static int __init smp_read_mpc(struct mp_config_table *mpc)
431 count += sizeof(*m); 348 count += sizeof(*m);
432 break; 349 break;
433 } 350 }
434 case MP_BUS: 351 case MP_BUS:
435 { 352 {
436 struct mpc_config_bus *m= 353 struct mpc_config_bus *m =
437 (struct mpc_config_bus *)mpt; 354 (struct mpc_config_bus *)mpt;
438 MP_bus_info(m); 355 MP_bus_info(m);
439 mpt += sizeof(*m); 356 mpt += sizeof(*m);
440 count += sizeof(*m); 357 count += sizeof(*m);
441 break; 358 break;
442 } 359 }
443 case MP_IOAPIC: 360 case MP_IOAPIC:
444 { 361 {
445 struct mpc_config_ioapic *m= 362#ifdef CONFIG_X86_IO_APIC
446 (struct mpc_config_ioapic *)mpt; 363 struct mpc_config_ioapic *m =
364 (struct mpc_config_ioapic *)mpt;
447 MP_ioapic_info(m); 365 MP_ioapic_info(m);
448 mpt+=sizeof(*m); 366#endif
449 count+=sizeof(*m); 367 mpt += sizeof(struct mpc_config_ioapic);
368 count += sizeof(struct mpc_config_ioapic);
450 break; 369 break;
451 } 370 }
452 case MP_INTSRC: 371 case MP_INTSRC:
453 { 372 {
454 struct mpc_config_intsrc *m= 373#ifdef CONFIG_X86_IO_APIC
455 (struct mpc_config_intsrc *)mpt; 374 struct mpc_config_intsrc *m =
375 (struct mpc_config_intsrc *)mpt;
456 376
457 MP_intsrc_info(m); 377 MP_intsrc_info(m);
458 mpt+=sizeof(*m); 378#endif
459 count+=sizeof(*m); 379 mpt += sizeof(struct mpc_config_intsrc);
380 count += sizeof(struct mpc_config_intsrc);
460 break; 381 break;
461 } 382 }
462 case MP_LINTSRC: 383 case MP_LINTSRC:
463 { 384 {
464 struct mpc_config_lintsrc *m= 385 struct mpc_config_lintsrc *m =
465 (struct mpc_config_lintsrc *)mpt; 386 (struct mpc_config_lintsrc *)mpt;
466 MP_lintsrc_info(m); 387 MP_lintsrc_info(m);
467 mpt+=sizeof(*m); 388 mpt += sizeof(*m);
468 count+=sizeof(*m); 389 count += sizeof(*m);
469 break;
470 }
471 default:
472 {
473 count = mpc->mpc_length;
474 break; 390 break;
475 } 391 }
392 default:
393 /* wrong mptable */
394 printk(KERN_ERR "Your mptable is wrong, contact your HW vendor!\n");
395 printk(KERN_ERR "type %x\n", *mpt);
396 print_hex_dump(KERN_ERR, " ", DUMP_PREFIX_ADDRESS, 16,
397 1, mpc, mpc->mpc_length, 1);
398 count = mpc->mpc_length;
399 break;
476 } 400 }
401#ifdef CONFIG_X86_NUMAQ
477 ++mpc_record; 402 ++mpc_record;
403#endif
478 } 404 }
479 setup_apic_routing(); 405 setup_apic_routing();
480 if (!num_processors) 406 if (!num_processors)
481 printk(KERN_ERR "SMP mptable: no processors registered!\n"); 407 printk(KERN_ERR "MPTABLE: no processors registered!\n");
482 return num_processors; 408 return num_processors;
483} 409}
484 410
411#ifdef CONFIG_X86_IO_APIC
412
485static int __init ELCR_trigger(unsigned int irq) 413static int __init ELCR_trigger(unsigned int irq)
486{ 414{
487 unsigned int port; 415 unsigned int port;
@@ -497,7 +425,7 @@ static void __init construct_default_ioirq_mptable(int mpc_default_type)
497 int ELCR_fallback = 0; 425 int ELCR_fallback = 0;
498 426
499 intsrc.mpc_type = MP_INTSRC; 427 intsrc.mpc_type = MP_INTSRC;
500 intsrc.mpc_irqflag = 0; /* conforming */ 428 intsrc.mpc_irqflag = 0; /* conforming */
501 intsrc.mpc_srcbus = 0; 429 intsrc.mpc_srcbus = 0;
502 intsrc.mpc_dstapic = mp_ioapics[0].mpc_apicid; 430 intsrc.mpc_dstapic = mp_ioapics[0].mpc_apicid;
503 431
@@ -512,12 +440,16 @@ static void __init construct_default_ioirq_mptable(int mpc_default_type)
512 * If it does, we assume it's valid. 440 * If it does, we assume it's valid.
513 */ 441 */
514 if (mpc_default_type == 5) { 442 if (mpc_default_type == 5) {
515 printk(KERN_INFO "ISA/PCI bus type with no IRQ information... falling back to ELCR\n"); 443 printk(KERN_INFO "ISA/PCI bus type with no IRQ information... "
444 "falling back to ELCR\n");
516 445
517 if (ELCR_trigger(0) || ELCR_trigger(1) || ELCR_trigger(2) || ELCR_trigger(13)) 446 if (ELCR_trigger(0) || ELCR_trigger(1) || ELCR_trigger(2) ||
518 printk(KERN_WARNING "ELCR contains invalid data... not using ELCR\n"); 447 ELCR_trigger(13))
448 printk(KERN_ERR "ELCR contains invalid data... "
449 "not using ELCR\n");
519 else { 450 else {
520 printk(KERN_INFO "Using ELCR to identify PCI interrupts\n"); 451 printk(KERN_INFO
452 "Using ELCR to identify PCI interrupts\n");
521 ELCR_fallback = 1; 453 ELCR_fallback = 1;
522 } 454 }
523 } 455 }
@@ -546,21 +478,25 @@ static void __init construct_default_ioirq_mptable(int mpc_default_type)
546 } 478 }
547 479
548 intsrc.mpc_srcbusirq = i; 480 intsrc.mpc_srcbusirq = i;
549 intsrc.mpc_dstirq = i ? i : 2; /* IRQ0 to INTIN2 */ 481 intsrc.mpc_dstirq = i ? i : 2; /* IRQ0 to INTIN2 */
550 MP_intsrc_info(&intsrc); 482 MP_intsrc_info(&intsrc);
551 } 483 }
552 484
553 intsrc.mpc_irqtype = mp_ExtINT; 485 intsrc.mpc_irqtype = mp_ExtINT;
554 intsrc.mpc_srcbusirq = 0; 486 intsrc.mpc_srcbusirq = 0;
555 intsrc.mpc_dstirq = 0; /* 8259A to INTIN0 */ 487 intsrc.mpc_dstirq = 0; /* 8259A to INTIN0 */
556 MP_intsrc_info(&intsrc); 488 MP_intsrc_info(&intsrc);
557} 489}
558 490
491#endif
492
559static inline void __init construct_default_ISA_mptable(int mpc_default_type) 493static inline void __init construct_default_ISA_mptable(int mpc_default_type)
560{ 494{
561 struct mpc_config_processor processor; 495 struct mpc_config_processor processor;
562 struct mpc_config_bus bus; 496 struct mpc_config_bus bus;
497#ifdef CONFIG_X86_IO_APIC
563 struct mpc_config_ioapic ioapic; 498 struct mpc_config_ioapic ioapic;
499#endif
564 struct mpc_config_lintsrc lintsrc; 500 struct mpc_config_lintsrc lintsrc;
565 int linttypes[2] = { mp_ExtINT, mp_NMI }; 501 int linttypes[2] = { mp_ExtINT, mp_NMI };
566 int i; 502 int i;
@@ -578,8 +514,7 @@ static inline void __init construct_default_ISA_mptable(int mpc_default_type)
578 processor.mpc_apicver = mpc_default_type > 4 ? 0x10 : 0x01; 514 processor.mpc_apicver = mpc_default_type > 4 ? 0x10 : 0x01;
579 processor.mpc_cpuflag = CPU_ENABLED; 515 processor.mpc_cpuflag = CPU_ENABLED;
580 processor.mpc_cpufeature = (boot_cpu_data.x86 << 8) | 516 processor.mpc_cpufeature = (boot_cpu_data.x86 << 8) |
581 (boot_cpu_data.x86_model << 4) | 517 (boot_cpu_data.x86_model << 4) | boot_cpu_data.x86_mask;
582 boot_cpu_data.x86_mask;
583 processor.mpc_featureflag = boot_cpu_data.x86_capability[0]; 518 processor.mpc_featureflag = boot_cpu_data.x86_capability[0];
584 processor.mpc_reserved[0] = 0; 519 processor.mpc_reserved[0] = 0;
585 processor.mpc_reserved[1] = 0; 520 processor.mpc_reserved[1] = 0;
@@ -591,23 +526,22 @@ static inline void __init construct_default_ISA_mptable(int mpc_default_type)
591 bus.mpc_type = MP_BUS; 526 bus.mpc_type = MP_BUS;
592 bus.mpc_busid = 0; 527 bus.mpc_busid = 0;
593 switch (mpc_default_type) { 528 switch (mpc_default_type) {
594 default: 529 default:
595 printk("???\n"); 530 printk(KERN_ERR "???\nUnknown standard configuration %d\n",
596 printk(KERN_ERR "Unknown standard configuration %d\n", 531 mpc_default_type);
597 mpc_default_type); 532 /* fall through */
598 /* fall through */ 533 case 1:
599 case 1: 534 case 5:
600 case 5: 535 memcpy(bus.mpc_bustype, "ISA ", 6);
601 memcpy(bus.mpc_bustype, "ISA ", 6); 536 break;
602 break; 537 case 2:
603 case 2: 538 case 6:
604 case 6: 539 case 3:
605 case 3: 540 memcpy(bus.mpc_bustype, "EISA ", 6);
606 memcpy(bus.mpc_bustype, "EISA ", 6); 541 break;
607 break; 542 case 4:
608 case 4: 543 case 7:
609 case 7: 544 memcpy(bus.mpc_bustype, "MCA ", 6);
610 memcpy(bus.mpc_bustype, "MCA ", 6);
611 } 545 }
612 MP_bus_info(&bus); 546 MP_bus_info(&bus);
613 if (mpc_default_type > 4) { 547 if (mpc_default_type > 4) {
@@ -616,6 +550,7 @@ static inline void __init construct_default_ISA_mptable(int mpc_default_type)
616 MP_bus_info(&bus); 550 MP_bus_info(&bus);
617 } 551 }
618 552
553#ifdef CONFIG_X86_IO_APIC
619 ioapic.mpc_type = MP_IOAPIC; 554 ioapic.mpc_type = MP_IOAPIC;
620 ioapic.mpc_apicid = 2; 555 ioapic.mpc_apicid = 2;
621 ioapic.mpc_apicver = mpc_default_type > 4 ? 0x10 : 0x01; 556 ioapic.mpc_apicver = mpc_default_type > 4 ? 0x10 : 0x01;
@@ -627,9 +562,9 @@ static inline void __init construct_default_ISA_mptable(int mpc_default_type)
627 * We set up most of the low 16 IO-APIC pins according to MPS rules. 562 * We set up most of the low 16 IO-APIC pins according to MPS rules.
628 */ 563 */
629 construct_default_ioirq_mptable(mpc_default_type); 564 construct_default_ioirq_mptable(mpc_default_type);
630 565#endif
631 lintsrc.mpc_type = MP_LINTSRC; 566 lintsrc.mpc_type = MP_LINTSRC;
632 lintsrc.mpc_irqflag = 0; /* conforming */ 567 lintsrc.mpc_irqflag = 0; /* conforming */
633 lintsrc.mpc_srcbusid = 0; 568 lintsrc.mpc_srcbusid = 0;
634 lintsrc.mpc_srcbusirq = 0; 569 lintsrc.mpc_srcbusirq = 0;
635 lintsrc.mpc_destapic = MP_APIC_ALL; 570 lintsrc.mpc_destapic = MP_APIC_ALL;
@@ -645,36 +580,49 @@ static struct intel_mp_floating *mpf_found;
645/* 580/*
646 * Scan the memory blocks for an SMP configuration block. 581 * Scan the memory blocks for an SMP configuration block.
647 */ 582 */
648void __init get_smp_config (void) 583static void __init __get_smp_config(unsigned early)
649{ 584{
650 struct intel_mp_floating *mpf = mpf_found; 585 struct intel_mp_floating *mpf = mpf_found;
651 586
587 if (acpi_lapic && early)
588 return;
652 /* 589 /*
653 * ACPI supports both logical (e.g. Hyper-Threading) and physical 590 * ACPI supports both logical (e.g. Hyper-Threading) and physical
654 * processors, where MPS only supports physical. 591 * processors, where MPS only supports physical.
655 */ 592 */
656 if (acpi_lapic && acpi_ioapic) { 593 if (acpi_lapic && acpi_ioapic) {
657 printk(KERN_INFO "Using ACPI (MADT) for SMP configuration information\n"); 594 printk(KERN_INFO "Using ACPI (MADT) for SMP configuration "
595 "information\n");
658 return; 596 return;
659 } 597 } else if (acpi_lapic)
660 else if (acpi_lapic) 598 printk(KERN_INFO "Using ACPI for processor (LAPIC) "
661 printk(KERN_INFO "Using ACPI for processor (LAPIC) configuration information\n"); 599 "configuration information\n");
662 600
663 printk(KERN_INFO "Intel MultiProcessor Specification v1.%d\n", mpf->mpf_specification); 601 printk(KERN_INFO "Intel MultiProcessor Specification v1.%d\n",
664 if (mpf->mpf_feature2 & (1<<7)) { 602 mpf->mpf_specification);
603#ifdef CONFIG_X86_32
604 if (mpf->mpf_feature2 & (1 << 7)) {
665 printk(KERN_INFO " IMCR and PIC compatibility mode.\n"); 605 printk(KERN_INFO " IMCR and PIC compatibility mode.\n");
666 pic_mode = 1; 606 pic_mode = 1;
667 } else { 607 } else {
668 printk(KERN_INFO " Virtual Wire compatibility mode.\n"); 608 printk(KERN_INFO " Virtual Wire compatibility mode.\n");
669 pic_mode = 0; 609 pic_mode = 0;
670 } 610 }
671 611#endif
672 /* 612 /*
673 * Now see if we need to read further. 613 * Now see if we need to read further.
674 */ 614 */
675 if (mpf->mpf_feature1 != 0) { 615 if (mpf->mpf_feature1 != 0) {
616 if (early) {
617 /*
618 * local APIC has default address
619 */
620 mp_lapic_addr = APIC_DEFAULT_PHYS_BASE;
621 return;
622 }
676 623
677 printk(KERN_INFO "Default MP configuration #%d\n", mpf->mpf_feature1); 624 printk(KERN_INFO "Default MP configuration #%d\n",
625 mpf->mpf_feature1);
678 construct_default_ISA_mptable(mpf->mpf_feature1); 626 construct_default_ISA_mptable(mpf->mpf_feature1);
679 627
680 } else if (mpf->mpf_physptr) { 628 } else if (mpf->mpf_physptr) {
@@ -683,12 +631,18 @@ void __init get_smp_config (void)
683 * Read the physical hardware table. Anything here will 631 * Read the physical hardware table. Anything here will
684 * override the defaults. 632 * override the defaults.
685 */ 633 */
686 if (!smp_read_mpc(phys_to_virt(mpf->mpf_physptr))) { 634 if (!smp_read_mpc(phys_to_virt(mpf->mpf_physptr), early)) {
687 smp_found_config = 0; 635 smp_found_config = 0;
688 printk(KERN_ERR "BIOS bug, MP table errors detected!...\n"); 636 printk(KERN_ERR
689 printk(KERN_ERR "... disabling SMP support. (tell your hw vendor)\n"); 637 "BIOS bug, MP table errors detected!...\n");
638 printk(KERN_ERR "... disabling SMP support. "
639 "(tell your hw vendor)\n");
690 return; 640 return;
691 } 641 }
642
643 if (early)
644 return;
645#ifdef CONFIG_X86_IO_APIC
692 /* 646 /*
693 * If there are no explicit MP IRQ entries, then we are 647 * If there are no explicit MP IRQ entries, then we are
694 * broken. We set up most of the low 16 IO-APIC pins to 648 * broken. We set up most of the low 16 IO-APIC pins to
@@ -697,7 +651,9 @@ void __init get_smp_config (void)
697 if (!mp_irq_entries) { 651 if (!mp_irq_entries) {
698 struct mpc_config_bus bus; 652 struct mpc_config_bus bus;
699 653
700 printk(KERN_ERR "BIOS bug, no explicit IRQ entries, using default mptable. (tell your hw vendor)\n"); 654 printk(KERN_ERR "BIOS bug, no explicit IRQ entries, "
655 "using default mptable. "
656 "(tell your hw vendor)\n");
701 657
702 bus.mpc_type = MP_BUS; 658 bus.mpc_type = MP_BUS;
703 bus.mpc_busid = 0; 659 bus.mpc_busid = 0;
@@ -706,36 +662,51 @@ void __init get_smp_config (void)
706 662
707 construct_default_ioirq_mptable(0); 663 construct_default_ioirq_mptable(0);
708 } 664 }
709 665#endif
710 } else 666 } else
711 BUG(); 667 BUG();
712 668
713 printk(KERN_INFO "Processors: %d\n", num_processors); 669 if (!early)
670 printk(KERN_INFO "Processors: %d\n", num_processors);
714 /* 671 /*
715 * Only use the first configuration found. 672 * Only use the first configuration found.
716 */ 673 */
717} 674}
718 675
719static int __init smp_scan_config (unsigned long base, unsigned long length) 676void __init early_get_smp_config(void)
677{
678 __get_smp_config(1);
679}
680
681void __init get_smp_config(void)
720{ 682{
721 unsigned long *bp = phys_to_virt(base); 683 __get_smp_config(0);
684}
685
686static int __init smp_scan_config(unsigned long base, unsigned long length,
687 unsigned reserve)
688{
689 extern void __bad_mpf_size(void);
690 unsigned int *bp = phys_to_virt(base);
722 struct intel_mp_floating *mpf; 691 struct intel_mp_floating *mpf;
723 692
724 printk(KERN_INFO "Scan SMP from %p for %ld bytes.\n", bp,length); 693 Dprintk("Scan SMP from %p for %ld bytes.\n", bp, length);
725 if (sizeof(*mpf) != 16) 694 if (sizeof(*mpf) != 16)
726 printk("Error: MPF size\n"); 695 __bad_mpf_size();
727 696
728 while (length > 0) { 697 while (length > 0) {
729 mpf = (struct intel_mp_floating *)bp; 698 mpf = (struct intel_mp_floating *)bp;
730 if ((*bp == SMP_MAGIC_IDENT) && 699 if ((*bp == SMP_MAGIC_IDENT) &&
731 (mpf->mpf_length == 1) && 700 (mpf->mpf_length == 1) &&
732 !mpf_checksum((unsigned char *)bp, 16) && 701 !mpf_checksum((unsigned char *)bp, 16) &&
733 ((mpf->mpf_specification == 1) 702 ((mpf->mpf_specification == 1)
734 || (mpf->mpf_specification == 4)) ) { 703 || (mpf->mpf_specification == 4))) {
735 704
736 smp_found_config = 1; 705 smp_found_config = 1;
706 mpf_found = mpf;
707#ifdef CONFIG_X86_32
737 printk(KERN_INFO "found SMP MP-table at [%p] %08lx\n", 708 printk(KERN_INFO "found SMP MP-table at [%p] %08lx\n",
738 mpf, virt_to_phys(mpf)); 709 mpf, virt_to_phys(mpf));
739 reserve_bootmem(virt_to_phys(mpf), PAGE_SIZE, 710 reserve_bootmem(virt_to_phys(mpf), PAGE_SIZE,
740 BOOTMEM_DEFAULT); 711 BOOTMEM_DEFAULT);
741 if (mpf->mpf_physptr) { 712 if (mpf->mpf_physptr) {
@@ -756,8 +727,16 @@ static int __init smp_scan_config (unsigned long base, unsigned long length)
756 BOOTMEM_DEFAULT); 727 BOOTMEM_DEFAULT);
757 } 728 }
758 729
759 mpf_found = mpf; 730#else
760 return 1; 731 if (!reserve)
732 return 1;
733
734 reserve_bootmem_generic(virt_to_phys(mpf), PAGE_SIZE);
735 if (mpf->mpf_physptr)
736 reserve_bootmem_generic(mpf->mpf_physptr,
737 PAGE_SIZE);
738#endif
739 return 1;
761 } 740 }
762 bp += 4; 741 bp += 4;
763 length -= 16; 742 length -= 16;
@@ -765,7 +744,7 @@ static int __init smp_scan_config (unsigned long base, unsigned long length)
765 return 0; 744 return 0;
766} 745}
767 746
768void __init find_smp_config (void) 747static void __init __find_smp_config(unsigned reserve)
769{ 748{
770 unsigned int address; 749 unsigned int address;
771 750
@@ -777,9 +756,9 @@ void __init find_smp_config (void)
777 * 2) Scan the top 1K of base RAM 756 * 2) Scan the top 1K of base RAM
778 * 3) Scan the 64K of bios 757 * 3) Scan the 64K of bios
779 */ 758 */
780 if (smp_scan_config(0x0,0x400) || 759 if (smp_scan_config(0x0, 0x400, reserve) ||
781 smp_scan_config(639*0x400,0x400) || 760 smp_scan_config(639 * 0x400, 0x400, reserve) ||
782 smp_scan_config(0xF0000,0x10000)) 761 smp_scan_config(0xF0000, 0x10000, reserve))
783 return; 762 return;
784 /* 763 /*
785 * If it is an SMP machine we should know now, unless the 764 * If it is an SMP machine we should know now, unless the
@@ -800,144 +779,113 @@ void __init find_smp_config (void)
800 779
801 address = get_bios_ebda(); 780 address = get_bios_ebda();
802 if (address) 781 if (address)
803 smp_scan_config(address, 0x400); 782 smp_scan_config(address, 0x400, reserve);
804} 783}
805 784
806int es7000_plat; 785void __init early_find_smp_config(void)
807
808/* --------------------------------------------------------------------------
809 ACPI-based MP Configuration
810 -------------------------------------------------------------------------- */
811
812#ifdef CONFIG_ACPI
813
814void __init mp_register_lapic_address(u64 address)
815{ 786{
816 mp_lapic_addr = (unsigned long) address; 787 __find_smp_config(0);
817
818 set_fixmap_nocache(FIX_APIC_BASE, mp_lapic_addr);
819
820 if (boot_cpu_physical_apicid == -1U)
821 boot_cpu_physical_apicid = GET_APIC_ID(apic_read(APIC_ID));
822
823 Dprintk("Boot CPU = %d\n", boot_cpu_physical_apicid);
824} 788}
825 789
826void __cpuinit mp_register_lapic (u8 id, u8 enabled) 790void __init find_smp_config(void)
827{ 791{
828 struct mpc_config_processor processor; 792 __find_smp_config(1);
829 int boot_cpu = 0; 793}
830
831 if (MAX_APICS - id <= 0) {
832 printk(KERN_WARNING "Processor #%d invalid (max %d)\n",
833 id, MAX_APICS);
834 return;
835 }
836
837 if (id == boot_cpu_physical_apicid)
838 boot_cpu = 1;
839 794
840 processor.mpc_type = MP_PROCESSOR; 795/* --------------------------------------------------------------------------
841 processor.mpc_apicid = id; 796 ACPI-based MP Configuration
842 processor.mpc_apicver = GET_APIC_VERSION(apic_read(APIC_LVR)); 797 -------------------------------------------------------------------------- */
843 processor.mpc_cpuflag = (enabled ? CPU_ENABLED : 0);
844 processor.mpc_cpuflag |= (boot_cpu ? CPU_BOOTPROCESSOR : 0);
845 processor.mpc_cpufeature = (boot_cpu_data.x86 << 8) |
846 (boot_cpu_data.x86_model << 4) | boot_cpu_data.x86_mask;
847 processor.mpc_featureflag = boot_cpu_data.x86_capability[0];
848 processor.mpc_reserved[0] = 0;
849 processor.mpc_reserved[1] = 0;
850 798
851 MP_processor_info(&processor); 799#ifdef CONFIG_ACPI
852}
853 800
854#ifdef CONFIG_X86_IO_APIC 801#ifdef CONFIG_X86_IO_APIC
855 802
856#define MP_ISA_BUS 0 803#define MP_ISA_BUS 0
857#define MP_MAX_IOAPIC_PIN 127 804#define MP_MAX_IOAPIC_PIN 127
858 805
859static struct mp_ioapic_routing { 806extern struct mp_ioapic_routing mp_ioapic_routing[MAX_IO_APICS];
860 int apic_id;
861 int gsi_base;
862 int gsi_end;
863 u32 pin_programmed[4];
864} mp_ioapic_routing[MAX_IO_APICS];
865 807
866static int mp_find_ioapic (int gsi) 808static int mp_find_ioapic(int gsi)
867{ 809{
868 int i = 0; 810 int i = 0;
869 811
870 /* Find the IOAPIC that manages this GSI. */ 812 /* Find the IOAPIC that manages this GSI. */
871 for (i = 0; i < nr_ioapics; i++) { 813 for (i = 0; i < nr_ioapics; i++) {
872 if ((gsi >= mp_ioapic_routing[i].gsi_base) 814 if ((gsi >= mp_ioapic_routing[i].gsi_base)
873 && (gsi <= mp_ioapic_routing[i].gsi_end)) 815 && (gsi <= mp_ioapic_routing[i].gsi_end))
874 return i; 816 return i;
875 } 817 }
876 818
877 printk(KERN_ERR "ERROR: Unable to locate IOAPIC for GSI %d\n", gsi); 819 printk(KERN_ERR "ERROR: Unable to locate IOAPIC for GSI %d\n", gsi);
878
879 return -1; 820 return -1;
880} 821}
881 822
882void __init mp_register_ioapic(u8 id, u32 address, u32 gsi_base) 823static u8 uniq_ioapic_id(u8 id)
824{
825#ifdef CONFIG_X86_32
826 if ((boot_cpu_data.x86_vendor == X86_VENDOR_INTEL) &&
827 !APIC_XAPIC(apic_version[boot_cpu_physical_apicid]))
828 return io_apic_get_unique_id(nr_ioapics, id);
829 else
830 return id;
831#else
832 int i;
833 DECLARE_BITMAP(used, 256);
834 bitmap_zero(used, 256);
835 for (i = 0; i < nr_ioapics; i++) {
836 struct mpc_config_ioapic *ia = &mp_ioapics[i];
837 __set_bit(ia->mpc_apicid, used);
838 }
839 if (!test_bit(id, used))
840 return id;
841 return find_first_zero_bit(used, 256);
842#endif
843}
844
845void __init mp_register_ioapic(int id, u32 address, u32 gsi_base)
883{ 846{
884 int idx = 0; 847 int idx = 0;
885 int tmpid;
886 848
887 if (nr_ioapics >= MAX_IO_APICS) { 849 if (bad_ioapic(address))
888 printk(KERN_ERR "ERROR: Max # of I/O APICs (%d) exceeded "
889 "(found %d)\n", MAX_IO_APICS, nr_ioapics);
890 panic("Recompile kernel with bigger MAX_IO_APICS!\n");
891 }
892 if (!address) {
893 printk(KERN_ERR "WARNING: Bogus (zero) I/O APIC address"
894 " found in MADT table, skipping!\n");
895 return; 850 return;
896 }
897 851
898 idx = nr_ioapics++; 852 idx = nr_ioapics;
899 853
900 mp_ioapics[idx].mpc_type = MP_IOAPIC; 854 mp_ioapics[idx].mpc_type = MP_IOAPIC;
901 mp_ioapics[idx].mpc_flags = MPC_APIC_USABLE; 855 mp_ioapics[idx].mpc_flags = MPC_APIC_USABLE;
902 mp_ioapics[idx].mpc_apicaddr = address; 856 mp_ioapics[idx].mpc_apicaddr = address;
903 857
904 set_fixmap_nocache(FIX_IO_APIC_BASE_0 + idx, address); 858 set_fixmap_nocache(FIX_IO_APIC_BASE_0 + idx, address);
905 if ((boot_cpu_data.x86_vendor == X86_VENDOR_INTEL) 859 mp_ioapics[idx].mpc_apicid = uniq_ioapic_id(id);
906 && !APIC_XAPIC(apic_version[boot_cpu_physical_apicid])) 860#ifdef CONFIG_X86_32
907 tmpid = io_apic_get_unique_id(idx, id);
908 else
909 tmpid = id;
910 if (tmpid == -1) {
911 nr_ioapics--;
912 return;
913 }
914 mp_ioapics[idx].mpc_apicid = tmpid;
915 mp_ioapics[idx].mpc_apicver = io_apic_get_version(idx); 861 mp_ioapics[idx].mpc_apicver = io_apic_get_version(idx);
916 862#else
917 /* 863 mp_ioapics[idx].mpc_apicver = 0;
864#endif
865 /*
918 * Build basic GSI lookup table to facilitate gsi->io_apic lookups 866 * Build basic GSI lookup table to facilitate gsi->io_apic lookups
919 * and to prevent reprogramming of IOAPIC pins (PCI GSIs). 867 * and to prevent reprogramming of IOAPIC pins (PCI GSIs).
920 */ 868 */
921 mp_ioapic_routing[idx].apic_id = mp_ioapics[idx].mpc_apicid; 869 mp_ioapic_routing[idx].apic_id = mp_ioapics[idx].mpc_apicid;
922 mp_ioapic_routing[idx].gsi_base = gsi_base; 870 mp_ioapic_routing[idx].gsi_base = gsi_base;
923 mp_ioapic_routing[idx].gsi_end = gsi_base + 871 mp_ioapic_routing[idx].gsi_end = gsi_base +
924 io_apic_get_redir_entries(idx); 872 io_apic_get_redir_entries(idx);
925 873
926 printk("IOAPIC[%d]: apic_id %d, version %d, address 0x%x, " 874 printk(KERN_INFO "IOAPIC[%d]: apic_id %d, version %d, address 0x%x, "
927 "GSI %d-%d\n", idx, mp_ioapics[idx].mpc_apicid, 875 "GSI %d-%d\n", idx, mp_ioapics[idx].mpc_apicid,
928 mp_ioapics[idx].mpc_apicver, mp_ioapics[idx].mpc_apicaddr, 876 mp_ioapics[idx].mpc_apicver, mp_ioapics[idx].mpc_apicaddr,
929 mp_ioapic_routing[idx].gsi_base, 877 mp_ioapic_routing[idx].gsi_base, mp_ioapic_routing[idx].gsi_end);
930 mp_ioapic_routing[idx].gsi_end); 878
879 nr_ioapics++;
931} 880}
932 881
933void __init 882void __init mp_override_legacy_irq(u8 bus_irq, u8 polarity, u8 trigger, u32 gsi)
934mp_override_legacy_irq(u8 bus_irq, u8 polarity, u8 trigger, u32 gsi)
935{ 883{
936 struct mpc_config_intsrc intsrc; 884 struct mpc_config_intsrc intsrc;
937 int ioapic = -1; 885 int ioapic = -1;
938 int pin = -1; 886 int pin = -1;
939 887
940 /* 888 /*
941 * Convert 'gsi' to 'ioapic.pin'. 889 * Convert 'gsi' to 'ioapic.pin'.
942 */ 890 */
943 ioapic = mp_find_ioapic(gsi); 891 ioapic = mp_find_ioapic(gsi);
@@ -947,7 +895,7 @@ mp_override_legacy_irq(u8 bus_irq, u8 polarity, u8 trigger, u32 gsi)
947 895
948 /* 896 /*
949 * TBD: This check is for faulty timer entries, where the override 897 * TBD: This check is for faulty timer entries, where the override
950 * erroneously sets the trigger to level, resulting in a HUGE 898 * erroneously sets the trigger to level, resulting in a HUGE
951 * increase of timer interrupts! 899 * increase of timer interrupts!
952 */ 900 */
953 if ((bus_irq == 0) && (trigger == 3)) 901 if ((bus_irq == 0) && (trigger == 3))
@@ -957,13 +905,13 @@ mp_override_legacy_irq(u8 bus_irq, u8 polarity, u8 trigger, u32 gsi)
957 intsrc.mpc_irqtype = mp_INT; 905 intsrc.mpc_irqtype = mp_INT;
958 intsrc.mpc_irqflag = (trigger << 2) | polarity; 906 intsrc.mpc_irqflag = (trigger << 2) | polarity;
959 intsrc.mpc_srcbus = MP_ISA_BUS; 907 intsrc.mpc_srcbus = MP_ISA_BUS;
960 intsrc.mpc_srcbusirq = bus_irq; /* IRQ */ 908 intsrc.mpc_srcbusirq = bus_irq; /* IRQ */
961 intsrc.mpc_dstapic = mp_ioapics[ioapic].mpc_apicid; /* APIC ID */ 909 intsrc.mpc_dstapic = mp_ioapics[ioapic].mpc_apicid; /* APIC ID */
962 intsrc.mpc_dstirq = pin; /* INTIN# */ 910 intsrc.mpc_dstirq = pin; /* INTIN# */
963 911
964 Dprintk("Int: type %d, pol %d, trig %d, bus %d, irq %d, %d-%d\n", 912 Dprintk("Int: type %d, pol %d, trig %d, bus %d, irq %d, %d-%d\n",
965 intsrc.mpc_irqtype, intsrc.mpc_irqflag & 3, 913 intsrc.mpc_irqtype, intsrc.mpc_irqflag & 3,
966 (intsrc.mpc_irqflag >> 2) & 3, intsrc.mpc_srcbus, 914 (intsrc.mpc_irqflag >> 2) & 3, intsrc.mpc_srcbus,
967 intsrc.mpc_srcbusirq, intsrc.mpc_dstapic, intsrc.mpc_dstirq); 915 intsrc.mpc_srcbusirq, intsrc.mpc_dstapic, intsrc.mpc_dstirq);
968 916
969 mp_irqs[mp_irq_entries] = intsrc; 917 mp_irqs[mp_irq_entries] = intsrc;
@@ -971,16 +919,21 @@ mp_override_legacy_irq(u8 bus_irq, u8 polarity, u8 trigger, u32 gsi)
971 panic("Max # of irq sources exceeded!\n"); 919 panic("Max # of irq sources exceeded!\n");
972} 920}
973 921
974void __init mp_config_acpi_legacy_irqs (void) 922int es7000_plat;
923
924void __init mp_config_acpi_legacy_irqs(void)
975{ 925{
976 struct mpc_config_intsrc intsrc; 926 struct mpc_config_intsrc intsrc;
977 int i = 0; 927 int i = 0;
978 int ioapic = -1; 928 int ioapic = -1;
979 929
980 /* 930#if defined (CONFIG_MCA) || defined (CONFIG_EISA)
931 /*
981 * Fabricate the legacy ISA bus (bus #31). 932 * Fabricate the legacy ISA bus (bus #31).
982 */ 933 */
983 mp_bus_id_to_type[MP_ISA_BUS] = MP_BUS_ISA; 934 mp_bus_id_to_type[MP_ISA_BUS] = MP_BUS_ISA;
935#endif
936 set_bit(MP_ISA_BUS, mp_bus_not_pci);
984 Dprintk("Bus #%d is ISA\n", MP_ISA_BUS); 937 Dprintk("Bus #%d is ISA\n", MP_ISA_BUS);
985 938
986 /* 939 /*
@@ -989,19 +942,20 @@ void __init mp_config_acpi_legacy_irqs (void)
989 if (es7000_plat == 1) 942 if (es7000_plat == 1)
990 return; 943 return;
991 944
992 /* 945 /*
993 * Locate the IOAPIC that manages the ISA IRQs (0-15). 946 * Locate the IOAPIC that manages the ISA IRQs (0-15).
994 */ 947 */
995 ioapic = mp_find_ioapic(0); 948 ioapic = mp_find_ioapic(0);
996 if (ioapic < 0) 949 if (ioapic < 0)
997 return; 950 return;
998 951
999 intsrc.mpc_type = MP_INTSRC; 952 intsrc.mpc_type = MP_INTSRC;
1000 intsrc.mpc_irqflag = 0; /* Conforming */ 953 intsrc.mpc_irqflag = 0; /* Conforming */
1001 intsrc.mpc_srcbus = MP_ISA_BUS; 954 intsrc.mpc_srcbus = MP_ISA_BUS;
955#ifdef CONFIG_X86_IO_APIC
1002 intsrc.mpc_dstapic = mp_ioapics[ioapic].mpc_apicid; 956 intsrc.mpc_dstapic = mp_ioapics[ioapic].mpc_apicid;
1003 957#endif
1004 /* 958 /*
1005 * Use the default configuration for the IRQs 0-15. Unless 959 * Use the default configuration for the IRQs 0-15. Unless
1006 * overridden by (MADT) interrupt source override entries. 960 * overridden by (MADT) interrupt source override entries.
1007 */ 961 */
@@ -1012,28 +966,29 @@ void __init mp_config_acpi_legacy_irqs (void)
1012 struct mpc_config_intsrc *irq = mp_irqs + idx; 966 struct mpc_config_intsrc *irq = mp_irqs + idx;
1013 967
1014 /* Do we already have a mapping for this ISA IRQ? */ 968 /* Do we already have a mapping for this ISA IRQ? */
1015 if (irq->mpc_srcbus == MP_ISA_BUS && irq->mpc_srcbusirq == i) 969 if (irq->mpc_srcbus == MP_ISA_BUS
970 && irq->mpc_srcbusirq == i)
1016 break; 971 break;
1017 972
1018 /* Do we already have a mapping for this IOAPIC pin */ 973 /* Do we already have a mapping for this IOAPIC pin */
1019 if ((irq->mpc_dstapic == intsrc.mpc_dstapic) && 974 if ((irq->mpc_dstapic == intsrc.mpc_dstapic) &&
1020 (irq->mpc_dstirq == i)) 975 (irq->mpc_dstirq == i))
1021 break; 976 break;
1022 } 977 }
1023 978
1024 if (idx != mp_irq_entries) { 979 if (idx != mp_irq_entries) {
1025 printk(KERN_DEBUG "ACPI: IRQ%d used by override.\n", i); 980 printk(KERN_DEBUG "ACPI: IRQ%d used by override.\n", i);
1026 continue; /* IRQ already used */ 981 continue; /* IRQ already used */
1027 } 982 }
1028 983
1029 intsrc.mpc_irqtype = mp_INT; 984 intsrc.mpc_irqtype = mp_INT;
1030 intsrc.mpc_srcbusirq = i; /* Identity mapped */ 985 intsrc.mpc_srcbusirq = i; /* Identity mapped */
1031 intsrc.mpc_dstirq = i; 986 intsrc.mpc_dstirq = i;
1032 987
1033 Dprintk("Int: type %d, pol %d, trig %d, bus %d, irq %d, " 988 Dprintk("Int: type %d, pol %d, trig %d, bus %d, irq %d, "
1034 "%d-%d\n", intsrc.mpc_irqtype, intsrc.mpc_irqflag & 3, 989 "%d-%d\n", intsrc.mpc_irqtype, intsrc.mpc_irqflag & 3,
1035 (intsrc.mpc_irqflag >> 2) & 3, intsrc.mpc_srcbus, 990 (intsrc.mpc_irqflag >> 2) & 3, intsrc.mpc_srcbus,
1036 intsrc.mpc_srcbusirq, intsrc.mpc_dstapic, 991 intsrc.mpc_srcbusirq, intsrc.mpc_dstapic,
1037 intsrc.mpc_dstirq); 992 intsrc.mpc_dstirq);
1038 993
1039 mp_irqs[mp_irq_entries] = intsrc; 994 mp_irqs[mp_irq_entries] = intsrc;
@@ -1042,21 +997,27 @@ void __init mp_config_acpi_legacy_irqs (void)
1042 } 997 }
1043} 998}
1044 999
1045#define MAX_GSI_NUM 4096
1046#define IRQ_COMPRESSION_START 64
1047
1048int mp_register_gsi(u32 gsi, int triggering, int polarity) 1000int mp_register_gsi(u32 gsi, int triggering, int polarity)
1049{ 1001{
1050 int ioapic = -1; 1002 int ioapic = -1;
1051 int ioapic_pin = 0; 1003 int ioapic_pin = 0;
1052 int idx, bit = 0; 1004 int idx, bit = 0;
1005#ifdef CONFIG_X86_32
1006#define MAX_GSI_NUM 4096
1007#define IRQ_COMPRESSION_START 64
1008
1053 static int pci_irq = IRQ_COMPRESSION_START; 1009 static int pci_irq = IRQ_COMPRESSION_START;
1054 /* 1010 /*
1055 * Mapping between Global System Interrupts, which 1011 * Mapping between Global System Interrupts, which
1056 * represent all possible interrupts, and IRQs 1012 * represent all possible interrupts, and IRQs
1057 * assigned to actual devices. 1013 * assigned to actual devices.
1058 */ 1014 */
1059 static int gsi_to_irq[MAX_GSI_NUM]; 1015 static int gsi_to_irq[MAX_GSI_NUM];
1016#else
1017
1018 if (acpi_irq_model != ACPI_IRQ_MODEL_IOAPIC)
1019 return gsi;
1020#endif
1060 1021
1061 /* Don't set up the ACPI SCI because it's already set up */ 1022 /* Don't set up the ACPI SCI because it's already set up */
1062 if (acpi_gbl_FADT.sci_interrupt == gsi) 1023 if (acpi_gbl_FADT.sci_interrupt == gsi)
@@ -1070,11 +1031,13 @@ int mp_register_gsi(u32 gsi, int triggering, int polarity)
1070 1031
1071 ioapic_pin = gsi - mp_ioapic_routing[ioapic].gsi_base; 1032 ioapic_pin = gsi - mp_ioapic_routing[ioapic].gsi_base;
1072 1033
1034#ifdef CONFIG_X86_32
1073 if (ioapic_renumber_irq) 1035 if (ioapic_renumber_irq)
1074 gsi = ioapic_renumber_irq(ioapic, gsi); 1036 gsi = ioapic_renumber_irq(ioapic, gsi);
1037#endif
1075 1038
1076 /* 1039 /*
1077 * Avoid pin reprogramming. PRTs typically include entries 1040 * Avoid pin reprogramming. PRTs typically include entries
1078 * with redundant pin->gsi mappings (but unique PCI devices); 1041 * with redundant pin->gsi mappings (but unique PCI devices);
1079 * we only program the IOAPIC on the first. 1042 * we only program the IOAPIC on the first.
1080 */ 1043 */
@@ -1082,23 +1045,27 @@ int mp_register_gsi(u32 gsi, int triggering, int polarity)
1082 idx = (ioapic_pin < 32) ? 0 : (ioapic_pin / 32); 1045 idx = (ioapic_pin < 32) ? 0 : (ioapic_pin / 32);
1083 if (idx > 3) { 1046 if (idx > 3) {
1084 printk(KERN_ERR "Invalid reference to IOAPIC pin " 1047 printk(KERN_ERR "Invalid reference to IOAPIC pin "
1085 "%d-%d\n", mp_ioapic_routing[ioapic].apic_id, 1048 "%d-%d\n", mp_ioapic_routing[ioapic].apic_id,
1086 ioapic_pin); 1049 ioapic_pin);
1087 return gsi; 1050 return gsi;
1088 } 1051 }
1089 if ((1<<bit) & mp_ioapic_routing[ioapic].pin_programmed[idx]) { 1052 if ((1 << bit) & mp_ioapic_routing[ioapic].pin_programmed[idx]) {
1090 Dprintk(KERN_DEBUG "Pin %d-%d already programmed\n", 1053 Dprintk(KERN_DEBUG "Pin %d-%d already programmed\n",
1091 mp_ioapic_routing[ioapic].apic_id, ioapic_pin); 1054 mp_ioapic_routing[ioapic].apic_id, ioapic_pin);
1055#ifdef CONFIG_X86_32
1092 return (gsi < IRQ_COMPRESSION_START ? gsi : gsi_to_irq[gsi]); 1056 return (gsi < IRQ_COMPRESSION_START ? gsi : gsi_to_irq[gsi]);
1057#else
1058 return gsi;
1059#endif
1093 } 1060 }
1094 1061
1095 mp_ioapic_routing[ioapic].pin_programmed[idx] |= (1<<bit); 1062 mp_ioapic_routing[ioapic].pin_programmed[idx] |= (1 << bit);
1096 1063#ifdef CONFIG_X86_32
1097 /* 1064 /*
1098 * For GSI >= 64, use IRQ compression 1065 * For GSI >= 64, use IRQ compression
1099 */ 1066 */
1100 if ((gsi >= IRQ_COMPRESSION_START) 1067 if ((gsi >= IRQ_COMPRESSION_START)
1101 && (triggering == ACPI_LEVEL_SENSITIVE)) { 1068 && (triggering == ACPI_LEVEL_SENSITIVE)) {
1102 /* 1069 /*
1103 * For PCI devices assign IRQs in order, avoiding gaps 1070 * For PCI devices assign IRQs in order, avoiding gaps
1104 * due to unused I/O APIC pins. 1071 * due to unused I/O APIC pins.
@@ -1115,8 +1082,7 @@ int mp_register_gsi(u32 gsi, int triggering, int polarity)
1115 * So test for this condition, and if necessary, avoid 1082 * So test for this condition, and if necessary, avoid
1116 * the pin collision. 1083 * the pin collision.
1117 */ 1084 */
1118 if (gsi > 15 || (gsi == 0 && !timer_uses_ioapic_pin_0)) 1085 gsi = pci_irq++;
1119 gsi = pci_irq++;
1120 /* 1086 /*
1121 * Don't assign IRQ used by ACPI SCI 1087 * Don't assign IRQ used by ACPI SCI
1122 */ 1088 */
@@ -1128,10 +1094,10 @@ int mp_register_gsi(u32 gsi, int triggering, int polarity)
1128 return gsi; 1094 return gsi;
1129 } 1095 }
1130 } 1096 }
1131 1097#endif
1132 io_apic_set_pci_routing(ioapic, ioapic_pin, gsi, 1098 io_apic_set_pci_routing(ioapic, ioapic_pin, gsi,
1133 triggering == ACPI_EDGE_SENSITIVE ? 0 : 1, 1099 triggering == ACPI_EDGE_SENSITIVE ? 0 : 1,
1134 polarity == ACPI_ACTIVE_HIGH ? 0 : 1); 1100 polarity == ACPI_ACTIVE_HIGH ? 0 : 1);
1135 return gsi; 1101 return gsi;
1136} 1102}
1137 1103
diff --git a/arch/x86/kernel/mpparse_64.c b/arch/x86/kernel/mpparse_64.c
deleted file mode 100644
index 72ab1403fed7..000000000000
--- a/arch/x86/kernel/mpparse_64.c
+++ /dev/null
@@ -1,867 +0,0 @@
1/*
2 * Intel Multiprocessor Specification 1.1 and 1.4
3 * compliant MP-table parsing routines.
4 *
5 * (c) 1995 Alan Cox, Building #3 <alan@redhat.com>
6 * (c) 1998, 1999, 2000 Ingo Molnar <mingo@redhat.com>
7 *
8 * Fixes
9 * Erich Boleyn : MP v1.4 and additional changes.
10 * Alan Cox : Added EBDA scanning
11 * Ingo Molnar : various cleanups and rewrites
12 * Maciej W. Rozycki: Bits for default MP configurations
13 * Paul Diefenbaugh: Added full ACPI support
14 */
15
16#include <linux/mm.h>
17#include <linux/init.h>
18#include <linux/delay.h>
19#include <linux/bootmem.h>
20#include <linux/kernel_stat.h>
21#include <linux/mc146818rtc.h>
22#include <linux/acpi.h>
23#include <linux/module.h>
24
25#include <asm/smp.h>
26#include <asm/mtrr.h>
27#include <asm/mpspec.h>
28#include <asm/pgalloc.h>
29#include <asm/io_apic.h>
30#include <asm/proto.h>
31#include <asm/acpi.h>
32
33/* Have we found an MP table */
34int smp_found_config;
35
36/*
37 * Various Linux-internal data structures created from the
38 * MP-table.
39 */
40DECLARE_BITMAP(mp_bus_not_pci, MAX_MP_BUSSES);
41int mp_bus_id_to_pci_bus [MAX_MP_BUSSES] = { [0 ... MAX_MP_BUSSES-1] = -1 };
42
43static int mp_current_pci_id = 0;
44/* I/O APIC entries */
45struct mpc_config_ioapic mp_ioapics[MAX_IO_APICS];
46
47/* # of MP IRQ source entries */
48struct mpc_config_intsrc mp_irqs[MAX_IRQ_SOURCES];
49
50/* MP IRQ source entries */
51int mp_irq_entries;
52
53int nr_ioapics;
54unsigned long mp_lapic_addr = 0;
55
56
57
58/* Processor that is doing the boot up */
59unsigned int boot_cpu_id = -1U;
60EXPORT_SYMBOL(boot_cpu_id);
61
62/* Internal processor count */
63unsigned int num_processors;
64
65unsigned disabled_cpus __cpuinitdata;
66
67/* Bitmask of physically existing CPUs */
68physid_mask_t phys_cpu_present_map = PHYSID_MASK_NONE;
69
70u16 x86_bios_cpu_apicid_init[NR_CPUS] __initdata
71 = { [0 ... NR_CPUS-1] = BAD_APICID };
72void *x86_bios_cpu_apicid_early_ptr;
73DEFINE_PER_CPU(u16, x86_bios_cpu_apicid) = BAD_APICID;
74EXPORT_PER_CPU_SYMBOL(x86_bios_cpu_apicid);
75
76
77/*
78 * Intel MP BIOS table parsing routines:
79 */
80
81/*
82 * Checksum an MP configuration block.
83 */
84
85static int __init mpf_checksum(unsigned char *mp, int len)
86{
87 int sum = 0;
88
89 while (len--)
90 sum += *mp++;
91
92 return sum & 0xFF;
93}
94
95static void __cpuinit MP_processor_info(struct mpc_config_processor *m)
96{
97 int cpu;
98 cpumask_t tmp_map;
99 char *bootup_cpu = "";
100
101 if (!(m->mpc_cpuflag & CPU_ENABLED)) {
102 disabled_cpus++;
103 return;
104 }
105 if (m->mpc_cpuflag & CPU_BOOTPROCESSOR) {
106 bootup_cpu = " (Bootup-CPU)";
107 boot_cpu_id = m->mpc_apicid;
108 }
109
110 printk(KERN_INFO "Processor #%d%s\n", m->mpc_apicid, bootup_cpu);
111
112 if (num_processors >= NR_CPUS) {
113 printk(KERN_WARNING "WARNING: NR_CPUS limit of %i reached."
114 " Processor ignored.\n", NR_CPUS);
115 return;
116 }
117
118 num_processors++;
119 cpus_complement(tmp_map, cpu_present_map);
120 cpu = first_cpu(tmp_map);
121
122 physid_set(m->mpc_apicid, phys_cpu_present_map);
123 if (m->mpc_cpuflag & CPU_BOOTPROCESSOR) {
124 /*
125 * x86_bios_cpu_apicid is required to have processors listed
126 * in same order as logical cpu numbers. Hence the first
127 * entry is BSP, and so on.
128 */
129 cpu = 0;
130 }
131 /* are we being called early in kernel startup? */
132 if (x86_cpu_to_apicid_early_ptr) {
133 u16 *cpu_to_apicid = x86_cpu_to_apicid_early_ptr;
134 u16 *bios_cpu_apicid = x86_bios_cpu_apicid_early_ptr;
135
136 cpu_to_apicid[cpu] = m->mpc_apicid;
137 bios_cpu_apicid[cpu] = m->mpc_apicid;
138 } else {
139 per_cpu(x86_cpu_to_apicid, cpu) = m->mpc_apicid;
140 per_cpu(x86_bios_cpu_apicid, cpu) = m->mpc_apicid;
141 }
142
143 cpu_set(cpu, cpu_possible_map);
144 cpu_set(cpu, cpu_present_map);
145}
146
147static void __init MP_bus_info (struct mpc_config_bus *m)
148{
149 char str[7];
150
151 memcpy(str, m->mpc_bustype, 6);
152 str[6] = 0;
153 Dprintk("Bus #%d is %s\n", m->mpc_busid, str);
154
155 if (strncmp(str, "ISA", 3) == 0) {
156 set_bit(m->mpc_busid, mp_bus_not_pci);
157 } else if (strncmp(str, "PCI", 3) == 0) {
158 clear_bit(m->mpc_busid, mp_bus_not_pci);
159 mp_bus_id_to_pci_bus[m->mpc_busid] = mp_current_pci_id;
160 mp_current_pci_id++;
161 } else {
162 printk(KERN_ERR "Unknown bustype %s\n", str);
163 }
164}
165
166static int bad_ioapic(unsigned long address)
167{
168 if (nr_ioapics >= MAX_IO_APICS) {
169 printk(KERN_ERR "ERROR: Max # of I/O APICs (%d) exceeded "
170 "(found %d)\n", MAX_IO_APICS, nr_ioapics);
171 panic("Recompile kernel with bigger MAX_IO_APICS!\n");
172 }
173 if (!address) {
174 printk(KERN_ERR "WARNING: Bogus (zero) I/O APIC address"
175 " found in table, skipping!\n");
176 return 1;
177 }
178 return 0;
179}
180
181static void __init MP_ioapic_info (struct mpc_config_ioapic *m)
182{
183 if (!(m->mpc_flags & MPC_APIC_USABLE))
184 return;
185
186 printk("I/O APIC #%d at 0x%X.\n",
187 m->mpc_apicid, m->mpc_apicaddr);
188
189 if (bad_ioapic(m->mpc_apicaddr))
190 return;
191
192 mp_ioapics[nr_ioapics] = *m;
193 nr_ioapics++;
194}
195
196static void __init MP_intsrc_info (struct mpc_config_intsrc *m)
197{
198 mp_irqs [mp_irq_entries] = *m;
199 Dprintk("Int: type %d, pol %d, trig %d, bus %d,"
200 " IRQ %02x, APIC ID %x, APIC INT %02x\n",
201 m->mpc_irqtype, m->mpc_irqflag & 3,
202 (m->mpc_irqflag >> 2) & 3, m->mpc_srcbus,
203 m->mpc_srcbusirq, m->mpc_dstapic, m->mpc_dstirq);
204 if (++mp_irq_entries >= MAX_IRQ_SOURCES)
205 panic("Max # of irq sources exceeded!!\n");
206}
207
208static void __init MP_lintsrc_info (struct mpc_config_lintsrc *m)
209{
210 Dprintk("Lint: type %d, pol %d, trig %d, bus %d,"
211 " IRQ %02x, APIC ID %x, APIC LINT %02x\n",
212 m->mpc_irqtype, m->mpc_irqflag & 3,
213 (m->mpc_irqflag >> 2) &3, m->mpc_srcbusid,
214 m->mpc_srcbusirq, m->mpc_destapic, m->mpc_destapiclint);
215}
216
217/*
218 * Read/parse the MPC
219 */
220
221static int __init smp_read_mpc(struct mp_config_table *mpc)
222{
223 char str[16];
224 int count=sizeof(*mpc);
225 unsigned char *mpt=((unsigned char *)mpc)+count;
226
227 if (memcmp(mpc->mpc_signature,MPC_SIGNATURE,4)) {
228 printk("MPTABLE: bad signature [%c%c%c%c]!\n",
229 mpc->mpc_signature[0],
230 mpc->mpc_signature[1],
231 mpc->mpc_signature[2],
232 mpc->mpc_signature[3]);
233 return 0;
234 }
235 if (mpf_checksum((unsigned char *)mpc,mpc->mpc_length)) {
236 printk("MPTABLE: checksum error!\n");
237 return 0;
238 }
239 if (mpc->mpc_spec!=0x01 && mpc->mpc_spec!=0x04) {
240 printk(KERN_ERR "MPTABLE: bad table version (%d)!!\n",
241 mpc->mpc_spec);
242 return 0;
243 }
244 if (!mpc->mpc_lapic) {
245 printk(KERN_ERR "MPTABLE: null local APIC address!\n");
246 return 0;
247 }
248 memcpy(str,mpc->mpc_oem,8);
249 str[8] = 0;
250 printk(KERN_INFO "MPTABLE: OEM ID: %s ",str);
251
252 memcpy(str,mpc->mpc_productid,12);
253 str[12] = 0;
254 printk("MPTABLE: Product ID: %s ",str);
255
256 printk("MPTABLE: APIC at: 0x%X\n",mpc->mpc_lapic);
257
258 /* save the local APIC address, it might be non-default */
259 if (!acpi_lapic)
260 mp_lapic_addr = mpc->mpc_lapic;
261
262 /*
263 * Now process the configuration blocks.
264 */
265 while (count < mpc->mpc_length) {
266 switch(*mpt) {
267 case MP_PROCESSOR:
268 {
269 struct mpc_config_processor *m=
270 (struct mpc_config_processor *)mpt;
271 if (!acpi_lapic)
272 MP_processor_info(m);
273 mpt += sizeof(*m);
274 count += sizeof(*m);
275 break;
276 }
277 case MP_BUS:
278 {
279 struct mpc_config_bus *m=
280 (struct mpc_config_bus *)mpt;
281 MP_bus_info(m);
282 mpt += sizeof(*m);
283 count += sizeof(*m);
284 break;
285 }
286 case MP_IOAPIC:
287 {
288 struct mpc_config_ioapic *m=
289 (struct mpc_config_ioapic *)mpt;
290 MP_ioapic_info(m);
291 mpt += sizeof(*m);
292 count += sizeof(*m);
293 break;
294 }
295 case MP_INTSRC:
296 {
297 struct mpc_config_intsrc *m=
298 (struct mpc_config_intsrc *)mpt;
299
300 MP_intsrc_info(m);
301 mpt += sizeof(*m);
302 count += sizeof(*m);
303 break;
304 }
305 case MP_LINTSRC:
306 {
307 struct mpc_config_lintsrc *m=
308 (struct mpc_config_lintsrc *)mpt;
309 MP_lintsrc_info(m);
310 mpt += sizeof(*m);
311 count += sizeof(*m);
312 break;
313 }
314 }
315 }
316 setup_apic_routing();
317 if (!num_processors)
318 printk(KERN_ERR "MPTABLE: no processors registered!\n");
319 return num_processors;
320}
321
322static int __init ELCR_trigger(unsigned int irq)
323{
324 unsigned int port;
325
326 port = 0x4d0 + (irq >> 3);
327 return (inb(port) >> (irq & 7)) & 1;
328}
329
330static void __init construct_default_ioirq_mptable(int mpc_default_type)
331{
332 struct mpc_config_intsrc intsrc;
333 int i;
334 int ELCR_fallback = 0;
335
336 intsrc.mpc_type = MP_INTSRC;
337 intsrc.mpc_irqflag = 0; /* conforming */
338 intsrc.mpc_srcbus = 0;
339 intsrc.mpc_dstapic = mp_ioapics[0].mpc_apicid;
340
341 intsrc.mpc_irqtype = mp_INT;
342
343 /*
344 * If true, we have an ISA/PCI system with no IRQ entries
345 * in the MP table. To prevent the PCI interrupts from being set up
346 * incorrectly, we try to use the ELCR. The sanity check to see if
347 * there is good ELCR data is very simple - IRQ0, 1, 2 and 13 can
348 * never be level sensitive, so we simply see if the ELCR agrees.
349 * If it does, we assume it's valid.
350 */
351 if (mpc_default_type == 5) {
352 printk(KERN_INFO "ISA/PCI bus type with no IRQ information... falling back to ELCR\n");
353
354 if (ELCR_trigger(0) || ELCR_trigger(1) || ELCR_trigger(2) || ELCR_trigger(13))
355 printk(KERN_ERR "ELCR contains invalid data... not using ELCR\n");
356 else {
357 printk(KERN_INFO "Using ELCR to identify PCI interrupts\n");
358 ELCR_fallback = 1;
359 }
360 }
361
362 for (i = 0; i < 16; i++) {
363 switch (mpc_default_type) {
364 case 2:
365 if (i == 0 || i == 13)
366 continue; /* IRQ0 & IRQ13 not connected */
367 /* fall through */
368 default:
369 if (i == 2)
370 continue; /* IRQ2 is never connected */
371 }
372
373 if (ELCR_fallback) {
374 /*
375 * If the ELCR indicates a level-sensitive interrupt, we
376 * copy that information over to the MP table in the
377 * irqflag field (level sensitive, active high polarity).
378 */
379 if (ELCR_trigger(i))
380 intsrc.mpc_irqflag = 13;
381 else
382 intsrc.mpc_irqflag = 0;
383 }
384
385 intsrc.mpc_srcbusirq = i;
386 intsrc.mpc_dstirq = i ? i : 2; /* IRQ0 to INTIN2 */
387 MP_intsrc_info(&intsrc);
388 }
389
390 intsrc.mpc_irqtype = mp_ExtINT;
391 intsrc.mpc_srcbusirq = 0;
392 intsrc.mpc_dstirq = 0; /* 8259A to INTIN0 */
393 MP_intsrc_info(&intsrc);
394}
395
396static inline void __init construct_default_ISA_mptable(int mpc_default_type)
397{
398 struct mpc_config_processor processor;
399 struct mpc_config_bus bus;
400 struct mpc_config_ioapic ioapic;
401 struct mpc_config_lintsrc lintsrc;
402 int linttypes[2] = { mp_ExtINT, mp_NMI };
403 int i;
404
405 /*
406 * local APIC has default address
407 */
408 mp_lapic_addr = APIC_DEFAULT_PHYS_BASE;
409
410 /*
411 * 2 CPUs, numbered 0 & 1.
412 */
413 processor.mpc_type = MP_PROCESSOR;
414 processor.mpc_apicver = 0;
415 processor.mpc_cpuflag = CPU_ENABLED;
416 processor.mpc_cpufeature = 0;
417 processor.mpc_featureflag = 0;
418 processor.mpc_reserved[0] = 0;
419 processor.mpc_reserved[1] = 0;
420 for (i = 0; i < 2; i++) {
421 processor.mpc_apicid = i;
422 MP_processor_info(&processor);
423 }
424
425 bus.mpc_type = MP_BUS;
426 bus.mpc_busid = 0;
427 switch (mpc_default_type) {
428 default:
429 printk(KERN_ERR "???\nUnknown standard configuration %d\n",
430 mpc_default_type);
431 /* fall through */
432 case 1:
433 case 5:
434 memcpy(bus.mpc_bustype, "ISA ", 6);
435 break;
436 }
437 MP_bus_info(&bus);
438 if (mpc_default_type > 4) {
439 bus.mpc_busid = 1;
440 memcpy(bus.mpc_bustype, "PCI ", 6);
441 MP_bus_info(&bus);
442 }
443
444 ioapic.mpc_type = MP_IOAPIC;
445 ioapic.mpc_apicid = 2;
446 ioapic.mpc_apicver = 0;
447 ioapic.mpc_flags = MPC_APIC_USABLE;
448 ioapic.mpc_apicaddr = 0xFEC00000;
449 MP_ioapic_info(&ioapic);
450
451 /*
452 * We set up most of the low 16 IO-APIC pins according to MPS rules.
453 */
454 construct_default_ioirq_mptable(mpc_default_type);
455
456 lintsrc.mpc_type = MP_LINTSRC;
457 lintsrc.mpc_irqflag = 0; /* conforming */
458 lintsrc.mpc_srcbusid = 0;
459 lintsrc.mpc_srcbusirq = 0;
460 lintsrc.mpc_destapic = MP_APIC_ALL;
461 for (i = 0; i < 2; i++) {
462 lintsrc.mpc_irqtype = linttypes[i];
463 lintsrc.mpc_destapiclint = i;
464 MP_lintsrc_info(&lintsrc);
465 }
466}
467
468static struct intel_mp_floating *mpf_found;
469
470/*
471 * Scan the memory blocks for an SMP configuration block.
472 */
473void __init get_smp_config (void)
474{
475 struct intel_mp_floating *mpf = mpf_found;
476
477 /*
478 * ACPI supports both logical (e.g. Hyper-Threading) and physical
479 * processors, where MPS only supports physical.
480 */
481 if (acpi_lapic && acpi_ioapic) {
482 printk(KERN_INFO "Using ACPI (MADT) for SMP configuration information\n");
483 return;
484 }
485 else if (acpi_lapic)
486 printk(KERN_INFO "Using ACPI for processor (LAPIC) configuration information\n");
487
488 printk("Intel MultiProcessor Specification v1.%d\n", mpf->mpf_specification);
489
490 /*
491 * Now see if we need to read further.
492 */
493 if (mpf->mpf_feature1 != 0) {
494
495 printk(KERN_INFO "Default MP configuration #%d\n", mpf->mpf_feature1);
496 construct_default_ISA_mptable(mpf->mpf_feature1);
497
498 } else if (mpf->mpf_physptr) {
499
500 /*
501 * Read the physical hardware table. Anything here will
502 * override the defaults.
503 */
504 if (!smp_read_mpc(phys_to_virt(mpf->mpf_physptr))) {
505 smp_found_config = 0;
506 printk(KERN_ERR "BIOS bug, MP table errors detected!...\n");
507 printk(KERN_ERR "... disabling SMP support. (tell your hw vendor)\n");
508 return;
509 }
510 /*
511 * If there are no explicit MP IRQ entries, then we are
512 * broken. We set up most of the low 16 IO-APIC pins to
513 * ISA defaults and hope it will work.
514 */
515 if (!mp_irq_entries) {
516 struct mpc_config_bus bus;
517
518 printk(KERN_ERR "BIOS bug, no explicit IRQ entries, using default mptable. (tell your hw vendor)\n");
519
520 bus.mpc_type = MP_BUS;
521 bus.mpc_busid = 0;
522 memcpy(bus.mpc_bustype, "ISA ", 6);
523 MP_bus_info(&bus);
524
525 construct_default_ioirq_mptable(0);
526 }
527
528 } else
529 BUG();
530
531 printk(KERN_INFO "Processors: %d\n", num_processors);
532 /*
533 * Only use the first configuration found.
534 */
535}
536
537static int __init smp_scan_config (unsigned long base, unsigned long length)
538{
539 extern void __bad_mpf_size(void);
540 unsigned int *bp = phys_to_virt(base);
541 struct intel_mp_floating *mpf;
542
543 Dprintk("Scan SMP from %p for %ld bytes.\n", bp,length);
544 if (sizeof(*mpf) != 16)
545 __bad_mpf_size();
546
547 while (length > 0) {
548 mpf = (struct intel_mp_floating *)bp;
549 if ((*bp == SMP_MAGIC_IDENT) &&
550 (mpf->mpf_length == 1) &&
551 !mpf_checksum((unsigned char *)bp, 16) &&
552 ((mpf->mpf_specification == 1)
553 || (mpf->mpf_specification == 4)) ) {
554
555 smp_found_config = 1;
556 reserve_bootmem_generic(virt_to_phys(mpf), PAGE_SIZE);
557 if (mpf->mpf_physptr)
558 reserve_bootmem_generic(mpf->mpf_physptr, PAGE_SIZE);
559 mpf_found = mpf;
560 return 1;
561 }
562 bp += 4;
563 length -= 16;
564 }
565 return 0;
566}
567
568void __init find_smp_config(void)
569{
570 unsigned int address;
571
572 /*
573 * FIXME: Linux assumes you have 640K of base ram..
574 * this continues the error...
575 *
576 * 1) Scan the bottom 1K for a signature
577 * 2) Scan the top 1K of base RAM
578 * 3) Scan the 64K of bios
579 */
580 if (smp_scan_config(0x0,0x400) ||
581 smp_scan_config(639*0x400,0x400) ||
582 smp_scan_config(0xF0000,0x10000))
583 return;
584 /*
585 * If it is an SMP machine we should know now.
586 *
587 * there is a real-mode segmented pointer pointing to the
588 * 4K EBDA area at 0x40E, calculate and scan it here.
589 *
590 * NOTE! There are Linux loaders that will corrupt the EBDA
591 * area, and as such this kind of SMP config may be less
592 * trustworthy, simply because the SMP table may have been
593 * stomped on during early boot. These loaders are buggy and
594 * should be fixed.
595 */
596
597 address = *(unsigned short *)phys_to_virt(0x40E);
598 address <<= 4;
599 if (smp_scan_config(address, 0x1000))
600 return;
601
602 /* If we have come this far, we did not find an MP table */
603 printk(KERN_INFO "No mptable found.\n");
604}
605
606/* --------------------------------------------------------------------------
607 ACPI-based MP Configuration
608 -------------------------------------------------------------------------- */
609
610#ifdef CONFIG_ACPI
611
612void __init mp_register_lapic_address(u64 address)
613{
614 mp_lapic_addr = (unsigned long) address;
615 set_fixmap_nocache(FIX_APIC_BASE, mp_lapic_addr);
616 if (boot_cpu_id == -1U)
617 boot_cpu_id = GET_APIC_ID(apic_read(APIC_ID));
618}
619
620void __cpuinit mp_register_lapic (u8 id, u8 enabled)
621{
622 struct mpc_config_processor processor;
623 int boot_cpu = 0;
624
625 if (id == boot_cpu_id)
626 boot_cpu = 1;
627
628 processor.mpc_type = MP_PROCESSOR;
629 processor.mpc_apicid = id;
630 processor.mpc_apicver = 0;
631 processor.mpc_cpuflag = (enabled ? CPU_ENABLED : 0);
632 processor.mpc_cpuflag |= (boot_cpu ? CPU_BOOTPROCESSOR : 0);
633 processor.mpc_cpufeature = 0;
634 processor.mpc_featureflag = 0;
635 processor.mpc_reserved[0] = 0;
636 processor.mpc_reserved[1] = 0;
637
638 MP_processor_info(&processor);
639}
640
641#define MP_ISA_BUS 0
642#define MP_MAX_IOAPIC_PIN 127
643
644static struct mp_ioapic_routing {
645 int apic_id;
646 int gsi_start;
647 int gsi_end;
648 u32 pin_programmed[4];
649} mp_ioapic_routing[MAX_IO_APICS];
650
651static int mp_find_ioapic(int gsi)
652{
653 int i = 0;
654
655 /* Find the IOAPIC that manages this GSI. */
656 for (i = 0; i < nr_ioapics; i++) {
657 if ((gsi >= mp_ioapic_routing[i].gsi_start)
658 && (gsi <= mp_ioapic_routing[i].gsi_end))
659 return i;
660 }
661
662 printk(KERN_ERR "ERROR: Unable to locate IOAPIC for GSI %d\n", gsi);
663 return -1;
664}
665
666static u8 uniq_ioapic_id(u8 id)
667{
668 int i;
669 DECLARE_BITMAP(used, 256);
670 bitmap_zero(used, 256);
671 for (i = 0; i < nr_ioapics; i++) {
672 struct mpc_config_ioapic *ia = &mp_ioapics[i];
673 __set_bit(ia->mpc_apicid, used);
674 }
675 if (!test_bit(id, used))
676 return id;
677 return find_first_zero_bit(used, 256);
678}
679
680void __init mp_register_ioapic(u8 id, u32 address, u32 gsi_base)
681{
682 int idx = 0;
683
684 if (bad_ioapic(address))
685 return;
686
687 idx = nr_ioapics;
688
689 mp_ioapics[idx].mpc_type = MP_IOAPIC;
690 mp_ioapics[idx].mpc_flags = MPC_APIC_USABLE;
691 mp_ioapics[idx].mpc_apicaddr = address;
692
693 set_fixmap_nocache(FIX_IO_APIC_BASE_0 + idx, address);
694 mp_ioapics[idx].mpc_apicid = uniq_ioapic_id(id);
695 mp_ioapics[idx].mpc_apicver = 0;
696
697 /*
698 * Build basic IRQ lookup table to facilitate gsi->io_apic lookups
699 * and to prevent reprogramming of IOAPIC pins (PCI IRQs).
700 */
701 mp_ioapic_routing[idx].apic_id = mp_ioapics[idx].mpc_apicid;
702 mp_ioapic_routing[idx].gsi_start = gsi_base;
703 mp_ioapic_routing[idx].gsi_end = gsi_base +
704 io_apic_get_redir_entries(idx);
705
706 printk(KERN_INFO "IOAPIC[%d]: apic_id %d, address 0x%x, "
707 "GSI %d-%d\n", idx, mp_ioapics[idx].mpc_apicid,
708 mp_ioapics[idx].mpc_apicaddr,
709 mp_ioapic_routing[idx].gsi_start,
710 mp_ioapic_routing[idx].gsi_end);
711
712 nr_ioapics++;
713}
714
715void __init
716mp_override_legacy_irq(u8 bus_irq, u8 polarity, u8 trigger, u32 gsi)
717{
718 struct mpc_config_intsrc intsrc;
719 int ioapic = -1;
720 int pin = -1;
721
722 /*
723 * Convert 'gsi' to 'ioapic.pin'.
724 */
725 ioapic = mp_find_ioapic(gsi);
726 if (ioapic < 0)
727 return;
728 pin = gsi - mp_ioapic_routing[ioapic].gsi_start;
729
730 /*
731 * TBD: This check is for faulty timer entries, where the override
732 * erroneously sets the trigger to level, resulting in a HUGE
733 * increase of timer interrupts!
734 */
735 if ((bus_irq == 0) && (trigger == 3))
736 trigger = 1;
737
738 intsrc.mpc_type = MP_INTSRC;
739 intsrc.mpc_irqtype = mp_INT;
740 intsrc.mpc_irqflag = (trigger << 2) | polarity;
741 intsrc.mpc_srcbus = MP_ISA_BUS;
742 intsrc.mpc_srcbusirq = bus_irq; /* IRQ */
743 intsrc.mpc_dstapic = mp_ioapics[ioapic].mpc_apicid; /* APIC ID */
744 intsrc.mpc_dstirq = pin; /* INTIN# */
745
746 Dprintk("Int: type %d, pol %d, trig %d, bus %d, irq %d, %d-%d\n",
747 intsrc.mpc_irqtype, intsrc.mpc_irqflag & 3,
748 (intsrc.mpc_irqflag >> 2) & 3, intsrc.mpc_srcbus,
749 intsrc.mpc_srcbusirq, intsrc.mpc_dstapic, intsrc.mpc_dstirq);
750
751 mp_irqs[mp_irq_entries] = intsrc;
752 if (++mp_irq_entries == MAX_IRQ_SOURCES)
753 panic("Max # of irq sources exceeded!\n");
754}
755
756void __init mp_config_acpi_legacy_irqs(void)
757{
758 struct mpc_config_intsrc intsrc;
759 int i = 0;
760 int ioapic = -1;
761
762 /*
763 * Fabricate the legacy ISA bus (bus #31).
764 */
765 set_bit(MP_ISA_BUS, mp_bus_not_pci);
766
767 /*
768 * Locate the IOAPIC that manages the ISA IRQs (0-15).
769 */
770 ioapic = mp_find_ioapic(0);
771 if (ioapic < 0)
772 return;
773
774 intsrc.mpc_type = MP_INTSRC;
775 intsrc.mpc_irqflag = 0; /* Conforming */
776 intsrc.mpc_srcbus = MP_ISA_BUS;
777 intsrc.mpc_dstapic = mp_ioapics[ioapic].mpc_apicid;
778
779 /*
780 * Use the default configuration for the IRQs 0-15. Unless
781 * overridden by (MADT) interrupt source override entries.
782 */
783 for (i = 0; i < 16; i++) {
784 int idx;
785
786 for (idx = 0; idx < mp_irq_entries; idx++) {
787 struct mpc_config_intsrc *irq = mp_irqs + idx;
788
789 /* Do we already have a mapping for this ISA IRQ? */
790 if (irq->mpc_srcbus == MP_ISA_BUS && irq->mpc_srcbusirq == i)
791 break;
792
793 /* Do we already have a mapping for this IOAPIC pin */
794 if ((irq->mpc_dstapic == intsrc.mpc_dstapic) &&
795 (irq->mpc_dstirq == i))
796 break;
797 }
798
799 if (idx != mp_irq_entries) {
800 printk(KERN_DEBUG "ACPI: IRQ%d used by override.\n", i);
801 continue; /* IRQ already used */
802 }
803
804 intsrc.mpc_irqtype = mp_INT;
805 intsrc.mpc_srcbusirq = i; /* Identity mapped */
806 intsrc.mpc_dstirq = i;
807
808 Dprintk("Int: type %d, pol %d, trig %d, bus %d, irq %d, "
809 "%d-%d\n", intsrc.mpc_irqtype, intsrc.mpc_irqflag & 3,
810 (intsrc.mpc_irqflag >> 2) & 3, intsrc.mpc_srcbus,
811 intsrc.mpc_srcbusirq, intsrc.mpc_dstapic,
812 intsrc.mpc_dstirq);
813
814 mp_irqs[mp_irq_entries] = intsrc;
815 if (++mp_irq_entries == MAX_IRQ_SOURCES)
816 panic("Max # of irq sources exceeded!\n");
817 }
818}
819
820int mp_register_gsi(u32 gsi, int triggering, int polarity)
821{
822 int ioapic = -1;
823 int ioapic_pin = 0;
824 int idx, bit = 0;
825
826 if (acpi_irq_model != ACPI_IRQ_MODEL_IOAPIC)
827 return gsi;
828
829 /* Don't set up the ACPI SCI because it's already set up */
830 if (acpi_gbl_FADT.sci_interrupt == gsi)
831 return gsi;
832
833 ioapic = mp_find_ioapic(gsi);
834 if (ioapic < 0) {
835 printk(KERN_WARNING "No IOAPIC for GSI %u\n", gsi);
836 return gsi;
837 }
838
839 ioapic_pin = gsi - mp_ioapic_routing[ioapic].gsi_start;
840
841 /*
842 * Avoid pin reprogramming. PRTs typically include entries
843 * with redundant pin->gsi mappings (but unique PCI devices);
844 * we only program the IOAPIC on the first.
845 */
846 bit = ioapic_pin % 32;
847 idx = (ioapic_pin < 32) ? 0 : (ioapic_pin / 32);
848 if (idx > 3) {
849 printk(KERN_ERR "Invalid reference to IOAPIC pin "
850 "%d-%d\n", mp_ioapic_routing[ioapic].apic_id,
851 ioapic_pin);
852 return gsi;
853 }
854 if ((1<<bit) & mp_ioapic_routing[ioapic].pin_programmed[idx]) {
855 Dprintk(KERN_DEBUG "Pin %d-%d already programmed\n",
856 mp_ioapic_routing[ioapic].apic_id, ioapic_pin);
857 return gsi;
858 }
859
860 mp_ioapic_routing[ioapic].pin_programmed[idx] |= (1<<bit);
861
862 io_apic_set_pci_routing(ioapic, ioapic_pin, gsi,
863 triggering == ACPI_EDGE_SENSITIVE ? 0 : 1,
864 polarity == ACPI_ACTIVE_HIGH ? 0 : 1);
865 return gsi;
866}
867#endif /*CONFIG_ACPI*/
diff --git a/arch/x86/kernel/msr.c b/arch/x86/kernel/msr.c
index af51ea8400b2..4dfb40530057 100644
--- a/arch/x86/kernel/msr.c
+++ b/arch/x86/kernel/msr.c
@@ -65,8 +65,8 @@ static loff_t msr_seek(struct file *file, loff_t offset, int orig)
65 return ret; 65 return ret;
66} 66}
67 67
68static ssize_t msr_read(struct file *file, char __user * buf, 68static ssize_t msr_read(struct file *file, char __user *buf,
69 size_t count, loff_t * ppos) 69 size_t count, loff_t *ppos)
70{ 70{
71 u32 __user *tmp = (u32 __user *) buf; 71 u32 __user *tmp = (u32 __user *) buf;
72 u32 data[2]; 72 u32 data[2];
diff --git a/arch/x86/kernel/nmi_32.c b/arch/x86/kernel/nmi_32.c
index 6a0aa7038685..8421d0ac6f22 100644
--- a/arch/x86/kernel/nmi_32.c
+++ b/arch/x86/kernel/nmi_32.c
@@ -22,9 +22,11 @@
22#include <linux/cpumask.h> 22#include <linux/cpumask.h>
23#include <linux/kernel_stat.h> 23#include <linux/kernel_stat.h>
24#include <linux/kdebug.h> 24#include <linux/kdebug.h>
25#include <linux/slab.h>
25 26
26#include <asm/smp.h> 27#include <asm/smp.h>
27#include <asm/nmi.h> 28#include <asm/nmi.h>
29#include <asm/timer.h>
28 30
29#include "mach_traps.h" 31#include "mach_traps.h"
30 32
@@ -67,7 +69,7 @@ static __init void nmi_cpu_busy(void *data)
67} 69}
68#endif 70#endif
69 71
70static int __init check_nmi_watchdog(void) 72int __init check_nmi_watchdog(void)
71{ 73{
72 unsigned int *prev_nmi_count; 74 unsigned int *prev_nmi_count;
73 int cpu; 75 int cpu;
@@ -80,7 +82,7 @@ static int __init check_nmi_watchdog(void)
80 82
81 prev_nmi_count = kmalloc(NR_CPUS * sizeof(int), GFP_KERNEL); 83 prev_nmi_count = kmalloc(NR_CPUS * sizeof(int), GFP_KERNEL);
82 if (!prev_nmi_count) 84 if (!prev_nmi_count)
83 return -1; 85 goto error;
84 86
85 printk(KERN_INFO "Testing NMI watchdog ... "); 87 printk(KERN_INFO "Testing NMI watchdog ... ");
86 88
@@ -117,7 +119,7 @@ static int __init check_nmi_watchdog(void)
117 if (!atomic_read(&nmi_active)) { 119 if (!atomic_read(&nmi_active)) {
118 kfree(prev_nmi_count); 120 kfree(prev_nmi_count);
119 atomic_set(&nmi_active, -1); 121 atomic_set(&nmi_active, -1);
120 return -1; 122 goto error;
121 } 123 }
122 printk("OK.\n"); 124 printk("OK.\n");
123 125
@@ -128,9 +130,11 @@ static int __init check_nmi_watchdog(void)
128 130
129 kfree(prev_nmi_count); 131 kfree(prev_nmi_count);
130 return 0; 132 return 0;
133error:
134 timer_ack = !cpu_has_tsc;
135
136 return -1;
131} 137}
132/* This needs to happen later in boot so counters are working */
133late_initcall(check_nmi_watchdog);
134 138
135static int __init setup_nmi_watchdog(char *str) 139static int __init setup_nmi_watchdog(char *str)
136{ 140{
diff --git a/arch/x86/kernel/nmi_64.c b/arch/x86/kernel/nmi_64.c
index 9a4fde74bee1..11f9130ac513 100644
--- a/arch/x86/kernel/nmi_64.c
+++ b/arch/x86/kernel/nmi_64.c
@@ -26,6 +26,8 @@
26#include <asm/proto.h> 26#include <asm/proto.h>
27#include <asm/mce.h> 27#include <asm/mce.h>
28 28
29#include <mach_traps.h>
30
29int unknown_nmi_panic; 31int unknown_nmi_panic;
30int nmi_watchdog_enabled; 32int nmi_watchdog_enabled;
31int panic_on_unrecovered_nmi; 33int panic_on_unrecovered_nmi;
diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c
index 075962cc75ab..3733412d1357 100644
--- a/arch/x86/kernel/paravirt.c
+++ b/arch/x86/kernel/paravirt.c
@@ -206,13 +206,6 @@ static struct resource reserve_ioports = {
206 .flags = IORESOURCE_IO | IORESOURCE_BUSY, 206 .flags = IORESOURCE_IO | IORESOURCE_BUSY,
207}; 207};
208 208
209static struct resource reserve_iomem = {
210 .start = 0,
211 .end = -1,
212 .name = "paravirt-iomem",
213 .flags = IORESOURCE_MEM | IORESOURCE_BUSY,
214};
215
216/* 209/*
217 * Reserve the whole legacy IO space to prevent any legacy drivers 210 * Reserve the whole legacy IO space to prevent any legacy drivers
218 * from wasting time probing for their hardware. This is a fairly 211 * from wasting time probing for their hardware. This is a fairly
@@ -222,16 +215,7 @@ static struct resource reserve_iomem = {
222 */ 215 */
223int paravirt_disable_iospace(void) 216int paravirt_disable_iospace(void)
224{ 217{
225 int ret; 218 return request_resource(&ioport_resource, &reserve_ioports);
226
227 ret = request_resource(&ioport_resource, &reserve_ioports);
228 if (ret == 0) {
229 ret = request_resource(&iomem_resource, &reserve_iomem);
230 if (ret)
231 release_resource(&reserve_ioports);
232 }
233
234 return ret;
235} 219}
236 220
237static DEFINE_PER_CPU(enum paravirt_lazy_mode, paravirt_lazy_mode) = PARAVIRT_LAZY_NONE; 221static DEFINE_PER_CPU(enum paravirt_lazy_mode, paravirt_lazy_mode) = PARAVIRT_LAZY_NONE;
diff --git a/arch/x86/kernel/pci-dma_64.c b/arch/x86/kernel/pci-dma_64.c
index 375cb2bc45be..ada5a0604992 100644
--- a/arch/x86/kernel/pci-dma_64.c
+++ b/arch/x86/kernel/pci-dma_64.c
@@ -232,32 +232,32 @@ static __init int iommu_setup(char *p)
232 return -EINVAL; 232 return -EINVAL;
233 233
234 while (*p) { 234 while (*p) {
235 if (!strncmp(p,"off",3)) 235 if (!strncmp(p, "off", 3))
236 no_iommu = 1; 236 no_iommu = 1;
237 /* gart_parse_options has more force support */ 237 /* gart_parse_options has more force support */
238 if (!strncmp(p,"force",5)) 238 if (!strncmp(p, "force", 5))
239 force_iommu = 1; 239 force_iommu = 1;
240 if (!strncmp(p,"noforce",7)) { 240 if (!strncmp(p, "noforce", 7)) {
241 iommu_merge = 0; 241 iommu_merge = 0;
242 force_iommu = 0; 242 force_iommu = 0;
243 } 243 }
244 244
245 if (!strncmp(p, "biomerge",8)) { 245 if (!strncmp(p, "biomerge", 8)) {
246 iommu_bio_merge = 4096; 246 iommu_bio_merge = 4096;
247 iommu_merge = 1; 247 iommu_merge = 1;
248 force_iommu = 1; 248 force_iommu = 1;
249 } 249 }
250 if (!strncmp(p, "panic",5)) 250 if (!strncmp(p, "panic", 5))
251 panic_on_overflow = 1; 251 panic_on_overflow = 1;
252 if (!strncmp(p, "nopanic",7)) 252 if (!strncmp(p, "nopanic", 7))
253 panic_on_overflow = 0; 253 panic_on_overflow = 0;
254 if (!strncmp(p, "merge",5)) { 254 if (!strncmp(p, "merge", 5)) {
255 iommu_merge = 1; 255 iommu_merge = 1;
256 force_iommu = 1; 256 force_iommu = 1;
257 } 257 }
258 if (!strncmp(p, "nomerge",7)) 258 if (!strncmp(p, "nomerge", 7))
259 iommu_merge = 0; 259 iommu_merge = 0;
260 if (!strncmp(p, "forcesac",8)) 260 if (!strncmp(p, "forcesac", 8))
261 iommu_sac_force = 1; 261 iommu_sac_force = 1;
262 if (!strncmp(p, "allowdac", 8)) 262 if (!strncmp(p, "allowdac", 8))
263 forbid_dac = 0; 263 forbid_dac = 0;
@@ -265,7 +265,7 @@ static __init int iommu_setup(char *p)
265 forbid_dac = -1; 265 forbid_dac = -1;
266 266
267#ifdef CONFIG_SWIOTLB 267#ifdef CONFIG_SWIOTLB
268 if (!strncmp(p, "soft",4)) 268 if (!strncmp(p, "soft", 4))
269 swiotlb = 1; 269 swiotlb = 1;
270#endif 270#endif
271 271
diff --git a/arch/x86/kernel/process_32.c b/arch/x86/kernel/process_32.c
index 43930e73f657..3903a8f2eb97 100644
--- a/arch/x86/kernel/process_32.c
+++ b/arch/x86/kernel/process_32.c
@@ -113,20 +113,13 @@ void default_idle(void)
113 113
114 local_irq_disable(); 114 local_irq_disable();
115 if (!need_resched()) { 115 if (!need_resched()) {
116 ktime_t t0, t1;
117 u64 t0n, t1n;
118
119 t0 = ktime_get();
120 t0n = ktime_to_ns(t0);
121 safe_halt(); /* enables interrupts racelessly */ 116 safe_halt(); /* enables interrupts racelessly */
122 local_irq_disable(); 117 local_irq_disable();
123 t1 = ktime_get();
124 t1n = ktime_to_ns(t1);
125 sched_clock_idle_wakeup_event(t1n - t0n);
126 } 118 }
127 local_irq_enable(); 119 local_irq_enable();
128 current_thread_info()->status |= TS_POLLING; 120 current_thread_info()->status |= TS_POLLING;
129 } else { 121 } else {
122 local_irq_enable();
130 /* loop is done by the caller */ 123 /* loop is done by the caller */
131 cpu_relax(); 124 cpu_relax();
132 } 125 }
@@ -142,6 +135,7 @@ EXPORT_SYMBOL(default_idle);
142 */ 135 */
143static void poll_idle(void) 136static void poll_idle(void)
144{ 137{
138 local_irq_enable();
145 cpu_relax(); 139 cpu_relax();
146} 140}
147 141
@@ -248,8 +242,11 @@ void mwait_idle_with_hints(unsigned long ax, unsigned long cx)
248 __monitor((void *)&current_thread_info()->flags, 0, 0); 242 __monitor((void *)&current_thread_info()->flags, 0, 0);
249 smp_mb(); 243 smp_mb();
250 if (!need_resched()) 244 if (!need_resched())
251 __mwait(ax, cx); 245 __sti_mwait(ax, cx);
252 } 246 else
247 local_irq_enable();
248 } else
249 local_irq_enable();
253} 250}
254 251
255/* Default MONITOR/MWAIT with no hints, used for default C1 state */ 252/* Default MONITOR/MWAIT with no hints, used for default C1 state */
@@ -332,7 +329,7 @@ void __show_registers(struct pt_regs *regs, int all)
332 init_utsname()->version); 329 init_utsname()->version);
333 330
334 printk("EIP: %04x:[<%08lx>] EFLAGS: %08lx CPU: %d\n", 331 printk("EIP: %04x:[<%08lx>] EFLAGS: %08lx CPU: %d\n",
335 0xffff & regs->cs, regs->ip, regs->flags, 332 (u16)regs->cs, regs->ip, regs->flags,
336 smp_processor_id()); 333 smp_processor_id());
337 print_symbol("EIP is at %s\n", regs->ip); 334 print_symbol("EIP is at %s\n", regs->ip);
338 335
@@ -341,8 +338,7 @@ void __show_registers(struct pt_regs *regs, int all)
341 printk("ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n", 338 printk("ESI: %08lx EDI: %08lx EBP: %08lx ESP: %08lx\n",
342 regs->si, regs->di, regs->bp, sp); 339 regs->si, regs->di, regs->bp, sp);
343 printk(" DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n", 340 printk(" DS: %04x ES: %04x FS: %04x GS: %04x SS: %04x\n",
344 regs->ds & 0xffff, regs->es & 0xffff, 341 (u16)regs->ds, (u16)regs->es, (u16)regs->fs, gs, ss);
345 regs->fs & 0xffff, gs, ss);
346 342
347 if (!all) 343 if (!all)
348 return; 344 return;
@@ -513,6 +509,21 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long sp,
513 return err; 509 return err;
514} 510}
515 511
512void
513start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp)
514{
515 __asm__("movl %0, %%gs" :: "r"(0));
516 regs->fs = 0;
517 set_fs(USER_DS);
518 regs->ds = __USER_DS;
519 regs->es = __USER_DS;
520 regs->ss = __USER_DS;
521 regs->cs = __USER_CS;
522 regs->ip = new_ip;
523 regs->sp = new_sp;
524}
525EXPORT_SYMBOL_GPL(start_thread);
526
516#ifdef CONFIG_SECCOMP 527#ifdef CONFIG_SECCOMP
517static void hard_disable_TSC(void) 528static void hard_disable_TSC(void)
518{ 529{
@@ -550,12 +561,12 @@ __switch_to_xtra(struct task_struct *prev_p, struct task_struct *next_p,
550 /* we clear debugctl to make sure DS 561 /* we clear debugctl to make sure DS
551 * is not in use when we change it */ 562 * is not in use when we change it */
552 debugctl = 0; 563 debugctl = 0;
553 wrmsrl(MSR_IA32_DEBUGCTLMSR, 0); 564 update_debugctlmsr(0);
554 wrmsr(MSR_IA32_DS_AREA, next->ds_area_msr, 0); 565 wrmsr(MSR_IA32_DS_AREA, next->ds_area_msr, 0);
555 } 566 }
556 567
557 if (next->debugctlmsr != debugctl) 568 if (next->debugctlmsr != debugctl)
558 wrmsr(MSR_IA32_DEBUGCTLMSR, next->debugctlmsr, 0); 569 update_debugctlmsr(next->debugctlmsr);
559 570
560 if (test_tsk_thread_flag(next_p, TIF_DEBUG)) { 571 if (test_tsk_thread_flag(next_p, TIF_DEBUG)) {
561 set_debugreg(next->debugreg0, 0); 572 set_debugreg(next->debugreg0, 0);
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index 46c4c546b499..e75ccc8a2b87 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -107,16 +107,8 @@ void default_idle(void)
107 smp_mb(); 107 smp_mb();
108 local_irq_disable(); 108 local_irq_disable();
109 if (!need_resched()) { 109 if (!need_resched()) {
110 ktime_t t0, t1;
111 u64 t0n, t1n;
112
113 t0 = ktime_get();
114 t0n = ktime_to_ns(t0);
115 safe_halt(); /* enables interrupts racelessly */ 110 safe_halt(); /* enables interrupts racelessly */
116 local_irq_disable(); 111 local_irq_disable();
117 t1 = ktime_get();
118 t1n = ktime_to_ns(t1);
119 sched_clock_idle_wakeup_event(t1n - t0n);
120 } 112 }
121 local_irq_enable(); 113 local_irq_enable();
122 current_thread_info()->status |= TS_POLLING; 114 current_thread_info()->status |= TS_POLLING;
@@ -528,6 +520,21 @@ out:
528 return err; 520 return err;
529} 521}
530 522
523void
524start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp)
525{
526 asm volatile("movl %0, %%fs; movl %0, %%es; movl %0, %%ds" :: "r"(0));
527 load_gs_index(0);
528 regs->ip = new_ip;
529 regs->sp = new_sp;
530 write_pda(oldrsp, new_sp);
531 regs->cs = __USER_CS;
532 regs->ss = __USER_DS;
533 regs->flags = 0x200;
534 set_fs(USER_DS);
535}
536EXPORT_SYMBOL_GPL(start_thread);
537
531/* 538/*
532 * This special macro can be used to load a debugging register 539 * This special macro can be used to load a debugging register
533 */ 540 */
@@ -548,12 +555,12 @@ static inline void __switch_to_xtra(struct task_struct *prev_p,
548 /* we clear debugctl to make sure DS 555 /* we clear debugctl to make sure DS
549 * is not in use when we change it */ 556 * is not in use when we change it */
550 debugctl = 0; 557 debugctl = 0;
551 wrmsrl(MSR_IA32_DEBUGCTLMSR, 0); 558 update_debugctlmsr(0);
552 wrmsrl(MSR_IA32_DS_AREA, next->ds_area_msr); 559 wrmsrl(MSR_IA32_DS_AREA, next->ds_area_msr);
553 } 560 }
554 561
555 if (next->debugctlmsr != debugctl) 562 if (next->debugctlmsr != debugctl)
556 wrmsrl(MSR_IA32_DEBUGCTLMSR, next->debugctlmsr); 563 update_debugctlmsr(next->debugctlmsr);
557 564
558 if (test_tsk_thread_flag(next_p, TIF_DEBUG)) { 565 if (test_tsk_thread_flag(next_p, TIF_DEBUG)) {
559 loaddebug(next, 0); 566 loaddebug(next, 0);
diff --git a/arch/x86/kernel/ptrace.c b/arch/x86/kernel/ptrace.c
index eb92ccbb3502..559c1b027417 100644
--- a/arch/x86/kernel/ptrace.c
+++ b/arch/x86/kernel/ptrace.c
@@ -1456,7 +1456,6 @@ void send_sigtrap(struct task_struct *tsk, struct pt_regs *regs, int error_code)
1456/* notification of system call entry/exit 1456/* notification of system call entry/exit
1457 * - triggered by current->work.syscall_trace 1457 * - triggered by current->work.syscall_trace
1458 */ 1458 */
1459__attribute__((regparm(3)))
1460int do_syscall_trace(struct pt_regs *regs, int entryexit) 1459int do_syscall_trace(struct pt_regs *regs, int entryexit)
1461{ 1460{
1462 int is_sysemu = test_thread_flag(TIF_SYSCALL_EMU); 1461 int is_sysemu = test_thread_flag(TIF_SYSCALL_EMU);
diff --git a/arch/x86/kernel/reboot.c b/arch/x86/kernel/reboot.c
index 484c4a80d38a..9692202d3bfb 100644
--- a/arch/x86/kernel/reboot.c
+++ b/arch/x86/kernel/reboot.c
@@ -1,5 +1,4 @@
1#include <linux/module.h> 1#include <linux/module.h>
2#include <linux/init.h>
3#include <linux/reboot.h> 2#include <linux/reboot.h>
4#include <linux/init.h> 3#include <linux/init.h>
5#include <linux/pm.h> 4#include <linux/pm.h>
@@ -412,12 +411,12 @@ static void native_machine_shutdown(void)
412#ifdef CONFIG_X86_32 411#ifdef CONFIG_X86_32
413 /* See if there has been given a command line override */ 412 /* See if there has been given a command line override */
414 if ((reboot_cpu != -1) && (reboot_cpu < NR_CPUS) && 413 if ((reboot_cpu != -1) && (reboot_cpu < NR_CPUS) &&
415 cpu_isset(reboot_cpu, cpu_online_map)) 414 cpu_online(reboot_cpu))
416 reboot_cpu_id = reboot_cpu; 415 reboot_cpu_id = reboot_cpu;
417#endif 416#endif
418 417
419 /* Make certain the cpu I'm about to reboot on is online */ 418 /* Make certain the cpu I'm about to reboot on is online */
420 if (!cpu_isset(reboot_cpu_id, cpu_online_map)) 419 if (!cpu_online(reboot_cpu_id))
421 reboot_cpu_id = smp_processor_id(); 420 reboot_cpu_id = smp_processor_id();
422 421
423 /* Make certain I only run on the appropriate processor */ 422 /* Make certain I only run on the appropriate processor */
diff --git a/arch/x86/kernel/relocate_kernel_32.S b/arch/x86/kernel/relocate_kernel_32.S
index f151d6fae462..c30fe25d470d 100644
--- a/arch/x86/kernel/relocate_kernel_32.S
+++ b/arch/x86/kernel/relocate_kernel_32.S
@@ -9,18 +9,19 @@
9#include <linux/linkage.h> 9#include <linux/linkage.h>
10#include <asm/page.h> 10#include <asm/page.h>
11#include <asm/kexec.h> 11#include <asm/kexec.h>
12#include <asm/processor-flags.h>
13#include <asm/pgtable.h>
12 14
13/* 15/*
14 * Must be relocatable PIC code callable as a C function 16 * Must be relocatable PIC code callable as a C function
15 */ 17 */
16 18
17#define PTR(x) (x << 2) 19#define PTR(x) (x << 2)
18#define PAGE_ALIGNED (1 << PAGE_SHIFT) 20#define PAGE_ATTR (_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY)
19#define PAGE_ATTR 0x63 /* _PAGE_PRESENT|_PAGE_RW|_PAGE_ACCESSED|_PAGE_DIRTY */ 21#define PAE_PGD_ATTR (_PAGE_PRESENT)
20#define PAE_PGD_ATTR 0x01 /* _PAGE_PRESENT */
21 22
22 .text 23 .text
23 .align PAGE_ALIGNED 24 .align PAGE_SIZE
24 .globl relocate_kernel 25 .globl relocate_kernel
25relocate_kernel: 26relocate_kernel:
26 movl 8(%esp), %ebp /* list of pages */ 27 movl 8(%esp), %ebp /* list of pages */
@@ -155,7 +156,7 @@ relocate_new_kernel:
155 movl %eax, %cr3 156 movl %eax, %cr3
156 157
157 /* setup a new stack at the end of the physical control page */ 158 /* setup a new stack at the end of the physical control page */
158 lea 4096(%edi), %esp 159 lea PAGE_SIZE(%edi), %esp
159 160
160 /* jump to identity mapped page */ 161 /* jump to identity mapped page */
161 movl %edi, %eax 162 movl %edi, %eax
@@ -168,16 +169,16 @@ identity_mapped:
168 pushl %edx 169 pushl %edx
169 170
170 /* Set cr0 to a known state: 171 /* Set cr0 to a known state:
171 * 31 0 == Paging disabled 172 * - Paging disabled
172 * 18 0 == Alignment check disabled 173 * - Alignment check disabled
173 * 16 0 == Write protect disabled 174 * - Write protect disabled
174 * 3 0 == No task switch 175 * - No task switch
175 * 2 0 == Don't do FP software emulation. 176 * - Don't do FP software emulation.
