aboutsummaryrefslogtreecommitdiffstats
path: root/arch
diff options
context:
space:
mode:
Diffstat (limited to 'arch')
-rw-r--r--arch/alpha/include/asm/fcntl.h2
-rw-r--r--arch/alpha/include/asm/futex.h29
-rw-r--r--arch/alpha/include/asm/rwsem.h36
-rw-r--r--arch/alpha/kernel/osf_sys.c36
-rw-r--r--arch/alpha/kernel/time.c8
-rw-r--r--arch/arm/include/asm/futex.h29
-rw-r--r--arch/arm/kernel/time.c4
-rw-r--r--arch/arm/mach-clps711x/include/mach/time.h2
-rw-r--r--arch/blackfin/kernel/time.c6
-rw-r--r--arch/cris/arch-v10/kernel/time.c4
-rw-r--r--arch/cris/arch-v32/kernel/smp.c4
-rw-r--r--arch/cris/arch-v32/kernel/time.c6
-rw-r--r--arch/frv/include/asm/futex.h5
-rw-r--r--arch/frv/kernel/futex.c14
-rw-r--r--arch/frv/kernel/time.c14
-rw-r--r--arch/h8300/kernel/time.c4
-rw-r--r--arch/h8300/kernel/timer/timer8.c2
-rw-r--r--arch/ia64/include/asm/futex.h15
-rw-r--r--arch/ia64/include/asm/rwsem.h37
-rw-r--r--arch/ia64/include/asm/xen/hypercall.h2
-rw-r--r--arch/ia64/kernel/time.c19
-rw-r--r--arch/ia64/xen/suspend.c9
-rw-r--r--arch/ia64/xen/time.c13
-rw-r--r--arch/m32r/kernel/time.c5
-rw-r--r--arch/m68k/bvme6000/config.c4
-rw-r--r--arch/m68k/kernel/time.c4
-rw-r--r--arch/m68k/mvme147/config.c4
-rw-r--r--arch/m68k/mvme16x/config.c4
-rw-r--r--arch/m68k/sun3/sun3ints.c2
-rw-r--r--arch/m68knommu/kernel/time.c8
-rw-r--r--arch/microblaze/include/asm/futex.h31
-rw-r--r--arch/microblaze/include/asm/pci-bridge.h12
-rw-r--r--arch/microblaze/include/asm/prom.h15
-rw-r--r--arch/microblaze/kernel/prom_parse.c77
-rw-r--r--arch/microblaze/pci/pci-common.c1
-rw-r--r--arch/mips/include/asm/futex.h39
-rw-r--r--arch/mn10300/kernel/time.c6
-rw-r--r--arch/parisc/hpux/sys_hpux.c65
-rw-r--r--arch/parisc/include/asm/fcntl.h2
-rw-r--r--arch/parisc/include/asm/futex.h24
-rw-r--r--arch/parisc/kernel/time.c7
-rw-r--r--arch/powerpc/include/asm/futex.h27
-rw-r--r--arch/powerpc/include/asm/pci-bridge.h10
-rw-r--r--arch/powerpc/include/asm/prom.h15
-rw-r--r--arch/powerpc/include/asm/rwsem.h51
-rw-r--r--arch/powerpc/kernel/pci-common.c1
-rw-r--r--arch/powerpc/kernel/prom_parse.c84
-rw-r--r--arch/powerpc/platforms/cell/spufs/syscalls.c2
-rw-r--r--arch/s390/include/asm/futex.h12
-rw-r--r--arch/s390/include/asm/rwsem.h63
-rw-r--r--arch/s390/include/asm/uaccess.h4
-rw-r--r--arch/s390/lib/uaccess.h8
-rw-r--r--arch/s390/lib/uaccess_pt.c17
-rw-r--r--arch/s390/lib/uaccess_std.c8
-rw-r--r--arch/sh/include/asm/futex-irq.h24
-rw-r--r--arch/sh/include/asm/futex.h11
-rw-r--r--arch/sh/include/asm/rwsem.h56
-rw-r--r--arch/sparc/include/asm/fcntl.h2
-rw-r--r--arch/sparc/include/asm/futex_64.h20
-rw-r--r--arch/sparc/include/asm/rwsem.h46
-rw-r--r--arch/sparc/kernel/pcic.c4
-rw-r--r--arch/sparc/kernel/time_32.c9
-rw-r--r--arch/sparc/lib/atomic32.c2
-rw-r--r--arch/tile/include/asm/futex.h27