176 * 0 1 == Proctected mode enabled 177 * - Proctected mode enabled
177 */ 178 */
178 movl %cr0, %eax 179 movl %cr0, %eax
179 andl $~((1<<31)|(1<<18)|(1<<16)|(1<<3)|(1<<2)), %eax 180 andl $~(X86_CR0_PG | X86_CR0_AM | X86_CR0_WP | X86_CR0_TS | X86_CR0_EM), %eax
180 orl $(1<<0), %eax 181 orl $(X86_CR0_PE), %eax
181 movl %eax, %cr0 182 movl %eax, %cr0
182 183
183 /* clear cr4 if applicable */ 184 /* clear cr4 if applicable */
@@ -186,8 +187,7 @@ identity_mapped:
186 /* Set cr4 to a known state: 187 /* Set cr4 to a known state:
187 * Setting everything to zero seems safe. 188 * Setting everything to zero seems safe.
188 */ 189 */
189 movl %cr4, %eax 190 xorl %eax, %eax
190 andl $0, %eax
191 movl %eax, %cr4 191 movl %eax, %cr4
192 192
193 jmp 1f 193 jmp 1f
diff --git a/arch/x86/kernel/relocate_kernel_64.S b/arch/x86/kernel/relocate_kernel_64.S
index 14e95872c6a3..f5afe665a82b 100644
--- a/arch/x86/kernel/relocate_kernel_64.S
+++ b/arch/x86/kernel/relocate_kernel_64.S
@@ -9,17 +9,18 @@
9#include <linux/linkage.h> 9#include <linux/linkage.h>
10#include <asm/page.h> 10#include <asm/page.h>
11#include <asm/kexec.h> 11#include <asm/kexec.h>
12#include <asm/processor-flags.h>
13#include <asm/pgtable.h>
12 14
13/* 15/*
14 * Must be relocatable PIC code callable as a C function 16 * Must be relocatable PIC code callable as a C function
15 */ 17 */
16 18
17#define PTR(x) (x << 3) 19#define PTR(x) (x << 3)
18#define PAGE_ALIGNED (1 << PAGE_SHIFT) 20#define PAGE_ATTR (_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY)
19#define PAGE_ATTR 0x63 /* _PAGE_PRESENT|_PAGE_RW|_PAGE_ACCESSED|_PAGE_DIRTY */
20 21
21 .text 22 .text
22 .align PAGE_ALIGNED 23 .align PAGE_SIZE
23 .code64 24 .code64
24 .globl relocate_kernel 25 .globl relocate_kernel
25relocate_kernel: 26relocate_kernel:
@@ -160,7 +161,7 @@ relocate_new_kernel:
160 movq %r9, %cr3 161 movq %r9, %cr3
161 162
162 /* setup a new stack at the end of the physical control page */ 163 /* setup a new stack at the end of the physical control page */
163 lea 4096(%r8), %rsp 164 lea PAGE_SIZE(%r8), %rsp
164 165
165 /* jump to identity mapped page */ 166 /* jump to identity mapped page */
166 addq $(identity_mapped - relocate_kernel), %r8 167 addq $(identity_mapped - relocate_kernel), %r8
@@ -172,33 +173,22 @@ identity_mapped:
172 pushq %rdx 173 pushq %rdx
173 174
174 /* Set cr0 to a known state: 175 /* Set cr0 to a known state:
175 * 31 1 == Paging enabled 176 * - Paging enabled
176 * 18 0 == Alignment check disabled 177 * - Alignment check disabled
177 * 16 0 == Write protect disabled 178 * - Write protect disabled
178 * 3 0 == No task switch 179 * - No task switch
179 * 2 0 == Don't do FP software emulation. 180 * - Don't do FP software emulation.
180 * 0 1 == Proctected mode enabled 181 * - Proctected mode enabled
181 */ 182 */
182 movq %cr0, %rax 183 movq %cr0, %rax
183 andq $~((1<<18)|(1<<16)|(1<<3)|(1<<2)), %rax 184 andq $~(X86_CR0_AM | X86_CR0_WP | X86_CR0_TS | X86_CR0_EM), %rax
184 orl $((1<<31)|(1<<0)), %eax 185 orl $(X86_CR0_PG | X86_CR0_PE), %eax
185 movq %rax, %cr0 186 movq %rax, %cr0
186 187
187 /* Set cr4 to a known state: 188 /* Set cr4 to a known state:
188 * 10 0 == xmm exceptions disabled 189 * - physical address extension enabled
189 * 9 0 == xmm registers instructions disabled
190 * 8 0 == performance monitoring counter disabled
191 * 7 0 == page global disabled
192 * 6 0 == machine check exceptions disabled
193 * 5 1 == physical address extension enabled
194 * 4 0 == page size extensions disabled
195 * 3 0 == Debug extensions disabled
196 * 2 0 == Time stamp disable (disabled)
197 * 1 0 == Protected mode virtual interrupts disabled
198 * 0 0 == VME disabled
199 */ 190 */
200 191 movq $X86_CR4_PAE, %rax
201 movq $((1<<5)), %rax
202 movq %rax, %cr4 192 movq %rax, %cr4
203 193
204 jmp 1f 194 jmp 1f
diff --git a/arch/x86/kernel/rtc.c b/arch/x86/kernel/rtc.c
index eb9b1a198f5e..9615eee9b775 100644
--- a/arch/x86/kernel/rtc.c
+++ b/arch/x86/kernel/rtc.c
@@ -9,7 +9,6 @@
9#include <asm/vsyscall.h> 9#include <asm/vsyscall.h>
10 10
11#ifdef CONFIG_X86_32 11#ifdef CONFIG_X86_32
12# define CMOS_YEARS_OFFS 1900
13/* 12/*
14 * This is a special lock that is owned by the CPU and holds the index 13 * This is a special lock that is owned by the CPU and holds the index
15 * register we are working with. It is required for NMI access to the 14 * register we are working with. It is required for NMI access to the
@@ -17,14 +16,11 @@
17 */ 16 */
18volatile unsigned long cmos_lock = 0; 17volatile unsigned long cmos_lock = 0;
19EXPORT_SYMBOL(cmos_lock); 18EXPORT_SYMBOL(cmos_lock);
20#else
21/*
22 * x86-64 systems only exists since 2002.
23 * This will work up to Dec 31, 2100
24 */
25# define CMOS_YEARS_OFFS 2000
26#endif 19#endif
27 20
21/* For two digit years assume time is always after that */
22#define CMOS_YEARS_OFFS 2000
23
28DEFINE_SPINLOCK(rtc_lock); 24DEFINE_SPINLOCK(rtc_lock);
29EXPORT_SYMBOL(rtc_lock); 25EXPORT_SYMBOL(rtc_lock);
30 26
@@ -98,7 +94,7 @@ int mach_set_rtc_mmss(unsigned long nowtime)
98 94
99unsigned long mach_get_cmos_time(void) 95unsigned long mach_get_cmos_time(void)
100{ 96{
101 unsigned int year, mon, day, hour, min, sec, century = 0; 97 unsigned int status, year, mon, day, hour, min, sec, century = 0;
102 98
103 /* 99 /*
104 * If UIP is clear, then we have >= 244 microseconds before 100 * If UIP is clear, then we have >= 244 microseconds before
@@ -116,14 +112,16 @@ unsigned long mach_get_cmos_time(void)
116 mon = CMOS_READ(RTC_MONTH); 112 mon = CMOS_READ(RTC_MONTH);
117 year = CMOS_READ(RTC_YEAR); 113 year = CMOS_READ(RTC_YEAR);
118 114
119#if defined(CONFIG_ACPI) && defined(CONFIG_X86_64) 115#ifdef CONFIG_ACPI
120 /* CHECKME: Is this really 64bit only ??? */
121 if (acpi_gbl_FADT.header.revision >= FADT2_REVISION_ID && 116 if (acpi_gbl_FADT.header.revision >= FADT2_REVISION_ID &&
122 acpi_gbl_FADT.century) 117 acpi_gbl_FADT.century)
123 century = CMOS_READ(acpi_gbl_FADT.century); 118 century = CMOS_READ(acpi_gbl_FADT.century);
124#endif 119#endif
125 120
126 if (RTC_ALWAYS_BCD || !(CMOS_READ(RTC_CONTROL) & RTC_DM_BINARY)) { 121 status = CMOS_READ(RTC_CONTROL);
122 WARN_ON_ONCE(RTC_ALWAYS_BCD && (status & RTC_DM_BINARY));
123
124 if (RTC_ALWAYS_BCD || !(status & RTC_DM_BINARY)) {
127 BCD_TO_BIN(sec); 125 BCD_TO_BIN(sec);
128 BCD_TO_BIN(min); 126 BCD_TO_BIN(min);
129 BCD_TO_BIN(hour); 127 BCD_TO_BIN(hour);
@@ -136,11 +134,8 @@ unsigned long mach_get_cmos_time(void)
136 BCD_TO_BIN(century); 134 BCD_TO_BIN(century);
137 year += century * 100; 135 year += century * 100;
138 printk(KERN_INFO "Extended CMOS year: %d\n", century * 100); 136 printk(KERN_INFO "Extended CMOS year: %d\n", century * 100);
139 } else { 137 } else
140 year += CMOS_YEARS_OFFS; 138 year += CMOS_YEARS_OFFS;
141 if (year < 1970)
142 year += 100;
143 }
144 139
145 return mktime(year, mon, day, hour, min, sec); 140 return mktime(year, mon, day, hour, min, sec);
146} 141}
@@ -151,8 +146,8 @@ unsigned char rtc_cmos_read(unsigned char addr)
151 unsigned char val; 146 unsigned char val;
152 147
153 lock_cmos_prefix(addr); 148 lock_cmos_prefix(addr);
154 outb_p(addr, RTC_PORT(0)); 149 outb(addr, RTC_PORT(0));
155 val = inb_p(RTC_PORT(1)); 150 val = inb(RTC_PORT(1));
156 lock_cmos_suffix(addr); 151 lock_cmos_suffix(addr);
157 return val; 152 return val;
158} 153}
@@ -161,8 +156,8 @@ EXPORT_SYMBOL(rtc_cmos_read);
161void rtc_cmos_write(unsigned char val, unsigned char addr) 156void rtc_cmos_write(unsigned char val, unsigned char addr)
162{ 157{
163 lock_cmos_prefix(addr); 158 lock_cmos_prefix(addr);
164 outb_p(addr, RTC_PORT(0)); 159 outb(addr, RTC_PORT(0));
165 outb_p(val, RTC_PORT(1)); 160 outb(val, RTC_PORT(1));
166 lock_cmos_suffix(addr); 161 lock_cmos_suffix(addr);
167} 162}
168EXPORT_SYMBOL(rtc_cmos_write); 163EXPORT_SYMBOL(rtc_cmos_write);
diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c
new file mode 100644
index 000000000000..ed157c90412e
--- /dev/null
+++ b/arch/x86/kernel/setup.c
@@ -0,0 +1,113 @@
1#include <linux/kernel.h>
2#include <linux/module.h>
3#include <linux/init.h>
4#include <linux/bootmem.h>
5#include <linux/percpu.h>
6#include <asm/smp.h>
7#include <asm/percpu.h>
8#include <asm/sections.h>
9#include <asm/processor.h>
10#include <asm/setup.h>
11#include <asm/topology.h>
12#include <asm/mpspec.h>
13#include <asm/apicdef.h>
14
15unsigned int num_processors;
16unsigned disabled_cpus __cpuinitdata;
17/* Processor that is doing the boot up */
18unsigned int boot_cpu_physical_apicid = -1U;
19EXPORT_SYMBOL(boot_cpu_physical_apicid);
20
21physid_mask_t phys_cpu_present_map;
22
23DEFINE_PER_CPU(u16, x86_cpu_to_apicid) = BAD_APICID;
24EXPORT_PER_CPU_SYMBOL(x86_cpu_to_apicid);
25
26/* Bitmask of physically existing CPUs */
27physid_mask_t phys_cpu_present_map;
28
29#if defined(CONFIG_HAVE_SETUP_PER_CPU_AREA) && defined(CONFIG_SMP)
30/*
31 * Copy data used in early init routines from the initial arrays to the
32 * per cpu data areas. These arrays then become expendable and the
33 * *_early_ptr's are zeroed indicating that the static arrays are gone.
34 */
35static void __init setup_per_cpu_maps(void)
36{
37 int cpu;
38
39 for_each_possible_cpu(cpu) {
40 per_cpu(x86_cpu_to_apicid, cpu) = x86_cpu_to_apicid_init[cpu];
41 per_cpu(x86_bios_cpu_apicid, cpu) =
42 x86_bios_cpu_apicid_init[cpu];
43#ifdef CONFIG_NUMA
44 per_cpu(x86_cpu_to_node_map, cpu) =
45 x86_cpu_to_node_map_init[cpu];
46#endif
47 }
48
49 /* indicate the early static arrays will soon be gone */
50 x86_cpu_to_apicid_early_ptr = NULL;
51 x86_bios_cpu_apicid_early_ptr = NULL;
52#ifdef CONFIG_NUMA
53 x86_cpu_to_node_map_early_ptr = NULL;
54#endif
55}
56
57#ifdef CONFIG_X86_32
58/*
59 * Great future not-so-futuristic plan: make i386 and x86_64 do it
60 * the same way
61 */
62unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
63EXPORT_SYMBOL(__per_cpu_offset);
64#endif
65
66/*
67 * Great future plan:
68 * Declare PDA itself and support (irqstack,tss,pgd) as per cpu data.
69 * Always point %gs to its beginning
70 */
71void __init setup_per_cpu_areas(void)
72{
73 int i;
74 unsigned long size;
75
76#ifdef CONFIG_HOTPLUG_CPU
77 prefill_possible_map();
78#endif
79
80 /* Copy section for each CPU (we discard the original) */
81 size = PERCPU_ENOUGH_ROOM;
82 printk(KERN_INFO "PERCPU: Allocating %lu bytes of per cpu data\n",
83 size);
84
85 for_each_possible_cpu(i) {
86 char *ptr;
87#ifndef CONFIG_NEED_MULTIPLE_NODES
88 ptr = alloc_bootmem_pages(size);
89#else
90 int node = early_cpu_to_node(i);
91 if (!node_online(node) || !NODE_DATA(node)) {
92 ptr = alloc_bootmem_pages(size);
93 printk(KERN_INFO
94 "cpu %d has no node or node-local memory\n", i);
95 }
96 else
97 ptr = alloc_bootmem_pages_node(NODE_DATA(node), size);
98#endif
99 if (!ptr)
100 panic("Cannot allocate cpu data for CPU %d\n", i);
101#ifdef CONFIG_X86_64
102 cpu_pda(i)->data_offset = ptr - __per_cpu_start;
103#else
104 __per_cpu_offset[i] = ptr - __per_cpu_start;
105#endif
106 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
107 }
108
109 /* Setup percpu data maps */
110 setup_per_cpu_maps();
111}
112
113#endif
diff --git a/arch/x86/kernel/setup64.c b/arch/x86/kernel/setup64.c
index 143aa78c566b..9042fb0e36f5 100644
--- a/arch/x86/kernel/setup64.c
+++ b/arch/x86/kernel/setup64.c
@@ -24,6 +24,7 @@
24#include <asm/proto.h> 24#include <asm/proto.h>
25#include <asm/sections.h> 25#include <asm/sections.h>
26#include <asm/setup.h> 26#include <asm/setup.h>
27#include <asm/genapic.h>
27 28
28#ifndef CONFIG_DEBUG_BOOT_PARAMS 29#ifndef CONFIG_DEBUG_BOOT_PARAMS
29struct boot_params __initdata boot_params; 30struct boot_params __initdata boot_params;
@@ -86,83 +87,6 @@ static int __init nonx32_setup(char *str)
86} 87}
87__setup("noexec32=", nonx32_setup); 88__setup("noexec32=", nonx32_setup);
88 89
89/*
90 * Copy data used in early init routines from the initial arrays to the
91 * per cpu data areas. These arrays then become expendable and the
92 * *_early_ptr's are zeroed indicating that the static arrays are gone.
93 */
94static void __init setup_per_cpu_maps(void)
95{
96 int cpu;
97
98 for_each_possible_cpu(cpu) {
99#ifdef CONFIG_SMP
100 if (per_cpu_offset(cpu)) {
101#endif
102 per_cpu(x86_cpu_to_apicid, cpu) =
103 x86_cpu_to_apicid_init[cpu];
104 per_cpu(x86_bios_cpu_apicid, cpu) =
105 x86_bios_cpu_apicid_init[cpu];
106#ifdef CONFIG_NUMA
107 per_cpu(x86_cpu_to_node_map, cpu) =
108 x86_cpu_to_node_map_init[cpu];
109#endif
110#ifdef CONFIG_SMP
111 }
112 else
113 printk(KERN_NOTICE "per_cpu_offset zero for cpu %d\n",
114 cpu);
115#endif
116 }
117
118 /* indicate the early static arrays will soon be gone */
119 x86_cpu_to_apicid_early_ptr = NULL;
120 x86_bios_cpu_apicid_early_ptr = NULL;
121#ifdef CONFIG_NUMA
122 x86_cpu_to_node_map_early_ptr = NULL;
123#endif
124}
125
126/*
127 * Great future plan:
128 * Declare PDA itself and support (irqstack,tss,pgd) as per cpu data.
129 * Always point %gs to its beginning
130 */
131void __init setup_per_cpu_areas(void)
132{
133 int i;
134 unsigned long size;
135
136#ifdef CONFIG_HOTPLUG_CPU
137 prefill_possible_map();
138#endif
139
140 /* Copy section for each CPU (we discard the original) */
141 size = PERCPU_ENOUGH_ROOM;
142
143 printk(KERN_INFO "PERCPU: Allocating %lu bytes of per cpu data\n", size);
144 for_each_cpu_mask (i, cpu_possible_map) {
145 char *ptr;
146#ifndef CONFIG_NEED_MULTIPLE_NODES
147 ptr = alloc_bootmem_pages(size);
148#else
149 int node = early_cpu_to_node(i);
150
151 if (!node_online(node) || !NODE_DATA(node))
152 ptr = alloc_bootmem_pages(size);
153 else
154 ptr = alloc_bootmem_pages_node(NODE_DATA(node), size);
155#endif
156 if (!ptr)
157 panic("Cannot allocate cpu data for CPU %d\n", i);
158 cpu_pda(i)->data_offset = ptr - __per_cpu_start;
159 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
160 }
161
162 /* setup percpu data maps early */
163 setup_per_cpu_maps();
164}
165
166void pda_init(int cpu) 90void pda_init(int cpu)
167{ 91{
168 struct x8664_pda *pda = cpu_pda(cpu); 92 struct x8664_pda *pda = cpu_pda(cpu);
@@ -357,4 +281,7 @@ void __cpuinit cpu_init (void)
357 fpu_init(); 281 fpu_init();
358 282
359 raw_local_save_flags(kernel_eflags); 283 raw_local_save_flags(kernel_eflags);
284
285 if (is_uv_system())
286 uv_cpu_init();
360} 287}
diff --git a/arch/x86/kernel/setup_32.c b/arch/x86/kernel/setup_32.c
index 2b3e5d45176b..5b0bffb7fcc9 100644
--- a/arch/x86/kernel/setup_32.c
+++ b/arch/x86/kernel/setup_32.c
@@ -62,8 +62,9 @@
62#include <asm/io.h> 62#include <asm/io.h>
63#include <asm/vmi.h> 63#include <asm/vmi.h>
64#include <setup_arch.h> 64#include <setup_arch.h>
65#include <bios_ebda.h> 65#include <asm/bios_ebda.h>
66#include <asm/cacheflush.h> 66#include <asm/cacheflush.h>
67#include <asm/processor.h>
67 68
68/* This value is set up by the early boot code to point to the value 69/* This value is set up by the early boot code to point to the value
69 immediately after the boot time page tables. It contains a *physical* 70 immediately after the boot time page tables. It contains a *physical*
@@ -154,6 +155,8 @@ struct cpuinfo_x86 new_cpu_data __cpuinitdata = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
154struct cpuinfo_x86 boot_cpu_data __read_mostly = { 0, 0, 0, 0, -1, 1, 0, 0, -1 }; 155struct cpuinfo_x86 boot_cpu_data __read_mostly = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
155EXPORT_SYMBOL(boot_cpu_data); 156EXPORT_SYMBOL(boot_cpu_data);
156 157
158unsigned int def_to_bigsmp;
159
157#ifndef CONFIG_X86_PAE 160#ifndef CONFIG_X86_PAE
158unsigned long mmu_cr4_features; 161unsigned long mmu_cr4_features;
159#else 162#else
@@ -189,7 +192,7 @@ EXPORT_SYMBOL(ist_info);
189extern void early_cpu_init(void); 192extern void early_cpu_init(void);
190extern int root_mountflags; 193extern int root_mountflags;
191 194
192unsigned long saved_videomode; 195unsigned long saved_video_mode;
193 196
194#define RAMDISK_IMAGE_START_MASK 0x07FF 197#define RAMDISK_IMAGE_START_MASK 0x07FF
195#define RAMDISK_PROMPT_FLAG 0x8000 198#define RAMDISK_PROMPT_FLAG 0x8000
@@ -227,7 +230,7 @@ static inline void copy_edd(void)
227} 230}
228#endif 231#endif
229 232
230int __initdata user_defined_memmap = 0; 233int __initdata user_defined_memmap;
231 234
232/* 235/*
233 * "mem=nopentium" disables the 4MB page tables. 236 * "mem=nopentium" disables the 4MB page tables.
@@ -385,15 +388,56 @@ unsigned long __init find_max_low_pfn(void)
385 return max_low_pfn; 388 return max_low_pfn;
386} 389}
387 390
391#define BIOS_EBDA_SEGMENT 0x40E
392#define BIOS_LOWMEM_KILOBYTES 0x413
393
388/* 394/*
389 * workaround for Dell systems that neglect to reserve EBDA 395 * The BIOS places the EBDA/XBDA at the top of conventional
396 * memory, and usually decreases the reported amount of
397 * conventional memory (int 0x12) too. This also contains a
398 * workaround for Dell systems that neglect to reserve EBDA.
399 * The same workaround also avoids a problem with the AMD768MPX
400 * chipset: reserve a page before VGA to prevent PCI prefetch
401 * into it (errata #56). Usually the page is reserved anyways,
402 * unless you have no PS/2 mouse plugged in.
390 */ 403 */
391static void __init reserve_ebda_region(void) 404static void __init reserve_ebda_region(void)
392{ 405{
393 unsigned int addr; 406 unsigned int lowmem, ebda_addr;
394 addr = get_bios_ebda(); 407
395 if (addr) 408 /* To determine the position of the EBDA and the */
396 reserve_bootmem(addr, PAGE_SIZE, BOOTMEM_DEFAULT); 409 /* end of conventional memory, we need to look at */
410 /* the BIOS data area. In a paravirtual environment */
411 /* that area is absent. We'll just have to assume */
412 /* that the paravirt case can handle memory setup */
413 /* correctly, without our help. */
414 if (paravirt_enabled())
415 return;
416
417 /* end of low (conventional) memory */
418 lowmem = *(unsigned short *)__va(BIOS_LOWMEM_KILOBYTES);
419 lowmem <<= 10;
420
421 /* start of EBDA area */
422 ebda_addr = *(unsigned short *)__va(BIOS_EBDA_SEGMENT);
423 ebda_addr <<= 4;
424
425 /* Fixup: bios puts an EBDA in the top 64K segment */
426 /* of conventional memory, but does not adjust lowmem. */
427 if ((lowmem - ebda_addr) <= 0x10000)
428 lowmem = ebda_addr;
429
430 /* Fixup: bios does not report an EBDA at all. */
431 /* Some old Dells seem to need 4k anyhow (bugzilla 2990) */
432 if ((ebda_addr == 0) && (lowmem >= 0x9f000))
433 lowmem = 0x9f000;
434
435 /* Paranoia: should never happen, but... */
436 if ((lowmem == 0) || (lowmem >= 0x100000))
437 lowmem = 0x9f000;
438
439 /* reserve all memory between lowmem and the 1MB mark */
440 reserve_bootmem(lowmem, 0x100000 - lowmem, BOOTMEM_DEFAULT);
397} 441}
398 442
399#ifndef CONFIG_NEED_MULTIPLE_NODES 443#ifndef CONFIG_NEED_MULTIPLE_NODES
@@ -617,16 +661,9 @@ void __init setup_bootmem_allocator(void)
617 */ 661 */
618 reserve_bootmem(0, PAGE_SIZE, BOOTMEM_DEFAULT); 662 reserve_bootmem(0, PAGE_SIZE, BOOTMEM_DEFAULT);
619 663
620 /* reserve EBDA region, it's a 4K region */ 664 /* reserve EBDA region */
621 reserve_ebda_region(); 665 reserve_ebda_region();
622 666
623 /* could be an AMD 768MPX chipset. Reserve a page before VGA to prevent
624 PCI prefetch into it (errata #56). Usually the page is reserved anyways,
625 unless you have no PS/2 mouse plugged in. */
626 if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD &&
627 boot_cpu_data.x86 == 6)
628 reserve_bootmem(0xa0000 - 4096, 4096, BOOTMEM_DEFAULT);
629
630#ifdef CONFIG_SMP 667#ifdef CONFIG_SMP
631 /* 668 /*
632 * But first pinch a few for the stack/trampoline stuff 669 * But first pinch a few for the stack/trampoline stuff
@@ -687,6 +724,18 @@ char * __init __attribute__((weak)) memory_setup(void)
687 return machine_specific_memory_setup(); 724 return machine_specific_memory_setup();
688} 725}
689 726
727#ifdef CONFIG_NUMA
728/*
729 * In the golden day, when everything among i386 and x86_64 will be
730 * integrated, this will not live here
731 */
732void *x86_cpu_to_node_map_early_ptr;
733int x86_cpu_to_node_map_init[NR_CPUS] = {
734 [0 ... NR_CPUS-1] = NUMA_NO_NODE
735};
736DEFINE_PER_CPU(int, x86_cpu_to_node_map) = NUMA_NO_NODE;
737#endif
738
690/* 739/*
691 * Determine if we were loaded by an EFI loader. If so, then we have also been 740 * Determine if we were loaded by an EFI loader. If so, then we have also been
692 * passed the efi memmap, systab, etc., so we should use these data structures 741 * passed the efi memmap, systab, etc., so we should use these data structures
@@ -714,7 +763,7 @@ void __init setup_arch(char **cmdline_p)
714 edid_info = boot_params.edid_info; 763 edid_info = boot_params.edid_info;
715 apm_info.bios = boot_params.apm_bios_info; 764 apm_info.bios = boot_params.apm_bios_info;
716 ist_info = boot_params.ist_info; 765 ist_info = boot_params.ist_info;
717 saved_videomode = boot_params.hdr.vid_mode; 766 saved_video_mode = boot_params.hdr.vid_mode;
718 if( boot_params.sys_desc_table.length != 0 ) { 767 if( boot_params.sys_desc_table.length != 0 ) {
719 set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2); 768 set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2);
720 machine_id = boot_params.sys_desc_table.table[0]; 769 machine_id = boot_params.sys_desc_table.table[0];
@@ -820,6 +869,18 @@ void __init setup_arch(char **cmdline_p)
820 869
821 io_delay_init(); 870 io_delay_init();
822 871
872#ifdef CONFIG_X86_SMP
873 /*
874 * setup to use the early static init tables during kernel startup
875 * X86_SMP will exclude sub-arches that don't deal well with it.
876 */
877 x86_cpu_to_apicid_early_ptr = (void *)x86_cpu_to_apicid_init;
878 x86_bios_cpu_apicid_early_ptr = (void *)x86_bios_cpu_apicid_init;
879#ifdef CONFIG_NUMA
880 x86_cpu_to_node_map_early_ptr = (void *)x86_cpu_to_node_map_init;
881#endif
882#endif
883
823#ifdef CONFIG_X86_GENERICARCH 884#ifdef CONFIG_X86_GENERICARCH
824 generic_apic_probe(); 885 generic_apic_probe();
825#endif 886#endif
diff --git a/arch/x86/kernel/setup_64.c b/arch/x86/kernel/setup_64.c
index f4f7ecfb898c..674ef3510cdf 100644
--- a/arch/x86/kernel/setup_64.c
+++ b/arch/x86/kernel/setup_64.c
@@ -58,7 +58,6 @@
58#include <asm/mmu_context.h> 58#include <asm/mmu_context.h>
59#include <asm/proto.h> 59#include <asm/proto.h>
60#include <asm/setup.h> 60#include <asm/setup.h>
61#include <asm/mach_apic.h>
62#include <asm/numa.h> 61#include <asm/numa.h>
63#include <asm/sections.h> 62#include <asm/sections.h>
64#include <asm/dmi.h> 63#include <asm/dmi.h>
@@ -66,7 +65,9 @@
66#include <asm/mce.h> 65#include <asm/mce.h>
67#include <asm/ds.h> 66#include <asm/ds.h>
68#include <asm/topology.h> 67#include <asm/topology.h>
68#include <asm/trampoline.h>
69 69
70#include <mach_apic.h>
70#ifdef CONFIG_PARAVIRT 71#ifdef CONFIG_PARAVIRT
71#include <asm/paravirt.h> 72#include <asm/paravirt.h>
72#else 73#else
@@ -248,6 +249,7 @@ static void __init reserve_crashkernel(void)
248 (unsigned long)(total_mem >> 20)); 249 (unsigned long)(total_mem >> 20));
249 crashk_res.start = crash_base; 250 crashk_res.start = crash_base;
250 crashk_res.end = crash_base + crash_size - 1; 251 crashk_res.end = crash_base + crash_size - 1;
252 insert_resource(&iomem_resource, &crashk_res);
251 } 253 }
252} 254}
253#else 255#else
@@ -322,6 +324,11 @@ void __init setup_arch(char **cmdline_p)
322 324
323 finish_e820_parsing(); 325 finish_e820_parsing();
324 326
327 /* after parse_early_param, so could debug it */
328 insert_resource(&iomem_resource, &code_resource);
329 insert_resource(&iomem_resource, &data_resource);
330 insert_resource(&iomem_resource, &bss_resource);
331
325 early_gart_iommu_check(); 332 early_gart_iommu_check();
326 333
327 e820_register_active_regions(0, 0, -1UL); 334 e820_register_active_regions(0, 0, -1UL);
@@ -341,10 +348,12 @@ void __init setup_arch(char **cmdline_p)
341 348
342 check_efer(); 349 check_efer();
343 350
344 init_memory_mapping(0, (end_pfn_map << PAGE_SHIFT)); 351 max_pfn_mapped = init_memory_mapping(0, (max_pfn_mapped << PAGE_SHIFT));
345 if (efi_enabled) 352 if (efi_enabled)
346 efi_init(); 353 efi_init();
347 354
355 vsmp_init();
356
348 dmi_scan_machine(); 357 dmi_scan_machine();
349 358
350 io_delay_init(); 359 io_delay_init();
@@ -450,7 +459,7 @@ void __init setup_arch(char **cmdline_p)
450 /* 459 /*
451 * We trust e820 completely. No explicit ROM probing in memory. 460 * We trust e820 completely. No explicit ROM probing in memory.
452 */ 461 */
453 e820_reserve_resources(&code_resource, &data_resource, &bss_resource); 462 e820_reserve_resources();
454 e820_mark_nosave_regions(); 463 e820_mark_nosave_regions();
455 464
456 /* request I/O space for devices used on all i[345]86 PCs */ 465 /* request I/O space for devices used on all i[345]86 PCs */
@@ -552,9 +561,9 @@ static void __cpuinit amd_detect_cmp(struct cpuinfo_x86 *c)
552 bits = c->x86_coreid_bits; 561 bits = c->x86_coreid_bits;
553 562
554 /* Low order bits define the core id (index of core in socket) */ 563 /* Low order bits define the core id (index of core in socket) */
555 c->cpu_core_id = c->phys_proc_id & ((1 << bits)-1); 564 c->cpu_core_id = c->initial_apicid & ((1 << bits)-1);
556 /* Convert the APIC ID into the socket ID */ 565 /* Convert the initial APIC ID into the socket ID */
557 c->phys_proc_id = phys_pkg_id(bits); 566 c->phys_proc_id = c->initial_apicid >> bits;
558 567
559#ifdef CONFIG_NUMA 568#ifdef CONFIG_NUMA
560 node = c->phys_proc_id; 569 node = c->phys_proc_id;
@@ -571,7 +580,7 @@ static void __cpuinit amd_detect_cmp(struct cpuinfo_x86 *c)
571 If that doesn't result in a usable node fall back to the 580 If that doesn't result in a usable node fall back to the
572 path for the previous case. */ 581 path for the previous case. */
573 582
574 int ht_nodeid = apicid - (cpu_data(0).phys_proc_id << bits); 583 int ht_nodeid = c->initial_apicid;
575 584
576 if (ht_nodeid >= 0 && 585 if (ht_nodeid >= 0 &&
577 apicid_to_node[ht_nodeid] != NUMA_NO_NODE) 586 apicid_to_node[ht_nodeid] != NUMA_NO_NODE)
@@ -677,7 +686,7 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c)
677 686
678 /* Bit 31 in normal CPUID used for nonstandard 3DNow ID; 687 /* Bit 31 in normal CPUID used for nonstandard 3DNow ID;
679 3DNow is IDd by bit 31 in extended CPUID (1*32+31) anyway */ 688 3DNow is IDd by bit 31 in extended CPUID (1*32+31) anyway */
680 clear_bit(0*32+31, (unsigned long *)&c->x86_capability); 689 clear_cpu_cap(c, 0*32+31);
681 690
682 /* On C+ stepping K8 rep microcode works well for copy/memset */ 691 /* On C+ stepping K8 rep microcode works well for copy/memset */
683 level = cpuid_eax(1); 692 level = cpuid_eax(1);
@@ -721,6 +730,19 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c)
721 730
722 if (amd_apic_timer_broken()) 731 if (amd_apic_timer_broken())
723 disable_apic_timer = 1; 732 disable_apic_timer = 1;
733
734 if (c == &boot_cpu_data && c->x86 >= 0xf && c->x86 <= 0x11) {
735 unsigned long long tseg;
736
737 /*
738 * Split up direct mapping around the TSEG SMM area.
739 * Don't do it for gbpages because there seems very little
740 * benefit in doing so.
741 */
742 if (!rdmsrl_safe(MSR_K8_TSEG_ADDR, &tseg) &&
743 (tseg >> PMD_SHIFT) < (max_pfn_mapped >> (PMD_SHIFT-PAGE_SHIFT)))
744 set_memory_4k((unsigned long)__va(tseg), 1);
745 }
724} 746}
725 747
726void __cpuinit detect_ht(struct cpuinfo_x86 *c) 748void __cpuinit detect_ht(struct cpuinfo_x86 *c)
@@ -813,7 +835,7 @@ static void __cpuinit early_init_intel(struct cpuinfo_x86 *c)
813{ 835{
814 if ((c->x86 == 0xf && c->x86_model >= 0x03) || 836 if ((c->x86 == 0xf && c->x86_model >= 0x03) ||
815 (c->x86 == 0x6 && c->x86_model >= 0x0e)) 837 (c->x86 == 0x6 && c->x86_model >= 0x0e))
816 set_bit(X86_FEATURE_CONSTANT_TSC, &c->x86_capability); 838 set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
817} 839}
818 840
819static void __cpuinit init_intel(struct cpuinfo_x86 *c) 841static void __cpuinit init_intel(struct cpuinfo_x86 *c)
@@ -856,9 +878,6 @@ static void __cpuinit init_intel(struct cpuinfo_x86 *c)
856 878
857 if (c->x86 == 15) 879 if (c->x86 == 15)
858 c->x86_cache_alignment = c->x86_clflush_size * 2; 880 c->x86_cache_alignment = c->x86_clflush_size * 2;
859 if ((c->x86 == 0xf && c->x86_model >= 0x03) ||
860 (c->x86 == 0x6 && c->x86_model >= 0x0e))
861 set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
862 if (c->x86 == 6) 881 if (c->x86 == 6)
863 set_cpu_cap(c, X86_FEATURE_REP_GOOD); 882 set_cpu_cap(c, X86_FEATURE_REP_GOOD);
864 set_cpu_cap(c, X86_FEATURE_LFENCE_RDTSC); 883 set_cpu_cap(c, X86_FEATURE_LFENCE_RDTSC);
@@ -867,6 +886,32 @@ static void __cpuinit init_intel(struct cpuinfo_x86 *c)
867 srat_detect_node(); 886 srat_detect_node();
868} 887}
869 888
889static void __cpuinit early_init_centaur(struct cpuinfo_x86 *c)
890{
891 if (c->x86 == 0x6 && c->x86_model >= 0xf)
892 set_bit(X86_FEATURE_CONSTANT_TSC, &c->x86_capability);
893}
894
895static void __cpuinit init_centaur(struct cpuinfo_x86 *c)
896{
897 /* Cache sizes */
898 unsigned n;
899
900 n = c->extended_cpuid_level;
901 if (n >= 0x80000008) {
902 unsigned eax = cpuid_eax(0x80000008);
903 c->x86_virt_bits = (eax >> 8) & 0xff;
904 c->x86_phys_bits = eax & 0xff;
905 }
906
907 if (c->x86 == 0x6 && c->x86_model >= 0xf) {
908 c->x86_cache_alignment = c->x86_clflush_size * 2;
909 set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
910 set_cpu_cap(c, X86_FEATURE_REP_GOOD);
911 }
912 set_cpu_cap(c, X86_FEATURE_LFENCE_RDTSC);
913}
914
870static void __cpuinit get_cpu_vendor(struct cpuinfo_x86 *c) 915static void __cpuinit get_cpu_vendor(struct cpuinfo_x86 *c)
871{ 916{
872 char *v = c->x86_vendor_id; 917 char *v = c->x86_vendor_id;
@@ -875,6 +920,8 @@ static void __cpuinit get_cpu_vendor(struct cpuinfo_x86 *c)
875 c->x86_vendor = X86_VENDOR_AMD; 920 c->x86_vendor = X86_VENDOR_AMD;
876 else if (!strcmp(v, "GenuineIntel")) 921 else if (!strcmp(v, "GenuineIntel"))
877 c->x86_vendor = X86_VENDOR_INTEL; 922 c->x86_vendor = X86_VENDOR_INTEL;
923 else if (!strcmp(v, "CentaurHauls"))
924 c->x86_vendor = X86_VENDOR_CENTAUR;
878 else 925 else
879 c->x86_vendor = X86_VENDOR_UNKNOWN; 926 c->x86_vendor = X86_VENDOR_UNKNOWN;
880} 927}
@@ -922,15 +969,16 @@ static void __cpuinit early_identify_cpu(struct cpuinfo_x86 *c)
922 c->x86 += (tfms >> 20) & 0xff; 969 c->x86 += (tfms >> 20) & 0xff;
923 if (c->x86 >= 0x6) 970 if (c->x86 >= 0x6)
924 c->x86_model += ((tfms >> 16) & 0xF) << 4; 971 c->x86_model += ((tfms >> 16) & 0xF) << 4;
925 if (c->x86_capability[0] & (1<<19)) 972 if (test_cpu_cap(c, X86_FEATURE_CLFLSH))
926 c->x86_clflush_size = ((misc >> 8) & 0xff) * 8; 973 c->x86_clflush_size = ((misc >> 8) & 0xff) * 8;
927 } else { 974 } else {
928 /* Have CPUID level 0 only - unheard of */ 975 /* Have CPUID level 0 only - unheard of */
929 c->x86 = 4; 976 c->x86 = 4;
930 } 977 }
931 978
979 c->initial_apicid = (cpuid_ebx(1) >> 24) & 0xff;
932#ifdef CONFIG_SMP 980#ifdef CONFIG_SMP
933 c->phys_proc_id = (cpuid_ebx(1) >> 24) & 0xff; 981 c->phys_proc_id = c->initial_apicid;
934#endif 982#endif
935 /* AMD-defined flags: level 0x80000001 */ 983 /* AMD-defined flags: level 0x80000001 */
936 xlvl = cpuid_eax(0x80000000); 984 xlvl = cpuid_eax(0x80000000);
@@ -956,12 +1004,22 @@ static void __cpuinit early_identify_cpu(struct cpuinfo_x86 *c)
956 if (c->extended_cpuid_level >= 0x80000007) 1004 if (c->extended_cpuid_level >= 0x80000007)
957 c->x86_power = cpuid_edx(0x80000007); 1005 c->x86_power = cpuid_edx(0x80000007);
958 1006
1007
1008 clear_cpu_cap(c, X86_FEATURE_PAT);
1009
959 switch (c->x86_vendor) { 1010 switch (c->x86_vendor) {
960 case X86_VENDOR_AMD: 1011 case X86_VENDOR_AMD:
961 early_init_amd(c); 1012 early_init_amd(c);
1013 if (c->x86 >= 0xf && c->x86 <= 0x11)
1014 set_cpu_cap(c, X86_FEATURE_PAT);
962 break; 1015 break;
963 case X86_VENDOR_INTEL: 1016 case X86_VENDOR_INTEL:
964 early_init_intel(c); 1017 early_init_intel(c);
1018 if (c->x86 == 0xF || (c->x86 == 6 && c->x86_model >= 15))
1019 set_cpu_cap(c, X86_FEATURE_PAT);
1020 break;
1021 case X86_VENDOR_CENTAUR:
1022 early_init_centaur(c);
965 break; 1023 break;
966 } 1024 }
967 1025
@@ -999,6 +1057,10 @@ void __cpuinit identify_cpu(struct cpuinfo_x86 *c)
999 init_intel(c); 1057 init_intel(c);
1000 break; 1058 break;
1001 1059
1060 case X86_VENDOR_CENTAUR:
1061 init_centaur(c);
1062 break;
1063
1002 case X86_VENDOR_UNKNOWN: 1064 case X86_VENDOR_UNKNOWN:
1003 default: 1065 default:
1004 display_cacheinfo(c); 1066 display_cacheinfo(c);
@@ -1028,14 +1090,24 @@ void __cpuinit identify_cpu(struct cpuinfo_x86 *c)
1028#endif 1090#endif
1029 select_idle_routine(c); 1091 select_idle_routine(c);
1030 1092
1031 if (c != &boot_cpu_data)
1032 mtrr_ap_init();
1033#ifdef CONFIG_NUMA 1093#ifdef CONFIG_NUMA
1034 numa_add_cpu(smp_processor_id()); 1094 numa_add_cpu(smp_processor_id());
1035#endif 1095#endif
1036 1096
1037} 1097}
1038 1098
1099void __cpuinit identify_boot_cpu(void)
1100{
1101 identify_cpu(&boot_cpu_data);
1102}
1103
1104void __cpuinit identify_secondary_cpu(struct cpuinfo_x86 *c)
1105{
1106 BUG_ON(c == &boot_cpu_data);
1107 identify_cpu(c);
1108 mtrr_ap_init();
1109}
1110
1039static __init int setup_noclflush(char *arg) 1111static __init int setup_noclflush(char *arg)
1040{ 1112{
1041 setup_clear_cpu_cap(X86_FEATURE_CLFLSH); 1113 setup_clear_cpu_cap(X86_FEATURE_CLFLSH);
@@ -1064,123 +1136,3 @@ static __init int setup_disablecpuid(char *arg)
1064 return 1; 1136 return 1;
1065} 1137}
1066__setup("clearcpuid=", setup_disablecpuid); 1138__setup("clearcpuid=", setup_disablecpuid);
1067
1068/*
1069 * Get CPU information for use by the procfs.
1070 */
1071
1072static int show_cpuinfo(struct seq_file *m, void *v)
1073{
1074 struct cpuinfo_x86 *c = v;
1075 int cpu = 0, i;
1076
1077#ifdef CONFIG_SMP
1078 cpu = c->cpu_index;
1079#endif
1080
1081 seq_printf(m, "processor\t: %u\n"
1082 "vendor_id\t: %s\n"
1083 "cpu family\t: %d\n"
1084 "model\t\t: %d\n"
1085 "model name\t: %s\n",
1086 (unsigned)cpu,
1087 c->x86_vendor_id[0] ? c->x86_vendor_id : "unknown",
1088 c->x86,
1089 (int)c->x86_model,
1090 c->x86_model_id[0] ? c->x86_model_id : "unknown");
1091
1092 if (c->x86_mask || c->cpuid_level >= 0)
1093 seq_printf(m, "stepping\t: %d\n", c->x86_mask);
1094 else
1095 seq_printf(m, "stepping\t: unknown\n");
1096
1097 if (cpu_has(c, X86_FEATURE_TSC)) {
1098 unsigned int freq = cpufreq_quick_get((unsigned)cpu);
1099
1100 if (!freq)
1101 freq = cpu_khz;
1102 seq_printf(m, "cpu MHz\t\t: %u.%03u\n",
1103 freq / 1000, (freq % 1000));
1104 }
1105
1106 /* Cache size */
1107 if (c->x86_cache_size >= 0)
1108 seq_printf(m, "cache size\t: %d KB\n", c->x86_cache_size);
1109
1110#ifdef CONFIG_SMP
1111 if (smp_num_siblings * c->x86_max_cores > 1) {
1112 seq_printf(m, "physical id\t: %d\n", c->phys_proc_id);
1113 seq_printf(m, "siblings\t: %d\n",
1114 cpus_weight(per_cpu(cpu_core_map, cpu)));
1115 seq_printf(m, "core id\t\t: %d\n", c->cpu_core_id);
1116 seq_printf(m, "cpu cores\t: %d\n", c->booted_cores);
1117 }
1118#endif
1119
1120 seq_printf(m,
1121 "fpu\t\t: yes\n"
1122 "fpu_exception\t: yes\n"
1123 "cpuid level\t: %d\n"
1124 "wp\t\t: yes\n"
1125 "flags\t\t:",
1126 c->cpuid_level);
1127
1128 for (i = 0; i < 32*NCAPINTS; i++)
1129 if (cpu_has(c, i) && x86_cap_flags[i] != NULL)
1130 seq_printf(m, " %s", x86_cap_flags[i]);
1131
1132 seq_printf(m, "\nbogomips\t: %lu.%02lu\n",
1133 c->loops_per_jiffy/(500000/HZ),
1134 (c->loops_per_jiffy/(5000/HZ)) % 100);
1135
1136 if (c->x86_tlbsize > 0)
1137 seq_printf(m, "TLB size\t: %d 4K pages\n", c->x86_tlbsize);
1138 seq_printf(m, "clflush size\t: %d\n", c->x86_clflush_size);
1139 seq_printf(m, "cache_alignment\t: %d\n", c->x86_cache_alignment);
1140
1141 seq_printf(m, "address sizes\t: %u bits physical, %u bits virtual\n",
1142 c->x86_phys_bits, c->x86_virt_bits);
1143
1144 seq_printf(m, "power management:");
1145 for (i = 0; i < 32; i++) {
1146 if (c->x86_power & (1 << i)) {
1147 if (i < ARRAY_SIZE(x86_power_flags) &&
1148 x86_power_flags[i])
1149 seq_printf(m, "%s%s",
1150 x86_power_flags[i][0]?" ":"",
1151 x86_power_flags[i]);
1152 else
1153 seq_printf(m, " [%d]", i);
1154 }
1155 }
1156
1157 seq_printf(m, "\n\n");
1158
1159 return 0;
1160}
1161
1162static void *c_start(struct seq_file *m, loff_t *pos)
1163{
1164 if (*pos == 0) /* just in case, cpu 0 is not the first */
1165 *pos = first_cpu(cpu_online_map);
1166 if ((*pos) < NR_CPUS && cpu_online(*pos))
1167 return &cpu_data(*pos);
1168 return NULL;
1169}
1170
1171static void *c_next(struct seq_file *m, void *v, loff_t *pos)
1172{
1173 *pos = next_cpu(*pos, cpu_online_map);
1174 return c_start(m, pos);
1175}
1176
1177static void c_stop(struct seq_file *m, void *v)
1178{
1179}
1180
1181const struct seq_operations cpuinfo_op = {
1182 .start = c_start,
1183 .next = c_next,
1184 .stop = c_stop,
1185 .show = show_cpuinfo,
1186};
diff --git a/arch/x86/kernel/sigframe_32.h b/arch/x86/kernel/sigframe.h
index 0b2221711dad..72bbb519d2dc 100644
--- a/arch/x86/kernel/sigframe_32.h
+++ b/arch/x86/kernel/sigframe.h
@@ -1,5 +1,5 @@
1struct sigframe 1#ifdef CONFIG_X86_32
2{ 2struct sigframe {
3 char __user *pretcode; 3 char __user *pretcode;
4 int sig; 4 int sig;
5 struct sigcontext sc; 5 struct sigcontext sc;
@@ -8,8 +8,7 @@ struct sigframe
8 char retcode[8]; 8 char retcode[8];
9}; 9};
10 10
11struct rt_sigframe 11struct rt_sigframe {
12{
13 char __user *pretcode; 12 char __user *pretcode;
14 int sig; 13 int sig;
15 struct siginfo __user *pinfo; 14 struct siginfo __user *pinfo;
@@ -19,3 +18,10 @@ struct rt_sigframe
19 struct _fpstate fpstate; 18 struct _fpstate fpstate;
20 char retcode[8]; 19 char retcode[8];
21}; 20};
21#else
22struct rt_sigframe {
23 char __user *pretcode;
24 struct ucontext uc;
25 struct siginfo info;
26};
27#endif
diff --git a/arch/x86/kernel/signal_32.c b/arch/x86/kernel/signal_32.c
index 0157a6f0f41f..f1b117930837 100644
--- a/arch/x86/kernel/signal_32.c
+++ b/arch/x86/kernel/signal_32.c
@@ -4,32 +4,44 @@
4 * 1997-11-28 Modified for POSIX.1b signals by Richard Henderson 4 * 1997-11-28 Modified for POSIX.1b signals by Richard Henderson
5 * 2000-06-20 Pentium III FXSR, SSE support by Gareth Hughes 5 * 2000-06-20 Pentium III FXSR, SSE support by Gareth Hughes
6 */ 6 */
7#include <linux/list.h>
7 8
8#include <linux/sched.h> 9#include <linux/personality.h>
9#include <linux/mm.h> 10#include <linux/binfmts.h>
10#include <linux/smp.h> 11#include <linux/suspend.h>
11#include <linux/kernel.h> 12#include <linux/kernel.h>
13#include <linux/ptrace.h>
12#include <linux/signal.h> 14#include <linux/signal.h>
15#include <linux/stddef.h>
16#include <linux/unistd.h>
13#include <linux/errno.h> 17#include <linux/errno.h>
18#include <linux/sched.h>
14#include <linux/wait.h> 19#include <linux/wait.h>
15#include <linux/unistd.h>
16#include <linux/stddef.h>
17#include <linux/personality.h>
18#include <linux/suspend.h>
19#include <linux/ptrace.h>
20#include <linux/elf.h> 20#include <linux/elf.h>
21#include <linux/binfmts.h> 21#include <linux/smp.h>
22#include <linux/mm.h>
23
22#include <asm/processor.h> 24#include <asm/processor.h>
23#include <asm/ucontext.h> 25#include <asm/ucontext.h>
24#include <asm/uaccess.h> 26#include <asm/uaccess.h>
25#include <asm/i387.h> 27#include <asm/i387.h>
26#include <asm/vdso.h> 28#include <asm/vdso.h>
27#include "sigframe_32.h"
28 29
29#define DEBUG_SIG 0 30#include "sigframe.h"
30 31
31#define _BLOCKABLE (~(sigmask(SIGKILL) | sigmask(SIGSTOP))) 32#define _BLOCKABLE (~(sigmask(SIGKILL) | sigmask(SIGSTOP)))
32 33
34#define __FIX_EFLAGS (X86_EFLAGS_AC | X86_EFLAGS_OF | \
35 X86_EFLAGS_DF | X86_EFLAGS_TF | X86_EFLAGS_SF | \
36 X86_EFLAGS_ZF | X86_EFLAGS_AF | X86_EFLAGS_PF | \
37 X86_EFLAGS_CF)
38
39#ifdef CONFIG_X86_32
40# define FIX_EFLAGS (__FIX_EFLAGS | X86_EFLAGS_RF)
41#else
42# define FIX_EFLAGS __FIX_EFLAGS
43#endif
44
33/* 45/*
34 * Atomically swap in the new signal mask, and wait for a signal. 46 * Atomically swap in the new signal mask, and wait for a signal.
35 */ 47 */
@@ -46,10 +58,11 @@ sys_sigsuspend(int history0, int history1, old_sigset_t mask)
46 current->state = TASK_INTERRUPTIBLE; 58 current->state = TASK_INTERRUPTIBLE;
47 schedule(); 59 schedule();
48 set_thread_flag(TIF_RESTORE_SIGMASK); 60 set_thread_flag(TIF_RESTORE_SIGMASK);
61
49 return -ERESTARTNOHAND; 62 return -ERESTARTNOHAND;
50} 63}
51 64
52asmlinkage int 65asmlinkage int
53sys_sigaction(int sig, const struct old_sigaction __user *act, 66sys_sigaction(int sig, const struct old_sigaction __user *act,
54 struct old_sigaction __user *oact) 67 struct old_sigaction __user *oact)
55{ 68{
@@ -58,10 +71,12 @@ sys_sigaction(int sig, const struct old_sigaction __user *act,
58 71
59 if (act) { 72 if (act) {
60 old_sigset_t mask; 73 old_sigset_t mask;
74
61 if (!access_ok(VERIFY_READ, act, sizeof(*act)) || 75 if (!access_ok(VERIFY_READ, act, sizeof(*act)) ||
62 __get_user(new_ka.sa.sa_handler, &act->sa_handler) || 76 __get_user(new_ka.sa.sa_handler, &act->sa_handler) ||
63 __get_user(new_ka.sa.sa_restorer, &act->sa_restorer)) 77 __get_user(new_ka.sa.sa_restorer, &act->sa_restorer))
64 return -EFAULT; 78 return -EFAULT;
79
65 __get_user(new_ka.sa.sa_flags, &act->sa_flags); 80 __get_user(new_ka.sa.sa_flags, &act->sa_flags);
66 __get_user(mask, &act->sa_mask); 81 __get_user(mask, &act->sa_mask);
67 siginitset(&new_ka.sa.sa_mask, mask); 82 siginitset(&new_ka.sa.sa_mask, mask);
@@ -74,6 +89,7 @@ sys_sigaction(int sig, const struct old_sigaction __user *act,
74 __put_user(old_ka.sa.sa_handler, &oact->sa_handler) || 89 __put_user(old_ka.sa.sa_handler, &oact->sa_handler) ||
75 __put_user(old_ka.sa.sa_restorer, &oact->sa_restorer)) 90 __put_user(old_ka.sa.sa_restorer, &oact->sa_restorer))
76 return -EFAULT; 91 return -EFAULT;
92
77 __put_user(old_ka.sa.sa_flags, &oact->sa_flags); 93 __put_user(old_ka.sa.sa_flags, &oact->sa_flags);
78 __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask); 94 __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask);
79 } 95 }
@@ -81,10 +97,12 @@ sys_sigaction(int sig, const struct old_sigaction __user *act,
81 return ret; 97 return ret;
82} 98}
83 99
84asmlinkage int 100asmlinkage int sys_sigaltstack(unsigned long bx)
85sys_sigaltstack(unsigned long bx)
86{ 101{
87 /* This is needed to make gcc realize it doesn't own the "struct pt_regs" */ 102 /*
103 * This is needed to make gcc realize it doesn't own the
104 * "struct pt_regs"
105 */
88 struct pt_regs *regs = (struct pt_regs *)&bx; 106 struct pt_regs *regs = (struct pt_regs *)&bx;
89 const stack_t __user *uss = (const stack_t __user *)bx; 107 const stack_t __user *uss = (const stack_t __user *)bx;
90 stack_t __user *uoss = (stack_t __user *)regs->cx; 108 stack_t __user *uoss = (stack_t __user *)regs->cx;
@@ -96,9 +114,9 @@ sys_sigaltstack(unsigned long bx)
96/* 114/*
97 * Do a signal return; undo the signal stack. 115 * Do a signal return; undo the signal stack.
98 */ 116 */
99
100static int 117static int
101restore_sigcontext(struct pt_regs *regs, struct sigcontext __user *sc, int *peax) 118restore_sigcontext(struct pt_regs *regs, struct sigcontext __user *sc,
119 unsigned long *pax)
102{ 120{
103 unsigned int err = 0; 121 unsigned int err = 0;
104 122
@@ -120,37 +138,29 @@ restore_sigcontext(struct pt_regs *regs, struct sigcontext __user *sc, int *peax
120#define GET_SEG(seg) \ 138#define GET_SEG(seg) \
121 { unsigned short tmp; \ 139 { unsigned short tmp; \
122 err |= __get_user(tmp, &sc->seg); \ 140 err |= __get_user(tmp, &sc->seg); \
123 loadsegment(seg,tmp); } 141 loadsegment(seg, tmp); }
124
125#define FIX_EFLAGS (X86_EFLAGS_AC | X86_EFLAGS_RF | \
126 X86_EFLAGS_OF | X86_EFLAGS_DF | \
127 X86_EFLAGS_TF | X86_EFLAGS_SF | X86_EFLAGS_ZF | \
128 X86_EFLAGS_AF | X86_EFLAGS_PF | X86_EFLAGS_CF)
129 142
130 GET_SEG(gs); 143 GET_SEG(gs);
131 COPY_SEG(fs); 144 COPY_SEG(fs);
132 COPY_SEG(es); 145 COPY_SEG(es);
133 COPY_SEG(ds); 146 COPY_SEG(ds);
134 COPY(di); 147 COPY(di); COPY(si); COPY(bp); COPY(sp); COPY(bx);
135 COPY(si); 148 COPY(dx); COPY(cx); COPY(ip);
136 COPY(bp);
137 COPY(sp);
138 COPY(bx);
139 COPY(dx);
140 COPY(cx);
141 COPY(ip);
142 COPY_SEG_STRICT(cs); 149 COPY_SEG_STRICT(cs);
143 COPY_SEG_STRICT(ss); 150 COPY_SEG_STRICT(ss);
144 151
145 { 152 {
146 unsigned int tmpflags; 153 unsigned int tmpflags;
154
147 err |= __get_user(tmpflags, &sc->flags); 155 err |= __get_user(tmpflags, &sc->flags);
148 regs->flags = (regs->flags & ~FIX_EFLAGS) | (tmpflags & FIX_EFLAGS); 156 regs->flags = (regs->flags & ~FIX_EFLAGS) |
157 (tmpflags & FIX_EFLAGS);
149 regs->orig_ax = -1; /* disable syscall checks */ 158 regs->orig_ax = -1; /* disable syscall checks */
150 } 159 }
151 160
152 { 161 {
153 struct _fpstate __user * buf; 162 struct _fpstate __user *buf;
163
154 err |= __get_user(buf, &sc->fpstate); 164 err |= __get_user(buf, &sc->fpstate);
155 if (buf) { 165 if (buf) {
156 if (!access_ok(VERIFY_READ, buf, sizeof(*buf))) 166 if (!access_ok(VERIFY_READ, buf, sizeof(*buf)))
@@ -158,6 +168,7 @@ restore_sigcontext(struct pt_regs *regs, struct sigcontext __user *sc, int *peax
158 err |= restore_i387(buf); 168 err |= restore_i387(buf);
159 } else { 169 } else {
160 struct task_struct *me = current; 170 struct task_struct *me = current;
171
161 if (used_math()) { 172 if (used_math()) {
162 clear_fpu(me); 173 clear_fpu(me);
163 clear_used_math(); 174 clear_used_math();
@@ -165,24 +176,26 @@ restore_sigcontext(struct pt_regs *regs, struct sigcontext __user *sc, int *peax
165 } 176 }
166 } 177 }
167 178
168 err |= __get_user(*peax, &sc->ax); 179 err |= __get_user(*pax, &sc->ax);
169 return err; 180 return err;
170 181
171badframe: 182badframe:
172 return 1; 183 return 1;
173} 184}
174 185
175asmlinkage int sys_sigreturn(unsigned long __unused) 186asmlinkage unsigned long sys_sigreturn(unsigned long __unused)
176{ 187{
177 struct pt_regs *regs = (struct pt_regs *) &__unused; 188 struct sigframe __user *frame;
178 struct sigframe __user *frame = (struct sigframe __user *)(regs->sp - 8); 189 struct pt_regs *regs;
190 unsigned long ax;
179 sigset_t set; 191 sigset_t set;
180 int ax; 192
193 regs = (struct pt_regs *) &__unused;
194 frame = (struct sigframe __user *)(regs->sp - 8);
181 195
182 if (!access_ok(VERIFY_READ, frame, sizeof(*frame))) 196 if (!access_ok(VERIFY_READ, frame, sizeof(*frame)))
183 goto badframe; 197 goto badframe;
184 if (__get_user(set.sig[0], &frame->sc.oldmask) 198 if (__get_user(set.sig[0], &frame->sc.oldmask) || (_NSIG_WORDS > 1
185 || (_NSIG_WORDS > 1
186 && __copy_from_user(&set.sig[1], &frame->extramask, 199 && __copy_from_user(&set.sig[1], &frame->extramask,
187 sizeof(frame->extramask)))) 200 sizeof(frame->extramask))))
188 goto badframe; 201 goto badframe;
@@ -192,33 +205,35 @@ asmlinkage int sys_sigreturn(unsigned long __unused)
192 current->blocked = set; 205 current->blocked = set;
193 recalc_sigpending(); 206 recalc_sigpending();
194 spin_unlock_irq(&current->sighand->siglock); 207 spin_unlock_irq(&current->sighand->siglock);
195 208
196 if (restore_sigcontext(regs, &frame->sc, &ax)) 209 if (restore_sigcontext(regs, &frame->sc, &ax))
197 goto badframe; 210 goto badframe;
198 return ax; 211 return ax;
199 212
200badframe: 213badframe:
201 if (show_unhandled_signals && printk_ratelimit()) { 214 if (show_unhandled_signals && printk_ratelimit()) {
202 printk("%s%s[%d] bad frame in sigreturn frame:%p ip:%lx" 215 printk(KERN_INFO "%s%s[%d] bad frame in sigreturn frame:"
203 " sp:%lx oeax:%lx", 216 "%p ip:%lx sp:%lx oeax:%lx",
204 task_pid_nr(current) > 1 ? KERN_INFO : KERN_EMERG, 217 task_pid_nr(current) > 1 ? KERN_INFO : KERN_EMERG,
205 current->comm, task_pid_nr(current), frame, regs->ip, 218 current->comm, task_pid_nr(current), frame, regs->ip,
206 regs->sp, regs->orig_ax); 219 regs->sp, regs->orig_ax);
207 print_vma_addr(" in ", regs->ip); 220 print_vma_addr(" in ", regs->ip);
208 printk("\n"); 221 printk(KERN_CONT "\n");
209 } 222 }
210 223
211 force_sig(SIGSEGV, current); 224 force_sig(SIGSEGV, current);
225
212 return 0; 226 return 0;
213} 227}
214 228
215asmlinkage int sys_rt_sigreturn(unsigned long __unused) 229asmlinkage int sys_rt_sigreturn(unsigned long __unused)
216{ 230{
217 struct pt_regs *regs = (struct pt_regs *) &__unused; 231 struct pt_regs *regs = (struct pt_regs *)&__unused;
218 struct rt_sigframe __user *frame = (struct rt_sigframe __user *)(regs->sp - 4); 232 struct rt_sigframe __user *frame;
233 unsigned long ax;
219 sigset_t set; 234 sigset_t set;
220 int ax;
221 235
236 frame = (struct rt_sigframe __user *)(regs->sp - sizeof(long));
222 if (!access_ok(VERIFY_READ, frame, sizeof(*frame))) 237 if (!access_ok(VERIFY_READ, frame, sizeof(*frame)))
223 goto badframe; 238 goto badframe;
224 if (__copy_from_user(&set, &frame->uc.uc_sigmask, sizeof(set))) 239 if (__copy_from_user(&set, &frame->uc.uc_sigmask, sizeof(set)))
@@ -229,7 +244,7 @@ asmlinkage int sys_rt_sigreturn(unsigned long __unused)
229 current->blocked = set; 244 current->blocked = set;
230 recalc_sigpending(); 245 recalc_sigpending();
231 spin_unlock_irq(&current->sighand->siglock); 246 spin_unlock_irq(&current->sighand->siglock);
232 247
233 if (restore_sigcontext(regs, &frame->uc.uc_mcontext, &ax)) 248 if (restore_sigcontext(regs, &frame->uc.uc_mcontext, &ax))
234 goto badframe; 249 goto badframe;
235 250
@@ -241,12 +256,11 @@ asmlinkage int sys_rt_sigreturn(unsigned long __unused)
241badframe: 256badframe:
242 force_sig(SIGSEGV, current); 257 force_sig(SIGSEGV, current);
243 return 0; 258 return 0;
244} 259}
245 260
246/* 261/*
247 * Set up a signal frame. 262 * Set up a signal frame.
248 */ 263 */
249
250static int 264static int
251setup_sigcontext(struct sigcontext __user *sc, struct _fpstate __user *fpstate, 265setup_sigcontext(struct sigcontext __user *sc, struct _fpstate __user *fpstate,
252 struct pt_regs *regs, unsigned long mask) 266 struct pt_regs *regs, unsigned long mask)
@@ -277,9 +291,9 @@ setup_sigcontext(struct sigcontext __user *sc, struct _fpstate __user *fpstate,
277 291
278 tmp = save_i387(fpstate); 292 tmp = save_i387(fpstate);
279 if (tmp < 0) 293 if (tmp < 0)
280 err = 1; 294 err = 1;
281 else 295 else
282 err |= __put_user(tmp ? fpstate : NULL, &sc->fpstate); 296 err |= __put_user(tmp ? fpstate : NULL, &sc->fpstate);
283 297
284 /* non-iBCS2 extensions.. */ 298 /* non-iBCS2 extensions.. */
285 err |= __put_user(mask, &sc->oldmask); 299 err |= __put_user(mask, &sc->oldmask);
@@ -292,7 +306,7 @@ setup_sigcontext(struct sigcontext __user *sc, struct _fpstate __user *fpstate,
292 * Determine which stack to use.. 306 * Determine which stack to use..
293 */ 307 */
294static inline void __user * 308static inline void __user *
295get_sigframe(struct k_sigaction *ka, struct pt_regs * regs, size_t frame_size) 309get_sigframe(struct k_sigaction *ka, struct pt_regs *regs, size_t frame_size)
296{ 310{
297 unsigned long sp; 311 unsigned long sp;
298 312
@@ -310,32 +324,30 @@ get_sigframe(struct k_sigaction *ka, struct pt_regs * regs, size_t frame_size)
310 if (ka->sa.sa_flags & SA_ONSTACK) { 324 if (ka->sa.sa_flags & SA_ONSTACK) {
311 if (sas_ss_flags(sp) == 0) 325 if (sas_ss_flags(sp) == 0)
312 sp = current->sas_ss_sp + current->sas_ss_size; 326 sp = current->sas_ss_sp + current->sas_ss_size;
313 } 327 } else {
314 328 /* This is the legacy signal stack switching. */
315 /* This is the legacy signal stack switching. */ 329 if ((regs->ss & 0xffff) != __USER_DS &&
316 else if ((regs->ss & 0xffff) != __USER_DS && 330 !(ka->sa.sa_flags & SA_RESTORER) &&
317 !(ka->sa.sa_flags & SA_RESTORER) && 331 ka->sa.sa_restorer)
318 ka->sa.sa_restorer) { 332 sp = (unsigned long) ka->sa.sa_restorer;
319 sp = (unsigned long) ka->sa.sa_restorer;
320 } 333 }
321 334
322 sp -= frame_size; 335 sp -= frame_size;
323 /* Align the stack pointer according to the i386 ABI, 336 /*
324 * i.e. so that on function entry ((sp + 4) & 15) == 0. */ 337 * Align the stack pointer according to the i386 ABI,
338 * i.e. so that on function entry ((sp + 4) & 15) == 0.
339 */
325 sp = ((sp + 4) & -16ul) - 4; 340 sp = ((sp + 4) & -16ul) - 4;
341
326 return (void __user *) sp; 342 return (void __user *) sp;
327} 343}
328 344
329/* These symbols are defined with the addresses in the vsyscall page. 345static int
330 See vsyscall-sigreturn.S. */ 346setup_frame(int sig, struct k_sigaction *ka, sigset_t *set,
331extern void __user __kernel_sigreturn; 347 struct pt_regs *regs)
332extern void __user __kernel_rt_sigreturn;
333
334static int setup_frame(int sig, struct k_sigaction *ka,
335 sigset_t *set, struct pt_regs * regs)
336{ 348{
337 void __user *restorer;
338 struct sigframe __user *frame; 349 struct sigframe __user *frame;
350 void __user *restorer;
339 int err = 0; 351 int err = 0;
340 int usig; 352 int usig;
341 353
@@ -365,7 +377,7 @@ static int setup_frame(int sig, struct k_sigaction *ka,
365 goto give_sigsegv; 377 goto give_sigsegv;
366 } 378 }
367 379
368 if (current->binfmt->hasvdso) 380 if (current->mm->context.vdso)
369 restorer = VDSO32_SYMBOL(current->mm->context.vdso, sigreturn); 381 restorer = VDSO32_SYMBOL(current->mm->context.vdso, sigreturn);
370 else 382 else
371 restorer = &frame->retcode; 383 restorer = &frame->retcode;
@@ -374,9 +386,9 @@ static int setup_frame(int sig, struct k_sigaction *ka,
374 386
375 /* Set up to return from userspace. */ 387 /* Set up to return from userspace. */
376 err |= __put_user(restorer, &frame->pretcode); 388 err |= __put_user(restorer, &frame->pretcode);
377 389
378 /* 390 /*
379 * This is popl %eax ; movl $,%eax ; int $0x80 391 * This is popl %eax ; movl $__NR_sigreturn, %eax ; int $0x80
380 * 392 *
381 * WE DO NOT USE IT ANY MORE! It's only left here for historical 393 * WE DO NOT USE IT ANY MORE! It's only left here for historical
382 * reasons and because gdb uses it as a signature to notice 394 * reasons and because gdb uses it as a signature to notice
@@ -390,11 +402,11 @@ static int setup_frame(int sig, struct k_sigaction *ka,
390 goto give_sigsegv; 402 goto give_sigsegv;
391 403
392 /* Set up registers for signal handler */ 404 /* Set up registers for signal handler */
393 regs->sp = (unsigned long) frame; 405 regs->sp = (unsigned long)frame;
394 regs->ip = (unsigned long) ka->sa.sa_handler; 406 regs->ip = (unsigned long)ka->sa.sa_handler;
395 regs->ax = (unsigned long) sig; 407 regs->ax = (unsigned long)sig;
396 regs->dx = (unsigned long) 0; 408 regs->dx = 0;
397 regs->cx = (unsigned long) 0; 409 regs->cx = 0;
398 410
399 regs->ds = __USER_DS; 411 regs->ds = __USER_DS;
400 regs->es = __USER_DS; 412 regs->es = __USER_DS;
@@ -407,15 +419,10 @@ static int setup_frame(int sig, struct k_sigaction *ka,
407 * The tracer may want to single-step inside the 419 * The tracer may want to single-step inside the
408 * handler too. 420 * handler too.
409 */ 421 */
410 regs->flags &= ~(TF_MASK | X86_EFLAGS_DF); 422 regs->flags &= ~(X86_EFLAGS_TF | X86_EFLAGS_DF);
411 if (test_thread_flag(TIF_SINGLESTEP)) 423 if (test_thread_flag(TIF_SINGLESTEP))
412 ptrace_notify(SIGTRAP); 424 ptrace_notify(SIGTRAP);
413 425
414#if DEBUG_SIG
415 printk("SIG deliver (%s:%d): sp=%p pc=%p ra=%p\n",
416 current->comm, current->pid, frame, regs->ip, frame->pretcode);
417#endif
418
419 return 0; 426 return 0;
420 427
421give_sigsegv: 428give_sigsegv:
@@ -424,10 +431,10 @@ give_sigsegv:
424} 431}
425 432
426static int setup_rt_frame(int sig, struct k_sigaction *ka, siginfo_t *info, 433static int setup_rt_frame(int sig, struct k_sigaction *ka, siginfo_t *info,
427 sigset_t *set, struct pt_regs * regs) 434 sigset_t *set, struct pt_regs *regs)
428{ 435{
429 void __user *restorer;
430 struct rt_sigframe __user *frame; 436 struct rt_sigframe __user *frame;
437 void __user *restorer;
431 int err = 0; 438 int err = 0;
432 int usig; 439 int usig;
433 440
@@ -457,7 +464,7 @@ static int setup_rt_frame(int sig, struct k_sigaction *ka, siginfo_t *info,
457 &frame->uc.uc_stack.ss_flags); 464 &frame->uc.uc_stack.ss_flags);
458 err |= __put_user(current->sas_ss_size, &frame->uc.uc_stack.ss_size); 465 err |= __put_user(current->sas_ss_size, &frame->uc.uc_stack.ss_size);
459 err |= setup_sigcontext(&frame->uc.uc_mcontext, &frame->fpstate, 466 err |= setup_sigcontext(&frame->uc.uc_mcontext, &frame->fpstate,
460 regs, set->sig[0]); 467 regs, set->sig[0]);
461 err |= __copy_to_user(&frame->uc.uc_sigmask, set, sizeof(*set)); 468 err |= __copy_to_user(&frame->uc.uc_sigmask, set, sizeof(*set));
462 if (err) 469 if (err)
463 goto give_sigsegv; 470 goto give_sigsegv;
@@ -467,9 +474,9 @@ static int setup_rt_frame(int sig, struct k_sigaction *ka, siginfo_t *info,
467 if (ka->sa.sa_flags & SA_RESTORER) 474 if (ka->sa.sa_flags & SA_RESTORER)
468 restorer = ka->sa.sa_restorer; 475 restorer = ka->sa.sa_restorer;
469 err |= __put_user(restorer, &frame->pretcode); 476 err |= __put_user(restorer, &frame->pretcode);
470 477
471 /* 478 /*
472 * This is movl $,%ax ; int $0x80 479 * This is movl $__NR_rt_sigreturn, %ax ; int $0x80
473 * 480 *
474 * WE DO NOT USE IT ANY MORE! It's only left here for historical 481 * WE DO NOT USE IT ANY MORE! It's only left here for historical
475 * reasons and because gdb uses it as a signature to notice 482 * reasons and because gdb uses it as a signature to notice
@@ -483,11 +490,11 @@ static int setup_rt_frame(int sig, struct k_sigaction *ka, siginfo_t *info,
483 goto give_sigsegv; 490 goto give_sigsegv;
484 491
485 /* Set up registers for signal handler */ 492 /* Set up registers for signal handler */
486 regs->sp = (unsigned long) frame; 493 regs->sp = (unsigned long)frame;
487 regs->ip = (unsigned long) ka->sa.sa_handler; 494 regs->ip = (unsigned long)ka->sa.sa_handler;
488 regs->ax = (unsigned long) usig; 495 regs->ax = (unsigned long)usig;
489 regs->dx = (unsigned long) &frame->info; 496 regs->dx = (unsigned long)&frame->info;
490 regs->cx = (unsigned long) &frame->uc; 497 regs->cx = (unsigned long)&frame->uc;
491 498
492 regs->ds = __USER_DS; 499 regs->ds = __USER_DS;
493 regs->es = __USER_DS; 500 regs->es = __USER_DS;
@@ -500,15 +507,10 @@ static int setup_rt_frame(int sig, struct k_sigaction *ka, siginfo_t *info,
500 * The tracer may want to single-step inside the 507 * The tracer may want to single-step inside the
501 * handler too. 508 * handler too.
502 */ 509 */
503 regs->flags &= ~(TF_MASK | X86_EFLAGS_DF); 510 regs->flags &= ~(X86_EFLAGS_TF | X86_EFLAGS_DF);
504 if (test_thread_flag(TIF_SINGLESTEP)) 511 if (test_thread_flag(TIF_SINGLESTEP))
505 ptrace_notify(SIGTRAP); 512 ptrace_notify(SIGTRAP);
506 513
507#if DEBUG_SIG
508 printk("SIG deliver (%s:%d): sp=%p pc=%p ra=%p\n",
509 current->comm, current->pid, frame, regs->ip, frame->pretcode);
510#endif
511
512 return 0; 514 return 0;
513 515
514give_sigsegv: 516give_sigsegv:
@@ -517,33 +519,33 @@ give_sigsegv:
517} 519}
518 520
519/* 521/*
520 * OK, we're invoking a handler 522 * OK, we're invoking a handler:
521 */ 523 */
522
523static int 524static int
524handle_signal(unsigned long sig, siginfo_t *info, struct k_sigaction *ka, 525handle_signal(unsigned long sig, siginfo_t *info, struct k_sigaction *ka,
525 sigset_t *oldset, struct pt_regs * regs) 526 sigset_t *oldset, struct pt_regs *regs)
526{ 527{
527 int ret; 528 int ret;
528 529
529 /* Are we from a system call? */ 530 /* Are we from a system call? */
530 if (regs->orig_ax >= 0) { 531 if ((long)regs->orig_ax >= 0) {
531 /* If so, check system call restarting.. */ 532 /* If so, check system call restarting.. */
532 switch (regs->ax) { 533 switch (regs->ax) {
533 case -ERESTART_RESTARTBLOCK: 534 case -ERESTART_RESTARTBLOCK:
534 case -ERESTARTNOHAND: 535 case -ERESTARTNOHAND:
536 regs->ax = -EINTR;
537 break;
538
539 case -ERESTARTSYS:
540 if (!(ka->sa.sa_flags & SA_RESTART)) {
535 regs->ax = -EINTR; 541 regs->ax = -EINTR;
536 break; 542 break;
537 543 }
538 case -ERESTARTSYS: 544 /* fallthrough */
539 if (!(ka->sa.sa_flags & SA_RESTART)) { 545 case -ERESTARTNOINTR:
540 regs->ax = -EINTR; 546 regs->ax = regs->orig_ax;
541 break; 547 regs->ip -= 2;
542 } 548 break;
543 /* fallthrough */
544 case -ERESTARTNOINTR:
545 regs->ax = regs->orig_ax;
546 regs->ip -= 2;
547 } 549 }
548 } 550 }
549 551
@@ -561,16 +563,17 @@ handle_signal(unsigned long sig, siginfo_t *info, struct k_sigaction *ka,
561 else 563 else
562 ret = setup_frame(sig, ka, oldset, regs); 564 ret = setup_frame(sig, ka, oldset, regs);
563 565
564 if (ret == 0) { 566 if (ret)
565 spin_lock_irq(&current->sighand->siglock); 567 return ret;
566 sigorsets(&current->blocked,&current->blocked,&ka->sa.sa_mask);
567 if (!(ka->sa.sa_flags & SA_NODEFER))
568 sigaddset(&current->blocked,sig);
569 recalc_sigpending();
570 spin_unlock_irq(&current->sighand->siglock);
571 }
572 568
573 return ret; 569 spin_lock_irq(&current->sighand->siglock);
570 sigorsets(&current->blocked, &current->blocked, &ka->sa.sa_mask);
571 if (!(ka->sa.sa_flags & SA_NODEFER))
572 sigaddset(&current->blocked, sig);
573 recalc_sigpending();
574 spin_unlock_irq(&current->sighand->siglock);
575
576 return 0;
574} 577}
575 578
576/* 579/*
@@ -580,18 +583,17 @@ handle_signal(unsigned long sig, siginfo_t *info, struct k_sigaction *ka,
580 */ 583 */
581static void do_signal(struct pt_regs *regs) 584static void do_signal(struct pt_regs *regs)
582{ 585{
586 struct k_sigaction ka;
583 siginfo_t info; 587 siginfo_t info;
584 int signr; 588 int signr;
585 struct k_sigaction ka;
586 sigset_t *oldset; 589 sigset_t *oldset;
587 590
588 /* 591 /*
589 * We want the common case to go fast, which 592 * We want the common case to go fast, which is why we may in certain
590 * is why we may in certain cases get here from 593 * cases get here from kernel mode. Just return without doing anything
591 * kernel mode. Just return without doing anything 594 * if so.