-rw-r--r--arch/um/Kconfig.common1
-rw-r--r--arch/um/Kconfig.x865
-rw-r--r--arch/um/drivers/mconsole_kern.c21
-rw-r--r--arch/um/drivers/ubd_kern.c2
-rw-r--r--arch/um/kernel/irq.c31
-rw-r--r--arch/x86/Kconfig30
-rw-r--r--arch/x86/Kconfig.cpu5
-rw-r--r--arch/x86/ia32/ia32entry.S32
-rw-r--r--arch/x86/include/asm/acpi.h10
-rw-r--r--arch/x86/include/asm/amd_nb.h8
-rw-r--r--arch/x86/include/asm/apic.h42
-rw-r--r--arch/x86/include/asm/apicdef.h12
-rw-r--r--arch/x86/include/asm/bootparam.h1
-rw-r--r--arch/x86/include/asm/cpufeature.h2
-rw-r--r--arch/x86/include/asm/e820.h2
-rw-r--r--arch/x86/include/asm/entry_arch.h5
-rw-r--r--arch/x86/include/asm/frame.h6
-rw-r--r--arch/x86/include/asm/futex.h22
-rw-r--r--arch/x86/include/asm/hw_irq.h24
-rw-r--r--arch/x86/include/asm/init.h6
-rw-r--r--arch/x86/include/asm/io_apic.h44
-rw-r--r--arch/x86/include/asm/ipi.h8
-rw-r--r--arch/x86/include/asm/irq.h3
-rw-r--r--arch/x86/include/asm/irq_controller.h12
-rw-r--r--arch/x86/include/asm/irq_vectors.h45
-rw-r--r--arch/x86/include/asm/kdebug.h1
-rw-r--r--arch/x86/include/asm/mpspec.h3
-rw-r--r--arch/x86/include/asm/msr-index.h3
-rw-r--r--arch/x86/include/asm/nmi.h1
-rw-r--r--arch/x86/include/asm/numa.h52
-rw-r--r--arch/x86/include/asm/numa_32.h7
-rw-r--r--arch/x86/include/asm/numa_64.h23
-rw-r--r--arch/x86/include/asm/olpc_ofw.h14
-rw-r--r--arch/x86/include/asm/page_types.h9
-rw-r--r--arch/x86/include/asm/processor.h4
-rw-r--r--arch/x86/include/asm/prom.h70
-rw-r--r--arch/x86/include/asm/rwsem.h80
-rw-r--r--arch/x86/include/asm/smp.h20
-rw-r--r--arch/x86/include/asm/system.h2
-rw-r--r--arch/x86/include/asm/topology.h19
-rw-r--r--arch/x86/include/asm/unistd_32.h5
-rw-r--r--arch/x86/include/asm/unistd_64.h6
-rw-r--r--arch/x86/include/asm/x86_init.h2
-rw-r--r--arch/x86/include/asm/xen/hypercall.h15
-rw-r--r--arch/x86/include/asm/xen/page.h47
-rw-r--r--arch/x86/include/asm/xen/pci.h8
-rw-r--r--arch/x86/kernel/Makefile5
-rw-r--r--arch/x86/kernel/acpi/boot.c8
-rw-r--r--arch/x86/kernel/apb_timer.c60
-rw-r--r--arch/x86/kernel/aperture_64.c33
-rw-r--r--arch/x86/kernel/apic/apic.c150
-rw-r--r--arch/x86/kernel/apic/apic_flat_64.c4
-rw-r--r--arch/x86/kernel/apic/apic_noop.c26
-rw-r--r--arch/x86/kernel/apic/bigsmp_32.c34
-rw-r--r--arch/x86/kernel/apic/es7000_32.c35
-rw-r--r--arch/x86/kernel/apic/hw_nmi.c1
-rw-r--r--arch/x86/kernel/apic/io_apic.c388
-rw-r--r--arch/x86/kernel/apic/ipi.c12
-rw-r--r--arch/x86/kernel/apic/numaq_32.c21
-rw-r--r--arch/x86/kernel/apic/probe_32.c10
-rw-r--r--arch/x86/kernel/apic/summit_32.c47
-rw-r--r--arch/x86/kernel/apic/x2apic_cluster.c2
-rw-r--r--arch/x86/kernel/apic/x2apic_phys.c2
-rw-r--r--arch/x86/kernel/apic/x2apic_uv_x.c2
-rw-r--r--arch/x86/kernel/asm-offsets.c65
-rw-r--r--arch/x86/kernel/asm-offsets_32.c69
-rw-r--r--arch/x86/kernel/asm-offsets_64.c90
-rw-r--r--arch/x86/kernel/cpu/amd.c51
-rw-r--r--arch/x86/kernel/cpu/common.c6
-rw-r--r--arch/x86/kernel/cpu/intel.c5
-rw-r--r--arch/x86/kernel/cpu/intel_cacheinfo.c4
-rw-r--r--arch/x86/kernel/cpu/mcheck/mce_amd.c7
-rw-r--r--arch/x86/kernel/cpu/perf_event.c170
-rw-r--r--arch/x86/kernel/cpu/perf_event_amd.c175
-rw-r--r--arch/x86/kernel/cpu/perf_event_intel.c417
-rw-r--r--arch/x86/kernel/cpu/perf_event_intel_ds.c97
-rw-r--r--arch/x86/kernel/cpu/perf_event_p4.c8
-rw-r--r--arch/x86/kernel/cpu/perf_event_p6.c4
-rw-r--r--arch/x86/kernel/cpu/perfctr-watchdog.c4
-rw-r--r--arch/x86/kernel/devicetree.c441
-rw-r--r--arch/x86/kernel/dumpstack.c25
-rw-r--r--arch/x86/kernel/e820.c18
-rw-r--r--arch/x86/kernel/entry_32.S11
-rw-r--r--arch/x86/kernel/entry_64.S13
-rw-r--r--arch/x86/kernel/ftrace.c15
-rw-r--r--arch/x86/kernel/head_32.S10
-rw-r--r--arch/x86/kernel/hpet.c2
-rw-r--r--arch/x86/kernel/i8259.c2
-rw-r--r--arch/x86/kernel/ioport.c20
-rw-r--r--arch/x86/kernel/irq.c91
-rw-r--r--arch/x86/kernel/irqinit.c92
-rw-r--r--arch/x86/kernel/kgdb.c9
-rw-r--r--arch/x86/kernel/kprobes.c8
-rw-r--r--arch/x86/kernel/microcode_amd.c188
-rw-r--r--arch/x86/kernel/microcode_core.c6
-rw-r--r--arch/x86/kernel/process.c9
-rw-r--r--arch/x86/kernel/rtc.c3
-rw-r--r--arch/x86/kernel/setup.c76
-rw-r--r--arch/x86/kernel/setup_percpu.c11
-rw-r--r--arch/x86/kernel/smpboot.c123
-rw-r--r--arch/x86/kernel/syscall_table_32.S3
-rw-r--r--arch/x86/kernel/vmlinux.lds.S3
-rw-r--r--arch/x86/kernel/x8664_ksyms_64.c1
-rw-r--r--arch/x86/kernel/x86_init.c1
-rw-r--r--arch/x86/kvm/trace.h8
-rw-r--r--arch/x86/lguest/boot.c4
-rw-r--r--arch/x86/lib/atomic64_386_32.S6
-rw-r--r--arch/x86/lib/atomic64_cx8_32.S6
-rw-r--r--arch/x86/lib/checksum_32.S63
-rw-r--r--arch/x86/lib/memmove_64.S197
-rw-r--r--arch/x86/lib/memmove_64.c192
-rw-r--r--arch/x86/lib/rwsem_64.S56
-rw-r--r--arch/x86/lib/semaphore_32.S38
-rw-r--r--arch/x86/lib/thunk_32.S18
-rw-r--r--arch/x86/lib/thunk_64.S27
-rw-r--r--arch/x86/mm/Makefile1
-rw-r--r--arch/x86/mm/amdtopology_64.c142
-rw-r--r--arch/x86/mm/init.c56
-rw-r--r--arch/x86/mm/init_32.c11
-rw-r--r--arch/x86/mm/init_64.c72
-rw-r--r--arch/x86/mm/numa.c212
-rw-r--r--arch/x86/mm/numa_32.c10
-rw-r--r--arch/x86/mm/numa_64.c984
-rw-r--r--arch/x86/mm/numa_emulation.c494
-rw-r--r--arch/x86/mm/numa_internal.h31
-rw-r--r--arch/x86/mm/srat_32.c6
-rw-r--r--arch/x86/mm/srat_64.c367
-rw-r--r--arch/x86/mm/tlb.c14
-rw-r--r--arch/x86/pci/ce4100.c2
-rw-r--r--arch/x86/pci/xen.c159
-rw-r--r--arch/x86/platform/ce4100/ce4100.c24
-rw-r--r--arch/x86/platform/ce4100/falconfalls.dts428
-rw-r--r--arch/x86/platform/mrst/mrst.c2
-rw-r--r--arch/x86/platform/mrst/vrtc.c16
-rw-r--r--arch/x86/platform/olpc/Makefile4
-rw-r--r--arch/x86/platform/uv/uv_irq.c4