592 * if so. vm86 regs switched out by assembly code 595 * X86_32: vm86 regs switched out by assembly code before reaching
593 * before reaching here, so testing against kernel 596 * here, so testing against kernel CS suffices.
594 * CS suffices.
595 */ 597 */
596 if (!user_mode(regs)) 598 if (!user_mode(regs))
597 return; 599 return;
@@ -603,29 +605,31 @@ static void do_signal(struct pt_regs *regs)
603 605
604 signr = get_signal_to_deliver(&info, &ka, regs, NULL); 606 signr = get_signal_to_deliver(&info, &ka, regs, NULL);
605 if (signr > 0) { 607 if (signr > 0) {
606 /* Re-enable any watchpoints before delivering the 608 /*
609 * Re-enable any watchpoints before delivering the
607 * signal to user space. The processor register will 610 * signal to user space. The processor register will
608 * have been cleared if the watchpoint triggered 611 * have been cleared if the watchpoint triggered
609 * inside the kernel. 612 * inside the kernel.
610 */ 613 */
611 if (unlikely(current->thread.debugreg7)) 614 if (current->thread.debugreg7)
612 set_debugreg(current->thread.debugreg7, 7); 615 set_debugreg(current->thread.debugreg7, 7);
613 616
614 /* Whee! Actually deliver the signal. */ 617 /* Whee! Actually deliver the signal. */
615 if (handle_signal(signr, &info, &ka, oldset, regs) == 0) { 618 if (handle_signal(signr, &info, &ka, oldset, regs) == 0) {
616 /* a signal was successfully delivered; the saved 619 /*
620 * a signal was successfully delivered; the saved
617 * sigmask will have been stored in the signal frame, 621 * sigmask will have been stored in the signal frame,
618 * and will be restored by sigreturn, so we can simply 622 * and will be restored by sigreturn, so we can simply
619 * clear the TIF_RESTORE_SIGMASK flag */ 623 * clear the TIF_RESTORE_SIGMASK flag
624 */
620 if (test_thread_flag(TIF_RESTORE_SIGMASK)) 625 if (test_thread_flag(TIF_RESTORE_SIGMASK))
621 clear_thread_flag(TIF_RESTORE_SIGMASK); 626 clear_thread_flag(TIF_RESTORE_SIGMASK);
622 } 627 }
623
624 return; 628 return;
625 } 629 }
626 630
627 /* Did we come from a system call? */ 631 /* Did we come from a system call? */
628 if (regs->orig_ax >= 0) { 632 if ((long)regs->orig_ax >= 0) {
629 /* Restart the system call - no handlers present */ 633 /* Restart the system call - no handlers present */
630 switch (regs->ax) { 634 switch (regs->ax) {
631 case -ERESTARTNOHAND: 635 case -ERESTARTNOHAND:
@@ -642,8 +646,10 @@ static void do_signal(struct pt_regs *regs)
642 } 646 }
643 } 647 }
644 648
645 /* if there's no signal to deliver, we just put the saved sigmask 649 /*
646 * back */ 650 * If there's no signal to deliver, we just put the saved sigmask
651 * back.
652 */
647 if (test_thread_flag(TIF_RESTORE_SIGMASK)) { 653 if (test_thread_flag(TIF_RESTORE_SIGMASK)) {
648 clear_thread_flag(TIF_RESTORE_SIGMASK); 654 clear_thread_flag(TIF_RESTORE_SIGMASK);
649 sigprocmask(SIG_SETMASK, &current->saved_sigmask, NULL); 655 sigprocmask(SIG_SETMASK, &current->saved_sigmask, NULL);
@@ -654,13 +660,12 @@ static void do_signal(struct pt_regs *regs)
654 * notification of userspace execution resumption 660 * notification of userspace execution resumption
655 * - triggered by the TIF_WORK_MASK flags 661 * - triggered by the TIF_WORK_MASK flags
656 */ 662 */
657__attribute__((regparm(3))) 663void
658void do_notify_resume(struct pt_regs *regs, void *_unused, 664do_notify_resume(struct pt_regs *regs, void *unused, __u32 thread_info_flags)
659 __u32 thread_info_flags)
660{ 665{
661 /* Pending single-step? */ 666 /* Pending single-step? */
662 if (thread_info_flags & _TIF_SINGLESTEP) { 667 if (thread_info_flags & _TIF_SINGLESTEP) {
663 regs->flags |= TF_MASK; 668 regs->flags |= X86_EFLAGS_TF;
664 clear_thread_flag(TIF_SINGLESTEP); 669 clear_thread_flag(TIF_SINGLESTEP);
665 } 670 }
666 671
@@ -670,6 +675,6 @@ void do_notify_resume(struct pt_regs *regs, void *_unused,
670 675
671 if (thread_info_flags & _TIF_HRTICK_RESCHED) 676 if (thread_info_flags & _TIF_HRTICK_RESCHED)
672 hrtick_resched(); 677 hrtick_resched();
673 678
674 clear_thread_flag(TIF_IRET); 679 clear_thread_flag(TIF_IRET);
675} 680}
diff --git a/arch/x86/kernel/signal_64.c b/arch/x86/kernel/signal_64.c
index 1c83e5124c65..827179c5b32a 100644
--- a/arch/x86/kernel/signal_64.c
+++ b/arch/x86/kernel/signal_64.c
@@ -19,17 +19,28 @@
19#include <linux/stddef.h> 19#include <linux/stddef.h>
20#include <linux/personality.h> 20#include <linux/personality.h>
21#include <linux/compiler.h> 21#include <linux/compiler.h>
22#include <asm/processor.h>
22#include <asm/ucontext.h> 23#include <asm/ucontext.h>
23#include <asm/uaccess.h> 24#include <asm/uaccess.h>
24#include <asm/i387.h> 25#include <asm/i387.h>
25#include <asm/proto.h> 26#include <asm/proto.h>
26#include <asm/ia32_unistd.h> 27#include <asm/ia32_unistd.h>
27#include <asm/mce.h> 28#include <asm/mce.h>
28 29#include "sigframe.h"
29/* #define DEBUG_SIG 1 */
30 30
31#define _BLOCKABLE (~(sigmask(SIGKILL) | sigmask(SIGSTOP))) 31#define _BLOCKABLE (~(sigmask(SIGKILL) | sigmask(SIGSTOP)))
32 32
33#define __FIX_EFLAGS (X86_EFLAGS_AC | X86_EFLAGS_OF | \
34 X86_EFLAGS_DF | X86_EFLAGS_TF | X86_EFLAGS_SF | \
35 X86_EFLAGS_ZF | X86_EFLAGS_AF | X86_EFLAGS_PF | \
36 X86_EFLAGS_CF)
37
38#ifdef CONFIG_X86_32
39# define FIX_EFLAGS (__FIX_EFLAGS | X86_EFLAGS_RF)
40#else
41# define FIX_EFLAGS __FIX_EFLAGS
42#endif
43
33int ia32_setup_rt_frame(int sig, struct k_sigaction *ka, siginfo_t *info, 44int ia32_setup_rt_frame(int sig, struct k_sigaction *ka, siginfo_t *info,
34 sigset_t *set, struct pt_regs * regs); 45 sigset_t *set, struct pt_regs * regs);
35int ia32_setup_frame(int sig, struct k_sigaction *ka, 46int ia32_setup_frame(int sig, struct k_sigaction *ka,
@@ -46,16 +57,9 @@ sys_sigaltstack(const stack_t __user *uss, stack_t __user *uoss,
46/* 57/*
47 * Do a signal return; undo the signal stack. 58 * Do a signal return; undo the signal stack.
48 */ 59 */
49
50struct rt_sigframe
51{
52 char __user *pretcode;
53 struct ucontext uc;
54 struct siginfo info;
55};
56
57static int 60static int
58restore_sigcontext(struct pt_regs *regs, struct sigcontext __user *sc, unsigned long *prax) 61restore_sigcontext(struct pt_regs *regs, struct sigcontext __user *sc,
62 unsigned long *pax)
59{ 63{
60 unsigned int err = 0; 64 unsigned int err = 0;
61 65
@@ -87,7 +91,7 @@ restore_sigcontext(struct pt_regs *regs, struct sigcontext __user *sc, unsigned
87 { 91 {
88 unsigned int tmpflags; 92 unsigned int tmpflags;
89 err |= __get_user(tmpflags, &sc->flags); 93 err |= __get_user(tmpflags, &sc->flags);
90 regs->flags = (regs->flags & ~0x40DD5) | (tmpflags & 0x40DD5); 94 regs->flags = (regs->flags & ~FIX_EFLAGS) | (tmpflags & FIX_EFLAGS);
91 regs->orig_ax = -1; /* disable syscall checks */ 95 regs->orig_ax = -1; /* disable syscall checks */
92 } 96 }
93 97
@@ -108,7 +112,7 @@ restore_sigcontext(struct pt_regs *regs, struct sigcontext __user *sc, unsigned
108 } 112 }
109 } 113 }
110 114
111 err |= __get_user(*prax, &sc->ax); 115 err |= __get_user(*pax, &sc->ax);
112 return err; 116 return err;
113 117
114badframe: 118badframe:
@@ -121,13 +125,11 @@ asmlinkage long sys_rt_sigreturn(struct pt_regs *regs)
121 sigset_t set; 125 sigset_t set;
122 unsigned long ax; 126 unsigned long ax;
123 127
124 frame = (struct rt_sigframe __user *)(regs->sp - 8); 128 frame = (struct rt_sigframe __user *)(regs->sp - sizeof(long));
125 if (!access_ok(VERIFY_READ, frame, sizeof(*frame))) { 129 if (!access_ok(VERIFY_READ, frame, sizeof(*frame)))
126 goto badframe; 130 goto badframe;
127 } 131 if (__copy_from_user(&set, &frame->uc.uc_sigmask, sizeof(set)))
128 if (__copy_from_user(&set, &frame->uc.uc_sigmask, sizeof(set))) {
129 goto badframe; 132 goto badframe;
130 }
131 133
132 sigdelsetmask(&set, ~_BLOCKABLE); 134 sigdelsetmask(&set, ~_BLOCKABLE);
133 spin_lock_irq(&current->sighand->siglock); 135 spin_lock_irq(&current->sighand->siglock);
@@ -138,10 +140,6 @@ asmlinkage long sys_rt_sigreturn(struct pt_regs *regs)
138 if (restore_sigcontext(regs, &frame->uc.uc_mcontext, &ax)) 140 if (restore_sigcontext(regs, &frame->uc.uc_mcontext, &ax))
139 goto badframe; 141 goto badframe;
140 142
141#ifdef DEBUG_SIG
142 printk("%d sigreturn ip:%lx sp:%lx frame:%p ax:%lx\n",current->pid,regs->ip,regs->sp,frame,ax);
143#endif
144
145 if (do_sigaltstack(&frame->uc.uc_stack, NULL, regs->sp) == -EFAULT) 143 if (do_sigaltstack(&frame->uc.uc_stack, NULL, regs->sp) == -EFAULT)
146 goto badframe; 144 goto badframe;
147 145
@@ -270,10 +268,6 @@ static int setup_rt_frame(int sig, struct k_sigaction *ka, siginfo_t *info,
270 if (err) 268 if (err)
271 goto give_sigsegv; 269 goto give_sigsegv;
272 270
273#ifdef DEBUG_SIG
274 printk("%d old ip %lx old sp %lx old ax %lx\n", current->pid,regs->ip,regs->sp,regs->ax);
275#endif
276
277 /* Set up registers for signal handler */ 271 /* Set up registers for signal handler */
278 regs->di = sig; 272 regs->di = sig;
279 /* In case the signal handler was declared without prototypes */ 273 /* In case the signal handler was declared without prototypes */
@@ -298,10 +292,6 @@ static int setup_rt_frame(int sig, struct k_sigaction *ka, siginfo_t *info,
298 regs->flags &= ~(X86_EFLAGS_TF | X86_EFLAGS_DF); 292 regs->flags &= ~(X86_EFLAGS_TF | X86_EFLAGS_DF);
299 if (test_thread_flag(TIF_SINGLESTEP)) 293 if (test_thread_flag(TIF_SINGLESTEP))
300 ptrace_notify(SIGTRAP); 294 ptrace_notify(SIGTRAP);
301#ifdef DEBUG_SIG
302 printk("SIG deliver (%s:%d): sp=%p pc=%lx ra=%p\n",
303 current->comm, current->pid, frame, regs->ip, frame->pretcode);
304#endif
305 295
306 return 0; 296 return 0;
307 297
@@ -345,35 +335,29 @@ static long current_syscall_ret(struct pt_regs *regs)
345 335
346static int 336static int
347handle_signal(unsigned long sig, siginfo_t *info, struct k_sigaction *ka, 337handle_signal(unsigned long sig, siginfo_t *info, struct k_sigaction *ka,
348 sigset_t *oldset, struct pt_regs *regs) 338 sigset_t *oldset, struct pt_regs *regs)
349{ 339{
350 int ret; 340 int ret;
351 341
352#ifdef DEBUG_SIG
353 printk("handle_signal pid:%d sig:%lu ip:%lx sp:%lx regs=%p\n",
354 current->pid, sig,
355 regs->ip, regs->sp, regs);
356#endif
357
358 /* Are we from a system call? */ 342 /* Are we from a system call? */
359 if (current_syscall(regs) >= 0) { 343 if (current_syscall(regs) >= 0) {
360 /* If so, check system call restarting.. */ 344 /* If so, check system call restarting.. */
361 switch (current_syscall_ret(regs)) { 345 switch (current_syscall_ret(regs)) {
362 case -ERESTART_RESTARTBLOCK: 346 case -ERESTART_RESTARTBLOCK:
363 case -ERESTARTNOHAND: 347 case -ERESTARTNOHAND:
364 regs->ax = -EINTR; 348 regs->ax = -EINTR;
365 break; 349 break;
366 350
367 case -ERESTARTSYS: 351 case -ERESTARTSYS:
368 if (!(ka->sa.sa_flags & SA_RESTART)) { 352 if (!(ka->sa.sa_flags & SA_RESTART)) {
369 regs->ax = -EINTR; 353 regs->ax = -EINTR;
370 break;
371 }
372 /* fallthrough */
373 case -ERESTARTNOINTR:
374 regs->ax = regs->orig_ax;
375 regs->ip -= 2;
376 break; 354 break;
355 }
356 /* fallthrough */
357 case -ERESTARTNOINTR:
358 regs->ax = regs->orig_ax;
359 regs->ip -= 2;
360 break;
377 } 361 }
378 } 362 }
379 363
@@ -420,10 +404,11 @@ static void do_signal(struct pt_regs *regs)
420 sigset_t *oldset; 404 sigset_t *oldset;
421 405
422 /* 406 /*
423 * We want the common case to go fast, which 407 * We want the common case to go fast, which is why we may in certain
424 * is why we may in certain cases get here from 408 * cases get here from kernel mode. Just return without doing anything
425 * kernel mode. Just return without doing anything
426 * if so. 409 * if so.
410 * X86_32: vm86 regs switched out by assembly code before reaching
411 * here, so testing against kernel CS suffices.
427 */ 412 */
428 if (!user_mode(regs)) 413 if (!user_mode(regs))
429 return; 414 return;
@@ -473,22 +458,19 @@ static void do_signal(struct pt_regs *regs)
473 } 458 }
474 } 459 }
475 460
476 /* if there's no signal to deliver, we just put the saved sigmask 461 /*
477 back. */ 462 * If there's no signal to deliver, we just put the saved sigmask
463 * back.
464 */
478 if (test_thread_flag(TIF_RESTORE_SIGMASK)) { 465 if (test_thread_flag(TIF_RESTORE_SIGMASK)) {
479 clear_thread_flag(TIF_RESTORE_SIGMASK); 466 clear_thread_flag(TIF_RESTORE_SIGMASK);
480 sigprocmask(SIG_SETMASK, &current->saved_sigmask, NULL); 467 sigprocmask(SIG_SETMASK, &current->saved_sigmask, NULL);
481 } 468 }
482} 469}
483 470
484void 471void do_notify_resume(struct pt_regs *regs, void *unused,
485do_notify_resume(struct pt_regs *regs, void *unused, __u32 thread_info_flags) 472 __u32 thread_info_flags)
486{ 473{
487#ifdef DEBUG_SIG
488 printk("do_notify_resume flags:%x ip:%lx sp:%lx caller:%p pending:%x\n",
489 thread_info_flags, regs->ip, regs->sp, __builtin_return_address(0),signal_pending(current));
490#endif
491
492 /* Pending single-step? */ 474 /* Pending single-step? */
493 if (thread_info_flags & _TIF_SINGLESTEP) { 475 if (thread_info_flags & _TIF_SINGLESTEP) {
494 regs->flags |= X86_EFLAGS_TF; 476 regs->flags |= X86_EFLAGS_TF;
@@ -502,7 +484,7 @@ do_notify_resume(struct pt_regs *regs, void *unused, __u32 thread_info_flags)
502#endif /* CONFIG_X86_MCE */ 484#endif /* CONFIG_X86_MCE */
503 485
504 /* deal with pending signal delivery */ 486 /* deal with pending signal delivery */
505 if (thread_info_flags & (_TIF_SIGPENDING|_TIF_RESTORE_SIGMASK)) 487 if (thread_info_flags & (_TIF_SIGPENDING | _TIF_RESTORE_SIGMASK))
506 do_signal(regs); 488 do_signal(regs);
507 489
508 if (thread_info_flags & _TIF_HRTICK_RESCHED) 490 if (thread_info_flags & _TIF_HRTICK_RESCHED)
diff --git a/arch/x86/kernel/smp.c b/arch/x86/kernel/smp.c
new file mode 100644
index 000000000000..8f75893a6467
--- /dev/null
+++ b/arch/x86/kernel/smp.c
@@ -0,0 +1,343 @@
1/*
2 * Intel SMP support routines.
3 *
4 * (c) 1995 Alan Cox, Building #3 <alan@redhat.com>
5 * (c) 1998-99, 2000 Ingo Molnar <mingo@redhat.com>
6 * (c) 2002,2003 Andi Kleen, SuSE Labs.
7 *
8 * i386 and x86_64 integration by Glauber Costa <gcosta@redhat.com>
9 *
10 * This code is released under the GNU General Public License version 2 or
11 * later.
12 */
13
14#include <linux/init.h>
15
16#include <linux/mm.h>
17#include <linux/delay.h>
18#include <linux/spinlock.h>
19#include <linux/kernel_stat.h>
20#include <linux/mc146818rtc.h>
21#include <linux/cache.h>
22#include <linux/interrupt.h>
23#include <linux/cpu.h>
24
25#include <asm/mtrr.h>
26#include <asm/tlbflush.h>
27#include <asm/mmu_context.h>
28#include <asm/proto.h>
29#include <mach_ipi.h>
30#include <mach_apic.h>
31/*
32 * Some notes on x86 processor bugs affecting SMP operation:
33 *
34 * Pentium, Pentium Pro, II, III (and all CPUs) have bugs.
35 * The Linux implications for SMP are handled as follows:
36 *
37 * Pentium III / [Xeon]
38 * None of the E1AP-E3AP errata are visible to the user.
39 *
40 * E1AP. see PII A1AP
41 * E2AP. see PII A2AP
42 * E3AP. see PII A3AP
43 *
44 * Pentium II / [Xeon]
45 * None of the A1AP-A3AP errata are visible to the user.
46 *
47 * A1AP. see PPro 1AP
48 * A2AP. see PPro 2AP
49 * A3AP. see PPro 7AP
50 *
51 * Pentium Pro
52 * None of 1AP-9AP errata are visible to the normal user,
53 * except occasional delivery of 'spurious interrupt' as trap #15.
54 * This is very rare and a non-problem.
55 *
56 * 1AP. Linux maps APIC as non-cacheable
57 * 2AP. worked around in hardware
58 * 3AP. fixed in C0 and above steppings microcode update.
59 * Linux does not use excessive STARTUP_IPIs.
60 * 4AP. worked around in hardware
61 * 5AP. symmetric IO mode (normal Linux operation) not affected.
62 * 'noapic' mode has vector 0xf filled out properly.
63 * 6AP. 'noapic' mode might be affected - fixed in later steppings
64 * 7AP. We do not assume writes to the LVT deassering IRQs
65 * 8AP. We do not enable low power mode (deep sleep) during MP bootup
66 * 9AP. We do not use mixed mode
67 *
68 * Pentium
69 * There is a marginal case where REP MOVS on 100MHz SMP
70 * machines with B stepping processors can fail. XXX should provide
71 * an L1cache=Writethrough or L1cache=off option.
72 *
73 * B stepping CPUs may hang. There are hardware work arounds
74 * for this. We warn about it in case your board doesn't have the work
75 * arounds. Basically that's so I can tell anyone with a B stepping
76 * CPU and SMP problems "tough".
77 *
78 * Specific items [From Pentium Processor Specification Update]
79 *
80 * 1AP. Linux doesn't use remote read
81 * 2AP. Linux doesn't trust APIC errors
82 * 3AP. We work around this
83 * 4AP. Linux never generated 3 interrupts of the same priority
84 * to cause a lost local interrupt.
85 * 5AP. Remote read is never used
86 * 6AP. not affected - worked around in hardware
87 * 7AP. not affected - worked around in hardware
88 * 8AP. worked around in hardware - we get explicit CS errors if not
89 * 9AP. only 'noapic' mode affected. Might generate spurious
90 * interrupts, we log only the first one and count the
91 * rest silently.
92 * 10AP. not affected - worked around in hardware
93 * 11AP. Linux reads the APIC between writes to avoid this, as per
94 * the documentation. Make sure you preserve this as it affects
95 * the C stepping chips too.
96 * 12AP. not affected - worked around in hardware
97 * 13AP. not affected - worked around in hardware
98 * 14AP. we always deassert INIT during bootup
99 * 15AP. not affected - worked around in hardware
100 * 16AP. not affected - worked around in hardware
101 * 17AP. not affected - worked around in hardware
102 * 18AP. not affected - worked around in hardware
103 * 19AP. not affected - worked around in BIOS
104 *
105 * If this sounds worrying believe me these bugs are either ___RARE___,
106 * or are signal timing bugs worked around in hardware and there's
107 * about nothing of note with C stepping upwards.
108 */
109
110/*
111 * this function sends a 'reschedule' IPI to another CPU.
112 * it goes straight through and wastes no time serializing
113 * anything. Worst case is that we lose a reschedule ...
114 */
115static void native_smp_send_reschedule(int cpu)
116{
117 if (unlikely(cpu_is_offline(cpu))) {
118 WARN_ON(1);
119 return;
120 }
121 send_IPI_mask(cpumask_of_cpu(cpu), RESCHEDULE_VECTOR);
122}
123
124/*
125 * Structure and data for smp_call_function(). This is designed to minimise
126 * static memory requirements. It also looks cleaner.
127 */
128static DEFINE_SPINLOCK(call_lock);
129
130struct call_data_struct {
131 void (*func) (void *info);
132 void *info;
133 atomic_t started;
134 atomic_t finished;
135 int wait;
136};
137
138void lock_ipi_call_lock(void)
139{
140 spin_lock_irq(&call_lock);
141}
142
143void unlock_ipi_call_lock(void)
144{
145 spin_unlock_irq(&call_lock);
146}
147
148static struct call_data_struct *call_data;
149
150static void __smp_call_function(void (*func) (void *info), void *info,
151 int nonatomic, int wait)
152{
153 struct call_data_struct data;
154 int cpus = num_online_cpus() - 1;
155
156 if (!cpus)
157 return;
158
159 data.func = func;
160 data.info = info;
161 atomic_set(&data.started, 0);
162 data.wait = wait;
163 if (wait)
164 atomic_set(&data.finished, 0);
165
166 call_data = &data;
167 mb();
168
169 /* Send a message to all other CPUs and wait for them to respond */
170 send_IPI_allbutself(CALL_FUNCTION_VECTOR);
171
172 /* Wait for response */
173 while (atomic_read(&data.started) != cpus)
174 cpu_relax();
175
176 if (wait)
177 while (atomic_read(&data.finished) != cpus)
178 cpu_relax();
179}
180
181
182/**
183 * smp_call_function_mask(): Run a function on a set of other CPUs.
184 * @mask: The set of cpus to run on. Must not include the current cpu.
185 * @func: The function to run. This must be fast and non-blocking.
186 * @info: An arbitrary pointer to pass to the function.
187 * @wait: If true, wait (atomically) until function has completed on other CPUs.
188 *
189 * Returns 0 on success, else a negative status code.
190 *
191 * If @wait is true, then returns once @func has returned; otherwise
192 * it returns just before the target cpu calls @func.
193 *
194 * You must not call this function with disabled interrupts or from a
195 * hardware interrupt handler or from a bottom half handler.
196 */
197static int
198native_smp_call_function_mask(cpumask_t mask,
199 void (*func)(void *), void *info,
200 int wait)
201{
202 struct call_data_struct data;
203 cpumask_t allbutself;
204 int cpus;
205
206 /* Can deadlock when called with interrupts disabled */
207 WARN_ON(irqs_disabled());
208
209 /* Holding any lock stops cpus from going down. */
210 spin_lock(&call_lock);
211
212 allbutself = cpu_online_map;
213 cpu_clear(smp_processor_id(), allbutself);
214
215 cpus_and(mask, mask, allbutself);
216 cpus = cpus_weight(mask);
217
218 if (!cpus) {
219 spin_unlock(&call_lock);
220 return 0;
221 }
222
223 data.func = func;
224 data.info = info;
225 atomic_set(&data.started, 0);
226 data.wait = wait;
227 if (wait)
228 atomic_set(&data.finished, 0);
229
230 call_data = &data;
231 wmb();
232
233 /* Send a message to other CPUs */
234 if (cpus_equal(mask, allbutself))
235 send_IPI_allbutself(CALL_FUNCTION_VECTOR);
236 else
237 send_IPI_mask(mask, CALL_FUNCTION_VECTOR);
238
239 /* Wait for response */
240 while (atomic_read(&data.started) != cpus)
241 cpu_relax();
242
243 if (wait)
244 while (atomic_read(&data.finished) != cpus)
245 cpu_relax();
246 spin_unlock(&call_lock);
247
248 return 0;
249}
250
251static void stop_this_cpu(void *dummy)
252{
253 local_irq_disable();
254 /*
255 * Remove this CPU:
256 */
257 cpu_clear(smp_processor_id(), cpu_online_map);
258 disable_local_APIC();
259 if (hlt_works(smp_processor_id()))
260 for (;;) halt();
261 for (;;);
262}
263
264/*
265 * this function calls the 'stop' function on all other CPUs in the system.
266 */
267
268static void native_smp_send_stop(void)
269{
270 int nolock;
271 unsigned long flags;
272
273 if (reboot_force)
274 return;
275
276 /* Don't deadlock on the call lock in panic */
277 nolock = !spin_trylock(&call_lock);
278 local_irq_save(flags);
279 __smp_call_function(stop_this_cpu, NULL, 0, 0);
280 if (!nolock)
281 spin_unlock(&call_lock);
282 disable_local_APIC();
283 local_irq_restore(flags);
284}
285
286/*
287 * Reschedule call back. Nothing to do,
288 * all the work is done automatically when
289 * we return from the interrupt.
290 */
291void smp_reschedule_interrupt(struct pt_regs *regs)
292{
293 ack_APIC_irq();
294#ifdef CONFIG_X86_32
295 __get_cpu_var(irq_stat).irq_resched_count++;
296#else
297 add_pda(irq_resched_count, 1);
298#endif
299}
300
301void smp_call_function_interrupt(struct pt_regs *regs)
302{
303 void (*func) (void *info) = call_data->func;
304 void *info = call_data->info;
305 int wait = call_data->wait;
306
307 ack_APIC_irq();
308 /*
309 * Notify initiating CPU that I've grabbed the data and am
310 * about to execute the function
311 */
312 mb();
313 atomic_inc(&call_data->started);
314 /*
315 * At this point the info structure may be out of scope unless wait==1
316 */
317 irq_enter();
318 (*func)(info);
319#ifdef CONFIG_X86_32
320 __get_cpu_var(irq_stat).irq_call_count++;
321#else
322 add_pda(irq_call_count, 1);
323#endif
324 irq_exit();
325
326 if (wait) {
327 mb();
328 atomic_inc(&call_data->finished);
329 }
330}
331
332struct smp_ops smp_ops = {
333 .smp_prepare_boot_cpu = native_smp_prepare_boot_cpu,
334 .smp_prepare_cpus = native_smp_prepare_cpus,
335 .cpu_up = native_cpu_up,
336 .smp_cpus_done = native_smp_cpus_done,
337
338 .smp_send_stop = native_smp_send_stop,
339 .smp_send_reschedule = native_smp_send_reschedule,
340 .smp_call_function_mask = native_smp_call_function_mask,
341};
342EXPORT_SYMBOL_GPL(smp_ops);
343
diff --git a/arch/x86/kernel/smp_32.c b/arch/x86/kernel/smp_32.c
deleted file mode 100644
index dc0cde9d16fb..000000000000
--- a/arch/x86/kernel/smp_32.c
+++ /dev/null
@@ -1,712 +0,0 @@
1/*
2 * Intel SMP support routines.
3 *
4 * (c) 1995 Alan Cox, Building #3 <alan@redhat.com>
5 * (c) 1998-99, 2000 Ingo Molnar <mingo@redhat.com>
6 *
7 * This code is released under the GNU General Public License version 2 or
8 * later.
9 */
10
11#include <linux/init.h>
12
13#include <linux/mm.h>
14#include <linux/delay.h>
15#include <linux/spinlock.h>
16#include <linux/kernel_stat.h>
17#include <linux/mc146818rtc.h>
18#include <linux/cache.h>
19#include <linux/interrupt.h>
20#include <linux/cpu.h>
21#include <linux/module.h>
22
23#include <asm/mtrr.h>
24#include <asm/tlbflush.h>
25#include <asm/mmu_context.h>
26#include <mach_apic.h>
27
28/*
29 * Some notes on x86 processor bugs affecting SMP operation:
30 *
31 * Pentium, Pentium Pro, II, III (and all CPUs) have bugs.
32 * The Linux implications for SMP are handled as follows:
33 *
34 * Pentium III / [Xeon]
35 * None of the E1AP-E3AP errata are visible to the user.
36 *
37 * E1AP. see PII A1AP
38 * E2AP. see PII A2AP
39 * E3AP. see PII A3AP
40 *
41 * Pentium II / [Xeon]
42 * None of the A1AP-A3AP errata are visible to the user.
43 *
44 * A1AP. see PPro 1AP
45 * A2AP. see PPro 2AP
46 * A3AP. see PPro 7AP
47 *
48 * Pentium Pro
49 * None of 1AP-9AP errata are visible to the normal user,
50 * except occasional delivery of 'spurious interrupt' as trap #15.
51 * This is very rare and a non-problem.
52 *
53 * 1AP. Linux maps APIC as non-cacheable
54 * 2AP. worked around in hardware
55 * 3AP. fixed in C0 and above steppings microcode update.
56 * Linux does not use excessive STARTUP_IPIs.
57 * 4AP. worked around in hardware
58 * 5AP. symmetric IO mode (normal Linux operation) not affected.
59 * 'noapic' mode has vector 0xf filled out properly.
60 * 6AP. 'noapic' mode might be affected - fixed in later steppings
61 * 7AP. We do not assume writes to the LVT deassering IRQs
62 * 8AP. We do not enable low power mode (deep sleep) during MP bootup
63 * 9AP. We do not use mixed mode
64 *
65 * Pentium
66 * There is a marginal case where REP MOVS on 100MHz SMP
67 * machines with B stepping processors can fail. XXX should provide
68 * an L1cache=Writethrough or L1cache=off option.
69 *
70 * B stepping CPUs may hang. There are hardware work arounds
71 * for this. We warn about it in case your board doesn't have the work
72 * arounds. Basically that's so I can tell anyone with a B stepping
73 * CPU and SMP problems "tough".
74 *
75 * Specific items [From Pentium Processor Specification Update]
76 *
77 * 1AP. Linux doesn't use remote read
78 * 2AP. Linux doesn't trust APIC errors
79 * 3AP. We work around this
80 * 4AP. Linux never generated 3 interrupts of the same priority
81 * to cause a lost local interrupt.
82 * 5AP. Remote read is never used
83 * 6AP. not affected - worked around in hardware
84 * 7AP. not affected - worked around in hardware
85 * 8AP. worked around in hardware - we get explicit CS errors if not
86 * 9AP. only 'noapic' mode affected. Might generate spurious
87 * interrupts, we log only the first one and count the
88 * rest silently.
89 * 10AP. not affected - worked around in hardware
90 * 11AP. Linux reads the APIC between writes to avoid this, as per
91 * the documentation. Make sure you preserve this as it affects
92 * the C stepping chips too.
93 * 12AP. not affected - worked around in hardware
94 * 13AP. not affected - worked around in hardware
95 * 14AP. we always deassert INIT during bootup
96 * 15AP. not affected - worked around in hardware
97 * 16AP. not affected - worked around in hardware
98 * 17AP. not affected - worked around in hardware
99 * 18AP. not affected - worked around in hardware
100 * 19AP. not affected - worked around in BIOS
101 *
102 * If this sounds worrying believe me these bugs are either ___RARE___,
103 * or are signal timing bugs worked around in hardware and there's
104 * about nothing of note with C stepping upwards.
105 */
106
107DEFINE_PER_CPU(struct tlb_state, cpu_tlbstate) ____cacheline_aligned = { &init_mm, 0, };
108
109/*
110 * the following functions deal with sending IPIs between CPUs.
111 *
112 * We use 'broadcast', CPU->CPU IPIs and self-IPIs too.
113 */
114
115static inline int __prepare_ICR (unsigned int shortcut, int vector)
116{
117 unsigned int icr = shortcut | APIC_DEST_LOGICAL;
118
119 switch (vector) {
120 default:
121 icr |= APIC_DM_FIXED | vector;
122 break;
123 case NMI_VECTOR:
124 icr |= APIC_DM_NMI;
125 break;
126 }
127 return icr;
128}
129
130static inline int __prepare_ICR2 (unsigned int mask)
131{
132 return SET_APIC_DEST_FIELD(mask);
133}
134
135void __send_IPI_shortcut(unsigned int shortcut, int vector)
136{
137 /*
138 * Subtle. In the case of the 'never do double writes' workaround
139 * we have to lock out interrupts to be safe. As we don't care
140 * of the value read we use an atomic rmw access to avoid costly
141 * cli/sti. Otherwise we use an even cheaper single atomic write
142 * to the APIC.
143 */
144 unsigned int cfg;
145
146 /*
147 * Wait for idle.
148 */
149 apic_wait_icr_idle();
150
151 /*
152 * No need to touch the target chip field
153 */
154 cfg = __prepare_ICR(shortcut, vector);
155
156 /*
157 * Send the IPI. The write to APIC_ICR fires this off.
158 */
159 apic_write_around(APIC_ICR, cfg);
160}
161
162void send_IPI_self(int vector)
163{
164 __send_IPI_shortcut(APIC_DEST_SELF, vector);
165}
166
167/*
168 * This is used to send an IPI with no shorthand notation (the destination is
169 * specified in bits 56 to 63 of the ICR).
170 */
171static inline void __send_IPI_dest_field(unsigned long mask, int vector)
172{
173 unsigned long cfg;
174
175 /*
176 * Wait for idle.
177 */
178 if (unlikely(vector == NMI_VECTOR))
179 safe_apic_wait_icr_idle();
180 else
181 apic_wait_icr_idle();
182
183 /*
184 * prepare target chip field
185 */
186 cfg = __prepare_ICR2(mask);
187 apic_write_around(APIC_ICR2, cfg);
188
189 /*
190 * program the ICR
191 */
192 cfg = __prepare_ICR(0, vector);
193
194 /*
195 * Send the IPI. The write to APIC_ICR fires this off.
196 */
197 apic_write_around(APIC_ICR, cfg);
198}
199
200/*
201 * This is only used on smaller machines.
202 */
203void send_IPI_mask_bitmask(cpumask_t cpumask, int vector)
204{
205 unsigned long mask = cpus_addr(cpumask)[0];
206 unsigned long flags;
207
208 local_irq_save(flags);
209 WARN_ON(mask & ~cpus_addr(cpu_online_map)[0]);
210 __send_IPI_dest_field(mask, vector);
211 local_irq_restore(flags);
212}
213
214void send_IPI_mask_sequence(cpumask_t mask, int vector)
215{
216 unsigned long flags;
217 unsigned int query_cpu;
218
219 /*
220 * Hack. The clustered APIC addressing mode doesn't allow us to send
221 * to an arbitrary mask, so I do a unicasts to each CPU instead. This
222 * should be modified to do 1 message per cluster ID - mbligh
223 */
224
225 local_irq_save(flags);
226 for_each_possible_cpu(query_cpu) {
227 if (cpu_isset(query_cpu, mask)) {
228 __send_IPI_dest_field(cpu_to_logical_apicid(query_cpu),
229 vector);
230 }
231 }
232 local_irq_restore(flags);
233}
234
235#include <mach_ipi.h> /* must come after the send_IPI functions above for inlining */
236
237/*
238 * Smarter SMP flushing macros.
239 * c/o Linus Torvalds.
240 *
241 * These mean you can really definitely utterly forget about
242 * writing to user space from interrupts. (Its not allowed anyway).
243 *
244 * Optimizations Manfred Spraul <manfred@colorfullife.com>
245 */
246
247static cpumask_t flush_cpumask;
248static struct mm_struct * flush_mm;
249static unsigned long flush_va;
250static DEFINE_SPINLOCK(tlbstate_lock);
251
252/*
253 * We cannot call mmdrop() because we are in interrupt context,
254 * instead update mm->cpu_vm_mask.
255 *
256 * We need to reload %cr3 since the page tables may be going
257 * away from under us..
258 */
259void leave_mm(int cpu)
260{
261 if (per_cpu(cpu_tlbstate, cpu).state == TLBSTATE_OK)
262 BUG();
263 cpu_clear(cpu, per_cpu(cpu_tlbstate, cpu).active_mm->cpu_vm_mask);
264 load_cr3(swapper_pg_dir);
265}
266EXPORT_SYMBOL_GPL(leave_mm);
267
268/*
269 *
270 * The flush IPI assumes that a thread switch happens in this order:
271 * [cpu0: the cpu that switches]
272 * 1) switch_mm() either 1a) or 1b)
273 * 1a) thread switch to a different mm
274 * 1a1) cpu_clear(cpu, old_mm->cpu_vm_mask);
275 * Stop ipi delivery for the old mm. This is not synchronized with
276 * the other cpus, but smp_invalidate_interrupt ignore flush ipis
277 * for the wrong mm, and in the worst case we perform a superfluous
278 * tlb flush.
279 * 1a2) set cpu_tlbstate to TLBSTATE_OK
280 * Now the smp_invalidate_interrupt won't call leave_mm if cpu0
281 * was in lazy tlb mode.
282 * 1a3) update cpu_tlbstate[].active_mm
283 * Now cpu0 accepts tlb flushes for the new mm.
284 * 1a4) cpu_set(cpu, new_mm->cpu_vm_mask);
285 * Now the other cpus will send tlb flush ipis.
286 * 1a4) change cr3.
287 * 1b) thread switch without mm change
288 * cpu_tlbstate[].active_mm is correct, cpu0 already handles
289 * flush ipis.
290 * 1b1) set cpu_tlbstate to TLBSTATE_OK
291 * 1b2) test_and_set the cpu bit in cpu_vm_mask.
292 * Atomically set the bit [other cpus will start sending flush ipis],
293 * and test the bit.
294 * 1b3) if the bit was 0: leave_mm was called, flush the tlb.
295 * 2) switch %%esp, ie current
296 *
297 * The interrupt must handle 2 special cases:
298 * - cr3 is changed before %%esp, ie. it cannot use current->{active_,}mm.
299 * - the cpu performs speculative tlb reads, i.e. even if the cpu only
300 * runs in kernel space, the cpu could load tlb entries for user space
301 * pages.
302 *
303 * The good news is that cpu_tlbstate is local to each cpu, no
304 * write/read ordering problems.
305 */
306
307/*
308 * TLB flush IPI:
309 *
310 * 1) Flush the tlb entries if the cpu uses the mm that's being flushed.
311 * 2) Leave the mm if we are in the lazy tlb mode.
312 */
313
314void smp_invalidate_interrupt(struct pt_regs *regs)
315{
316 unsigned long cpu;
317
318 cpu = get_cpu();
319
320 if (!cpu_isset(cpu, flush_cpumask))
321 goto out;
322 /*
323 * This was a BUG() but until someone can quote me the
324 * line from the intel manual that guarantees an IPI to
325 * multiple CPUs is retried _only_ on the erroring CPUs
326 * its staying as a return
327 *
328 * BUG();
329 */
330
331 if (flush_mm == per_cpu(cpu_tlbstate, cpu).active_mm) {
332 if (per_cpu(cpu_tlbstate, cpu).state == TLBSTATE_OK) {
333 if (flush_va == TLB_FLUSH_ALL)
334 local_flush_tlb();
335 else
336 __flush_tlb_one(flush_va);
337 } else
338 leave_mm(cpu);
339 }
340 ack_APIC_irq();
341 smp_mb__before_clear_bit();
342 cpu_clear(cpu, flush_cpumask);
343 smp_mb__after_clear_bit();
344out:
345 put_cpu_no_resched();
346 __get_cpu_var(irq_stat).irq_tlb_count++;
347}
348
349void native_flush_tlb_others(const cpumask_t *cpumaskp, struct mm_struct *mm,
350 unsigned long va)
351{
352 cpumask_t cpumask = *cpumaskp;
353
354 /*
355 * A couple of (to be removed) sanity checks:
356 *
357 * - current CPU must not be in mask
358 * - mask must exist :)
359 */
360 BUG_ON(cpus_empty(cpumask));
361 BUG_ON(cpu_isset(smp_processor_id(), cpumask));
362 BUG_ON(!mm);
363
364#ifdef CONFIG_HOTPLUG_CPU
365 /* If a CPU which we ran on has gone down, OK. */
366 cpus_and(cpumask, cpumask, cpu_online_map);
367 if (unlikely(cpus_empty(cpumask)))
368 return;
369#endif
370
371 /*
372 * i'm not happy about this global shared spinlock in the
373 * MM hot path, but we'll see how contended it is.
374 * AK: x86-64 has a faster method that could be ported.
375 */
376 spin_lock(&tlbstate_lock);
377
378 flush_mm = mm;
379 flush_va = va;
380 cpus_or(flush_cpumask, cpumask, flush_cpumask);
381 /*
382 * We have to send the IPI only to
383 * CPUs affected.
384 */
385 send_IPI_mask(cpumask, INVALIDATE_TLB_VECTOR);
386
387 while (!cpus_empty(flush_cpumask))
388 /* nothing. lockup detection does not belong here */
389 cpu_relax();
390
391 flush_mm = NULL;
392 flush_va = 0;
393 spin_unlock(&tlbstate_lock);
394}
395
396void flush_tlb_current_task(void)
397{
398 struct mm_struct *mm = current->mm;
399 cpumask_t cpu_mask;
400
401 preempt_disable();
402 cpu_mask = mm->cpu_vm_mask;
403 cpu_clear(smp_processor_id(), cpu_mask);
404
405 local_flush_tlb();
406 if (!cpus_empty(cpu_mask))
407 flush_tlb_others(cpu_mask, mm, TLB_FLUSH_ALL);
408 preempt_enable();
409}
410
411void flush_tlb_mm (struct mm_struct * mm)
412{
413 cpumask_t cpu_mask;
414
415 preempt_disable();
416 cpu_mask = mm->cpu_vm_mask;
417 cpu_clear(smp_processor_id(), cpu_mask);
418
419 if (current->active_mm == mm) {
420 if (current->mm)
421 local_flush_tlb();
422 else
423 leave_mm(smp_processor_id());
424 }
425 if (!cpus_empty(cpu_mask))
426 flush_tlb_others(cpu_mask, mm, TLB_FLUSH_ALL);
427
428 preempt_enable();
429}
430
431void flush_tlb_page(struct vm_area_struct * vma, unsigned long va)
432{
433 struct mm_struct *mm = vma->vm_mm;
434 cpumask_t cpu_mask;
435
436 preempt_disable();
437 cpu_mask = mm->cpu_vm_mask;
438 cpu_clear(smp_processor_id(), cpu_mask);
439
440 if (current->active_mm == mm) {
441 if(current->mm)
442 __flush_tlb_one(va);
443 else
444 leave_mm(smp_processor_id());
445 }
446
447 if (!cpus_empty(cpu_mask))
448 flush_tlb_others(cpu_mask, mm, va);
449
450 preempt_enable();
451}
452EXPORT_SYMBOL(flush_tlb_page);
453
454static void do_flush_tlb_all(void* info)
455{
456 unsigned long cpu = smp_processor_id();
457
458 __flush_tlb_all();
459 if (per_cpu(cpu_tlbstate, cpu).state == TLBSTATE_LAZY)
460 leave_mm(cpu);
461}
462
463void flush_tlb_all(void)
464{
465 on_each_cpu(do_flush_tlb_all, NULL, 1, 1);
466}
467
468/*
469 * this function sends a 'reschedule' IPI to another CPU.
470 * it goes straight through and wastes no time serializing
471 * anything. Worst case is that we lose a reschedule ...
472 */
473static void native_smp_send_reschedule(int cpu)
474{
475 WARN_ON(cpu_is_offline(cpu));
476 send_IPI_mask(cpumask_of_cpu(cpu), RESCHEDULE_VECTOR);
477}
478
479/*
480 * Structure and data for smp_call_function(). This is designed to minimise
481 * static memory requirements. It also looks cleaner.
482 */
483static DEFINE_SPINLOCK(call_lock);
484
485struct call_data_struct {
486 void (*func) (void *info);
487 void *info;
488 atomic_t started;
489 atomic_t finished;
490 int wait;
491};
492
493void lock_ipi_call_lock(void)
494{
495 spin_lock_irq(&call_lock);
496}
497
498void unlock_ipi_call_lock(void)
499{
500 spin_unlock_irq(&call_lock);
501}
502
503static struct call_data_struct *call_data;
504
505static void __smp_call_function(void (*func) (void *info), void *info,
506 int nonatomic, int wait)
507{
508 struct call_data_struct data;
509 int cpus = num_online_cpus() - 1;
510
511 if (!cpus)
512 return;
513
514 data.func = func;
515 data.info = info;
516 atomic_set(&data.started, 0);
517 data.wait = wait;
518 if (wait)
519 atomic_set(&data.finished, 0);
520
521 call_data = &data;
522 mb();
523
524 /* Send a message to all other CPUs and wait for them to respond */
525 send_IPI_allbutself(CALL_FUNCTION_VECTOR);
526
527 /* Wait for response */
528 while (atomic_read(&data.started) != cpus)
529 cpu_relax();
530
531 if (wait)
532 while (atomic_read(&data.finished) != cpus)
533 cpu_relax();
534}
535
536
537/**
538 * smp_call_function_mask(): Run a function on a set of other CPUs.
539 * @mask: The set of cpus to run on. Must not include the current cpu.
540 * @func: The function to run. This must be fast and non-blocking.
541 * @info: An arbitrary pointer to pass to the function.
542 * @wait: If true, wait (atomically) until function has completed on other CPUs.
543 *
544 * Returns 0 on success, else a negative status code.
545 *
546 * If @wait is true, then returns once @func has returned; otherwise
547 * it returns just before the target cpu calls @func.
548 *
549 * You must not call this function with disabled interrupts or from a
550 * hardware interrupt handler or from a bottom half handler.
551 */
552static int
553native_smp_call_function_mask(cpumask_t mask,
554 void (*func)(void *), void *info,
555 int wait)
556{
557 struct call_data_struct data;
558 cpumask_t allbutself;
559 int cpus;
560
561 /* Can deadlock when called with interrupts disabled */
562 WARN_ON(irqs_disabled());
563
564 /* Holding any lock stops cpus from going down. */
565 spin_lock(&call_lock);
566
567 allbutself = cpu_online_map;
568 cpu_clear(smp_processor_id(), allbutself);
569
570 cpus_and(mask, mask, allbutself);
571 cpus = cpus_weight(mask);
572
573 if (!cpus) {
574 spin_unlock(&call_lock);
575 return 0;
576 }
577
578 data.func = func;
579 data.info = info;
580 atomic_set(&data.started, 0);
581 data.wait = wait;
582 if (wait)
583 atomic_set(&data.finished, 0);
584
585 call_data = &data;
586 mb();
587
588 /* Send a message to other CPUs */
589 if (cpus_equal(mask, allbutself))
590 send_IPI_allbutself(CALL_FUNCTION_VECTOR);
591 else
592 send_IPI_mask(mask, CALL_FUNCTION_VECTOR);
593
594 /* Wait for response */
595 while (atomic_read(&data.started) != cpus)
596 cpu_relax();
597
598 if (wait)
599 while (atomic_read(&data.finished) != cpus)
600 cpu_relax();
601 spin_unlock(&call_lock);
602
603 return 0;
604}
605
606static void stop_this_cpu (void * dummy)
607{
608 local_irq_disable();
609 /*
610 * Remove this CPU:
611 */
612 cpu_clear(smp_processor_id(), cpu_online_map);
613 disable_local_APIC();
614 if (cpu_data(smp_processor_id()).hlt_works_ok)
615 for(;;) halt();
616 for (;;);
617}
618
619/*
620 * this function calls the 'stop' function on all other CPUs in the system.
621 */
622
623static void native_smp_send_stop(void)
624{
625 /* Don't deadlock on the call lock in panic */
626 int nolock = !spin_trylock(&call_lock);
627 unsigned long flags;
628
629 local_irq_save(flags);
630 __smp_call_function(stop_this_cpu, NULL, 0, 0);
631 if (!nolock)
632 spin_unlock(&call_lock);
633 disable_local_APIC();
634 local_irq_restore(flags);
635}
636
637/*
638 * Reschedule call back. Nothing to do,
639 * all the work is done automatically when
640 * we return from the interrupt.
641 */
642void smp_reschedule_interrupt(struct pt_regs *regs)
643{
644 ack_APIC_irq();
645 __get_cpu_var(irq_stat).irq_resched_count++;
646}
647
648void smp_call_function_interrupt(struct pt_regs *regs)
649{
650 void (*func) (void *info) = call_data->func;
651 void *info = call_data->info;
652 int wait = call_data->wait;
653
654 ack_APIC_irq();
655 /*
656 * Notify initiating CPU that I've grabbed the data and am
657 * about to execute the function
658 */
659 mb();
660 atomic_inc(&call_data->started);
661 /*
662 * At this point the info structure may be out of scope unless wait==1
663 */
664 irq_enter();
665 (*func)(info);
666 __get_cpu_var(irq_stat).irq_call_count++;
667 irq_exit();
668
669 if (wait) {
670 mb();
671 atomic_inc(&call_data->finished);
672 }
673}
674
675static int convert_apicid_to_cpu(int apic_id)
676{
677 int i;
678
679 for_each_possible_cpu(i) {
680 if (per_cpu(x86_cpu_to_apicid, i) == apic_id)
681 return i;
682 }
683 return -1;
684}
685
686int safe_smp_processor_id(void)
687{
688 int apicid, cpuid;
689
690 if (!boot_cpu_has(X86_FEATURE_APIC))
691 return 0;
692
693 apicid = hard_smp_processor_id();
694 if (apicid == BAD_APICID)
695 return 0;
696
697 cpuid = convert_apicid_to_cpu(apicid);
698
699 return cpuid >= 0 ? cpuid : 0;
700}
701
702struct smp_ops smp_ops = {
703 .smp_prepare_boot_cpu = native_smp_prepare_boot_cpu,
704 .smp_prepare_cpus = native_smp_prepare_cpus,
705 .cpu_up = native_cpu_up,
706 .smp_cpus_done = native_smp_cpus_done,
707
708 .smp_send_stop = native_smp_send_stop,
709 .smp_send_reschedule = native_smp_send_reschedule,
710 .smp_call_function_mask = native_smp_call_function_mask,
711};
712EXPORT_SYMBOL_GPL(smp_ops);
diff --git a/arch/x86/kernel/smpboot_32.c b/arch/x86/kernel/smpboot.c
index 579b9b740c7c..e6abe8a49b1f 100644
--- a/arch/x86/kernel/smpboot_32.c
+++ b/arch/x86/kernel/smpboot.c
@@ -3,6 +3,7 @@
3 * 3 *
4 * (c) 1995 Alan Cox, Building #3 <alan@redhat.com> 4 * (c) 1995 Alan Cox, Building #3 <alan@redhat.com>
5 * (c) 1998, 1999, 2000 Ingo Molnar <mingo@redhat.com> 5 * (c) 1998, 1999, 2000 Ingo Molnar <mingo@redhat.com>
6 * Copyright 2001 Andi Kleen, SuSE Labs.
6 * 7 *
7 * Much of the core SMP work is based on previous work by Thomas Radke, to 8 * Much of the core SMP work is based on previous work by Thomas Radke, to
8 * whom a great many thanks are extended. 9 * whom a great many thanks are extended.
@@ -29,53 +30,90 @@
29 * Ingo Molnar : various cleanups and rewrites 30 * Ingo Molnar : various cleanups and rewrites
30 * Tigran Aivazian : fixed "0.00 in /proc/uptime on SMP" bug. 31 * Tigran Aivazian : fixed "0.00 in /proc/uptime on SMP" bug.
31 * Maciej W. Rozycki : Bits for genuine 82489DX APICs 32 * Maciej W. Rozycki : Bits for genuine 82489DX APICs
33 * Andi Kleen : Changed for SMP boot into long mode.
32 * Martin J. Bligh : Added support for multi-quad systems 34 * Martin J. Bligh : Added support for multi-quad systems
33 * Dave Jones : Report invalid combinations of Athlon CPUs. 35 * Dave Jones : Report invalid combinations of Athlon CPUs.
34* Rusty Russell : Hacked into shape for new "hotplug" boot process. */ 36 * Rusty Russell : Hacked into shape for new "hotplug" boot process.
37 * Andi Kleen : Converted to new state machine.
38 * Ashok Raj : CPU hotplug support
39 * Glauber Costa : i386 and x86_64 integration
40 */
35 41
36#include <linux/module.h>
37#include <linux/init.h> 42#include <linux/init.h>
38#include <linux/kernel.h> 43#include <linux/smp.h>
39 44#include <linux/module.h>
40#include <linux/mm.h>
41#include <linux/sched.h> 45#include <linux/sched.h>
42#include <linux/kernel_stat.h>
43#include <linux/bootmem.h>
44#include <linux/notifier.h>
45#include <linux/cpu.h>
46#include <linux/percpu.h> 46#include <linux/percpu.h>
47#include <linux/bootmem.h>
48#include <linux/err.h>
47#include <linux/nmi.h> 49#include <linux/nmi.h>
48 50
49#include <linux/delay.h> 51#include <asm/acpi.h>
50#include <linux/mc146818rtc.h>
51#include <asm/tlbflush.h>
52#include <asm/desc.h> 52#include <asm/desc.h>
53#include <asm/arch_hooks.h>
54#include <asm/nmi.h> 53#include <asm/nmi.h>
54#include <asm/irq.h>
55#include <asm/smp.h>
56#include <asm/trampoline.h>
57#include <asm/cpu.h>
58#include <asm/numa.h>
59#include <asm/pgtable.h>
60#include <asm/tlbflush.h>
61#include <asm/mtrr.h>
62#include <asm/nmi.h>
63#include <asm/vmi.h>
64#include <linux/mc146818rtc.h>
55 65
56#include <mach_apic.h> 66#include <mach_apic.h>
57#include <mach_wakecpu.h> 67#include <mach_wakecpu.h>
58#include <smpboot_hooks.h> 68#include <smpboot_hooks.h>
59#include <asm/vmi.h>
60#include <asm/mtrr.h>
61 69
62/* Set if we find a B stepping CPU */ 70/*
63static int __cpuinitdata smp_b_stepping; 71 * FIXME: For x86_64, those are defined in other files. But moving them here,
72 * would make the setup areas dependent on smp, which is a loss. When we
73 * integrate apic between arches, we can probably do a better job, but
74 * right now, they'll stay here -- glommer
75 */
76
77/* which logical CPU number maps to which CPU (physical APIC ID) */
78u16 x86_cpu_to_apicid_init[NR_CPUS] __initdata =
79 { [0 ... NR_CPUS-1] = BAD_APICID };
80void *x86_cpu_to_apicid_early_ptr;
81
82u16 x86_bios_cpu_apicid_init[NR_CPUS] __initdata
83 = { [0 ... NR_CPUS-1] = BAD_APICID };
84void *x86_bios_cpu_apicid_early_ptr;
85
86#ifdef CONFIG_X86_32
87u8 apicid_2_node[MAX_APICID];
88#endif
89
90/* State of each CPU */
91DEFINE_PER_CPU(int, cpu_state) = { 0 };
92
93/* Store all idle threads, this can be reused instead of creating
94* a new thread. Also avoids complicated thread destroy functionality
95* for idle threads.
96*/
97#ifdef CONFIG_HOTPLUG_CPU
98/*
99 * Needed only for CONFIG_HOTPLUG_CPU because __cpuinitdata is
100 * removed after init for !CONFIG_HOTPLUG_CPU.
101 */
102static DEFINE_PER_CPU(struct task_struct *, idle_thread_array);
103#define get_idle_for_cpu(x) (per_cpu(idle_thread_array, x))
104#define set_idle_for_cpu(x, p) (per_cpu(idle_thread_array, x) = (p))
105#else
106struct task_struct *idle_thread_array[NR_CPUS] __cpuinitdata ;
107#define get_idle_for_cpu(x) (idle_thread_array[(x)])
108#define set_idle_for_cpu(x, p) (idle_thread_array[(x)] = (p))
109#endif
64 110
65/* Number of siblings per CPU package */ 111/* Number of siblings per CPU package */
66int smp_num_siblings = 1; 112int smp_num_siblings = 1;
67EXPORT_SYMBOL(smp_num_siblings); 113EXPORT_SYMBOL(smp_num_siblings);
68 114
69/* Last level cache ID of each logical CPU */ 115/* Last level cache ID of each logical CPU */
70DEFINE_PER_CPU(u8, cpu_llc_id) = BAD_APICID; 116DEFINE_PER_CPU(u16, cpu_llc_id) = BAD_APICID;
71
72/* representing HT siblings of each logical CPU */
73DEFINE_PER_CPU(cpumask_t, cpu_sibling_map);
74EXPORT_PER_CPU_SYMBOL(cpu_sibling_map);
75
76/* representing HT and core siblings of each logical CPU */
77DEFINE_PER_CPU(cpumask_t, cpu_core_map);
78EXPORT_PER_CPU_SYMBOL(cpu_core_map);
79 117
80/* bitmap of online cpus */ 118/* bitmap of online cpus */
81cpumask_t cpu_online_map __read_mostly; 119cpumask_t cpu_online_map __read_mostly;
@@ -85,126 +123,94 @@ cpumask_t cpu_callin_map;
85cpumask_t cpu_callout_map; 123cpumask_t cpu_callout_map;
86cpumask_t cpu_possible_map; 124cpumask_t cpu_possible_map;
87EXPORT_SYMBOL(cpu_possible_map); 125EXPORT_SYMBOL(cpu_possible_map);
88static cpumask_t smp_commenced_mask; 126
127/* representing HT siblings of each logical CPU */
128DEFINE_PER_CPU(cpumask_t, cpu_sibling_map);
129EXPORT_PER_CPU_SYMBOL(cpu_sibling_map);
130
131/* representing HT and core siblings of each logical CPU */
132DEFINE_PER_CPU(cpumask_t, cpu_core_map);
133EXPORT_PER_CPU_SYMBOL(cpu_core_map);
89 134
90/* Per CPU bogomips and other parameters */ 135/* Per CPU bogomips and other parameters */
91DEFINE_PER_CPU_SHARED_ALIGNED(struct cpuinfo_x86, cpu_info); 136DEFINE_PER_CPU_SHARED_ALIGNED(struct cpuinfo_x86, cpu_info);
92EXPORT_PER_CPU_SYMBOL(cpu_info); 137EXPORT_PER_CPU_SYMBOL(cpu_info);
93 138
94/* which logical CPU number maps to which CPU (physical APIC ID) */ 139static atomic_t init_deasserted;
95u8 x86_cpu_to_apicid_init[NR_CPUS] __initdata =
96 { [0 ... NR_CPUS-1] = BAD_APICID };
97void *x86_cpu_to_apicid_early_ptr;
98DEFINE_PER_CPU(u8, x86_cpu_to_apicid) = BAD_APICID;
99EXPORT_PER_CPU_SYMBOL(x86_cpu_to_apicid);
100
101u8 apicid_2_node[MAX_APICID];
102 140
103/* 141static int boot_cpu_logical_apicid;
104 * Trampoline 80x86 program as an array.
105 */
106 142
107extern const unsigned char trampoline_data []; 143/* representing cpus for which sibling maps can be computed */
108extern const unsigned char trampoline_end []; 144static cpumask_t cpu_sibling_setup_map;
109static unsigned char *trampoline_base;
110 145
111static void map_cpu_to_logical_apicid(void); 146/* Set if we find a B stepping CPU */
147int __cpuinitdata smp_b_stepping;
112 148
113/* State of each CPU. */ 149#if defined(CONFIG_NUMA) && defined(CONFIG_X86_32)
114DEFINE_PER_CPU(int, cpu_state) = { 0 };
115 150
116/* 151/* which logical CPUs are on which nodes */
117 * Currently trivial. Write the real->protected mode 152cpumask_t node_to_cpumask_map[MAX_NUMNODES] __read_mostly =
118 * bootstrap into the page concerned. The caller 153 { [0 ... MAX_NUMNODES-1] = CPU_MASK_NONE };
119 * has made sure it's suitably aligned. 154EXPORT_SYMBOL(node_to_cpumask_map);
120 */ 155/* which node each logical CPU is on */
156int cpu_to_node_map[NR_CPUS] __read_mostly = { [0 ... NR_CPUS-1] = 0 };
157EXPORT_SYMBOL(cpu_to_node_map);
121 158
122static unsigned long __cpuinit setup_trampoline(void) 159/* set up a mapping between cpu and node. */
160static void map_cpu_to_node(int cpu, int node)
123{ 161{
124 memcpy(trampoline_base, trampoline_data, trampoline_end - trampoline_data); 162 printk(KERN_INFO "Mapping cpu %d to node %d\n", cpu, node);
125 return virt_to_phys(trampoline_base); 163 cpu_set(cpu, node_to_cpumask_map[node]);
164 cpu_to_node_map[cpu] = node;
126} 165}
127 166
128/* 167/* undo a mapping between cpu and node. */
129 * We are called very early to get the low memory for the 168static void unmap_cpu_to_node(int cpu)
130 * SMP bootup trampoline page.
131 */
132void __init smp_alloc_memory(void)
133{ 169{
134 trampoline_base = alloc_bootmem_low_pages(PAGE_SIZE); 170 int node;
135 /* 171
136 * Has to be in very low memory so we can execute 172 printk(KERN_INFO "Unmapping cpu %d from all nodes\n", cpu);
137 * real-mode AP code. 173 for (node = 0; node < MAX_NUMNODES; node++)
138 */ 174 cpu_clear(cpu, node_to_cpumask_map[node]);
139 if (__pa(trampoline_base) >= 0x9F000) 175 cpu_to_node_map[cpu] = 0;
140 BUG();
141} 176}
177#else /* !(CONFIG_NUMA && CONFIG_X86_32) */
178#define map_cpu_to_node(cpu, node) ({})
179#define unmap_cpu_to_node(cpu) ({})
180#endif
142 181
143/* 182#ifdef CONFIG_X86_32
144 * The bootstrap kernel entry code has set these up. Save them for 183u8 cpu_2_logical_apicid[NR_CPUS] __read_mostly =
145 * a given CPU 184 { [0 ... NR_CPUS-1] = BAD_APICID };
146 */
147 185
148void __cpuinit smp_store_cpu_info(int id) 186void map_cpu_to_logical_apicid(void)
149{ 187{
150 struct cpuinfo_x86 *c = &cpu_data(id); 188 int cpu = smp_processor_id();
151 189 int apicid = logical_smp_processor_id();
152 *c = boot_cpu_data; 190 int node = apicid_to_node(apicid);
153 c->cpu_index = id;
154 if (id!=0)
155 identify_secondary_cpu(c);
156 /*
157 * Mask B, Pentium, but not Pentium MMX
158 */
159 if (c->x86_vendor == X86_VENDOR_INTEL &&
160 c->x86 == 5 &&
161 c->x86_mask >= 1 && c->x86_mask <= 4 &&
162 c->x86_model <= 3)
163 /*
164 * Remember we have B step Pentia with bugs
165 */
166 smp_b_stepping = 1;
167
168 /*
169 * Certain Athlons might work (for various values of 'work') in SMP
170 * but they are not certified as MP capable.
171 */
172 if ((c->x86_vendor == X86_VENDOR_AMD) && (c->x86 == 6)) {
173
174 if (num_possible_cpus() == 1)
175 goto valid_k7;
176
177 /* Athlon 660/661 is valid. */
178 if ((c->x86_model==6) && ((c->x86_mask==0) || (c->x86_mask==1)))
179 goto valid_k7;
180
181 /* Duron 670 is valid */
182 if ((c->x86_model==7) && (c->x86_mask==0))
183 goto valid_k7;
184
185 /*
186 * Athlon 662, Duron 671, and Athlon >model 7 have capability bit.
187 * It's worth noting that the A5 stepping (662) of some Athlon XP's
188 * have the MP bit set.
189 * See http://www.heise.de/newsticker/data/jow-18.10.01-000 for more.
190 */
191 if (((c->x86_model==6) && (c->x86_mask>=2)) ||
192 ((c->x86_model==7) && (c->x86_mask>=1)) ||
193 (c->x86_model> 7))
194 if (cpu_has_mp)
195 goto valid_k7;
196 191
197 /* If we get here, it's not a certified SMP capable AMD system. */ 192 if (!node_online(node))
198 add_taint(TAINT_UNSAFE_SMP); 193 node = first_online_node;
199 }
200 194
201valid_k7: 195 cpu_2_logical_apicid[cpu] = apicid;
202 ; 196 map_cpu_to_node(cpu, node);
203} 197}
204 198
205static atomic_t init_deasserted; 199void unmap_cpu_to_logical_apicid(int cpu)
200{
201 cpu_2_logical_apicid[cpu] = BAD_APICID;
202 unmap_cpu_to_node(cpu);
203}
204#else
205#define unmap_cpu_to_logical_apicid(cpu) do {} while (0)
206#define map_cpu_to_logical_apicid() do {} while (0)
207#endif
206 208
207static void __cpuinit smp_callin(void) 209/*
210 * Report back to the Boot Processor.
211 * Running on AP.
212 */
213void __cpuinit smp_callin(void)
208{ 214{
209 int cpuid, phys_id; 215 int cpuid, phys_id;
210 unsigned long timeout; 216 unsigned long timeout;
@@ -220,12 +226,11 @@ static void __cpuinit smp_callin(void)
220 /* 226 /*
221 * (This works even if the APIC is not enabled.) 227 * (This works even if the APIC is not enabled.)
222 */ 228 */
223 phys_id = GET_APIC_ID(apic_read(APIC_ID)); 229 phys_id = GET_APIC_ID(read_apic_id());
224 cpuid = smp_processor_id(); 230 cpuid = smp_processor_id();
225 if (cpu_isset(cpuid, cpu_callin_map)) { 231 if (cpu_isset(cpuid, cpu_callin_map)) {
226 printk("huh, phys CPU#%d, CPU#%d already present??\n", 232 panic("%s: phys CPU#%d, CPU#%d already present??\n", __func__,
227 phys_id, cpuid); 233 phys_id, cpuid);
228 BUG();
229 } 234 }
230 Dprintk("CPU#%d (phys ID: %d) waiting for CALLOUT\n", cpuid, phys_id); 235 Dprintk("CPU#%d (phys ID: %d) waiting for CALLOUT\n", cpuid, phys_id);
231 236
@@ -247,13 +252,12 @@ static void __cpuinit smp_callin(void)
247 */ 252 */
248 if (cpu_isset(cpuid, cpu_callout_map)) 253 if (cpu_isset(cpuid, cpu_callout_map))
249 break; 254 break;
250 rep_nop(); 255 cpu_relax();
251 } 256 }
252 257
253 if (!time_before(jiffies, timeout)) { 258 if (!time_before(jiffies, timeout)) {
254 printk("BUG: CPU%d started up but did not get a callout!\n", 259 panic("%s: CPU%d started up but did not get a callout!\n",
255 cpuid); 260 __func__, cpuid);
256 BUG();
257 } 261 }
258 262
259 /* 263 /*
@@ -266,13 +270,19 @@ static void __cpuinit smp_callin(void)
266 Dprintk("CALLIN, before setup_local_APIC().\n"); 270 Dprintk("CALLIN, before setup_local_APIC().\n");
267 smp_callin_clear_local_apic(); 271 smp_callin_clear_local_apic();
268 setup_local_APIC(); 272 setup_local_APIC();
273 end_local_APIC_setup();
269 map_cpu_to_logical_apicid(); 274 map_cpu_to_logical_apicid();
270 275
271 /* 276 /*
272 * Get our bogomips. 277 * Get our bogomips.
278 *
279 * Need to enable IRQs because it can take longer and then
280 * the NMI watchdog might kill us.
273 */ 281 */
282 local_irq_enable();
274 calibrate_delay(); 283 calibrate_delay();
275 Dprintk("Stack at about %p\n",&cpuid); 284 local_irq_disable();
285 Dprintk("Stack at about %p\n", &cpuid);
276 286
277 /* 287 /*
278 * Save our processor parameters 288 * Save our processor parameters
@@ -285,91 +295,10 @@ static void __cpuinit smp_callin(void)
285 cpu_set(cpuid, cpu_callin_map); 295 cpu_set(cpuid, cpu_callin_map);
286} 296}
287 297
288static int cpucount;
289
290/* maps the cpu to the sched domain representing multi-core */
291cpumask_t cpu_coregroup_map(int cpu)
292{
293 struct cpuinfo_x86 *c = &cpu_data(cpu);
294 /*
295 * For perf, we return last level cache shared map.
296 * And for power savings, we return cpu_core_map
297 */
298 if (sched_mc_power_savings || sched_smt_power_savings)
299 return per_cpu(cpu_core_map, cpu);
300 else
301 return c->llc_shared_map;
302}
303
304/* representing cpus for which sibling maps can be computed */
305static cpumask_t cpu_sibling_setup_map;
306
307void __cpuinit set_cpu_sibling_map(int cpu)
308{
309 int i;
310 struct cpuinfo_x86 *c = &cpu_data(cpu);
311
312 cpu_set(cpu, cpu_sibling_setup_map);
313
314 if (smp_num_siblings > 1) {
315 for_each_cpu_mask(i, cpu_sibling_setup_map) {
316 if (c->phys_proc_id == cpu_data(i).phys_proc_id &&
317 c->cpu_core_id == cpu_data(i).cpu_core_id) {
318 cpu_set(i, per_cpu(cpu_sibling_map, cpu));
319 cpu_set(cpu, per_cpu(cpu_sibling_map, i));
320 cpu_set(i, per_cpu(cpu_core_map, cpu));
321 cpu_set(cpu, per_cpu(cpu_core_map, i));
322 cpu_set(i, c->llc_shared_map);
323 cpu_set(cpu, cpu_data(i).llc_shared_map);
324 }
325 }
326 } else {
327 cpu_set(cpu, per_cpu(cpu_sibling_map, cpu));
328 }
329
330 cpu_set(cpu, c->llc_shared_map);
331
332 if (current_cpu_data.x86_max_cores == 1) {
333 per_cpu(cpu_core_map, cpu) = per_cpu(cpu_sibling_map, cpu);
334 c->booted_cores = 1;
335 return;
336 }
337
338 for_each_cpu_mask(i, cpu_sibling_setup_map) {
339 if (per_cpu(cpu_llc_id, cpu) != BAD_APICID &&
340 per_cpu(cpu_llc_id, cpu) == per_cpu(cpu_llc_id, i)) {
341 cpu_set(i, c->llc_shared_map);
342 cpu_set(cpu, cpu_data(i).llc_shared_map);
343 }
344 if (c->phys_proc_id == cpu_data(i).phys_proc_id) {
345 cpu_set(i, per_cpu(cpu_core_map, cpu));
346 cpu_set(cpu, per_cpu(cpu_core_map, i));
347 /*
348 * Does this new cpu bringup a new core?
349 */
350 if (cpus_weight(per_cpu(cpu_sibling_map, cpu)) == 1) {
351 /*
352 * for each core in package, increment
353 * the booted_cores for this new cpu
354 */
355 if (first_cpu(per_cpu(cpu_sibling_map, i)) == i)
356 c->booted_cores++;
357 /*
358 * increment the core count for all
359 * the other cpus in this package
360 */
361 if (i != cpu)
362 cpu_data(i).booted_cores++;
363 } else if (i != cpu && !c->booted_cores)
364 c->booted_cores = cpu_data(i).booted_cores;
365 }
366 }
367}
368
369/* 298/*
370 * Activate a secondary processor. 299 * Activate a secondary processor.
371 */ 300 */
372static void __cpuinit start_secondary(void *unused) 301void __cpuinit start_secondary(void *unused)
373{ 302{
374 /* 303 /*
375 * Don't put *anything* before cpu_init(), SMP booting is too 304 * Don't put *anything* before cpu_init(), SMP booting is too
@@ -382,24 +311,19 @@ static void __cpuinit start_secondary(void *unused)
382 cpu_init(); 311 cpu_init();
383 preempt_disable(); 312 preempt_disable();
384 smp_callin(); 313 smp_callin();
385 while (!cpu_isset(smp_processor_id(), smp_commenced_mask)) 314
386 rep_nop(); 315 /* otherwise gcc will move up smp_processor_id before the cpu_init */
316 barrier();
387 /* 317 /*
388 * Check TSC synchronization with the BP: 318 * Check TSC synchronization with the BP:
389 */ 319 */
390 check_tsc_sync_target(); 320 check_tsc_sync_target();
391 321
392 setup_secondary_clock();
393 if (nmi_watchdog == NMI_IO_APIC) { 322 if (nmi_watchdog == NMI_IO_APIC) {
394 disable_8259A_irq(0); 323 disable_8259A_irq(0);
395 enable_NMI_through_LVT0(); 324 enable_NMI_through_LVT0();
396 enable_8259A_irq(0); 325 enable_8259A_irq(0);
397 } 326 }
398 /*
399 * low-memory mappings have been cleared, flush them from
400 * the local TLBs too.
401 */
402 local_flush_tlb();
403 327
404 /* This must be done before setting cpu_online_map */ 328 /* This must be done before setting cpu_online_map */
405 set_cpu_sibling_map(raw_smp_processor_id()); 329 set_cpu_sibling_map(raw_smp_processor_id());
@@ -414,17 +338,27 @@ static void __cpuinit start_secondary(void *unused)
414 * smp_call_function(). 338 * smp_call_function().
415 */ 339 */
416 lock_ipi_call_lock(); 340 lock_ipi_call_lock();
341#ifdef CONFIG_X86_64
342 spin_lock(&vector_lock);
343
344 /* Setup the per cpu irq handling data structures */
345 __setup_vector_irq(smp_processor_id());
346 /*
347 * Allow the master to continue.
348 */
349 spin_unlock(&vector_lock);
350#endif
417 cpu_set(smp_processor_id(), cpu_online_map); 351 cpu_set(smp_processor_id(), cpu_online_map);
418 unlock_ipi_call_lock(); 352 unlock_ipi_call_lock();
419 per_cpu(cpu_state, smp_processor_id()) = CPU_ONLINE; 353 per_cpu(cpu_state, smp_processor_id()) = CPU_ONLINE;
420 354
421 /* We can take interrupts now: we're officially "up". */ 355 setup_secondary_clock();
422 local_irq_enable();
423 356
424 wmb(); 357 wmb();
425 cpu_idle(); 358 cpu_idle();
426} 359}
427 360
361#ifdef CONFIG_X86_32
428/* 362/*
429 * Everything has been set up for the secondary 363 * Everything has been set up for the secondary
430 * CPUs - they just need to reload everything 364 * CPUs - they just need to reload everything
@@ -442,89 +376,233 @@ void __devinit initialize_secondary(void)
442 "movl %0,%%esp\n\t" 376 "movl %0,%%esp\n\t"
443 "jmp *%1" 377 "jmp *%1"
444 : 378 :
445 :"m" (current->thread.sp),"m" (current->thread.ip)); 379 :"m" (current->thread.sp), "m" (current->thread.ip));
446} 380}
381#endif
447 382
448/* Static state in head.S used to set up a CPU */ 383static void __cpuinit smp_apply_quirks(struct cpuinfo_x86 *c)
449extern struct { 384{
450 void * sp; 385#ifdef CONFIG_X86_32
451 unsigned short ss; 386 /*
452} stack_start; 387 * Mask B, Pentium, but not Pentium MMX
388 */
389 if (c->x86_vendor == X86_VENDOR_INTEL &&
390 c->x86 == 5 &&
391 c->x86_mask >= 1 && c->x86_mask <= 4 &&
392 c->x86_model <= 3)
393 /*
394 * Remember we have B step Pentia with bugs
395 */
396 smp_b_stepping = 1;
453 397
454#ifdef CONFIG_NUMA 398 /*
399 * Certain Athlons might work (for various values of 'work') in SMP
400 * but they are not certified as MP capable.
401 */
402 if ((c->x86_vendor == X86_VENDOR_AMD) && (c->x86 == 6)) {
455 403
456/* which logical CPUs are on which nodes */ 404 if (num_possible_cpus() == 1)
457cpumask_t node_to_cpumask_map[MAX_NUMNODES] __read_mostly = 405 goto valid_k7;
458 { [0 ... MAX_NUMNODES-1] = CPU_MASK_NONE };
459EXPORT_SYMBOL(node_to_cpumask_map);
460/* which node each logical CPU is on */
461int cpu_to_node_map[NR_CPUS] __read_mostly = { [0 ... NR_CPUS-1] = 0 };
462EXPORT_SYMBOL(cpu_to_node_map);
463 406
464/* set up a mapping between cpu and node. */ 407 /* Athlon 660/661 is valid. */
465static inline void map_cpu_to_node(int cpu, int node) 408 if ((c->x86_model == 6) && ((c->x86_mask == 0) ||
466{ 409 (c->x86_mask == 1)))
467 printk("Mapping cpu %d to node %d\n", cpu, node); 410 goto valid_k7;
468 cpu_set(cpu, node_to_cpumask_map[node]); 411
469 cpu_to_node_map[cpu] = node; 412 /* Duron 670 is valid */
413 if ((c->x86_model == 7) && (c->x86_mask == 0))
414 goto valid_k7;
415
416 /*
417 * Athlon 662, Duron 671, and Athlon >model 7 have capability
418 * bit. It's worth noting that the A5 stepping (662) of some
419 * Athlon XP's have the MP bit set.
420 * See http://www.heise.de/newsticker/data/jow-18.10.01-000 for
421 * more.
422 */
423 if (((c->x86_model == 6) && (c->x86_mask >= 2)) ||
424 ((c->x86_model == 7) && (c->x86_mask >= 1)) ||
425 (c->x86_model > 7))
426 if (cpu_has_mp)
427 goto valid_k7;
428
429 /* If we get here, not a certified SMP capable AMD system. */
430 add_taint(TAINT_UNSAFE_SMP);
431 }
432
433valid_k7:
434 ;
435#endif
470} 436}
471 437
472/* undo a mapping between cpu and node. */ 438void __cpuinit smp_checks(void)
473static inline void unmap_cpu_to_node(int cpu)
474{ 439{
475 int node; 440 if (smp_b_stepping)
441 printk(KERN_WARNING "WARNING: SMP operation may be unreliable"
442 "with B stepping processors.\n");
476 443
477 printk("Unmapping cpu %d from all nodes\n", cpu); 444 /*
478 for (node = 0; node < MAX_NUMNODES; node ++) 445 * Don't taint if we are running SMP kernel on a single non-MP
479 cpu_clear(cpu, node_to_cpumask_map[node]); 446 * approved Athlon
480 cpu_to_node_map[cpu] = 0; 447 */
448 if (tainted & TAINT_UNSAFE_SMP) {
449 if (num_online_cpus())
450 printk(KERN_INFO "WARNING: This combination of AMD"
451 "processors is not suitable for SMP.\n");
452 else
453 tainted &= ~TAINT_UNSAFE_SMP;
454 }
481} 455}
482#else /* !CONFIG_NUMA */
483 456
484#define map_cpu_to_node(cpu, node) ({}) 457/*
485#define unmap_cpu_to_node(cpu) ({}) 458 * The bootstrap kernel entry code has set these up. Save them for
459 * a given CPU
460 */
486 461
487#endif /* CONFIG_NUMA */ 462void __cpuinit smp_store_cpu_info(int id)
463{
464 struct cpuinfo_x86 *c = &cpu_data(id);
465
466 *c = boot_cpu_data;
467 c->cpu_index = id;
468 if (id != 0)
469 identify_secondary_cpu(c);
470 smp_apply_quirks(c);
471}
488 472
489u8 cpu_2_logical_apicid[NR_CPUS] __read_mostly = { [0 ... NR_CPUS-1] = BAD_APICID };
490 473
491static void map_cpu_to_logical_apicid(void) 474void __cpuinit set_cpu_sibling_map(int cpu)
492{ 475{
493 int cpu = smp_processor_id(); 476 int i;
494 int apicid = logical_smp_processor_id(); 477 struct cpuinfo_x86 *c = &cpu_data(cpu);
495 int node = apicid_to_node(apicid);
496 478
497 if (!node_online(node)) 479 cpu_set(cpu, cpu_sibling_setup_map);
498 node = first_online_node;
499 480
500 cpu_2_logical_apicid[cpu] = apicid; 481 if (smp_num_siblings > 1) {
501 map_cpu_to_node(cpu, node); 482 for_each_cpu_mask(i, cpu_sibling_setup_map) {
483 if (c->phys_proc_id == cpu_data(i).phys_proc_id &&
484 c->cpu_core_id == cpu_data(i).cpu_core_id) {
485 cpu_set(i, per_cpu(cpu_sibling_map, cpu));
486 cpu_set(cpu, per_cpu(cpu_sibling_map, i));
487 cpu_set(i, per_cpu(cpu_core_map, cpu));
488 cpu_set(cpu, per_cpu(cpu_core_map, i));
489 cpu_set(i, c->llc_shared_map);
490 cpu_set(cpu, cpu_data(i).llc_shared_map);
491 }
492 }
493 } else {
494 cpu_set(cpu, per_cpu(cpu_sibling_map, cpu));
495 }
496
497 cpu_set(cpu, c->llc_shared_map);
498
499 if (current_cpu_data.x86_max_cores == 1) {
500 per_cpu(cpu_core_map, cpu) = per_cpu(cpu_sibling_map, cpu);
501 c->booted_cores = 1;
502 return;
503 }
504
505 for_each_cpu_mask(i, cpu_sibling_setup_map) {
506 if (per_cpu(cpu_llc_id, cpu) != BAD_APICID &&
507 per_cpu(cpu_llc_id, cpu) == per_cpu(cpu_llc_id, i)) {
508 cpu_set(i, c->llc_shared_map);
509 cpu_set(cpu, cpu_data(i).llc_shared_map);
510 }
511 if (c->phys_proc_id == cpu_data(i).phys_proc_id) {
512 cpu_set(i, per_cpu(cpu_core_map, cpu));
513 cpu_set(cpu, per_cpu(cpu_core_map, i));
514 /*
515 * Does this new cpu bringup a new core?
516 */
517 if (cpus_weight(per_cpu(cpu_sibling_map, cpu)) == 1) {
518 /*
519 * for each core in package, increment
520 * the booted_cores for this new cpu
521 */
522 if (first_cpu(per_cpu(cpu_sibling_map, i)) == i)
523 c->booted_cores++;
524 /*
525 * increment the core count for all
526 * the other cpus in this package
527 */
528 if (i != cpu)
529 cpu_data(i).booted_cores++;
530 } else if (i != cpu && !c->booted_cores)
531 c->booted_cores = cpu_data(i).booted_cores;
532 }
533 }
502} 534}
503 535
504static void unmap_cpu_to_logical_apicid(int cpu) 536/* maps the cpu to the sched domain representing multi-core */
537cpumask_t cpu_coregroup_map(int cpu)
505{ 538{
506 cpu_2_logical_apicid[cpu] = BAD_APICID; 539 struct cpuinfo_x86 *c = &cpu_data(cpu);
507 unmap_cpu_to_node(cpu); 540 /*
541 * For perf, we return last level cache shared map.
542 * And for power savings, we return cpu_core_map
543 */
544 if (sched_mc_power_savings || sched_smt_power_savings)
545 return per_cpu(cpu_core_map, cpu);
546 else
547 return c->llc_shared_map;
548}
549
550#ifdef CONFIG_X86_32
551/*
552 * We are called very early to get the low memory for the
553 * SMP bootup trampoline page.
554 */
555void __init smp_alloc_memory(void)
556{
557 trampoline_base = alloc_bootmem_low_pages(PAGE_SIZE);
558 /*
559 * Has to be in very low memory so we can execute
560 * real-mode AP code.
561 */
562 if (__pa(trampoline_base) >= 0x9F000)
563 BUG();
564}
565#endif
566
567void impress_friends(void)
568{
569 int cpu;
570 unsigned long bogosum = 0;
571 /*
572 * Allow the user to impress friends.
573 */
574 Dprintk("Before bogomips.\n");
575 for_each_possible_cpu(cpu)
576 if (cpu_isset(cpu, cpu_callout_map))
577 bogosum += cpu_data(cpu).loops_per_jiffy;
578 printk(KERN_INFO
579 "Total of %d processors activated (%lu.%02lu BogoMIPS).\n",
580 num_online_cpus(),
581 bogosum/(500000/HZ),
582 (bogosum/(5000/HZ))%100);
583
584 Dprintk("Before bogocount - setting activated=1.\n");
508} 585}
509 586
510static inline void __inquire_remote_apic(int apicid) 587static inline void __inquire_remote_apic(int apicid)
511{ 588{
512 int i, regs[] = { APIC_ID >> 4, APIC_LVR >> 4, APIC_SPIV >> 4 }; 589 unsigned i, regs[] = { APIC_ID >> 4, APIC_LVR >> 4, APIC_SPIV >> 4 };
513 char *names[] = { "ID", "VERSION", "SPIV" }; 590 char *names[] = { "ID", "VERSION", "SPIV" };
514 int timeout; 591 int timeout;
515 unsigned long status; 592 u32 status;
516 593
517 printk("Inquiring remote APIC #%d...\n", apicid); 594 printk(KERN_INFO "Inquiring remote APIC #%d...\n", apicid);
518 595
519 for (i = 0; i < ARRAY_SIZE(regs); i++) { 596 for (i = 0; i < ARRAY_SIZE(regs); i++) {
520 printk("... APIC #%d %s: ", apicid, names[i]); 597 printk(KERN_INFO "... APIC #%d %s: ", apicid, names[i]);
521 598
522 /* 599 /*
523 * Wait for idle. 600 * Wait for idle.
524 */ 601 */
525 status = safe_apic_wait_icr_idle(); 602 status = safe_apic_wait_icr_idle();
526 if (status) 603 if (status)
527 printk("a previous APIC delivery may have failed\n"); 604 printk(KERN_CONT
605 "a previous APIC delivery may have failed\n");
528 606
529 apic_write_around(APIC_ICR2, SET_APIC_DEST_FIELD(apicid)); 607 apic_write_around(APIC_ICR2, SET_APIC_DEST_FIELD(apicid));
530 apic_write_around(APIC_ICR, APIC_DM_REMRD | regs[i]); 608 apic_write_around(APIC_ICR, APIC_DM_REMRD | regs[i]);
@@ -538,16 +616,16 @@ static inline void __inquire_remote_apic(int apicid)
538 switch (status) { 616 switch (status) {
539 case APIC_ICR_RR_VALID: 617 case APIC_ICR_RR_VALID:
540 status = apic_read(APIC_RRR); 618 status = apic_read(APIC_RRR);
541 printk("%lx\n", status); 619 printk(KERN_CONT "%08x\n", status);
542 break; 620 break;
543 default: 621 default:
544 printk("failed\n"); 622 printk(KERN_CONT "failed\n");
545 } 623 }
546 } 624 }
547} 625}
548 626
549#ifdef WAKE_SECONDARY_VIA_NMI 627#ifdef WAKE_SECONDARY_VIA_NMI
550/* 628/*
551 * Poke the other CPU in the eye via NMI to wake it up. Remember that the normal 629 * Poke the other CPU in the eye via NMI to wake it up. Remember that the normal
552 * INIT, INIT, STARTUP sequence will reset the chip hard for us, and this 630 * INIT, INIT, STARTUP sequence will reset the chip hard for us, and this
553 * won't ... remember to clear down the APIC, etc later. 631 * won't ... remember to clear down the APIC, etc later.