-rw-r--r--arch/x86/platform/visws/visws_quirks.c4
-rw-r--r--arch/x86/xen/Kconfig8
-rw-r--r--arch/x86/xen/enlighten.c8
-rw-r--r--arch/x86/xen/mmu.c74
-rw-r--r--arch/x86/xen/p2m.c330
-rw-r--r--arch/x86/xen/setup.c68
-rw-r--r--arch/x86/xen/smp.c38
-rw-r--r--arch/x86/xen/suspend.c8
-rw-r--r--arch/x86/xen/time.c4
-rw-r--r--arch/x86/xen/xen-head.S4
-rw-r--r--arch/x86/xen/xen-ops.h2
-rw-r--r--arch/xtensa/include/asm/rwsem.h37
-rw-r--r--arch/xtensa/kernel/time.c6
213 files changed, 5356 insertions, 3965 deletions
diff --git a/arch/alpha/include/asm/fcntl.h b/arch/alpha/include/asm/fcntl.h
index 70145cbb21cb..1b71ca70c9f6 100644
--- a/arch/alpha/include/asm/fcntl.h
+++ b/arch/alpha/include/asm/fcntl.h
@@ -31,6 +31,8 @@
31#define __O_SYNC 020000000 31#define __O_SYNC 020000000
32#define O_SYNC (__O_SYNC|O_DSYNC) 32#define O_SYNC (__O_SYNC|O_DSYNC)
33 33
34#define O_PATH 040000000
35
34#define F_GETLK 7 36#define F_GETLK 7
35#define F_SETLK 8 37#define F_SETLK 8
36#define F_SETLKW 9 38#define F_SETLKW 9
diff --git a/arch/alpha/include/asm/futex.h b/arch/alpha/include/asm/futex.h
index 945de222ab91..e8a761aee088 100644
--- a/arch/alpha/include/asm/futex.h
+++ b/arch/alpha/include/asm/futex.h
@@ -29,7 +29,7 @@
29 : "r" (uaddr), "r"(oparg) \ 29 : "r" (uaddr), "r"(oparg) \
30 : "memory") 30 : "memory")
31 31
32static inline int futex_atomic_op_inuser (int encoded_op, int __user *uaddr) 32static inline int futex_atomic_op_inuser (int encoded_op, u32 __user *uaddr)
33{ 33{
34 int op = (encoded_op >> 28) & 7; 34 int op = (encoded_op >> 28) & 7;
35 int cmp = (encoded_op >> 24) & 15; 35 int cmp = (encoded_op >> 24) & 15;
@@ -39,7 +39,7 @@ static inline int futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
39 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28)) 39 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
40 oparg = 1 << oparg; 40 oparg = 1 << oparg;
41 41
42 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int))) 42 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
43 return -EFAULT; 43 return -EFAULT;
44 44
45 pagefault_disable(); 45 pagefault_disable();
@@ -81,21 +81,23 @@ static inline int futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
81} 81}
82 82
83static inline int 83static inline int
84futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval) 84futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
85 u32 oldval, u32 newval)
85{ 86{
86 int prev, cmp; 87 int ret = 0, cmp;
88 u32 prev;
87 89
88 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int))) 90 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
89 return -EFAULT; 91 return -EFAULT;
90 92
91 __asm__ __volatile__ ( 93 __asm__ __volatile__ (
92 __ASM_SMP_MB 94 __ASM_SMP_MB
93 "1: ldl_l %0,0(%2)\n" 95 "1: ldl_l %1,0(%3)\n"
94 " cmpeq %0,%3,%1\n" 96 " cmpeq %1,%4,%2\n"
95 " beq %1,3f\n" 97 " beq %2,3f\n"
96 " mov %4,%1\n" 98 " mov %5,%2\n"
97 "2: stl_c %1,0(%2)\n" 99 "2: stl_c %2,0(%3)\n"
98 " beq %1,4f\n" 100 " beq %2,4f\n"
99 "3: .subsection 2\n" 101 "3: .subsection 2\n"
100 "4: br 1b\n" 102 "4: br 1b\n"
101 " .previous\n" 103 " .previous\n"
@@ -105,11 +107,12 @@ futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval)
105 " .long 2b-.\n" 107 " .long 2b-.\n"
106 " lda $31,3b-2b(%0)\n" 108 " lda $31,3b-2b(%0)\n"
107 " .previous\n" 109 " .previous\n"
108 : "=&r"(prev), "=&r"(cmp) 110 : "+r"(ret), "=&r"(prev), "=&r"(cmp)
109 : "r"(uaddr), "r"((long)oldval), "r"(newval) 111 : "r"(uaddr), "r"((long)oldval), "r"(newval)
110 : "memory"); 112 : "memory");
111 113
112 return prev; 114 *uval = prev;
115 return ret;
113} 116}
114 117
115#endif /* __KERNEL__ */ 118#endif /* __KERNEL__ */
diff --git a/arch/alpha/include/asm/rwsem.h b/arch/alpha/include/asm/rwsem.h
index 1570c0b54336..a83bbea62c67 100644
--- a/arch/alpha/include/asm/rwsem.h
+++ b/arch/alpha/include/asm/rwsem.h
@@ -13,44 +13,13 @@
13#ifdef __KERNEL__ 13#ifdef __KERNEL__
14 14
15#include <linux/compiler.h> 15#include <linux/compiler.h>
16#include <linux/list.h>
17#include <linux/spinlock.h>
18 16
19struct rwsem_waiter;
20
21extern struct rw_semaphore *rwsem_down_read_failed(struct rw_semaphore *sem);
22extern struct rw_semaphore *rwsem_down_write_failed(struct rw_semaphore *sem);
23extern struct rw_semaphore *rwsem_wake(struct rw_semaphore *);
24extern struct rw_semaphore *rwsem_downgrade_wake(struct rw_semaphore *sem);
25
26/*
27 * the semaphore definition
28 */
29struct rw_semaphore {
30 long count;
31#define RWSEM_UNLOCKED_VALUE 0x0000000000000000L 17#define RWSEM_UNLOCKED_VALUE 0x0000000000000000L
32#define RWSEM_ACTIVE_BIAS 0x0000000000000001L 18#define RWSEM_ACTIVE_BIAS 0x0000000000000001L
33#define RWSEM_ACTIVE_MASK 0x00000000ffffffffL 19#define RWSEM_ACTIVE_MASK 0x00000000ffffffffL
34#define RWSEM_WAITING_BIAS (-0x0000000100000000L) 20#define RWSEM_WAITING_BIAS (-0x0000000100000000L)
35#define RWSEM_ACTIVE_READ_BIAS RWSEM_ACTIVE_BIAS 21#define RWSEM_ACTIVE_READ_BIAS RWSEM_ACTIVE_BIAS
36#define RWSEM_ACTIVE_WRITE_BIAS (RWSEM_WAITING_BIAS + RWSEM_ACTIVE_BIAS) 22#define RWSEM_ACTIVE_WRITE_BIAS (RWSEM_WAITING_BIAS + RWSEM_ACTIVE_BIAS)
37 spinlock_t wait_lock;
38 struct list_head wait_list;
39};
40
41#define __RWSEM_INITIALIZER(name) \
42 { RWSEM_UNLOCKED_VALUE, SPIN_LOCK_UNLOCKED, \
43 LIST_HEAD_INIT((name).wait_list) }
44
45#define DECLARE_RWSEM(name) \
46 struct rw_semaphore name = __RWSEM_INITIALIZER(name)
47
48static inline void init_rwsem(struct rw_semaphore *sem)
49{
50 sem->count = RWSEM_UNLOCKED_VALUE;
51 spin_lock_init(&sem->wait_lock);
52 INIT_LIST_HEAD(&sem->wait_list);