@@ -584,9 +662,9 @@ wakeup_secondary_cpu(int logical_apicid, unsigned long start_eip)
584 Dprintk("NMI sent.\n"); 662 Dprintk("NMI sent.\n");
585 663
586 if (send_status) 664 if (send_status)
587 printk("APIC never delivered???\n"); 665 printk(KERN_ERR "APIC never delivered???\n");
588 if (accept_status) 666 if (accept_status)
589 printk("APIC delivery error (%lx).\n", accept_status); 667 printk(KERN_ERR "APIC delivery error (%lx).\n", accept_status);
590 668
591 return (send_status | accept_status); 669 return (send_status | accept_status);
592} 670}
@@ -637,6 +715,7 @@ wakeup_secondary_cpu(int phys_apicid, unsigned long start_eip)
637 Dprintk("Waiting for send to finish...\n"); 715 Dprintk("Waiting for send to finish...\n");
638 send_status = safe_apic_wait_icr_idle(); 716 send_status = safe_apic_wait_icr_idle();
639 717
718 mb();
640 atomic_set(&init_deasserted, 1); 719 atomic_set(&init_deasserted, 1);
641 720
642 /* 721 /*
@@ -655,7 +734,11 @@ wakeup_secondary_cpu(int phys_apicid, unsigned long start_eip)
655 * target processor state. 734 * target processor state.
656 */ 735 */
657 startup_ipi_hook(phys_apicid, (unsigned long) start_secondary, 736 startup_ipi_hook(phys_apicid, (unsigned long) start_secondary,
658 (unsigned long) stack_start.sp); 737#ifdef CONFIG_X86_64
738 (unsigned long)init_rsp);
739#else
740 (unsigned long)stack_start.sp);
741#endif
659 742
660 /* 743 /*
661 * Run STARTUP IPI loop. 744 * Run STARTUP IPI loop.
@@ -665,7 +748,7 @@ wakeup_secondary_cpu(int phys_apicid, unsigned long start_eip)
665 maxlvt = lapic_get_maxlvt(); 748 maxlvt = lapic_get_maxlvt();
666 749
667 for (j = 1; j <= num_starts; j++) { 750 for (j = 1; j <= num_starts; j++) {
668 Dprintk("Sending STARTUP #%d.\n",j); 751 Dprintk("Sending STARTUP #%d.\n", j);
669 apic_read_around(APIC_SPIV); 752 apic_read_around(APIC_SPIV);
670 apic_write(APIC_ESR, 0); 753 apic_write(APIC_ESR, 0);
671 apic_read(APIC_ESR); 754 apic_read(APIC_ESR);
@@ -711,49 +794,29 @@ wakeup_secondary_cpu(int phys_apicid, unsigned long start_eip)
711 Dprintk("After Startup.\n"); 794 Dprintk("After Startup.\n");
712 795
713 if (send_status) 796 if (send_status)
714 printk("APIC never delivered???\n"); 797 printk(KERN_ERR "APIC never delivered???\n");
715 if (accept_status) 798 if (accept_status)
716 printk("APIC delivery error (%lx).\n", accept_status); 799 printk(KERN_ERR "APIC delivery error (%lx).\n", accept_status);
717 800
718 return (send_status | accept_status); 801 return (send_status | accept_status);
719} 802}
720#endif /* WAKE_SECONDARY_VIA_INIT */ 803#endif /* WAKE_SECONDARY_VIA_INIT */
721 804
722extern cpumask_t cpu_initialized; 805struct create_idle {
723static inline int alloc_cpu_id(void) 806 struct work_struct work;
724{ 807 struct task_struct *idle;
725 cpumask_t tmp_map; 808 struct completion done;
726 int cpu; 809 int cpu;
727 cpus_complement(tmp_map, cpu_present_map); 810};
728 cpu = first_cpu(tmp_map);
729 if (cpu >= NR_CPUS)
730 return -ENODEV;
731 return cpu;
732}
733 811
734#ifdef CONFIG_HOTPLUG_CPU 812static void __cpuinit do_fork_idle(struct work_struct *work)
735static struct task_struct * __cpuinitdata cpu_idle_tasks[NR_CPUS];
736static inline struct task_struct * __cpuinit alloc_idle_task(int cpu)
737{ 813{
738 struct task_struct *idle; 814 struct create_idle *c_idle =
815 container_of(work, struct create_idle, work);
739 816
740 if ((idle = cpu_idle_tasks[cpu]) != NULL) { 817 c_idle->idle = fork_idle(c_idle->cpu);
741 /* initialize thread_struct. we really want to avoid destroy 818 complete(&c_idle->done);
742 * idle tread
743 */
744 idle->thread.sp = (unsigned long)task_pt_regs(idle);
745 init_idle(idle, cpu);
746 return idle;
747 }
748 idle = fork_idle(cpu);
749
750 if (!IS_ERR(idle))
751 cpu_idle_tasks[cpu] = idle;
752 return idle;
753} 819}
754#else
755#define alloc_idle_task(cpu) fork_idle(cpu)
756#endif
757 820
758static int __cpuinit do_boot_cpu(int apicid, int cpu) 821static int __cpuinit do_boot_cpu(int apicid, int cpu)
759/* 822/*
@@ -762,45 +825,92 @@ static int __cpuinit do_boot_cpu(int apicid, int cpu)
762 * Returns zero if CPU booted OK, else error code from wakeup_secondary_cpu. 825 * Returns zero if CPU booted OK, else error code from wakeup_secondary_cpu.
763 */ 826 */
764{ 827{
765 struct task_struct *idle; 828 unsigned long boot_error = 0;
766 unsigned long boot_error;
767 int timeout; 829 int timeout;
768 unsigned long start_eip; 830 unsigned long start_ip;
769 unsigned short nmi_high = 0, nmi_low = 0; 831 unsigned short nmi_high = 0, nmi_low = 0;
832 struct create_idle c_idle = {
833 .cpu = cpu,
834 .done = COMPLETION_INITIALIZER_ONSTACK(c_idle.done),
835 };
836 INIT_WORK(&c_idle.work, do_fork_idle);
837#ifdef CONFIG_X86_64
838 /* allocate memory for gdts of secondary cpus. Hotplug is considered */
839 if (!cpu_gdt_descr[cpu].address &&
840 !(cpu_gdt_descr[cpu].address = get_zeroed_page(GFP_KERNEL))) {
841 printk(KERN_ERR "Failed to allocate GDT for CPU %d\n", cpu);
842 return -1;
843 }
770 844
771 /* 845 /* Allocate node local memory for AP pdas */
772 * Save current MTRR state in case it was changed since early boot 846 if (cpu_pda(cpu) == &boot_cpu_pda[cpu]) {
773 * (e.g. by the ACPI SMI) to initialize new CPUs with MTRRs in sync: 847 struct x8664_pda *newpda, *pda;
774 */ 848 int node = cpu_to_node(cpu);
775 mtrr_save_state(); 849 pda = cpu_pda(cpu);
850 newpda = kmalloc_node(sizeof(struct x8664_pda), GFP_ATOMIC,
851 node);
852 if (newpda) {
853 memcpy(newpda, pda, sizeof(struct x8664_pda));
854 cpu_pda(cpu) = newpda;
855 } else
856 printk(KERN_ERR
857 "Could not allocate node local PDA for CPU %d on node %d\n",
858 cpu, node);
859 }
860#endif
861
862 alternatives_smp_switch(1);
863
864 c_idle.idle = get_idle_for_cpu(cpu);
776 865
777 /* 866 /*
778 * We can't use kernel_thread since we must avoid to 867 * We can't use kernel_thread since we must avoid to
779 * reschedule the child. 868 * reschedule the child.
780 */ 869 */
781 idle = alloc_idle_task(cpu); 870 if (c_idle.idle) {
782 if (IS_ERR(idle)) 871 c_idle.idle->thread.sp = (unsigned long) (((struct pt_regs *)
783 panic("failed fork for CPU %d", cpu); 872 (THREAD_SIZE + task_stack_page(c_idle.idle))) - 1);
873 init_idle(c_idle.idle, cpu);
874 goto do_rest;
875 }
784 876
877 if (!keventd_up() || current_is_keventd())
878 c_idle.work.func(&c_idle.work);
879 else {
880 schedule_work(&c_idle.work);
881 wait_for_completion(&c_idle.done);
882 }
883
884 if (IS_ERR(c_idle.idle)) {
885 printk("failed fork for CPU %d\n", cpu);
886 return PTR_ERR(c_idle.idle);
887 }
888
889 set_idle_for_cpu(cpu, c_idle.idle);
890do_rest:
891#ifdef CONFIG_X86_32
892 per_cpu(current_task, cpu) = c_idle.idle;
785 init_gdt(cpu); 893 init_gdt(cpu);
786 per_cpu(current_task, cpu) = idle;
787 early_gdt_descr.address = (unsigned long)get_cpu_gdt_table(cpu); 894 early_gdt_descr.address = (unsigned long)get_cpu_gdt_table(cpu);
895 c_idle.idle->thread.ip = (unsigned long) start_secondary;
896 /* Stack for startup_32 can be just as for start_secondary onwards */
897 stack_start.sp = (void *) c_idle.idle->thread.sp;
898 irq_ctx_init(cpu);
899#else
900 cpu_pda(cpu)->pcurrent = c_idle.idle;
901 init_rsp = c_idle.idle->thread.sp;
902 load_sp0(&per_cpu(init_tss, cpu), &c_idle.idle->thread);
903 initial_code = (unsigned long)start_secondary;
904 clear_tsk_thread_flag(c_idle.idle, TIF_FORK);
905#endif
788 906
789 idle->thread.ip = (unsigned long) start_secondary; 907 /* start_ip had better be page-aligned! */
790 /* start_eip had better be page-aligned! */ 908 start_ip = setup_trampoline();
791 start_eip = setup_trampoline();
792
793 ++cpucount;
794 alternatives_smp_switch(1);
795 909
796 /* So we see what's up */ 910 /* So we see what's up */
797 printk("Booting processor %d/%d ip %lx\n", cpu, apicid, start_eip); 911 printk(KERN_INFO "Booting processor %d/%d ip %lx\n",
798 /* Stack for startup_32 can be just as for start_secondary onwards */ 912 cpu, apicid, start_ip);
799 stack_start.sp = (void *) idle->thread.sp;
800 913
801 irq_ctx_init(cpu);
802
803 per_cpu(x86_cpu_to_apicid, cpu) = apicid;
804 /* 914 /*
805 * This grunge runs the startup process for 915 * This grunge runs the startup process for
806 * the targeted processor. 916 * the targeted processor.
@@ -812,12 +922,17 @@ static int __cpuinit do_boot_cpu(int apicid, int cpu)
812 922
813 store_NMI_vector(&nmi_high, &nmi_low); 923 store_NMI_vector(&nmi_high, &nmi_low);
814 924
815 smpboot_setup_warm_reset_vector(start_eip); 925 smpboot_setup_warm_reset_vector(start_ip);
926 /*
927 * Be paranoid about clearing APIC errors.
928 */
929 apic_write(APIC_ESR, 0);
930 apic_read(APIC_ESR);
816 931
817 /* 932 /*
818 * Starting actual IPI sequence... 933 * Starting actual IPI sequence...
819 */ 934 */
820 boot_error = wakeup_secondary_cpu(apicid, start_eip); 935 boot_error = wakeup_secondary_cpu(apicid, start_ip);
821 936
822 if (!boot_error) { 937 if (!boot_error) {
823 /* 938 /*
@@ -839,18 +954,18 @@ static int __cpuinit do_boot_cpu(int apicid, int cpu)
839 if (cpu_isset(cpu, cpu_callin_map)) { 954 if (cpu_isset(cpu, cpu_callin_map)) {
840 /* number CPUs logically, starting from 1 (BSP is 0) */ 955 /* number CPUs logically, starting from 1 (BSP is 0) */
841 Dprintk("OK.\n"); 956 Dprintk("OK.\n");
842 printk("CPU%d: ", cpu); 957 printk(KERN_INFO "CPU%d: ", cpu);
843 print_cpu_info(&cpu_data(cpu)); 958 print_cpu_info(&cpu_data(cpu));
844 Dprintk("CPU has booted.\n"); 959 Dprintk("CPU has booted.\n");
845 } else { 960 } else {
846 boot_error= 1; 961 boot_error = 1;
847 if (*((volatile unsigned char *)trampoline_base) 962 if (*((volatile unsigned char *)trampoline_base)
848 == 0xA5) 963 == 0xA5)
849 /* trampoline started but...? */ 964 /* trampoline started but...? */
850 printk("Stuck ??\n"); 965 printk(KERN_ERR "Stuck ??\n");
851 else 966 else
852 /* trampoline code not run */ 967 /* trampoline code not run */
853 printk("Not responding.\n"); 968 printk(KERN_ERR "Not responding.\n");
854 inquire_remote_apic(apicid); 969 inquire_remote_apic(apicid);
855 } 970 }
856 } 971 }
@@ -858,156 +973,159 @@ static int __cpuinit do_boot_cpu(int apicid, int cpu)
858 if (boot_error) { 973 if (boot_error) {
859 /* Try to put things back the way they were before ... */ 974 /* Try to put things back the way they were before ... */
860 unmap_cpu_to_logical_apicid(cpu); 975 unmap_cpu_to_logical_apicid(cpu);
861 cpu_clear(cpu, cpu_callout_map); /* was set here (do_boot_cpu()) */ 976#ifdef CONFIG_X86_64
977 clear_node_cpumask(cpu); /* was set by numa_add_cpu */
978#endif
979 cpu_clear(cpu, cpu_callout_map); /* was set by do_boot_cpu() */
862 cpu_clear(cpu, cpu_initialized); /* was set by cpu_init() */ 980 cpu_clear(cpu, cpu_initialized); /* was set by cpu_init() */
863 cpucount--; 981 cpu_clear(cpu, cpu_possible_map);
864 } else { 982 cpu_clear(cpu, cpu_present_map);
865 per_cpu(x86_cpu_to_apicid, cpu) = apicid; 983 per_cpu(x86_cpu_to_apicid, cpu) = BAD_APICID;
866 cpu_set(cpu, cpu_present_map);
867 } 984 }
868 985
869 /* mark "stuck" area as not stuck */ 986 /* mark "stuck" area as not stuck */
870 *((volatile unsigned long *)trampoline_base) = 0; 987 *((volatile unsigned long *)trampoline_base) = 0;
871 988
989 /*
990 * Cleanup possible dangling ends...
991 */
992 smpboot_restore_warm_reset_vector();
993
872 return boot_error; 994 return boot_error;
873} 995}
874 996
875#ifdef CONFIG_HOTPLUG_CPU 997int __cpuinit native_cpu_up(unsigned int cpu)
876void cpu_exit_clear(void)
877{ 998{
878 int cpu = raw_smp_processor_id(); 999 int apicid = cpu_present_to_apicid(cpu);
879 1000 unsigned long flags;
880 idle_task_exit(); 1001 int err;
881
882 cpucount --;
883 cpu_uninit();
884 irq_ctx_exit(cpu);
885
886 cpu_clear(cpu, cpu_callout_map);
887 cpu_clear(cpu, cpu_callin_map);
888 1002
889 cpu_clear(cpu, smp_commenced_mask); 1003 WARN_ON(irqs_disabled());
890 unmap_cpu_to_logical_apicid(cpu);
891}
892 1004
893struct warm_boot_cpu_info { 1005 Dprintk("++++++++++++++++++++=_---CPU UP %u\n", cpu);
894 struct completion *complete;
895 struct work_struct task;
896 int apicid;
897 int cpu;
898};
899 1006
900static void __cpuinit do_warm_boot_cpu(struct work_struct *work) 1007 if (apicid == BAD_APICID || apicid == boot_cpu_physical_apicid ||
901{ 1008 !physid_isset(apicid, phys_cpu_present_map)) {
902 struct warm_boot_cpu_info *info = 1009 printk(KERN_ERR "%s: bad cpu %d\n", __func__, cpu);
903 container_of(work, struct warm_boot_cpu_info, task); 1010 return -EINVAL;
904 do_boot_cpu(info->apicid, info->cpu); 1011 }
905 complete(info->complete);
906}
907 1012
908static int __cpuinit __smp_prepare_cpu(int cpu) 1013 /*
909{ 1014 * Already booted CPU?
910 DECLARE_COMPLETION_ONSTACK(done); 1015 */
911 struct warm_boot_cpu_info info; 1016 if (cpu_isset(cpu, cpu_callin_map)) {
912 int apicid, ret; 1017 Dprintk("do_boot_cpu %d Already started\n", cpu);
913 1018 return -ENOSYS;
914 apicid = per_cpu(x86_cpu_to_apicid, cpu);
915 if (apicid == BAD_APICID) {
916 ret = -ENODEV;
917 goto exit;
918 } 1019 }
919 1020
920 info.complete = &done; 1021 /*
921 info.apicid = apicid; 1022 * Save current MTRR state in case it was changed since early boot
922 info.cpu = cpu; 1023 * (e.g. by the ACPI SMI) to initialize new CPUs with MTRRs in sync:
923 INIT_WORK(&info.task, do_warm_boot_cpu); 1024 */
1025 mtrr_save_state();
1026
1027 per_cpu(cpu_state, cpu) = CPU_UP_PREPARE;
924 1028
1029#ifdef CONFIG_X86_32
925 /* init low mem mapping */ 1030 /* init low mem mapping */
926 clone_pgd_range(swapper_pg_dir, swapper_pg_dir + USER_PGD_PTRS, 1031 clone_pgd_range(swapper_pg_dir, swapper_pg_dir + USER_PGD_PTRS,
927 min_t(unsigned long, KERNEL_PGD_PTRS, USER_PGD_PTRS)); 1032 min_t(unsigned long, KERNEL_PGD_PTRS, USER_PGD_PTRS));
928 flush_tlb_all(); 1033 flush_tlb_all();
929 schedule_work(&info.task);
930 wait_for_completion(&done);
931
932 zap_low_mappings();
933 ret = 0;
934exit:
935 return ret;
936}
937#endif
938
939/*
940 * Cycle through the processors sending APIC IPIs to boot each.
941 */
942
943static int boot_cpu_logical_apicid;
944/* Where the IO area was mapped on multiquad, always 0 otherwise */
945void *xquad_portio;
946#ifdef CONFIG_X86_NUMAQ
947EXPORT_SYMBOL(xquad_portio);
948#endif 1034#endif
949 1035
950static void __init smp_boot_cpus(unsigned int max_cpus) 1036 err = do_boot_cpu(apicid, cpu);
951{ 1037 if (err < 0) {
952 int apicid, cpu, bit, kicked; 1038 Dprintk("do_boot_cpu failed %d\n", err);
953 unsigned long bogosum = 0; 1039 return err;
1040 }
954 1041
955 /* 1042 /*
956 * Setup boot CPU information 1043 * Check TSC synchronization with the AP (keep irqs disabled
1044 * while doing so):
957 */ 1045 */
958 smp_store_cpu_info(0); /* Final full version of the data */ 1046 local_irq_save(flags);
959 printk("CPU%d: ", 0); 1047 check_tsc_sync_source(cpu);
960 print_cpu_info(&cpu_data(0)); 1048 local_irq_restore(flags);
961 1049
962 boot_cpu_physical_apicid = GET_APIC_ID(apic_read(APIC_ID)); 1050 while (!cpu_isset(cpu, cpu_online_map)) {
963 boot_cpu_logical_apicid = logical_smp_processor_id(); 1051 cpu_relax();
964 per_cpu(x86_cpu_to_apicid, 0) = boot_cpu_physical_apicid; 1052 touch_nmi_watchdog();
1053 }
965 1054
966 current_thread_info()->cpu = 0; 1055 return 0;
1056}
967 1057
968 set_cpu_sibling_map(0); 1058/*
1059 * Fall back to non SMP mode after errors.
1060 *
1061 * RED-PEN audit/test this more. I bet there is more state messed up here.
1062 */
1063static __init void disable_smp(void)
1064{
1065 cpu_present_map = cpumask_of_cpu(0);
1066 cpu_possible_map = cpumask_of_cpu(0);
1067#ifdef CONFIG_X86_32
1068 smpboot_clear_io_apic_irqs();
1069#endif
1070 if (smp_found_config)
1071 phys_cpu_present_map =
1072 physid_mask_of_physid(boot_cpu_physical_apicid);
1073 else
1074 phys_cpu_present_map = physid_mask_of_physid(0);
1075 map_cpu_to_logical_apicid();
1076 cpu_set(0, per_cpu(cpu_sibling_map, 0));
1077 cpu_set(0, per_cpu(cpu_core_map, 0));
1078}
1079
1080/*
1081 * Various sanity checks.
1082 */
1083static int __init smp_sanity_check(unsigned max_cpus)
1084{
1085 preempt_disable();
1086 if (!physid_isset(hard_smp_processor_id(), phys_cpu_present_map)) {
1087 printk(KERN_WARNING "weird, boot CPU (#%d) not listed"
1088 "by the BIOS.\n", hard_smp_processor_id());
1089 physid_set(hard_smp_processor_id(), phys_cpu_present_map);
1090 }
969 1091
970 /* 1092 /*
971 * If we couldn't find an SMP configuration at boot time, 1093 * If we couldn't find an SMP configuration at boot time,
972 * get out of here now! 1094 * get out of here now!
973 */ 1095 */
974 if (!smp_found_config && !acpi_lapic) { 1096 if (!smp_found_config && !acpi_lapic) {
1097 preempt_enable();
975 printk(KERN_NOTICE "SMP motherboard not detected.\n"); 1098 printk(KERN_NOTICE "SMP motherboard not detected.\n");
976 smpboot_clear_io_apic_irqs(); 1099 disable_smp();
977 phys_cpu_present_map = physid_mask_of_physid(0);
978 if (APIC_init_uniprocessor()) 1100 if (APIC_init_uniprocessor())
979 printk(KERN_NOTICE "Local APIC not detected." 1101 printk(KERN_NOTICE "Local APIC not detected."
980 " Using dummy APIC emulation.\n"); 1102 " Using dummy APIC emulation.\n");
981 map_cpu_to_logical_apicid(); 1103 return -1;
982 cpu_set(0, per_cpu(cpu_sibling_map, 0));
983 cpu_set(0, per_cpu(cpu_core_map, 0));
984 return;
985 } 1104 }
986 1105
987 /* 1106 /*
988 * Should not be necessary because the MP table should list the boot 1107 * Should not be necessary because the MP table should list the boot
989 * CPU too, but we do it for the sake of robustness anyway. 1108 * CPU too, but we do it for the sake of robustness anyway.
990 * Makes no sense to do this check in clustered apic mode, so skip it
991 */ 1109 */
992 if (!check_phys_apicid_present(boot_cpu_physical_apicid)) { 1110 if (!check_phys_apicid_present(boot_cpu_physical_apicid)) {
993 printk("weird, boot CPU (#%d) not listed by the BIOS.\n", 1111 printk(KERN_NOTICE
994 boot_cpu_physical_apicid); 1112 "weird, boot CPU (#%d) not listed by the BIOS.\n",
1113 boot_cpu_physical_apicid);
995 physid_set(hard_smp_processor_id(), phys_cpu_present_map); 1114 physid_set(hard_smp_processor_id(), phys_cpu_present_map);
996 } 1115 }
1116 preempt_enable();
997 1117
998 /* 1118 /*
999 * If we couldn't find a local APIC, then get out of here now! 1119 * If we couldn't find a local APIC, then get out of here now!
1000 */ 1120 */
1001 if (APIC_INTEGRATED(apic_version[boot_cpu_physical_apicid]) && !cpu_has_apic) { 1121 if (APIC_INTEGRATED(apic_version[boot_cpu_physical_apicid]) &&
1122 !cpu_has_apic) {
1002 printk(KERN_ERR "BIOS bug, local APIC #%d not detected!...\n", 1123 printk(KERN_ERR "BIOS bug, local APIC #%d not detected!...\n",
1003 boot_cpu_physical_apicid); 1124 boot_cpu_physical_apicid);
1004 printk(KERN_ERR "... forcing use of dummy APIC emulation. (tell your hw vendor)\n"); 1125 printk(KERN_ERR "... forcing use of dummy APIC emulation."
1005 smpboot_clear_io_apic_irqs(); 1126 "(tell your hw vendor)\n");
1006 phys_cpu_present_map = physid_mask_of_physid(0); 1127 smpboot_clear_io_apic();
1007 map_cpu_to_logical_apicid(); 1128 return -1;
1008 cpu_set(0, per_cpu(cpu_sibling_map, 0));
1009 cpu_set(0, per_cpu(cpu_core_map, 0));
1010 return;
1011 } 1129 }
1012 1130
1013 verify_local_APIC(); 1131 verify_local_APIC();
@@ -1016,137 +1134,148 @@ static void __init smp_boot_cpus(unsigned int max_cpus)
1016 * If SMP should be disabled, then really disable it! 1134 * If SMP should be disabled, then really disable it!
1017 */ 1135 */
1018 if (!max_cpus) { 1136 if (!max_cpus) {
1019 smp_found_config = 0; 1137 printk(KERN_INFO "SMP mode deactivated,"
1020 printk(KERN_INFO "SMP mode deactivated, forcing use of dummy APIC emulation.\n"); 1138 "forcing use of dummy APIC emulation.\n");
1021 1139 smpboot_clear_io_apic();
1140#ifdef CONFIG_X86_32
1022 if (nmi_watchdog == NMI_LOCAL_APIC) { 1141 if (nmi_watchdog == NMI_LOCAL_APIC) {
1023 printk(KERN_INFO "activating minimal APIC for NMI watchdog use.\n"); 1142 printk(KERN_INFO "activating minimal APIC for"
1143 "NMI watchdog use.\n");
1024 connect_bsp_APIC(); 1144 connect_bsp_APIC();
1025 setup_local_APIC(); 1145 setup_local_APIC();
1146 end_local_APIC_setup();
1026 } 1147 }
1027 smpboot_clear_io_apic_irqs(); 1148#endif
1028 phys_cpu_present_map = physid_mask_of_physid(0); 1149 return -1;
1029 map_cpu_to_logical_apicid();
1030 cpu_set(0, per_cpu(cpu_sibling_map, 0));
1031 cpu_set(0, per_cpu(cpu_core_map, 0));
1032 return;
1033 } 1150 }
1034 1151
1035 connect_bsp_APIC(); 1152 return 0;
1036 setup_local_APIC(); 1153}
1037 map_cpu_to_logical_apicid();
1038 1154
1155static void __init smp_cpu_index_default(void)
1156{
1157 int i;
1158 struct cpuinfo_x86 *c;
1039 1159
1040 setup_portio_remap(); 1160 for_each_cpu_mask(i, cpu_possible_map) {
1161 c = &cpu_data(i);
1162 /* mark all to hotplug */
1163 c->cpu_index = NR_CPUS;
1164 }
1165}
1041 1166
1167/*
1168 * Prepare for SMP bootup. The MP table or ACPI has been read
1169 * earlier. Just do some sanity checking here and enable APIC mode.
1170 */
1171void __init native_smp_prepare_cpus(unsigned int max_cpus)
1172{
1173 nmi_watchdog_default();
1174 smp_cpu_index_default();
1175 current_cpu_data = boot_cpu_data;
1176 cpu_callin_map = cpumask_of_cpu(0);
1177 mb();
1042 /* 1178 /*
1043 * Scan the CPU present map and fire up the other CPUs via do_boot_cpu 1179 * Setup boot CPU information
1044 *
1045 * In clustered apic mode, phys_cpu_present_map is a constructed thus:
1046 * bits 0-3 are quad0, 4-7 are quad1, etc. A perverse twist on the
1047 * clustered apic ID.
1048 */ 1180 */
1049 Dprintk("CPU present map: %lx\n", physids_coerce(phys_cpu_present_map)); 1181 smp_store_cpu_info(0); /* Final full version of the data */
1050 1182 boot_cpu_logical_apicid = logical_smp_processor_id();
1051 kicked = 1; 1183 current_thread_info()->cpu = 0; /* needed? */
1052 for (bit = 0; kicked < NR_CPUS && bit < MAX_APICS; bit++) { 1184 set_cpu_sibling_map(0);
1053 apicid = cpu_present_to_apicid(bit);
1054 /*
1055 * Don't even attempt to start the boot CPU!
1056 */
1057 if ((apicid == boot_cpu_apicid) || (apicid == BAD_APICID))
1058 continue;
1059 1185
1060 if (!check_apicid_present(bit)) 1186 if (smp_sanity_check(max_cpus) < 0) {
1061 continue; 1187 printk(KERN_INFO "SMP disabled\n");
1062 if (max_cpus <= cpucount+1) 1188 disable_smp();
1063 continue; 1189 return;
1190 }
1064 1191
1065 if (((cpu = alloc_cpu_id()) <= 0) || do_boot_cpu(apicid, cpu)) 1192 preempt_disable();
1066 printk("CPU #%d not responding - cannot use it.\n", 1193 if (GET_APIC_ID(read_apic_id()) != boot_cpu_physical_apicid) {
1067 apicid); 1194 panic("Boot APIC ID in local APIC unexpected (%d vs %d)",
1068 else 1195 GET_APIC_ID(read_apic_id()), boot_cpu_physical_apicid);
1069 ++kicked; 1196 /* Or can we switch back to PIC here? */
1070 } 1197 }
1198 preempt_enable();
1071 1199
1200#ifdef CONFIG_X86_32
1201 connect_bsp_APIC();
1202#endif
1072 /* 1203 /*
1073 * Cleanup possible dangling ends... 1204 * Switch from PIC to APIC mode.
1074 */ 1205 */
1075 smpboot_restore_warm_reset_vector(); 1206 setup_local_APIC();
1076 1207
1208#ifdef CONFIG_X86_64
1077 /* 1209 /*
1078 * Allow the user to impress friends. 1210 * Enable IO APIC before setting up error vector
1079 */ 1211 */
1080 Dprintk("Before bogomips.\n"); 1212 if (!skip_ioapic_setup && nr_ioapics)
1081 for_each_possible_cpu(cpu) 1213 enable_IO_APIC();
1082 if (cpu_isset(cpu, cpu_callout_map)) 1214#endif
1083 bogosum += cpu_data(cpu).loops_per_jiffy; 1215 end_local_APIC_setup();
1084 printk(KERN_INFO
1085 "Total of %d processors activated (%lu.%02lu BogoMIPS).\n",
1086 cpucount+1,
1087 bogosum/(500000/HZ),
1088 (bogosum/(5000/HZ))%100);
1089
1090 Dprintk("Before bogocount - setting activated=1.\n");
1091
1092 if (smp_b_stepping)
1093 printk(KERN_WARNING "WARNING: SMP operation may be unreliable with B stepping processors.\n");
1094 1216
1095 /* 1217 map_cpu_to_logical_apicid();
1096 * Don't taint if we are running SMP kernel on a single non-MP
1097 * approved Athlon
1098 */
1099 if (tainted & TAINT_UNSAFE_SMP) {
1100 if (cpucount)
1101 printk (KERN_INFO "WARNING: This combination of AMD processors is not suitable for SMP.\n");
1102 else
1103 tainted &= ~TAINT_UNSAFE_SMP;
1104 }
1105 1218
1106 Dprintk("Boot done.\n"); 1219 setup_portio_remap();
1107 1220
1221 smpboot_setup_io_apic();
1108 /* 1222 /*
1109 * construct cpu_sibling_map, so that we can tell sibling CPUs 1223 * Set up local APIC timer on boot CPU.
1110 * efficiently.
1111 */ 1224 */
1112 for_each_possible_cpu(cpu) {
1113 cpus_clear(per_cpu(cpu_sibling_map, cpu));
1114 cpus_clear(per_cpu(cpu_core_map, cpu));
1115 }
1116
1117 cpu_set(0, per_cpu(cpu_sibling_map, 0));
1118 cpu_set(0, per_cpu(cpu_core_map, 0));
1119
1120 smpboot_setup_io_apic();
1121 1225
1226 printk(KERN_INFO "CPU%d: ", 0);
1227 print_cpu_info(&cpu_data(0));
1122 setup_boot_clock(); 1228 setup_boot_clock();
1123} 1229}
1230/*
1231 * Early setup to make printk work.
1232 */
1233void __init native_smp_prepare_boot_cpu(void)
1234{
1235 int me = smp_processor_id();
1236#ifdef CONFIG_X86_32
1237 init_gdt(me);
1238 switch_to_new_gdt();
1239#endif
1240 /* already set me in cpu_online_map in boot_cpu_init() */
1241 cpu_set(me, cpu_callout_map);
1242 per_cpu(cpu_state, me) = CPU_ONLINE;
1243}
1124 1244
1125/* These are wrappers to interface to the new boot process. Someone 1245void __init native_smp_cpus_done(unsigned int max_cpus)
1126 who understands all this stuff should rewrite it properly. --RR 15/Jul/02 */
1127void __init native_smp_prepare_cpus(unsigned int max_cpus)
1128{ 1246{
1129 smp_commenced_mask = cpumask_of_cpu(0); 1247 Dprintk("Boot done.\n");
1130 cpu_callin_map = cpumask_of_cpu(0); 1248
1131 mb(); 1249 impress_friends();
1132 smp_boot_cpus(max_cpus); 1250 smp_checks();
1251#ifdef CONFIG_X86_IO_APIC
1252 setup_ioapic_dest();
1253#endif
1254 check_nmi_watchdog();
1255#ifdef CONFIG_X86_32
1256 zap_low_mappings();
1257#endif
1133} 1258}
1134 1259
1135void __init native_smp_prepare_boot_cpu(void) 1260#ifdef CONFIG_HOTPLUG_CPU
1261
1262# ifdef CONFIG_X86_32
1263void cpu_exit_clear(void)
1136{ 1264{
1137 unsigned int cpu = smp_processor_id(); 1265 int cpu = raw_smp_processor_id();
1138 1266
1139 init_gdt(cpu); 1267 idle_task_exit();
1140 switch_to_new_gdt(); 1268
1269 cpu_uninit();
1270 irq_ctx_exit(cpu);
1271
1272 cpu_clear(cpu, cpu_callout_map);
1273 cpu_clear(cpu, cpu_callin_map);
1141 1274
1142 cpu_set(cpu, cpu_online_map); 1275 unmap_cpu_to_logical_apicid(cpu);
1143 cpu_set(cpu, cpu_callout_map);
1144 cpu_set(cpu, cpu_present_map);
1145 cpu_set(cpu, cpu_possible_map);
1146 __get_cpu_var(cpu_state) = CPU_ONLINE;
1147} 1276}
1277# endif /* CONFIG_X86_32 */
1148 1278
1149#ifdef CONFIG_HOTPLUG_CPU
1150void remove_siblinginfo(int cpu) 1279void remove_siblinginfo(int cpu)
1151{ 1280{
1152 int sibling; 1281 int sibling;
@@ -1160,7 +1289,7 @@ void remove_siblinginfo(int cpu)
1160 if (cpus_weight(per_cpu(cpu_sibling_map, cpu)) == 1) 1289 if (cpus_weight(per_cpu(cpu_sibling_map, cpu)) == 1)
1161 cpu_data(sibling).booted_cores--; 1290 cpu_data(sibling).booted_cores--;
1162 } 1291 }
1163 1292
1164 for_each_cpu_mask(sibling, per_cpu(cpu_sibling_map, cpu)) 1293 for_each_cpu_mask(sibling, per_cpu(cpu_sibling_map, cpu))
1165 cpu_clear(cpu, per_cpu(cpu_sibling_map, sibling)); 1294 cpu_clear(cpu, per_cpu(cpu_sibling_map, sibling));
1166 cpus_clear(per_cpu(cpu_sibling_map, cpu)); 1295 cpus_clear(per_cpu(cpu_sibling_map, cpu));
@@ -1170,35 +1299,99 @@ void remove_siblinginfo(int cpu)
1170 cpu_clear(cpu, cpu_sibling_setup_map); 1299 cpu_clear(cpu, cpu_sibling_setup_map);
1171} 1300}
1172 1301
1302int additional_cpus __initdata = -1;
1303
1304static __init int setup_additional_cpus(char *s)
1305{
1306 return s && get_option(&s, &additional_cpus) ? 0 : -EINVAL;
1307}
1308early_param("additional_cpus", setup_additional_cpus);
1309
1310/*
1311 * cpu_possible_map should be static, it cannot change as cpu's
1312 * are onlined, or offlined. The reason is per-cpu data-structures
1313 * are allocated by some modules at init time, and dont expect to
1314 * do this dynamically on cpu arrival/departure.
1315 * cpu_present_map on the other hand can change dynamically.
1316 * In case when cpu_hotplug is not compiled, then we resort to current
1317 * behaviour, which is cpu_possible == cpu_present.
1318 * - Ashok Raj
1319 *
1320 * Three ways to find out the number of additional hotplug CPUs:
1321 * - If the BIOS specified disabled CPUs in ACPI/mptables use that.
1322 * - The user can overwrite it with additional_cpus=NUM
1323 * - Otherwise don't reserve additional CPUs.
1324 * We do this because additional CPUs waste a lot of memory.
1325 * -AK
1326 */
1327__init void prefill_possible_map(void)
1328{
1329 int i;
1330 int possible;
1331
1332 if (additional_cpus == -1) {
1333 if (disabled_cpus > 0)
1334 additional_cpus = disabled_cpus;
1335 else
1336 additional_cpus = 0;
1337 }
1338 possible = num_processors + additional_cpus;
1339 if (possible > NR_CPUS)
1340 possible = NR_CPUS;
1341
1342 printk(KERN_INFO "SMP: Allowing %d CPUs, %d hotplug CPUs\n",
1343 possible, max_t(int, possible - num_processors, 0));
1344
1345 for (i = 0; i < possible; i++)
1346 cpu_set(i, cpu_possible_map);
1347}
1348
1349static void __ref remove_cpu_from_maps(int cpu)
1350{
1351 cpu_clear(cpu, cpu_online_map);
1352#ifdef CONFIG_X86_64
1353 cpu_clear(cpu, cpu_callout_map);
1354 cpu_clear(cpu, cpu_callin_map);
1355 /* was set by cpu_init() */
1356 clear_bit(cpu, (unsigned long *)&cpu_initialized);
1357 clear_node_cpumask(cpu);
1358#endif
1359}
1360
1173int __cpu_disable(void) 1361int __cpu_disable(void)
1174{ 1362{
1175 cpumask_t map = cpu_online_map;
1176 int cpu = smp_processor_id(); 1363 int cpu = smp_processor_id();
1177 1364
1178 /* 1365 /*
1179 * Perhaps use cpufreq to drop frequency, but that could go 1366 * Perhaps use cpufreq to drop frequency, but that could go
1180 * into generic code. 1367 * into generic code.
1181 * 1368 *
1182 * We won't take down the boot processor on i386 due to some 1369 * We won't take down the boot processor on i386 due to some
1183 * interrupts only being able to be serviced by the BSP. 1370 * interrupts only being able to be serviced by the BSP.
1184 * Especially so if we're not using an IOAPIC -zwane 1371 * Especially so if we're not using an IOAPIC -zwane
1185 */ 1372 */
1186 if (cpu == 0) 1373 if (cpu == 0)
1187 return -EBUSY; 1374 return -EBUSY;
1375
1188 if (nmi_watchdog == NMI_LOCAL_APIC) 1376 if (nmi_watchdog == NMI_LOCAL_APIC)
1189 stop_apic_nmi_watchdog(NULL); 1377 stop_apic_nmi_watchdog(NULL);
1190 clear_local_APIC(); 1378 clear_local_APIC();
1191 /* Allow any queued timer interrupts to get serviced */ 1379
1380 /*
1381 * HACK:
1382 * Allow any queued timer interrupts to get serviced
1383 * This is only a temporary solution until we cleanup
1384 * fixup_irqs as we do for IA64.
1385 */
1192 local_irq_enable(); 1386 local_irq_enable();
1193 mdelay(1); 1387 mdelay(1);
1194 local_irq_disable();
1195 1388
1389 local_irq_disable();
1196 remove_siblinginfo(cpu); 1390 remove_siblinginfo(cpu);
1197 1391
1198 cpu_clear(cpu, map);
1199 fixup_irqs(map);
1200 /* It's now safe to remove this processor from the online map */ 1392 /* It's now safe to remove this processor from the online map */
1201 cpu_clear(cpu, cpu_online_map); 1393 remove_cpu_from_maps(cpu);
1394 fixup_irqs(cpu_online_map);
1202 return 0; 1395 return 0;
1203} 1396}
1204 1397
@@ -1210,14 +1403,14 @@ void __cpu_die(unsigned int cpu)
1210 for (i = 0; i < 10; i++) { 1403 for (i = 0; i < 10; i++) {
1211 /* They ack this in play_dead by setting CPU_DEAD */ 1404 /* They ack this in play_dead by setting CPU_DEAD */
1212 if (per_cpu(cpu_state, cpu) == CPU_DEAD) { 1405 if (per_cpu(cpu_state, cpu) == CPU_DEAD) {
1213 printk ("CPU %d is now offline\n", cpu); 1406 printk(KERN_INFO "CPU %d is now offline\n", cpu);
1214 if (1 == num_online_cpus()) 1407 if (1 == num_online_cpus())
1215 alternatives_smp_switch(0); 1408 alternatives_smp_switch(0);
1216 return; 1409 return;
1217 } 1410 }
1218 msleep(100); 1411 msleep(100);
1219 } 1412 }
1220 printk(KERN_ERR "CPU %u didn't die...\n", cpu); 1413 printk(KERN_ERR "CPU %u didn't die...\n", cpu);
1221} 1414}
1222#else /* ... !CONFIG_HOTPLUG_CPU */ 1415#else /* ... !CONFIG_HOTPLUG_CPU */
1223int __cpu_disable(void) 1416int __cpu_disable(void)
@@ -1230,81 +1423,7 @@ void __cpu_die(unsigned int cpu)
1230 /* We said "no" in __cpu_disable */ 1423 /* We said "no" in __cpu_disable */
1231 BUG(); 1424 BUG();
1232} 1425}
1233#endif /* CONFIG_HOTPLUG_CPU */
1234
1235int __cpuinit native_cpu_up(unsigned int cpu)
1236{
1237 unsigned long flags;
1238#ifdef CONFIG_HOTPLUG_CPU
1239 int ret = 0;
1240
1241 /*
1242 * We do warm boot only on cpus that had booted earlier
1243 * Otherwise cold boot is all handled from smp_boot_cpus().
1244 * cpu_callin_map is set during AP kickstart process. Its reset
1245 * when a cpu is taken offline from cpu_exit_clear().
1246 */
1247 if (!cpu_isset(cpu, cpu_callin_map))
1248 ret = __smp_prepare_cpu(cpu);
1249
1250 if (ret)
1251 return -EIO;
1252#endif
1253
1254 /* In case one didn't come up */
1255 if (!cpu_isset(cpu, cpu_callin_map)) {
1256 printk(KERN_DEBUG "skipping cpu%d, didn't come online\n", cpu);
1257 return -EIO;
1258 }
1259
1260 per_cpu(cpu_state, cpu) = CPU_UP_PREPARE;
1261 /* Unleash the CPU! */
1262 cpu_set(cpu, smp_commenced_mask);
1263
1264 /*
1265 * Check TSC synchronization with the AP (keep irqs disabled
1266 * while doing so):
1267 */
1268 local_irq_save(flags);
1269 check_tsc_sync_source(cpu);
1270 local_irq_restore(flags);
1271
1272 while (!cpu_isset(cpu, cpu_online_map)) {
1273 cpu_relax();
1274 touch_nmi_watchdog();
1275 }
1276
1277 return 0;
1278}
1279
1280void __init native_smp_cpus_done(unsigned int max_cpus)
1281{
1282#ifdef CONFIG_X86_IO_APIC
1283 setup_ioapic_dest();
1284#endif 1426#endif
1285 zap_low_mappings();
1286}
1287
1288void __init smp_intr_init(void)
1289{
1290 /*
1291 * IRQ0 must be given a fixed assignment and initialized,
1292 * because it's used before the IO-APIC is set up.
1293 */
1294 set_intr_gate(FIRST_DEVICE_VECTOR, interrupt[0]);
1295
1296 /*
1297 * The reschedule interrupt is a CPU-to-CPU reschedule-helper
1298 * IPI, driven by wakeup.
1299 */
1300 set_intr_gate(RESCHEDULE_VECTOR, reschedule_interrupt);
1301
1302 /* IPI for invalidation */
1303 set_intr_gate(INVALIDATE_TLB_VECTOR, invalidate_interrupt);
1304
1305 /* IPI for generic function call */
1306 set_intr_gate(CALL_FUNCTION_VECTOR, call_function_interrupt);
1307}
1308 1427
1309/* 1428/*
1310 * If the BIOS enumerates physical processors before logical, 1429 * If the BIOS enumerates physical processors before logical,
diff --git a/arch/x86/kernel/smpboot_64.c b/arch/x86/kernel/smpboot_64.c
deleted file mode 100644
index 0880f2c388a9..000000000000
--- a/arch/x86/kernel/smpboot_64.c
+++ /dev/null
@@ -1,1108 +0,0 @@
1/*
2 * x86 SMP booting functions
3 *
4 * (c) 1995 Alan Cox, Building #3 <alan@redhat.com>
5 * (c) 1998, 1999, 2000 Ingo Molnar <mingo@redhat.com>
6 * Copyright 2001 Andi Kleen, SuSE Labs.
7 *
8 * Much of the core SMP work is based on previous work by Thomas Radke, to
9 * whom a great many thanks are extended.
10 *
11 * Thanks to Intel for making available several different Pentium,
12 * Pentium Pro and Pentium-II/Xeon MP machines.
13 * Original development of Linux SMP code supported by Caldera.
14 *
15 * This code is released under the GNU General Public License version 2
16 *
17 * Fixes
18 * Felix Koop : NR_CPUS used properly
19 * Jose Renau : Handle single CPU case.
20 * Alan Cox : By repeated request 8) - Total BogoMIP report.
21 * Greg Wright : Fix for kernel stacks panic.
22 * Erich Boleyn : MP v1.4 and additional changes.
23 * Matthias Sattler : Changes for 2.1 kernel map.
24 * Michel Lespinasse : Changes for 2.1 kernel map.
25 * Michael Chastain : Change trampoline.S to gnu as.
26 * Alan Cox : Dumb bug: 'B' step PPro's are fine
27 * Ingo Molnar : Added APIC timers, based on code
28 * from Jose Renau
29 * Ingo Molnar : various cleanups and rewrites
30 * Tigran Aivazian : fixed "0.00 in /proc/uptime on SMP" bug.
31 * Maciej W. Rozycki : Bits for genuine 82489DX APICs
32 * Andi Kleen : Changed for SMP boot into long mode.
33 * Rusty Russell : Hacked into shape for new "hotplug" boot process.
34 * Andi Kleen : Converted to new state machine.
35 * Various cleanups.
36 * Probably mostly hotplug CPU ready now.
37 * Ashok Raj : CPU hotplug support
38 */
39
40
41#include <linux/init.h>
42
43#include <linux/mm.h>
44#include <linux/kernel_stat.h>
45#include <linux/bootmem.h>
46#include <linux/thread_info.h>
47#include <linux/module.h>
48#include <linux/delay.h>
49#include <linux/mc146818rtc.h>
50#include <linux/smp.h>
51#include <linux/kdebug.h>
52
53#include <asm/mtrr.h>
54#include <asm/pgalloc.h>
55#include <asm/desc.h>
56#include <asm/tlbflush.h>
57#include <asm/proto.h>
58#include <asm/nmi.h>
59#include <asm/irq.h>
60#include <asm/hw_irq.h>
61#include <asm/numa.h>
62
63/* Number of siblings per CPU package */
64int smp_num_siblings = 1;
65EXPORT_SYMBOL(smp_num_siblings);
66
67/* Last level cache ID of each logical CPU */
68DEFINE_PER_CPU(u16, cpu_llc_id) = BAD_APICID;
69
70/* Bitmask of currently online CPUs */
71cpumask_t cpu_online_map __read_mostly;
72
73EXPORT_SYMBOL(cpu_online_map);
74
75/*
76 * Private maps to synchronize booting between AP and BP.
77 * Probably not needed anymore, but it makes for easier debugging. -AK
78 */
79cpumask_t cpu_callin_map;
80cpumask_t cpu_callout_map;
81cpumask_t cpu_possible_map;
82EXPORT_SYMBOL(cpu_possible_map);
83
84/* Per CPU bogomips and other parameters */
85DEFINE_PER_CPU_SHARED_ALIGNED(struct cpuinfo_x86, cpu_info);
86EXPORT_PER_CPU_SYMBOL(cpu_info);
87
88/* Set when the idlers are all forked */
89int smp_threads_ready;
90
91/* representing HT siblings of each logical CPU */
92DEFINE_PER_CPU(cpumask_t, cpu_sibling_map);
93EXPORT_PER_CPU_SYMBOL(cpu_sibling_map);
94
95/* representing HT and core siblings of each logical CPU */
96DEFINE_PER_CPU(cpumask_t, cpu_core_map);
97EXPORT_PER_CPU_SYMBOL(cpu_core_map);
98
99/*
100 * Trampoline 80x86 program as an array.
101 */
102
103extern const unsigned char trampoline_data[];
104extern const unsigned char trampoline_end[];
105
106/* State of each CPU */
107DEFINE_PER_CPU(int, cpu_state) = { 0 };
108
109/*
110 * Store all idle threads, this can be reused instead of creating
111 * a new thread. Also avoids complicated thread destroy functionality
112 * for idle threads.
113 */
114#ifdef CONFIG_HOTPLUG_CPU
115/*
116 * Needed only for CONFIG_HOTPLUG_CPU because __cpuinitdata is
117 * removed after init for !CONFIG_HOTPLUG_CPU.
118 */
119static DEFINE_PER_CPU(struct task_struct *, idle_thread_array);
120#define get_idle_for_cpu(x) (per_cpu(idle_thread_array, x))
121#define set_idle_for_cpu(x,p) (per_cpu(idle_thread_array, x) = (p))
122#else
123struct task_struct *idle_thread_array[NR_CPUS] __cpuinitdata ;
124#define get_idle_for_cpu(x) (idle_thread_array[(x)])
125#define set_idle_for_cpu(x,p) (idle_thread_array[(x)] = (p))
126#endif
127
128
129/*
130 * Currently trivial. Write the real->protected mode
131 * bootstrap into the page concerned. The caller
132 * has made sure it's suitably aligned.
133 */
134
135static unsigned long __cpuinit setup_trampoline(void)
136{
137 void *tramp = __va(SMP_TRAMPOLINE_BASE);
138 memcpy(tramp, trampoline_data, trampoline_end - trampoline_data);
139 return virt_to_phys(tramp);
140}
141
142/*
143 * The bootstrap kernel entry code has set these up. Save them for
144 * a given CPU
145 */
146
147static void __cpuinit smp_store_cpu_info(int id)
148{
149 struct cpuinfo_x86 *c = &cpu_data(id);
150
151 *c = boot_cpu_data;
152 c->cpu_index = id;
153 identify_cpu(c);
154 print_cpu_info(c);
155}
156
157static atomic_t init_deasserted __cpuinitdata;
158
159/*
160 * Report back to the Boot Processor.
161 * Running on AP.
162 */
163void __cpuinit smp_callin(void)
164{
165 int cpuid, phys_id;
166 unsigned long timeout;
167
168 /*
169 * If waken up by an INIT in an 82489DX configuration
170 * we may get here before an INIT-deassert IPI reaches
171 * our local APIC. We have to wait for the IPI or we'll
172 * lock up on an APIC access.
173 */
174 while (!atomic_read(&init_deasserted))
175 cpu_relax();
176
177 /*
178 * (This works even if the APIC is not enabled.)
179 */
180 phys_id = GET_APIC_ID(apic_read(APIC_ID));
181 cpuid = smp_processor_id();
182 if (cpu_isset(cpuid, cpu_callin_map)) {
183 panic("smp_callin: phys CPU#%d, CPU#%d already present??\n",
184 phys_id, cpuid);
185 }
186 Dprintk("CPU#%d (phys ID: %d) waiting for CALLOUT\n", cpuid, phys_id);
187
188 /*
189 * STARTUP IPIs are fragile beasts as they might sometimes
190 * trigger some glue motherboard logic. Complete APIC bus
191 * silence for 1 second, this overestimates the time the
192 * boot CPU is spending to send the up to 2 STARTUP IPIs
193 * by a factor of two. This should be enough.
194 */
195
196 /*
197 * Waiting 2s total for startup (udelay is not yet working)
198 */
199 timeout = jiffies + 2*HZ;
200 while (time_before(jiffies, timeout)) {
201 /*
202 * Has the boot CPU finished it's STARTUP sequence?
203 */
204 if (cpu_isset(cpuid, cpu_callout_map))
205 break;
206 cpu_relax();
207 }
208
209 if (!time_before(jiffies, timeout)) {
210 panic("smp_callin: CPU%d started up but did not get a callout!\n",
211 cpuid);
212 }
213
214 /*
215 * the boot CPU has finished the init stage and is spinning
216 * on callin_map until we finish. We are free to set up this
217 * CPU, first the APIC. (this is probably redundant on most
218 * boards)
219 */
220
221 Dprintk("CALLIN, before setup_local_APIC().\n");
222 setup_local_APIC();
223 end_local_APIC_setup();
224
225 /*
226 * Get our bogomips.
227 *
228 * Need to enable IRQs because it can take longer and then
229 * the NMI watchdog might kill us.
230 */
231 local_irq_enable();
232 calibrate_delay();
233 local_irq_disable();
234 Dprintk("Stack at about %p\n",&cpuid);
235
236 /*
237 * Save our processor parameters
238 */
239 smp_store_cpu_info(cpuid);
240
241 /*
242 * Allow the master to continue.
243 */
244 cpu_set(cpuid, cpu_callin_map);
245}
246
247/* maps the cpu to the sched domain representing multi-core */
248cpumask_t cpu_coregroup_map(int cpu)
249{
250 struct cpuinfo_x86 *c = &cpu_data(cpu);
251 /*
252 * For perf, we return last level cache shared map.
253 * And for power savings, we return cpu_core_map
254 */
255 if (sched_mc_power_savings || sched_smt_power_savings)
256 return per_cpu(cpu_core_map, cpu);
257 else
258 return c->llc_shared_map;
259}
260
261/* representing cpus for which sibling maps can be computed */
262static cpumask_t cpu_sibling_setup_map;
263
264static inline void set_cpu_sibling_map(int cpu)
265{
266 int i;
267 struct cpuinfo_x86 *c = &cpu_data(cpu);
268
269 cpu_set(cpu, cpu_sibling_setup_map);
270
271 if (smp_num_siblings > 1) {
272 for_each_cpu_mask(i, cpu_sibling_setup_map) {
273 if (c->phys_proc_id == cpu_data(i).phys_proc_id &&
274 c->cpu_core_id == cpu_data(i).cpu_core_id) {
275 cpu_set(i, per_cpu(cpu_sibling_map, cpu));
276 cpu_set(cpu, per_cpu(cpu_sibling_map, i));
277 cpu_set(i, per_cpu(cpu_core_map, cpu));
278 cpu_set(cpu, per_cpu(cpu_core_map, i));
279 cpu_set(i, c->llc_shared_map);
280 cpu_set(cpu, cpu_data(i).llc_shared_map);
281 }
282 }
283 } else {
284 cpu_set(cpu, per_cpu(cpu_sibling_map, cpu));
285 }
286
287 cpu_set(cpu, c->llc_shared_map);
288
289 if (current_cpu_data.x86_max_cores == 1) {
290 per_cpu(cpu_core_map, cpu) = per_cpu(cpu_sibling_map, cpu);
291 c->booted_cores = 1;
292 return;
293 }
294
295 for_each_cpu_mask(i, cpu_sibling_setup_map) {
296 if (per_cpu(cpu_llc_id, cpu) != BAD_APICID &&
297 per_cpu(cpu_llc_id, cpu) == per_cpu(cpu_llc_id, i)) {
298 cpu_set(i, c->llc_shared_map);
299 cpu_set(cpu, cpu_data(i).llc_shared_map);
300 }
301 if (c->phys_proc_id == cpu_data(i).phys_proc_id) {
302 cpu_set(i, per_cpu(cpu_core_map, cpu));
303 cpu_set(cpu, per_cpu(cpu_core_map, i));
304 /*
305 * Does this new cpu bringup a new core?
306 */
307 if (cpus_weight(per_cpu(cpu_sibling_map, cpu)) == 1) {
308 /*
309 * for each core in package, increment
310 * the booted_cores for this new cpu
311 */
312 if (first_cpu(per_cpu(cpu_sibling_map, i)) == i)
313 c->booted_cores++;
314 /*
315 * increment the core count for all
316 * the other cpus in this package
317 */
318 if (i != cpu)
319 cpu_data(i).booted_cores++;
320 } else if (i != cpu && !c->booted_cores)
321 c->booted_cores = cpu_data(i).booted_cores;
322 }
323 }
324}
325
326/*
327 * Setup code on secondary processor (after comming out of the trampoline)
328 */
329void __cpuinit start_secondary(void)
330{
331 /*
332 * Dont put anything before smp_callin(), SMP
333 * booting is too fragile that we want to limit the
334 * things done here to the most necessary things.
335 */
336 cpu_init();
337 preempt_disable();
338 smp_callin();
339
340 /* otherwise gcc will move up the smp_processor_id before the cpu_init */
341 barrier();
342
343 /*
344 * Check TSC sync first:
345 */
346 check_tsc_sync_target();
347
348 if (nmi_watchdog == NMI_IO_APIC) {
349 disable_8259A_irq(0);
350 enable_NMI_through_LVT0();
351 enable_8259A_irq(0);
352 }
353
354 /*
355 * The sibling maps must be set before turing the online map on for
356 * this cpu
357 */
358 set_cpu_sibling_map(smp_processor_id());
359
360 /*
361 * We need to hold call_lock, so there is no inconsistency
362 * between the time smp_call_function() determines number of
363 * IPI recipients, and the time when the determination is made
364 * for which cpus receive the IPI in genapic_flat.c. Holding this
365 * lock helps us to not include this cpu in a currently in progress
366 * smp_call_function().
367 */
368 lock_ipi_call_lock();
369 spin_lock(&vector_lock);
370
371 /* Setup the per cpu irq handling data structures */
372 __setup_vector_irq(smp_processor_id());
373 /*
374 * Allow the master to continue.
375 */
376 cpu_set(smp_processor_id(), cpu_online_map);
377 per_cpu(cpu_state, smp_processor_id()) = CPU_ONLINE;
378 spin_unlock(&vector_lock);
379
380 unlock_ipi_call_lock();
381
382 setup_secondary_clock();
383
384 cpu_idle();
385}
386
387extern volatile unsigned long init_rsp;
388extern void (*initial_code)(void);
389
390#ifdef APIC_DEBUG
391static void inquire_remote_apic(int apicid)
392{
393 unsigned i, regs[] = { APIC_ID >> 4, APIC_LVR >> 4, APIC_SPIV >> 4 };
394 char *names[] = { "ID", "VERSION", "SPIV" };
395 int timeout;
396 u32 status;
397
398 printk(KERN_INFO "Inquiring remote APIC #%d...\n", apicid);
399
400 for (i = 0; i < ARRAY_SIZE(regs); i++) {
401 printk(KERN_INFO "... APIC #%d %s: ", apicid, names[i]);
402
403 /*
404 * Wait for idle.
405 */
406 status = safe_apic_wait_icr_idle();
407 if (status)
408 printk(KERN_CONT
409 "a previous APIC delivery may have failed\n");
410
411 apic_write(APIC_ICR2, SET_APIC_DEST_FIELD(apicid));
412 apic_write(APIC_ICR, APIC_DM_REMRD | regs[i]);
413
414 timeout = 0;
415 do {
416 udelay(100);
417 status = apic_read(APIC_ICR) & APIC_ICR_RR_MASK;
418 } while (status == APIC_ICR_RR_INPROG && timeout++ < 1000);
419
420 switch (status) {
421 case APIC_ICR_RR_VALID:
422 status = apic_read(APIC_RRR);
423 printk(KERN_CONT "%08x\n", status);
424 break;
425 default:
426 printk(KERN_CONT "failed\n");
427 }
428 }
429}
430#endif
431
432/*
433 * Kick the secondary to wake up.
434 */
435static int __cpuinit wakeup_secondary_via_INIT(int phys_apicid, unsigned int start_rip)
436{
437 unsigned long send_status, accept_status = 0;
438 int maxlvt, num_starts, j;
439
440 Dprintk("Asserting INIT.\n");
441
442 /*
443 * Turn INIT on target chip
444 */
445 apic_write(APIC_ICR2, SET_APIC_DEST_FIELD(phys_apicid));
446
447 /*
448 * Send IPI
449 */
450 apic_write(APIC_ICR, APIC_INT_LEVELTRIG | APIC_INT_ASSERT
451 | APIC_DM_INIT);
452
453 Dprintk("Waiting for send to finish...\n");
454 send_status = safe_apic_wait_icr_idle();
455
456 mdelay(10);
457
458 Dprintk("Deasserting INIT.\n");
459
460 /* Target chip */
461 apic_write(APIC_ICR2, SET_APIC_DEST_FIELD(phys_apicid));
462
463 /* Send IPI */
464 apic_write(APIC_ICR, APIC_INT_LEVELTRIG | APIC_DM_INIT);
465
466 Dprintk("Waiting for send to finish...\n");
467 send_status = safe_apic_wait_icr_idle();
468
469 mb();
470 atomic_set(&init_deasserted, 1);
471
472 num_starts = 2;
473
474 /*
475 * Run STARTUP IPI loop.
476 */
477 Dprintk("#startup loops: %d.\n", num_starts);
478
479 maxlvt = lapic_get_maxlvt();
480
481 for (j = 1; j <= num_starts; j++) {
482 Dprintk("Sending STARTUP #%d.\n",j);
483 apic_write(APIC_ESR, 0);
484 apic_read(APIC_ESR);
485 Dprintk("After apic_write.\n");
486
487 /*
488 * STARTUP IPI
489 */
490
491 /* Target chip */
492 apic_write(APIC_ICR2, SET_APIC_DEST_FIELD(phys_apicid));
493
494 /* Boot on the stack */
495 /* Kick the second */
496 apic_write(APIC_ICR, APIC_DM_STARTUP | (start_rip >> 12));
497
498 /*
499 * Give the other CPU some time to accept the IPI.
500 */
501 udelay(300);
502
503 Dprintk("Startup point 1.\n");
504
505 Dprintk("Waiting for send to finish...\n");
506 send_status = safe_apic_wait_icr_idle();
507
508 /*
509 * Give the other CPU some time to accept the IPI.
510 */
511 udelay(200);
512 /*
513 * Due to the Pentium erratum 3AP.
514 */
515 if (maxlvt > 3) {
516 apic_write(APIC_ESR, 0);
517 }
518 accept_status = (apic_read(APIC_ESR) & 0xEF);
519 if (send_status || accept_status)
520 break;
521 }
522 Dprintk("After Startup.\n");
523
524 if (send_status)
525 printk(KERN_ERR "APIC never delivered???\n");
526 if (accept_status)
527 printk(KERN_ERR "APIC delivery error (%lx).\n", accept_status);
528
529 return (send_status | accept_status);
530}
531
532struct create_idle {
533 struct work_struct work;
534 struct task_struct *idle;
535 struct completion done;
536 int cpu;
537};
538
539static void __cpuinit do_fork_idle(struct work_struct *work)
540{
541 struct create_idle *c_idle =
542 container_of(work, struct create_idle, work);
543
544 c_idle->idle = fork_idle(c_idle->cpu);
545 complete(&c_idle->done);
546}
547
548/*
549 * Boot one CPU.
550 */
551static int __cpuinit do_boot_cpu(int cpu, int apicid)
552{
553 unsigned long boot_error;
554 int timeout;
555 unsigned long start_rip;
556 struct create_idle c_idle = {
557 .cpu = cpu,
558 .done = COMPLETION_INITIALIZER_ONSTACK(c_idle.done),
559 };
560 INIT_WORK(&c_idle.work, do_fork_idle);
561
562 /* allocate memory for gdts of secondary cpus. Hotplug is considered */
563 if (!cpu_gdt_descr[cpu].address &&
564 !(cpu_gdt_descr[cpu].address = get_zeroed_page(GFP_KERNEL))) {
565 printk(KERN_ERR "Failed to allocate GDT for CPU %d\n", cpu);
566 return -1;
567 }
568
569 /* Allocate node local memory for AP pdas */
570 if (cpu_pda(cpu) == &boot_cpu_pda[cpu]) {
571 struct x8664_pda *newpda, *pda;
572 int node = cpu_to_node(cpu);
573 pda = cpu_pda(cpu);
574 newpda = kmalloc_node(sizeof (struct x8664_pda), GFP_ATOMIC,
575 node);
576 if (newpda) {
577 memcpy(newpda, pda, sizeof (struct x8664_pda));
578 cpu_pda(cpu) = newpda;
579 } else
580 printk(KERN_ERR
581 "Could not allocate node local PDA for CPU %d on node %d\n",
582 cpu, node);
583 }
584
585 alternatives_smp_switch(1);
586
587 c_idle.idle = get_idle_for_cpu(cpu);
588
589 if (c_idle.idle) {
590 c_idle.idle->thread.sp = (unsigned long) (((struct pt_regs *)
591 (THREAD_SIZE + task_stack_page(c_idle.idle))) - 1);
592 init_idle(c_idle.idle, cpu);
593 goto do_rest;
594 }
595
596 /*
597 * During cold boot process, keventd thread is not spun up yet.
598 * When we do cpu hot-add, we create idle threads on the fly, we should
599 * not acquire any attributes from the calling context. Hence the clean
600 * way to create kernel_threads() is to do that from keventd().
601 * We do the current_is_keventd() due to the fact that ACPI notifier
602 * was also queuing to keventd() and when the caller is already running
603 * in context of keventd(), we would end up with locking up the keventd
604 * thread.
605 */
606 if (!keventd_up() || current_is_keventd())
607 c_idle.work.func(&c_idle.work);
608 else {
609 schedule_work(&c_idle.work);
610 wait_for_completion(&c_idle.done);
611 }
612
613 if (IS_ERR(c_idle.idle)) {
614 printk("failed fork for CPU %d\n", cpu);
615 return PTR_ERR(c_idle.idle);
616 }
617
618 set_idle_for_cpu(cpu, c_idle.idle);
619
620do_rest:
621
622 cpu_pda(cpu)->pcurrent = c_idle.idle;
623
624 start_rip = setup_trampoline();
625
626 init_rsp = c_idle.idle->thread.sp;
627 load_sp0(&per_cpu(init_tss, cpu), &c_idle.idle->thread);
628 initial_code = start_secondary;
629 clear_tsk_thread_flag(c_idle.idle, TIF_FORK);
630
631 printk(KERN_INFO "Booting processor %d/%d APIC 0x%x\n", cpu,
632 cpus_weight(cpu_present_map),
633 apicid);
634
635 /*
636 * This grunge runs the startup process for
637 * the targeted processor.
638 */
639
640 atomic_set(&init_deasserted, 0);
641
642 Dprintk("Setting warm reset code and vector.\n");
643
644 CMOS_WRITE(0xa, 0xf);
645 local_flush_tlb();
646 Dprintk("1.\n");
647 *((volatile unsigned short *) phys_to_virt(0x469)) = start_rip >> 4;
648 Dprintk("2.\n");
649 *((volatile unsigned short *) phys_to_virt(0x467)) = start_rip & 0xf;
650 Dprintk("3.\n");
651
652 /*
653 * Be paranoid about clearing APIC errors.
654 */
655 apic_write(APIC_ESR, 0);
656 apic_read(APIC_ESR);
657
658 /*
659 * Status is now clean
660 */
661 boot_error = 0;
662
663 /*
664 * Starting actual IPI sequence...
665 */
666 boot_error = wakeup_secondary_via_INIT(apicid, start_rip);
667
668 if (!boot_error) {
669 /*
670 * allow APs to start initializing.
671 */
672 Dprintk("Before Callout %d.\n", cpu);
673 cpu_set(cpu, cpu_callout_map);
674 Dprintk("After Callout %d.\n", cpu);
675
676 /*
677 * Wait 5s total for a response
678 */
679 for (timeout = 0; timeout < 50000; timeout++) {
680 if (cpu_isset(cpu, cpu_callin_map))
681 break; /* It has booted */
682 udelay(100);
683 }
684
685 if (cpu_isset(cpu, cpu_callin_map)) {
686 /* number CPUs logically, starting from 1 (BSP is 0) */
687 Dprintk("CPU has booted.\n");
688 } else {
689 boot_error = 1;
690 if (*((volatile unsigned char *)phys_to_virt(SMP_TRAMPOLINE_BASE))
691 == 0xA5)
692 /* trampoline started but...? */
693 printk("Stuck ??\n");
694 else
695 /* trampoline code not run */
696 printk("Not responding.\n");
697#ifdef APIC_DEBUG
698 inquire_remote_apic(apicid);
699#endif
700 }
701 }
702 if (boot_error) {
703 cpu_clear(cpu, cpu_callout_map); /* was set here (do_boot_cpu()) */
704 clear_bit(cpu, (unsigned long *)&cpu_initialized); /* was set by cpu_init() */
705 clear_node_cpumask(cpu); /* was set by numa_add_cpu */
706 cpu_clear(cpu, cpu_present_map);
707 cpu_clear(cpu, cpu_possible_map);
708 per_cpu(x86_cpu_to_apicid, cpu) = BAD_APICID;
709 return -EIO;
710 }
711
712 return 0;
713}
714
715cycles_t cacheflush_time;
716unsigned long cache_decay_ticks;
717
718/*
719 * Cleanup possible dangling ends...
720 */
721static __cpuinit void smp_cleanup_boot(void)
722{
723 /*
724 * Paranoid: Set warm reset code and vector here back
725 * to default values.
726 */
727 CMOS_WRITE(0, 0xf);
728
729 /*
730 * Reset trampoline flag
731 */
732 *((volatile int *) phys_to_virt(0x467)) = 0;
733}
734
735/*
736 * Fall back to non SMP mode after errors.
737 *
738 * RED-PEN audit/test this more. I bet there is more state messed up here.
739 */
740static __init void disable_smp(void)
741{
742 cpu_present_map = cpumask_of_cpu(0);
743 cpu_possible_map = cpumask_of_cpu(0);
744 if (smp_found_config)
745 phys_cpu_present_map = physid_mask_of_physid(boot_cpu_id);
746 else
747 phys_cpu_present_map = physid_mask_of_physid(0);
748 cpu_set(0, per_cpu(cpu_sibling_map, 0));
749 cpu_set(0, per_cpu(cpu_core_map, 0));
750}
751
752#ifdef CONFIG_HOTPLUG_CPU
753
754int additional_cpus __initdata = -1;
755
756/*
757 * cpu_possible_map should be static, it cannot change as cpu's
758 * are onlined, or offlined. The reason is per-cpu data-structures
759 * are allocated by some modules at init time, and dont expect to
760 * do this dynamically on cpu arrival/departure.
761 * cpu_present_map on the other hand can change dynamically.
762 * In case when cpu_hotplug is not compiled, then we resort to current
763 * behaviour, which is cpu_possible == cpu_present.
764 * - Ashok Raj
765 *
766 * Three ways to find out the number of additional hotplug CPUs:
767 * - If the BIOS specified disabled CPUs in ACPI/mptables use that.
768 * - The user can overwrite it with additional_cpus=NUM
769 * - Otherwise don't reserve additional CPUs.
770 * We do this because additional CPUs waste a lot of memory.
771 * -AK
772 */
773__init void prefill_possible_map(void)
774{
775 int i;
776 int possible;
777
778 if (additional_cpus == -1) {
779 if (disabled_cpus > 0)
780 additional_cpus = disabled_cpus;
781 else
782 additional_cpus = 0;
783 }
784 possible = num_processors + additional_cpus;
785 if (possible > NR_CPUS)
786 possible = NR_CPUS;
787
788 printk(KERN_INFO "SMP: Allowing %d CPUs, %d hotplug CPUs\n",
789 possible,
790 max_t(int, possible - num_processors, 0));
791
792 for (i = 0; i < possible; i++)
793 cpu_set(i, cpu_possible_map);
794}
795#endif
796
797/*
798 * Various sanity checks.
799 */
800static int __init smp_sanity_check(unsigned max_cpus)
801{
802 if (!physid_isset(hard_smp_processor_id(), phys_cpu_present_map)) {
803 printk("weird, boot CPU (#%d) not listed by the BIOS.\n",
804 hard_smp_processor_id());
805 physid_set(hard_smp_processor_id(), phys_cpu_present_map);
806 }
807
808 /*
809 * If we couldn't find an SMP configuration at boot time,
810 * get out of here now!
811 */
812 if (!smp_found_config) {
813 printk(KERN_NOTICE "SMP motherboard not detected.\n");
814 disable_smp();
815 if (APIC_init_uniprocessor())
816 printk(KERN_NOTICE "Local APIC not detected."
817 " Using dummy APIC emulation.\n");
818 return -1;
819 }
820
821 /*
822 * Should not be necessary because the MP table should list the boot
823 * CPU too, but we do it for the sake of robustness anyway.
824 */
825 if (!physid_isset(boot_cpu_id, phys_cpu_present_map)) {
826 printk(KERN_NOTICE "weird, boot CPU (#%d) not listed by the BIOS.\n",
827 boot_cpu_id);
828 physid_set(hard_smp_processor_id(), phys_cpu_present_map);
829 }
830
831 /*
832 * If we couldn't find a local APIC, then get out of here now!
833 */
834 if (!cpu_has_apic) {
835 printk(KERN_ERR "BIOS bug, local APIC #%d not detected!...\n",
836 boot_cpu_id);
837 printk(KERN_ERR "... forcing use of dummy APIC emulation. (tell your hw vendor)\n");
838 nr_ioapics = 0;
839 return -1;
840 }
841
842 /*
843 * If SMP should be disabled, then really disable it!
844 */
845 if (!max_cpus) {
846 printk(KERN_INFO "SMP mode deactivated, forcing use of dummy APIC emulation.\n");
847 nr_ioapics = 0;
848 return -1;
849 }
850
851 return 0;
852}
853
854static void __init smp_cpu_index_default(void)
855{
856 int i;
857 struct cpuinfo_x86 *c;
858
859 for_each_cpu_mask(i, cpu_possible_map) {
860 c = &cpu_data(i);
861 /* mark all to hotplug */
862 c->cpu_index = NR_CPUS;
863 }
864}
865
866/*
867 * Prepare for SMP bootup. The MP table or ACPI has been read
868 * earlier. Just do some sanity checking here and enable APIC mode.
869 */
870void __init smp_prepare_cpus(unsigned int max_cpus)
871{
872 nmi_watchdog_default();
873 smp_cpu_index_default();
874 current_cpu_data = boot_cpu_data;
875 current_thread_info()->cpu = 0; /* needed? */
876 set_cpu_sibling_map(0);
877
878 if (smp_sanity_check(max_cpus) < 0) {
879 printk(KERN_INFO "SMP disabled\n");
880 disable_smp();
881 return;
882 }
883
884
885 /*
886 * Switch from PIC to APIC mode.
887 */
888 setup_local_APIC();
889
890 /*
891 * Enable IO APIC before setting up error vector
892 */
893 if (!skip_ioapic_setup && nr_ioapics)
894 enable_IO_APIC();
895 end_local_APIC_setup();
896
897 if (GET_APIC_ID(apic_read(APIC_ID)) != boot_cpu_id) {
898 panic("Boot APIC ID in local APIC unexpected (%d vs %d)",
899 GET_APIC_ID(apic_read(APIC_ID)), boot_cpu_id);
900 /* Or can we switch back to PIC here? */
901 }
902
903 /*
904 * Now start the IO-APICs
905 */
906 if (!skip_ioapic_setup && nr_ioapics)
907 setup_IO_APIC();
908 else
909 nr_ioapics = 0;
910
911 /*
912 * Set up local APIC timer on boot CPU.
913 */
914
915 setup_boot_clock();
916}
917
918/*
919 * Early setup to make printk work.
920 */
921void __init smp_prepare_boot_cpu(void)
922{
923 int me = smp_processor_id();
924 /* already set me in cpu_online_map in boot_cpu_init() */
925 cpu_set(me, cpu_callout_map);
926 per_cpu(cpu_state, me) = CPU_ONLINE;
927}
928
929/*
930 * Entry point to boot a CPU.
931 */
932int __cpuinit __cpu_up(unsigned int cpu)
933{
934 int apicid = cpu_present_to_apicid(cpu);
935 unsigned long flags;
936 int err;
937
938 WARN_ON(irqs_disabled());
939
940 Dprintk("++++++++++++++++++++=_---CPU UP %u\n", cpu);
941
942 if (apicid == BAD_APICID || apicid == boot_cpu_id ||
943 !physid_isset(apicid, phys_cpu_present_map)) {
944 printk("__cpu_up: bad cpu %d\n", cpu);
945 return -EINVAL;
946 }
947
948 /*
949 * Already booted CPU?
950 */
951 if (cpu_isset(cpu, cpu_callin_map)) {
952 Dprintk("do_boot_cpu %d Already started\n", cpu);
953 return -ENOSYS;
954 }
955
956 /*
957 * Save current MTRR state in case it was changed since early boot
958 * (e.g. by the ACPI SMI) to initialize new CPUs with MTRRs in sync:
959 */
960 mtrr_save_state();
961
962 per_cpu(cpu_state, cpu) = CPU_UP_PREPARE;
963 /* Boot it! */
964 err = do_boot_cpu(cpu, apicid);
965 if (err < 0) {
966 Dprintk("do_boot_cpu failed %d\n", err);
967 return err;
968 }
969
970 /* Unleash the CPU! */
971 Dprintk("waiting for cpu %d\n", cpu);
972
973 /*
974 * Make sure and check TSC sync:
975 */
976 local_irq_save(flags);
977 check_tsc_sync_source(cpu);
978 local_irq_restore(flags);
979
980 while (!cpu_isset(cpu, cpu_online_map))
981 cpu_relax();
982 err = 0;
983
984 return err;
985}
986
987/*
988 * Finish the SMP boot.
989 */
990void __init smp_cpus_done(unsigned int max_cpus)
991{
992 smp_cleanup_boot();
993 setup_ioapic_dest();
994 check_nmi_watchdog();
995}
996
997#ifdef CONFIG_HOTPLUG_CPU
998
999static void remove_siblinginfo(int cpu)
1000{
1001 int sibling;
1002 struct cpuinfo_x86 *c = &cpu_data(cpu);
1003
1004 for_each_cpu_mask(sibling, per_cpu(cpu_core_map, cpu)) {
1005 cpu_clear(cpu, per_cpu(cpu_core_map, sibling));
1006 /*
1007 * last thread sibling in this cpu core going down
1008 */
1009 if (cpus_weight(per_cpu(cpu_sibling_map, cpu)) == 1)
1010 cpu_data(sibling).booted_cores--;
1011 }
1012
1013 for_each_cpu_mask(sibling, per_cpu(cpu_sibling_map, cpu))
1014 cpu_clear(cpu, per_cpu(cpu_sibling_map, sibling));
1015 cpus_clear(per_cpu(cpu_sibling_map, cpu));
1016 cpus_clear(per_cpu(cpu_core_map, cpu));
1017 c->phys_proc_id = 0;
1018 c->cpu_core_id = 0;
1019 cpu_clear(cpu, cpu_sibling_setup_map);
1020}
1021
1022static void __ref remove_cpu_from_maps(void)
1023{
1024 int cpu = smp_processor_id();
1025
1026 cpu_clear(cpu, cpu_callout_map);
1027 cpu_clear(cpu, cpu_callin_map);
1028 clear_bit(cpu, (unsigned long *)&cpu_initialized); /* was set by cpu_init() */
1029 clear_node_cpumask(cpu);
1030}
1031
1032int __cpu_disable(void)
1033{
1034 int cpu = smp_processor_id();
1035
1036 /*
1037 * Perhaps use cpufreq to drop frequency, but that could go
1038 * into generic code.
1039 *
1040 * We won't take down the boot processor on i386 due to some
1041 * interrupts only being able to be serviced by the BSP.
1042 * Especially so if we're not using an IOAPIC -zwane
1043 */
1044 if (cpu == 0)
1045 return -EBUSY;
1046
1047 if (nmi_watchdog == NMI_LOCAL_APIC)
1048 stop_apic_nmi_watchdog(NULL);
1049 clear_local_APIC();
1050
1051 /*
1052 * HACK:
1053 * Allow any queued timer interrupts to get serviced
1054 * This is only a temporary solution until we cleanup
1055 * fixup_irqs as we do for IA64.
1056 */
1057 local_irq_enable();
1058 mdelay(1);
1059
1060 local_irq_disable();
1061 remove_siblinginfo(cpu);
1062
1063 spin_lock(&vector_lock);
1064 /* It's now safe to remove this processor from the online map */
1065 cpu_clear(cpu, cpu_online_map);
1066 spin_unlock(&vector_lock);
1067 remove_cpu_from_maps();
1068 fixup_irqs(cpu_online_map);
1069 return 0;
1070}
1071
1072void __cpu_die(unsigned int cpu)
1073{
1074 /* We don't do anything here: idle task is faking death itself. */
1075 unsigned int i;
1076
1077 for (i = 0; i < 10; i++) {
1078 /* They ack this in play_dead by setting CPU_DEAD */
1079 if (per_cpu(cpu_state, cpu) == CPU_DEAD) {
1080 printk ("CPU %d is now offline\n", cpu);
1081 if (1 == num_online_cpus())
1082 alternatives_smp_switch(0);
1083 return;
1084 }
1085 msleep(100);
1086 }
1087 printk(KERN_ERR "CPU %u didn't die...\n", cpu);
1088}
1089
1090static __init int setup_additional_cpus(char *s)
1091{
1092 return s && get_option(&s, &additional_cpus) ? 0 : -EINVAL;
1093}
1094early_param("additional_cpus", setup_additional_cpus);
1095
1096#else /* ... !CONFIG_HOTPLUG_CPU */
1097
1098int __cpu_disable(void)
1099{
1100 return -ENOSYS;
1101}
1102
1103void __cpu_die(unsigned int cpu)
1104{
1105 /* We said "no" in __cpu_disable */
1106 BUG();
1107}
1108#endif /* CONFIG_HOTPLUG_CPU */
diff --git a/arch/x86/kernel/smpcommon.c b/arch/x86/kernel/smpcommon.c
new file mode 100644
index 000000000000..3449064d141a
--- /dev/null
+++ b/arch/x86/kernel/smpcommon.c
@@ -0,0 +1,83 @@
1/*
2 * SMP stuff which is common to all sub-architectures.
3 */
4#include <linux/module.h>
5#include <asm/smp.h>
6
7#ifdef CONFIG_X86_32
8DEFINE_PER_CPU(unsigned long, this_cpu_off);
9EXPORT_PER_CPU_SYMBOL(this_cpu_off);
10
11/* Initialize the CPU's GDT. This is either the boot CPU doing itself
12 (still using the master per-cpu area), or a CPU doing it for a
13 secondary which will soon come up. */
14__cpuinit void init_gdt(int cpu)
15{
16 struct desc_struct *gdt = get_cpu_gdt_table(cpu);
17
18 pack_descriptor(&gdt[GDT_ENTRY_PERCPU],
19 __per_cpu_offset[cpu], 0xFFFFF,
20 0x2 | DESCTYPE_S, 0x8);
21
22 gdt[GDT_ENTRY_PERCPU].s = 1;
23
24 per_cpu(this_cpu_off, cpu) = __per_cpu_offset[cpu];
25 per_cpu(cpu_number, cpu) = cpu;
26}
27#endif
28
29/**
30 * smp_call_function(): Run a function on all other CPUs.
31 * @func: The function to run. This must be fast and non-blocking.
32 * @info: An arbitrary pointer to pass to the function.
33 * @nonatomic: Unused.
34 * @wait: If true, wait (atomically) until function has completed on other CPUs.
35 *
36 * Returns 0 on success, else a negative status code.
37 *
38 * If @wait is true, then returns once @func has returned; otherwise
39 * it returns just before the target cpu calls @func.
40 *
41 * You must not call this function with disabled interrupts or from a
42 * hardware interrupt handler or from a bottom half handler.
43 */
44int smp_call_function(void (*func) (void *info), void *info, int nonatomic,
45 int wait)
46{
47 return smp_call_function_mask(cpu_online_map, func, info, wait);
48}
49EXPORT_SYMBOL(smp_call_function);
50
51/**
52 * smp_call_function_single - Run a function on a specific CPU
53 * @cpu: The target CPU. Cannot be the calling CPU.
54 * @func: The function to run. This must be fast and non-blocking.
55 * @info: An arbitrary pointer to pass to the function.
56 * @nonatomic: Unused.
57 * @wait: If true, wait until function has completed on other CPUs.
58 *
59 * Returns 0 on success, else a negative status code.
60 *
61 * If @wait is true, then returns once @func has returned; otherwise
62 * it returns just before the target cpu calls @func.
63 */
64int smp_call_function_single(int cpu, void (*func) (void *info), void *info,
65 int nonatomic, int wait)
66{
67 /* prevent preemption and reschedule on another processor */
68 int ret;
69 int me = get_cpu();
70 if (cpu == me) {
71 local_irq_disable();
72 func(info);
73 local_irq_enable();
74 put_cpu();
75 return 0;
76 }
77
78 ret = smp_call_function_mask(cpumask_of_cpu(cpu), func, info, wait);
79
80 put_cpu();
81 return ret;
82}
83EXPORT_SYMBOL(smp_call_function_single);
diff --git a/arch/x86/kernel/smpcommon_32.c b/arch/x86/kernel/smpcommon_32.c
index 8bc38af29aef..8b137891791f 100644
--- a/arch/x86/kernel/smpcommon_32.c
+++ b/arch/x86/kernel/smpcommon_32.c
@@ -1,82 +1 @@
1/*
2 * SMP stuff which is common to all sub-architectures.
3 */
4#include <linux/module.h>
5#include <asm/smp.h>
6
7DEFINE_PER_CPU(unsigned long, this_cpu_off);
8EXPORT_PER_CPU_SYMBOL(this_cpu_off);
9
10/* Initialize the CPU's GDT. This is either the boot CPU doing itself
11 (still using the master per-cpu area), or a CPU doing it for a
12 secondary which will soon come up. */
13__cpuinit void init_gdt(int cpu)
14{
15 struct desc_struct *gdt = get_cpu_gdt_table(cpu);
16
17 pack_descriptor(&gdt[GDT_ENTRY_PERCPU],
18 __per_cpu_offset[cpu], 0xFFFFF,
19 0x2 | DESCTYPE_S, 0x8);
20
21 gdt[GDT_ENTRY_PERCPU].s = 1;
22
23 per_cpu(this_cpu_off, cpu) = __per_cpu_offset[cpu];
24 per_cpu(cpu_number, cpu) = cpu;
25}
26
27
28/**
29 * smp_call_function(): Run a function on all other CPUs.
30 * @func: The function to run. This must be fast and non-blocking.
31 * @info: An arbitrary pointer to pass to the function.
32 * @nonatomic: Unused.
33 * @wait: If true, wait (atomically) until function has completed on other CPUs.
34 *
35 * Returns 0 on success, else a negative status code.
36 *
37 * If @wait is true, then returns once @func has returned; otherwise
38 * it returns just before the target cpu calls @func.
39 *
40 * You must not call this function with disabled interrupts or from a
41 * hardware interrupt handler or from a bottom half handler.
42 */
43int smp_call_function(void (*func) (void *info), void *info, int nonatomic,
44 int wait)
45{
46 return smp_call_function_mask(cpu_online_map, func, info, wait);
47}
48EXPORT_SYMBOL(smp_call_function);
49
50/**
51 * smp_call_function_single - Run a function on a specific CPU
52 * @cpu: The target CPU. Cannot be the calling CPU.
53 * @func: The function to run. This must be fast and non-blocking.
54 * @info: An arbitrary pointer to pass to the function.
55 * @nonatomic: Unused.
56 * @wait: If true, wait until function has completed on other CPUs.
57 *
58 * Returns 0 on success, else a negative status code.
59 *
60 * If @wait is true, then returns once @func has returned; otherwise
61 * it returns just before the target cpu calls @func.