53}
54 23
55static inline void __down_read(struct rw_semaphore *sem) 24static inline void __down_read(struct rw_semaphore *sem)
56{ 25{
@@ -250,10 +219,5 @@ static inline long rwsem_atomic_update(long val, struct rw_semaphore *sem)
250#endif 219#endif
251} 220}
252 221
253static inline int rwsem_is_locked(struct rw_semaphore *sem)
254{
255 return (sem->count != 0);
256}
257
258#endif /* __KERNEL__ */ 222#endif /* __KERNEL__ */
259#endif /* _ALPHA_RWSEM_H */ 223#endif /* _ALPHA_RWSEM_H */
diff --git a/arch/alpha/kernel/osf_sys.c b/arch/alpha/kernel/osf_sys.c
index fe698b5045e9..376f22130791 100644
--- a/arch/alpha/kernel/osf_sys.c
+++ b/arch/alpha/kernel/osf_sys.c
@@ -230,44 +230,24 @@ linux_to_osf_statfs(struct kstatfs *linux_stat, struct osf_statfs __user *osf_st
230 return copy_to_user(osf_stat, &tmp_stat, bufsiz) ? -EFAULT : 0; 230 return copy_to_user(osf_stat, &tmp_stat, bufsiz) ? -EFAULT : 0;
231} 231}
232 232
233static int 233SYSCALL_DEFINE3(osf_statfs, const char __user *, pathname,
234do_osf_statfs(struct path *path, struct osf_statfs __user *buffer, 234 struct osf_statfs __user *, buffer, unsigned long, bufsiz)
235 unsigned long bufsiz)
236{ 235{
237 struct kstatfs linux_stat; 236 struct kstatfs linux_stat;
238 int error = vfs_statfs(path, &linux_stat); 237 int error = user_statfs(pathname, &linux_stat);
239 if (!error) 238 if (!error)
240 error = linux_to_osf_statfs(&linux_stat, buffer, bufsiz); 239 error = linux_to_osf_statfs(&linux_stat, buffer, bufsiz);
241 return error; 240 return error;
242} 241}
243 242
244SYSCALL_DEFINE3(osf_statfs, const char __user *, pathname,
245 struct osf_statfs __user *, buffer, unsigned long, bufsiz)
246{
247 struct path path;
248 int retval;
249
250 retval = user_path(pathname, &path);
251 if (!retval) {
252 retval = do_osf_statfs(&path, buffer, bufsiz);
253 path_put(&path);
254 }
255 return retval;
256}
257
258SYSCALL_DEFINE3(osf_fstatfs, unsigned long, fd, 243SYSCALL_DEFINE3(osf_fstatfs, unsigned long, fd,
259 struct osf_statfs __user *, buffer, unsigned long, bufsiz) 244 struct osf_statfs __user *, buffer, unsigned long, bufsiz)
260{ 245{
261 struct file *file; 246 struct kstatfs linux_stat;
262 int retval; 247 int error = fd_statfs(fd, &linux_stat);
263 248 if (!error)
264 retval = -EBADF; 249 error = linux_to_osf_statfs(&linux_stat, buffer, bufsiz);
265 file = fget(fd); 250 return error;
266 if (file) {
267 retval = do_osf_statfs(&file->f_path, buffer, bufsiz);
268 fput(file);
269 }
270 return retval;
271} 251}
272 252
273/* 253/*
diff --git a/arch/alpha/kernel/time.c b/arch/alpha/kernel/time.c
index c1f3e7cb82a4..a58e84f1a63b 100644
--- a/arch/alpha/kernel/time.c
+++ b/arch/alpha/kernel/time.c
@@ -159,7 +159,7 @@ void read_persistent_clock(struct timespec *ts)