62 */
63int smp_call_function_single(int cpu, void (*func) (void *info), void *info,
64 int nonatomic, int wait)
65{
66 /* prevent preemption and reschedule on another processor */
67 int ret;
68 int me = get_cpu();
69 if (cpu == me) {
70 local_irq_disable();
71 func(info);
72 local_irq_enable();
73 put_cpu();
74 return 0;
75 }
76
77 ret = smp_call_function_mask(cpumask_of_cpu(cpu), func, info, wait);
78
79 put_cpu();
80 return ret;
81}
82EXPORT_SYMBOL(smp_call_function_single);
diff --git a/arch/x86/kernel/srat_32.c b/arch/x86/kernel/srat_32.c
index b72e61359c36..70e4a374b4e8 100644
--- a/arch/x86/kernel/srat_32.c
+++ b/arch/x86/kernel/srat_32.c
@@ -277,14 +277,14 @@ int __init get_memcfg_from_srat(void)
277 rsdp_address = acpi_os_get_root_pointer(); 277 rsdp_address = acpi_os_get_root_pointer();
278 if (!rsdp_address) { 278 if (!rsdp_address) {
279 printk("%s: System description tables not found\n", 279 printk("%s: System description tables not found\n",
280 __FUNCTION__); 280 __func__);
281 goto out_err; 281 goto out_err;
282 } 282 }
283 283
284 printk("%s: assigning address to rsdp\n", __FUNCTION__); 284 printk("%s: assigning address to rsdp\n", __func__);
285 rsdp = (struct acpi_table_rsdp *)(u32)rsdp_address; 285 rsdp = (struct acpi_table_rsdp *)(u32)rsdp_address;
286 if (!rsdp) { 286 if (!rsdp) {
287 printk("%s: Didn't find ACPI root!\n", __FUNCTION__); 287 printk("%s: Didn't find ACPI root!\n", __func__);
288 goto out_err; 288 goto out_err;
289 } 289 }
290 290
@@ -292,7 +292,7 @@ int __init get_memcfg_from_srat(void)
292 rsdp->oem_id); 292 rsdp->oem_id);
293 293
294 if (strncmp(rsdp->signature, ACPI_SIG_RSDP,strlen(ACPI_SIG_RSDP))) { 294 if (strncmp(rsdp->signature, ACPI_SIG_RSDP,strlen(ACPI_SIG_RSDP))) {
295 printk(KERN_WARNING "%s: RSDP table signature incorrect\n", __FUNCTION__); 295 printk(KERN_WARNING "%s: RSDP table signature incorrect\n", __func__);
296 goto out_err; 296 goto out_err;
297 } 297 }
298 298
@@ -302,7 +302,7 @@ int __init get_memcfg_from_srat(void)
302 if (!rsdt) { 302 if (!rsdt) {
303 printk(KERN_WARNING 303 printk(KERN_WARNING
304 "%s: ACPI: Invalid root system description tables (RSDT)\n", 304 "%s: ACPI: Invalid root system description tables (RSDT)\n",
305 __FUNCTION__); 305 __func__);
306 goto out_err; 306 goto out_err;
307 } 307 }
308 308
diff --git a/arch/x86/kernel/step.c b/arch/x86/kernel/step.c
index 071ff4798236..92c20fee6781 100644
--- a/arch/x86/kernel/step.c
+++ b/arch/x86/kernel/step.c
@@ -148,7 +148,7 @@ static void write_debugctlmsr(struct task_struct *child, unsigned long val)
148 if (child != current) 148 if (child != current)
149 return; 149 return;
150 150
151 wrmsrl(MSR_IA32_DEBUGCTLMSR, val); 151 update_debugctlmsr(val);
152} 152}
153 153
154/* 154/*
diff --git a/arch/x86/kernel/summit_32.c b/arch/x86/kernel/summit_32.c
index 72f463401592..6878a9c2df5d 100644
--- a/arch/x86/kernel/summit_32.c
+++ b/arch/x86/kernel/summit_32.c
@@ -35,43 +35,47 @@ static struct rio_table_hdr *rio_table_hdr __initdata;
35static struct scal_detail *scal_devs[MAX_NUMNODES] __initdata; 35static struct scal_detail *scal_devs[MAX_NUMNODES] __initdata;
36static struct rio_detail *rio_devs[MAX_NUMNODES*4] __initdata; 36static struct rio_detail *rio_devs[MAX_NUMNODES*4] __initdata;
37 37
38static int mp_bus_id_to_node[MAX_MP_BUSSES] __initdata;
39
38static int __init setup_pci_node_map_for_wpeg(int wpeg_num, int last_bus) 40static int __init setup_pci_node_map_for_wpeg(int wpeg_num, int last_bus)
39{ 41{
40 int twister = 0, node = 0; 42 int twister = 0, node = 0;
41 int i, bus, num_buses; 43 int i, bus, num_buses;
42 44
43 for(i = 0; i < rio_table_hdr->num_rio_dev; i++){ 45 for (i = 0; i < rio_table_hdr->num_rio_dev; i++) {
44 if (rio_devs[i]->node_id == rio_devs[wpeg_num]->owner_id){ 46 if (rio_devs[i]->node_id == rio_devs[wpeg_num]->owner_id) {
45 twister = rio_devs[i]->owner_id; 47 twister = rio_devs[i]->owner_id;
46 break; 48 break;
47 } 49 }
48 } 50 }
49 if (i == rio_table_hdr->num_rio_dev){ 51 if (i == rio_table_hdr->num_rio_dev) {
50 printk(KERN_ERR "%s: Couldn't find owner Cyclone for Winnipeg!\n", __FUNCTION__); 52 printk(KERN_ERR "%s: Couldn't find owner Cyclone for Winnipeg!\n", __func__);
51 return last_bus; 53 return last_bus;
52 } 54 }
53 55
54 for(i = 0; i < rio_table_hdr->num_scal_dev; i++){ 56 for (i = 0; i < rio_table_hdr->num_scal_dev; i++) {
55 if (scal_devs[i]->node_id == twister){ 57 if (scal_devs[i]->node_id == twister) {
56 node = scal_devs[i]->node_id; 58 node = scal_devs[i]->node_id;
57 break; 59 break;
58 } 60 }
59 } 61 }
60 if (i == rio_table_hdr->num_scal_dev){ 62 if (i == rio_table_hdr->num_scal_dev) {
61 printk(KERN_ERR "%s: Couldn't find owner Twister for Cyclone!\n", __FUNCTION__); 63 printk(KERN_ERR "%s: Couldn't find owner Twister for Cyclone!\n", __func__);
62 return last_bus; 64 return last_bus;
63 } 65 }
64 66
65 switch (rio_devs[wpeg_num]->type){ 67 switch (rio_devs[wpeg_num]->type) {
66 case CompatWPEG: 68 case CompatWPEG:
67 /* The Compatibility Winnipeg controls the 2 legacy buses, 69 /*
70 * The Compatibility Winnipeg controls the 2 legacy buses,
68 * the 66MHz PCI bus [2 slots] and the 2 "extra" buses in case 71 * the 66MHz PCI bus [2 slots] and the 2 "extra" buses in case
69 * a PCI-PCI bridge card is used in either slot: total 5 buses. 72 * a PCI-PCI bridge card is used in either slot: total 5 buses.
70 */ 73 */
71 num_buses = 5; 74 num_buses = 5;
72 break; 75 break;
73 case AltWPEG: 76 case AltWPEG:
74 /* The Alternate Winnipeg controls the 2 133MHz buses [1 slot 77 /*
78 * The Alternate Winnipeg controls the 2 133MHz buses [1 slot
75 * each], their 2 "extra" buses, the 100MHz bus [2 slots] and 79 * each], their 2 "extra" buses, the 100MHz bus [2 slots] and
76 * the "extra" buses for each of those slots: total 7 buses. 80 * the "extra" buses for each of those slots: total 7 buses.
77 */ 81 */
@@ -79,17 +83,18 @@ static int __init setup_pci_node_map_for_wpeg(int wpeg_num, int last_bus)
79 break; 83 break;
80 case LookOutAWPEG: 84 case LookOutAWPEG:
81 case LookOutBWPEG: 85 case LookOutBWPEG:
82 /* A Lookout Winnipeg controls 3 100MHz buses [2 slots each] 86 /*
87 * A Lookout Winnipeg controls 3 100MHz buses [2 slots each]
83 * & the "extra" buses for each of those slots: total 9 buses. 88 * & the "extra" buses for each of those slots: total 9 buses.
84 */ 89 */
85 num_buses = 9; 90 num_buses = 9;
86 break; 91 break;
87 default: 92 default:
88 printk(KERN_INFO "%s: Unsupported Winnipeg type!\n", __FUNCTION__); 93 printk(KERN_INFO "%s: Unsupported Winnipeg type!\n", __func__);
89 return last_bus; 94 return last_bus;
90 } 95 }
91 96
92 for(bus = last_bus; bus < last_bus + num_buses; bus++) 97 for (bus = last_bus; bus < last_bus + num_buses; bus++)
93 mp_bus_id_to_node[bus] = node; 98 mp_bus_id_to_node[bus] = node;
94 return bus; 99 return bus;
95} 100}
@@ -99,14 +104,14 @@ static int __init build_detail_arrays(void)
99 unsigned long ptr; 104 unsigned long ptr;
100 int i, scal_detail_size, rio_detail_size; 105 int i, scal_detail_size, rio_detail_size;
101 106
102 if (rio_table_hdr->num_scal_dev > MAX_NUMNODES){ 107 if (rio_table_hdr->num_scal_dev > MAX_NUMNODES) {
103 printk(KERN_WARNING "%s: MAX_NUMNODES too low! Defined as %d, but system has %d nodes.\n", __FUNCTION__, MAX_NUMNODES, rio_table_hdr->num_scal_dev); 108 printk(KERN_WARNING "%s: MAX_NUMNODES too low! Defined as %d, but system has %d nodes.\n", __func__, MAX_NUMNODES, rio_table_hdr->num_scal_dev);
104 return 0; 109 return 0;
105 } 110 }
106 111
107 switch (rio_table_hdr->version){ 112 switch (rio_table_hdr->version) {
108 default: 113 default:
109 printk(KERN_WARNING "%s: Invalid Rio Grande Table Version: %d\n", __FUNCTION__, rio_table_hdr->version); 114 printk(KERN_WARNING "%s: Invalid Rio Grande Table Version: %d\n", __func__, rio_table_hdr->version);
110 return 0; 115 return 0;
111 case 2: 116 case 2:
112 scal_detail_size = 11; 117 scal_detail_size = 11;
@@ -119,10 +124,10 @@ static int __init build_detail_arrays(void)
119 } 124 }
120 125
121 ptr = (unsigned long)rio_table_hdr + 3; 126 ptr = (unsigned long)rio_table_hdr + 3;
122 for(i = 0; i < rio_table_hdr->num_scal_dev; i++, ptr += scal_detail_size) 127 for (i = 0; i < rio_table_hdr->num_scal_dev; i++, ptr += scal_detail_size)
123 scal_devs[i] = (struct scal_detail *)ptr; 128 scal_devs[i] = (struct scal_detail *)ptr;
124 129
125 for(i = 0; i < rio_table_hdr->num_rio_dev; i++, ptr += rio_detail_size) 130 for (i = 0; i < rio_table_hdr->num_rio_dev; i++, ptr += rio_detail_size)
126 rio_devs[i] = (struct rio_detail *)ptr; 131 rio_devs[i] = (struct rio_detail *)ptr;
127 132
128 return 1; 133 return 1;
@@ -140,9 +145,9 @@ void __init setup_summit(void)
140 145
141 rio_table_hdr = NULL; 146 rio_table_hdr = NULL;
142 offset = 0x180; 147 offset = 0x180;
143 while (offset){ 148 while (offset) {
144 /* The block id is stored in the 2nd word */ 149 /* The block id is stored in the 2nd word */
145 if (*((unsigned short *)(ptr + offset + 2)) == 0x4752){ 150 if (*((unsigned short *)(ptr + offset + 2)) == 0x4752) {
146 /* set the pointer past the offset & block id */ 151 /* set the pointer past the offset & block id */
147 rio_table_hdr = (struct rio_table_hdr *)(ptr + offset + 4); 152 rio_table_hdr = (struct rio_table_hdr *)(ptr + offset + 4);
148 break; 153 break;
@@ -150,8 +155,8 @@ void __init setup_summit(void)
150 /* The next offset is stored in the 1st word. 0 means no more */ 155 /* The next offset is stored in the 1st word. 0 means no more */
151 offset = *((unsigned short *)(ptr + offset)); 156 offset = *((unsigned short *)(ptr + offset));
152 } 157 }
153 if (!rio_table_hdr){ 158 if (!rio_table_hdr) {
154 printk(KERN_ERR "%s: Unable to locate Rio Grande Table in EBDA - bailing!\n", __FUNCTION__); 159 printk(KERN_ERR "%s: Unable to locate Rio Grande Table in EBDA - bailing!\n", __func__);
155 return; 160 return;
156 } 161 }
157 162
@@ -161,8 +166,8 @@ void __init setup_summit(void)
161 /* The first Winnipeg we're looking for has an index of 0 */ 166 /* The first Winnipeg we're looking for has an index of 0 */
162 next_wpeg = 0; 167 next_wpeg = 0;
163 do { 168 do {
164 for(i = 0; i < rio_table_hdr->num_rio_dev; i++){ 169 for (i = 0; i < rio_table_hdr->num_rio_dev; i++) {
165 if (is_WPEG(rio_devs[i]) && rio_devs[i]->WP_index == next_wpeg){ 170 if (is_WPEG(rio_devs[i]) && rio_devs[i]->WP_index == next_wpeg) {
166 /* It's the Winnipeg we're looking for! */ 171 /* It's the Winnipeg we're looking for! */
167 next_bus = setup_pci_node_map_for_wpeg(i, next_bus); 172 next_bus = setup_pci_node_map_for_wpeg(i, next_bus);
168 next_wpeg++; 173 next_wpeg++;
diff --git a/arch/x86/kernel/syscall_64.c b/arch/x86/kernel/syscall_64.c
index 9d498c2f8eea..170d43c17487 100644
--- a/arch/x86/kernel/syscall_64.c
+++ b/arch/x86/kernel/syscall_64.c
@@ -1,4 +1,4 @@
1/* System call table for x86-64. */ 1/* System call table for x86-64. */
2 2
3#include <linux/linkage.h> 3#include <linux/linkage.h>
4#include <linux/sys.h> 4#include <linux/sys.h>
@@ -7,20 +7,23 @@
7 7
8#define __NO_STUBS 8#define __NO_STUBS
9 9
10#define __SYSCALL(nr, sym) extern asmlinkage void sym(void) ; 10#define __SYSCALL(nr, sym) extern asmlinkage void sym(void) ;
11#undef _ASM_X86_64_UNISTD_H_ 11#undef _ASM_X86_64_UNISTD_H_
12#include <asm/unistd_64.h> 12#include <asm/unistd_64.h>
13 13
14#undef __SYSCALL 14#undef __SYSCALL
15#define __SYSCALL(nr, sym) [ nr ] = sym, 15#define __SYSCALL(nr, sym) [nr] = sym,
16#undef _ASM_X86_64_UNISTD_H_ 16#undef _ASM_X86_64_UNISTD_H_
17 17
18typedef void (*sys_call_ptr_t)(void); 18typedef void (*sys_call_ptr_t)(void);
19 19
20extern void sys_ni_syscall(void); 20extern void sys_ni_syscall(void);
21 21
22const sys_call_ptr_t sys_call_table[__NR_syscall_max+1] = { 22const sys_call_ptr_t sys_call_table[__NR_syscall_max+1] = {
23 /* Smells like a like a compiler bug -- it doesn't work when the & below is removed. */ 23 /*
24 *Smells like a like a compiler bug -- it doesn't work
25 *when the & below is removed.
26 */
24 [0 ... __NR_syscall_max] = &sys_ni_syscall, 27 [0 ... __NR_syscall_max] = &sys_ni_syscall,
25#include <asm/unistd_64.h> 28#include <asm/unistd_64.h>
26}; 29};
diff --git a/arch/x86/kernel/test_nx.c b/arch/x86/kernel/test_nx.c
index 10b8a6f69f84..787a5e499dd1 100644
--- a/arch/x86/kernel/test_nx.c
+++ b/arch/x86/kernel/test_nx.c
@@ -11,6 +11,8 @@
11 */ 11 */
12#include <linux/module.h> 12#include <linux/module.h>
13#include <linux/sort.h> 13#include <linux/sort.h>
14#include <linux/slab.h>
15
14#include <asm/uaccess.h> 16#include <asm/uaccess.h>
15#include <asm/asm.h> 17#include <asm/asm.h>
16 18
diff --git a/arch/x86/kernel/tlb_32.c b/arch/x86/kernel/tlb_32.c
new file mode 100644
index 000000000000..9bb2363851af
--- /dev/null
+++ b/arch/x86/kernel/tlb_32.c
@@ -0,0 +1,243 @@
1#include <linux/spinlock.h>
2#include <linux/cpu.h>
3#include <linux/interrupt.h>
4
5#include <asm/tlbflush.h>
6
7DEFINE_PER_CPU(struct tlb_state, cpu_tlbstate)
8 ____cacheline_aligned = { &init_mm, 0, };
9
10/* must come after the send_IPI functions above for inlining */
11#include <mach_ipi.h>
12
13/*
14 * Smarter SMP flushing macros.
15 * c/o Linus Torvalds.
16 *
17 * These mean you can really definitely utterly forget about
18 * writing to user space from interrupts. (Its not allowed anyway).
19 *
20 * Optimizations Manfred Spraul <manfred@colorfullife.com>
21 */
22
23static cpumask_t flush_cpumask;
24static struct mm_struct *flush_mm;
25static unsigned long flush_va;
26static DEFINE_SPINLOCK(tlbstate_lock);
27
28/*
29 * We cannot call mmdrop() because we are in interrupt context,
30 * instead update mm->cpu_vm_mask.
31 *
32 * We need to reload %cr3 since the page tables may be going
33 * away from under us..
34 */
35void leave_mm(int cpu)
36{
37 if (per_cpu(cpu_tlbstate, cpu).state == TLBSTATE_OK)
38 BUG();
39 cpu_clear(cpu, per_cpu(cpu_tlbstate, cpu).active_mm->cpu_vm_mask);
40 load_cr3(swapper_pg_dir);
41}
42EXPORT_SYMBOL_GPL(leave_mm);
43
44/*
45 *
46 * The flush IPI assumes that a thread switch happens in this order:
47 * [cpu0: the cpu that switches]
48 * 1) switch_mm() either 1a) or 1b)
49 * 1a) thread switch to a different mm
50 * 1a1) cpu_clear(cpu, old_mm->cpu_vm_mask);
51 * Stop ipi delivery for the old mm. This is not synchronized with
52 * the other cpus, but smp_invalidate_interrupt ignore flush ipis
53 * for the wrong mm, and in the worst case we perform a superfluous
54 * tlb flush.
55 * 1a2) set cpu_tlbstate to TLBSTATE_OK
56 * Now the smp_invalidate_interrupt won't call leave_mm if cpu0
57 * was in lazy tlb mode.
58 * 1a3) update cpu_tlbstate[].active_mm
59 * Now cpu0 accepts tlb flushes for the new mm.
60 * 1a4) cpu_set(cpu, new_mm->cpu_vm_mask);
61 * Now the other cpus will send tlb flush ipis.
62 * 1a4) change cr3.
63 * 1b) thread switch without mm change
64 * cpu_tlbstate[].active_mm is correct, cpu0 already handles
65 * flush ipis.
66 * 1b1) set cpu_tlbstate to TLBSTATE_OK
67 * 1b2) test_and_set the cpu bit in cpu_vm_mask.
68 * Atomically set the bit [other cpus will start sending flush ipis],
69 * and test the bit.
70 * 1b3) if the bit was 0: leave_mm was called, flush the tlb.
71 * 2) switch %%esp, ie current
72 *
73 * The interrupt must handle 2 special cases:
74 * - cr3 is changed before %%esp, ie. it cannot use current->{active_,}mm.
75 * - the cpu performs speculative tlb reads, i.e. even if the cpu only
76 * runs in kernel space, the cpu could load tlb entries for user space
77 * pages.
78 *
79 * The good news is that cpu_tlbstate is local to each cpu, no
80 * write/read ordering problems.
81 */
82
83/*
84 * TLB flush IPI:
85 *
86 * 1) Flush the tlb entries if the cpu uses the mm that's being flushed.
87 * 2) Leave the mm if we are in the lazy tlb mode.
88 */
89
90void smp_invalidate_interrupt(struct pt_regs *regs)
91{
92 unsigned long cpu;
93
94 cpu = get_cpu();
95
96 if (!cpu_isset(cpu, flush_cpumask))
97 goto out;
98 /*
99 * This was a BUG() but until someone can quote me the
100 * line from the intel manual that guarantees an IPI to
101 * multiple CPUs is retried _only_ on the erroring CPUs
102 * its staying as a return
103 *
104 * BUG();
105 */
106
107 if (flush_mm == per_cpu(cpu_tlbstate, cpu).active_mm) {
108 if (per_cpu(cpu_tlbstate, cpu).state == TLBSTATE_OK) {
109 if (flush_va == TLB_FLUSH_ALL)
110 local_flush_tlb();
111 else
112 __flush_tlb_one(flush_va);
113 } else
114 leave_mm(cpu);
115 }
116 ack_APIC_irq();
117 smp_mb__before_clear_bit();
118 cpu_clear(cpu, flush_cpumask);
119 smp_mb__after_clear_bit();
120out:
121 put_cpu_no_resched();
122 __get_cpu_var(irq_stat).irq_tlb_count++;
123}
124
125void native_flush_tlb_others(const cpumask_t *cpumaskp, struct mm_struct *mm,
126 unsigned long va)
127{
128 cpumask_t cpumask = *cpumaskp;
129
130 /*
131 * A couple of (to be removed) sanity checks:
132 *
133 * - current CPU must not be in mask
134 * - mask must exist :)
135 */
136 BUG_ON(cpus_empty(cpumask));
137 BUG_ON(cpu_isset(smp_processor_id(), cpumask));
138 BUG_ON(!mm);
139
140#ifdef CONFIG_HOTPLUG_CPU
141 /* If a CPU which we ran on has gone down, OK. */
142 cpus_and(cpumask, cpumask, cpu_online_map);
143 if (unlikely(cpus_empty(cpumask)))
144 return;
145#endif
146
147 /*
148 * i'm not happy about this global shared spinlock in the
149 * MM hot path, but we'll see how contended it is.
150 * AK: x86-64 has a faster method that could be ported.
151 */
152 spin_lock(&tlbstate_lock);
153
154 flush_mm = mm;
155 flush_va = va;
156 cpus_or(flush_cpumask, cpumask, flush_cpumask);
157 /*
158 * We have to send the IPI only to
159 * CPUs affected.
160 */
161 send_IPI_mask(cpumask, INVALIDATE_TLB_VECTOR);
162
163 while (!cpus_empty(flush_cpumask))
164 /* nothing. lockup detection does not belong here */
165 cpu_relax();
166
167 flush_mm = NULL;
168 flush_va = 0;
169 spin_unlock(&tlbstate_lock);
170}
171
172void flush_tlb_current_task(void)
173{
174 struct mm_struct *mm = current->mm;
175 cpumask_t cpu_mask;
176
177 preempt_disable();
178 cpu_mask = mm->cpu_vm_mask;
179 cpu_clear(smp_processor_id(), cpu_mask);
180
181 local_flush_tlb();
182 if (!cpus_empty(cpu_mask))
183 flush_tlb_others(cpu_mask, mm, TLB_FLUSH_ALL);
184 preempt_enable();
185}
186
187void flush_tlb_mm(struct mm_struct *mm)
188{
189 cpumask_t cpu_mask;
190
191 preempt_disable();
192 cpu_mask = mm->cpu_vm_mask;
193 cpu_clear(smp_processor_id(), cpu_mask);
194
195 if (current->active_mm == mm) {
196 if (current->mm)
197 local_flush_tlb();
198 else
199 leave_mm(smp_processor_id());
200 }
201 if (!cpus_empty(cpu_mask))
202 flush_tlb_others(cpu_mask, mm, TLB_FLUSH_ALL);
203
204 preempt_enable();
205}
206
207void flush_tlb_page(struct vm_area_struct *vma, unsigned long va)
208{
209 struct mm_struct *mm = vma->vm_mm;
210 cpumask_t cpu_mask;
211
212 preempt_disable();
213 cpu_mask = mm->cpu_vm_mask;
214 cpu_clear(smp_processor_id(), cpu_mask);
215
216 if (current->active_mm == mm) {
217 if (current->mm)
218 __flush_tlb_one(va);
219 else
220 leave_mm(smp_processor_id());
221 }
222
223 if (!cpus_empty(cpu_mask))
224 flush_tlb_others(cpu_mask, mm, va);
225
226 preempt_enable();
227}
228EXPORT_SYMBOL(flush_tlb_page);
229
230static void do_flush_tlb_all(void *info)
231{
232 unsigned long cpu = smp_processor_id();
233
234 __flush_tlb_all();
235 if (per_cpu(cpu_tlbstate, cpu).state == TLBSTATE_LAZY)
236 leave_mm(cpu);
237}
238
239void flush_tlb_all(void)
240{
241 on_each_cpu(do_flush_tlb_all, NULL, 1, 1);
242}
243
diff --git a/arch/x86/kernel/smp_64.c b/arch/x86/kernel/tlb_64.c
index 2fd74b06db67..1558e513757e 100644
--- a/arch/x86/kernel/smp_64.c
+++ b/arch/x86/kernel/tlb_64.c
@@ -1,14 +1,3 @@
1/*
2 * Intel SMP support routines.
3 *
4 * (c) 1995 Alan Cox, Building #3 <alan@redhat.com>
5 * (c) 1998-99, 2000 Ingo Molnar <mingo@redhat.com>
6 * (c) 2002,2003 Andi Kleen, SuSE Labs.
7 *
8 * This code is released under the GNU General Public License version 2 or
9 * later.
10 */
11
12#include <linux/init.h> 1#include <linux/init.h>
13 2
14#include <linux/mm.h> 3#include <linux/mm.h>
@@ -22,12 +11,12 @@
22#include <asm/mtrr.h> 11#include <asm/mtrr.h>
23#include <asm/pgalloc.h> 12#include <asm/pgalloc.h>
24#include <asm/tlbflush.h> 13#include <asm/tlbflush.h>
25#include <asm/mach_apic.h>
26#include <asm/mmu_context.h> 14#include <asm/mmu_context.h>
27#include <asm/proto.h> 15#include <asm/proto.h>
28#include <asm/apicdef.h> 16#include <asm/apicdef.h>
29#include <asm/idle.h> 17#include <asm/idle.h>
30 18
19#include <mach_ipi.h>
31/* 20/*
32 * Smarter SMP flushing macros. 21 * Smarter SMP flushing macros.
33 * c/o Linus Torvalds. 22 * c/o Linus Torvalds.
@@ -228,7 +217,7 @@ void flush_tlb_current_task(void)
228 preempt_enable(); 217 preempt_enable();
229} 218}
230 219
231void flush_tlb_mm (struct mm_struct * mm) 220void flush_tlb_mm(struct mm_struct *mm)
232{ 221{
233 cpumask_t cpu_mask; 222 cpumask_t cpu_mask;
234 223
@@ -248,7 +237,7 @@ void flush_tlb_mm (struct mm_struct * mm)
248 preempt_enable(); 237 preempt_enable();
249} 238}
250 239
251void flush_tlb_page(struct vm_area_struct * vma, unsigned long va) 240void flush_tlb_page(struct vm_area_struct *vma, unsigned long va)
252{ 241{
253 struct mm_struct *mm = vma->vm_mm; 242 struct mm_struct *mm = vma->vm_mm;
254 cpumask_t cpu_mask; 243 cpumask_t cpu_mask;
@@ -258,7 +247,7 @@ void flush_tlb_page(struct vm_area_struct * vma, unsigned long va)
258 cpu_clear(smp_processor_id(), cpu_mask); 247 cpu_clear(smp_processor_id(), cpu_mask);
259 248
260 if (current->active_mm == mm) { 249 if (current->active_mm == mm) {
261 if(current->mm) 250 if (current->mm)
262 __flush_tlb_one(va); 251 __flush_tlb_one(va);
263 else 252 else
264 leave_mm(smp_processor_id()); 253 leave_mm(smp_processor_id());
@@ -270,7 +259,7 @@ void flush_tlb_page(struct vm_area_struct * vma, unsigned long va)
270 preempt_enable(); 259 preempt_enable();
271} 260}
272 261
273static void do_flush_tlb_all(void* info) 262static void do_flush_tlb_all(void *info)
274{ 263{
275 unsigned long cpu = smp_processor_id(); 264 unsigned long cpu = smp_processor_id();
276 265
@@ -283,248 +272,3 @@ void flush_tlb_all(void)
283{ 272{
284 on_each_cpu(do_flush_tlb_all, NULL, 1, 1); 273 on_each_cpu(do_flush_tlb_all, NULL, 1, 1);
285} 274}
286
287/*
288 * this function sends a 'reschedule' IPI to another CPU.
289 * it goes straight through and wastes no time serializing
290 * anything. Worst case is that we lose a reschedule ...
291 */
292
293void smp_send_reschedule(int cpu)
294{
295 send_IPI_mask(cpumask_of_cpu(cpu), RESCHEDULE_VECTOR);
296}
297
298/*
299 * Structure and data for smp_call_function(). This is designed to minimise
300 * static memory requirements. It also looks cleaner.
301 */
302static DEFINE_SPINLOCK(call_lock);
303
304struct call_data_struct {
305 void (*func) (void *info);
306 void *info;
307 atomic_t started;
308 atomic_t finished;
309 int wait;
310};
311
312static struct call_data_struct * call_data;
313
314void lock_ipi_call_lock(void)
315{
316 spin_lock_irq(&call_lock);
317}
318
319void unlock_ipi_call_lock(void)
320{
321 spin_unlock_irq(&call_lock);
322}
323
324/*
325 * this function sends a 'generic call function' IPI to all other CPU
326 * of the system defined in the mask.
327 */
328static int __smp_call_function_mask(cpumask_t mask,
329 void (*func)(void *), void *info,
330 int wait)
331{
332 struct call_data_struct data;
333 cpumask_t allbutself;
334 int cpus;
335
336 allbutself = cpu_online_map;
337 cpu_clear(smp_processor_id(), allbutself);
338
339 cpus_and(mask, mask, allbutself);
340 cpus = cpus_weight(mask);
341
342 if (!cpus)
343 return 0;
344
345 data.func = func;
346 data.info = info;
347 atomic_set(&data.started, 0);
348 data.wait = wait;
349 if (wait)
350 atomic_set(&data.finished, 0);
351
352 call_data = &data;
353 wmb();
354
355 /* Send a message to other CPUs */
356 if (cpus_equal(mask, allbutself))
357 send_IPI_allbutself(CALL_FUNCTION_VECTOR);
358 else
359 send_IPI_mask(mask, CALL_FUNCTION_VECTOR);
360
361 /* Wait for response */
362 while (atomic_read(&data.started) != cpus)
363 cpu_relax();
364
365 if (!wait)
366 return 0;
367
368 while (atomic_read(&data.finished) != cpus)
369 cpu_relax();
370
371 return 0;
372}
373/**
374 * smp_call_function_mask(): Run a function on a set of other CPUs.
375 * @mask: The set of cpus to run on. Must not include the current cpu.
376 * @func: The function to run. This must be fast and non-blocking.
377 * @info: An arbitrary pointer to pass to the function.
378 * @wait: If true, wait (atomically) until function has completed on other CPUs.
379 *
380 * Returns 0 on success, else a negative status code.
381 *
382 * If @wait is true, then returns once @func has returned; otherwise
383 * it returns just before the target cpu calls @func.
384 *
385 * You must not call this function with disabled interrupts or from a
386 * hardware interrupt handler or from a bottom half handler.
387 */
388int smp_call_function_mask(cpumask_t mask,
389 void (*func)(void *), void *info,
390 int wait)
391{
392 int ret;
393
394 /* Can deadlock when called with interrupts disabled */
395 WARN_ON(irqs_disabled());
396
397 spin_lock(&call_lock);
398 ret = __smp_call_function_mask(mask, func, info, wait);
399 spin_unlock(&call_lock);
400 return ret;
401}
402EXPORT_SYMBOL(smp_call_function_mask);
403
404/*
405 * smp_call_function_single - Run a function on a specific CPU
406 * @func: The function to run. This must be fast and non-blocking.
407 * @info: An arbitrary pointer to pass to the function.
408 * @nonatomic: Currently unused.
409 * @wait: If true, wait until function has completed on other CPUs.
410 *
411 * Retrurns 0 on success, else a negative status code.
412 *
413 * Does not return until the remote CPU is nearly ready to execute <func>
414 * or is or has executed.
415 */
416
417int smp_call_function_single (int cpu, void (*func) (void *info), void *info,
418 int nonatomic, int wait)
419{
420 /* prevent preemption and reschedule on another processor */
421 int ret, me = get_cpu();
422
423 /* Can deadlock when called with interrupts disabled */
424 WARN_ON(irqs_disabled());
425
426 if (cpu == me) {
427 local_irq_disable();
428 func(info);
429 local_irq_enable();
430 put_cpu();
431 return 0;
432 }
433
434 ret = smp_call_function_mask(cpumask_of_cpu(cpu), func, info, wait);
435
436 put_cpu();
437 return ret;
438}
439EXPORT_SYMBOL(smp_call_function_single);
440
441/*
442 * smp_call_function - run a function on all other CPUs.
443 * @func: The function to run. This must be fast and non-blocking.
444 * @info: An arbitrary pointer to pass to the function.
445 * @nonatomic: currently unused.
446 * @wait: If true, wait (atomically) until function has completed on other
447 * CPUs.
448 *
449 * Returns 0 on success, else a negative status code. Does not return until
450 * remote CPUs are nearly ready to execute func or are or have executed.
451 *
452 * You must not call this function with disabled interrupts or from a
453 * hardware interrupt handler or from a bottom half handler.
454 * Actually there are a few legal cases, like panic.
455 */
456int smp_call_function (void (*func) (void *info), void *info, int nonatomic,
457 int wait)
458{
459 return smp_call_function_mask(cpu_online_map, func, info, wait);
460}
461EXPORT_SYMBOL(smp_call_function);
462
463static void stop_this_cpu(void *dummy)
464{
465 local_irq_disable();
466 /*
467 * Remove this CPU:
468 */
469 cpu_clear(smp_processor_id(), cpu_online_map);
470 disable_local_APIC();
471 for (;;)
472 halt();
473}
474
475void smp_send_stop(void)
476{
477 int nolock;
478 unsigned long flags;
479
480 if (reboot_force)
481 return;
482
483 /* Don't deadlock on the call lock in panic */
484 nolock = !spin_trylock(&call_lock);
485 local_irq_save(flags);
486 __smp_call_function_mask(cpu_online_map, stop_this_cpu, NULL, 0);
487 if (!nolock)
488 spin_unlock(&call_lock);
489 disable_local_APIC();
490 local_irq_restore(flags);
491}
492
493/*
494 * Reschedule call back. Nothing to do,
495 * all the work is done automatically when
496 * we return from the interrupt.
497 */
498asmlinkage void smp_reschedule_interrupt(void)
499{
500 ack_APIC_irq();
501 add_pda(irq_resched_count, 1);
502}
503
504asmlinkage void smp_call_function_interrupt(void)
505{
506 void (*func) (void *info) = call_data->func;
507 void *info = call_data->info;
508 int wait = call_data->wait;
509
510 ack_APIC_irq();
511 /*
512 * Notify initiating CPU that I've grabbed the data and am
513 * about to execute the function
514 */
515 mb();
516 atomic_inc(&call_data->started);
517 /*
518 * At this point the info structure may be out of scope unless wait==1
519 */
520 exit_idle();
521 irq_enter();
522 (*func)(info);
523 add_pda(irq_call_count, 1);
524 irq_exit();
525 if (wait) {
526 mb();
527 atomic_inc(&call_data->finished);
528 }
529}
530
diff --git a/arch/x86/kernel/trampoline.c b/arch/x86/kernel/trampoline.c
new file mode 100644
index 000000000000..abbf199adebb
--- /dev/null
+++ b/arch/x86/kernel/trampoline.c
@@ -0,0 +1,18 @@
1#include <linux/io.h>
2
3#include <asm/trampoline.h>
4
5/* ready for x86_64, no harm for x86, since it will overwrite after alloc */
6unsigned char *trampoline_base = __va(TRAMPOLINE_BASE);
7
8/*
9 * Currently trivial. Write the real->protected mode
10 * bootstrap into the page concerned. The caller
11 * has made sure it's suitably aligned.
12 */
13unsigned long setup_trampoline(void)
14{
15 memcpy(trampoline_base, trampoline_data,
16 trampoline_end - trampoline_data);
17 return virt_to_phys(trampoline_base);
18}
diff --git a/arch/x86/kernel/trampoline_64.S b/arch/x86/kernel/trampoline_64.S
index 4aedd0bcee4c..894293c598db 100644
--- a/arch/x86/kernel/trampoline_64.S
+++ b/arch/x86/kernel/trampoline_64.S
@@ -30,12 +30,7 @@
30#include <asm/msr.h> 30#include <asm/msr.h>
31#include <asm/segment.h> 31#include <asm/segment.h>
32 32
33/* We can free up trampoline after bootup if cpu hotplug is not supported. */
34#ifndef CONFIG_HOTPLUG_CPU
35.section .init.data, "aw", @progbits
36#else
37.section .rodata, "a", @progbits 33.section .rodata, "a", @progbits
38#endif
39 34
40.code16 35.code16
41 36
diff --git a/arch/x86/kernel/traps_32.c b/arch/x86/kernel/traps_32.c
index c5421f30d678..65791ca2824a 100644
--- a/arch/x86/kernel/traps_32.c
+++ b/arch/x86/kernel/traps_32.c
@@ -9,26 +9,28 @@
9 * 'Traps.c' handles hardware traps and faults after we have saved some 9 * 'Traps.c' handles hardware traps and faults after we have saved some
10 * state in 'asm.s'. 10 * state in 'asm.s'.
11 */ 11 */
12#include <linux/sched.h> 12#include <linux/interrupt.h>
13#include <linux/kallsyms.h>
14#include <linux/spinlock.h>
15#include <linux/highmem.h>
16#include <linux/kprobes.h>
17#include <linux/uaccess.h>
18#include <linux/utsname.h>
19#include <linux/kdebug.h>
13#include <linux/kernel.h> 20#include <linux/kernel.h>
21#include <linux/module.h>
22#include <linux/ptrace.h>
14#include <linux/string.h> 23#include <linux/string.h>
24#include <linux/unwind.h>
25#include <linux/delay.h>
15#include <linux/errno.h> 26#include <linux/errno.h>
27#include <linux/kexec.h>
28#include <linux/sched.h>
16#include <linux/timer.h> 29#include <linux/timer.h>
17#include <linux/mm.h>
18#include <linux/init.h> 30#include <linux/init.h>
19#include <linux/delay.h>
20#include <linux/spinlock.h>
21#include <linux/interrupt.h>
22#include <linux/highmem.h>
23#include <linux/kallsyms.h>
24#include <linux/ptrace.h>
25#include <linux/utsname.h>
26#include <linux/kprobes.h>
27#include <linux/kexec.h>
28#include <linux/unwind.h>
29#include <linux/uaccess.h>
30#include <linux/nmi.h>
31#include <linux/bug.h> 31#include <linux/bug.h>
32#include <linux/nmi.h>
33#include <linux/mm.h>
32 34
33#ifdef CONFIG_EISA 35#ifdef CONFIG_EISA
34#include <linux/ioport.h> 36#include <linux/ioport.h>
@@ -43,21 +45,18 @@
43#include <linux/edac.h> 45#include <linux/edac.h>
44#endif 46#endif
45 47
48#include <asm/arch_hooks.h>
49#include <asm/stacktrace.h>
46#include <asm/processor.h> 50#include <asm/processor.h>
47#include <asm/system.h>
48#include <asm/io.h>
49#include <asm/atomic.h>
50#include <asm/debugreg.h> 51#include <asm/debugreg.h>
52#include <asm/atomic.h>
53#include <asm/system.h>
54#include <asm/unwind.h>
51#include <asm/desc.h> 55#include <asm/desc.h>
52#include <asm/i387.h> 56#include <asm/i387.h>
53#include <asm/nmi.h> 57#include <asm/nmi.h>
54#include <asm/unwind.h>
55#include <asm/smp.h> 58#include <asm/smp.h>
56#include <asm/arch_hooks.h> 59#include <asm/io.h>
57#include <linux/kdebug.h>
58#include <asm/stacktrace.h>
59
60#include <linux/module.h>
61 60
62#include "mach_traps.h" 61#include "mach_traps.h"
63 62
@@ -69,7 +68,7 @@ EXPORT_SYMBOL_GPL(used_vectors);
69asmlinkage int system_call(void); 68asmlinkage int system_call(void);
70 69
71/* Do we ignore FPU interrupts ? */ 70/* Do we ignore FPU interrupts ? */
72char ignore_fpu_irq = 0; 71char ignore_fpu_irq;
73 72
74/* 73/*
75 * The IDT has to be page-aligned to simplify the Pentium 74 * The IDT has to be page-aligned to simplify the Pentium
@@ -105,12 +104,13 @@ static unsigned int code_bytes = 64;
105void printk_address(unsigned long address, int reliable) 104void printk_address(unsigned long address, int reliable)
106{ 105{
107#ifdef CONFIG_KALLSYMS 106#ifdef CONFIG_KALLSYMS
108 unsigned long offset = 0, symsize; 107 char namebuf[KSYM_NAME_LEN];
108 unsigned long offset = 0;
109 unsigned long symsize;
109 const char *symname; 110 const char *symname;
110 char *modname;
111 char *delim = ":";
112 char namebuf[128];
113 char reliab[4] = ""; 111 char reliab[4] = "";
112 char *delim = ":";
113 char *modname;
114 114
115 symname = kallsyms_lookup(address, &symsize, &offset, 115 symname = kallsyms_lookup(address, &symsize, &offset,
116 &modname, namebuf); 116 &modname, namebuf);
@@ -138,13 +138,14 @@ static inline int valid_stack_ptr(struct thread_info *tinfo, void *p, unsigned s
138 138
139/* The form of the top of the frame on the stack */ 139/* The form of the top of the frame on the stack */
140struct stack_frame { 140struct stack_frame {
141 struct stack_frame *next_frame; 141 struct stack_frame *next_frame;
142 unsigned long return_address; 142 unsigned long return_address;
143}; 143};
144 144
145static inline unsigned long print_context_stack(struct thread_info *tinfo, 145static inline unsigned long
146 unsigned long *stack, unsigned long bp, 146print_context_stack(struct thread_info *tinfo,
147 const struct stacktrace_ops *ops, void *data) 147 unsigned long *stack, unsigned long bp,
148 const struct stacktrace_ops *ops, void *data)
148{ 149{
149 struct stack_frame *frame = (struct stack_frame *)bp; 150 struct stack_frame *frame = (struct stack_frame *)bp;
150 151
@@ -166,7 +167,7 @@ static inline unsigned long print_context_stack(struct thread_info *tinfo,
166 return bp; 167 return bp;
167} 168}
168 169
169#define MSG(msg) ops->warning(data, msg) 170#define MSG(msg) ops->warning(data, msg)
170 171
171void dump_trace(struct task_struct *task, struct pt_regs *regs, 172void dump_trace(struct task_struct *task, struct pt_regs *regs,
172 unsigned long *stack, unsigned long bp, 173 unsigned long *stack, unsigned long bp,
@@ -177,6 +178,7 @@ void dump_trace(struct task_struct *task, struct pt_regs *regs,
177 178
178 if (!stack) { 179 if (!stack) {
179 unsigned long dummy; 180 unsigned long dummy;
181
180 stack = &dummy; 182 stack = &dummy;
181 if (task != current) 183 if (task != current)
182 stack = (unsigned long *)task->thread.sp; 184 stack = (unsigned long *)task->thread.sp;
@@ -186,7 +188,7 @@ void dump_trace(struct task_struct *task, struct pt_regs *regs,
186 if (!bp) { 188 if (!bp) {
187 if (task == current) { 189 if (task == current) {
188 /* Grab bp right from our regs */ 190 /* Grab bp right from our regs */
189 asm ("movl %%ebp, %0" : "=r" (bp) : ); 191 asm("movl %%ebp, %0" : "=r" (bp) :);
190 } else { 192 } else {
191 /* bp is the last reg pushed by switch_to */ 193 /* bp is the last reg pushed by switch_to */
192 bp = *(unsigned long *) task->thread.sp; 194 bp = *(unsigned long *) task->thread.sp;
@@ -196,15 +198,18 @@ void dump_trace(struct task_struct *task, struct pt_regs *regs,
196 198
197 while (1) { 199 while (1) {
198 struct thread_info *context; 200 struct thread_info *context;
201
199 context = (struct thread_info *) 202 context = (struct thread_info *)
200 ((unsigned long)stack & (~(THREAD_SIZE - 1))); 203 ((unsigned long)stack & (~(THREAD_SIZE - 1)));
201 bp = print_context_stack(context, stack, bp, ops, data); 204 bp = print_context_stack(context, stack, bp, ops, data);
202 /* Should be after the line below, but somewhere 205 /*
203 in early boot context comes out corrupted and we 206 * Should be after the line below, but somewhere
204 can't reference it -AK */ 207 * in early boot context comes out corrupted and we
208 * can't reference it:
209 */
205 if (ops->stack(data, "IRQ") < 0) 210 if (ops->stack(data, "IRQ") < 0)
206 break; 211 break;
207 stack = (unsigned long*)context->previous_esp; 212 stack = (unsigned long *)context->previous_esp;
208 if (!stack) 213 if (!stack)
209 break; 214 break;
210 touch_nmi_watchdog(); 215 touch_nmi_watchdog();
@@ -243,15 +248,15 @@ static void print_trace_address(void *data, unsigned long addr, int reliable)
243} 248}
244 249
245static const struct stacktrace_ops print_trace_ops = { 250static const struct stacktrace_ops print_trace_ops = {
246 .warning = print_trace_warning, 251 .warning = print_trace_warning,
247 .warning_symbol = print_trace_warning_symbol, 252 .warning_symbol = print_trace_warning_symbol,
248 .stack = print_trace_stack, 253 .stack = print_trace_stack,
249 .address = print_trace_address, 254 .address = print_trace_address,
250}; 255};
251 256
252static void 257static void
253show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs, 258show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs,
254 unsigned long *stack, unsigned long bp, char *log_lvl) 259 unsigned long *stack, unsigned long bp, char *log_lvl)
255{ 260{
256 dump_trace(task, regs, stack, bp, &print_trace_ops, log_lvl); 261 dump_trace(task, regs, stack, bp, &print_trace_ops, log_lvl);
257 printk("%s =======================\n", log_lvl); 262 printk("%s =======================\n", log_lvl);
@@ -263,21 +268,22 @@ void show_trace(struct task_struct *task, struct pt_regs *regs,
263 show_trace_log_lvl(task, regs, stack, bp, ""); 268 show_trace_log_lvl(task, regs, stack, bp, "");
264} 269}
265 270
266static void show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs, 271static void
267 unsigned long *sp, unsigned long bp, char *log_lvl) 272show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs,
273 unsigned long *sp, unsigned long bp, char *log_lvl)
268{ 274{
269 unsigned long *stack; 275 unsigned long *stack;
270 int i; 276 int i;
271 277
272 if (sp == NULL) { 278 if (sp == NULL) {
273 if (task) 279 if (task)
274 sp = (unsigned long*)task->thread.sp; 280 sp = (unsigned long *)task->thread.sp;
275 else 281 else
276 sp = (unsigned long *)&sp; 282 sp = (unsigned long *)&sp;
277 } 283 }
278 284
279 stack = sp; 285 stack = sp;
280 for(i = 0; i < kstack_depth_to_print; i++) { 286 for (i = 0; i < kstack_depth_to_print; i++) {
281 if (kstack_end(stack)) 287 if (kstack_end(stack))
282 break; 288 break;
283 if (i && ((i % 8) == 0)) 289 if (i && ((i % 8) == 0))
@@ -285,6 +291,7 @@ static void show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs,
285 printk("%08lx ", *stack++); 291 printk("%08lx ", *stack++);
286 } 292 }
287 printk("\n%sCall Trace:\n", log_lvl); 293 printk("\n%sCall Trace:\n", log_lvl);
294
288 show_trace_log_lvl(task, regs, sp, bp, log_lvl); 295 show_trace_log_lvl(task, regs, sp, bp, log_lvl);
289} 296}
290 297
@@ -299,8 +306,8 @@ void show_stack(struct task_struct *task, unsigned long *sp)
299 */ 306 */
300void dump_stack(void) 307void dump_stack(void)
301{ 308{
302 unsigned long stack;
303 unsigned long bp = 0; 309 unsigned long bp = 0;
310 unsigned long stack;
304 311
305#ifdef CONFIG_FRAME_POINTER 312#ifdef CONFIG_FRAME_POINTER
306 if (!bp) 313 if (!bp)
@@ -312,6 +319,7 @@ void dump_stack(void)
312 init_utsname()->release, 319 init_utsname()->release,
313 (int)strcspn(init_utsname()->version, " "), 320 (int)strcspn(init_utsname()->version, " "),
314 init_utsname()->version); 321 init_utsname()->version);
322
315 show_trace(current, NULL, &stack, bp); 323 show_trace(current, NULL, &stack, bp);
316} 324}
317 325
@@ -323,6 +331,7 @@ void show_registers(struct pt_regs *regs)
323 331
324 print_modules(); 332 print_modules();
325 __show_registers(regs, 0); 333 __show_registers(regs, 0);
334
326 printk(KERN_EMERG "Process %.*s (pid: %d, ti=%p task=%p task.ti=%p)", 335 printk(KERN_EMERG "Process %.*s (pid: %d, ti=%p task=%p task.ti=%p)",
327 TASK_COMM_LEN, current->comm, task_pid_nr(current), 336 TASK_COMM_LEN, current->comm, task_pid_nr(current),
328 current_thread_info(), current, task_thread_info(current)); 337 current_thread_info(), current, task_thread_info(current));
@@ -331,10 +340,10 @@ void show_registers(struct pt_regs *regs)
331 * time of the fault.. 340 * time of the fault..
332 */ 341 */
333 if (!user_mode_vm(regs)) { 342 if (!user_mode_vm(regs)) {
334 u8 *ip;
335 unsigned int code_prologue = code_bytes * 43 / 64; 343 unsigned int code_prologue = code_bytes * 43 / 64;
336 unsigned int code_len = code_bytes; 344 unsigned int code_len = code_bytes;
337 unsigned char c; 345 unsigned char c;
346 u8 *ip;
338 347
339 printk("\n" KERN_EMERG "Stack: "); 348 printk("\n" KERN_EMERG "Stack: ");
340 show_stack_log_lvl(NULL, regs, &regs->sp, 0, KERN_EMERG); 349 show_stack_log_lvl(NULL, regs, &regs->sp, 0, KERN_EMERG);
@@ -361,7 +370,7 @@ void show_registers(struct pt_regs *regs)
361 } 370 }
362 } 371 }
363 printk("\n"); 372 printk("\n");
364} 373}
365 374
366int is_valid_bugaddr(unsigned long ip) 375int is_valid_bugaddr(unsigned long ip)
367{ 376{
@@ -377,10 +386,10 @@ int is_valid_bugaddr(unsigned long ip)
377 386
378static int die_counter; 387static int die_counter;
379 388
380int __kprobes __die(const char * str, struct pt_regs * regs, long err) 389int __kprobes __die(const char *str, struct pt_regs *regs, long err)
381{ 390{
382 unsigned long sp;
383 unsigned short ss; 391 unsigned short ss;
392 unsigned long sp;
384 393
385 printk(KERN_EMERG "%s: %04lx [#%d] ", str, err & 0xffff, ++die_counter); 394 printk(KERN_EMERG "%s: %04lx [#%d] ", str, err & 0xffff, ++die_counter);
386#ifdef CONFIG_PREEMPT 395#ifdef CONFIG_PREEMPT
@@ -395,8 +404,8 @@ int __kprobes __die(const char * str, struct pt_regs * regs, long err)
395 printk("\n"); 404 printk("\n");
396 405
397 if (notify_die(DIE_OOPS, str, regs, err, 406 if (notify_die(DIE_OOPS, str, regs, err,
398 current->thread.trap_no, SIGSEGV) != 407 current->thread.trap_no, SIGSEGV) != NOTIFY_STOP) {
399 NOTIFY_STOP) { 408
400 show_registers(regs); 409 show_registers(regs);
401 /* Executive summary in case the oops scrolled away */ 410 /* Executive summary in case the oops scrolled away */
402 sp = (unsigned long) (&regs->sp); 411 sp = (unsigned long) (&regs->sp);
@@ -408,17 +417,18 @@ int __kprobes __die(const char * str, struct pt_regs * regs, long err)
408 printk(KERN_EMERG "EIP: [<%08lx>] ", regs->ip); 417 printk(KERN_EMERG "EIP: [<%08lx>] ", regs->ip);
409 print_symbol("%s", regs->ip); 418 print_symbol("%s", regs->ip);
410 printk(" SS:ESP %04x:%08lx\n", ss, sp); 419 printk(" SS:ESP %04x:%08lx\n", ss, sp);
420
411 return 0; 421 return 0;
412 } else {
413 return 1;
414 } 422 }
423
424 return 1;
415} 425}
416 426
417/* 427/*
418 * This is gone through when something in the kernel has done something bad and 428 * This is gone through when something in the kernel has done something bad
419 * is about to be terminated. 429 * and is about to be terminated:
420 */ 430 */
421void die(const char * str, struct pt_regs * regs, long err) 431void die(const char *str, struct pt_regs *regs, long err)
422{ 432{
423 static struct { 433 static struct {
424 raw_spinlock_t lock; 434 raw_spinlock_t lock;
@@ -440,8 +450,9 @@ void die(const char * str, struct pt_regs * regs, long err)
440 die.lock_owner = smp_processor_id(); 450 die.lock_owner = smp_processor_id();
441 die.lock_owner_depth = 0; 451 die.lock_owner_depth = 0;
442 bust_spinlocks(1); 452 bust_spinlocks(1);
443 } else 453 } else {
444 raw_local_irq_save(flags); 454 raw_local_irq_save(flags);
455 }
445 456
446 if (++die.lock_owner_depth < 3) { 457 if (++die.lock_owner_depth < 3) {
447 report_bug(regs->ip, regs); 458 report_bug(regs->ip, regs);
@@ -474,19 +485,20 @@ void die(const char * str, struct pt_regs * regs, long err)
474 do_exit(SIGSEGV); 485 do_exit(SIGSEGV);
475} 486}
476 487
477static inline void die_if_kernel(const char * str, struct pt_regs * regs, long err) 488static inline void
489die_if_kernel(const char *str, struct pt_regs *regs, long err)
478{ 490{
479 if (!user_mode_vm(regs)) 491 if (!user_mode_vm(regs))
480 die(str, regs, err); 492 die(str, regs, err);
481} 493}
482 494
483static void __kprobes do_trap(int trapnr, int signr, char *str, int vm86, 495static void __kprobes
484 struct pt_regs * regs, long error_code, 496do_trap(int trapnr, int signr, char *str, int vm86, struct pt_regs *regs,
485 siginfo_t *info) 497 long error_code, siginfo_t *info)
486{ 498{
487 struct task_struct *tsk = current; 499 struct task_struct *tsk = current;
488 500
489 if (regs->flags & VM_MASK) { 501 if (regs->flags & X86_VM_MASK) {
490 if (vm86) 502 if (vm86)
491 goto vm86_trap; 503 goto vm86_trap;
492 goto trap_signal; 504 goto trap_signal;
@@ -495,111 +507,112 @@ static void __kprobes do_trap(int trapnr, int signr, char *str, int vm86,
495 if (!user_mode(regs)) 507 if (!user_mode(regs))
496 goto kernel_trap; 508 goto kernel_trap;
497 509
498 trap_signal: { 510trap_signal:
499 /* 511 /*
500 * We want error_code and trap_no set for userspace faults and 512 * We want error_code and trap_no set for userspace faults and
501 * kernelspace faults which result in die(), but not 513 * kernelspace faults which result in die(), but not
502 * kernelspace faults which are fixed up. die() gives the 514 * kernelspace faults which are fixed up. die() gives the
503 * process no chance to handle the signal and notice the 515 * process no chance to handle the signal and notice the
504 * kernel fault information, so that won't result in polluting 516 * kernel fault information, so that won't result in polluting
505 * the information about previously queued, but not yet 517 * the information about previously queued, but not yet
506 * delivered, faults. See also do_general_protection below. 518 * delivered, faults. See also do_general_protection below.
507 */ 519 */
508 tsk->thread.error_code = error_code; 520 tsk->thread.error_code = error_code;
509 tsk->thread.trap_no = trapnr; 521 tsk->thread.trap_no = trapnr;
510 522
511 if (info) 523 if (info)
512 force_sig_info(signr, info, tsk); 524 force_sig_info(signr, info, tsk);
513 else 525 else
514 force_sig(signr, tsk); 526 force_sig(signr, tsk);
515 return; 527 return;
516 }
517 528
518 kernel_trap: { 529kernel_trap:
519 if (!fixup_exception(regs)) { 530 if (!fixup_exception(regs)) {
520 tsk->thread.error_code = error_code; 531 tsk->thread.error_code = error_code;
521 tsk->thread.trap_no = trapnr; 532 tsk->thread.trap_no = trapnr;
522 die(str, regs, error_code); 533 die(str, regs, error_code);
523 }
524 return;
525 } 534 }
535 return;
526 536
527 vm86_trap: { 537vm86_trap:
528 int ret = handle_vm86_trap((struct kernel_vm86_regs *) regs, error_code, trapnr); 538 if (handle_vm86_trap((struct kernel_vm86_regs *) regs,
529 if (ret) goto trap_signal; 539 error_code, trapnr))
530 return; 540 goto trap_signal;
531 } 541 return;
532} 542}
533 543
534#define DO_ERROR(trapnr, signr, str, name) \ 544#define DO_ERROR(trapnr, signr, str, name) \
535void do_##name(struct pt_regs * regs, long error_code) \ 545void do_##name(struct pt_regs *regs, long error_code) \
536{ \ 546{ \
537 if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \ 547 if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \
538 == NOTIFY_STOP) \ 548 == NOTIFY_STOP) \
539 return; \ 549 return; \
540 do_trap(trapnr, signr, str, 0, regs, error_code, NULL); \ 550 do_trap(trapnr, signr, str, 0, regs, error_code, NULL); \
541} 551}
542 552
543#define DO_ERROR_INFO(trapnr, signr, str, name, sicode, siaddr, irq) \ 553#define DO_ERROR_INFO(trapnr, signr, str, name, sicode, siaddr, irq) \
544void do_##name(struct pt_regs * regs, long error_code) \ 554void do_##name(struct pt_regs *regs, long error_code) \
545{ \ 555{ \
546 siginfo_t info; \ 556 siginfo_t info; \
547 if (irq) \ 557 if (irq) \
548 local_irq_enable(); \ 558 local_irq_enable(); \
549 info.si_signo = signr; \ 559 info.si_signo = signr; \
550 info.si_errno = 0; \ 560 info.si_errno = 0; \
551 info.si_code = sicode; \ 561 info.si_code = sicode; \
552 info.si_addr = (void __user *)siaddr; \ 562 info.si_addr = (void __user *)siaddr; \
553 if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \ 563 if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \
554 == NOTIFY_STOP) \ 564 == NOTIFY_STOP) \
555 return; \ 565 return; \
556 do_trap(trapnr, signr, str, 0, regs, error_code, &info); \ 566 do_trap(trapnr, signr, str, 0, regs, error_code, &info); \
557} 567}
558 568
559#define DO_VM86_ERROR(trapnr, signr, str, name) \ 569#define DO_VM86_ERROR(trapnr, signr, str, name) \
560void do_##name(struct pt_regs * regs, long error_code) \ 570void do_##name(struct pt_regs *regs, long error_code) \
561{ \ 571{ \
562 if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \ 572 if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \
563 == NOTIFY_STOP) \ 573 == NOTIFY_STOP) \
564 return; \ 574 return; \
565 do_trap(trapnr, signr, str, 1, regs, error_code, NULL); \ 575 do_trap(trapnr, signr, str, 1, regs, error_code, NULL); \
566} 576}
567 577
568#define DO_VM86_ERROR_INFO(trapnr, signr, str, name, sicode, siaddr) \ 578#define DO_VM86_ERROR_INFO(trapnr, signr, str, name, sicode, siaddr) \
569void do_##name(struct pt_regs * regs, long error_code) \ 579void do_##name(struct pt_regs *regs, long error_code) \
570{ \ 580{ \
571 siginfo_t info; \ 581 siginfo_t info; \
572 info.si_signo = signr; \ 582 info.si_signo = signr; \
573 info.si_errno = 0; \ 583 info.si_errno = 0; \
574 info.si_code = sicode; \ 584 info.si_code = sicode; \
575 info.si_addr = (void __user *)siaddr; \ 585 info.si_addr = (void __user *)siaddr; \
576 trace_hardirqs_fixup(); \ 586 trace_hardirqs_fixup(); \
577 if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \ 587 if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \
578 == NOTIFY_STOP) \ 588 == NOTIFY_STOP) \
579 return; \ 589 return; \
580 do_trap(trapnr, signr, str, 1, regs, error_code, &info); \ 590 do_trap(trapnr, signr, str, 1, regs, error_code, &info); \
581} 591}
582 592
583DO_VM86_ERROR_INFO( 0, SIGFPE, "divide error", divide_error, FPE_INTDIV, regs->ip) 593DO_VM86_ERROR_INFO(0, SIGFPE, "divide error", divide_error, FPE_INTDIV, regs->ip)
584#ifndef CONFIG_KPROBES 594#ifndef CONFIG_KPROBES
585DO_VM86_ERROR( 3, SIGTRAP, "int3", int3) 595DO_VM86_ERROR(3, SIGTRAP, "int3", int3)
586#endif 596#endif
587DO_VM86_ERROR( 4, SIGSEGV, "overflow", overflow) 597DO_VM86_ERROR(4, SIGSEGV, "overflow", overflow)
588DO_VM86_ERROR( 5, SIGSEGV, "bounds", bounds) 598DO_VM86_ERROR(5, SIGSEGV, "bounds", bounds)
589DO_ERROR_INFO( 6, SIGILL, "invalid opcode", invalid_op, ILL_ILLOPN, regs->ip, 0) 599DO_ERROR_INFO(6, SIGILL, "invalid opcode", invalid_op, ILL_ILLOPN, regs->ip, 0)
590DO_ERROR( 9, SIGFPE, "coprocessor segment overrun", coprocessor_segment_overrun) 600DO_ERROR(9, SIGFPE, "coprocessor segment overrun", coprocessor_segment_overrun)
591DO_ERROR(10, SIGSEGV, "invalid TSS", invalid_TSS) 601DO_ERROR(10, SIGSEGV, "invalid TSS", invalid_TSS)
592DO_ERROR(11, SIGBUS, "segment not present", segment_not_present) 602DO_ERROR(11, SIGBUS, "segment not present", segment_not_present)
593DO_ERROR(12, SIGBUS, "stack segment", stack_segment) 603DO_ERROR(12, SIGBUS, "stack segment", stack_segment)
594DO_ERROR_INFO(17, SIGBUS, "alignment check", alignment_check, BUS_ADRALN, 0, 0) 604DO_ERROR_INFO(17, SIGBUS, "alignment check", alignment_check, BUS_ADRALN, 0, 0)
595DO_ERROR_INFO(32, SIGSEGV, "iret exception", iret_error, ILL_BADSTK, 0, 1) 605DO_ERROR_INFO(32, SIGSEGV, "iret exception", iret_error, ILL_BADSTK, 0, 1)
596 606
597void __kprobes do_general_protection(struct pt_regs * regs, 607void __kprobes do_general_protection(struct pt_regs *regs, long error_code)
598 long error_code)
599{ 608{
600 int cpu = get_cpu(); 609 struct thread_struct *thread;
601 struct tss_struct *tss = &per_cpu(init_tss, cpu); 610 struct tss_struct *tss;
602 struct thread_struct *thread = &current->thread; 611 int cpu;
612
613 cpu = get_cpu();
614 tss = &per_cpu(init_tss, cpu);
615 thread = &current->thread;
603 616
604 /* 617 /*
605 * Perform the lazy TSS's I/O bitmap copy. If the TSS has an 618 * Perform the lazy TSS's I/O bitmap copy. If the TSS has an
@@ -616,19 +629,21 @@ void __kprobes do_general_protection(struct pt_regs * regs,
616 * If the previously set map was extending to higher ports 629 * If the previously set map was extending to higher ports
617 * than the current one, pad extra space with 0xff (no access). 630 * than the current one, pad extra space with 0xff (no access).
618 */ 631 */
619 if (thread->io_bitmap_max < tss->io_bitmap_max) 632 if (thread->io_bitmap_max < tss->io_bitmap_max) {
620 memset((char *) tss->io_bitmap + 633 memset((char *) tss->io_bitmap +
621 thread->io_bitmap_max, 0xff, 634 thread->io_bitmap_max, 0xff,
622 tss->io_bitmap_max - thread->io_bitmap_max); 635 tss->io_bitmap_max - thread->io_bitmap_max);
636 }
623 tss->io_bitmap_max = thread->io_bitmap_max; 637 tss->io_bitmap_max = thread->io_bitmap_max;
624 tss->x86_tss.io_bitmap_base = IO_BITMAP_OFFSET; 638 tss->x86_tss.io_bitmap_base = IO_BITMAP_OFFSET;
625 tss->io_bitmap_owner = thread; 639 tss->io_bitmap_owner = thread;
626 put_cpu(); 640 put_cpu();
641
627 return; 642 return;
628 } 643 }
629 put_cpu(); 644 put_cpu();
630 645
631 if (regs->flags & VM_MASK) 646 if (regs->flags & X86_VM_MASK)
632 goto gp_in_vm86; 647 goto gp_in_vm86;
633 648
634 if (!user_mode(regs)) 649 if (!user_mode(regs))
@@ -636,6 +651,7 @@ void __kprobes do_general_protection(struct pt_regs * regs,
636 651
637 current->thread.error_code = error_code; 652 current->thread.error_code = error_code;
638 current->thread.trap_no = 13; 653 current->thread.trap_no = 13;
654
639 if (show_unhandled_signals && unhandled_signal(current, SIGSEGV) && 655 if (show_unhandled_signals && unhandled_signal(current, SIGSEGV) &&
640 printk_ratelimit()) { 656 printk_ratelimit()) {
641 printk(KERN_INFO 657 printk(KERN_INFO
@@ -666,21 +682,24 @@ gp_in_kernel:
666} 682}
667 683
668static __kprobes void 684static __kprobes void
669mem_parity_error(unsigned char reason, struct pt_regs * regs) 685mem_parity_error(unsigned char reason, struct pt_regs *regs)
670{ 686{
671 printk(KERN_EMERG "Uhhuh. NMI received for unknown reason %02x on " 687 printk(KERN_EMERG
672 "CPU %d.\n", reason, smp_processor_id()); 688 "Uhhuh. NMI received for unknown reason %02x on CPU %d.\n",
673 printk(KERN_EMERG "You have some hardware problem, likely on the PCI bus.\n"); 689 reason, smp_processor_id());
690
691 printk(KERN_EMERG
692 "You have some hardware problem, likely on the PCI bus.\n");
674 693
675#if defined(CONFIG_EDAC) 694#if defined(CONFIG_EDAC)
676 if(edac_handler_set()) { 695 if (edac_handler_set()) {
677 edac_atomic_assert_error(); 696 edac_atomic_assert_error();
678 return; 697 return;
679 } 698 }
680#endif 699#endif
681 700
682 if (panic_on_unrecovered_nmi) 701 if (panic_on_unrecovered_nmi)
683 panic("NMI: Not continuing"); 702 panic("NMI: Not continuing");
684 703
685 printk(KERN_EMERG "Dazed and confused, but trying to continue\n"); 704 printk(KERN_EMERG "Dazed and confused, but trying to continue\n");
686 705
@@ -689,7 +708,7 @@ mem_parity_error(unsigned char reason, struct pt_regs * regs)
689} 708}
690 709
691static __kprobes void 710static __kprobes void
692io_check_error(unsigned char reason, struct pt_regs * regs) 711io_check_error(unsigned char reason, struct pt_regs *regs)
693{ 712{
694 unsigned long i; 713 unsigned long i;
695 714
@@ -699,30 +718,37 @@ io_check_error(unsigned char reason, struct pt_regs * regs)
699 /* Re-enable the IOCK line, wait for a few seconds */ 718 /* Re-enable the IOCK line, wait for a few seconds */
700 reason = (reason & 0xf) | 8; 719 reason = (reason & 0xf) | 8;
701 outb(reason, 0x61); 720 outb(reason, 0x61);
721
702 i = 2000; 722 i = 2000;
703 while (--i) udelay(1000); 723 while (--i)
724 udelay(1000);
725
704 reason &= ~8; 726 reason &= ~8;
705 outb(reason, 0x61); 727 outb(reason, 0x61);
706} 728}
707 729
708static __kprobes void 730static __kprobes void
709unknown_nmi_error(unsigned char reason, struct pt_regs * regs) 731unknown_nmi_error(unsigned char reason, struct pt_regs *regs)
710{ 732{
711 if (notify_die(DIE_NMIUNKNOWN, "nmi", regs, reason, 2, SIGINT) == NOTIFY_STOP) 733 if (notify_die(DIE_NMIUNKNOWN, "nmi", regs, reason, 2, SIGINT) == NOTIFY_STOP)
712 return; 734 return;
713#ifdef CONFIG_MCA 735#ifdef CONFIG_MCA
714 /* Might actually be able to figure out what the guilty party 736 /*
715 * is. */ 737 * Might actually be able to figure out what the guilty party
716 if( MCA_bus ) { 738 * is:
739 */
740 if (MCA_bus) {
717 mca_handle_nmi(); 741 mca_handle_nmi();
718 return; 742 return;
719 } 743 }
720#endif 744#endif
721 printk(KERN_EMERG "Uhhuh. NMI received for unknown reason %02x on " 745 printk(KERN_EMERG
722 "CPU %d.\n", reason, smp_processor_id()); 746 "Uhhuh. NMI received for unknown reason %02x on CPU %d.\n",
747 reason, smp_processor_id());
748
723 printk(KERN_EMERG "Do you have a strange power saving mode enabled?\n"); 749 printk(KERN_EMERG "Do you have a strange power saving mode enabled?\n");
724 if (panic_on_unrecovered_nmi) 750 if (panic_on_unrecovered_nmi)
725 panic("NMI: Not continuing"); 751 panic("NMI: Not continuing");
726 752
727 printk(KERN_EMERG "Dazed and confused, but trying to continue\n"); 753 printk(KERN_EMERG "Dazed and confused, but trying to continue\n");
728} 754}
@@ -731,14 +757,13 @@ static DEFINE_SPINLOCK(nmi_print_lock);
731 757
732void __kprobes die_nmi(struct pt_regs *regs, const char *msg) 758void __kprobes die_nmi(struct pt_regs *regs, const char *msg)
733{ 759{
734 if (notify_die(DIE_NMIWATCHDOG, msg, regs, 0, 2, SIGINT) == 760 if (notify_die(DIE_NMIWATCHDOG, msg, regs, 0, 2, SIGINT) == NOTIFY_STOP)
735 NOTIFY_STOP)
736 return; 761 return;
737 762
738 spin_lock(&nmi_print_lock); 763 spin_lock(&nmi_print_lock);
739 /* 764 /*
740 * We are in trouble anyway, lets at least try 765 * We are in trouble anyway, lets at least try
741 * to get a message out. 766 * to get a message out:
742 */ 767 */
743 bust_spinlocks(1); 768 bust_spinlocks(1);
744 printk(KERN_EMERG "%s", msg); 769 printk(KERN_EMERG "%s", msg);
@@ -749,9 +774,10 @@ void __kprobes die_nmi(struct pt_regs *regs, const char *msg)
749 spin_unlock(&nmi_print_lock); 774 spin_unlock(&nmi_print_lock);
750 bust_spinlocks(0); 775 bust_spinlocks(0);
751 776
752 /* If we are in kernel we are probably nested up pretty bad 777 /*
753 * and might aswell get out now while we still can. 778 * If we are in kernel we are probably nested up pretty bad
754 */ 779 * and might aswell get out now while we still can:
780 */
755 if (!user_mode_vm(regs)) { 781 if (!user_mode_vm(regs)) {
756 current->thread.trap_no = 2; 782 current->thread.trap_no = 2;
757 crash_kexec(regs); 783 crash_kexec(regs);
@@ -760,14 +786,14 @@ void __kprobes die_nmi(struct pt_regs *regs, const char *msg)
760 do_exit(SIGSEGV); 786 do_exit(SIGSEGV);
761} 787}
762 788
763static __kprobes void default_do_nmi(struct pt_regs * regs) 789static __kprobes void default_do_nmi(struct pt_regs *regs)
764{ 790{
765 unsigned char reason = 0; 791 unsigned char reason = 0;
766 792
767 /* Only the BSP gets external NMIs from the system. */ 793 /* Only the BSP gets external NMIs from the system: */
768 if (!smp_processor_id()) 794 if (!smp_processor_id())
769 reason = get_nmi_reason(); 795 reason = get_nmi_reason();
770 796
771 if (!(reason & 0xc0)) { 797 if (!(reason & 0xc0)) {
772 if (notify_die(DIE_NMI_IPI, "nmi_ipi", regs, reason, 2, SIGINT) 798 if (notify_die(DIE_NMI_IPI, "nmi_ipi", regs, reason, 2, SIGINT)
773 == NOTIFY_STOP) 799 == NOTIFY_STOP)
@@ -780,8 +806,10 @@ static __kprobes void default_do_nmi(struct pt_regs * regs)
780 if (nmi_watchdog_tick(regs, reason)) 806 if (nmi_watchdog_tick(regs, reason))
781 return; 807 return;
782 if (!do_nmi_callback(regs, smp_processor_id())) 808 if (!do_nmi_callback(regs, smp_processor_id()))
783#endif
784 unknown_nmi_error(reason, regs); 809 unknown_nmi_error(reason, regs);
810#else
811 unknown_nmi_error(reason, regs);
812#endif
785 813
786 return; 814 return;
787 } 815 }
@@ -793,14 +821,14 @@ static __kprobes void default_do_nmi(struct pt_regs * regs)
793 io_check_error(reason, regs); 821 io_check_error(reason, regs);
794 /* 822 /*
795 * Reassert NMI in case it became active meanwhile 823 * Reassert NMI in case it became active meanwhile
796 * as it's edge-triggered. 824 * as it's edge-triggered:
797 */ 825 */
798 reassert_nmi(); 826 reassert_nmi();
799} 827}
800 828
801static int ignore_nmis; 829static int ignore_nmis;
802 830
803__kprobes void do_nmi(struct pt_regs * regs, long error_code) 831__kprobes void do_nmi(struct pt_regs *regs, long error_code)
804{ 832{
805 int cpu; 833 int cpu;
806 834
@@ -836,9 +864,12 @@ void __kprobes do_int3(struct pt_regs *regs, long error_code)
836 if (notify_die(DIE_INT3, "int3", regs, error_code, 3, SIGTRAP) 864 if (notify_die(DIE_INT3, "int3", regs, error_code, 3, SIGTRAP)
837 == NOTIFY_STOP) 865 == NOTIFY_STOP)
838 return; 866 return;
839 /* This is an interrupt gate, because kprobes wants interrupts 867 /*
840 disabled. Normal trap handlers don't. */ 868 * This is an interrupt gate, because kprobes wants interrupts
869 * disabled. Normal trap handlers don't.
870 */
841 restore_interrupts(regs); 871 restore_interrupts(regs);
872
842 do_trap(3, SIGTRAP, "int3", 1, regs, error_code, NULL); 873 do_trap(3, SIGTRAP, "int3", 1, regs, error_code, NULL);
843} 874}
844#endif 875#endif
@@ -853,7 +884,7 @@ void __kprobes do_int3(struct pt_regs *regs, long error_code)
853 * from user space. Such code must not hold kernel locks (since it 884 * from user space. Such code must not hold kernel locks (since it
854 * can equally take a page fault), therefore it is safe to call 885 * can equally take a page fault), therefore it is safe to call
855 * force_sig_info even though that claims and releases locks. 886 * force_sig_info even though that claims and releases locks.
856 * 887 *
857 * Code in ./signal.c ensures that the debug control register 888 * Code in ./signal.c ensures that the debug control register
858 * is restored before we deliver any signal, and therefore that 889 * is restored before we deliver any signal, and therefore that
859 * user code runs with the correct debug control register even though 890 * user code runs with the correct debug control register even though
@@ -865,10 +896,10 @@ void __kprobes do_int3(struct pt_regs *regs, long error_code)
865 * find every occurrence of the TF bit that could be saved away even 896 * find every occurrence of the TF bit that could be saved away even
866 * by user code) 897 * by user code)
867 */ 898 */
868void __kprobes do_debug(struct pt_regs * regs, long error_code) 899void __kprobes do_debug(struct pt_regs *regs, long error_code)
869{ 900{
870 unsigned int condition;
871 struct task_struct *tsk = current; 901 struct task_struct *tsk = current;
902 unsigned int condition;
872 903
873 trace_hardirqs_fixup(); 904 trace_hardirqs_fixup();
874 905
@@ -893,7 +924,7 @@ void __kprobes do_debug(struct pt_regs * regs, long error_code)
893 goto clear_dr7; 924 goto clear_dr7;
894 } 925 }
895 926
896 if (regs->flags & VM_MASK) 927 if (regs->flags & X86_VM_MASK)
897 goto debug_vm86; 928 goto debug_vm86;
898 929
899 /* Save debug status register where ptrace can see it */ 930 /* Save debug status register where ptrace can see it */
@@ -916,7 +947,8 @@ void __kprobes do_debug(struct pt_regs * regs, long error_code)
916 /* Ok, finally something we can handle */ 947 /* Ok, finally something we can handle */
917 send_sigtrap(tsk, regs, error_code); 948 send_sigtrap(tsk, regs, error_code);
918 949
919 /* Disable additional traps. They'll be re-enabled when 950 /*
951 * Disable additional traps. They'll be re-enabled when
920 * the signal is delivered. 952 * the signal is delivered.
921 */ 953 */
922clear_dr7: 954clear_dr7:
@@ -929,7 +961,7 @@ debug_vm86:
929 961
930clear_TF_reenable: 962clear_TF_reenable:
931 set_tsk_thread_flag(tsk, TIF_SINGLESTEP); 963 set_tsk_thread_flag(tsk, TIF_SINGLESTEP);
932 regs->flags &= ~TF_MASK; 964 regs->flags &= ~X86_EFLAGS_TF;
933 return; 965 return;
934} 966}
935 967
@@ -940,9 +972,10 @@ clear_TF_reenable:
940 */ 972 */
941void math_error(void __user *ip) 973void math_error(void __user *ip)
942{ 974{
943 struct task_struct * task; 975 struct task_struct *task;
976 unsigned short cwd;
977 unsigned short swd;
944 siginfo_t info; 978 siginfo_t info;
945 unsigned short cwd, swd;
946 979
947 /* 980 /*
948 * Save the info for the exception handler and clear the error. 981 * Save the info for the exception handler and clear the error.
@@ -968,36 +1001,36 @@ void math_error(void __user *ip)
968 cwd = get_fpu_cwd(task); 1001 cwd = get_fpu_cwd(task);
969 swd = get_fpu_swd(task); 1002 swd = get_fpu_swd(task);
970 switch (swd & ~cwd & 0x3f) { 1003 switch (swd & ~cwd & 0x3f) {
971 case 0x000: /* No unmasked exception */ 1004 case 0x000: /* No unmasked exception */
972 return; 1005 return;
973 default: /* Multiple exceptions */ 1006 default: /* Multiple exceptions */
974 break; 1007 break;
975 case 0x001: /* Invalid Op */ 1008 case 0x001: /* Invalid Op */
976 /* 1009 /*
977 * swd & 0x240 == 0x040: Stack Underflow 1010 * swd & 0x240 == 0x040: Stack Underflow
978 * swd & 0x240 == 0x240: Stack Overflow 1011 * swd & 0x240 == 0x240: Stack Overflow
979 * User must clear the SF bit (0x40) if set 1012 * User must clear the SF bit (0x40) if set
980 */ 1013 */
981 info.si_code = FPE_FLTINV; 1014 info.si_code = FPE_FLTINV;
982 break; 1015 break;
983 case 0x002: /* Denormalize */ 1016 case 0x002: /* Denormalize */
984 case 0x010: /* Underflow */ 1017 case 0x010: /* Underflow */
985 info.si_code = FPE_FLTUND; 1018 info.si_code = FPE_FLTUND;
986 break; 1019 break;
987 case 0x004: /* Zero Divide */ 1020 case 0x004: /* Zero Divide */
988 info.si_code = FPE_FLTDIV; 1021 info.si_code = FPE_FLTDIV;
989 break; 1022 break;
990 case 0x008: /* Overflow */ 1023 case 0x008: /* Overflow */
991 info.si_code = FPE_FLTOVF; 1024 info.si_code = FPE_FLTOVF;
992 break; 1025 break;
993 case 0x020: /* Precision */ 1026 case 0x020: /* Precision */
994 info.si_code = FPE_FLTRES; 1027 info.si_code = FPE_FLTRES;
995 break; 1028 break;
996 } 1029 }
997 force_sig_info(SIGFPE, &info, task); 1030 force_sig_info(SIGFPE, &info, task);
998} 1031}
999 1032
1000void do_coprocessor_error(struct pt_regs * regs, long error_code) 1033void do_coprocessor_error(struct pt_regs *regs, long error_code)
1001{ 1034{
1002 ignore_fpu_irq = 1; 1035 ignore_fpu_irq = 1;
1003 math_error((void __user *)regs->ip); 1036 math_error((void __user *)regs->ip);
@@ -1005,9 +1038,9 @@ void do_coprocessor_error(struct pt_regs * regs, long error_code)
1005 1038
1006static void simd_math_error(void __user *ip) 1039static void simd_math_error(void __user *ip)
1007{ 1040{
1008 struct task_struct * task; 1041 struct task_struct *task;
1009 siginfo_t info;
1010 unsigned short mxcsr; 1042 unsigned short mxcsr;
1043 siginfo_t info;
1011 1044
1012 /* 1045 /*
1013 * Save the info for the exception handler and clear the error. 1046 * Save the info for the exception handler and clear the error.
@@ -1028,82 +1061,80 @@ static void simd_math_error(void __user *ip)
1028 */ 1061 */
1029 mxcsr = get_fpu_mxcsr(task); 1062 mxcsr = get_fpu_mxcsr(task);
1030 switch (~((mxcsr & 0x1f80) >> 7) & (mxcsr & 0x3f)) { 1063 switch (~((mxcsr & 0x1f80) >> 7) & (mxcsr & 0x3f)) {
1031 case 0x000: 1064 case 0x000:
1032 default: 1065 default:
1033 break; 1066 break;
1034 case 0x001: /* Invalid Op */ 1067 case 0x001: /* Invalid Op */
1035 info.si_code = FPE_FLTINV; 1068 info.si_code = FPE_FLTINV;
1036 break; 1069 break;
1037 case 0x002: /* Denormalize */ 1070 case 0x002: /* Denormalize */
1038 case 0x010: /* Underflow */ 1071 case 0x010: /* Underflow */
1039 info.si_code = FPE_FLTUND; 1072 info.si_code = FPE_FLTUND;
1040 break; 1073 break;
1041 case 0x004: /* Zero Divide */ 1074 case 0x004: /* Zero Divide */
1042 info.si_code = FPE_FLTDIV; 1075 info.si_code = FPE_FLTDIV;
1043 break; 1076 break;
1044 case 0x008: /* Overflow */ 1077 case 0x008: /* Overflow */
1045 info.si_code = FPE_FLTOVF; 1078 info.si_code = FPE_FLTOVF;
1046 break; 1079 break;
1047 case 0x020: /* Precision */ 1080 case 0x020: /* Precision */
1048 info.si_code = FPE_FLTRES; 1081 info.si_code = FPE_FLTRES;
1049 break; 1082 break;
1050 } 1083 }
1051 force_sig_info(SIGFPE, &info, task); 1084 force_sig_info(SIGFPE, &info, task);
1052} 1085}
1053 1086
1054void do_simd_coprocessor_error(struct pt_regs * regs, 1087void do_simd_coprocessor_error(struct pt_regs *regs, long error_code)
1055 long error_code)
1056{ 1088{
1057 if (cpu_has_xmm) { 1089 if (cpu_has_xmm) {
1058 /* Handle SIMD FPU exceptions on PIII+ processors. */ 1090 /* Handle SIMD FPU exceptions on PIII+ processors. */
1059 ignore_fpu_irq = 1; 1091 ignore_fpu_irq = 1;
1060 simd_math_error((void __user *)regs->ip); 1092 simd_math_error((void __user *)regs->ip);
1061 } else { 1093 return;
1062 /* 1094 }
1063 * Handle strange cache flush from user space exception 1095 /*
1064 * in all other cases. This is undocumented behaviour. 1096 * Handle strange cache flush from user space exception
1065 */ 1097 * in all other cases. This is undocumented behaviour.
1066 if (regs->flags & VM_MASK) { 1098 */
1067 handle_vm86_fault((struct kernel_vm86_regs *)regs, 1099 if (regs->flags & X86_VM_MASK) {
1068 error_code); 1100 handle_vm86_fault((struct kernel_vm86_regs *)regs, error_code);
1069 return; 1101 return;
1070 }
1071 current->thread.trap_no = 19;
1072 current->thread.error_code = error_code;
1073 die_if_kernel("cache flush denied", regs, error_code);
1074 force_sig(SIGSEGV, current);
1075 } 1102 }
1103 current->thread.trap_no = 19;
1104 current->thread.error_code = error_code;
1105 die_if_kernel("cache flush denied", regs, error_code);
1106 force_sig(SIGSEGV, current);
1076} 1107}
1077 1108
1078void do_spurious_interrupt_bug(struct pt_regs * regs, 1109void do_spurious_interrupt_bug(struct pt_regs *regs, long error_code)
1079 long error_code)
1080{ 1110{
1081#if 0 1111#if 0
1082 /* No need to warn about this any longer. */ 1112 /* No need to warn about this any longer. */
1083 printk("Ignoring P6 Local APIC Spurious Interrupt Bug...\n"); 1113 printk(KERN_INFO "Ignoring P6 Local APIC Spurious Interrupt Bug...\n");
1084#endif 1114#endif
1085} 1115}
1086 1116
1087unsigned long patch_espfix_desc(unsigned long uesp, 1117unsigned long patch_espfix_desc(unsigned long uesp, unsigned long kesp)
1088 unsigned long kesp)
1089{ 1118{
1090 struct desc_struct *gdt = __get_cpu_var(gdt_page).gdt; 1119 struct desc_struct *gdt = __get_cpu_var(gdt_page).gdt;
1091 unsigned long base = (kesp - uesp) & -THREAD_SIZE; 1120 unsigned long base = (kesp - uesp) & -THREAD_SIZE;
1092 unsigned long new_kesp = kesp - base; 1121 unsigned long new_kesp = kesp - base;
1093 unsigned long lim_pages = (new_kesp | (THREAD_SIZE - 1)) >> PAGE_SHIFT; 1122 unsigned long lim_pages = (new_kesp | (THREAD_SIZE - 1)) >> PAGE_SHIFT;
1094 __u64 desc = *(__u64 *)&gdt[GDT_ENTRY_ESPFIX_SS]; 1123 __u64 desc = *(__u64 *)&gdt[GDT_ENTRY_ESPFIX_SS];
1124
1095 /* Set up base for espfix segment */ 1125 /* Set up base for espfix segment */
1096 desc &= 0x00f0ff0000000000ULL; 1126 desc &= 0x00f0ff0000000000ULL;
1097 desc |= ((((__u64)base) << 16) & 0x000000ffffff0000ULL) | 1127 desc |= ((((__u64)base) << 16) & 0x000000ffffff0000ULL) |
1098 ((((__u64)base) << 32) & 0xff00000000000000ULL) | 1128 ((((__u64)base) << 32) & 0xff00000000000000ULL) |
1099 ((((__u64)lim_pages) << 32) & 0x000f000000000000ULL) | 1129 ((((__u64)lim_pages) << 32) & 0x000f000000000000ULL) |
1100 (lim_pages & 0xffff); 1130 (lim_pages & 0xffff);
1101 *(__u64 *)&gdt[GDT_ENTRY_ESPFIX_SS] = desc; 1131 *(__u64 *)&gdt[GDT_ENTRY_ESPFIX_SS] = desc;
1132
1102 return new_kesp; 1133 return new_kesp;
1103} 1134}
1104 1135
1105/* 1136/*
1106 * 'math_state_restore()' saves the current math information in the 1137 * 'math_state_restore()' saves the current math information in the
1107 * old math state array, and gets the new ones from the current task 1138 * old math state array, and gets the new ones from the current task
1108 * 1139 *
1109 * Careful.. There are problems with IBM-designed IRQ13 behaviour. 1140 * Careful.. There are problems with IBM-designed IRQ13 behaviour.
@@ -1117,7 +1148,7 @@ asmlinkage void math_state_restore(void)
1117 struct thread_info *thread = current_thread_info(); 1148 struct thread_info *thread = current_thread_info();
1118 struct task_struct *tsk = thread->task; 1149 struct task_struct *tsk = thread->task;
1119 1150
1120 clts(); /* Allow maths ops (or we recurse) */ 1151 clts(); /* Allow maths ops (or we recurse) */
1121 if (!tsk_used_math(tsk)) 1152 if (!tsk_used_math(tsk))
1122 init_fpu(tsk); 1153 init_fpu(tsk);
1123 restore_fpu(tsk); 1154 restore_fpu(tsk);
@@ -1130,53 +1161,52 @@ EXPORT_SYMBOL_GPL(math_state_restore);
1130 1161
1131asmlinkage void math_emulate(long arg) 1162asmlinkage void math_emulate(long arg)
1132{ 1163{
1133 printk(KERN_EMERG "math-emulation not enabled and no coprocessor found.\n"); 1164 printk(KERN_EMERG
1134 printk(KERN_EMERG "killing %s.\n",current->comm); 1165 "math-emulation not enabled and no coprocessor found.\n");
1135 force_sig(SIGFPE,current); 1166 printk(KERN_EMERG "killing %s.\n", current->comm);
1167 force_sig(SIGFPE, current);
1136 schedule(); 1168 schedule();
1137} 1169}
1138 1170
1139#endif /* CONFIG_MATH_EMULATION */ 1171#endif /* CONFIG_MATH_EMULATION */
1140 1172
1141
1142void __init trap_init(void) 1173void __init trap_init(void)
1143{ 1174{
1144 int i; 1175 int i;
1145 1176
1146#ifdef CONFIG_EISA 1177#ifdef CONFIG_EISA
1147 void __iomem *p = early_ioremap(0x0FFFD9, 4); 1178 void __iomem *p = early_ioremap(0x0FFFD9, 4);
1148 if (readl(p) == 'E'+('I'<<8)+('S'<<16)+('A'<<24)) { 1179
1180 if (readl(p) == 'E' + ('I'<<8) + ('S'<<16) + ('A'<<24))
1149 EISA_bus = 1; 1181 EISA_bus = 1;
1150 }
1151 early_iounmap(p, 4); 1182 early_iounmap(p, 4);
1152#endif 1183#endif
1153 1184
1154#ifdef CONFIG_X86_LOCAL_APIC 1185#ifdef CONFIG_X86_LOCAL_APIC
1155 init_apic_mappings(); 1186 init_apic_mappings();
1156#endif 1187#endif
1157 1188 set_trap_gate(0, &divide_error);
1158 set_trap_gate(0,&divide_error); 1189 set_intr_gate(1, &debug);
1159 set_intr_gate(1,&debug); 1190 set_intr_gate(2, &nmi);
1160 set_intr_gate(2,&nmi);
1161 set_system_intr_gate(3, &int3); /* int3/4 can be called from all */ 1191 set_system_intr_gate(3, &int3); /* int3/4 can be called from all */
1162 set_system_gate(4,&overflow); 1192 set_system_gate(4, &overflow);
1163 set_trap_gate(5,&bounds); 1193 set_trap_gate(5, &bounds);
1164 set_trap_gate(6,&invalid_op); 1194 set_trap_gate(6, &invalid_op);
1165 set_trap_gate(7,&device_not_available); 1195 set_trap_gate(7, &device_not_available);
1166 set_task_gate(8,GDT_ENTRY_DOUBLEFAULT_TSS); 1196 set_task_gate(8, GDT_ENTRY_DOUBLEFAULT_TSS);
1167 set_trap_gate(9,&coprocessor_segment_overrun); 1197 set_trap_gate(9, &coprocessor_segment_overrun);
1168 set_trap_gate(10,&invalid_TSS); 1198 set_trap_gate(10, &invalid_TSS);
1169 set_trap_gate(11,&segment_not_present); 1199 set_trap_gate(11, &segment_not_present);
1170 set_trap_gate(12,&stack_segment); 1200 set_trap_gate(12, &stack_segment);
1171 set_trap_gate(13,&general_protection); 1201 set_trap_gate(13, &general_protection);
1172 set_intr_gate(14,&page_fault); 1202 set_intr_gate(14, &page_fault);
1173 set_trap_gate(15,&spurious_interrupt_bug); 1203 set_trap_gate(15, &spurious_interrupt_bug);
1174 set_trap_gate(16,&coprocessor_error); 1204 set_trap_gate(16, &coprocessor_error);
1175 set_trap_gate(17,&alignment_check); 1205 set_trap_gate(17, &alignment_check);
1176#ifdef CONFIG_X86_MCE 1206#ifdef CONFIG_X86_MCE
1177 set_trap_gate(18,&machine_check); 1207 set_trap_gate(18, &machine_check);
1178#endif 1208#endif
1179 set_trap_gate(19,&simd_coprocessor_error); 1209 set_trap_gate(19, &simd_coprocessor_error);
1180 1210
1181 /* 1211 /*
1182 * Verify that the FXSAVE/FXRSTOR data will be 16-byte aligned. 1212 * Verify that the FXSAVE/FXRSTOR data will be 16-byte aligned.
@@ -1189,21 +1219,22 @@ void __init trap_init(void)
1189 printk("done.\n"); 1219 printk("done.\n");
1190 } 1220 }
1191 if (cpu_has_xmm) { 1221 if (cpu_has_xmm) {
1192 printk(KERN_INFO "Enabling unmasked SIMD FPU exception " 1222 printk(KERN_INFO
1193 "support... "); 1223 "Enabling unmasked SIMD FPU exception support... ");
1194 set_in_cr4(X86_CR4_OSXMMEXCPT); 1224 set_in_cr4(X86_CR4_OSXMMEXCPT);
1195 printk("done.\n"); 1225 printk("done.\n");
1196 } 1226 }
1197 1227
1198 set_system_gate(SYSCALL_VECTOR,&system_call); 1228 set_system_gate(SYSCALL_VECTOR, &system_call);
1199 1229
1200 /* Reserve all the builtin and the syscall vector. */ 1230 /* Reserve all the builtin and the syscall vector: */
1201 for (i = 0; i < FIRST_EXTERNAL_VECTOR; i++) 1231 for (i = 0; i < FIRST_EXTERNAL_VECTOR; i++)
1202 set_bit(i, used_vectors); 1232 set_bit(i, used_vectors);
1233
1203 set_bit(SYSCALL_VECTOR, used_vectors); 1234 set_bit(SYSCALL_VECTOR, used_vectors);
1204 1235
1205 /* 1236 /*
1206 * Should be a barrier for any external CPU state. 1237 * Should be a barrier for any external CPU state:
1207 */ 1238 */
1208 cpu_init(); 1239 cpu_init();
1209 1240
@@ -1213,6 +1244,7 @@ void __init trap_init(void)
1213static int __init kstack_setup(char *s) 1244static int __init kstack_setup(char *s)
1214{ 1245{
1215 kstack_depth_to_print = simple_strtoul(s, NULL, 0); 1246 kstack_depth_to_print = simple_strtoul(s, NULL, 0);
1247
1216 return 1; 1248 return 1;
1217} 1249}
1218__setup("kstack=", kstack_setup); 1250__setup("kstack=", kstack_setup);
diff --git a/arch/x86/kernel/traps_64.c b/arch/x86/kernel/traps_64.c
index 4e073320e70a..79aa6fc0815c 100644
--- a/arch/x86/kernel/traps_64.c
+++ b/arch/x86/kernel/traps_64.c
@@ -33,6 +33,8 @@
33#include <linux/kdebug.h> 33#include <linux/kdebug.h>
34#include <linux/utsname.h> 34#include <linux/utsname.h>
35 35
36#include <mach_traps.h>
37
36#if defined(CONFIG_EDAC) 38#if defined(CONFIG_EDAC)
37#include <linux/edac.h> 39#include <linux/edac.h>
38#endif 40#endif
diff --git a/arch/x86/kernel/tsc_32.c b/arch/x86/kernel/tsc_32.c
index c2241e04ea5f..3d7e6e9fa6c2 100644
--- a/arch/x86/kernel/tsc_32.c
+++ b/arch/x86/kernel/tsc_32.c
@@ -84,8 +84,8 @@ DEFINE_PER_CPU(unsigned long, cyc2ns);
84 84
85static void set_cyc2ns_scale(unsigned long cpu_khz, int cpu) 85static void set_cyc2ns_scale(unsigned long cpu_khz, int cpu)
86{ 86{
87 unsigned long flags, prev_scale, *scale;
88 unsigned long long tsc_now, ns_now; 87 unsigned long long tsc_now, ns_now;
88 unsigned long flags, *scale;
89 89
90 local_irq_save(flags); 90 local_irq_save(flags);
91 sched_clock_idle_sleep_event(); 91 sched_clock_idle_sleep_event();
@@ -95,7 +95,6 @@ static void set_cyc2ns_scale(unsigned long cpu_khz, int cpu)
95 rdtscll(tsc_now); 95 rdtscll(tsc_now);
96 ns_now = __cycles_2_ns(tsc_now); 96 ns_now = __cycles_2_ns(tsc_now);
97 97
98 prev_scale = *scale;
99 if (cpu_khz) 98 if (cpu_khz)
100 *scale = (NSEC_PER_MSEC << CYC2NS_SCALE_FACTOR)/cpu_khz; 99 *scale = (NSEC_PER_MSEC << CYC2NS_SCALE_FACTOR)/cpu_khz;
101 100
@@ -392,13 +391,15 @@ void __init tsc_init(void)
392 int cpu; 391 int cpu;
393 392
394 if (!cpu_has_tsc) 393 if (!cpu_has_tsc)
395 goto out_no_tsc; 394 return;
396 395
397 cpu_khz = calculate_cpu_khz(); 396 cpu_khz = calculate_cpu_khz();
398 tsc_khz = cpu_khz; 397 tsc_khz = cpu_khz;
399 398
400 if (!cpu_khz) 399 if (!cpu_khz) {
401 goto out_no_tsc; 400 mark_tsc_unstable("could not calculate TSC khz");
401 return;
402 }
402 403
403 printk("Detected %lu.%03lu MHz processor.\n", 404 printk("Detected %lu.%03lu MHz processor.\n",
404 (unsigned long)cpu_khz / 1000, 405 (unsigned long)cpu_khz / 1000,
@@ -431,9 +432,4 @@ void __init tsc_init(void)
431 tsc_enabled = 1; 432 tsc_enabled = 1;
432 433
433 clocksource_register(&clocksource_tsc); 434 clocksource_register(&clocksource_tsc);
434
435 return;
436
437out_no_tsc:
438 setup_clear_cpu_cap(X86_FEATURE_TSC);
439} 435}
diff --git a/arch/x86/kernel/tsc_64.c b/arch/x86/kernel/tsc_64.c
index d3bebaaad842..ceeba01e7f47 100644
--- a/arch/x86/kernel/tsc_64.c
+++ b/arch/x86/kernel/tsc_64.c
@@ -44,8 +44,8 @@ DEFINE_PER_CPU(unsigned long, cyc2ns);
44 44
45static void set_cyc2ns_scale(unsigned long cpu_khz, int cpu) 45static void set_cyc2ns_scale(unsigned long cpu_khz, int cpu)
46{ 46{
47 unsigned long flags, prev_scale, *scale;
48 unsigned long long tsc_now, ns_now; 47 unsigned long long tsc_now, ns_now;
48 unsigned long flags, *scale;
49 49
50 local_irq_save(flags); 50 local_irq_save(flags);
51 sched_clock_idle_sleep_event(); 51 sched_clock_idle_sleep_event();
@@ -55,7 +55,6 @@ static void set_cyc2ns_scale(unsigned long cpu_khz, int cpu)
55 rdtscll(tsc_now); 55 rdtscll(tsc_now);
56 ns_now = __cycles_2_ns(tsc_now); 56 ns_now = __cycles_2_ns(tsc_now);
57 57
58 prev_scale = *scale;
59 if (cpu_khz) 58 if (cpu_khz)
60 *scale = (NSEC_PER_MSEC << CYC2NS_SCALE_FACTOR)/cpu_khz; 59 *scale = (NSEC_PER_MSEC << CYC2NS_SCALE_FACTOR)/cpu_khz;
61 60
diff --git a/arch/x86/kernel/vm86_32.c b/arch/x86/kernel/vm86_32.c
index 738c2104df30..38f566fa27d2 100644
--- a/arch/x86/kernel/vm86_32.c
+++ b/arch/x86/kernel/vm86_32.c
@@ -64,7 +64,7 @@
64 64
65 65
66#define KVM86 ((struct kernel_vm86_struct *)regs) 66#define KVM86 ((struct kernel_vm86_struct *)regs)
67#define VMPI KVM86->vm86plus 67#define VMPI KVM86->vm86plus
68 68
69 69
70/* 70/*
@@ -81,7 +81,7 @@
81#define VFLAGS (*(unsigned short *)&(current->thread.v86flags)) 81#define VFLAGS (*(unsigned short *)&(current->thread.v86flags))
82#define VEFLAGS (current->thread.v86flags) 82#define VEFLAGS (current->thread.v86flags)
83 83
84#define set_flags(X,new,mask) \ 84#define set_flags(X, new, mask) \
85((X) = ((X) & ~(mask)) | ((new) & (mask))) 85((X) = ((X) & ~(mask)) | ((new) & (mask)))
86 86
87#define SAFE_MASK (0xDD5) 87#define SAFE_MASK (0xDD5)
@@ -93,8 +93,10 @@ static int copy_vm86_regs_to_user(struct vm86_regs __user *user,
93{ 93{
94 int ret = 0; 94 int ret = 0;
95 95
96 /* kernel_vm86_regs is missing gs, so copy everything up to 96 /*
97 (but not including) orig_eax, and then rest including orig_eax. */ 97 * kernel_vm86_regs is missing gs, so copy everything up to
98 * (but not including) orig_eax, and then rest including orig_eax.
99 */
98 ret += copy_to_user(user, regs, offsetof(struct kernel_vm86_regs, pt.orig_ax)); 100 ret += copy_to_user(user, regs, offsetof(struct kernel_vm86_regs, pt.orig_ax));
99 ret += copy_to_user(&user->orig_eax, &regs->pt.orig_ax, 101 ret += copy_to_user(&user->orig_eax, &regs->pt.orig_ax,
100 sizeof(struct kernel_vm86_regs) - 102 sizeof(struct kernel_vm86_regs) -
@@ -120,7 +122,7 @@ static int copy_vm86_regs_from_user(struct kernel_vm86_regs *regs,
120 return ret; 122 return ret;
121} 123}
122 124
123struct pt_regs * save_v86_state(struct kernel_vm86_regs * regs) 125struct pt_regs *save_v86_state(struct kernel_vm86_regs *regs)
124{ 126{
125 struct tss_struct *tss; 127 struct tss_struct *tss;
126 struct pt_regs *ret; 128 struct pt_regs *ret;
@@ -137,9 +139,9 @@ struct pt_regs * save_v86_state(struct kernel_vm86_regs * regs)
137 printk("no vm86_info: BAD\n"); 139 printk("no vm86_info: BAD\n");
138 do_exit(SIGSEGV); 140 do_exit(SIGSEGV);
139 } 141 }
140 set_flags(regs->pt.flags, VEFLAGS, VIF_MASK | current->thread.v86mask); 142 set_flags(regs->pt.flags, VEFLAGS, X86_EFLAGS_VIF | current->thread.v86mask);
141 tmp = copy_vm86_regs_to_user(&current->thread.vm86_info->regs,regs); 143 tmp = copy_vm86_regs_to_user(&current->thread.vm86_info->regs, regs);
142 tmp += put_user(current->thread.screen_bitmap,&current->thread.vm86_info->screen_bitmap); 144 tmp += put_user(current->thread.screen_bitmap, &current->thread.vm86_info->screen_bitmap);
143 if (tmp) { 145 if (tmp) {
144 printk("vm86: could not access userspace vm86_info\n"); 146 printk("vm86: could not access userspace vm86_info\n");
145 do_exit(SIGSEGV); 147 do_exit(SIGSEGV);
@@ -237,20 +239,21 @@ asmlinkage int sys_vm86(struct pt_regs regs)
237 239
238 tsk = current; 240 tsk = current;
239 switch (regs.bx) { 241 switch (regs.bx) {
240 case VM86_REQUEST_IRQ: 242 case VM86_REQUEST_IRQ:
241 case VM86_FREE_IRQ: 243 case VM86_FREE_IRQ:
242 case VM86_GET_IRQ_BITS: 244 case VM86_GET_IRQ_BITS:
243 case VM86_GET_AND_RESET_IRQ: 245 case VM86_GET_AND_RESET_IRQ:
244 ret = do_vm86_irq_handling(regs.bx, (int)regs.cx); 246 ret = do_vm86_irq_handling(regs.bx, (int)regs.cx);
245 goto out; 247 goto out;
246 case VM86_PLUS_INSTALL_CHECK: 248 case VM86_PLUS_INSTALL_CHECK:
247 /* NOTE: on old vm86 stuff this will return the error 249 /*
248 from access_ok(), because the subfunction is 250 * NOTE: on old vm86 stuff this will return the error
249 interpreted as (invalid) address to vm86_struct. 251 * from access_ok(), because the subfunction is
250 So the installation check works. 252 * interpreted as (invalid) address to vm86_struct.
251 */ 253 * So the installation check works.
252 ret = 0; 254 */
253 goto out; 255 ret = 0;
256 goto out;
254 } 257 }
255 258
256 /* we come here only for functions VM86_ENTER, VM86_ENTER_NO_BYPASS */ 259 /* we come here only for functions VM86_ENTER, VM86_ENTER_NO_BYPASS */
@@ -296,21 +299,21 @@ static void do_sys_vm86(struct kernel_vm86_struct *info, struct task_struct *tsk
296 VEFLAGS = info->regs.pt.flags; 299 VEFLAGS = info->regs.pt.flags;
297 info->regs.pt.flags &= SAFE_MASK; 300 info->regs.pt.flags &= SAFE_MASK;
298 info->regs.pt.flags |= info->regs32->flags & ~SAFE_MASK; 301 info->regs.pt.flags |= info->regs32->flags & ~SAFE_MASK;
299 info->regs.pt.flags |= VM_MASK; 302 info->regs.pt.flags |= X86_VM_MASK;
300 303
301 switch (info->cpu_type) { 304 switch (info->cpu_type) {
302 case CPU_286: 305 case CPU_286:
303 tsk->thread.v86mask = 0; 306 tsk->thread.v86mask = 0;
304 break; 307 break;
305 case CPU_386: 308 case CPU_386:
306 tsk->thread.v86mask = NT_MASK | IOPL_MASK; 309 tsk->thread.v86mask = X86_EFLAGS_NT | X86_EFLAGS_IOPL;
307 break; 310 break;
308 case CPU_486: 311 case CPU_486:
309 tsk->thread.v86mask = AC_MASK | NT_MASK | IOPL_MASK; 312 tsk->thread.v86mask = X86_EFLAGS_AC | X86_EFLAGS_NT | X86_EFLAGS_IOPL;
310 break; 313 break;
311 default: 314 default:
312 tsk->thread.v86mask = ID_MASK | AC_MASK | NT_MASK | IOPL_MASK; 315 tsk->thread.v86mask = X86_EFLAGS_ID | X86_EFLAGS_AC | X86_EFLAGS_NT | X86_EFLAGS_IOPL;
313 break; 316 break;
314 } 317 }
315 318
316/* 319/*
@@ -346,9 +349,9 @@ static void do_sys_vm86(struct kernel_vm86_struct *info, struct task_struct *tsk
346 /* we never return here */ 349 /* we never return here */
347} 350}
348 351
349static inline void return_to_32bit(struct kernel_vm86_regs * regs16, int retval) 352static inline void return_to_32bit(struct kernel_vm86_regs *regs16, int retval)
350{ 353{
351 struct pt_regs * regs32; 354 struct pt_regs *regs32;
352 355
353 regs32 = save_v86_state(regs16); 356 regs32 = save_v86_state(regs16);
354 regs32->ax = retval; 357 regs32->ax = retval;
@@ -358,29 +361,30 @@ static inline void return_to_32bit(struct kernel_vm86_regs * regs16, int retval)
358 : : "r" (regs32), "r" (current_thread_info())); 361 : : "r" (regs32), "r" (current_thread_info()));
359} 362}
360 363
361static inline void set_IF(struct kernel_vm86_regs * regs) 364static inline void set_IF(struct kernel_vm86_regs *regs)
362{ 365{
363 VEFLAGS |= VIF_MASK; 366 VEFLAGS |= X86_EFLAGS_VIF;
364 if (VEFLAGS & VIP_MASK) 367 if (VEFLAGS & X86_EFLAGS_VIP)
365 return_to_32bit(regs, VM86_STI); 368 return_to_32bit(regs, VM86_STI);
366} 369}
367 370
368static inline void clear_IF(struct kernel_vm86_regs * regs) 371static inline void clear_IF(struct kernel_vm86_regs *regs)
369{ 372{
370 VEFLAGS &= ~VIF_MASK; 373 VEFLAGS &= ~X86_EFLAGS_VIF;
371} 374}
372 375
373static inline void clear_TF(struct kernel_vm86_regs * regs) 376static inline void clear_TF(struct kernel_vm86_regs *regs)
374{ 377{
375 regs->pt.flags &= ~TF_MASK; 378 regs->pt.flags &= ~X86_EFLAGS_TF;
376} 379}
377 380
378static inline void clear_AC(struct kernel_vm86_regs * regs) 381static inline void clear_AC(struct kernel_vm86_regs *regs)
379{ 382{
380 regs->pt.flags &= ~AC_MASK; 383 regs->pt.flags &= ~X86_EFLAGS_AC;
381} 384}
382 385
383/* It is correct to call set_IF(regs) from the set_vflags_* 386/*
387 * It is correct to call set_IF(regs) from the set_vflags_*
384 * functions. However someone forgot to call clear_IF(regs) 388 * functions. However someone forgot to call clear_IF(regs)
385 * in the opposite case. 389 * in the opposite case.
386 * After the command sequence CLI PUSHF STI POPF you should 390 * After the command sequence CLI PUSHF STI POPF you should
@@ -391,41 +395,41 @@ static inline void clear_AC(struct kernel_vm86_regs * regs)
391 * [KD] 395 * [KD]
392 */ 396 */
393 397
394static inline void set_vflags_long(unsigned long flags, struct kernel_vm86_regs * regs) 398static inline void set_vflags_long(unsigned long flags, struct kernel_vm86_regs *regs)
395{ 399{
396 set_flags(VEFLAGS, flags, current->thread.v86mask); 400 set_flags(VEFLAGS, flags, current->thread.v86mask);
397 set_flags(regs->pt.flags, flags, SAFE_MASK); 401 set_flags(regs->pt.flags, flags, SAFE_MASK);
398 if (flags & IF_MASK) 402 if (flags & X86_EFLAGS_IF)
399 set_IF(regs); 403 set_IF(regs);
400 else 404 else
401 clear_IF(regs); 405 clear_IF(regs);
402} 406}
403 407
404static inline void set_vflags_short(unsigned short flags, struct kernel_vm86_regs * regs) 408static inline void set_vflags_short(unsigned short flags, struct kernel_vm86_regs *regs)
405{ 409{
406 set_flags(VFLAGS, flags, current->thread.v86mask); 410 set_flags(VFLAGS, flags, current->thread.v86mask);
407 set_flags(regs->pt.flags, flags, SAFE_MASK); 411 set_flags(regs->pt.flags, flags, SAFE_MASK);
408 if (flags & IF_MASK) 412 if (flags & X86_EFLAGS_IF)
409 set_IF(regs); 413 set_IF(regs);
410 else 414 else
411 clear_IF(regs); 415 clear_IF(regs);
412} 416}
413 417
414static inline unsigned long get_vflags(struct kernel_vm86_regs * regs) 418static inline unsigned long get_vflags(struct kernel_vm86_regs *regs)
415{ 419{
416 unsigned long flags = regs->pt.flags & RETURN_MASK; 420 unsigned long flags = regs->pt.flags & RETURN_MASK;
417 421
418 if (VEFLAGS & VIF_MASK) 422 if (VEFLAGS & X86_EFLAGS_VIF)
419 flags |= IF_MASK; 423 flags |= X86_EFLAGS_IF;
420 flags |= IOPL_MASK; 424 flags |= X86_EFLAGS_IOPL;
421 return flags | (VEFLAGS & current->thread.v86mask); 425 return flags | (VEFLAGS & current->thread.v86mask);
422} 426}
423 427
424static inline int is_revectored(int nr, struct revectored_struct * bitmap) 428static inline int is_revectored(int nr, struct revectored_struct *bitmap)
425{ 429{
426 __asm__ __volatile__("btl %2,%1\n\tsbbl %0,%0" 430 __asm__ __volatile__("btl %2,%1\n\tsbbl %0,%0"
427 :"=r" (nr) 431 :"=r" (nr)
428 :"m" (*bitmap),"r" (nr)); 432 :"m" (*bitmap), "r" (nr));
429 return nr; 433 return nr;
430} 434}
431 435
@@ -437,7 +441,7 @@ static inline int is_revectored(int nr, struct revectored_struct * bitmap)
437 ptr--; \ 441 ptr--; \
438 if (put_user(__val, base + ptr) < 0) \ 442 if (put_user(__val, base + ptr) < 0) \
439 goto err_label; \ 443 goto err_label; \
440 } while(0) 444 } while (0)
441 445
442#define pushw(base, ptr, val, err_label) \ 446#define pushw(base, ptr, val, err_label) \
443 do { \ 447 do { \
@@ -448,7 +452,7 @@ static inline int is_revectored(int nr, struct revectored_struct * bitmap)
448 ptr--; \ 452 ptr--; \
449 if (put_user(val_byte(__val, 0), base + ptr) < 0) \ 453 if (put_user(val_byte(__val, 0), base + ptr) < 0) \
450 goto err_label; \ 454 goto err_label; \
451 } while(0) 455 } while (0)
452 456
453#define pushl(base, ptr, val, err_label) \ 457#define pushl(base, ptr, val, err_label) \
454 do { \ 458 do { \
@@ -465,7 +469,7 @@ static inline int is_revectored(int nr, struct revectored_struct * bitmap)
465 ptr--; \ 469 ptr--; \
466 if (put_user(val_byte(__val, 0), base + ptr) < 0) \ 470 if (put_user(val_byte(__val, 0), base + ptr) < 0) \
467 goto err_label; \ 471 goto err_label; \
468 } while(0) 472 } while (0)
469 473
470#define popb(base, ptr, err_label) \ 474#define popb(base, ptr, err_label) \
471 ({ \ 475 ({ \
@@ -512,7 +516,7 @@ static inline int is_revectored(int nr, struct revectored_struct * bitmap)
512 * in userspace is always better than an Oops anyway.) [KD] 516 * in userspace is always better than an Oops anyway.) [KD]
513 */ 517 */
514static void do_int(struct kernel_vm86_regs *regs, int i, 518static void do_int(struct kernel_vm86_regs *regs, int i,
515 unsigned char __user * ssp, unsigned short sp) 519 unsigned char __user *ssp, unsigned short sp)
516{ 520{
517 unsigned long __user *intr_ptr; 521 unsigned long __user *intr_ptr;
518 unsigned long segoffs; 522 unsigned long segoffs;
@@ -521,7 +525,7 @@ static void do_int(struct kernel_vm86_regs *regs, int i,
521 goto cannot_handle; 525 goto cannot_handle;
522 if (is_revectored(i, &KVM86->int_revectored)) 526 if (is_revectored(i, &KVM86->int_revectored))
523 goto cannot_handle; 527 goto cannot_handle;
524 if (i==0x21 && is_revectored(AH(regs),&KVM86->int21_revectored)) 528 if (i == 0x21 && is_revectored(AH(regs), &KVM86->int21_revectored))
525 goto cannot_handle; 529 goto cannot_handle;
526 intr_ptr = (unsigned long __user *) (i << 2); 530 intr_ptr = (unsigned long __user *) (i << 2);
527 if (get_user(segoffs, intr_ptr)) 531 if (get_user(segoffs, intr_ptr))
@@ -543,30 +547,23 @@ cannot_handle:
543 return_to_32bit(regs, VM86_INTx + (i << 8)); 547 return_to_32bit(regs, VM86_INTx + (i << 8));
544} 548}
545 549
546int handle_vm86_trap(struct kernel_vm86_regs * regs, long error_code, int trapno) 550int handle_vm86_trap(struct kernel_vm86_regs *regs, long error_code, int trapno)
547{ 551{
548 if (VMPI.is_vm86pus) { 552 if (VMPI.is_vm86pus) {
549 if ( (trapno==3) || (trapno==1) ) 553 if ((trapno == 3) || (trapno == 1))
550 return_to_32bit(regs, VM86_TRAP + (trapno << 8)); 554 return_to_32bit(regs, VM86_TRAP + (trapno << 8));
551 do_int(regs, trapno, (unsigned char __user *) (regs->pt.ss << 4), SP(regs)); 555 do_int(regs, trapno, (unsigned char __user *) (regs->pt.ss << 4), SP(regs));
552 return 0; 556 return 0;
553 } 557 }
554 if (trapno !=1) 558 if (trapno != 1)
555 return 1; /* we let this handle by the calling routine */ 559 return 1; /* we let this handle by the calling routine */
556 if (current->ptrace & PT_PTRACED) {
557 unsigned long flags;
558 spin_lock_irqsave(&current->sighand->siglock, flags);
559 sigdelset(&current->blocked, SIGTRAP);
560 recalc_sigpending();
561 spin_unlock_irqrestore(&current->sighand->siglock, flags);
562 }
563 send_sig(SIGTRAP, current, 1);
564 current->thread.trap_no = trapno; 560 current->thread.trap_no = trapno;
565 current->thread.error_code = error_code; 561 current->thread.error_code = error_code;
562 force_sig(SIGTRAP, current);
566 return 0; 563 return 0;
567} 564}
568 565
569void handle_vm86_fault(struct kernel_vm86_regs * regs, long error_code) 566void handle_vm86_fault(struct kernel_vm86_regs *regs, long error_code)
570{ 567{
571 unsigned char opcode; 568 unsigned char opcode;
572 unsigned char __user *csp; 569 unsigned char __user *csp;
@@ -576,11 +573,11 @@ void handle_vm86_fault(struct kernel_vm86_regs * regs, long error_code)
576 573
577#define CHECK_IF_IN_TRAP \ 574#define CHECK_IF_IN_TRAP \
578 if (VMPI.vm86dbg_active && VMPI.vm86dbg_TFpendig) \ 575 if (VMPI.vm86dbg_active && VMPI.vm86dbg_TFpendig) \
579 newflags |= TF_MASK 576 newflags |= X86_EFLAGS_TF
580#define VM86_FAULT_RETURN do { \ 577#define VM86_FAULT_RETURN do { \
581 if (VMPI.force_return_for_pic && (VEFLAGS & (IF_MASK | VIF_MASK))) \ 578 if (VMPI.force_return_for_pic && (VEFLAGS & (X86_EFLAGS_IF | X86_EFLAGS_VIF))) \
582 return_to_32bit(regs, VM86_PICRETURN); \ 579 return_to_32bit(regs, VM86_PICRETURN); \
583 if (orig_flags & TF_MASK) \ 580 if (orig_flags & X86_EFLAGS_TF) \
584 handle_vm86_trap(regs, 0, 1); \ 581 handle_vm86_trap(regs, 0, 1); \
585 return; } while (0) 582 return; } while (0)
586 583
@@ -595,17 +592,17 @@ void handle_vm86_fault(struct kernel_vm86_regs * regs, long error_code)
595 pref_done = 0; 592 pref_done = 0;
596 do { 593 do {
597 switch (opcode = popb(csp, ip, simulate_sigsegv)) { 594 switch (opcode = popb(csp, ip, simulate_sigsegv)) {
598 case 0x66: /* 32-bit data */ data32=1; break; 595 case 0x66: /* 32-bit data */ data32 = 1; break;
599 case 0x67: /* 32-bit address */ break; 596 case 0x67: /* 32-bit address */ break;
600 case 0x2e: /* CS */ break; 597 case 0x2e: /* CS */ break;
601 case 0x3e: /* DS */ break; 598 case 0x3e: /* DS */ break;
602 case 0x26: /* ES */ break; 599 case 0x26: /* ES */ break;
603 case 0x36: /* SS */ break; 600 case 0x36: /* SS */ break;
604 case 0x65: /* GS */ break; 601 case 0x65: /* GS */ break;
605 case 0x64: /* FS */ break; 602 case 0x64: /* FS */ break;
606 case 0xf2: /* repnz */ break; 603 case 0xf2: /* repnz */ break;
607 case 0xf3: /* rep */ break; 604 case 0xf3: /* rep */ break;
608 default: pref_done = 1; 605 default: pref_done = 1;
609 } 606 }
610 } while (!pref_done); 607 } while (!pref_done);
611 608
@@ -628,7 +625,7 @@ void handle_vm86_fault(struct kernel_vm86_regs * regs, long error_code)
628 { 625 {
629 unsigned long newflags; 626 unsigned long newflags;
630 if (data32) { 627 if (data32) {
631 newflags=popl(ssp, sp, simulate_sigsegv); 628 newflags = popl(ssp, sp, simulate_sigsegv);
632 SP(regs) += 4; 629 SP(regs) += 4;
633 } else { 630 } else {
634 newflags = popw(ssp, sp, simulate_sigsegv); 631 newflags = popw(ssp, sp, simulate_sigsegv);
@@ -636,20 +633,20 @@ void handle_vm86_fault(struct kernel_vm86_regs * regs, long error_code)
636 } 633 }
637 IP(regs) = ip; 634 IP(regs) = ip;
638 CHECK_IF_IN_TRAP; 635 CHECK_IF_IN_TRAP;
639 if (data32) { 636 if (data32)
640 set_vflags_long(newflags, regs); 637 set_vflags_long(newflags, regs);
641 } else { 638 else
642 set_vflags_short(newflags, regs); 639 set_vflags_short(newflags, regs);
643 } 640
644 VM86_FAULT_RETURN; 641 VM86_FAULT_RETURN;
645 } 642 }
646 643
647 /* int xx */ 644 /* int xx */
648 case 0xcd: { 645 case 0xcd: {
649 int intno=popb(csp, ip, simulate_sigsegv); 646 int intno = popb(csp, ip, simulate_sigsegv);
650 IP(regs) = ip; 647 IP(regs) = ip;
651 if (VMPI.vm86dbg_active) { 648 if (VMPI.vm86dbg_active) {
652 if ( (1 << (intno &7)) & VMPI.vm86dbg_intxxtab[intno >> 3] ) 649 if ((1 << (intno & 7)) & VMPI.vm86dbg_intxxtab[intno >> 3])
653 return_to_32bit(regs, VM86_INTx + (intno << 8)); 650 return_to_32bit(regs, VM86_INTx + (intno << 8));
654 } 651 }
655 do_int(regs, intno, ssp, sp); 652 do_int(regs, intno, ssp, sp);
@@ -663,9 +660,9 @@ void handle_vm86_fault(struct kernel_vm86_regs * regs, long error_code)
663 unsigned long newcs; 660 unsigned long newcs;
664 unsigned long newflags; 661 unsigned long newflags;
665 if (data32) { 662 if (data32) {
666 newip=popl(ssp, sp, simulate_sigsegv); 663 newip = popl(ssp, sp, simulate_sigsegv);
667 newcs=popl(ssp, sp, simulate_sigsegv); 664 newcs = popl(ssp, sp, simulate_sigsegv);
668 newflags=popl(ssp, sp, simulate_sigsegv); 665 newflags = popl(ssp, sp, simulate_sigsegv);
669 SP(regs) += 12; 666 SP(regs) += 12;
670 } else { 667 } else {
671 newip = popw(ssp, sp, simulate_sigsegv); 668 newip = popw(ssp, sp, simulate_sigsegv);
@@ -734,18 +731,18 @@ static struct vm86_irqs {
734static DEFINE_SPINLOCK(irqbits_lock); 731static DEFINE_SPINLOCK(irqbits_lock);
735static int irqbits; 732static int irqbits;
736 733
737#define ALLOWED_SIGS ( 1 /* 0 = don't send a signal */ \ 734#define ALLOWED_SIGS (1 /* 0 = don't send a signal */ \
738 | (1 << SIGUSR1) | (1 << SIGUSR2) | (1 << SIGIO) | (1 << SIGURG) \ 735 | (1 << SIGUSR1) | (1 << SIGUSR2) | (1 << SIGIO) | (1 << SIGURG) \
739 | (1 << SIGUNUSED) ) 736 | (1 << SIGUNUSED))
740 737
741static irqreturn_t irq_handler(int intno, void *dev_id) 738static irqreturn_t irq_handler(int intno, void *dev_id)
742{ 739{
743 int irq_bit; 740 int irq_bit;
744 unsigned long flags; 741 unsigned long flags;
745 742
746 spin_lock_irqsave(&irqbits_lock, flags); 743 spin_lock_irqsave(&irqbits_lock, flags);
747 irq_bit = 1 << intno; 744 irq_bit = 1 << intno;
748 if ((irqbits & irq_bit) || ! vm86_irqs[intno].tsk) 745 if ((irqbits & irq_bit) || !vm86_irqs[intno].tsk)
749 goto out; 746 goto out;
750 irqbits |= irq_bit; 747 irqbits |= irq_bit;
751 if (vm86_irqs[intno].sig) 748 if (vm86_irqs[intno].sig)
@@ -759,7 +756,7 @@ static irqreturn_t irq_handler(int intno, void *dev_id)
759 return IRQ_HANDLED; 756 return IRQ_HANDLED;
760 757
761out: 758out:
762 spin_unlock_irqrestore(&irqbits_lock, flags); 759 spin_unlock_irqrestore(&irqbits_lock, flags);
763 return IRQ_NONE; 760 return IRQ_NONE;
764} 761}
765 762
@@ -770,9 +767,9 @@ static inline void free_vm86_irq(int irqnumber)
770 free_irq(irqnumber, NULL); 767 free_irq(irqnumber, NULL);
771 vm86_irqs[irqnumber].tsk = NULL; 768 vm86_irqs[irqnumber].tsk = NULL;
772 769
773 spin_lock_irqsave(&irqbits_lock, flags); 770 spin_lock_irqsave(&irqbits_lock, flags);
774 irqbits &= ~(1 << irqnumber); 771 irqbits &= ~(1 << irqnumber);
775 spin_unlock_irqrestore(&irqbits_lock, flags); 772 spin_unlock_irqrestore(&irqbits_lock, flags);
776} 773}
777 774
778void release_vm86_irqs(struct task_struct *task) 775void release_vm86_irqs(struct task_struct *task)
@@ -788,10 +785,10 @@ static inline int get_and_reset_irq(int irqnumber)
788 int bit; 785 int bit;
789 unsigned long flags; 786 unsigned long flags;
790 int ret = 0; 787 int ret = 0;
791 788
792 if (invalid_vm86_irq(irqnumber)) return 0; 789 if (invalid_vm86_irq(irqnumber)) return 0;
793 if (vm86_irqs[irqnumber].tsk != current) return 0; 790 if (vm86_irqs[irqnumber].tsk != current) return 0;
794 spin_lock_irqsave(&irqbits_lock, flags); 791 spin_lock_irqsave(&irqbits_lock, flags);
795 bit = irqbits & (1 << irqnumber); 792 bit = irqbits & (1 << irqnumber);
796 irqbits &= ~bit; 793 irqbits &= ~bit;
797 if (bit) { 794 if (bit) {
@@ -799,7 +796,7 @@ static inline int get_and_reset_irq(int irqnumber)
799 ret = 1; 796 ret = 1;
800 } 797 }
801 798
802 spin_unlock_irqrestore(&irqbits_lock, flags); 799 spin_unlock_irqrestore(&irqbits_lock, flags);
803 return ret; 800 return ret;
804} 801}
805 802
diff --git a/arch/x86/kernel/vmlinux_32.lds.S b/arch/x86/kernel/vmlinux_32.lds.S
index 2ffa9656fe7a..ce5ed083a1e9 100644
--- a/arch/x86/kernel/vmlinux_32.lds.S
+++ b/arch/x86/kernel/vmlinux_32.lds.S
@@ -149,6 +149,11 @@ SECTIONS
149 *(.con_initcall.init) 149 *(.con_initcall.init)
150 __con_initcall_end = .; 150 __con_initcall_end = .;
151 } 151 }
152 .x86cpuvendor.init : AT(ADDR(.x86cpuvendor.init) - LOAD_OFFSET) {
153 __x86cpuvendor_start = .;
154 *(.x86cpuvendor.init)
155 __x86cpuvendor_end = .;
156 }
152 SECURITY_INIT 157 SECURITY_INIT
153 . = ALIGN(4); 158 . = ALIGN(4);
154 .altinstructions : AT(ADDR(.altinstructions) - LOAD_OFFSET) { 159 .altinstructions : AT(ADDR(.altinstructions) - LOAD_OFFSET) {
diff --git a/arch/x86/kernel/vmlinux_64.lds.S b/arch/x86/kernel/vmlinux_64.lds.S
index fab132299735..b7ab3c335fae 100644
--- a/arch/x86/kernel/vmlinux_64.lds.S
+++ b/arch/x86/kernel/vmlinux_64.lds.S
@@ -177,6 +177,11 @@ SECTIONS
177 *(.con_initcall.init) 177 *(.con_initcall.init)
178 } 178 }
179 __con_initcall_end = .; 179 __con_initcall_end = .;
180 __x86cpuvendor_start = .;
181 .x86cpuvendor.init : AT(ADDR(.x86cpuvendor.init) - LOAD_OFFSET) {
182 *(.x86cpuvendor.init)
183 }
184 __x86cpuvendor_end = .;
180 SECURITY_INIT 185 SECURITY_INIT
181 186
182 . = ALIGN(8); 187 . = ALIGN(8);
@@ -247,3 +252,9 @@ SECTIONS
247 252
248 DWARF_DEBUG 253 DWARF_DEBUG
249} 254}
255
256/*
257 * Build-time check on the image size:
258 */
259ASSERT((_end - _text <= KERNEL_IMAGE_SIZE),
260 "kernel image bigger than KERNEL_IMAGE_SIZE")
diff --git a/arch/x86/kernel/vsmp_64.c b/arch/x86/kernel/vsmp_64.c
index d971210a6d36..caf2a26f5cfd 100644
--- a/arch/x86/kernel/vsmp_64.c
+++ b/arch/x86/kernel/vsmp_64.c
@@ -8,6 +8,8 @@
8 * 8 *
9 * Ravikiran Thirumalai <kiran@scalemp.com>, 9 * Ravikiran Thirumalai <kiran@scalemp.com>,
10 * Shai Fultheim <shai@scalemp.com> 10 * Shai Fultheim <shai@scalemp.com>
11 * Paravirt ops integration: Glauber de Oliveira Costa <gcosta@redhat.com>,
12 * Ravikiran Thirumalai <kiran@scalemp.com>
11 */ 13 */
12 14
13#include <linux/init.h> 15#include <linux/init.h>
@@ -15,38 +17,137 @@
15#include <linux/pci_regs.h> 17#include <linux/pci_regs.h>
16#include <asm/pci-direct.h> 18#include <asm/pci-direct.h>
17#include <asm/io.h> 19#include <asm/io.h>
20#include <asm/paravirt.h>
18 21
19static int __init vsmp_init(void) 22#if defined CONFIG_PCI && defined CONFIG_PARAVIRT
23/*
24 * Interrupt control on vSMPowered systems:
25 * ~AC is a shadow of IF. If IF is 'on' AC should be 'off'
26 * and vice versa.
27 */
28
29static unsigned long vsmp_save_fl(void)
20{ 30{
21 void *address; 31 unsigned long flags = native_save_fl();
22 unsigned int cap, ctl;
23 32
24 if (!early_pci_allowed()) 33 if (!(flags & X86_EFLAGS_IF) || (flags & X86_EFLAGS_AC))
25 return 0; 34 flags &= ~X86_EFLAGS_IF;
35 return flags;
36}
26 37
27 /* Check if we are running on a ScaleMP vSMP box */ 38static void vsmp_restore_fl(unsigned long flags)
28 if ((read_pci_config_16(0, 0x1f, 0, PCI_VENDOR_ID) != 39{
29 PCI_VENDOR_ID_SCALEMP) || 40 if (flags & X86_EFLAGS_IF)
30 (read_pci_config_16(0, 0x1f, 0, PCI_DEVICE_ID) != 41 flags &= ~X86_EFLAGS_AC;
31 PCI_DEVICE_ID_SCALEMP_VSMP_CTL)) 42 else
32 return 0; 43 flags |= X86_EFLAGS_AC;
44 native_restore_fl(flags);
45}
46
47static void vsmp_irq_disable(void)
48{
49 unsigned long flags = native_save_fl();
50
51 native_restore_fl((flags & ~X86_EFLAGS_IF) | X86_EFLAGS_AC);
52}
53
54static void vsmp_irq_enable(void)
55{
56 unsigned long flags = native_save_fl();
57
58 native_restore_fl((flags | X86_EFLAGS_IF) & (~X86_EFLAGS_AC));
59}
60
61static unsigned __init vsmp_patch(u8 type, u16 clobbers, void *ibuf,
62 unsigned long addr, unsigned len)
63{
64 switch (type) {
65 case PARAVIRT_PATCH(pv_irq_ops.irq_enable):
66 case PARAVIRT_PATCH(pv_irq_ops.irq_disable):
67 case PARAVIRT_PATCH(pv_irq_ops.save_fl):
68 case PARAVIRT_PATCH(pv_irq_ops.restore_fl):
69 return paravirt_patch_default(type, clobbers, ibuf, addr, len);
70 default:
71 return native_patch(type, clobbers, ibuf, addr, len);
72 }
73
74}
75
76static void __init set_vsmp_pv_ops(void)
77{
78 void *address;
79 unsigned int cap, ctl, cfg;
33 80
34 /* set vSMP magic bits to indicate vSMP capable kernel */ 81 /* set vSMP magic bits to indicate vSMP capable kernel */
35 address = ioremap(read_pci_config(0, 0x1f, 0, PCI_BASE_ADDRESS_0), 8); 82 cfg = read_pci_config(0, 0x1f, 0, PCI_BASE_ADDRESS_0);
83 address = early_ioremap(cfg, 8);
36 cap = readl(address); 84 cap = readl(address);
37 ctl = readl(address + 4); 85 ctl = readl(address + 4);
38 printk(KERN_INFO "vSMP CTL: capabilities:0x%08x control:0x%08x\n", 86 printk(KERN_INFO "vSMP CTL: capabilities:0x%08x control:0x%08x\n",
39 cap, ctl); 87 cap, ctl);
40 if (cap & ctl & (1 << 4)) { 88 if (cap & ctl & (1 << 4)) {
41 /* Turn on vSMP IRQ fastpath handling (see system.h) */ 89 /* Setup irq ops and turn on vSMP IRQ fastpath handling */
90 pv_irq_ops.irq_disable = vsmp_irq_disable;
91 pv_irq_ops.irq_enable = vsmp_irq_enable;
92 pv_irq_ops.save_fl = vsmp_save_fl;
93 pv_irq_ops.restore_fl = vsmp_restore_fl;
94 pv_init_ops.patch = vsmp_patch;
95
42 ctl &= ~(1 << 4); 96 ctl &= ~(1 << 4);
43 writel(ctl, address + 4); 97 writel(ctl, address + 4);
44 ctl = readl(address + 4); 98 ctl = readl(address + 4);
45 printk(KERN_INFO "vSMP CTL: control set to:0x%08x\n", ctl); 99 printk(KERN_INFO "vSMP CTL: control set to:0x%08x\n", ctl);
46 } 100 }
47 101
48 iounmap(address); 102 early_iounmap(address, 8);
103}
104#else
105static void __init set_vsmp_pv_ops(void)
106{
107}
108#endif
109
110#ifdef CONFIG_PCI
111static int is_vsmp = -1;
112
113static void __init detect_vsmp_box(void)
114{
115 is_vsmp = 0;
116
117 if (!early_pci_allowed())
118 return;
119
120 /* Check if we are running on a ScaleMP vSMPowered box */
121 if (read_pci_config(0, 0x1f, 0, PCI_VENDOR_ID) ==
122 (PCI_VENDOR_ID_SCALEMP | (PCI_DEVICE_ID_SCALEMP_VSMP_CTL << 16)))
123 is_vsmp = 1;
124}
125
126int is_vsmp_box(void)
127{
128 if (is_vsmp != -1)
129 return is_vsmp;
130 else {
131 WARN_ON_ONCE(1);
132 return 0;
133 }
134}
135#else
136static int __init detect_vsmp_box(void)
137{
138}
139int is_vsmp_box(void)
140{
49 return 0; 141 return 0;
50} 142}
143#endif
51 144
52core_initcall(vsmp_init); 145void __init vsmp_init(void)
146{
147 detect_vsmp_box();
148 if (!is_vsmp_box())
149 return;
150
151 set_vsmp_pv_ops();
152 return;
153}
diff --git a/arch/x86/kernel/x8664_ksyms_64.c b/arch/x86/kernel/x8664_ksyms_64.c
index a66e9c1a0537..58882f9f2637 100644
--- a/arch/x86/kernel/x8664_ksyms_64.c
+++ b/arch/x86/kernel/x8664_ksyms_64.c
@@ -4,7 +4,6 @@
4#include <linux/module.h> 4#include <linux/module.h>
5#include <linux/smp.h> 5#include <linux/smp.h>
6 6
7#include <asm/semaphore.h>
8#include <asm/processor.h> 7#include <asm/processor.h>
9#include <asm/uaccess.h> 8#include <asm/uaccess.h>
10#include <asm/pgtable.h> 9#include <asm/pgtable.h>
@@ -12,11 +11,6 @@
12 11
13EXPORT_SYMBOL(kernel_thread); 12EXPORT_SYMBOL(kernel_thread);
14 13
15EXPORT_SYMBOL(__down_failed);
16EXPORT_SYMBOL(__down_failed_interruptible);
17EXPORT_SYMBOL(__down_failed_trylock);
18EXPORT_SYMBOL(__up_wakeup);
19
20EXPORT_SYMBOL(__get_user_1); 14EXPORT_SYMBOL(__get_user_1);
21EXPORT_SYMBOL(__get_user_2); 15EXPORT_SYMBOL(__get_user_2);
22EXPORT_SYMBOL(__get_user_4); 16EXPORT_SYMBOL(__get_user_4);
@@ -35,15 +29,17 @@ EXPORT_SYMBOL(__copy_from_user_inatomic);
35EXPORT_SYMBOL(copy_page); 29EXPORT_SYMBOL(copy_page);
36EXPORT_SYMBOL(clear_page); 30EXPORT_SYMBOL(clear_page);
37 31
38/* Export string functions. We normally rely on gcc builtin for most of these, 32/*
39 but gcc sometimes decides not to inline them. */ 33 * Export string functions. We normally rely on gcc builtin for most of these,
34 * but gcc sometimes decides not to inline them.
35 */
40#undef memcpy 36#undef memcpy
41#undef memset 37#undef memset
42#undef memmove 38#undef memmove
43 39
44extern void * memset(void *,int,__kernel_size_t); 40extern void *memset(void *, int, __kernel_size_t);
45extern void * memcpy(void *,const void *,__kernel_size_t); 41extern void *memcpy(void *, const void *, __kernel_size_t);
46extern void * __memcpy(void *,const void *,__kernel_size_t); 42extern void *__memcpy(void *, const void *, __kernel_size_t);
47 43
48EXPORT_SYMBOL(memset); 44EXPORT_SYMBOL(memset);
49EXPORT_SYMBOL(memcpy); 45EXPORT_SYMBOL(memcpy);
diff --git a/arch/x86/lguest/boot.c b/arch/x86/lguest/boot.c
index 3335b4595efd..af65b2da3ba0 100644
--- a/arch/x86/lguest/boot.c
+++ b/arch/x86/lguest/boot.c
@@ -661,7 +661,7 @@ static int lguest_clockevent_set_next_event(unsigned long delta,
661 if (delta < LG_CLOCK_MIN_DELTA) { 661 if (delta < LG_CLOCK_MIN_DELTA) {
662 if (printk_ratelimit()) 662 if (printk_ratelimit())
663 printk(KERN_DEBUG "%s: small delta %lu ns\n", 663 printk(KERN_DEBUG "%s: small delta %lu ns\n",
664 __FUNCTION__, delta); 664 __func__, delta);
665 return -ETIME; 665 return -ETIME;
666 } 666 }
667 667
diff --git a/arch/x86/lib/memcpy_32.c b/arch/x86/lib/memcpy_32.c
index 37756b6fb329..5415a9d06f53 100644
--- a/arch/x86/lib/memcpy_32.c
+++ b/arch/x86/lib/memcpy_32.c
@@ -25,7 +25,7 @@ void *memmove(void *dest, const void *src, size_t n)
25 int d0, d1, d2; 25 int d0, d1, d2;
26 26
27 if (dest < src) { 27 if (dest < src) {
28 memcpy(dest,src,n); 28 memcpy(dest, src, n);
29 } else { 29 } else {
30 __asm__ __volatile__( 30 __asm__ __volatile__(
31 "std\n\t" 31 "std\n\t"
diff --git a/arch/x86/lib/memmove_64.c b/arch/x86/lib/memmove_64.c
index 80175e47b190..0a33909bf122 100644
--- a/arch/x86/lib/memmove_64.c
+++ b/arch/x86/lib/memmove_64.c
@@ -6,10 +6,10 @@
6#include <linux/module.h> 6#include <linux/module.h>
7 7
8#undef memmove 8#undef memmove
9void *memmove(void * dest,const void *src,size_t count) 9void *memmove(void *dest, const void *src, size_t count)
10{ 10{
11 if (dest < src) { 11 if (dest < src) {
12 return memcpy(dest,src,count); 12 return memcpy(dest, src, count);
13 } else { 13 } else {
14 char *p = dest + count; 14 char *p = dest + count;
15 const char *s = src + count; 15 const char *s = src + count;
@@ -17,5 +17,5 @@ void *memmove(void * dest,const void *src,size_t count)
17 *--p = *--s; 17 *--p = *--s;
18 } 18 }
19 return dest; 19 return dest;
20} 20}
21EXPORT_SYMBOL(memmove); 21EXPORT_SYMBOL(memmove);
diff --git a/arch/x86/lib/mmx_32.c b/arch/x86/lib/mmx_32.c
index cc9b4a4450f3..c9f2d9ba8dd8 100644
--- a/arch/x86/lib/mmx_32.c
+++ b/arch/x86/lib/mmx_32.c
@@ -1,32 +1,30 @@
1#include <linux/types.h>
2#include <linux/string.h>
3#include <linux/sched.h>
4#include <linux/hardirq.h>
5#include <linux/module.h>
6
7#include <asm/asm.h>
8#include <asm/i387.h>
9
10
11/* 1/*
12 * MMX 3DNow! library helper functions 2 * MMX 3DNow! library helper functions
13 * 3 *
14 * To do: 4 * To do:
15 * We can use MMX just for prefetch in IRQ's. This may be a win. 5 * We can use MMX just for prefetch in IRQ's. This may be a win.
16 * (reported so on K6-III) 6 * (reported so on K6-III)
17 * We should use a better code neutral filler for the short jump 7 * We should use a better code neutral filler for the short jump
18 * leal ebx. [ebx] is apparently best for K6-2, but Cyrix ?? 8 * leal ebx. [ebx] is apparently best for K6-2, but Cyrix ??
19 * We also want to clobber the filler register so we don't get any 9 * We also want to clobber the filler register so we don't get any
20 * register forwarding stalls on the filler. 10 * register forwarding stalls on the filler.
21 * 11 *
22 * Add *user handling. Checksums are not a win with MMX on any CPU 12 * Add *user handling. Checksums are not a win with MMX on any CPU
23 * tested so far for any MMX solution figured. 13 * tested so far for any MMX solution figured.
24 * 14 *
25 * 22/09/2000 - Arjan van de Ven 15 * 22/09/2000 - Arjan van de Ven
26 * Improved for non-egineering-sample Athlons 16 * Improved for non-egineering-sample Athlons
27 * 17 *
28 */ 18 */
29 19#include <linux/hardirq.h>
20#include <linux/string.h>
21#include <linux/module.h>
22#include <linux/sched.h>
23#include <linux/types.h>
24
25#include <asm/i387.h>
26#include <asm/asm.h>
27
30void *_mmx_memcpy(void *to, const void *from, size_t len) 28void *_mmx_memcpy(void *to, const void *from, size_t len)
31{ 29{
32 void *p; 30 void *p;
@@ -51,12 +49,10 @@ void *_mmx_memcpy(void *to, const void *from, size_t len)
51 "3: movw $0x1AEB, 1b\n" /* jmp on 26 bytes */ 49 "3: movw $0x1AEB, 1b\n" /* jmp on 26 bytes */
52 " jmp 2b\n" 50 " jmp 2b\n"
53 ".previous\n" 51 ".previous\n"
54 _ASM_EXTABLE(1b,3b) 52 _ASM_EXTABLE(1b, 3b)
55 : : "r" (from) ); 53 : : "r" (from));
56 54
57 55 for ( ; i > 5; i--) {
58 for(; i>5; i--)
59 {
60 __asm__ __volatile__ ( 56 __asm__ __volatile__ (
61 "1: prefetch 320(%0)\n" 57 "1: prefetch 320(%0)\n"
62 "2: movq (%0), %%mm0\n" 58 "2: movq (%0), %%mm0\n"
@@ -79,14 +75,14 @@ void *_mmx_memcpy(void *to, const void *from, size_t len)
79 "3: movw $0x05EB, 1b\n" /* jmp on 5 bytes */ 75 "3: movw $0x05EB, 1b\n" /* jmp on 5 bytes */
80 " jmp 2b\n" 76 " jmp 2b\n"
81 ".previous\n" 77 ".previous\n"
82 _ASM_EXTABLE(1b,3b) 78 _ASM_EXTABLE(1b, 3b)
83 : : "r" (from), "r" (to) : "memory"); 79 : : "r" (from), "r" (to) : "memory");
84 from+=64; 80
85 to+=64; 81 from += 64;
82 to += 64;
86 } 83 }
87 84
88 for(; i>0; i--) 85 for ( ; i > 0; i--) {
89 {
90 __asm__ __volatile__ ( 86 __asm__ __volatile__ (
91 " movq (%0), %%mm0\n" 87 " movq (%0), %%mm0\n"
92 " movq 8(%0), %%mm1\n" 88 " movq 8(%0), %%mm1\n"
@@ -104,17 +100,20 @@ void *_mmx_memcpy(void *to, const void *from, size_t len)
104 " movq %%mm1, 40(%1)\n" 100 " movq %%mm1, 40(%1)\n"
105 " movq %%mm2, 48(%1)\n" 101 " movq %%mm2, 48(%1)\n"
106 " movq %%mm3, 56(%1)\n" 102 " movq %%mm3, 56(%1)\n"
107 : : "r" (from), "r" (to) : "memory"); 103 : : "r" (from), "r" (to) : "memory");
108 from+=64; 104
109 to+=64; 105 from += 64;
106 to += 64;
110 } 107 }
111 /* 108 /*
112 * Now do the tail of the block 109 * Now do the tail of the block:
113 */ 110 */
114 __memcpy(to, from, len&63); 111 __memcpy(to, from, len & 63);
115 kernel_fpu_end(); 112 kernel_fpu_end();
113
116 return p; 114 return p;
117} 115}
116EXPORT_SYMBOL(_mmx_memcpy);
118 117
119#ifdef CONFIG_MK7 118#ifdef CONFIG_MK7
120 119
@@ -128,13 +127,12 @@ static void fast_clear_page(void *page)
128 int i; 127 int i;
129 128
130 kernel_fpu_begin(); 129 kernel_fpu_begin();
131 130
132 __asm__ __volatile__ ( 131 __asm__ __volatile__ (
133 " pxor %%mm0, %%mm0\n" : : 132 " pxor %%mm0, %%mm0\n" : :
134 ); 133 );
135 134
136 for(i=0;i<4096/64;i++) 135 for (i = 0; i < 4096/64; i++) {
137 {
138 __asm__ __volatile__ ( 136 __asm__ __volatile__ (
139 " movntq %%mm0, (%0)\n" 137 " movntq %%mm0, (%0)\n"
140 " movntq %%mm0, 8(%0)\n" 138 " movntq %%mm0, 8(%0)\n"
@@ -145,14 +143,15 @@ static void fast_clear_page(void *page)
145 " movntq %%mm0, 48(%0)\n" 143 " movntq %%mm0, 48(%0)\n"
146 " movntq %%mm0, 56(%0)\n" 144 " movntq %%mm0, 56(%0)\n"
147 : : "r" (page) : "memory"); 145 : : "r" (page) : "memory");
148 page+=64; 146 page += 64;
149 } 147 }
150 /* since movntq is weakly-ordered, a "sfence" is needed to become 148
151 * ordered again. 149 /*
150 * Since movntq is weakly-ordered, a "sfence" is needed to become
151 * ordered again:
152 */ 152 */
153 __asm__ __volatile__ ( 153 __asm__ __volatile__("sfence\n"::);
154 " sfence \n" : : 154
155 );
156 kernel_fpu_end(); 155 kernel_fpu_end();
157} 156}
158 157
@@ -162,10 +161,11 @@ static void fast_copy_page(void *to, void *from)
162 161
163 kernel_fpu_begin(); 162 kernel_fpu_begin();
164 163
165 /* maybe the prefetch stuff can go before the expensive fnsave... 164 /*
165 * maybe the prefetch stuff can go before the expensive fnsave...
166 * but that is for later. -AV 166 * but that is for later. -AV
167 */ 167 */
168 __asm__ __volatile__ ( 168 __asm__ __volatile__(
169 "1: prefetch (%0)\n" 169 "1: prefetch (%0)\n"
170 " prefetch 64(%0)\n" 170 " prefetch 64(%0)\n"
171 " prefetch 128(%0)\n" 171 " prefetch 128(%0)\n"
@@ -176,11 +176,9 @@ static void fast_copy_page(void *to, void *from)
176 "3: movw $0x1AEB, 1b\n" /* jmp on 26 bytes */ 176 "3: movw $0x1AEB, 1b\n" /* jmp on 26 bytes */
177 " jmp 2b\n" 177 " jmp 2b\n"
178 ".previous\n" 178 ".previous\n"
179 _ASM_EXTABLE(1b,3b) 179 _ASM_EXTABLE(1b, 3b) : : "r" (from));
180 : : "r" (from) );
181 180
182 for(i=0; i<(4096-320)/64; i++) 181 for (i = 0; i < (4096-320)/64; i++) {
183 {
184 __asm__ __volatile__ ( 182 __asm__ __volatile__ (
185 "1: prefetch 320(%0)\n" 183 "1: prefetch 320(%0)\n"
186 "2: movq (%0), %%mm0\n" 184 "2: movq (%0), %%mm0\n"
@@ -203,13 +201,13 @@ static void fast_copy_page(void *to, void *from)
203 "3: movw $0x05EB, 1b\n" /* jmp on 5 bytes */ 201 "3: movw $0x05EB, 1b\n" /* jmp on 5 bytes */
204 " jmp 2b\n" 202 " jmp 2b\n"
205 ".previous\n" 203 ".previous\n"
206 _ASM_EXTABLE(1b,3b) 204 _ASM_EXTABLE(1b, 3b) : : "r" (from), "r" (to) : "memory");
207 : : "r" (from), "r" (to) : "memory"); 205
208 from+=64; 206 from += 64;
209 to+=64; 207 to += 64;
210 } 208 }
211 for(i=(4096-320)/64; i<4096/64; i++) 209
212 { 210 for (i = (4096-320)/64; i < 4096/64; i++) {
213 __asm__ __volatile__ ( 211 __asm__ __volatile__ (
214 "2: movq (%0), %%mm0\n" 212 "2: movq (%0), %%mm0\n"
215 " movntq %%mm0, (%1)\n" 213 " movntq %%mm0, (%1)\n"
@@ -227,37 +225,34 @@ static void fast_copy_page(void *to, void *from)
227 " movntq %%mm6, 48(%1)\n" 225 " movntq %%mm6, 48(%1)\n"
228 " movq 56(%0), %%mm7\n" 226 " movq 56(%0), %%mm7\n"
229 " movntq %%mm7, 56(%1)\n" 227 " movntq %%mm7, 56(%1)\n"
230 : : "r" (from), "r" (to) : "memory"); 228 : : "r" (from), "r" (to) : "memory");
231 from+=64; 229 from += 64;
232 to+=64; 230 to += 64;
233 } 231 }
234 /* since movntq is weakly-ordered, a "sfence" is needed to become 232 /*
235 * ordered again. 233 * Since movntq is weakly-ordered, a "sfence" is needed to become
234 * ordered again:
236 */ 235 */
237 __asm__ __volatile__ ( 236 __asm__ __volatile__("sfence \n"::);
238 " sfence \n" : :
239 );
240 kernel_fpu_end(); 237 kernel_fpu_end();
241} 238}
242 239
243#else 240#else /* CONFIG_MK7 */
244 241
245/* 242/*
246 * Generic MMX implementation without K7 specific streaming 243 * Generic MMX implementation without K7 specific streaming
247 */ 244 */
248
249static void fast_clear_page(void *page) 245static void fast_clear_page(void *page)
250{ 246{
251 int i; 247 int i;
252 248
253 kernel_fpu_begin(); 249 kernel_fpu_begin();
254 250
255 __asm__ __volatile__ ( 251 __asm__ __volatile__ (
256 " pxor %%mm0, %%mm0\n" : : 252 " pxor %%mm0, %%mm0\n" : :
257 ); 253 );
258 254
259 for(i=0;i<4096/128;i++) 255 for (i = 0; i < 4096/128; i++) {
260 {
261 __asm__ __volatile__ ( 256 __asm__ __volatile__ (
262 " movq %%mm0, (%0)\n" 257 " movq %%mm0, (%0)\n"
263 " movq %%mm0, 8(%0)\n" 258 " movq %%mm0, 8(%0)\n"
@@ -275,8 +270,8 @@ static void fast_clear_page(void *page)
275 " movq %%mm0, 104(%0)\n" 270 " movq %%mm0, 104(%0)\n"
276 " movq %%mm0, 112(%0)\n" 271 " movq %%mm0, 112(%0)\n"
277 " movq %%mm0, 120(%0)\n" 272 " movq %%mm0, 120(%0)\n"
278 : : "r" (page) : "memory"); 273 : : "r" (page) : "memory");
279 page+=128; 274 page += 128;
280 } 275 }
281 276
282 kernel_fpu_end(); 277 kernel_fpu_end();
@@ -285,8 +280,7 @@ static void fast_clear_page(void *page)
285static void fast_copy_page(void *to, void *from) 280static void fast_copy_page(void *to, void *from)
286{ 281{
287 int i; 282 int i;
288 283
289
290 kernel_fpu_begin(); 284 kernel_fpu_begin();
291 285
292 __asm__ __volatile__ ( 286 __asm__ __volatile__ (
@@ -300,11 +294,9 @@ static void fast_copy_page(void *to, void *from)
300 "3: movw $0x1AEB, 1b\n" /* jmp on 26 bytes */ 294 "3: movw $0x1AEB, 1b\n" /* jmp on 26 bytes */
301 " jmp 2b\n" 295 " jmp 2b\n"
302 ".previous\n" 296 ".previous\n"
303 _ASM_EXTABLE(1b,3b) 297 _ASM_EXTABLE(1b, 3b) : : "r" (from));
304 : : "r" (from) );
305 298
306 for(i=0; i<4096/64; i++) 299 for (i = 0; i < 4096/64; i++) {
307 {
308 __asm__ __volatile__ ( 300 __asm__ __volatile__ (
309 "1: prefetch 320(%0)\n" 301 "1: prefetch 320(%0)\n"
310 "2: movq (%0), %%mm0\n" 302 "2: movq (%0), %%mm0\n"
@@ -327,60 +319,59 @@ static void fast_copy_page(void *to, void *from)
327 "3: movw $0x05EB, 1b\n" /* jmp on 5 bytes */ 319 "3: movw $0x05EB, 1b\n" /* jmp on 5 bytes */
328 " jmp 2b\n" 320 " jmp 2b\n"
329 ".previous\n" 321 ".previous\n"
330 _ASM_EXTABLE(1b,3b) 322 _ASM_EXTABLE(1b, 3b)
331 : : "r" (from), "r" (to) : "memory"); 323 : : "r" (from), "r" (to) : "memory");
332 from+=64; 324
333 to+=64; 325 from += 64;
326 to += 64;
334 } 327 }
335 kernel_fpu_end(); 328 kernel_fpu_end();
336} 329}
337 330
338 331#endif /* !CONFIG_MK7 */
339#endif
340 332
341/* 333/*
342 * Favour MMX for page clear and copy. 334 * Favour MMX for page clear and copy:
343 */ 335 */
344 336static void slow_zero_page(void *page)
345static void slow_zero_page(void * page)
346{ 337{
347 int d0, d1; 338 int d0, d1;
348 __asm__ __volatile__( \ 339
349 "cld\n\t" \ 340 __asm__ __volatile__(
350 "rep ; stosl" \ 341 "cld\n\t"
351 : "=&c" (d0), "=&D" (d1) 342 "rep ; stosl"
352 :"a" (0),"1" (page),"0" (1024) 343
353 :"memory"); 344 : "=&c" (d0), "=&D" (d1)
345 :"a" (0), "1" (page), "0" (1024)
346 :"memory");
354} 347}
355 348
356void mmx_clear_page(void * page) 349void mmx_clear_page(void *page)
357{ 350{
358 if(unlikely(in_interrupt())) 351 if (unlikely(in_interrupt()))
359 slow_zero_page(page); 352 slow_zero_page(page);
360 else 353 else
361 fast_clear_page(page); 354 fast_clear_page(page);
362} 355}
356EXPORT_SYMBOL(mmx_clear_page);
363 357
364static void slow_copy_page(void *to, void *from) 358static void slow_copy_page(void *to, void *from)
365{ 359{
366 int d0, d1, d2; 360 int d0, d1, d2;
367 __asm__ __volatile__( \ 361
368 "cld\n\t" \ 362 __asm__ __volatile__(
369 "rep ; movsl" \ 363 "cld\n\t"
370 : "=&c" (d0), "=&D" (d1), "=&S" (d2) \ 364 "rep ; movsl"
371 : "0" (1024),"1" ((long) to),"2" ((long) from) \ 365 : "=&c" (d0), "=&D" (d1), "=&S" (d2)
366 : "0" (1024), "1" ((long) to), "2" ((long) from)
372 : "memory"); 367 : "memory");
373} 368}
374
375 369
376void mmx_copy_page(void *to, void *from) 370void mmx_copy_page(void *to, void *from)
377{ 371{
378 if(unlikely(in_interrupt())) 372 if (unlikely(in_interrupt()))
379 slow_copy_page(to, from); 373 slow_copy_page(to, from);
380 else 374 else
381 fast_copy_page(to, from); 375 fast_copy_page(to, from);
382} 376}
383
384EXPORT_SYMBOL(_mmx_memcpy);
385EXPORT_SYMBOL(mmx_clear_page);
386EXPORT_SYMBOL(mmx_copy_page); 377EXPORT_SYMBOL(mmx_copy_page);
diff --git a/arch/x86/lib/semaphore_32.S b/arch/x86/lib/semaphore_32.S
index 3899bd37fdf0..648fe4741782 100644
--- a/arch/x86/lib/semaphore_32.S
+++ b/arch/x86/lib/semaphore_32.S
@@ -30,89 +30,6 @@
30 * value or just clobbered.. 30 * value or just clobbered..
31 */ 31 */
32 .section .sched.text, "ax" 32 .section .sched.text, "ax"
33ENTRY(__down_failed)
34 CFI_STARTPROC
35 FRAME
36 pushl %edx
37 CFI_ADJUST_CFA_OFFSET 4
38 CFI_REL_OFFSET edx,0
39 pushl %ecx
40 CFI_ADJUST_CFA_OFFSET 4
41 CFI_REL_OFFSET ecx,0
42 call __down
43 popl %ecx
44 CFI_ADJUST_CFA_OFFSET -4
45 CFI_RESTORE ecx
46 popl %edx
47 CFI_ADJUST_CFA_OFFSET -4
48 CFI_RESTORE edx
49 ENDFRAME
50 ret
51 CFI_ENDPROC
52 ENDPROC(__down_failed)
53
54ENTRY(__down_failed_interruptible)
55 CFI_STARTPROC
56 FRAME
57 pushl %edx
58 CFI_ADJUST_CFA_OFFSET 4
59 CFI_REL_OFFSET edx,0
60 pushl %ecx
61 CFI_ADJUST_CFA_OFFSET 4
62 CFI_REL_OFFSET ecx,0
63 call __down_interruptible
64 popl %ecx
65 CFI_ADJUST_CFA_OFFSET -4
66 CFI_RESTORE ecx
67 popl %edx
68 CFI_ADJUST_CFA_OFFSET -4
69 CFI_RESTORE edx
70 ENDFRAME
71 ret
72 CFI_ENDPROC
73 ENDPROC(__down_failed_interruptible)
74
75ENTRY(__down_failed_trylock)
76 CFI_STARTPROC
77 FRAME
78 pushl %edx
79 CFI_ADJUST_CFA_OFFSET 4
80 CFI_REL_OFFSET edx,0
81 pushl %ecx
82 CFI_ADJUST_CFA_OFFSET 4
83 CFI_REL_OFFSET ecx,0
84 call __down_trylock
85 popl %ecx
86 CFI_ADJUST_CFA_OFFSET -4
87 CFI_RESTORE ecx
88 popl %edx
89 CFI_ADJUST_CFA_OFFSET -4
90 CFI_RESTORE edx
91 ENDFRAME
92 ret
93 CFI_ENDPROC
94 ENDPROC(__down_failed_trylock)
95
96ENTRY(__up_wakeup)
97 CFI_STARTPROC
98 FRAME
99 pushl %edx
100 CFI_ADJUST_CFA_OFFSET 4
101 CFI_REL_OFFSET edx,0
102 pushl %ecx
103 CFI_ADJUST_CFA_OFFSET 4
104 CFI_REL_OFFSET ecx,0
105 call __up
106 popl %ecx
107 CFI_ADJUST_CFA_OFFSET -4
108 CFI_RESTORE ecx
109 popl %edx
110 CFI_ADJUST_CFA_OFFSET -4
111 CFI_RESTORE edx
112 ENDFRAME
113 ret
114 CFI_ENDPROC
115 ENDPROC(__up_wakeup)
116 33
117/* 34/*
118 * rw spinlock fallbacks 35 * rw spinlock fallbacks
diff --git a/arch/x86/lib/string_32.c b/arch/x86/lib/string_32.c
index c2c0504a3071..94972e7c094d 100644
--- a/arch/x86/lib/string_32.c
+++ b/arch/x86/lib/string_32.c
@@ -14,25 +14,25 @@
14#include <linux/module.h> 14#include <linux/module.h>
15 15
16#ifdef __HAVE_ARCH_STRCPY 16#ifdef __HAVE_ARCH_STRCPY
17char *strcpy(char * dest,const char *src) 17char *strcpy(char *dest, const char *src)
18{ 18{
19 int d0, d1, d2; 19 int d0, d1, d2;
20 asm volatile( "1:\tlodsb\n\t" 20 asm volatile("1:\tlodsb\n\t"
21 "stosb\n\t" 21 "stosb\n\t"
22 "testb %%al,%%al\n\t" 22 "testb %%al,%%al\n\t"
23 "jne 1b" 23 "jne 1b"
24 : "=&S" (d0), "=&D" (d1), "=&a" (d2) 24 : "=&S" (d0), "=&D" (d1), "=&a" (d2)
25 :"0" (src),"1" (dest) : "memory"); 25 :"0" (src), "1" (dest) : "memory");
26 return dest; 26 return dest;
27} 27}
28EXPORT_SYMBOL(strcpy); 28EXPORT_SYMBOL(strcpy);
29#endif 29#endif
30 30
31#ifdef __HAVE_ARCH_STRNCPY 31#ifdef __HAVE_ARCH_STRNCPY
32char *strncpy(char * dest,const char *src,size_t count) 32char *strncpy(char *dest, const char *src, size_t count)
33{ 33{
34 int d0, d1, d2, d3; 34 int d0, d1, d2, d3;
35 asm volatile( "1:\tdecl %2\n\t" 35 asm volatile("1:\tdecl %2\n\t"
36 "js 2f\n\t" 36 "js 2f\n\t"
37 "lodsb\n\t" 37 "lodsb\n\t"
38 "stosb\n\t" 38 "stosb\n\t"
@@ -42,17 +42,17 @@ char *strncpy(char * dest,const char *src,size_t count)
42 "stosb\n" 42 "stosb\n"
43 "2:" 43 "2:"
44 : "=&S" (d0), "=&D" (d1), "=&c" (d2), "=&a" (d3) 44 : "=&S" (d0), "=&D" (d1), "=&c" (d2), "=&a" (d3)
45 :"0" (src),"1" (dest),"2" (count) : "memory"); 45 :"0" (src), "1" (dest), "2" (count) : "memory");
46 return dest; 46 return dest;
47} 47}
48EXPORT_SYMBOL(strncpy); 48EXPORT_SYMBOL(strncpy);
49#endif 49#endif
50 50
51#ifdef __HAVE_ARCH_STRCAT 51#ifdef __HAVE_ARCH_STRCAT
52char *strcat(char * dest,const char * src) 52char *strcat(char *dest, const char *src)
53{ 53{
54 int d0, d1, d2, d3; 54 int d0, d1, d2, d3;
55 asm volatile( "repne\n\t" 55 asm volatile("repne\n\t"
56 "scasb\n\t" 56 "scasb\n\t"
57 "decl %1\n" 57 "decl %1\n"
58 "1:\tlodsb\n\t" 58 "1:\tlodsb\n\t"
@@ -67,10 +67,10 @@ EXPORT_SYMBOL(strcat);
67#endif 67#endif
68 68
69#ifdef __HAVE_ARCH_STRNCAT 69#ifdef __HAVE_ARCH_STRNCAT
70char *strncat(char * dest,const char * src,size_t count) 70char *strncat(char *dest, const char *src, size_t count)
71{ 71{
72 int d0, d1, d2, d3; 72 int d0, d1, d2, d3;
73 asm volatile( "repne\n\t" 73 asm volatile("repne\n\t"
74 "scasb\n\t" 74 "scasb\n\t"
75 "decl %1\n\t" 75 "decl %1\n\t"
76 "movl %8,%3\n" 76 "movl %8,%3\n"
@@ -83,7 +83,7 @@ char *strncat(char * dest,const char * src,size_t count)
83 "2:\txorl %2,%2\n\t" 83 "2:\txorl %2,%2\n\t"
84 "stosb" 84 "stosb"
85 : "=&S" (d0), "=&D" (d1), "=&a" (d2), "=&c" (d3) 85 : "=&S" (d0), "=&D" (d1), "=&a" (d2), "=&c" (d3)
86 : "0" (src),"1" (dest),"2" (0),"3" (0xffffffffu), "g" (count) 86 : "0" (src), "1" (dest), "2" (0), "3" (0xffffffffu), "g" (count)
87 : "memory"); 87 : "memory");
88 return dest; 88 return dest;
89} 89}
@@ -91,11 +91,11 @@ EXPORT_SYMBOL(strncat);
91#endif 91#endif
92 92
93#ifdef __HAVE_ARCH_STRCMP 93#ifdef __HAVE_ARCH_STRCMP
94int strcmp(const char * cs,const char * ct) 94int strcmp(const char *cs, const char *ct)
95{ 95{
96 int d0, d1; 96 int d0, d1;
97 int res; 97 int res;
98 asm volatile( "1:\tlodsb\n\t" 98 asm volatile("1:\tlodsb\n\t"
99 "scasb\n\t" 99 "scasb\n\t"
100 "jne 2f\n\t" 100 "jne 2f\n\t"
101 "testb %%al,%%al\n\t" 101 "testb %%al,%%al\n\t"
@@ -106,7 +106,7 @@ int strcmp(const char * cs,const char * ct)
106 "orb $1,%%al\n" 106 "orb $1,%%al\n"
107 "3:" 107 "3:"
108 :"=a" (res), "=&S" (d0), "=&D" (d1) 108 :"=a" (res), "=&S" (d0), "=&D" (d1)
109 :"1" (cs),"2" (ct) 109 :"1" (cs), "2" (ct)
110 :"memory"); 110 :"memory");
111 return res; 111 return res;
112} 112}
@@ -114,11 +114,11 @@ EXPORT_SYMBOL(strcmp);
114#endif 114#endif
115 115
116#ifdef __HAVE_ARCH_STRNCMP 116#ifdef __HAVE_ARCH_STRNCMP
117int strncmp(const char * cs,const char * ct,size_t count) 117int strncmp(const char *cs, const char *ct, size_t count)
118{ 118{
119 int res; 119 int res;
120 int d0, d1, d2; 120 int d0, d1, d2;
121 asm volatile( "1:\tdecl %3\n\t" 121 asm volatile("1:\tdecl %3\n\t"
122 "js 2f\n\t" 122 "js 2f\n\t"
123 "lodsb\n\t" 123 "lodsb\n\t"
124 "scasb\n\t" 124 "scasb\n\t"
@@ -131,7 +131,7 @@ int strncmp(const char * cs,const char * ct,size_t count)
131 "orb $1,%%al\n" 131 "orb $1,%%al\n"
132 "4:" 132 "4:"
133 :"=a" (res), "=&S" (d0), "=&D" (d1), "=&c" (d2) 133 :"=a" (res), "=&S" (d0), "=&D" (d1), "=&c" (d2)
134 :"1" (cs),"2" (ct),"3" (count) 134 :"1" (cs), "2" (ct), "3" (count)
135 :"memory"); 135 :"memory");
136 return res; 136 return res;
137} 137}
@@ -139,11 +139,11 @@ EXPORT_SYMBOL(strncmp);
139#endif 139#endif
140 140
141#ifdef __HAVE_ARCH_STRCHR 141#ifdef __HAVE_ARCH_STRCHR
142char *strchr(const char * s, int c) 142char *strchr(const char *s, int c)
143{ 143{
144 int d0; 144 int d0;
145 char * res; 145 char *res;
146 asm volatile( "movb %%al,%%ah\n" 146 asm volatile("movb %%al,%%ah\n"
147 "1:\tlodsb\n\t" 147 "1:\tlodsb\n\t"
148 "cmpb %%ah,%%al\n\t" 148 "cmpb %%ah,%%al\n\t"
149 "je 2f\n\t" 149 "je 2f\n\t"
@@ -153,7 +153,7 @@ char *strchr(const char * s, int c)
153 "2:\tmovl %1,%0\n\t" 153 "2:\tmovl %1,%0\n\t"
154 "decl %0" 154 "decl %0"
155 :"=a" (res), "=&S" (d0) 155 :"=a" (res), "=&S" (d0)
156 :"1" (s),"0" (c) 156 :"1" (s), "0" (c)
157 :"memory"); 157 :"memory");
158 return res; 158 return res;
159} 159}
@@ -161,16 +161,16 @@ EXPORT_SYMBOL(strchr);
161#endif 161#endif
162 162
163#ifdef __HAVE_ARCH_STRLEN 163#ifdef __HAVE_ARCH_STRLEN
164size_t strlen(const char * s) 164size_t strlen(const char *s)
165{ 165{
166 int d0; 166 int d0;
167 int res; 167 int res;
168 asm volatile( "repne\n\t" 168 asm volatile("repne\n\t"
169 "scasb\n\t" 169 "scasb\n\t"
170 "notl %0\n\t" 170 "notl %0\n\t"
171 "decl %0" 171 "decl %0"
172 :"=c" (res), "=&D" (d0) 172 :"=c" (res), "=&D" (d0)
173 :"1" (s),"a" (0), "0" (0xffffffffu) 173 :"1" (s), "a" (0), "0" (0xffffffffu)
174 :"memory"); 174 :"memory");
175 return res; 175 return res;
176} 176}
@@ -178,19 +178,19 @@ EXPORT_SYMBOL(strlen);
178#endif 178#endif
179 179
180#ifdef __HAVE_ARCH_MEMCHR 180#ifdef __HAVE_ARCH_MEMCHR
181void *memchr(const void *cs,int c,size_t count) 181void *memchr(const void *cs, int c, size_t count)
182{ 182{
183 int d0; 183 int d0;
184 void *res; 184 void *res;
185 if (!count) 185 if (!count)
186 return NULL; 186 return NULL;
187 asm volatile( "repne\n\t" 187 asm volatile("repne\n\t"
188 "scasb\n\t" 188 "scasb\n\t"
189 "je 1f\n\t" 189 "je 1f\n\t"
190 "movl $1,%0\n" 190 "movl $1,%0\n"
191 "1:\tdecl %0" 191 "1:\tdecl %0"
192 :"=D" (res), "=&c" (d0) 192 :"=D" (res), "=&c" (d0)
193 :"a" (c),"0" (cs),"1" (count) 193 :"a" (c), "0" (cs), "1" (count)
194 :"memory"); 194 :"memory");
195 return res; 195 return res;
196} 196}
@@ -198,7 +198,7 @@ EXPORT_SYMBOL(memchr);
198#endif 198#endif
199 199
200#ifdef __HAVE_ARCH_MEMSCAN 200#ifdef __HAVE_ARCH_MEMSCAN
201void *memscan(void * addr, int c, size_t size) 201void *memscan(void *addr, int c, size_t size)
202{ 202{
203 if (!size) 203 if (!size)
204 return addr; 204 return addr;
@@ -219,7 +219,7 @@ size_t strnlen(const char *s, size_t count)
219{ 219{
220 int d0; 220 int d0;
221 int res; 221 int res;
222 asm volatile( "movl %2,%0\n\t" 222 asm volatile("movl %2,%0\n\t"
223 "jmp 2f\n" 223 "jmp 2f\n"
224 "1:\tcmpb $0,(%0)\n\t" 224 "1:\tcmpb $0,(%0)\n\t"
225 "je 3f\n\t" 225 "je 3f\n\t"
@@ -229,7 +229,7 @@ size_t strnlen(const char *s, size_t count)
229 "jne 1b\n" 229 "jne 1b\n"
230 "3:\tsubl %2,%0" 230 "3:\tsubl %2,%0"
231 :"=a" (res), "=&d" (d0) 231 :"=a" (res), "=&d" (d0)
232 :"c" (s),"1" (count) 232 :"c" (s), "1" (count)
233 :"memory"); 233 :"memory");
234 return res; 234 return res;
235} 235}
diff --git a/arch/x86/lib/strstr_32.c b/arch/x86/lib/strstr_32.c
index a3dafbf59dae..42e8a50303f3 100644
--- a/arch/x86/lib/strstr_32.c
+++ b/arch/x86/lib/strstr_32.c
@@ -1,9 +1,9 @@
1#include <linux/string.h> 1#include <linux/string.h>
2 2
3char * strstr(const char * cs,const char * ct) 3char *strstr(const char *cs, const char *ct)
4{ 4{
5int d0, d1; 5int d0, d1;
6register char * __res; 6register char *__res;
7__asm__ __volatile__( 7__asm__ __volatile__(
8 "movl %6,%%edi\n\t" 8 "movl %6,%%edi\n\t"
9 "repne\n\t" 9 "repne\n\t"
diff --git a/arch/x86/lib/thunk_64.S b/arch/x86/lib/thunk_64.S
index 8b92d428ab02..e009251d4e9f 100644
--- a/arch/x86/lib/thunk_64.S
+++ b/arch/x86/lib/thunk_64.S
@@ -41,11 +41,6 @@
41 thunk rwsem_downgrade_thunk,rwsem_downgrade_wake 41 thunk rwsem_downgrade_thunk,rwsem_downgrade_wake
42#endif 42#endif
43 43
44 thunk __down_failed,__down
45 thunk_retrax __down_failed_interruptible,__down_interruptible
46 thunk_retrax __down_failed_trylock,__down_trylock
47 thunk __up_wakeup,__up
48
49#ifdef CONFIG_TRACE_IRQFLAGS 44#ifdef CONFIG_TRACE_IRQFLAGS
50 thunk trace_hardirqs_on_thunk,trace_hardirqs_on 45 thunk trace_hardirqs_on_thunk,trace_hardirqs_on
51 thunk trace_hardirqs_off_thunk,trace_hardirqs_off 46 thunk trace_hardirqs_off_thunk,trace_hardirqs_off
diff --git a/arch/x86/lib/usercopy_32.c b/arch/x86/lib/usercopy_32.c
index e849b9998b0e..24e60944971a 100644
--- a/arch/x86/lib/usercopy_32.c
+++ b/arch/x86/lib/usercopy_32.c
@@ -1,4 +1,4 @@
1/* 1/*
2 * User address space access functions. 2 * User address space access functions.
3 * The non inlined parts of asm-i386/uaccess.h are here. 3 * The non inlined parts of asm-i386/uaccess.h are here.
4 * 4 *
@@ -22,14 +22,14 @@ static inline int __movsl_is_ok(unsigned long a1, unsigned long a2, unsigned lon
22#endif 22#endif
23 return 1; 23 return 1;
24} 24}
25#define movsl_is_ok(a1,a2,n) \ 25#define movsl_is_ok(a1, a2, n) \
26 __movsl_is_ok((unsigned long)(a1),(unsigned long)(a2),(n)) 26 __movsl_is_ok((unsigned long)(a1), (unsigned long)(a2), (n))
27 27
28/* 28/*
29 * Copy a null terminated string from userspace. 29 * Copy a null terminated string from userspace.
30 */ 30 */
31 31
32#define __do_strncpy_from_user(dst,src,count,res) \ 32#define __do_strncpy_from_user(dst, src, count, res) \
33do { \ 33do { \
34 int __d0, __d1, __d2; \ 34 int __d0, __d1, __d2; \
35 might_sleep(); \ 35 might_sleep(); \
@@ -61,7 +61,7 @@ do { \
61 * least @count bytes long. 61 * least @count bytes long.
62 * @src: Source address, in user space. 62 * @src: Source address, in user space.
63 * @count: Maximum number of bytes to copy, including the trailing NUL. 63 * @count: Maximum number of bytes to copy, including the trailing NUL.
64 * 64 *
65 * Copies a NUL-terminated string from userspace to kernel space. 65 * Copies a NUL-terminated string from userspace to kernel space.
66 * Caller must check the specified block with access_ok() before calling 66 * Caller must check the specified block with access_ok() before calling
67 * this function. 67 * this function.
@@ -90,7 +90,7 @@ EXPORT_SYMBOL(__strncpy_from_user);
90 * least @count bytes long. 90 * least @count bytes long.
91 * @src: Source address, in user space. 91 * @src: Source address, in user space.
92 * @count: Maximum number of bytes to copy, including the trailing NUL. 92 * @count: Maximum number of bytes to copy, including the trailing NUL.
93 * 93 *
94 * Copies a NUL-terminated string from userspace to kernel space. 94 * Copies a NUL-terminated string from userspace to kernel space.