159 159
160/* 160/*
161 * timer_interrupt() needs to keep up the real-time clock, 161 * timer_interrupt() needs to keep up the real-time clock,
162 * as well as call the "do_timer()" routine every clocktick 162 * as well as call the "xtime_update()" routine every clocktick
163 */ 163 */
164irqreturn_t timer_interrupt(int irq, void *dev) 164irqreturn_t timer_interrupt(int irq, void *dev)
165{ 165{
@@ -172,8 +172,6 @@ irqreturn_t timer_interrupt(int irq, void *dev)
172 profile_tick(CPU_PROFILING); 172 profile_tick(CPU_PROFILING);
173#endif 173#endif
174 174
175 write_seqlock(&xtime_lock);
176
177 /* 175 /*
178 * Calculate how many ticks have passed since the last update, 176 * Calculate how many ticks have passed since the last update,
179 * including any previous partial leftover. Save any resulting 177 * including any previous partial leftover. Save any resulting
@@ -187,9 +185,7 @@ irqreturn_t timer_interrupt(int irq, void *dev)
187 nticks = delta >> FIX_SHIFT; 185 nticks = delta >> FIX_SHIFT;
188 186
189 if (nticks) 187 if (nticks)
190 do_timer(nticks); 188 xtime_update(nticks);
191
192 write_sequnlock(&xtime_lock);
193 189
194 if (test_irq_work_pending()) { 190 if (test_irq_work_pending()) {
195 clear_irq_work_pending(); 191 clear_irq_work_pending();
diff --git a/arch/arm/include/asm/futex.h b/arch/arm/include/asm/futex.h
index b33fe7065b38..199a6b6de7f4 100644
--- a/arch/arm/include/asm/futex.h
+++ b/arch/arm/include/asm/futex.h
@@ -35,7 +35,7 @@
35 : "cc", "memory") 35 : "cc", "memory")
36 36
37static inline int 37static inline int
38futex_atomic_op_inuser (int encoded_op, int __user *uaddr) 38futex_atomic_op_inuser (int encoded_op, u32 __user *uaddr)
39{ 39{
40 int op = (encoded_op >> 28) & 7; 40 int op = (encoded_op >> 28) & 7;
41 int cmp = (encoded_op >> 24) & 15; 41 int cmp = (encoded_op >> 24) & 15;
@@ -46,7 +46,7 @@ futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
46 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28)) 46 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
47 oparg = 1 << oparg; 47 oparg = 1 << oparg;
48 48
49 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int))) 49 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
50 return -EFAULT; 50 return -EFAULT;
51 51
52 pagefault_disable(); /* implies preempt_disable() */ 52 pagefault_disable(); /* implies preempt_disable() */
@@ -88,36 +88,35 @@ futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
88} 88}
89 89
90static inline int 90static inline int
91futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval) 91futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
92 u32 oldval, u32 newval)
92{ 93{
93 int val; 94 int ret = 0;
95 u32 val;
94 96
95 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int))) 97 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