95 * 95 *
96 * On success, returns the length of the string (not including the trailing 96 * On success, returns the length of the string (not including the trailing
@@ -120,7 +120,7 @@ EXPORT_SYMBOL(strncpy_from_user);
120do { \ 120do { \
121 int __d0; \ 121 int __d0; \
122 might_sleep(); \ 122 might_sleep(); \
123 __asm__ __volatile__( \ 123 __asm__ __volatile__( \
124 "0: rep; stosl\n" \ 124 "0: rep; stosl\n" \
125 " movl %2,%0\n" \ 125 " movl %2,%0\n" \
126 "1: rep; stosb\n" \ 126 "1: rep; stosb\n" \
@@ -333,17 +333,17 @@ __copy_user_zeroing_intel(void *to, const void __user *from, unsigned long size)
333 __asm__ __volatile__( 333 __asm__ __volatile__(
334 " .align 2,0x90\n" 334 " .align 2,0x90\n"
335 "0: movl 32(%4), %%eax\n" 335 "0: movl 32(%4), %%eax\n"
336 " cmpl $67, %0\n" 336 " cmpl $67, %0\n"
337 " jbe 2f\n" 337 " jbe 2f\n"
338 "1: movl 64(%4), %%eax\n" 338 "1: movl 64(%4), %%eax\n"
339 " .align 2,0x90\n" 339 " .align 2,0x90\n"
340 "2: movl 0(%4), %%eax\n" 340 "2: movl 0(%4), %%eax\n"
341 "21: movl 4(%4), %%edx\n" 341 "21: movl 4(%4), %%edx\n"
342 " movl %%eax, 0(%3)\n" 342 " movl %%eax, 0(%3)\n"
343 " movl %%edx, 4(%3)\n" 343 " movl %%edx, 4(%3)\n"
344 "3: movl 8(%4), %%eax\n" 344 "3: movl 8(%4), %%eax\n"
345 "31: movl 12(%4),%%edx\n" 345 "31: movl 12(%4),%%edx\n"
346 " movl %%eax, 8(%3)\n" 346 " movl %%eax, 8(%3)\n"
347 " movl %%edx, 12(%3)\n" 347 " movl %%edx, 12(%3)\n"
348 "4: movl 16(%4), %%eax\n" 348 "4: movl 16(%4), %%eax\n"
349 "41: movl 20(%4), %%edx\n" 349 "41: movl 20(%4), %%edx\n"
@@ -369,38 +369,38 @@ __copy_user_zeroing_intel(void *to, const void __user *from, unsigned long size)
369 "91: movl 60(%4), %%edx\n" 369 "91: movl 60(%4), %%edx\n"
370 " movl %%eax, 56(%3)\n" 370 " movl %%eax, 56(%3)\n"
371 " movl %%edx, 60(%3)\n" 371 " movl %%edx, 60(%3)\n"
372 " addl $-64, %0\n" 372 " addl $-64, %0\n"
373 " addl $64, %4\n" 373 " addl $64, %4\n"
374 " addl $64, %3\n" 374 " addl $64, %3\n"
375 " cmpl $63, %0\n" 375 " cmpl $63, %0\n"
376 " ja 0b\n" 376 " ja 0b\n"
377 "5: movl %0, %%eax\n" 377 "5: movl %0, %%eax\n"
378 " shrl $2, %0\n" 378 " shrl $2, %0\n"
379 " andl $3, %%eax\n" 379 " andl $3, %%eax\n"
380 " cld\n" 380 " cld\n"
381 "6: rep; movsl\n" 381 "6: rep; movsl\n"
382 " movl %%eax,%0\n" 382 " movl %%eax,%0\n"
383 "7: rep; movsb\n" 383 "7: rep; movsb\n"
384 "8:\n" 384 "8:\n"
385 ".section .fixup,\"ax\"\n" 385 ".section .fixup,\"ax\"\n"
386 "9: lea 0(%%eax,%0,4),%0\n" 386 "9: lea 0(%%eax,%0,4),%0\n"
387 "16: pushl %0\n" 387 "16: pushl %0\n"
388 " pushl %%eax\n" 388 " pushl %%eax\n"
389 " xorl %%eax,%%eax\n" 389 " xorl %%eax,%%eax\n"
390 " rep; stosb\n" 390 " rep; stosb\n"
391 " popl %%eax\n" 391 " popl %%eax\n"
392 " popl %0\n" 392 " popl %0\n"
393 " jmp 8b\n" 393 " jmp 8b\n"
394 ".previous\n" 394 ".previous\n"
395 ".section __ex_table,\"a\"\n" 395 ".section __ex_table,\"a\"\n"
396 " .align 4\n" 396 " .align 4\n"
397 " .long 0b,16b\n" 397 " .long 0b,16b\n"
398 " .long 1b,16b\n" 398 " .long 1b,16b\n"
399 " .long 2b,16b\n" 399 " .long 2b,16b\n"
400 " .long 21b,16b\n" 400 " .long 21b,16b\n"
401 " .long 3b,16b\n" 401 " .long 3b,16b\n"
402 " .long 31b,16b\n" 402 " .long 31b,16b\n"
403 " .long 4b,16b\n" 403 " .long 4b,16b\n"
404 " .long 41b,16b\n" 404 " .long 41b,16b\n"
405 " .long 10b,16b\n" 405 " .long 10b,16b\n"
406 " .long 51b,16b\n" 406 " .long 51b,16b\n"
@@ -412,9 +412,9 @@ __copy_user_zeroing_intel(void *to, const void __user *from, unsigned long size)
412 " .long 81b,16b\n" 412 " .long 81b,16b\n"
413 " .long 14b,16b\n" 413 " .long 14b,16b\n"
414 " .long 91b,16b\n" 414 " .long 91b,16b\n"
415 " .long 6b,9b\n" 415 " .long 6b,9b\n"
416 " .long 7b,16b\n" 416 " .long 7b,16b\n"
417 ".previous" 417 ".previous"
418 : "=&c"(size), "=&D" (d0), "=&S" (d1) 418 : "=&c"(size), "=&D" (d0), "=&S" (d1)
419 : "1"(to), "2"(from), "0"(size) 419 : "1"(to), "2"(from), "0"(size)
420 : "eax", "edx", "memory"); 420 : "eax", "edx", "memory");
@@ -429,7 +429,7 @@ __copy_user_zeroing_intel(void *to, const void __user *from, unsigned long size)
429static unsigned long __copy_user_zeroing_intel_nocache(void *to, 429static unsigned long __copy_user_zeroing_intel_nocache(void *to,
430 const void __user *from, unsigned long size) 430 const void __user *from, unsigned long size)
431{ 431{
432 int d0, d1; 432 int d0, d1;
433 433
434 __asm__ __volatile__( 434 __asm__ __volatile__(
435 " .align 2,0x90\n" 435 " .align 2,0x90\n"
@@ -526,7 +526,7 @@ static unsigned long __copy_user_zeroing_intel_nocache(void *to,
526static unsigned long __copy_user_intel_nocache(void *to, 526static unsigned long __copy_user_intel_nocache(void *to,
527 const void __user *from, unsigned long size) 527 const void __user *from, unsigned long size)
528{ 528{
529 int d0, d1; 529 int d0, d1;
530 530
531 __asm__ __volatile__( 531 __asm__ __volatile__(
532 " .align 2,0x90\n" 532 " .align 2,0x90\n"
@@ -629,7 +629,7 @@ unsigned long __copy_user_zeroing_intel_nocache(void *to,
629#endif /* CONFIG_X86_INTEL_USERCOPY */ 629#endif /* CONFIG_X86_INTEL_USERCOPY */
630 630
631/* Generic arbitrary sized copy. */ 631/* Generic arbitrary sized copy. */
632#define __copy_user(to,from,size) \ 632#define __copy_user(to, from, size) \
633do { \ 633do { \
634 int __d0, __d1, __d2; \ 634 int __d0, __d1, __d2; \
635 __asm__ __volatile__( \ 635 __asm__ __volatile__( \
@@ -665,7 +665,7 @@ do { \
665 : "memory"); \ 665 : "memory"); \
666} while (0) 666} while (0)
667 667
668#define __copy_user_zeroing(to,from,size) \ 668#define __copy_user_zeroing(to, from, size) \
669do { \ 669do { \
670 int __d0, __d1, __d2; \ 670 int __d0, __d1, __d2; \
671 __asm__ __volatile__( \ 671 __asm__ __volatile__( \
@@ -712,7 +712,7 @@ unsigned long __copy_to_user_ll(void __user *to, const void *from,
712{ 712{
713#ifndef CONFIG_X86_WP_WORKS_OK 713#ifndef CONFIG_X86_WP_WORKS_OK
714 if (unlikely(boot_cpu_data.wp_works_ok == 0) && 714 if (unlikely(boot_cpu_data.wp_works_ok == 0) &&
715 ((unsigned long )to) < TASK_SIZE) { 715 ((unsigned long)to) < TASK_SIZE) {
716 /* 716 /*
717 * When we are in an atomic section (see 717 * When we are in an atomic section (see
718 * mm/filemap.c:file_read_actor), return the full 718 * mm/filemap.c:file_read_actor), return the full
@@ -721,26 +721,26 @@ unsigned long __copy_to_user_ll(void __user *to, const void *from,
721 if (in_atomic()) 721 if (in_atomic())
722 return n; 722 return n;
723 723
724 /* 724 /*
725 * CPU does not honor the WP bit when writing 725 * CPU does not honor the WP bit when writing
726 * from supervisory mode, and due to preemption or SMP, 726 * from supervisory mode, and due to preemption or SMP,
727 * the page tables can change at any time. 727 * the page tables can change at any time.
728 * Do it manually. Manfred <manfred@colorfullife.com> 728 * Do it manually. Manfred <manfred@colorfullife.com>
729 */ 729 */
730 while (n) { 730 while (n) {
731 unsigned long offset = ((unsigned long)to)%PAGE_SIZE; 731 unsigned long offset = ((unsigned long)to)%PAGE_SIZE;
732 unsigned long len = PAGE_SIZE - offset; 732 unsigned long len = PAGE_SIZE - offset;
733 int retval; 733 int retval;
734 struct page *pg; 734 struct page *pg;
735 void *maddr; 735 void *maddr;
736 736
737 if (len > n) 737 if (len > n)
738 len = n; 738 len = n;
739 739
740survive: 740survive:
741 down_read(&current->mm->mmap_sem); 741 down_read(&current->mm->mmap_sem);
742 retval = get_user_pages(current, current->mm, 742 retval = get_user_pages(current, current->mm,
743 (unsigned long )to, 1, 1, 0, &pg, NULL); 743 (unsigned long)to, 1, 1, 0, &pg, NULL);
744 744
745 if (retval == -ENOMEM && is_global_init(current)) { 745 if (retval == -ENOMEM && is_global_init(current)) {
746 up_read(&current->mm->mmap_sem); 746 up_read(&current->mm->mmap_sem);
@@ -750,8 +750,8 @@ survive:
750 750
751 if (retval != 1) { 751 if (retval != 1) {
752 up_read(&current->mm->mmap_sem); 752 up_read(&current->mm->mmap_sem);
753 break; 753 break;
754 } 754 }
755 755
756 maddr = kmap_atomic(pg, KM_USER0); 756 maddr = kmap_atomic(pg, KM_USER0);
757 memcpy(maddr + offset, from, len); 757 memcpy(maddr + offset, from, len);
@@ -802,12 +802,12 @@ unsigned long __copy_from_user_ll_nocache(void *to, const void __user *from,
802 unsigned long n) 802 unsigned long n)
803{ 803{
804#ifdef CONFIG_X86_INTEL_USERCOPY 804#ifdef CONFIG_X86_INTEL_USERCOPY
805 if ( n > 64 && cpu_has_xmm2) 805 if (n > 64 && cpu_has_xmm2)
806 n = __copy_user_zeroing_intel_nocache(to, from, n); 806 n = __copy_user_zeroing_intel_nocache(to, from, n);
807 else 807 else
808 __copy_user_zeroing(to, from, n); 808 __copy_user_zeroing(to, from, n);
809#else 809#else
810 __copy_user_zeroing(to, from, n); 810 __copy_user_zeroing(to, from, n);
811#endif 811#endif
812 return n; 812 return n;
813} 813}
@@ -817,12 +817,12 @@ unsigned long __copy_from_user_ll_nocache_nozero(void *to, const void __user *fr
817 unsigned long n) 817 unsigned long n)
818{ 818{
819#ifdef CONFIG_X86_INTEL_USERCOPY 819#ifdef CONFIG_X86_INTEL_USERCOPY
820 if ( n > 64 && cpu_has_xmm2) 820 if (n > 64 && cpu_has_xmm2)
821 n = __copy_user_intel_nocache(to, from, n); 821 n = __copy_user_intel_nocache(to, from, n);
822 else 822 else
823 __copy_user(to, from, n); 823 __copy_user(to, from, n);
824#else 824#else
825 __copy_user(to, from, n); 825 __copy_user(to, from, n);
826#endif 826#endif
827 return n; 827 return n;
828} 828}
diff --git a/arch/x86/mach-generic/bigsmp.c b/arch/x86/mach-generic/bigsmp.c
index 292a225edabe..95fc463056d0 100644
--- a/arch/x86/mach-generic/bigsmp.c
+++ b/arch/x86/mach-generic/bigsmp.c
@@ -1,4 +1,4 @@
1/* 1/*
2 * APIC driver for "bigsmp" XAPIC machines with more than 8 virtual CPUs. 2 * APIC driver for "bigsmp" XAPIC machines with more than 8 virtual CPUs.
3 * Drives the local APIC in "clustered mode". 3 * Drives the local APIC in "clustered mode".
4 */ 4 */
@@ -32,26 +32,26 @@ static int hp_ht_bigsmp(const struct dmi_system_id *d)
32 32
33 33
34static const struct dmi_system_id bigsmp_dmi_table[] = { 34static const struct dmi_system_id bigsmp_dmi_table[] = {
35 { hp_ht_bigsmp, "HP ProLiant DL760 G2", { 35 { hp_ht_bigsmp, "HP ProLiant DL760 G2",
36 DMI_MATCH(DMI_BIOS_VENDOR, "HP"), 36 { DMI_MATCH(DMI_BIOS_VENDOR, "HP"),
37 DMI_MATCH(DMI_BIOS_VERSION, "P44-"), 37 DMI_MATCH(DMI_BIOS_VERSION, "P44-"),}
38 }}, 38 },
39 39
40 { hp_ht_bigsmp, "HP ProLiant DL740", { 40 { hp_ht_bigsmp, "HP ProLiant DL740",
41 DMI_MATCH(DMI_BIOS_VENDOR, "HP"), 41 { DMI_MATCH(DMI_BIOS_VENDOR, "HP"),
42 DMI_MATCH(DMI_BIOS_VERSION, "P47-"), 42 DMI_MATCH(DMI_BIOS_VERSION, "P47-"),}
43 }}, 43 },
44 { } 44 { }
45}; 45};
46 46
47 47
48static int probe_bigsmp(void) 48static int probe_bigsmp(void)
49{ 49{
50 if (def_to_bigsmp) 50 if (def_to_bigsmp)
51 dmi_bigsmp = 1; 51 dmi_bigsmp = 1;
52 else 52 else
53 dmi_check_system(bigsmp_dmi_table); 53 dmi_check_system(bigsmp_dmi_table);
54 return dmi_bigsmp; 54 return dmi_bigsmp;
55} 55}
56 56
57struct genapic apic_bigsmp = APIC_INIT("bigsmp", probe_bigsmp); 57struct genapic apic_bigsmp = APIC_INIT("bigsmp", probe_bigsmp);
diff --git a/arch/x86/mach-generic/default.c b/arch/x86/mach-generic/default.c
index 1af0cc7648f0..9e835a11a13a 100644
--- a/arch/x86/mach-generic/default.c
+++ b/arch/x86/mach-generic/default.c
@@ -1,4 +1,4 @@
1/* 1/*
2 * Default generic APIC driver. This handles up to 8 CPUs. 2 * Default generic APIC driver. This handles up to 8 CPUs.
3 */ 3 */
4#define APIC_DEFINITION 1 4#define APIC_DEFINITION 1
@@ -19,8 +19,8 @@
19 19
20/* should be called last. */ 20/* should be called last. */
21static int probe_default(void) 21static int probe_default(void)
22{ 22{
23 return 1; 23 return 1;
24} 24}
25 25
26struct genapic apic_default = APIC_INIT("default", probe_default); 26struct genapic apic_default = APIC_INIT("default", probe_default);
diff --git a/arch/x86/mach-generic/probe.c b/arch/x86/mach-generic/probe.c
index f410d3cb5659..c5ae751b994a 100644
--- a/arch/x86/mach-generic/probe.c
+++ b/arch/x86/mach-generic/probe.c
@@ -1,8 +1,9 @@
1/* Copyright 2003 Andi Kleen, SuSE Labs. 1/*
2 * Subject to the GNU Public License, v.2 2 * Copyright 2003 Andi Kleen, SuSE Labs.
3 * 3 * Subject to the GNU Public License, v.2
4 *
4 * Generic x86 APIC driver probe layer. 5 * Generic x86 APIC driver probe layer.
5 */ 6 */
6#include <linux/threads.h> 7#include <linux/threads.h>
7#include <linux/cpumask.h> 8#include <linux/cpumask.h>
8#include <linux/string.h> 9#include <linux/string.h>
@@ -24,7 +25,7 @@ struct genapic *genapic = &apic_default;
24 25
25static struct genapic *apic_probe[] __initdata = { 26static struct genapic *apic_probe[] __initdata = {
26 &apic_summit, 27 &apic_summit,
27 &apic_bigsmp, 28 &apic_bigsmp,
28 &apic_es7000, 29 &apic_es7000,
29 &apic_default, /* must be last */ 30 &apic_default, /* must be last */
30 NULL, 31 NULL,
@@ -69,7 +70,7 @@ void __init generic_bigsmp_probe(void)
69} 70}
70 71
71void __init generic_apic_probe(void) 72void __init generic_apic_probe(void)
72{ 73{
73 if (!cmdline_apic) { 74 if (!cmdline_apic) {
74 int i; 75 int i;
75 for (i = 0; apic_probe[i]; i++) { 76 for (i = 0; apic_probe[i]; i++) {
@@ -83,40 +84,40 @@ void __init generic_apic_probe(void)
83 panic("Didn't find an APIC driver"); 84 panic("Didn't find an APIC driver");
84 } 85 }
85 printk(KERN_INFO "Using APIC driver %s\n", genapic->name); 86 printk(KERN_INFO "Using APIC driver %s\n", genapic->name);
86} 87}
87 88
88/* These functions can switch the APIC even after the initial ->probe() */ 89/* These functions can switch the APIC even after the initial ->probe() */
89 90
90int __init mps_oem_check(struct mp_config_table *mpc, char *oem, char *productid) 91int __init mps_oem_check(struct mp_config_table *mpc, char *oem, char *productid)
91{ 92{
92 int i; 93 int i;
93 for (i = 0; apic_probe[i]; ++i) { 94 for (i = 0; apic_probe[i]; ++i) {
94 if (apic_probe[i]->mps_oem_check(mpc,oem,productid)) { 95 if (apic_probe[i]->mps_oem_check(mpc, oem, productid)) {
95 if (!cmdline_apic) { 96 if (!cmdline_apic) {
96 genapic = apic_probe[i]; 97 genapic = apic_probe[i];
97 printk(KERN_INFO "Switched to APIC driver `%s'.\n", 98 printk(KERN_INFO "Switched to APIC driver `%s'.\n",
98 genapic->name); 99 genapic->name);
99 } 100 }
100 return 1; 101 return 1;
101 } 102 }
102 } 103 }
103 return 0; 104 return 0;
104} 105}
105 106
106int __init acpi_madt_oem_check(char *oem_id, char *oem_table_id) 107int __init acpi_madt_oem_check(char *oem_id, char *oem_table_id)
107{ 108{
108 int i; 109 int i;
109 for (i = 0; apic_probe[i]; ++i) { 110 for (i = 0; apic_probe[i]; ++i) {
110 if (apic_probe[i]->acpi_madt_oem_check(oem_id, oem_table_id)) { 111 if (apic_probe[i]->acpi_madt_oem_check(oem_id, oem_table_id)) {
111 if (!cmdline_apic) { 112 if (!cmdline_apic) {
112 genapic = apic_probe[i]; 113 genapic = apic_probe[i];
113 printk(KERN_INFO "Switched to APIC driver `%s'.\n", 114 printk(KERN_INFO "Switched to APIC driver `%s'.\n",
114 genapic->name); 115 genapic->name);
115 } 116 }
116 return 1; 117 return 1;
117 } 118 }
118 } 119 }
119 return 0; 120 return 0;
120} 121}
121 122
122int hard_smp_processor_id(void) 123int hard_smp_processor_id(void)
diff --git a/arch/x86/mach-generic/summit.c b/arch/x86/mach-generic/summit.c
index 74883ccb8f73..a97ea0f35b1e 100644
--- a/arch/x86/mach-generic/summit.c
+++ b/arch/x86/mach-generic/summit.c
@@ -1,4 +1,4 @@
1/* 1/*
2 * APIC driver for the IBM "Summit" chipset. 2 * APIC driver for the IBM "Summit" chipset.
3 */ 3 */
4#define APIC_DEFINITION 1 4#define APIC_DEFINITION 1
@@ -19,9 +19,9 @@
19#include <asm/mach-summit/mach_mpparse.h> 19#include <asm/mach-summit/mach_mpparse.h>
20 20
21static int probe_summit(void) 21static int probe_summit(void)
22{ 22{
23 /* probed later in mptable/ACPI hooks */ 23 /* probed later in mptable/ACPI hooks */
24 return 0; 24 return 0;
25} 25}
26 26
27struct genapic apic_summit = APIC_INIT("summit", probe_summit); 27struct genapic apic_summit = APIC_INIT("summit", probe_summit);
diff --git a/arch/x86/mach-rdc321x/Makefile b/arch/x86/mach-rdc321x/Makefile
index 1faac8125e3d..8325b4ca431c 100644
--- a/arch/x86/mach-rdc321x/Makefile
+++ b/arch/x86/mach-rdc321x/Makefile
@@ -1,5 +1,5 @@
1# 1#
2# Makefile for the RDC321x specific parts of the kernel 2# Makefile for the RDC321x specific parts of the kernel
3# 3#
4obj-$(CONFIG_X86_RDC321X) := gpio.o platform.o wdt.o 4obj-$(CONFIG_X86_RDC321X) := gpio.o platform.o
5 5
diff --git a/arch/x86/mach-rdc321x/wdt.c b/arch/x86/mach-rdc321x/wdt.c
deleted file mode 100644
index ec5625ae7061..000000000000
--- a/arch/x86/mach-rdc321x/wdt.c
+++ /dev/null
@@ -1,275 +0,0 @@
1/*
2 * RDC321x watchdog driver
3 *
4 * Copyright (C) 2007 Florian Fainelli <florian@openwrt.org>
5 *
6 * This driver is highly inspired from the cpu5_wdt driver
7 *
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License as published by
10 * the Free Software Foundation; either version 2 of the License, or
11 * (at your option) any later version.
12 *
13 * This program is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU General Public License for more details.
17 *
18 * You should have received a copy of the GNU General Public License
19 * along with this program; if not, write to the Free Software
20 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
21 *
22 */
23
24#include <linux/module.h>
25#include <linux/moduleparam.h>
26#include <linux/types.h>
27#include <linux/errno.h>
28#include <linux/miscdevice.h>
29#include <linux/fs.h>
30#include <linux/init.h>
31#include <linux/ioport.h>
32#include <linux/timer.h>
33#include <linux/completion.h>
34#include <linux/jiffies.h>
35#include <linux/platform_device.h>
36#include <linux/watchdog.h>
37#include <linux/io.h>
38#include <linux/uaccess.h>
39
40#include <asm/mach-rdc321x/rdc321x_defs.h>
41
42#define RDC_WDT_MASK 0x80000000 /* Mask */
43#define RDC_WDT_EN 0x00800000 /* Enable bit */
44#define RDC_WDT_WTI 0x00200000 /* Generate CPU reset/NMI/WDT on timeout */
45#define RDC_WDT_RST 0x00100000 /* Reset bit */
46#define RDC_WDT_WIF 0x00040000 /* WDT IRQ Flag */
47#define RDC_WDT_IRT 0x00000100 /* IRQ Routing table */
48#define RDC_WDT_CNT 0x00000001 /* WDT count */
49
50#define RDC_CLS_TMR 0x80003844 /* Clear timer */
51
52#define RDC_WDT_INTERVAL (HZ/10+1)
53
54int nowayout = WATCHDOG_NOWAYOUT;
55module_param(nowayout, int, 0);
56MODULE_PARM_DESC(nowayout, "Watchdog cannot be stopped once started (default=" __MODULE_STRING(WATCHDOG_NOWAYOUT) ")");
57
58static int ticks = 1000;
59
60/* some device data */
61
62static struct {
63 struct completion stop;
64 volatile int running;
65 struct timer_list timer;
66 volatile int queue;
67 int default_ticks;
68 unsigned long inuse;
69} rdc321x_wdt_device;
70
71/* generic helper functions */
72
73static void rdc321x_wdt_trigger(unsigned long unused)
74{
75 if (rdc321x_wdt_device.running)
76 ticks--;
77
78 /* keep watchdog alive */
79 outl(RDC_WDT_EN|inl(RDC3210_CFGREG_DATA), RDC3210_CFGREG_DATA);
80
81 /* requeue?? */
82 if (rdc321x_wdt_device.queue && ticks)
83 mod_timer(&rdc321x_wdt_device.timer,
84 jiffies + RDC_WDT_INTERVAL);
85 else {
86 /* ticks doesn't matter anyway */
87 complete(&rdc321x_wdt_device.stop);
88 }
89
90}
91
92static void rdc321x_wdt_reset(void)
93{
94 ticks = rdc321x_wdt_device.default_ticks;
95}
96
97static void rdc321x_wdt_start(void)
98{
99 if (!rdc321x_wdt_device.queue) {
100 rdc321x_wdt_device.queue = 1;
101
102 /* Clear the timer */
103 outl(RDC_CLS_TMR, RDC3210_CFGREG_ADDR);
104
105 /* Enable watchdog and set the timeout to 81.92 us */
106 outl(RDC_WDT_EN|RDC_WDT_CNT, RDC3210_CFGREG_DATA);
107
108 mod_timer(&rdc321x_wdt_device.timer,
109 jiffies + RDC_WDT_INTERVAL);
110 }
111
112 /* if process dies, counter is not decremented */
113 rdc321x_wdt_device.running++;
114}
115
116static int rdc321x_wdt_stop(void)
117{
118 if (rdc321x_wdt_device.running)
119 rdc321x_wdt_device.running = 0;
120
121 ticks = rdc321x_wdt_device.default_ticks;
122
123 return -EIO;
124}
125
126/* filesystem operations */
127
128static int rdc321x_wdt_open(struct inode *inode, struct file *file)
129{
130 if (test_and_set_bit(0, &rdc321x_wdt_device.inuse))
131 return -EBUSY;
132
133 return nonseekable_open(inode, file);
134}
135
136static int rdc321x_wdt_release(struct inode *inode, struct file *file)
137{
138 clear_bit(0, &rdc321x_wdt_device.inuse);
139 return 0;
140}
141
142static int rdc321x_wdt_ioctl(struct inode *inode, struct file *file,
143 unsigned int cmd, unsigned long arg)
144{
145 void __user *argp = (void __user *)arg;
146 unsigned int value;
147 static struct watchdog_info ident = {
148 .options = WDIOF_CARDRESET,
149 .identity = "RDC321x WDT",
150 };
151
152 switch (cmd) {
153 case WDIOC_KEEPALIVE:
154 rdc321x_wdt_reset();
155 break;
156 case WDIOC_GETSTATUS:
157 /* Read the value from the DATA register */
158 value = inl(RDC3210_CFGREG_DATA);
159 if (copy_to_user(argp, &value, sizeof(int)))
160 return -EFAULT;
161 break;
162 case WDIOC_GETSUPPORT:
163 if (copy_to_user(argp, &ident, sizeof(ident)))
164 return -EFAULT;
165 break;
166 case WDIOC_SETOPTIONS:
167 if (copy_from_user(&value, argp, sizeof(int)))
168 return -EFAULT;
169 switch (value) {
170 case WDIOS_ENABLECARD:
171 rdc321x_wdt_start();
172 break;
173 case WDIOS_DISABLECARD:
174 return rdc321x_wdt_stop();
175 default:
176 return -EINVAL;
177 }
178 break;
179 default:
180 return -ENOTTY;
181 }
182 return 0;
183}
184
185static ssize_t rdc321x_wdt_write(struct file *file, const char __user *buf,
186 size_t count, loff_t *ppos)
187{
188 if (!count)
189 return -EIO;
190
191 rdc321x_wdt_reset();
192
193 return count;
194}
195
196static const struct file_operations rdc321x_wdt_fops = {
197 .owner = THIS_MODULE,
198 .llseek = no_llseek,
199 .ioctl = rdc321x_wdt_ioctl,
200 .open = rdc321x_wdt_open,
201 .write = rdc321x_wdt_write,
202 .release = rdc321x_wdt_release,
203};
204
205static struct miscdevice rdc321x_wdt_misc = {
206 .minor = WATCHDOG_MINOR,
207 .name = "watchdog",
208 .fops = &rdc321x_wdt_fops,
209};
210
211static int __devinit rdc321x_wdt_probe(struct platform_device *pdev)
212{
213 int err;
214
215 err = misc_register(&rdc321x_wdt_misc);
216 if (err < 0) {
217 printk(KERN_ERR PFX "watchdog misc_register failed\n");
218 return err;
219 }
220
221 /* Reset the watchdog */
222 outl(RDC_WDT_RST, RDC3210_CFGREG_DATA);
223
224 init_completion(&rdc321x_wdt_device.stop);
225 rdc321x_wdt_device.queue = 0;
226
227 clear_bit(0, &rdc321x_wdt_device.inuse);
228
229 setup_timer(&rdc321x_wdt_device.timer, rdc321x_wdt_trigger, 0);
230
231 rdc321x_wdt_device.default_ticks = ticks;
232
233 printk(KERN_INFO PFX "watchdog init success\n");
234
235 return 0;
236}
237
238static int rdc321x_wdt_remove(struct platform_device *pdev)
239{
240 if (rdc321x_wdt_device.queue) {
241 rdc321x_wdt_device.queue = 0;
242 wait_for_completion(&rdc321x_wdt_device.stop);
243 }
244
245 misc_deregister(&rdc321x_wdt_misc);
246
247 return 0;
248}
249
250static struct platform_driver rdc321x_wdt_driver = {
251 .probe = rdc321x_wdt_probe,
252 .remove = rdc321x_wdt_remove,
253 .driver = {
254 .owner = THIS_MODULE,
255 .name = "rdc321x-wdt",
256 },
257};
258
259static int __init rdc321x_wdt_init(void)
260{
261 return platform_driver_register(&rdc321x_wdt_driver);
262}
263
264static void __exit rdc321x_wdt_exit(void)
265{
266 platform_driver_unregister(&rdc321x_wdt_driver);
267}
268
269module_init(rdc321x_wdt_init);
270module_exit(rdc321x_wdt_exit);
271
272MODULE_AUTHOR("Florian Fainelli <florian@openwrt.org>");
273MODULE_DESCRIPTION("RDC321x watchdog driver");
274MODULE_LICENSE("GPL");
275MODULE_ALIAS_MISCDEV(WATCHDOG_MINOR);
diff --git a/arch/x86/mach-voyager/voyager_smp.c b/arch/x86/mach-voyager/voyager_smp.c
index 3cc8eb2f36a9..be7235bf105d 100644
--- a/arch/x86/mach-voyager/voyager_smp.c
+++ b/arch/x86/mach-voyager/voyager_smp.c
@@ -27,6 +27,7 @@
27#include <asm/pgalloc.h> 27#include <asm/pgalloc.h>
28#include <asm/tlbflush.h> 28#include <asm/tlbflush.h>
29#include <asm/arch_hooks.h> 29#include <asm/arch_hooks.h>
30#include <asm/trampoline.h>
30 31
31/* TLB state -- visible externally, indexed physically */ 32/* TLB state -- visible externally, indexed physically */
32DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate) = { &init_mm, 0 }; 33DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate) = { &init_mm, 0 };
@@ -210,7 +211,7 @@ static int cpucount = 0;
210/* steal a page from the bottom of memory for the trampoline and 211/* steal a page from the bottom of memory for the trampoline and
211 * squirrel its address away here. This will be in kernel virtual 212 * squirrel its address away here. This will be in kernel virtual
212 * space */ 213 * space */
213static __u32 trampoline_base; 214unsigned char *trampoline_base;
214 215
215/* The per cpu profile stuff - used in smp_local_timer_interrupt */ 216/* The per cpu profile stuff - used in smp_local_timer_interrupt */
216static DEFINE_PER_CPU(int, prof_multiplier) = 1; 217static DEFINE_PER_CPU(int, prof_multiplier) = 1;
@@ -429,15 +430,15 @@ void __init smp_store_cpu_info(int id)
429} 430}
430 431
431/* set up the trampoline and return the physical address of the code */ 432/* set up the trampoline and return the physical address of the code */
432static __u32 __init setup_trampoline(void) 433unsigned long __init setup_trampoline(void)
433{ 434{
434 /* these two are global symbols in trampoline.S */ 435 /* these two are global symbols in trampoline.S */
435 extern const __u8 trampoline_end[]; 436 extern const __u8 trampoline_end[];
436 extern const __u8 trampoline_data[]; 437 extern const __u8 trampoline_data[];
437 438
438 memcpy((__u8 *) trampoline_base, trampoline_data, 439 memcpy(trampoline_base, trampoline_data,
439 trampoline_end - trampoline_data); 440 trampoline_end - trampoline_data);
440 return virt_to_phys((__u8 *) trampoline_base); 441 return virt_to_phys(trampoline_base);
441} 442}
442 443
443/* Routine initially called when a non-boot CPU is brought online */ 444/* Routine initially called when a non-boot CPU is brought online */
@@ -520,13 +521,6 @@ static void __init do_boot_cpu(__u8 cpu)
520 & ~(voyager_extended_vic_processors 521 & ~(voyager_extended_vic_processors
521 & voyager_allowed_boot_processors); 522 & voyager_allowed_boot_processors);
522 523
523 /* This is an area in head.S which was used to set up the
524 * initial kernel stack. We need to alter this to give the
525 * booting CPU a new stack (taken from its idle process) */
526 extern struct {
527 __u8 *sp;
528 unsigned short ss;
529 } stack_start;
530 /* This is the format of the CPI IDT gate (in real mode) which 524 /* This is the format of the CPI IDT gate (in real mode) which
531 * we're hijacking to boot the CPU */ 525 * we're hijacking to boot the CPU */
532 union IDTFormat { 526 union IDTFormat {
@@ -1166,7 +1160,7 @@ void flush_tlb_all(void)
1166 * is sorted out */ 1160 * is sorted out */
1167void __init smp_alloc_memory(void) 1161void __init smp_alloc_memory(void)
1168{ 1162{
1169 trampoline_base = (__u32) alloc_bootmem_low_pages(PAGE_SIZE); 1163 trampoline_base = alloc_bootmem_low_pages(PAGE_SIZE);
1170 if (__pa(trampoline_base) >= 0x93000) 1164 if (__pa(trampoline_base) >= 0x93000)
1171 BUG(); 1165 BUG();
1172} 1166}
diff --git a/arch/x86/math-emu/fpu_entry.c b/arch/x86/math-emu/fpu_entry.c
index 760baeea5f07..4bab3b145392 100644
--- a/arch/x86/math-emu/fpu_entry.c
+++ b/arch/x86/math-emu/fpu_entry.c
@@ -276,6 +276,7 @@ asmlinkage void math_emulate(long arg)
276 entry_sel_off.offset = FPU_ORIG_EIP; 276 entry_sel_off.offset = FPU_ORIG_EIP;
277 entry_sel_off.selector = FPU_CS; 277 entry_sel_off.selector = FPU_CS;
278 entry_sel_off.opcode = (byte1 << 8) | FPU_modrm; 278 entry_sel_off.opcode = (byte1 << 8) | FPU_modrm;
279 entry_sel_off.empty = 0;
279 280
280 FPU_rm = FPU_modrm & 7; 281 FPU_rm = FPU_modrm & 7;
281 282
diff --git a/arch/x86/math-emu/reg_ld_str.c b/arch/x86/math-emu/reg_ld_str.c
index 799d4af5be66..02af772a24db 100644
--- a/arch/x86/math-emu/reg_ld_str.c
+++ b/arch/x86/math-emu/reg_ld_str.c
@@ -383,15 +383,15 @@ int FPU_store_double(FPU_REG *st0_ptr, u_char st0_tag, double __user *dfloat)
383 int exp; 383 int exp;
384 FPU_REG tmp; 384 FPU_REG tmp;
385 385
386 l[0] = 0;
387 l[1] = 0;
386 if (st0_tag == TAG_Valid) { 388 if (st0_tag == TAG_Valid) {
387 reg_copy(st0_ptr, &tmp); 389 reg_copy(st0_ptr, &tmp);
388 exp = exponent(&tmp); 390 exp = exponent(&tmp);
389 391
390 if (exp < DOUBLE_Emin) { /* It may be a denormal */ 392 if (exp < DOUBLE_Emin) { /* It may be a denormal */
391 addexponent(&tmp, -DOUBLE_Emin + 52); /* largest exp to be 51 */ 393 addexponent(&tmp, -DOUBLE_Emin + 52); /* largest exp to be 51 */
392 394denormal_arg:
393 denormal_arg:
394
395 if ((precision_loss = FPU_round_to_int(&tmp, st0_tag))) { 395 if ((precision_loss = FPU_round_to_int(&tmp, st0_tag))) {
396#ifdef PECULIAR_486 396#ifdef PECULIAR_486
397 /* Did it round to a non-denormal ? */ 397 /* Did it round to a non-denormal ? */
@@ -477,8 +477,7 @@ int FPU_store_double(FPU_REG *st0_ptr, u_char st0_tag, double __user *dfloat)
477 477
478 /* This is a special case: see sec 16.2.5.1 of the 80486 book */ 478 /* This is a special case: see sec 16.2.5.1 of the 80486 book */
479 /* Overflow to infinity */ 479 /* Overflow to infinity */
480 l[0] = 0x00000000; /* Set to */ 480 l[1] = 0x7ff00000; /* Set to + INF */
481 l[1] = 0x7ff00000; /* + INF */
482 } else { 481 } else {
483 if (precision_loss) { 482 if (precision_loss) {
484 if (increment) 483 if (increment)
@@ -492,8 +491,6 @@ int FPU_store_double(FPU_REG *st0_ptr, u_char st0_tag, double __user *dfloat)
492 } 491 }
493 } else if (st0_tag == TAG_Zero) { 492 } else if (st0_tag == TAG_Zero) {
494 /* Number is zero */ 493 /* Number is zero */
495 l[0] = 0;
496 l[1] = 0;
497 } else if (st0_tag == TAG_Special) { 494 } else if (st0_tag == TAG_Special) {
498 st0_tag = FPU_Special(st0_ptr); 495 st0_tag = FPU_Special(st0_ptr);
499 if (st0_tag == TW_Denormal) { 496 if (st0_tag == TW_Denormal) {
@@ -508,7 +505,6 @@ int FPU_store_double(FPU_REG *st0_ptr, u_char st0_tag, double __user *dfloat)
508 reg_copy(st0_ptr, &tmp); 505 reg_copy(st0_ptr, &tmp);
509 goto denormal_arg; 506 goto denormal_arg;
510 } else if (st0_tag == TW_Infinity) { 507 } else if (st0_tag == TW_Infinity) {
511 l[0] = 0;
512 l[1] = 0x7ff00000; 508 l[1] = 0x7ff00000;
513 } else if (st0_tag == TW_NaN) { 509 } else if (st0_tag == TW_NaN) {
514 /* Is it really a NaN ? */ 510 /* Is it really a NaN ? */
@@ -532,7 +528,6 @@ int FPU_store_double(FPU_REG *st0_ptr, u_char st0_tag, double __user *dfloat)
532 EXCEPTION(EX_Invalid); 528 EXCEPTION(EX_Invalid);
533 if (!(control_word & CW_Invalid)) 529 if (!(control_word & CW_Invalid))
534 return 0; 530 return 0;
535 l[0] = 0;
536 l[1] = 0xfff80000; 531 l[1] = 0xfff80000;
537 } 532 }
538 } 533 }
diff --git a/arch/x86/mm/Makefile b/arch/x86/mm/Makefile
index 983291096848..20941d2954e2 100644
--- a/arch/x86/mm/Makefile
+++ b/arch/x86/mm/Makefile
@@ -1,5 +1,17 @@
1obj-y := init_$(BITS).o fault.o ioremap.o extable.o pageattr.o mmap.o \
2 pat.o
3
4obj-$(CONFIG_X86_32) += pgtable_32.o
5
6obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o
7obj-$(CONFIG_X86_PTDUMP) += dump_pagetables.o
8
9obj-$(CONFIG_HIGHMEM) += highmem_32.o
10
1ifeq ($(CONFIG_X86_32),y) 11ifeq ($(CONFIG_X86_32),y)
2include ${srctree}/arch/x86/mm/Makefile_32 12obj-$(CONFIG_NUMA) += discontig_32.o
3else 13else
4include ${srctree}/arch/x86/mm/Makefile_64 14obj-$(CONFIG_NUMA) += numa_64.o
15obj-$(CONFIG_K8_NUMA) += k8topology_64.o
16obj-$(CONFIG_ACPI_NUMA) += srat_64.o
5endif 17endif
diff --git a/arch/x86/mm/Makefile_32 b/arch/x86/mm/Makefile_32
deleted file mode 100644
index c36ae88bb543..000000000000
--- a/arch/x86/mm/Makefile_32
+++ /dev/null
@@ -1,9 +0,0 @@
1#
2# Makefile for the linux i386-specific parts of the memory manager.
3#
4
5obj-y := init_32.o pgtable_32.o fault.o ioremap.o extable.o pageattr.o mmap.o
6
7obj-$(CONFIG_NUMA) += discontig_32.o
8obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o
9obj-$(CONFIG_HIGHMEM) += highmem_32.o
diff --git a/arch/x86/mm/Makefile_64 b/arch/x86/mm/Makefile_64
deleted file mode 100644
index 688c8c28ac8f..000000000000
--- a/arch/x86/mm/Makefile_64
+++ /dev/null
@@ -1,9 +0,0 @@
1#
2# Makefile for the linux x86_64-specific parts of the memory manager.
3#
4
5obj-y := init_64.o fault.o ioremap.o extable.o pageattr.o mmap.o
6obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o
7obj-$(CONFIG_NUMA) += numa_64.o
8obj-$(CONFIG_K8_NUMA) += k8topology_64.o
9obj-$(CONFIG_ACPI_NUMA) += srat_64.o
diff --git a/arch/x86/mm/discontig_32.c b/arch/x86/mm/discontig_32.c
index 8e25e06ff730..eba0bbede7a6 100644
--- a/arch/x86/mm/discontig_32.c
+++ b/arch/x86/mm/discontig_32.c
@@ -37,7 +37,7 @@
37#include <asm/e820.h> 37#include <asm/e820.h>
38#include <asm/setup.h> 38#include <asm/setup.h>
39#include <asm/mmzone.h> 39#include <asm/mmzone.h>
40#include <bios_ebda.h> 40#include <asm/bios_ebda.h>
41 41
42struct pglist_data *node_data[MAX_NUMNODES] __read_mostly; 42struct pglist_data *node_data[MAX_NUMNODES] __read_mostly;
43EXPORT_SYMBOL(node_data); 43EXPORT_SYMBOL(node_data);
diff --git a/arch/x86/mm/dump_pagetables.c b/arch/x86/mm/dump_pagetables.c
new file mode 100644
index 000000000000..6791b8334bc6
--- /dev/null
+++ b/arch/x86/mm/dump_pagetables.c
@@ -0,0 +1,354 @@
1/*
2 * Debug helper to dump the current kernel pagetables of the system
3 * so that we can see what the various memory ranges are set to.
4 *
5 * (C) Copyright 2008 Intel Corporation
6 *
7 * Author: Arjan van de Ven <arjan@linux.intel.com>
8 *
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; version 2
12 * of the License.
13 */
14
15#include <linux/debugfs.h>
16#include <linux/mm.h>
17#include <linux/module.h>
18#include <linux/seq_file.h>
19
20#include <asm/pgtable.h>
21
22/*
23 * The dumper groups pagetable entries of the same type into one, and for
24 * that it needs to keep some state when walking, and flush this state
25 * when a "break" in the continuity is found.
26 */
27struct pg_state {
28 int level;
29 pgprot_t current_prot;
30 unsigned long start_address;
31 unsigned long current_address;
32 const struct addr_marker *marker;
33};
34
35struct addr_marker {
36 unsigned long start_address;
37 const char *name;
38};
39
40/* Address space markers hints */
41static struct addr_marker address_markers[] = {
42 { 0, "User Space" },
43#ifdef CONFIG_X86_64
44 { 0x8000000000000000UL, "Kernel Space" },
45 { 0xffff810000000000UL, "Low Kernel Mapping" },
46 { VMALLOC_START, "vmalloc() Area" },
47 { VMEMMAP_START, "Vmemmap" },
48 { __START_KERNEL_map, "High Kernel Mapping" },
49 { MODULES_VADDR, "Modules" },
50 { MODULES_END, "End Modules" },
51#else
52 { PAGE_OFFSET, "Kernel Mapping" },
53 { 0/* VMALLOC_START */, "vmalloc() Area" },
54 { 0/*VMALLOC_END*/, "vmalloc() End" },
55# ifdef CONFIG_HIGHMEM
56 { 0/*PKMAP_BASE*/, "Persisent kmap() Area" },
57# endif
58 { 0/*FIXADDR_START*/, "Fixmap Area" },
59#endif
60 { -1, NULL } /* End of list */
61};
62
63/* Multipliers for offsets within the PTEs */
64#define PTE_LEVEL_MULT (PAGE_SIZE)
65#define PMD_LEVEL_MULT (PTRS_PER_PTE * PTE_LEVEL_MULT)
66#define PUD_LEVEL_MULT (PTRS_PER_PMD * PMD_LEVEL_MULT)
67#define PGD_LEVEL_MULT (PTRS_PER_PUD * PUD_LEVEL_MULT)
68
69/*
70 * Print a readable form of a pgprot_t to the seq_file
71 */
72static void printk_prot(struct seq_file *m, pgprot_t prot, int level)
73{
74 pgprotval_t pr = pgprot_val(prot);
75 static const char * const level_name[] =
76 { "cr3", "pgd", "pud", "pmd", "pte" };
77
78 if (!pgprot_val(prot)) {
79 /* Not present */
80 seq_printf(m, " ");
81 } else {
82 if (pr & _PAGE_USER)
83 seq_printf(m, "USR ");
84 else
85 seq_printf(m, " ");
86 if (pr & _PAGE_RW)
87 seq_printf(m, "RW ");
88 else
89 seq_printf(m, "ro ");
90 if (pr & _PAGE_PWT)
91 seq_printf(m, "PWT ");
92 else
93 seq_printf(m, " ");
94 if (pr & _PAGE_PCD)
95 seq_printf(m, "PCD ");
96 else
97 seq_printf(m, " ");
98
99 /* Bit 9 has a different meaning on level 3 vs 4 */
100 if (level <= 3) {
101 if (pr & _PAGE_PSE)
102 seq_printf(m, "PSE ");
103 else
104 seq_printf(m, " ");
105 } else {
106 if (pr & _PAGE_PAT)
107 seq_printf(m, "pat ");
108 else
109 seq_printf(m, " ");
110 }
111 if (pr & _PAGE_GLOBAL)
112 seq_printf(m, "GLB ");
113 else
114 seq_printf(m, " ");
115 if (pr & _PAGE_NX)
116 seq_printf(m, "NX ");
117 else
118 seq_printf(m, "x ");
119 }
120 seq_printf(m, "%s\n", level_name[level]);
121}
122
123/*
124 * On 64 bits, sign-extend the 48 bit address to 64 bit
125 */
126static unsigned long normalize_addr(unsigned long u)
127{
128#ifdef CONFIG_X86_64
129 return (signed long)(u << 16) >> 16;
130#else
131 return u;
132#endif
133}
134
135/*
136 * This function gets called on a break in a continuous series
137 * of PTE entries; the next one is different so we need to
138 * print what we collected so far.
139 */
140static void note_page(struct seq_file *m, struct pg_state *st,
141 pgprot_t new_prot, int level)
142{
143 pgprotval_t prot, cur;
144 static const char units[] = "KMGTPE";
145
146 /*
147 * If we have a "break" in the series, we need to flush the state that
148 * we have now. "break" is either changing perms, levels or
149 * address space marker.
150 */
151 prot = pgprot_val(new_prot) & ~(PTE_MASK);
152 cur = pgprot_val(st->current_prot) & ~(PTE_MASK);
153
154 if (!st->level) {
155 /* First entry */
156 st->current_prot = new_prot;
157 st->level = level;
158 st->marker = address_markers;
159 seq_printf(m, "---[ %s ]---\n", st->marker->name);
160 } else if (prot != cur || level != st->level ||
161 st->current_address >= st->marker[1].start_address) {
162 const char *unit = units;
163 unsigned long delta;
164
165 /*
166 * Now print the actual finished series
167 */
168 seq_printf(m, "0x%p-0x%p ",
169 (void *)st->start_address,
170 (void *)st->current_address);
171
172 delta = (st->current_address - st->start_address) >> 10;
173 while (!(delta & 1023) && unit[1]) {
174 delta >>= 10;
175 unit++;
176 }
177 seq_printf(m, "%9lu%c ", delta, *unit);
178 printk_prot(m, st->current_prot, st->level);
179
180 /*
181 * We print markers for special areas of address space,
182 * such as the start of vmalloc space etc.
183 * This helps in the interpretation.
184 */
185 if (st->current_address >= st->marker[1].start_address) {
186 st->marker++;
187 seq_printf(m, "---[ %s ]---\n", st->marker->name);
188 }
189
190 st->start_address = st->current_address;
191 st->current_prot = new_prot;
192 st->level = level;
193 }
194}
195
196static void walk_pte_level(struct seq_file *m, struct pg_state *st, pmd_t addr,
197 unsigned long P)
198{
199 int i;
200 pte_t *start;
201
202 start = (pte_t *) pmd_page_vaddr(addr);
203 for (i = 0; i < PTRS_PER_PTE; i++) {
204 pgprot_t prot = pte_pgprot(*start);
205
206 st->current_address = normalize_addr(P + i * PTE_LEVEL_MULT);
207 note_page(m, st, prot, 4);
208 start++;
209 }
210}
211
212#if PTRS_PER_PMD > 1
213
214static void walk_pmd_level(struct seq_file *m, struct pg_state *st, pud_t addr,
215 unsigned long P)
216{
217 int i;
218 pmd_t *start;
219
220 start = (pmd_t *) pud_page_vaddr(addr);
221 for (i = 0; i < PTRS_PER_PMD; i++) {
222 st->current_address = normalize_addr(P + i * PMD_LEVEL_MULT);
223 if (!pmd_none(*start)) {
224 pgprotval_t prot = pmd_val(*start) & ~PTE_MASK;
225
226 if (pmd_large(*start) || !pmd_present(*start))
227 note_page(m, st, __pgprot(prot), 3);
228 else
229 walk_pte_level(m, st, *start,
230 P + i * PMD_LEVEL_MULT);
231 } else
232 note_page(m, st, __pgprot(0), 3);
233 start++;
234 }
235}
236
237#else
238#define walk_pmd_level(m,s,a,p) walk_pte_level(m,s,__pmd(pud_val(a)),p)
239#define pud_large(a) pmd_large(__pmd(pud_val(a)))
240#define pud_none(a) pmd_none(__pmd(pud_val(a)))
241#endif
242
243#if PTRS_PER_PUD > 1
244
245static void walk_pud_level(struct seq_file *m, struct pg_state *st, pgd_t addr,
246 unsigned long P)
247{
248 int i;
249 pud_t *start;
250
251 start = (pud_t *) pgd_page_vaddr(addr);
252
253 for (i = 0; i < PTRS_PER_PUD; i++) {
254 st->current_address = normalize_addr(P + i * PUD_LEVEL_MULT);
255 if (!pud_none(*start)) {
256 pgprotval_t prot = pud_val(*start) & ~PTE_MASK;
257
258 if (pud_large(*start) || !pud_present(*start))
259 note_page(m, st, __pgprot(prot), 2);
260 else
261 walk_pmd_level(m, st, *start,
262 P + i * PUD_LEVEL_MULT);
263 } else
264 note_page(m, st, __pgprot(0), 2);
265
266 start++;
267 }
268}
269
270#else
271#define walk_pud_level(m,s,a,p) walk_pmd_level(m,s,__pud(pgd_val(a)),p)
272#define pgd_large(a) pud_large(__pud(pgd_val(a)))
273#define pgd_none(a) pud_none(__pud(pgd_val(a)))
274#endif
275
276static void walk_pgd_level(struct seq_file *m)
277{
278#ifdef CONFIG_X86_64
279 pgd_t *start = (pgd_t *) &init_level4_pgt;
280#else
281 pgd_t *start = swapper_pg_dir;
282#endif
283 int i;
284 struct pg_state st;
285
286 memset(&st, 0, sizeof(st));
287
288 for (i = 0; i < PTRS_PER_PGD; i++) {
289 st.current_address = normalize_addr(i * PGD_LEVEL_MULT);
290 if (!pgd_none(*start)) {
291 pgprotval_t prot = pgd_val(*start) & ~PTE_MASK;
292
293 if (pgd_large(*start) || !pgd_present(*start))
294 note_page(m, &st, __pgprot(prot), 1);
295 else
296 walk_pud_level(m, &st, *start,
297 i * PGD_LEVEL_MULT);
298 } else
299 note_page(m, &st, __pgprot(0), 1);
300
301 start++;
302 }
303
304 /* Flush out the last page */
305 st.current_address = normalize_addr(PTRS_PER_PGD*PGD_LEVEL_MULT);
306 note_page(m, &st, __pgprot(0), 0);
307}
308
309static int ptdump_show(struct seq_file *m, void *v)
310{
311 walk_pgd_level(m);
312 return 0;
313}
314
315static int ptdump_open(struct inode *inode, struct file *filp)
316{
317 return single_open(filp, ptdump_show, NULL);
318}
319
320static const struct file_operations ptdump_fops = {
321 .open = ptdump_open,
322 .read = seq_read,
323 .llseek = seq_lseek,
324 .release = single_release,
325};
326
327int pt_dump_init(void)
328{
329 struct dentry *pe;
330
331#ifdef CONFIG_X86_32
332 /* Not a compile-time constant on x86-32 */
333 address_markers[2].start_address = VMALLOC_START;
334 address_markers[3].start_address = VMALLOC_END;
335# ifdef CONFIG_HIGHMEM
336 address_markers[4].start_address = PKMAP_BASE;
337 address_markers[5].start_address = FIXADDR_START;
338# else
339 address_markers[4].start_address = FIXADDR_START;
340# endif
341#endif
342
343 pe = debugfs_create_file("kernel_page_tables", 0600, NULL, NULL,
344 &ptdump_fops);
345 if (!pe)
346 return -ENOMEM;
347
348 return 0;
349}
350
351__initcall(pt_dump_init);
352MODULE_LICENSE("GPL");
353MODULE_AUTHOR("Arjan van de Ven <arjan@linux.intel.com>");
354MODULE_DESCRIPTION("Kernel debugging helper that dumps pagetables");
diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c
index ec08d8389850..fd7e1798c75a 100644
--- a/arch/x86/mm/fault.c
+++ b/arch/x86/mm/fault.c
@@ -639,7 +639,7 @@ void __kprobes do_page_fault(struct pt_regs *regs, unsigned long error_code)
639#ifdef CONFIG_X86_32 639#ifdef CONFIG_X86_32
640 /* It's safe to allow irq's after cr2 has been saved and the vmalloc 640 /* It's safe to allow irq's after cr2 has been saved and the vmalloc
641 fault has been handled. */ 641 fault has been handled. */
642 if (regs->flags & (X86_EFLAGS_IF|VM_MASK)) 642 if (regs->flags & (X86_EFLAGS_IF | X86_VM_MASK))
643 local_irq_enable(); 643 local_irq_enable();
644 644
645 /* 645 /*
@@ -976,9 +976,5 @@ void vmalloc_sync_all(void)
976 if (address == start) 976 if (address == start)
977 start = address + PGDIR_SIZE; 977 start = address + PGDIR_SIZE;
978 } 978 }
979 /* Check that there is no need to do the same for the modules area. */
980 BUILD_BUG_ON(!(MODULES_VADDR > __START_KERNEL));
981 BUILD_BUG_ON(!(((MODULES_END - 1) & PGDIR_MASK) ==
982 (__START_KERNEL & PGDIR_MASK)));
983#endif 979#endif
984} 980}
diff --git a/arch/x86/mm/init_32.c b/arch/x86/mm/init_32.c
index ee1091a46964..1500dc8d63e4 100644
--- a/arch/x86/mm/init_32.c
+++ b/arch/x86/mm/init_32.c
@@ -51,6 +51,8 @@
51 51
52unsigned int __VMALLOC_RESERVE = 128 << 20; 52unsigned int __VMALLOC_RESERVE = 128 << 20;
53 53
54unsigned long max_pfn_mapped;
55
54DEFINE_PER_CPU(struct mmu_gather, mmu_gathers); 56DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
55unsigned long highstart_pfn, highend_pfn; 57unsigned long highstart_pfn, highend_pfn;
56 58
@@ -179,8 +181,13 @@ static void __init kernel_physical_mapping_init(pgd_t *pgd_base)
179 /* 181 /*
180 * Map with big pages if possible, otherwise 182 * Map with big pages if possible, otherwise
181 * create normal page tables: 183 * create normal page tables:
184 *
185 * Don't use a large page for the first 2/4MB of memory
186 * because there are often fixed size MTRRs in there
187 * and overlapping MTRRs into large pages can cause
188 * slowdowns.
182 */ 189 */
183 if (cpu_has_pse) { 190 if (cpu_has_pse && !(pgd_idx == 0 && pmd_idx == 0)) {
184 unsigned int addr2; 191 unsigned int addr2;
185 pgprot_t prot = PAGE_KERNEL_LARGE; 192 pgprot_t prot = PAGE_KERNEL_LARGE;
186 193
@@ -194,6 +201,7 @@ static void __init kernel_physical_mapping_init(pgd_t *pgd_base)
194 set_pmd(pmd, pfn_pmd(pfn, prot)); 201 set_pmd(pmd, pfn_pmd(pfn, prot));
195 202
196 pfn += PTRS_PER_PTE; 203 pfn += PTRS_PER_PTE;
204 max_pfn_mapped = pfn;
197 continue; 205 continue;
198 } 206 }
199 pte = one_page_table_init(pmd); 207 pte = one_page_table_init(pmd);
@@ -208,6 +216,7 @@ static void __init kernel_physical_mapping_init(pgd_t *pgd_base)
208 216
209 set_pte(pte, pfn_pte(pfn, prot)); 217 set_pte(pte, pfn_pte(pfn, prot));
210 } 218 }
219 max_pfn_mapped = pfn;
211 } 220 }
212 } 221 }
213} 222}
@@ -723,25 +732,17 @@ void mark_rodata_ro(void)
723 unsigned long start = PFN_ALIGN(_text); 732 unsigned long start = PFN_ALIGN(_text);
724 unsigned long size = PFN_ALIGN(_etext) - start; 733 unsigned long size = PFN_ALIGN(_etext) - start;
725 734
726#ifndef CONFIG_KPROBES 735 set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT);
727#ifdef CONFIG_HOTPLUG_CPU 736 printk(KERN_INFO "Write protecting the kernel text: %luk\n",
728 /* It must still be possible to apply SMP alternatives. */ 737 size >> 10);
729 if (num_possible_cpus() <= 1)
730#endif
731 {
732 set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT);
733 printk(KERN_INFO "Write protecting the kernel text: %luk\n",
734 size >> 10);
735 738
736#ifdef CONFIG_CPA_DEBUG 739#ifdef CONFIG_CPA_DEBUG
737 printk(KERN_INFO "Testing CPA: Reverting %lx-%lx\n", 740 printk(KERN_INFO "Testing CPA: Reverting %lx-%lx\n",
738 start, start+size); 741 start, start+size);
739 set_pages_rw(virt_to_page(start), size>>PAGE_SHIFT); 742 set_pages_rw(virt_to_page(start), size>>PAGE_SHIFT);
740 743
741 printk(KERN_INFO "Testing CPA: write protecting again\n"); 744 printk(KERN_INFO "Testing CPA: write protecting again\n");
742 set_pages_ro(virt_to_page(start), size>>PAGE_SHIFT); 745 set_pages_ro(virt_to_page(start), size>>PAGE_SHIFT);
743#endif
744 }
745#endif 746#endif
746 start += size; 747 start += size;
747 size = (unsigned long)__end_rodata - start; 748 size = (unsigned long)__end_rodata - start;
diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c
index a02a14f0f324..1076097dcab2 100644
--- a/arch/x86/mm/init_64.c
+++ b/arch/x86/mm/init_64.c
@@ -54,6 +54,26 @@ static unsigned long dma_reserve __initdata;
54 54
55DEFINE_PER_CPU(struct mmu_gather, mmu_gathers); 55DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
56 56
57int direct_gbpages __meminitdata
58#ifdef CONFIG_DIRECT_GBPAGES
59 = 1
60#endif
61;
62
63static int __init parse_direct_gbpages_off(char *arg)
64{
65 direct_gbpages = 0;
66 return 0;
67}
68early_param("nogbpages", parse_direct_gbpages_off);
69
70static int __init parse_direct_gbpages_on(char *arg)
71{
72 direct_gbpages = 1;
73 return 0;
74}
75early_param("gbpages", parse_direct_gbpages_on);
76
57/* 77/*
58 * NOTE: pagetable_init alloc all the fixmap pagetables contiguous on the 78 * NOTE: pagetable_init alloc all the fixmap pagetables contiguous on the
59 * physical space so we can cache the place of the first one and move 79 * physical space so we can cache the place of the first one and move
@@ -69,9 +89,6 @@ void show_mem(void)
69 89
70 printk(KERN_INFO "Mem-info:\n"); 90 printk(KERN_INFO "Mem-info:\n");
71 show_free_areas(); 91 show_free_areas();
72 printk(KERN_INFO "Free swap: %6ldkB\n",
73 nr_swap_pages << (PAGE_SHIFT-10));
74
75 for_each_online_pgdat(pgdat) { 92 for_each_online_pgdat(pgdat) {
76 for (i = 0; i < pgdat->node_spanned_pages; ++i) { 93 for (i = 0; i < pgdat->node_spanned_pages; ++i) {
77 /* 94 /*
@@ -296,7 +313,7 @@ __meminit void early_iounmap(void *addr, unsigned long size)
296 __flush_tlb_all(); 313 __flush_tlb_all();
297} 314}
298 315
299static void __meminit 316static unsigned long __meminit
300phys_pmd_init(pmd_t *pmd_page, unsigned long address, unsigned long end) 317phys_pmd_init(pmd_t *pmd_page, unsigned long address, unsigned long end)
301{ 318{
302 int i = pmd_index(address); 319 int i = pmd_index(address);
@@ -318,21 +335,26 @@ phys_pmd_init(pmd_t *pmd_page, unsigned long address, unsigned long end)
318 set_pte((pte_t *)pmd, 335 set_pte((pte_t *)pmd,
319 pfn_pte(address >> PAGE_SHIFT, PAGE_KERNEL_LARGE)); 336 pfn_pte(address >> PAGE_SHIFT, PAGE_KERNEL_LARGE));
320 } 337 }
338 return address;
321} 339}
322 340
323static void __meminit 341static unsigned long __meminit
324phys_pmd_update(pud_t *pud, unsigned long address, unsigned long end) 342phys_pmd_update(pud_t *pud, unsigned long address, unsigned long end)
325{ 343{
326 pmd_t *pmd = pmd_offset(pud, 0); 344 pmd_t *pmd = pmd_offset(pud, 0);
345 unsigned long last_map_addr;
346
327 spin_lock(&init_mm.page_table_lock); 347 spin_lock(&init_mm.page_table_lock);
328 phys_pmd_init(pmd, address, end); 348 last_map_addr = phys_pmd_init(pmd, address, end);
329 spin_unlock(&init_mm.page_table_lock); 349 spin_unlock(&init_mm.page_table_lock);
330 __flush_tlb_all(); 350 __flush_tlb_all();
351 return last_map_addr;
331} 352}
332 353
333static void __meminit 354static unsigned long __meminit
334phys_pud_init(pud_t *pud_page, unsigned long addr, unsigned long end) 355phys_pud_init(pud_t *pud_page, unsigned long addr, unsigned long end)
335{ 356{
357 unsigned long last_map_addr = end;
336 int i = pud_index(addr); 358 int i = pud_index(addr);
337 359
338 for (; i < PTRS_PER_PUD; i++, addr = (addr & PUD_MASK) + PUD_SIZE) { 360 for (; i < PTRS_PER_PUD; i++, addr = (addr & PUD_MASK) + PUD_SIZE) {
@@ -350,7 +372,15 @@ phys_pud_init(pud_t *pud_page, unsigned long addr, unsigned long end)
350 } 372 }
351 373
352 if (pud_val(*pud)) { 374 if (pud_val(*pud)) {
353 phys_pmd_update(pud, addr, end); 375 if (!pud_large(*pud))
376 last_map_addr = phys_pmd_update(pud, addr, end);
377 continue;
378 }
379
380 if (direct_gbpages) {
381 set_pte((pte_t *)pud,
382 pfn_pte(addr >> PAGE_SHIFT, PAGE_KERNEL_LARGE));
383 last_map_addr = (addr & PUD_MASK) + PUD_SIZE;
354 continue; 384 continue;
355 } 385 }
356 386
@@ -358,12 +388,14 @@ phys_pud_init(pud_t *pud_page, unsigned long addr, unsigned long end)
358 388
359 spin_lock(&init_mm.page_table_lock); 389 spin_lock(&init_mm.page_table_lock);
360 set_pud(pud, __pud(pmd_phys | _KERNPG_TABLE)); 390 set_pud(pud, __pud(pmd_phys | _KERNPG_TABLE));
361 phys_pmd_init(pmd, addr, end); 391 last_map_addr = phys_pmd_init(pmd, addr, end);
362 spin_unlock(&init_mm.page_table_lock); 392 spin_unlock(&init_mm.page_table_lock);
363 393
364 unmap_low_page(pmd); 394 unmap_low_page(pmd);
365 } 395 }
366 __flush_tlb_all(); 396 __flush_tlb_all();
397
398 return last_map_addr >> PAGE_SHIFT;
367} 399}
368 400
369static void __init find_early_table_space(unsigned long end) 401static void __init find_early_table_space(unsigned long end)
@@ -371,9 +403,11 @@ static void __init find_early_table_space(unsigned long end)
371 unsigned long puds, pmds, tables, start; 403 unsigned long puds, pmds, tables, start;
372 404
373 puds = (end + PUD_SIZE - 1) >> PUD_SHIFT; 405 puds = (end + PUD_SIZE - 1) >> PUD_SHIFT;
374 pmds = (end + PMD_SIZE - 1) >> PMD_SHIFT; 406 tables = round_up(puds * sizeof(pud_t), PAGE_SIZE);
375 tables = round_up(puds * sizeof(pud_t), PAGE_SIZE) + 407 if (!direct_gbpages) {
376 round_up(pmds * sizeof(pmd_t), PAGE_SIZE); 408 pmds = (end + PMD_SIZE - 1) >> PMD_SHIFT;
409 tables += round_up(pmds * sizeof(pmd_t), PAGE_SIZE);
410 }
377 411
378 /* 412 /*
379 * RED-PEN putting page tables only on node 0 could 413 * RED-PEN putting page tables only on node 0 could
@@ -393,16 +427,135 @@ static void __init find_early_table_space(unsigned long end)
393 (table_start << PAGE_SHIFT) + tables); 427 (table_start << PAGE_SHIFT) + tables);
394} 428}
395 429
430static void __init init_gbpages(void)
431{
432 if (direct_gbpages && cpu_has_gbpages)
433 printk(KERN_INFO "Using GB pages for direct mapping\n");
434 else
435 direct_gbpages = 0;
436}
437
438#ifdef CONFIG_MEMTEST_BOOTPARAM
439
440static void __init memtest(unsigned long start_phys, unsigned long size,
441 unsigned pattern)
442{
443 unsigned long i;
444 unsigned long *start;
445 unsigned long start_bad;
446 unsigned long last_bad;
447 unsigned long val;
448 unsigned long start_phys_aligned;
449 unsigned long count;
450 unsigned long incr;
451
452 switch (pattern) {
453 case 0:
454 val = 0UL;
455 break;
456 case 1:
457 val = -1UL;
458 break;
459 case 2:
460 val = 0x5555555555555555UL;
461 break;
462 case 3:
463 val = 0xaaaaaaaaaaaaaaaaUL;
464 break;
465 default:
466 return;
467 }
468
469 incr = sizeof(unsigned long);
470 start_phys_aligned = ALIGN(start_phys, incr);
471 count = (size - (start_phys_aligned - start_phys))/incr;
472 start = __va(start_phys_aligned);
473 start_bad = 0;
474 last_bad = 0;
475
476 for (i = 0; i < count; i++)
477 start[i] = val;
478 for (i = 0; i < count; i++, start++, start_phys_aligned += incr) {
479 if (*start != val) {
480 if (start_phys_aligned == last_bad + incr) {
481 last_bad += incr;
482 } else {
483 if (start_bad) {
484 printk(KERN_CONT "\n %016lx bad mem addr %016lx - %016lx reserved",
485 val, start_bad, last_bad + incr);
486 reserve_early(start_bad, last_bad - start_bad, "BAD RAM");
487 }
488 start_bad = last_bad = start_phys_aligned;
489 }
490 }
491 }
492 if (start_bad) {
493 printk(KERN_CONT "\n %016lx bad mem addr %016lx - %016lx reserved",
494 val, start_bad, last_bad + incr);
495 reserve_early(start_bad, last_bad - start_bad, "BAD RAM");
496 }
497
498}
499
500static int memtest_pattern __initdata = CONFIG_MEMTEST_BOOTPARAM_VALUE;
501
502static int __init parse_memtest(char *arg)
503{
504 if (arg)
505 memtest_pattern = simple_strtoul(arg, NULL, 0);
506 return 0;
507}
508
509early_param("memtest", parse_memtest);
510
511static void __init early_memtest(unsigned long start, unsigned long end)
512{
513 unsigned long t_start, t_size;
514 unsigned pattern;
515
516 if (!memtest_pattern)
517 return;
518
519 printk(KERN_INFO "early_memtest: pattern num %d", memtest_pattern);
520 for (pattern = 0; pattern < memtest_pattern; pattern++) {
521 t_start = start;
522 t_size = 0;
523 while (t_start < end) {
524 t_start = find_e820_area_size(t_start, &t_size, 1);
525
526 /* done ? */
527 if (t_start >= end)
528 break;
529 if (t_start + t_size > end)
530 t_size = end - t_start;
531
532 printk(KERN_CONT "\n %016lx - %016lx pattern %d",
533 t_start, t_start + t_size, pattern);
534
535 memtest(t_start, t_size, pattern);
536
537 t_start += t_size;
538 }
539 }
540 printk(KERN_CONT "\n");
541}
542#else
543static void __init early_memtest(unsigned long start, unsigned long end)
544{
545}
546#endif
547
396/* 548/*
397 * Setup the direct mapping of the physical memory at PAGE_OFFSET. 549 * Setup the direct mapping of the physical memory at PAGE_OFFSET.
398 * This runs before bootmem is initialized and gets pages directly from 550 * This runs before bootmem is initialized and gets pages directly from
399 * the physical memory. To access them they are temporarily mapped. 551 * the physical memory. To access them they are temporarily mapped.
400 */ 552 */
401void __init_refok init_memory_mapping(unsigned long start, unsigned long end) 553unsigned long __init_refok init_memory_mapping(unsigned long start, unsigned long end)
402{ 554{
403 unsigned long next; 555 unsigned long next, last_map_addr = end;
556 unsigned long start_phys = start, end_phys = end;
404 557
405 pr_debug("init_memory_mapping\n"); 558 printk(KERN_INFO "init_memory_mapping\n");
406 559
407 /* 560 /*
408 * Find space for the kernel direct mapping tables. 561 * Find space for the kernel direct mapping tables.
@@ -411,8 +564,10 @@ void __init_refok init_memory_mapping(unsigned long start, unsigned long end)
411 * memory mapped. Unfortunately this is done currently before the 564 * memory mapped. Unfortunately this is done currently before the
412 * nodes are discovered. 565 * nodes are discovered.
413 */ 566 */
414 if (!after_bootmem) 567 if (!after_bootmem) {
568 init_gbpages();
415 find_early_table_space(end); 569 find_early_table_space(end);
570 }
416 571
417 start = (unsigned long)__va(start); 572 start = (unsigned long)__va(start);
418 end = (unsigned long)__va(end); 573 end = (unsigned long)__va(end);
@@ -430,7 +585,7 @@ void __init_refok init_memory_mapping(unsigned long start, unsigned long end)
430 next = start + PGDIR_SIZE; 585 next = start + PGDIR_SIZE;
431 if (next > end) 586 if (next > end)
432 next = end; 587 next = end;
433 phys_pud_init(pud, __pa(start), __pa(next)); 588 last_map_addr = phys_pud_init(pud, __pa(start), __pa(next));
434 if (!after_bootmem) 589 if (!after_bootmem)
435 set_pgd(pgd_offset_k(start), mk_kernel_pgd(pud_phys)); 590 set_pgd(pgd_offset_k(start), mk_kernel_pgd(pud_phys));
436 unmap_low_page(pud); 591 unmap_low_page(pud);
@@ -443,6 +598,11 @@ void __init_refok init_memory_mapping(unsigned long start, unsigned long end)
443 if (!after_bootmem) 598 if (!after_bootmem)
444 reserve_early(table_start << PAGE_SHIFT, 599 reserve_early(table_start << PAGE_SHIFT,
445 table_end << PAGE_SHIFT, "PGTABLE"); 600 table_end << PAGE_SHIFT, "PGTABLE");
601
602 if (!after_bootmem)
603 early_memtest(start_phys, end_phys);
604
605 return last_map_addr;
446} 606}
447 607
448#ifndef CONFIG_NUMA 608#ifndef CONFIG_NUMA
@@ -482,11 +642,13 @@ int arch_add_memory(int nid, u64 start, u64 size)
482{ 642{
483 struct pglist_data *pgdat = NODE_DATA(nid); 643 struct pglist_data *pgdat = NODE_DATA(nid);
484 struct zone *zone = pgdat->node_zones + ZONE_NORMAL; 644 struct zone *zone = pgdat->node_zones + ZONE_NORMAL;
485 unsigned long start_pfn = start >> PAGE_SHIFT; 645 unsigned long last_mapped_pfn, start_pfn = start >> PAGE_SHIFT;
486 unsigned long nr_pages = size >> PAGE_SHIFT; 646 unsigned long nr_pages = size >> PAGE_SHIFT;
487 int ret; 647 int ret;
488 648
489 init_memory_mapping(start, start + size-1); 649 last_mapped_pfn = init_memory_mapping(start, start + size-1);
650 if (last_mapped_pfn > max_pfn_mapped)
651 max_pfn_mapped = last_mapped_pfn;
490 652
491 ret = __add_pages(zone, start_pfn, nr_pages); 653 ret = __add_pages(zone, start_pfn, nr_pages);
492 WARN_ON(1); 654 WARN_ON(1);
@@ -596,24 +758,7 @@ EXPORT_SYMBOL_GPL(rodata_test_data);
596 758
597void mark_rodata_ro(void) 759void mark_rodata_ro(void)
598{ 760{
599 unsigned long start = (unsigned long)_stext, end; 761 unsigned long start = PFN_ALIGN(_stext), end = PFN_ALIGN(__end_rodata);
600
601#ifdef CONFIG_HOTPLUG_CPU
602 /* It must still be possible to apply SMP alternatives. */
603 if (num_possible_cpus() > 1)
604 start = (unsigned long)_etext;
605#endif
606
607#ifdef CONFIG_KPROBES
608 start = (unsigned long)__start_rodata;
609#endif
610
611 end = (unsigned long)__end_rodata;
612 start = (start + PAGE_SIZE - 1) & PAGE_MASK;
613 end &= PAGE_MASK;
614 if (end <= start)
615 return;
616
617 762
618 printk(KERN_INFO "Write protecting the kernel read-only data: %luk\n", 763 printk(KERN_INFO "Write protecting the kernel read-only data: %luk\n",
619 (end - start) >> 10); 764 (end - start) >> 10);
@@ -636,6 +781,7 @@ void mark_rodata_ro(void)
636 set_memory_ro(start, (end-start) >> PAGE_SHIFT); 781 set_memory_ro(start, (end-start) >> PAGE_SHIFT);
637#endif 782#endif
638} 783}
784
639#endif 785#endif
640 786
641#ifdef CONFIG_BLK_DEV_INITRD 787#ifdef CONFIG_BLK_DEV_INITRD
@@ -657,7 +803,7 @@ void __init reserve_bootmem_generic(unsigned long phys, unsigned len)
657 * This can happen with kdump kernels when accessing 803 * This can happen with kdump kernels when accessing
658 * firmware tables: 804 * firmware tables:
659 */ 805 */
660 if (pfn < end_pfn_map) 806 if (pfn < max_pfn_mapped)
661 return; 807 return;
662 808
663 printk(KERN_ERR "reserve_bootmem: illegal reserve %lx %u\n", 809 printk(KERN_ERR "reserve_bootmem: illegal reserve %lx %u\n",
diff --git a/arch/x86/mm/ioremap.c b/arch/x86/mm/ioremap.c
index 794895c6dcc9..c590fd200e29 100644
--- a/arch/x86/mm/ioremap.c
+++ b/arch/x86/mm/ioremap.c
@@ -19,11 +19,7 @@
19#include <asm/pgtable.h> 19#include <asm/pgtable.h>
20#include <asm/tlbflush.h> 20#include <asm/tlbflush.h>
21#include <asm/pgalloc.h> 21#include <asm/pgalloc.h>
22 22#include <asm/pat.h>
23enum ioremap_mode {
24 IOR_MODE_UNCACHED,
25 IOR_MODE_CACHED,
26};
27 23
28#ifdef CONFIG_X86_64 24#ifdef CONFIG_X86_64
29 25
@@ -35,11 +31,23 @@ unsigned long __phys_addr(unsigned long x)
35} 31}
36EXPORT_SYMBOL(__phys_addr); 32EXPORT_SYMBOL(__phys_addr);
37 33
34static inline int phys_addr_valid(unsigned long addr)
35{
36 return addr < (1UL << boot_cpu_data.x86_phys_bits);
37}
38
39#else
40
41static inline int phys_addr_valid(unsigned long addr)
42{
43 return 1;
44}
45
38#endif 46#endif
39 47
40int page_is_ram(unsigned long pagenr) 48int page_is_ram(unsigned long pagenr)
41{ 49{
42 unsigned long addr, end; 50 resource_size_t addr, end;
43 int i; 51 int i;
44 52
45 /* 53 /*
@@ -78,19 +86,22 @@ int page_is_ram(unsigned long pagenr)
78 * Fix up the linear direct mapping of the kernel to avoid cache attribute 86 * Fix up the linear direct mapping of the kernel to avoid cache attribute
79 * conflicts. 87 * conflicts.
80 */ 88 */
81static int ioremap_change_attr(unsigned long vaddr, unsigned long size, 89int ioremap_change_attr(unsigned long vaddr, unsigned long size,
82 enum ioremap_mode mode) 90 unsigned long prot_val)
83{ 91{
84 unsigned long nrpages = size >> PAGE_SHIFT; 92 unsigned long nrpages = size >> PAGE_SHIFT;
85 int err; 93 int err;
86 94
87 switch (mode) { 95 switch (prot_val) {
88 case IOR_MODE_UNCACHED: 96 case _PAGE_CACHE_UC:
89 default: 97 default:
90 err = set_memory_uc(vaddr, nrpages); 98 err = _set_memory_uc(vaddr, nrpages);
99 break;
100 case _PAGE_CACHE_WC:
101 err = _set_memory_wc(vaddr, nrpages);
91 break; 102 break;
92 case IOR_MODE_CACHED: 103 case _PAGE_CACHE_WB:
93 err = set_memory_wb(vaddr, nrpages); 104 err = _set_memory_wb(vaddr, nrpages);
94 break; 105 break;
95 } 106 }
96 107
@@ -107,17 +118,27 @@ static int ioremap_change_attr(unsigned long vaddr, unsigned long size,
107 * caller shouldn't need to know that small detail. 118 * caller shouldn't need to know that small detail.
108 */ 119 */
109static void __iomem *__ioremap(resource_size_t phys_addr, unsigned long size, 120static void __iomem *__ioremap(resource_size_t phys_addr, unsigned long size,
110 enum ioremap_mode mode) 121 unsigned long prot_val)
111{ 122{
112 unsigned long pfn, offset, last_addr, vaddr; 123 unsigned long pfn, offset, vaddr;
124 resource_size_t last_addr;
113 struct vm_struct *area; 125 struct vm_struct *area;
126 unsigned long new_prot_val;
114 pgprot_t prot; 127 pgprot_t prot;
128 int retval;
115 129
116 /* Don't allow wraparound or zero size */ 130 /* Don't allow wraparound or zero size */
117 last_addr = phys_addr + size - 1; 131 last_addr = phys_addr + size - 1;
118 if (!size || last_addr < phys_addr) 132 if (!size || last_addr < phys_addr)
119 return NULL; 133 return NULL;
120 134
135 if (!phys_addr_valid(phys_addr)) {
136 printk(KERN_WARNING "ioremap: invalid physical address %llx\n",
137 phys_addr);
138 WARN_ON_ONCE(1);
139 return NULL;
140 }
141
121 /* 142 /*
122 * Don't remap the low PCI/ISA area, it's always mapped.. 143 * Don't remap the low PCI/ISA area, it's always mapped..
123 */ 144 */
@@ -127,25 +148,14 @@ static void __iomem *__ioremap(resource_size_t phys_addr, unsigned long size,
127 /* 148 /*
128 * Don't allow anybody to remap normal RAM that we're using.. 149 * Don't allow anybody to remap normal RAM that we're using..
129 */ 150 */
130 for (pfn = phys_addr >> PAGE_SHIFT; pfn < max_pfn_mapped && 151 for (pfn = phys_addr >> PAGE_SHIFT;
131 (pfn << PAGE_SHIFT) < last_addr; pfn++) { 152 (pfn << PAGE_SHIFT) < last_addr; pfn++) {
132 if (page_is_ram(pfn) && pfn_valid(pfn) &&
133 !PageReserved(pfn_to_page(pfn)))
134 return NULL;
135 }
136 153
137 switch (mode) { 154 int is_ram = page_is_ram(pfn);
138 case IOR_MODE_UNCACHED: 155
139 default: 156 if (is_ram && pfn_valid(pfn) && !PageReserved(pfn_to_page(pfn)))
140 /* 157 return NULL;
141 * FIXME: we will use UC MINUS for now, as video fb drivers 158 WARN_ON_ONCE(is_ram);
142 * depend on it. Upcoming ioremap_wc() will fix this behavior.
143 */
144 prot = PAGE_KERNEL_UC_MINUS;
145 break;
146 case IOR_MODE_CACHED:
147 prot = PAGE_KERNEL;
148 break;
149 } 159 }
150 160
151 /* 161 /*
@@ -155,6 +165,49 @@ static void __iomem *__ioremap(resource_size_t phys_addr, unsigned long size,
155 phys_addr &= PAGE_MASK; 165 phys_addr &= PAGE_MASK;
156 size = PAGE_ALIGN(last_addr+1) - phys_addr; 166 size = PAGE_ALIGN(last_addr+1) - phys_addr;
157 167
168 retval = reserve_memtype(phys_addr, phys_addr + size,
169 prot_val, &new_prot_val);
170 if (retval) {
171 pr_debug("Warning: reserve_memtype returned %d\n", retval);
172 return NULL;
173 }
174
175 if (prot_val != new_prot_val) {
176 /*
177 * Do not fallback to certain memory types with certain
178 * requested type:
179 * - request is uncached, return cannot be write-back
180 * - request is uncached, return cannot be write-combine
181 * - request is write-combine, return cannot be write-back
182 */
183 if ((prot_val == _PAGE_CACHE_UC &&
184 (new_prot_val == _PAGE_CACHE_WB ||
185 new_prot_val == _PAGE_CACHE_WC)) ||
186 (prot_val == _PAGE_CACHE_WC &&
187 new_prot_val == _PAGE_CACHE_WB)) {
188 pr_debug(
189 "ioremap error for 0x%llx-0x%llx, requested 0x%lx, got 0x%lx\n",
190 phys_addr, phys_addr + size,
191 prot_val, new_prot_val);
192 free_memtype(phys_addr, phys_addr + size);
193 return NULL;
194 }
195 prot_val = new_prot_val;
196 }
197
198 switch (prot_val) {
199 case _PAGE_CACHE_UC:
200 default:
201 prot = PAGE_KERNEL_NOCACHE;
202 break;
203 case _PAGE_CACHE_WC:
204 prot = PAGE_KERNEL_WC;
205 break;
206 case _PAGE_CACHE_WB:
207 prot = PAGE_KERNEL;
208 break;
209 }
210
158 /* 211 /*
159 * Ok, go for it.. 212 * Ok, go for it..
160 */ 213 */
@@ -164,11 +217,13 @@ static void __iomem *__ioremap(resource_size_t phys_addr, unsigned long size,
164 area->phys_addr = phys_addr; 217 area->phys_addr = phys_addr;
165 vaddr = (unsigned long) area->addr; 218 vaddr = (unsigned long) area->addr;
166 if (ioremap_page_range(vaddr, vaddr + size, phys_addr, prot)) { 219 if (ioremap_page_range(vaddr, vaddr + size, phys_addr, prot)) {
220 free_memtype(phys_addr, phys_addr + size);
167 free_vm_area(area); 221 free_vm_area(area);
168 return NULL; 222 return NULL;
169 } 223 }
170 224
171 if (ioremap_change_attr(vaddr, size, mode) < 0) { 225 if (ioremap_change_attr(vaddr, size, prot_val) < 0) {
226 free_memtype(phys_addr, phys_addr + size);
172 vunmap(area->addr); 227 vunmap(area->addr);
173 return NULL; 228 return NULL;
174 } 229 }
@@ -199,13 +254,32 @@ static void __iomem *__ioremap(resource_size_t phys_addr, unsigned long size,
199 */ 254 */
200void __iomem *ioremap_nocache(resource_size_t phys_addr, unsigned long size) 255void __iomem *ioremap_nocache(resource_size_t phys_addr, unsigned long size)
201{ 256{
202 return __ioremap(phys_addr, size, IOR_MODE_UNCACHED); 257 return __ioremap(phys_addr, size, _PAGE_CACHE_UC);
203} 258}
204EXPORT_SYMBOL(ioremap_nocache); 259EXPORT_SYMBOL(ioremap_nocache);
205 260
261/**
262 * ioremap_wc - map memory into CPU space write combined
263 * @offset: bus address of the memory
264 * @size: size of the resource to map
265 *
266 * This version of ioremap ensures that the memory is marked write combining.
267 * Write combining allows faster writes to some hardware devices.
268 *
269 * Must be freed with iounmap.
270 */
271void __iomem *ioremap_wc(unsigned long phys_addr, unsigned long size)
272{
273 if (pat_wc_enabled)
274 return __ioremap(phys_addr, size, _PAGE_CACHE_WC);
275 else
276 return ioremap_nocache(phys_addr, size);
277}
278EXPORT_SYMBOL(ioremap_wc);
279
206void __iomem *ioremap_cache(resource_size_t phys_addr, unsigned long size) 280void __iomem *ioremap_cache(resource_size_t phys_addr, unsigned long size)
207{ 281{
208 return __ioremap(phys_addr, size, IOR_MODE_CACHED); 282 return __ioremap(phys_addr, size, _PAGE_CACHE_WB);
209} 283}
210EXPORT_SYMBOL(ioremap_cache); 284EXPORT_SYMBOL(ioremap_cache);
211 285
@@ -252,6 +326,8 @@ void iounmap(volatile void __iomem *addr)
252 return; 326 return;
253 } 327 }
254 328
329 free_memtype(p->phys_addr, p->phys_addr + get_vm_area_size(p));
330
255 /* Finally remove it */ 331 /* Finally remove it */
256 o = remove_vm_area((void *)addr); 332 o = remove_vm_area((void *)addr);
257 BUG_ON(p != o || o == NULL); 333 BUG_ON(p != o || o == NULL);
@@ -272,8 +348,8 @@ static int __init early_ioremap_debug_setup(char *str)
272early_param("early_ioremap_debug", early_ioremap_debug_setup); 348early_param("early_ioremap_debug", early_ioremap_debug_setup);
273 349
274static __initdata int after_paging_init; 350static __initdata int after_paging_init;
275static __initdata pte_t bm_pte[PAGE_SIZE/sizeof(pte_t)] 351static pte_t bm_pte[PAGE_SIZE/sizeof(pte_t)]
276 __attribute__((aligned(PAGE_SIZE))); 352 __section(.bss.page_aligned);
277 353
278static inline pmd_t * __init early_ioremap_pmd(unsigned long addr) 354static inline pmd_t * __init early_ioremap_pmd(unsigned long addr)
279{ 355{
diff --git a/arch/x86/mm/numa_64.c b/arch/x86/mm/numa_64.c
index 16b82ad34b96..2ea56f48f29b 100644
--- a/arch/x86/mm/numa_64.c
+++ b/arch/x86/mm/numa_64.c
@@ -31,13 +31,15 @@ bootmem_data_t plat_node_bdata[MAX_NUMNODES];
31 31
32struct memnode memnode; 32struct memnode memnode;
33 33
34#ifdef CONFIG_SMP
34int x86_cpu_to_node_map_init[NR_CPUS] = { 35int x86_cpu_to_node_map_init[NR_CPUS] = {
35 [0 ... NR_CPUS-1] = NUMA_NO_NODE 36 [0 ... NR_CPUS-1] = NUMA_NO_NODE
36}; 37};
37void *x86_cpu_to_node_map_early_ptr; 38void *x86_cpu_to_node_map_early_ptr;
39EXPORT_SYMBOL(x86_cpu_to_node_map_early_ptr);
40#endif
38DEFINE_PER_CPU(int, x86_cpu_to_node_map) = NUMA_NO_NODE; 41DEFINE_PER_CPU(int, x86_cpu_to_node_map) = NUMA_NO_NODE;
39EXPORT_PER_CPU_SYMBOL(x86_cpu_to_node_map); 42EXPORT_PER_CPU_SYMBOL(x86_cpu_to_node_map);
40EXPORT_SYMBOL(x86_cpu_to_node_map_early_ptr);
41 43
42s16 apicid_to_node[MAX_LOCAL_APIC] __cpuinitdata = { 44s16 apicid_to_node[MAX_LOCAL_APIC] __cpuinitdata = {
43 [0 ... MAX_LOCAL_APIC-1] = NUMA_NO_NODE 45 [0 ... MAX_LOCAL_APIC-1] = NUMA_NO_NODE
@@ -548,8 +550,6 @@ void __cpuinit numa_set_node(int cpu, int node)
548{ 550{
549 int *cpu_to_node_map = x86_cpu_to_node_map_early_ptr; 551 int *cpu_to_node_map = x86_cpu_to_node_map_early_ptr;
550 552
551 cpu_pda(cpu)->nodenumber = node;
552
553 if(cpu_to_node_map) 553 if(cpu_to_node_map)
554 cpu_to_node_map[cpu] = node; 554 cpu_to_node_map[cpu] = node;
555 else if(per_cpu_offset(cpu)) 555 else if(per_cpu_offset(cpu))
diff --git a/arch/x86/mm/pageattr.c b/arch/x86/mm/pageattr.c
index 7b79f6be4e7d..7d9517abc9af 100644
--- a/arch/x86/mm/pageattr.c
+++ b/arch/x86/mm/pageattr.c
@@ -9,6 +9,8 @@
9#include <linux/slab.h> 9#include <linux/slab.h>
10#include <linux/mm.h> 10#include <linux/mm.h>
11#include <linux/interrupt.h> 11#include <linux/interrupt.h>
12#include <linux/seq_file.h>
13#include <linux/debugfs.h>
12 14
13#include <asm/e820.h> 15#include <asm/e820.h>
14#include <asm/processor.h> 16#include <asm/processor.h>
@@ -17,6 +19,7 @@
17#include <asm/uaccess.h> 19#include <asm/uaccess.h>
18#include <asm/pgalloc.h> 20#include <asm/pgalloc.h>
19#include <asm/proto.h> 21#include <asm/proto.h>
22#include <asm/pat.h>
20 23
21/* 24/*
22 * The current flushing context - we pass it instead of 5 arguments: 25 * The current flushing context - we pass it instead of 5 arguments:
@@ -28,6 +31,7 @@ struct cpa_data {
28 int numpages; 31 int numpages;
29 int flushtlb; 32 int flushtlb;
30 unsigned long pfn; 33 unsigned long pfn;
34 unsigned force_split : 1;
31}; 35};
32 36
33#ifdef CONFIG_X86_64 37#ifdef CONFIG_X86_64
@@ -259,6 +263,9 @@ try_preserve_large_page(pte_t *kpte, unsigned long address,
259 int i, do_split = 1; 263 int i, do_split = 1;
260 unsigned int level; 264 unsigned int level;
261 265
266 if (cpa->force_split)
267 return 1;
268
262 spin_lock_irqsave(&pgd_lock, flags); 269 spin_lock_irqsave(&pgd_lock, flags);
263 /* 270 /*
264 * Check for races, another CPU might have split this page 271 * Check for races, another CPU might have split this page
@@ -693,7 +700,8 @@ static inline int cache_attr(pgprot_t attr)
693} 700}
694 701
695static int change_page_attr_set_clr(unsigned long addr, int numpages, 702static int change_page_attr_set_clr(unsigned long addr, int numpages,
696 pgprot_t mask_set, pgprot_t mask_clr) 703 pgprot_t mask_set, pgprot_t mask_clr,
704 int force_split)
697{ 705{
698 struct cpa_data cpa; 706 struct cpa_data cpa;
699 int ret, cache, checkalias; 707 int ret, cache, checkalias;
@@ -704,7 +712,7 @@ static int change_page_attr_set_clr(unsigned long addr, int numpages,
704 */ 712 */
705 mask_set = canon_pgprot(mask_set); 713 mask_set = canon_pgprot(mask_set);
706 mask_clr = canon_pgprot(mask_clr); 714 mask_clr = canon_pgprot(mask_clr);
707 if (!pgprot_val(mask_set) && !pgprot_val(mask_clr)) 715 if (!pgprot_val(mask_set) && !pgprot_val(mask_clr) && !force_split)
708 return 0; 716 return 0;
709 717
710 /* Ensure we are PAGE_SIZE aligned */ 718 /* Ensure we are PAGE_SIZE aligned */
@@ -721,6 +729,7 @@ static int change_page_attr_set_clr(unsigned long addr, int numpages,
721 cpa.mask_set = mask_set; 729 cpa.mask_set = mask_set;
722 cpa.mask_clr = mask_clr; 730 cpa.mask_clr = mask_clr;
723 cpa.flushtlb = 0; 731 cpa.flushtlb = 0;
732 cpa.force_split = force_split;
724 733
725 /* No alias checking for _NX bit modifications */ 734 /* No alias checking for _NX bit modifications */
726 checkalias = (pgprot_val(mask_set) | pgprot_val(mask_clr)) != _PAGE_NX; 735 checkalias = (pgprot_val(mask_set) | pgprot_val(mask_clr)) != _PAGE_NX;
@@ -759,26 +768,61 @@ out:
759static inline int change_page_attr_set(unsigned long addr, int numpages, 768static inline int change_page_attr_set(unsigned long addr, int numpages,
760 pgprot_t mask) 769 pgprot_t mask)
761{ 770{
762 return change_page_attr_set_clr(addr, numpages, mask, __pgprot(0)); 771 return change_page_attr_set_clr(addr, numpages, mask, __pgprot(0), 0);
763} 772}
764 773
765static inline int change_page_attr_clear(unsigned long addr, int numpages, 774static inline int change_page_attr_clear(unsigned long addr, int numpages,
766 pgprot_t mask) 775 pgprot_t mask)
767{ 776{
768 return change_page_attr_set_clr(addr, numpages, __pgprot(0), mask); 777 return change_page_attr_set_clr(addr, numpages, __pgprot(0), mask, 0);
769} 778}
770 779
771int set_memory_uc(unsigned long addr, int numpages) 780int _set_memory_uc(unsigned long addr, int numpages)
772{ 781{
773 return change_page_attr_set(addr, numpages, 782 return change_page_attr_set(addr, numpages,
774 __pgprot(_PAGE_PCD)); 783 __pgprot(_PAGE_CACHE_UC));
784}
785
786int set_memory_uc(unsigned long addr, int numpages)
787{
788 if (reserve_memtype(addr, addr + numpages * PAGE_SIZE,
789 _PAGE_CACHE_UC, NULL))
790 return -EINVAL;
791
792 return _set_memory_uc(addr, numpages);
775} 793}
776EXPORT_SYMBOL(set_memory_uc); 794EXPORT_SYMBOL(set_memory_uc);
777 795
778int set_memory_wb(unsigned long addr, int numpages) 796int _set_memory_wc(unsigned long addr, int numpages)
797{
798 return change_page_attr_set(addr, numpages,
799 __pgprot(_PAGE_CACHE_WC));
800}
801
802int set_memory_wc(unsigned long addr, int numpages)
803{
804 if (!pat_wc_enabled)
805 return set_memory_uc(addr, numpages);
806
807 if (reserve_memtype(addr, addr + numpages * PAGE_SIZE,
808 _PAGE_CACHE_WC, NULL))
809 return -EINVAL;
810
811 return _set_memory_wc(addr, numpages);
812}
813EXPORT_SYMBOL(set_memory_wc);
814
815int _set_memory_wb(unsigned long addr, int numpages)
779{ 816{
780 return change_page_attr_clear(addr, numpages, 817 return change_page_attr_clear(addr, numpages,
781 __pgprot(_PAGE_PCD | _PAGE_PWT)); 818 __pgprot(_PAGE_CACHE_MASK));
819}
820
821int set_memory_wb(unsigned long addr, int numpages)
822{
823 free_memtype(addr, addr + numpages * PAGE_SIZE);
824
825 return _set_memory_wb(addr, numpages);
782} 826}
783EXPORT_SYMBOL(set_memory_wb); 827EXPORT_SYMBOL(set_memory_wb);
784 828
@@ -809,6 +853,12 @@ int set_memory_np(unsigned long addr, int numpages)
809 return change_page_attr_clear(addr, numpages, __pgprot(_PAGE_PRESENT)); 853 return change_page_attr_clear(addr, numpages, __pgprot(_PAGE_PRESENT));
810} 854}
811 855
856int set_memory_4k(unsigned long addr, int numpages)
857{
858 return change_page_attr_set_clr(addr, numpages, __pgprot(0),
859 __pgprot(0), 1);
860}
861
812int set_pages_uc(struct page *page, int numpages) 862int set_pages_uc(struct page *page, int numpages)
813{ 863{
814 unsigned long addr = (unsigned long)page_address(page); 864 unsigned long addr = (unsigned long)page_address(page);
@@ -918,6 +968,45 @@ void kernel_map_pages(struct page *page, int numpages, int enable)
918 cpa_fill_pool(NULL); 968 cpa_fill_pool(NULL);
919} 969}
920 970
971#ifdef CONFIG_DEBUG_FS
972static int dpa_show(struct seq_file *m, void *v)
973{
974 seq_puts(m, "DEBUG_PAGEALLOC\n");
975 seq_printf(m, "pool_size : %lu\n", pool_size);
976 seq_printf(m, "pool_pages : %lu\n", pool_pages);
977 seq_printf(m, "pool_low : %lu\n", pool_low);
978 seq_printf(m, "pool_used : %lu\n", pool_used);
979 seq_printf(m, "pool_failed : %lu\n", pool_failed);
980
981 return 0;
982}
983
984static int dpa_open(struct inode *inode, struct file *filp)
985{
986 return single_open(filp, dpa_show, NULL);
987}
988
989static const struct file_operations dpa_fops = {
990 .open = dpa_open,
991 .read = seq_read,
992 .llseek = seq_lseek,
993 .release = single_release,
994};
995
996int __init debug_pagealloc_proc_init(void)
997{
998 struct dentry *de;
999
1000 de = debugfs_create_file("debug_pagealloc", 0600, NULL, NULL,
1001 &dpa_fops);
1002 if (!de)
1003 return -ENOMEM;
1004
1005 return 0;
1006}
1007__initcall(debug_pagealloc_proc_init);
1008#endif
1009
921#ifdef CONFIG_HIBERNATION 1010#ifdef CONFIG_HIBERNATION
922 1011
923bool kernel_page_present(struct page *page) 1012bool kernel_page_present(struct page *page)
diff --git a/arch/x86/mm/pat.c b/arch/x86/mm/pat.c
new file mode 100644
index 000000000000..72c0f6097402
--- /dev/null
+++ b/arch/x86/mm/pat.c
@@ -0,0 +1,421 @@
1/*
2 * Handle caching attributes in page tables (PAT)
3 *
4 * Authors: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
5 * Suresh B Siddha <suresh.b.siddha@intel.com>
6 *
7 * Loosely based on earlier PAT patchset from Eric Biederman and Andi Kleen.
8 */
9
10#include <linux/mm.h>
11#include <linux/kernel.h>
12#include <linux/gfp.h>
13#include <linux/fs.h>
14
15#include <asm/msr.h>
16#include <asm/tlbflush.h>
17#include <asm/processor.h>
18#include <asm/pgtable.h>
19#include <asm/pat.h>
20#include <asm/e820.h>
21#include <asm/cacheflush.h>
22#include <asm/fcntl.h>
23#include <asm/mtrr.h>
24
25int pat_wc_enabled = 1;
26
27static u64 __read_mostly boot_pat_state;
28
29static int nopat(char *str)
30{
31 pat_wc_enabled = 0;
32 printk(KERN_INFO "x86: PAT support disabled.\n");
33
34 return 0;
35}
36early_param("nopat", nopat);
37
38static int pat_known_cpu(void)
39{
40 if (!pat_wc_enabled)
41 return 0;
42
43 if (cpu_has_pat)
44 return 1;
45
46 pat_wc_enabled = 0;
47 printk(KERN_INFO "CPU and/or kernel does not support PAT.\n");
48 return 0;
49}
50
51enum {
52 PAT_UC = 0, /* uncached */
53 PAT_WC = 1, /* Write combining */
54 PAT_WT = 4, /* Write Through */
55 PAT_WP = 5, /* Write Protected */
56 PAT_WB = 6, /* Write Back (default) */
57 PAT_UC_MINUS = 7, /* UC, but can be overriden by MTRR */
58};
59
60#define PAT(x,y) ((u64)PAT_ ## y << ((x)*8))
61
62void pat_init(void)
63{
64 u64 pat;
65
66#ifndef CONFIG_X86_PAT
67 nopat(NULL);
68#endif
69
70 /* Boot CPU enables PAT based on CPU feature */
71 if (!smp_processor_id() && !pat_known_cpu())
72 return;
73
74 /* APs enable PAT iff boot CPU has enabled it before */
75 if (smp_processor_id() && !pat_wc_enabled)
76 return;
77
78 /* Set PWT to Write-Combining. All other bits stay the same */
79 /*
80 * PTE encoding used in Linux:
81 * PAT
82 * |PCD
83 * ||PWT
84 * |||
85 * 000 WB _PAGE_CACHE_WB
86 * 001 WC _PAGE_CACHE_WC
87 * 010 UC- _PAGE_CACHE_UC_MINUS
88 * 011 UC _PAGE_CACHE_UC
89 * PAT bit unused
90 */
91 pat = PAT(0,WB) | PAT(1,WC) | PAT(2,UC_MINUS) | PAT(3,UC) |
92 PAT(4,WB) | PAT(5,WC) | PAT(6,UC_MINUS) | PAT(7,UC);
93
94 /* Boot CPU check */
95 if (!smp_processor_id()) {
96 rdmsrl(MSR_IA32_CR_PAT, boot_pat_state);
97 }
98
99 wrmsrl(MSR_IA32_CR_PAT, pat);
100 printk(KERN_INFO "x86 PAT enabled: cpu %d, old 0x%Lx, new 0x%Lx\n",
101 smp_processor_id(), boot_pat_state, pat);
102}
103
104#undef PAT
105
106static char *cattr_name(unsigned long flags)
107{
108 switch (flags & _PAGE_CACHE_MASK) {
109 case _PAGE_CACHE_UC: return "uncached";
110 case _PAGE_CACHE_UC_MINUS: return "uncached-minus";
111 case _PAGE_CACHE_WB: return "write-back";
112 case _PAGE_CACHE_WC: return "write-combining";
113 default: return "broken";
114 }
115}
116
117/*
118 * The global memtype list keeps track of memory type for specific
119 * physical memory areas. Conflicting memory types in different
120 * mappings can cause CPU cache corruption. To avoid this we keep track.
121 *
122 * The list is sorted based on starting address and can contain multiple
123 * entries for each address (this allows reference counting for overlapping
124 * areas). All the aliases have the same cache attributes of course.
125 * Zero attributes are represented as holes.
126 *
127 * Currently the data structure is a list because the number of mappings
128 * are expected to be relatively small. If this should be a problem
129 * it could be changed to a rbtree or similar.
130 *
131 * memtype_lock protects the whole list.
132 */
133
134struct memtype {
135 u64 start;
136 u64 end;
137 unsigned long type;
138 struct list_head nd;
139};
140
141static LIST_HEAD(memtype_list);
142static DEFINE_SPINLOCK(memtype_lock); /* protects memtype list */
143
144/*
145 * Does intersection of PAT memory type and MTRR memory type and returns
146 * the resulting memory type as PAT understands it.
147 * (Type in pat and mtrr will not have same value)
148 * The intersection is based on "Effective Memory Type" tables in IA-32
149 * SDM vol 3a
150 */
151static int pat_x_mtrr_type(u64 start, u64 end, unsigned long prot,
152 unsigned long *ret_prot)
153{
154 unsigned long pat_type;
155 u8 mtrr_type;
156
157 mtrr_type = mtrr_type_lookup(start, end);
158 if (mtrr_type == 0xFF) { /* MTRR not enabled */
159 *ret_prot = prot;
160 return 0;
161 }
162 if (mtrr_type == 0xFE) { /* MTRR match error */
163 *ret_prot = _PAGE_CACHE_UC;
164 return -1;
165 }
166 if (mtrr_type != MTRR_TYPE_UNCACHABLE &&
167 mtrr_type != MTRR_TYPE_WRBACK &&
168 mtrr_type != MTRR_TYPE_WRCOMB) { /* MTRR type unhandled */
169 *ret_prot = _PAGE_CACHE_UC;
170 return -1;
171 }
172
173 pat_type = prot & _PAGE_CACHE_MASK;
174 prot &= (~_PAGE_CACHE_MASK);
175
176 /* Currently doing intersection by hand. Optimize it later. */
177 if (pat_type == _PAGE_CACHE_WC) {
178 *ret_prot = prot | _PAGE_CACHE_WC;
179 } else if (pat_type == _PAGE_CACHE_UC_MINUS) {
180 *ret_prot = prot | _PAGE_CACHE_UC_MINUS;
181 } else if (pat_type == _PAGE_CACHE_UC ||
182 mtrr_type == MTRR_TYPE_UNCACHABLE) {
183 *ret_prot = prot | _PAGE_CACHE_UC;
184 } else if (mtrr_type == MTRR_TYPE_WRCOMB) {
185 *ret_prot = prot | _PAGE_CACHE_WC;
186 } else {
187 *ret_prot = prot | _PAGE_CACHE_WB;
188 }
189
190 return 0;
191}
192
193int reserve_memtype(u64 start, u64 end, unsigned long req_type,
194 unsigned long *ret_type)
195{
196 struct memtype *new_entry = NULL;
197 struct memtype *parse;
198 unsigned long actual_type;
199 int err = 0;
200
201 /* Only track when pat_wc_enabled */
202 if (!pat_wc_enabled) {
203 if (ret_type)
204 *ret_type = req_type;
205
206 return 0;
207 }
208
209 /* Low ISA region is always mapped WB in page table. No need to track */
210 if (start >= ISA_START_ADDRESS && (end - 1) <= ISA_END_ADDRESS) {
211 if (ret_type)
212 *ret_type = _PAGE_CACHE_WB;
213
214 return 0;
215 }
216
217 req_type &= _PAGE_CACHE_MASK;
218 err = pat_x_mtrr_type(start, end, req_type, &actual_type);
219 if (err) {
220 if (ret_type)
221 *ret_type = actual_type;
222
223 return -EINVAL;
224 }
225
226 new_entry = kmalloc(sizeof(struct memtype), GFP_KERNEL);
227 if (!new_entry)
228 return -ENOMEM;
229
230 new_entry->start = start;
231 new_entry->end = end;
232 new_entry->type = actual_type;
233
234 if (ret_type)
235 *ret_type = actual_type;
236
237 spin_lock(&memtype_lock);
238
239 /* Search for existing mapping that overlaps the current range */
240 list_for_each_entry(parse, &memtype_list, nd) {
241 struct memtype *saved_ptr;
242
243 if (parse->start >= end) {
244 printk("New Entry\n");
245 list_add(&new_entry->nd, parse->nd.prev);
246 new_entry = NULL;
247 break;
248 }
249
250 if (start <= parse->start && end >= parse->start) {
251 if (actual_type != parse->type && ret_type) {
252 actual_type = parse->type;
253 *ret_type = actual_type;
254 new_entry->type = actual_type;
255 }
256
257 if (actual_type != parse->type) {
258 printk(
259 KERN_INFO "%s:%d conflicting memory types %Lx-%Lx %s<->%s\n",
260 current->comm, current->pid,
261 start, end,
262 cattr_name(actual_type),
263 cattr_name(parse->type));
264 err = -EBUSY;
265 break;
266 }
267
268 saved_ptr = parse;
269 /*
270 * Check to see whether the request overlaps more
271 * than one entry in the list
272 */
273 list_for_each_entry_continue(parse, &memtype_list, nd) {
274 if (end <= parse->start) {
275 break;
276 }
277
278 if (actual_type != parse->type) {
279 printk(
280 KERN_INFO "%s:%d conflicting memory types %Lx-%Lx %s<->%s\n",
281 current->comm, current->pid,
282 start, end,
283 cattr_name(actual_type),
284 cattr_name(parse->type));
285 err = -EBUSY;
286 break;
287 }
288 }
289
290 if (err) {
291 break;
292 }
293
294 printk("Overlap at 0x%Lx-0x%Lx\n",
295 saved_ptr->start, saved_ptr->end);
296 /* No conflict. Go ahead and add this new entry */
297 list_add(&new_entry->nd, saved_ptr->nd.prev);
298 new_entry = NULL;
299 break;
300 }
301
302 if (start < parse->end) {
303 if (actual_type != parse->type && ret_type) {
304 actual_type = parse->type;
305 *ret_type = actual_type;
306 new_entry->type = actual_type;
307 }
308
309 if (actual_type != parse->type) {
310 printk(
311 KERN_INFO "%s:%d conflicting memory types %Lx-%Lx %s<->%s\n",
312 current->comm, current->pid,
313 start, end,
314 cattr_name(actual_type),
315 cattr_name(parse->type));
316 err = -EBUSY;
317 break;
318 }
319
320 saved_ptr = parse;
321 /*
322 * Check to see whether the request overlaps more
323 * than one entry in the list
324 */
325 list_for_each_entry_continue(parse, &memtype_list, nd) {
326 if (end <= parse->start) {
327 break;
328 }
329
330 if (actual_type != parse->type) {
331 printk(
332 KERN_INFO "%s:%d conflicting memory types %Lx-%Lx %s<->%s\n",
333 current->comm, current->pid,
334 start, end,
335 cattr_name(actual_type),
336 cattr_name(parse->type));
337 err = -EBUSY;
338 break;
339 }
340 }
341
342 if (err) {
343 break;
344 }
345
346 printk("Overlap at 0x%Lx-0x%Lx\n",
347 saved_ptr->start, saved_ptr->end);
348 /* No conflict. Go ahead and add this new entry */
349 list_add(&new_entry->nd, &saved_ptr->nd);
350 new_entry = NULL;
351 break;
352 }
353 }
354
355 if (err) {
356 printk(
357 "reserve_memtype failed 0x%Lx-0x%Lx, track %s, req %s\n",
358 start, end, cattr_name(new_entry->type),
359 cattr_name(req_type));
360 kfree(new_entry);
361 spin_unlock(&memtype_lock);
362 return err;
363 }
364
365 if (new_entry) {
366 /* No conflict. Not yet added to the list. Add to the tail */
367 list_add_tail(&new_entry->nd, &memtype_list);
368 printk("New Entry\n");
369 }
370
371 if (ret_type) {
372 printk(
373 "reserve_memtype added 0x%Lx-0x%Lx, track %s, req %s, ret %s\n",
374 start, end, cattr_name(actual_type),
375 cattr_name(req_type), cattr_name(*ret_type));
376 } else {
377 printk(
378 "reserve_memtype added 0x%Lx-0x%Lx, track %s, req %s\n",
379 start, end, cattr_name(actual_type),
380 cattr_name(req_type));
381 }
382
383 spin_unlock(&memtype_lock);
384 return err;
385}
386
387int free_memtype(u64 start, u64 end)
388{
389 struct memtype *ml;
390 int err = -EINVAL;
391
392 /* Only track when pat_wc_enabled */
393 if (!pat_wc_enabled) {
394 return 0;
395 }
396
397 /* Low ISA region is always mapped WB. No need to track */
398 if (start >= ISA_START_ADDRESS && end <= ISA_END_ADDRESS) {
399 return 0;
400 }
401
402 spin_lock(&memtype_lock);
403 list_for_each_entry(ml, &memtype_list, nd) {
404 if (ml->start == start && ml->end == end) {
405 list_del(&ml->nd);
406 kfree(ml);
407 err = 0;
408 break;
409 }
410 }
411 spin_unlock(&memtype_lock);
412
413 if (err) {
414 printk(KERN_DEBUG "%s:%d freeing invalid memtype %Lx-%Lx\n",
415 current->comm, current->pid, start, end);
416 }
417
418 printk( "free_memtype request 0x%Lx-0x%Lx\n", start, end);
419 return err;
420}
421
diff --git a/arch/x86/mm/pgtable_32.c b/arch/x86/mm/pgtable_32.c
index 2f9e9afcb9f4..3165ec0672bd 100644
--- a/arch/x86/mm/pgtable_32.c
+++ b/arch/x86/mm/pgtable_32.c
@@ -36,7 +36,6 @@ void show_mem(void)
36 36
37 printk(KERN_INFO "Mem-info:\n"); 37 printk(KERN_INFO "Mem-info:\n");
38 show_free_areas(); 38 show_free_areas();
39 printk(KERN_INFO "Free swap: %6ldkB\n", nr_swap_pages<<(PAGE_SHIFT-10));
40 for_each_online_pgdat(pgdat) { 39 for_each_online_pgdat(pgdat) {
41 pgdat_resize_lock(pgdat, &flags); 40 pgdat_resize_lock(pgdat, &flags);
42 for (i = 0; i < pgdat->node_spanned_pages; ++i) { 41 for (i = 0; i < pgdat->node_spanned_pages; ++i) {
@@ -381,3 +380,10 @@ void __pmd_free_tlb(struct mmu_gather *tlb, pmd_t *pmd)
381} 380}
382 381
383#endif 382#endif
383
384int pmd_bad(pmd_t pmd)
385{
386 WARN_ON_ONCE(pmd_bad_v1(pmd) != pmd_bad_v2(pmd));
387
388 return pmd_bad_v1(pmd);
389}
diff --git a/arch/x86/mm/srat_64.c b/arch/x86/mm/srat_64.c
index 845001c617cc..1bae9c855ceb 100644
--- a/arch/x86/mm/srat_64.c
+++ b/arch/x86/mm/srat_64.c
@@ -20,6 +20,7 @@
20#include <asm/proto.h> 20#include <asm/proto.h>
21#include <asm/numa.h> 21#include <asm/numa.h>
22#include <asm/e820.h> 22#include <asm/e820.h>
23#include <asm/genapic.h>
23 24
24int acpi_numa __initdata; 25int acpi_numa __initdata;
25 26
@@ -132,7 +133,6 @@ acpi_numa_processor_affinity_init(struct acpi_srat_cpu_affinity *pa)
132 int pxm, node; 133 int pxm, node;
133 int apic_id; 134 int apic_id;
134 135
135 apic_id = pa->apic_id;
136 if (srat_disabled()) 136 if (srat_disabled())
137 return; 137 return;
138 if (pa->header.length != sizeof(struct acpi_srat_cpu_affinity)) { 138 if (pa->header.length != sizeof(struct acpi_srat_cpu_affinity)) {
@@ -148,6 +148,11 @@ acpi_numa_processor_affinity_init(struct acpi_srat_cpu_affinity *pa)
148 bad_srat(); 148 bad_srat();
149 return; 149 return;
150 } 150 }
151
152 if (is_uv_system())
153 apic_id = (pa->apic_id << 8) | pa->local_sapic_eid;
154 else
155 apic_id = pa->apic_id;
151 apicid_to_node[apic_id] = node; 156 apicid_to_node[apic_id] = node;
152 acpi_numa = 1; 157 acpi_numa = 1;
153 printk(KERN_INFO "SRAT: PXM %u -> APIC %u -> Node %u\n", 158 printk(KERN_INFO "SRAT: PXM %u -> APIC %u -> Node %u\n",
diff --git a/arch/x86/oprofile/init.c b/arch/x86/oprofile/init.c
index 5341d481d92f..cdfe4c54deca 100644
--- a/arch/x86/oprofile/init.c
+++ b/arch/x86/oprofile/init.c
@@ -10,18 +10,19 @@
10#include <linux/oprofile.h> 10#include <linux/oprofile.h>
11#include <linux/init.h> 11#include <linux/init.h>
12#include <linux/errno.h> 12#include <linux/errno.h>
13 13
14/* We support CPUs that have performance counters like the Pentium Pro 14/*
15 * We support CPUs that have performance counters like the Pentium Pro
15 * with the NMI mode driver. 16 * with the NMI mode driver.
16 */ 17 */
17 18
18extern int op_nmi_init(struct oprofile_operations * ops); 19extern int op_nmi_init(struct oprofile_operations *ops);
19extern int op_nmi_timer_init(struct oprofile_operations * ops); 20extern int op_nmi_timer_init(struct oprofile_operations *ops);
20extern void op_nmi_exit(void); 21extern void op_nmi_exit(void);
21extern void x86_backtrace(struct pt_regs * const regs, unsigned int depth); 22extern void x86_backtrace(struct pt_regs * const regs, unsigned int depth);
22 23
23 24
24int __init oprofile_arch_init(struct oprofile_operations * ops) 25int __init oprofile_arch_init(struct oprofile_operations *ops)
25{ 26{
26 int ret; 27 int ret;
27 28
diff --git a/arch/x86/oprofile/nmi_timer_int.c b/arch/x86/oprofile/nmi_timer_int.c
index 1418e36ae7ab..e3ecb71b5790 100644
--- a/arch/x86/oprofile/nmi_timer_int.c
+++ b/arch/x86/oprofile/nmi_timer_int.c
@@ -17,14 +17,14 @@
17#include <asm/nmi.h> 17#include <asm/nmi.h>
18#include <asm/apic.h> 18#include <asm/apic.h>
19#include <asm/ptrace.h> 19#include <asm/ptrace.h>
20 20
21static int profile_timer_exceptions_notify(struct notifier_block *self, 21static int profile_timer_exceptions_notify(struct notifier_block *self,
22 unsigned long val, void *data) 22 unsigned long val, void *data)
23{ 23{
24 struct die_args *args = (struct die_args *)data; 24 struct die_args *args = (struct die_args *)data;
25 int ret = NOTIFY_DONE; 25 int ret = NOTIFY_DONE;
26 26
27 switch(val) { 27 switch (val) {
28 case DIE_NMI: 28 case DIE_NMI:
29 oprofile_add_sample(args->regs, 0); 29 oprofile_add_sample(args->regs, 0);
30 ret = NOTIFY_STOP; 30 ret = NOTIFY_STOP;
@@ -56,7 +56,7 @@ static void timer_stop(void)
56} 56}
57 57
58 58
59int __init op_nmi_timer_init(struct oprofile_operations * ops) 59int __init op_nmi_timer_init(struct oprofile_operations *ops)
60{ 60{
61 if ((nmi_watchdog != NMI_IO_APIC) || (atomic_read(&nmi_active) <= 0)) 61 if ((nmi_watchdog != NMI_IO_APIC) || (atomic_read(&nmi_active) <= 0))
62 return -ENODEV; 62 return -ENODEV;
diff --git a/arch/x86/oprofile/op_model_athlon.c b/arch/x86/oprofile/op_model_athlon.c
index c3ee43333f26..3d534879a9dc 100644
--- a/arch/x86/oprofile/op_model_athlon.c
+++ b/arch/x86/oprofile/op_model_athlon.c
@@ -1,4 +1,4 @@
1/** 1/*
2 * @file op_model_athlon.h 2 * @file op_model_athlon.h
3 * athlon / K7 / K8 / Family 10h model-specific MSR operations 3 * athlon / K7 / K8 / Family 10h model-specific MSR operations
4 * 4 *
@@ -14,28 +14,28 @@
14#include <asm/ptrace.h> 14#include <asm/ptrace.h>
15#include <asm/msr.h> 15#include <asm/msr.h>
16#include <asm/nmi.h> 16#include <asm/nmi.h>
17 17
18#include "op_x86_model.h" 18#include "op_x86_model.h"
19#include "op_counter.h" 19#include "op_counter.h"
20 20
21#define NUM_COUNTERS 4 21#define NUM_COUNTERS 4
22#define NUM_CONTROLS 4 22#define NUM_CONTROLS 4
23 23
24#define CTR_IS_RESERVED(msrs,c) (msrs->counters[(c)].addr ? 1 : 0) 24#define CTR_IS_RESERVED(msrs, c) (msrs->counters[(c)].addr ? 1 : 0)
25#define CTR_READ(l,h,msrs,c) do {rdmsr(msrs->counters[(c)].addr, (l), (h));} while (0) 25#define CTR_READ(l, h, msrs, c) do {rdmsr(msrs->counters[(c)].addr, (l), (h)); } while (0)
26#define CTR_WRITE(l,msrs,c) do {wrmsr(msrs->counters[(c)].addr, -(unsigned int)(l), -1);} while (0) 26#define CTR_WRITE(l, msrs, c) do {wrmsr(msrs->counters[(c)].addr, -(unsigned int)(l), -1); } while (0)
27#define CTR_OVERFLOWED(n) (!((n) & (1U<<31))) 27#define CTR_OVERFLOWED(n) (!((n) & (1U<<31)))
28 28
29#define CTRL_IS_RESERVED(msrs,c) (msrs->controls[(c)].addr ? 1 : 0) 29#define CTRL_IS_RESERVED(msrs, c) (msrs->controls[(c)].addr ? 1 : 0)
30#define CTRL_READ(l,h,msrs,c) do {rdmsr(msrs->controls[(c)].addr, (l), (h));} while (0) 30#define CTRL_READ(l, h, msrs, c) do {rdmsr(msrs->controls[(c)].addr, (l), (h)); } while (0)
31#define CTRL_WRITE(l,h,msrs,c) do {wrmsr(msrs->controls[(c)].addr, (l), (h));} while (0) 31#define CTRL_WRITE(l, h, msrs, c) do {wrmsr(msrs->controls[(c)].addr, (l), (h)); } while (0)
32#define CTRL_SET_ACTIVE(n) (n |= (1<<22)) 32#define CTRL_SET_ACTIVE(n) (n |= (1<<22))
33#define CTRL_SET_INACTIVE(n) (n &= ~(1<<22)) 33#define CTRL_SET_INACTIVE(n) (n &= ~(1<<22))
34#define CTRL_CLEAR_LO(x) (x &= (1<<21)) 34#define CTRL_CLEAR_LO(x) (x &= (1<<21))
35#define CTRL_CLEAR_HI(x) (x &= 0xfffffcf0) 35#define CTRL_CLEAR_HI(x) (x &= 0xfffffcf0)
36#define CTRL_SET_ENABLE(val) (val |= 1<<20) 36#define CTRL_SET_ENABLE(val) (val |= 1<<20)
37#define CTRL_SET_USR(val,u) (val |= ((u & 1) << 16)) 37#define CTRL_SET_USR(val, u) (val |= ((u & 1) << 16))
38#define CTRL_SET_KERN(val,k) (val |= ((k & 1) << 17)) 38#define CTRL_SET_KERN(val, k) (val |= ((k & 1) << 17))
39#define CTRL_SET_UM(val, m) (val |= (m << 8)) 39#define CTRL_SET_UM(val, m) (val |= (m << 8))
40#define CTRL_SET_EVENT_LOW(val, e) (val |= (e & 0xff)) 40#define CTRL_SET_EVENT_LOW(val, e) (val |= (e & 0xff))
41#define CTRL_SET_EVENT_HIGH(val, e) (val |= ((e >> 8) & 0xf)) 41#define CTRL_SET_EVENT_HIGH(val, e) (val |= ((e >> 8) & 0xf))
@@ -43,19 +43,19 @@
43#define CTRL_SET_GUEST_ONLY(val, h) (val |= ((h & 1) << 8)) 43#define CTRL_SET_GUEST_ONLY(val, h) (val |= ((h & 1) << 8))
44 44
45static unsigned long reset_value[NUM_COUNTERS]; 45static unsigned long reset_value[NUM_COUNTERS];
46 46
47static void athlon_fill_in_addresses(struct op_msrs * const msrs) 47static void athlon_fill_in_addresses(struct op_msrs * const msrs)
48{ 48{
49 int i; 49 int i;
50 50
51 for (i=0; i < NUM_COUNTERS; i++) { 51 for (i = 0; i < NUM_COUNTERS; i++) {
52 if (reserve_perfctr_nmi(MSR_K7_PERFCTR0 + i)) 52 if (reserve_perfctr_nmi(MSR_K7_PERFCTR0 + i))
53 msrs->counters[i].addr = MSR_K7_PERFCTR0 + i; 53 msrs->counters[i].addr = MSR_K7_PERFCTR0 + i;
54 else 54 else
55 msrs->counters[i].addr = 0; 55 msrs->counters[i].addr = 0;
56 } 56 }
57 57
58 for (i=0; i < NUM_CONTROLS; i++) { 58 for (i = 0; i < NUM_CONTROLS; i++) {
59 if (reserve_evntsel_nmi(MSR_K7_EVNTSEL0 + i)) 59 if (reserve_evntsel_nmi(MSR_K7_EVNTSEL0 + i))
60 msrs->controls[i].addr = MSR_K7_EVNTSEL0 + i; 60 msrs->controls[i].addr = MSR_K7_EVNTSEL0 + i;
61 else 61 else
@@ -63,15 +63,15 @@ static void athlon_fill_in_addresses(struct op_msrs * const msrs)
63 } 63 }
64} 64}
65 65
66 66
67static void athlon_setup_ctrs(struct op_msrs const * const msrs) 67static void athlon_setup_ctrs(struct op_msrs const * const msrs)
68{ 68{
69 unsigned int low, high; 69 unsigned int low, high;
70 int i; 70 int i;
71 71
72 /* clear all counters */ 72 /* clear all counters */
73 for (i = 0 ; i < NUM_CONTROLS; ++i) { 73 for (i = 0 ; i < NUM_CONTROLS; ++i) {
74 if (unlikely(!CTRL_IS_RESERVED(msrs,i))) 74 if (unlikely(!CTRL_IS_RESERVED(msrs, i)))
75 continue; 75 continue;
76 CTRL_READ(low, high, msrs, i); 76 CTRL_READ(low, high, msrs, i);
77 CTRL_CLEAR_LO(low); 77 CTRL_CLEAR_LO(low);
@@ -81,14 +81,14 @@ static void athlon_setup_ctrs(struct op_msrs const * const msrs)
81 81
82 /* avoid a false detection of ctr overflows in NMI handler */ 82 /* avoid a false detection of ctr overflows in NMI handler */
83 for (i = 0; i < NUM_COUNTERS; ++i) { 83 for (i = 0; i < NUM_COUNTERS; ++i) {
84 if (unlikely(!CTR_IS_RESERVED(msrs,i))) 84 if (unlikely(!CTR_IS_RESERVED(msrs, i)))
85 continue; 85 continue;
86 CTR_WRITE(1, msrs, i); 86 CTR_WRITE(1, msrs, i);
87 } 87 }
88 88
89 /* enable active counters */ 89 /* enable active counters */
90 for (i = 0; i < NUM_COUNTERS; ++i) { 90 for (i = 0; i < NUM_COUNTERS; ++i) {
91 if ((counter_config[i].enabled) && (CTR_IS_RESERVED(msrs,i))) { 91 if ((counter_config[i].enabled) && (CTR_IS_RESERVED(msrs, i))) {
92 reset_value[i] = counter_config[i].count; 92 reset_value[i] = counter_config[i].count;
93 93
94 CTR_WRITE(counter_config[i].count, msrs, i); 94 CTR_WRITE(counter_config[i].count, msrs, i);
@@ -112,7 +112,7 @@ static void athlon_setup_ctrs(struct op_msrs const * const msrs)
112 } 112 }
113} 113}
114 114
115 115
116static int athlon_check_ctrs(struct pt_regs * const regs, 116static int athlon_check_ctrs(struct pt_regs * const regs,
117 struct op_msrs const * const msrs) 117 struct op_msrs const * const msrs)
118{ 118{
@@ -133,7 +133,7 @@ static int athlon_check_ctrs(struct pt_regs * const regs,
133 return 1; 133 return 1;
134} 134}
135 135
136 136
137static void athlon_start(struct op_msrs const * const msrs) 137static void athlon_start(struct op_msrs const * const msrs)
138{ 138{
139 unsigned int low, high; 139 unsigned int low, high;
@@ -150,7 +150,7 @@ static void athlon_start(struct op_msrs const * const msrs)
150 150
151static void athlon_stop(struct op_msrs const * const msrs) 151static void athlon_stop(struct op_msrs const * const msrs)
152{ 152{
153 unsigned int low,high; 153 unsigned int low, high;
154 int i; 154 int i;
155 155
156 /* Subtle: stop on all counters to avoid race with 156 /* Subtle: stop on all counters to avoid race with
@@ -169,11 +169,11 @@ static void athlon_shutdown(struct op_msrs const * const msrs)
169 int i; 169 int i;
170 170
171 for (i = 0 ; i < NUM_COUNTERS ; ++i) { 171 for (i = 0 ; i < NUM_COUNTERS ; ++i) {
172 if (CTR_IS_RESERVED(msrs,i)) 172 if (CTR_IS_RESERVED(msrs, i))
173 release_perfctr_nmi(MSR_K7_PERFCTR0 + i); 173 release_perfctr_nmi(MSR_K7_PERFCTR0 + i);
174 } 174 }
175 for (i = 0 ; i < NUM_CONTROLS ; ++i) { 175 for (i = 0 ; i < NUM_CONTROLS ; ++i) {
176 if (CTRL_IS_RESERVED(msrs,i)) 176 if (CTRL_IS_RESERVED(msrs, i))
177 release_evntsel_nmi(MSR_K7_EVNTSEL0 + i); 177 release_evntsel_nmi(MSR_K7_EVNTSEL0 + i);
178 } 178 }
179} 179}
diff --git a/arch/x86/oprofile/op_model_ppro.c b/arch/x86/oprofile/op_model_ppro.c
index c554f52cb808..eff431f6c57b 100644
--- a/arch/x86/oprofile/op_model_ppro.c
+++ b/arch/x86/oprofile/op_model_ppro.c
@@ -1,4 +1,4 @@
1/** 1/*
2 * @file op_model_ppro.h 2 * @file op_model_ppro.h
3 * pentium pro / P6 model-specific MSR operations 3 * pentium pro / P6 model-specific MSR operations
4 * 4 *
@@ -15,45 +15,45 @@
15#include <asm/msr.h> 15#include <asm/msr.h>
16#include <asm/apic.h> 16#include <asm/apic.h>
17#include <asm/nmi.h> 17#include <asm/nmi.h>
18 18
19#include "op_x86_model.h" 19#include "op_x86_model.h"
20#include "op_counter.h" 20#include "op_counter.h"
21 21
22#define NUM_COUNTERS 2 22#define NUM_COUNTERS 2
23#define NUM_CONTROLS 2 23#define NUM_CONTROLS 2
24 24
25#define CTR_IS_RESERVED(msrs,c) (msrs->counters[(c)].addr ? 1 : 0) 25#define CTR_IS_RESERVED(msrs, c) (msrs->counters[(c)].addr ? 1 : 0)
26#define CTR_READ(l,h,msrs,c) do {rdmsr(msrs->counters[(c)].addr, (l), (h));} while (0) 26#define CTR_READ(l, h, msrs, c) do {rdmsr(msrs->counters[(c)].addr, (l), (h)); } while (0)
27#define CTR_32BIT_WRITE(l,msrs,c) \ 27#define CTR_32BIT_WRITE(l, msrs, c) \
28 do {wrmsr(msrs->counters[(c)].addr, -(u32)(l), 0);} while (0) 28 do {wrmsr(msrs->counters[(c)].addr, -(u32)(l), 0); } while (0)
29#define CTR_OVERFLOWED(n) (!((n) & (1U<<31))) 29#define CTR_OVERFLOWED(n) (!((n) & (1U<<31)))
30 30
31#define CTRL_IS_RESERVED(msrs,c) (msrs->controls[(c)].addr ? 1 : 0) 31#define CTRL_IS_RESERVED(msrs, c) (msrs->controls[(c)].addr ? 1 : 0)
32#define CTRL_READ(l,h,msrs,c) do {rdmsr((msrs->controls[(c)].addr), (l), (h));} while (0) 32#define CTRL_READ(l, h, msrs, c) do {rdmsr((msrs->controls[(c)].addr), (l), (h)); } while (0)
33#define CTRL_WRITE(l,h,msrs,c) do {wrmsr((msrs->controls[(c)].addr), (l), (h));} while (0) 33#define CTRL_WRITE(l, h, msrs, c) do {wrmsr((msrs->controls[(c)].addr), (l), (h)); } while (0)
34#define CTRL_SET_ACTIVE(n) (n |= (1<<22)) 34#define CTRL_SET_ACTIVE(n) (n |= (1<<22))
35#define CTRL_SET_INACTIVE(n) (n &= ~(1<<22)) 35#define CTRL_SET_INACTIVE(n) (n &= ~(1<<22))
36#define CTRL_CLEAR(x) (x &= (1<<21)) 36#define CTRL_CLEAR(x) (x &= (1<<21))
37#define CTRL_SET_ENABLE(val) (val |= 1<<20) 37#define CTRL_SET_ENABLE(val) (val |= 1<<20)
38#define CTRL_SET_USR(val,u) (val |= ((u & 1) << 16)) 38#define CTRL_SET_USR(val, u) (val |= ((u & 1) << 16))
39#define CTRL_SET_KERN(val,k) (val |= ((k & 1) << 17)) 39#define CTRL_SET_KERN(val, k) (val |= ((k & 1) << 17))
40#define CTRL_SET_UM(val, m) (val |= (m << 8)) 40#define CTRL_SET_UM(val, m) (val |= (m << 8))
41#define CTRL_SET_EVENT(val, e) (val |= e) 41#define CTRL_SET_EVENT(val, e) (val |= e)
42 42
43static unsigned long reset_value[NUM_COUNTERS]; 43static unsigned long reset_value[NUM_COUNTERS];
44 44
45static void ppro_fill_in_addresses(struct op_msrs * const msrs) 45static void ppro_fill_in_addresses(struct op_msrs * const msrs)
46{ 46{
47 int i; 47 int i;
48 48
49 for (i=0; i < NUM_COUNTERS; i++) { 49 for (i = 0; i < NUM_COUNTERS; i++) {
50 if (reserve_perfctr_nmi(MSR_P6_PERFCTR0 + i)) 50 if (reserve_perfctr_nmi(MSR_P6_PERFCTR0 + i))
51 msrs->counters[i].addr = MSR_P6_PERFCTR0 + i; 51 msrs->counters[i].addr = MSR_P6_PERFCTR0 + i;
52 else 52 else
53 msrs->counters[i].addr = 0; 53 msrs->counters[i].addr = 0;
54 } 54 }
55 55
56 for (i=0; i < NUM_CONTROLS; i++) { 56 for (i = 0; i < NUM_CONTROLS; i++) {
57 if (reserve_evntsel_nmi(MSR_P6_EVNTSEL0 + i)) 57 if (reserve_evntsel_nmi(MSR_P6_EVNTSEL0 + i))
58 msrs->controls[i].addr = MSR_P6_EVNTSEL0 + i; 58 msrs->controls[i].addr = MSR_P6_EVNTSEL0 + i;
59 else 59 else
@@ -69,23 +69,23 @@ static void ppro_setup_ctrs(struct op_msrs const * const msrs)
69 69
70 /* clear all counters */ 70 /* clear all counters */
71 for (i = 0 ; i < NUM_CONTROLS; ++i) { 71 for (i = 0 ; i < NUM_CONTROLS; ++i) {
72 if (unlikely(!CTRL_IS_RESERVED(msrs,i))) 72 if (unlikely(!CTRL_IS_RESERVED(msrs, i)))
73 continue; 73 continue;
74 CTRL_READ(low, high, msrs, i); 74 CTRL_READ(low, high, msrs, i);
75 CTRL_CLEAR(low); 75 CTRL_CLEAR(low);
76 CTRL_WRITE(low, high, msrs, i); 76 CTRL_WRITE(low, high, msrs, i);
77 } 77 }
78 78
79 /* avoid a false detection of ctr overflows in NMI handler */ 79 /* avoid a false detection of ctr overflows in NMI handler */
80 for (i = 0; i < NUM_COUNTERS; ++i) { 80 for (i = 0; i < NUM_COUNTERS; ++i) {
81 if (unlikely(!CTR_IS_RESERVED(msrs,i))) 81 if (unlikely(!CTR_IS_RESERVED(msrs, i)))
82 continue; 82 continue;
83 CTR_32BIT_WRITE(1, msrs, i); 83 CTR_32BIT_WRITE(1, msrs, i);
84 } 84 }
85 85
86 /* enable active counters */ 86 /* enable active counters */
87 for (i = 0; i < NUM_COUNTERS; ++i) { 87 for (i = 0; i < NUM_COUNTERS; ++i) {
88 if ((counter_config[i].enabled) && (CTR_IS_RESERVED(msrs,i))) { 88 if ((counter_config[i].enabled) && (CTR_IS_RESERVED(msrs, i))) {
89 reset_value[i] = counter_config[i].count; 89 reset_value[i] = counter_config[i].count;
90 90
91 CTR_32BIT_WRITE(counter_config[i].count, msrs, i); 91 CTR_32BIT_WRITE(counter_config[i].count, msrs, i);
@@ -104,13 +104,13 @@ static void ppro_setup_ctrs(struct op_msrs const * const msrs)
104 } 104 }
105} 105}
106 106
107 107
108static int ppro_check_ctrs(struct pt_regs * const regs, 108static int ppro_check_ctrs(struct pt_regs * const regs,
109 struct op_msrs const * const msrs) 109 struct op_msrs const * const msrs)
110{ 110{
111 unsigned int low, high; 111 unsigned int low, high;
112 int i; 112 int i;
113 113
114 for (i = 0 ; i < NUM_COUNTERS; ++i) { 114 for (i = 0 ; i < NUM_COUNTERS; ++i) {
115 if (!reset_value[i]) 115 if (!reset_value[i])
116 continue; 116 continue;
@@ -135,10 +135,10 @@ static int ppro_check_ctrs(struct pt_regs * const regs,
135 return 1; 135 return 1;
136} 136}
137 137
138 138
139static void ppro_start(struct op_msrs const * const msrs) 139static void ppro_start(struct op_msrs const * const msrs)
140{ 140{
141 unsigned int low,high; 141 unsigned int low, high;
142 int i; 142 int i;
143 143
144 for (i = 0; i < NUM_COUNTERS; ++i) { 144 for (i = 0; i < NUM_COUNTERS; ++i) {
@@ -153,7 +153,7 @@ static void ppro_start(struct op_msrs const * const msrs)
153 153
154static void ppro_stop(struct op_msrs const * const msrs) 154static void ppro_stop(struct op_msrs const * const msrs)
155{ 155{
156 unsigned int low,high; 156 unsigned int low, high;
157 int i; 157 int i;
158 158
159 for (i = 0; i < NUM_COUNTERS; ++i) { 159 for (i = 0; i < NUM_COUNTERS; ++i) {
@@ -170,11 +170,11 @@ static void ppro_shutdown(struct op_msrs const * const msrs)
170 int i; 170 int i;
171 171
172 for (i = 0 ; i < NUM_COUNTERS ; ++i) { 172 for (i = 0 ; i < NUM_COUNTERS ; ++i) {
173 if (CTR_IS_RESERVED(msrs,i)) 173 if (CTR_IS_RESERVED(msrs, i))
174 release_perfctr_nmi(MSR_P6_PERFCTR0 + i); 174 release_perfctr_nmi(MSR_P6_PERFCTR0 + i);
175 } 175 }
176 for (i = 0 ; i < NUM_CONTROLS ; ++i) { 176 for (i = 0 ; i < NUM_CONTROLS ; ++i) {
177 if (CTRL_IS_RESERVED(msrs,i)) 177 if (CTRL_IS_RESERVED(msrs, i))
178 release_evntsel_nmi(MSR_P6_EVNTSEL0 + i); 178 release_evntsel_nmi(MSR_P6_EVNTSEL0 + i);
179 } 179 }
180} 180}
diff --git a/arch/x86/pci/i386.c b/arch/x86/pci/i386.c
index 103b9dff1213..2ead72363077 100644
--- a/arch/x86/pci/i386.c
+++ b/arch/x86/pci/i386.c
@@ -30,6 +30,9 @@
30#include <linux/init.h> 30#include <linux/init.h>
31#include <linux/ioport.h> 31#include <linux/ioport.h>
32#include <linux/errno.h> 32#include <linux/errno.h>
33#include <linux/bootmem.h>
34
35#include <asm/pat.h>
33 36
34#include "pci.h" 37#include "pci.h"
35 38
@@ -297,10 +300,35 @@ void pcibios_set_master(struct pci_dev *dev)
297 pci_write_config_byte(dev, PCI_LATENCY_TIMER, lat); 300 pci_write_config_byte(dev, PCI_LATENCY_TIMER, lat);
298} 301}
299 302
303static void pci_unmap_page_range(struct vm_area_struct *vma)
304{
305 u64 addr = (u64)vma->vm_pgoff << PAGE_SHIFT;
306 free_memtype(addr, addr + vma->vm_end - vma->vm_start);
307}
308
309static void pci_track_mmap_page_range(struct vm_area_struct *vma)
310{
311 u64 addr = (u64)vma->vm_pgoff << PAGE_SHIFT;
312 unsigned long flags = pgprot_val(vma->vm_page_prot)
313 & _PAGE_CACHE_MASK;
314
315 reserve_memtype(addr, addr + vma->vm_end - vma->vm_start, flags, NULL);
316}
317
318static struct vm_operations_struct pci_mmap_ops = {
319 .open = pci_track_mmap_page_range,
320 .close = pci_unmap_page_range,
321};
322
300int pci_mmap_page_range(struct pci_dev *dev, struct vm_area_struct *vma, 323int pci_mmap_page_range(struct pci_dev *dev, struct vm_area_struct *vma,
301 enum pci_mmap_state mmap_state, int write_combine) 324 enum pci_mmap_state mmap_state, int write_combine)
302{ 325{
303 unsigned long prot; 326 unsigned long prot;
327 u64 addr = vma->vm_pgoff << PAGE_SHIFT;
328 unsigned long len = vma->vm_end - vma->vm_start;
329 unsigned long flags;
330 unsigned long new_flags;
331 int retval;
304 332
305 /* I/O space cannot be accessed via normal processor loads and 333 /* I/O space cannot be accessed via normal processor loads and
306 * stores on this platform. 334 * stores on this platform.
@@ -308,21 +336,50 @@ int pci_mmap_page_range(struct pci_dev *dev, struct vm_area_struct *vma,
308 if (mmap_state == pci_mmap_io) 336 if (mmap_state == pci_mmap_io)
309 return -EINVAL; 337 return -EINVAL;
310 338
311 /* Leave vm_pgoff as-is, the PCI space address is the physical
312 * address on this platform.
313 */
314 prot = pgprot_val(vma->vm_page_prot); 339 prot = pgprot_val(vma->vm_page_prot);
315 if (boot_cpu_data.x86 > 3) 340 if (pat_wc_enabled && write_combine)
316 prot |= _PAGE_PCD | _PAGE_PWT; 341 prot |= _PAGE_CACHE_WC;
342 else if (boot_cpu_data.x86 > 3)
343 prot |= _PAGE_CACHE_UC;
344
317 vma->vm_page_prot = __pgprot(prot); 345 vma->vm_page_prot = __pgprot(prot);
318 346
319 /* Write-combine setting is ignored, it is changed via the mtrr 347 flags = pgprot_val(vma->vm_page_prot) & _PAGE_CACHE_MASK;
320 * interfaces on this platform. 348 retval = reserve_memtype(addr, addr + len, flags, &new_flags);
321 */ 349 if (retval)
350 return retval;
351
352 if (flags != new_flags) {
353 /*
354 * Do not fallback to certain memory types with certain
355 * requested type:
356 * - request is uncached, return cannot be write-back
357 * - request is uncached, return cannot be write-combine
358 * - request is write-combine, return cannot be write-back
359 */
360 if ((flags == _PAGE_CACHE_UC &&
361 (new_flags == _PAGE_CACHE_WB ||
362 new_flags == _PAGE_CACHE_WC)) ||
363 (flags == _PAGE_CACHE_WC &&
364 new_flags == _PAGE_CACHE_WB)) {
365 free_memtype(addr, addr+len);
366 return -EINVAL;
367 }
368 flags = new_flags;
369 }
370
371 if (vma->vm_pgoff <= max_pfn_mapped &&
372 ioremap_change_attr((unsigned long)__va(addr), len, flags)) {
373 free_memtype(addr, addr + len);
374 return -EINVAL;
375 }
376
322 if (io_remap_pfn_range(vma, vma->vm_start, vma->vm_pgoff, 377 if (io_remap_pfn_range(vma, vma->vm_start, vma->vm_pgoff,
323 vma->vm_end - vma->vm_start, 378 vma->vm_end - vma->vm_start,
324 vma->vm_page_prot)) 379 vma->vm_page_prot))
325 return -EAGAIN; 380 return -EAGAIN;
326 381
382 vma->vm_ops = &pci_mmap_ops;
383
327 return 0; 384 return 0;
328} 385}
diff --git a/arch/x86/pci/irq.c b/arch/x86/pci/irq.c
index a8715861877e..579745ca6b66 100644
--- a/arch/x86/pci/irq.c
+++ b/arch/x86/pci/irq.c
@@ -200,7 +200,7 @@ static int pirq_ali_get(struct pci_dev *router, struct pci_dev *dev, int pirq)
200{ 200{
201 static const unsigned char irqmap[16] = { 0, 9, 3, 10, 4, 5, 7, 6, 1, 11, 0, 12, 0, 14, 0, 15 }; 201 static const unsigned char irqmap[16] = { 0, 9, 3, 10, 4, 5, 7, 6, 1, 11, 0, 12, 0, 14, 0, 15 };
202 202
203 WARN_ON_ONCE(pirq >= 16); 203 WARN_ON_ONCE(pirq > 16);
204 return irqmap[read_config_nybble(router, 0x48, pirq-1)]; 204 return irqmap[read_config_nybble(router, 0x48, pirq-1)];
205} 205}
206 206
@@ -209,7 +209,7 @@ static int pirq_ali_set(struct pci_dev *router, struct pci_dev *dev, int pirq, i
209 static const unsigned char irqmap[16] = { 0, 8, 0, 2, 4, 5, 7, 6, 0, 1, 3, 9, 11, 0, 13, 15 }; 209 static const unsigned char irqmap[16] = { 0, 8, 0, 2, 4, 5, 7, 6, 0, 1, 3, 9, 11, 0, 13, 15 };
210 unsigned int val = irqmap[irq]; 210 unsigned int val = irqmap[irq];
211 211
212 WARN_ON_ONCE(pirq >= 16); 212 WARN_ON_ONCE(pirq > 16);
213 if (val) { 213 if (val) {
214 write_config_nybble(router, 0x48, pirq-1, val); 214 write_config_nybble(router, 0x48, pirq-1, val);
215 return 1; 215 return 1;
@@ -260,7 +260,7 @@ static int pirq_via586_get(struct pci_dev *router, struct pci_dev *dev, int pirq
260{ 260{
261 static const unsigned int pirqmap[5] = { 3, 2, 5, 1, 1 }; 261 static const unsigned int pirqmap[5] = { 3, 2, 5, 1, 1 };
262 262
263 WARN_ON_ONCE(pirq >= 5); 263 WARN_ON_ONCE(pirq > 5);
264 return read_config_nybble(router, 0x55, pirqmap[pirq-1]); 264 return read_config_nybble(router, 0x55, pirqmap[pirq-1]);
265} 265}
266 266
@@ -268,7 +268,7 @@ static int pirq_via586_set(struct pci_dev *router, struct pci_dev *dev, int pirq
268{ 268{
269 static const unsigned int pirqmap[5] = { 3, 2, 5, 1, 1 }; 269 static const unsigned int pirqmap[5] = { 3, 2, 5, 1, 1 };
270 270
271 WARN_ON_ONCE(pirq >= 5); 271 WARN_ON_ONCE(pirq > 5);
272 write_config_nybble(router, 0x55, pirqmap[pirq-1], irq); 272 write_config_nybble(router, 0x55, pirqmap[pirq-1], irq);
273 return 1; 273 return 1;
274} 274}
@@ -282,7 +282,7 @@ static int pirq_ite_get(struct pci_dev *router, struct pci_dev *dev, int pirq)
282{ 282{
283 static const unsigned char pirqmap[4] = { 1, 0, 2, 3 }; 283 static const unsigned char pirqmap[4] = { 1, 0, 2, 3 };
284 284
285 WARN_ON_ONCE(pirq >= 4); 285 WARN_ON_ONCE(pirq > 4);
286 return read_config_nybble(router,0x43, pirqmap[pirq-1]); 286 return read_config_nybble(router,0x43, pirqmap[pirq-1]);
287} 287}
288 288
@@ -290,7 +290,7 @@ static int pirq_ite_set(struct pci_dev *router, struct pci_dev *dev, int pirq, i
290{ 290{
291 static const unsigned char pirqmap[4] = { 1, 0, 2, 3 }; 291 static const unsigned char pirqmap[4] = { 1, 0, 2, 3 };
292 292
293 WARN_ON_ONCE(pirq >= 4); 293 WARN_ON_ONCE(pirq > 4);
294 write_config_nybble(router, 0x43, pirqmap[pirq-1], irq); 294 write_config_nybble(router, 0x43, pirqmap[pirq-1], irq);
295 return 1; 295 return 1;
296} 296}
diff --git a/arch/x86/pci/numa.c b/arch/x86/pci/numa.c
index 55270c26237c..d9afbae5092b 100644
--- a/arch/x86/pci/numa.c
+++ b/arch/x86/pci/numa.c
@@ -11,11 +11,41 @@
11#define XQUAD_PORTIO_BASE 0xfe400000 11#define XQUAD_PORTIO_BASE 0xfe400000
12#define XQUAD_PORTIO_QUAD 0x40000 /* 256k per quad. */ 12#define XQUAD_PORTIO_QUAD 0x40000 /* 256k per quad. */
13 13
14int mp_bus_id_to_node[MAX_MP_BUSSES];
14#define BUS2QUAD(global) (mp_bus_id_to_node[global]) 15#define BUS2QUAD(global) (mp_bus_id_to_node[global])
16
17int mp_bus_id_to_local[MAX_MP_BUSSES];
15#define BUS2LOCAL(global) (mp_bus_id_to_local[global]) 18#define BUS2LOCAL(global) (mp_bus_id_to_local[global])
19
20void mpc_oem_bus_info(struct mpc_config_bus *m, char *name,
21 struct mpc_config_translation *translation)
22{
23 int quad = translation->trans_quad;
24 int local = translation->trans_local;
25
26 mp_bus_id_to_node[m->mpc_busid] = quad;
27 mp_bus_id_to_local[m->mpc_busid] = local;
28 printk(KERN_INFO "Bus #%d is %s (node %d)\n",
29 m->mpc_busid, name, quad);
30}
31
32int quad_local_to_mp_bus_id [NR_CPUS/4][4];
16#define QUADLOCAL2BUS(quad,local) (quad_local_to_mp_bus_id[quad][local]) 33#define QUADLOCAL2BUS(quad,local) (quad_local_to_mp_bus_id[quad][local])
34void mpc_oem_pci_bus(struct mpc_config_bus *m,
35 struct mpc_config_translation *translation)
36{
37 int quad = translation->trans_quad;
38 int local = translation->trans_local;
39
40 quad_local_to_mp_bus_id[quad][local] = m->mpc_busid;
41}
42
43/* Where the IO area was mapped on multiquad, always 0 otherwise */
44void *xquad_portio;
45#ifdef CONFIG_X86_NUMAQ
46EXPORT_SYMBOL(xquad_portio);
47#endif
17 48
18extern void *xquad_portio; /* Where the IO area was mapped */
19#define XQUAD_PORT_ADDR(port, quad) (xquad_portio + (XQUAD_PORTIO_QUAD*quad) + port) 49#define XQUAD_PORT_ADDR(port, quad) (xquad_portio + (XQUAD_PORTIO_QUAD*quad) + port)
20 50
21#define PCI_CONF1_MQ_ADDRESS(bus, devfn, reg) \ 51#define PCI_CONF1_MQ_ADDRESS(bus, devfn, reg) \
diff --git a/arch/x86/power/cpu_32.c b/arch/x86/power/cpu_32.c
index 7f9c6da04a4c..7dc5d5cf50a2 100644
--- a/arch/x86/power/cpu_32.c
+++ b/arch/x86/power/cpu_32.c
@@ -27,17 +27,17 @@ static void __save_processor_state(struct saved_context *ctxt)
27 /* 27 /*
28 * descriptor tables 28 * descriptor tables
29 */ 29 */
30 store_gdt(&ctxt->gdt); 30 store_gdt(&ctxt->gdt);
31 store_idt(&ctxt->idt); 31 store_idt(&ctxt->idt);
32 store_tr(ctxt->tr); 32 store_tr(ctxt->tr);
33 33
34 /* 34 /*
35 * segment registers 35 * segment registers
36 */ 36 */
37 savesegment(es, ctxt->es); 37 savesegment(es, ctxt->es);
38 savesegment(fs, ctxt->fs); 38 savesegment(fs, ctxt->fs);
39 savesegment(gs, ctxt->gs); 39 savesegment(gs, ctxt->gs);
40 savesegment(ss, ctxt->ss); 40 savesegment(ss, ctxt->ss);
41 41
42 /* 42 /*
43 * control registers 43 * control registers
@@ -48,10 +48,12 @@ static void __save_processor_state(struct saved_context *ctxt)
48 ctxt->cr4 = read_cr4(); 48 ctxt->cr4 = read_cr4();
49} 49}
50 50
51/* Needed by apm.c */
51void save_processor_state(void) 52void save_processor_state(void)
52{ 53{
53 __save_processor_state(&saved_context); 54 __save_processor_state(&saved_context);
54} 55}
56EXPORT_SYMBOL(save_processor_state);
55 57
56static void do_fpu_end(void) 58static void do_fpu_end(void)
57{ 59{
@@ -64,9 +66,14 @@ static void do_fpu_end(void)
64static void fix_processor_context(void) 66static void fix_processor_context(void)
65{ 67{
66 int cpu = smp_processor_id(); 68 int cpu = smp_processor_id();
67 struct tss_struct * t = &per_cpu(init_tss, cpu); 69 struct tss_struct *t = &per_cpu(init_tss, cpu);
68 70
69 set_tss_desc(cpu,t); /* This just modifies memory; should not be necessary. But... This is necessary, because 386 hardware has concept of busy TSS or some similar stupidity. */ 71 set_tss_desc(cpu, t); /*
72 * This just modifies memory; should not be
73 * necessary. But... This is necessary, because
74 * 386 hardware has concept of busy TSS or some
75 * similar stupidity.
76 */
70 77
71 load_TR_desc(); /* This does ltr */ 78 load_TR_desc(); /* This does ltr */
72 load_LDT(&current->active_mm->context); /* This does lldt */ 79 load_LDT(&current->active_mm->context); /* This does lldt */
@@ -100,16 +107,16 @@ static void __restore_processor_state(struct saved_context *ctxt)
100 * now restore the descriptor tables to their proper values 107 * now restore the descriptor tables to their proper values
101 * ltr is done i fix_processor_context(). 108 * ltr is done i fix_processor_context().
102 */ 109 */
103 load_gdt(&ctxt->gdt); 110 load_gdt(&ctxt->gdt);
104 load_idt(&ctxt->idt); 111 load_idt(&ctxt->idt);
105 112
106 /* 113 /*
107 * segment registers 114 * segment registers
108 */ 115 */
109 loadsegment(es, ctxt->es); 116 loadsegment(es, ctxt->es);
110 loadsegment(fs, ctxt->fs); 117 loadsegment(fs, ctxt->fs);
111 loadsegment(gs, ctxt->gs); 118 loadsegment(gs, ctxt->gs);
112 loadsegment(ss, ctxt->ss); 119 loadsegment(ss, ctxt->ss);
113 120
114 /* 121 /*
115 * sysenter MSRs 122 * sysenter MSRs
@@ -123,11 +130,9 @@ static void __restore_processor_state(struct saved_context *ctxt)
123 mcheck_init(&boot_cpu_data); 130 mcheck_init(&boot_cpu_data);
124} 131}
125 132
133/* Needed by apm.c */
126void restore_processor_state(void) 134void restore_processor_state(void)
127{ 135{
128 __restore_processor_state(&saved_context); 136 __restore_processor_state(&saved_context);
129} 137}
130
131/* Needed by apm.c */
132EXPORT_SYMBOL(save_processor_state);
133EXPORT_SYMBOL(restore_processor_state); 138EXPORT_SYMBOL(restore_processor_state);
diff --git a/arch/x86/vdso/Makefile b/arch/x86/vdso/Makefile
index 0a8f4742ef51..17a6b057856b 100644
--- a/arch/x86/vdso/Makefile
+++ b/arch/x86/vdso/Makefile
@@ -39,7 +39,7 @@ $(obj)/%.so: $(obj)/%.so.dbg FORCE
39 39
40CFL := $(PROFILING) -mcmodel=small -fPIC -g0 -O2 -fasynchronous-unwind-tables -m64 40CFL := $(PROFILING) -mcmodel=small -fPIC -g0 -O2 -fasynchronous-unwind-tables -m64
41 41
42$(vobjs): KBUILD_CFLAGS = $(CFL) 42$(vobjs): KBUILD_CFLAGS += $(CFL)
43 43
44targets += vdso-syms.lds 44targets += vdso-syms.lds
45obj-$(VDSO64-y) += vdso-syms.lds 45obj-$(VDSO64-y) += vdso-syms.lds
diff --git a/arch/x86/vdso/vdso32-setup.c b/arch/x86/vdso/vdso32-setup.c
index 348f1341e1c8..e2af8eee80e3 100644
--- a/arch/x86/vdso/vdso32-setup.c
+++ b/arch/x86/vdso/vdso32-setup.c
@@ -210,8 +210,12 @@ static int use_sysenter __read_mostly = -1;
210/* May not be __init: called during resume */ 210/* May not be __init: called during resume */
211void syscall32_cpu_init(void) 211void syscall32_cpu_init(void)
212{ 212{
213 if (use_sysenter < 0) 213 if (use_sysenter < 0) {
214 use_sysenter = (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL); 214 if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL)
215 use_sysenter = 1;
216 if (boot_cpu_data.x86_vendor == X86_VENDOR_CENTAUR)
217 use_sysenter = 1;
218 }
215 219
216 /* Load these always in case some future AMD CPU supports 220 /* Load these always in case some future AMD CPU supports
217 SYSENTER from compat mode too. */ 221 SYSENTER from compat mode too. */
@@ -325,6 +329,9 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int exstack)
325 int ret = 0; 329 int ret = 0;
326 bool compat; 330 bool compat;
327 331
332 if (vdso_enabled == VDSO_DISABLED)
333 return 0;
334
328 down_write(&mm->mmap_sem); 335 down_write(&mm->mmap_sem);
329 336
330 /* Test compat mode once here, in case someone 337 /* Test compat mode once here, in case someone
diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
index 27ee26aedf94..c0388220cf97 100644
--- a/arch/x86/xen/enlighten.c
+++ b/arch/x86/xen/enlighten.c
@@ -25,6 +25,7 @@
25#include <linux/mm.h> 25#include <linux/mm.h>
26#include <linux/page-flags.h> 26#include <linux/page-flags.h>
27#include <linux/highmem.h> 27#include <linux/highmem.h>
28#include <linux/console.h>
28 29
29#include <xen/interface/xen.h> 30#include <xen/interface/xen.h>
30#include <xen/interface/physdev.h> 31#include <xen/interface/physdev.h>
@@ -889,7 +890,6 @@ void __init xen_setup_vcpu_info_placement(void)
889 pv_irq_ops.irq_disable = xen_irq_disable_direct; 890 pv_irq_ops.irq_disable = xen_irq_disable_direct;
890 pv_irq_ops.irq_enable = xen_irq_enable_direct; 891 pv_irq_ops.irq_enable = xen_irq_enable_direct;
891 pv_mmu_ops.read_cr2 = xen_read_cr2_direct; 892 pv_mmu_ops.read_cr2 = xen_read_cr2_direct;
892 pv_cpu_ops.iret = xen_iret_direct;
893 } 893 }
894} 894}
895 895
@@ -993,7 +993,7 @@ static const struct pv_cpu_ops xen_cpu_ops __initdata = {
993 .read_tsc = native_read_tsc, 993 .read_tsc = native_read_tsc,
994 .read_pmc = native_read_pmc, 994 .read_pmc = native_read_pmc,
995 995
996 .iret = (void *)&hypercall_page[__HYPERVISOR_iret], 996 .iret = xen_iret,
997 .irq_enable_syscall_ret = NULL, /* never called */ 997 .irq_enable_syscall_ret = NULL, /* never called */
998 998
999 .load_tr_desc = paravirt_nop, 999 .load_tr_desc = paravirt_nop,
@@ -1228,6 +1228,9 @@ asmlinkage void __init xen_start_kernel(void)
1228 ? __pa(xen_start_info->mod_start) : 0; 1228 ? __pa(xen_start_info->mod_start) : 0;
1229 boot_params.hdr.ramdisk_size = xen_start_info->mod_len; 1229 boot_params.hdr.ramdisk_size = xen_start_info->mod_len;
1230 1230
1231 if (!is_initial_xendomain())
1232 add_preferred_console("hvc", 0, NULL);
1233
1231 /* Start the world */ 1234 /* Start the world */
1232 start_kernel(); 1235 start_kernel();
1233} 1236}
diff --git a/arch/x86/xen/multicalls.c b/arch/x86/xen/multicalls.c
index 5e6f36f6d876..5791eb2e3750 100644
--- a/arch/x86/xen/multicalls.c
+++ b/arch/x86/xen/multicalls.c
@@ -76,7 +76,7 @@ void xen_mc_flush(void)
76 if (ret) { 76 if (ret) {
77 printk(KERN_ERR "%d multicall(s) failed: cpu %d\n", 77 printk(KERN_ERR "%d multicall(s) failed: cpu %d\n",
78 ret, smp_processor_id()); 78 ret, smp_processor_id());
79 for(i = 0; i < b->mcidx; i++) { 79 for (i = 0; i < b->mcidx; i++) {
80 printk(" call %2d/%d: op=%lu arg=[%lx] result=%ld\n", 80 printk(" call %2d/%d: op=%lu arg=[%lx] result=%ld\n",
81 i+1, b->mcidx, 81 i+1, b->mcidx,
82 b->debug[i].op, 82 b->debug[i].op,
@@ -93,7 +93,7 @@ void xen_mc_flush(void)
93 93
94 local_irq_restore(flags); 94 local_irq_restore(flags);
95 95
96 for(i = 0; i < b->cbidx; i++) { 96 for (i = 0; i < b->cbidx; i++) {
97 struct callback *cb = &b->callbacks[i]; 97 struct callback *cb = &b->callbacks[i];
98 98
99 (*cb->fn)(cb->data); 99 (*cb->fn)(cb->data);
diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c
index aafc54437403..e340ff92f6b6 100644
--- a/arch/x86/xen/smp.c
+++ b/arch/x86/xen/smp.c
@@ -35,7 +35,7 @@
35#include "xen-ops.h" 35#include "xen-ops.h"
36#include "mmu.h" 36#include "mmu.h"
37 37
38static cpumask_t cpu_initialized_map; 38static cpumask_t xen_cpu_initialized_map;
39static DEFINE_PER_CPU(int, resched_irq); 39static DEFINE_PER_CPU(int, resched_irq);
40static DEFINE_PER_CPU(int, callfunc_irq); 40static DEFINE_PER_CPU(int, callfunc_irq);
41 41
@@ -179,7 +179,7 @@ void __init xen_smp_prepare_cpus(unsigned int max_cpus)
179 if (xen_smp_intr_init(0)) 179 if (xen_smp_intr_init(0))
180 BUG(); 180 BUG();
181 181
182 cpu_initialized_map = cpumask_of_cpu(0); 182 xen_cpu_initialized_map = cpumask_of_cpu(0);
183 183
184 /* Restrict the possible_map according to max_cpus. */ 184 /* Restrict the possible_map according to max_cpus. */
185 while ((num_possible_cpus() > 1) && (num_possible_cpus() > max_cpus)) { 185 while ((num_possible_cpus() > 1) && (num_possible_cpus() > max_cpus)) {
@@ -210,7 +210,7 @@ cpu_initialize_context(unsigned int cpu, struct task_struct *idle)
210 struct vcpu_guest_context *ctxt; 210 struct vcpu_guest_context *ctxt;
211 struct gdt_page *gdt = &per_cpu(gdt_page, cpu); 211 struct gdt_page *gdt = &per_cpu(gdt_page, cpu);
212 212
213 if (cpu_test_and_set(cpu, cpu_initialized_map)) 213 if (cpu_test_and_set(cpu, xen_cpu_initialized_map))
214 return 0; 214 return 0;
215 215
216 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL); 216 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
diff --git a/arch/x86/xen/xen-asm.S b/arch/x86/xen/xen-asm.S
index 6b7190449d07..fe161ed4b01e 100644
--- a/arch/x86/xen/xen-asm.S
+++ b/arch/x86/xen/xen-asm.S
@@ -135,13 +135,8 @@ ENDPATCH(xen_restore_fl_direct)
135 current stack state in whatever form its in, we keep things 135 current stack state in whatever form its in, we keep things
136 simple by only using a single register which is pushed/popped 136 simple by only using a single register which is pushed/popped
137 on the stack. 137 on the stack.
138
139 Non-direct iret could be done in the same way, but it would
140 require an annoying amount of code duplication. We'll assume
141 that direct mode will be the common case once the hypervisor
142 support becomes commonplace.
143 */ 138 */
144ENTRY(xen_iret_direct) 139ENTRY(xen_iret)
145 /* test eflags for special cases */ 140 /* test eflags for special cases */
146 testl $(X86_EFLAGS_VM | XEN_EFLAGS_NMI), 8(%esp) 141 testl $(X86_EFLAGS_VM | XEN_EFLAGS_NMI), 8(%esp)
147 jnz hyper_iret 142 jnz hyper_iret
@@ -155,9 +150,9 @@ ENTRY(xen_iret_direct)
155 GET_THREAD_INFO(%eax) 150 GET_THREAD_INFO(%eax)
156 movl TI_cpu(%eax),%eax 151 movl TI_cpu(%eax),%eax
157 movl __per_cpu_offset(,%eax,4),%eax 152 movl __per_cpu_offset(,%eax,4),%eax
158 lea per_cpu__xen_vcpu_info(%eax),%eax 153 mov per_cpu__xen_vcpu(%eax),%eax
159#else 154#else
160 movl $per_cpu__xen_vcpu_info, %eax 155 movl per_cpu__xen_vcpu, %eax
161#endif 156#endif
162 157
163 /* check IF state we're restoring */ 158 /* check IF state we're restoring */
diff --git a/arch/x86/xen/xen-ops.h b/arch/x86/xen/xen-ops.h
index b02a909bfd4c..956a491ea998 100644
--- a/arch/x86/xen/xen-ops.h
+++ b/arch/x86/xen/xen-ops.h
@@ -63,5 +63,5 @@ DECL_ASM(void, xen_irq_disable_direct, void);
63DECL_ASM(unsigned long, xen_save_fl_direct, void); 63DECL_ASM(unsigned long, xen_save_fl_direct, void);
64DECL_ASM(void, xen_restore_fl_direct, unsigned long); 64DECL_ASM(void, xen_restore_fl_direct, unsigned long);
65 65
66void xen_iret_direct(void); 66void xen_iret(void);
67#endif /* XEN_OPS_H */ 67#endif /* XEN_OPS_H */
diff --git a/arch/xtensa/kernel/Makefile b/arch/xtensa/kernel/Makefile
index f582d6a24ec2..7419dbccf027 100644
--- a/arch/xtensa/kernel/Makefile
+++ b/arch/xtensa/kernel/Makefile
@@ -5,7 +5,7 @@
5extra-y := head.o vmlinux.lds 5extra-y := head.o vmlinux.lds
6 6
7 7
8obj-y := align.o entry.o irq.o coprocessor.o process.o ptrace.o semaphore.o \ 8obj-y := align.o entry.o irq.o coprocessor.o process.o ptrace.o \
9 setup.o signal.o syscall.o time.o traps.o vectors.o platform.o \ 9 setup.o signal.o syscall.o time.o traps.o vectors.o platform.o \
10 pci-dma.o init_task.o io.o 10 pci-dma.o init_task.o io.o
11 11
diff --git a/arch/xtensa/kernel/semaphore.c b/arch/xtensa/kernel/semaphore.c
deleted file mode 100644
index 995c6410ae10..000000000000
--- a/arch/xtensa/kernel/semaphore.c
+++ /dev/null
@@ -1,226 +0,0 @@
1/*
2 * arch/xtensa/kernel/semaphore.c
3 *
4 * Generic semaphore code. Buyer beware. Do your own specific changes
5 * in <asm/semaphore-helper.h>
6 *
7 * This file is subject to the terms and conditions of the GNU General Public
8 * License. See the file "COPYING" in the main directory of this archive
9 * for more details.
10 *
11 * Copyright (C) 2001 - 2005 Tensilica Inc.
12 *
13 * Joe Taylor <joe@tensilica.com, joetylr@yahoo.com>
14 * Chris Zankel <chris@zankel.net>
15 * Marc Gauthier<marc@tensilica.com, marc@alumni.uwaterloo.ca>
16 * Kevin Chea
17 */
18
19#include <linux/sched.h>
20#include <linux/wait.h>
21#include <linux/init.h>
22#include <asm/semaphore.h>
23#include <asm/errno.h>
24
25/*
26 * These two _must_ execute atomically wrt each other.
27 */
28
29static __inline__ void wake_one_more(struct semaphore * sem)
30{
31 atomic_inc((atomic_t *)&sem->sleepers);
32}
33
34static __inline__ int waking_non_zero(struct semaphore *sem)
35{
36 unsigned long flags;
37 int ret = 0;
38
39 spin_lock_irqsave(&semaphore_wake_lock, flags);
40 if (sem->sleepers > 0) {
41 sem->sleepers--;
42 ret = 1;
43 }
44 spin_unlock_irqrestore(&semaphore_wake_lock, flags);
45 return ret;
46}
47
48/*
49 * waking_non_zero_interruptible:
50 * 1 got the lock
51 * 0 go to sleep
52 * -EINTR interrupted
53 *
54 * We must undo the sem->count down_interruptible() increment while we are
55 * protected by the spinlock in order to make atomic this atomic_inc() with the
56 * atomic_read() in wake_one_more(), otherwise we can race. -arca
57 */
58
59static __inline__ int waking_non_zero_interruptible(struct semaphore *sem,
60 struct task_struct *tsk)
61{
62 unsigned long flags;
63 int ret = 0;
64
65 spin_lock_irqsave(&semaphore_wake_lock, flags);
66 if (sem->sleepers > 0) {
67 sem->sleepers--;
68 ret = 1;
69 } else if (signal_pending(tsk)) {
70 atomic_inc(&sem->count);
71 ret = -EINTR;
72 }
73 spin_unlock_irqrestore(&semaphore_wake_lock, flags);
74 return ret;
75}
76
77/*
78 * waking_non_zero_trylock:
79 * 1 failed to lock
80 * 0 got the lock
81 *
82 * We must undo the sem->count down_trylock() increment while we are
83 * protected by the spinlock in order to make atomic this atomic_inc() with the
84 * atomic_read() in wake_one_more(), otherwise we can race. -arca
85 */
86
87static __inline__ int waking_non_zero_trylock(struct semaphore *sem)
88{
89 unsigned long flags;
90 int ret = 1;
91
92 spin_lock_irqsave(&semaphore_wake_lock, flags);
93 if (sem->sleepers <= 0)
94 atomic_inc(&sem->count);
95 else {
96 sem->sleepers--;
97 ret = 0;
98 }
99 spin_unlock_irqrestore(&semaphore_wake_lock, flags);
100 return ret;
101}
102
103DEFINE_SPINLOCK(semaphore_wake_lock);
104
105/*
106 * Semaphores are implemented using a two-way counter:
107 * The "count" variable is decremented for each process
108 * that tries to sleep, while the "waking" variable is
109 * incremented when the "up()" code goes to wake up waiting
110 * processes.
111 *
112 * Notably, the inline "up()" and "down()" functions can
113 * efficiently test if they need to do any extra work (up
114 * needs to do something only if count was negative before
115 * the increment operation.
116 *
117 * waking_non_zero() (from asm/semaphore.h) must execute
118 * atomically.
119 *
120 * When __up() is called, the count was negative before
121 * incrementing it, and we need to wake up somebody.
122 *
123 * This routine adds one to the count of processes that need to
124 * wake up and exit. ALL waiting processes actually wake up but
125 * only the one that gets to the "waking" field first will gate
126 * through and acquire the semaphore. The others will go back
127 * to sleep.
128 *
129 * Note that these functions are only called when there is
130 * contention on the lock, and as such all this is the
131 * "non-critical" part of the whole semaphore business. The
132 * critical part is the inline stuff in <asm/semaphore.h>
133 * where we want to avoid any extra jumps and calls.
134 */
135
136void __up(struct semaphore *sem)
137{
138 wake_one_more(sem);
139 wake_up(&sem->wait);
140}
141
142/*
143 * Perform the "down" function. Return zero for semaphore acquired,
144 * return negative for signalled out of the function.
145 *
146 * If called from __down, the return is ignored and the wait loop is
147 * not interruptible. This means that a task waiting on a semaphore
148 * using "down()" cannot be killed until someone does an "up()" on
149 * the semaphore.
150 *
151 * If called from __down_interruptible, the return value gets checked
152 * upon return. If the return value is negative then the task continues
153 * with the negative value in the return register (it can be tested by
154 * the caller).
155 *
156 * Either form may be used in conjunction with "up()".
157 *
158 */
159
160#define DOWN_VAR \
161 struct task_struct *tsk = current; \
162 wait_queue_t wait; \
163 init_waitqueue_entry(&wait, tsk);
164
165#define DOWN_HEAD(task_state) \
166 \
167 \
168 tsk->state = (task_state); \
169 add_wait_queue(&sem->wait, &wait); \
170 \
171 /* \
172 * Ok, we're set up. sem->count is known to be less than zero \
173 * so we must wait. \
174 * \
175 * We can let go the lock for purposes of waiting. \
176 * We re-acquire it after awaking so as to protect \
177 * all semaphore operations. \
178 * \
179 * If "up()" is called before we call waking_non_zero() then \
180 * we will catch it right away. If it is called later then \
181 * we will have to go through a wakeup cycle to catch it. \
182 * \
183 * Multiple waiters contend for the semaphore lock to see \
184 * who gets to gate through and who has to wait some more. \
185 */ \
186 for (;;) {
187
188#define DOWN_TAIL(task_state) \
189 tsk->state = (task_state); \
190 } \
191 tsk->state = TASK_RUNNING; \
192 remove_wait_queue(&sem->wait, &wait);
193
194void __sched __down(struct semaphore * sem)
195{
196 DOWN_VAR
197 DOWN_HEAD(TASK_UNINTERRUPTIBLE)
198 if (waking_non_zero(sem))
199 break;
200 schedule();
201 DOWN_TAIL(TASK_UNINTERRUPTIBLE)
202}
203
204int __sched __down_interruptible(struct semaphore * sem)
205{
206 int ret = 0;
207 DOWN_VAR
208 DOWN_HEAD(TASK_INTERRUPTIBLE)
209
210 ret = waking_non_zero_interruptible(sem, tsk);
211 if (ret)
212 {
213 if (ret == 1)
214 /* ret != 0 only if we get interrupted -arca */
215 ret = 0;
216 break;
217 }
218 schedule();
219 DOWN_TAIL(TASK_INTERRUPTIBLE)
220 return ret;
221}
222
223int __down_trylock(struct semaphore * sem)
224{
225 return waking_non_zero_trylock(sem);
226}
diff --git a/arch/xtensa/kernel/xtensa_ksyms.c b/arch/xtensa/kernel/xtensa_ksyms.c
index 60dbdb43fb4c..6e52cdd6166f 100644
--- a/arch/xtensa/kernel/xtensa_ksyms.c
+++ b/arch/xtensa/kernel/xtensa_ksyms.c
@@ -26,7 +26,6 @@
26#include <asm/io.h> 26#include <asm/io.h>
27#include <asm/page.h> 27#include <asm/page.h>
28#include <asm/pgalloc.h> 28#include <asm/pgalloc.h>
29#include <asm/semaphore.h>
30#ifdef CONFIG_BLK_DEV_FD 29#ifdef CONFIG_BLK_DEV_FD
31#include <asm/floppy.h> 30#include <asm/floppy.h>
32#endif 31#endif
@@ -71,14 +70,6 @@ EXPORT_SYMBOL(__umodsi3);
71EXPORT_SYMBOL(__udivdi3); 70EXPORT_SYMBOL(__udivdi3);
72EXPORT_SYMBOL(__umoddi3); 71EXPORT_SYMBOL(__umoddi3);
73 72
74/*
75 * Semaphore operations
76 */
77EXPORT_SYMBOL(__down);
78EXPORT_SYMBOL(__down_interruptible);
79EXPORT_SYMBOL(__down_trylock);
80EXPORT_SYMBOL(__up);
81
82#ifdef CONFIG_NET 73#ifdef CONFIG_NET
83/* 74/*
84 * Networking support 75 * Networking support
generated by cgit v1.2.2 (git 2.25.0) at 2024-12-27 12:02:18 -0500