96 return -EFAULT; 98 return -EFAULT;
97 99
98 pagefault_disable(); /* implies preempt_disable() */
99
100 __asm__ __volatile__("@futex_atomic_cmpxchg_inatomic\n" 100 __asm__ __volatile__("@futex_atomic_cmpxchg_inatomic\n"
101 "1: " T(ldr) " %0, [%3]\n" 101 "1: " T(ldr) " %1, [%4]\n"
102 " teq %0, %1\n" 102 " teq %1, %2\n"
103 " it eq @ explicit IT needed for the 2b label\n" 103 " it eq @ explicit IT needed for the 2b label\n"
104 "2: " T(streq) " %2, [%3]\n" 104 "2: " T(streq) " %3, [%4]\n"
105 "3:\n" 105 "3:\n"
106 " .pushsection __ex_table,\"a\"\n" 106 " .pushsection __ex_table,\"a\"\n"
107 " .align 3\n" 107 " .align 3\n"
108 " .long 1b, 4f, 2b, 4f\n" 108 " .long 1b, 4f, 2b, 4f\n"
109 " .popsection\n" 109 " .popsection\n"
110 " .pushsection .fixup,\"ax\"\n" 110 " .pushsection .fixup,\"ax\"\n"
111 "4: mov %0, %4\n" 111 "4: mov %0, %5\n"
112 " b 3b\n" 112 " b 3b\n"
113 " .popsection" 113 " .popsection"
114 : "=&r" (val) 114 : "+r" (ret), "=&r" (val)
115 : "r" (oldval), "r" (newval), "r" (uaddr), "Ir" (-EFAULT) 115 : "r" (oldval), "r" (newval), "r" (uaddr), "Ir" (-EFAULT)
116 : "cc", "memory"); 116 : "cc", "memory");
117 117
118 pagefault_enable(); /* subsumes preempt_enable() */ 118 *uval = val;
119 119 return ret;
120 return val;
121} 120}
122 121
123#endif /* !SMP */ 122#endif /* !SMP */
diff --git a/arch/arm/kernel/time.c b/arch/arm/kernel/time.c
index 3d76bf233734..1ff46cabc7ef 100644
--- a/arch/arm/kernel/time.c
+++ b/arch/arm/kernel/time.c
@@ -107,9 +107,7 @@ void timer_tick(void)
107{ 107{
108 profile_tick(CPU_PROFILING); 108 profile_tick(CPU_PROFILING);
109 do_leds(); 109 do_leds();
110 write_seqlock(&xtime_lock); 110 xtime_update(1);
111 do_timer(1);
112 write_sequnlock(&xtime_lock);
113#ifndef CONFIG_SMP 111#ifndef CONFIG_SMP
114 update_process_times(user_mode(get_irq_regs())); 112 update_process_times(user_mode(get_irq_regs()));
115#endif 113#endif
diff --git a/arch/arm/mach-clps711x/include/mach/time.h b/arch/arm/mach-clps711x/include/mach/time.h
index 8fe283ccd1f3..61fef9129c6a 100644
--- a/arch/arm/mach-clps711x/include/mach/time.h
+++ b/arch/arm/mach-clps711x/include/mach/time.h
@@ -30,7 +30,7 @@ p720t_timer_interrupt(int irq, void *dev_id)
30{ 30{
31 struct pt_regs *regs = get_irq_regs(); 31 struct pt_regs *regs = get_irq_regs();
32 do_leds(); 32 do_leds();
33 do_timer(1); 33 xtime_update(1);
34#ifndef CONFIG_SMP 34#ifndef CONFIG_SMP
35 update_process_times(user_mode(regs)); 35 update_process_times(user_mode(regs));
36#endif 36#endif
diff --git a/arch/blackfin/kernel/time.c b/arch/blackfin/kernel/time.c
index c9113619029f..8d73724c0092 100644
--- a/arch/blackfin/kernel/time.c
+++ b/arch/blackfin/kernel/time.c
@@ -114,16 +114,14 @@ u32 arch_gettimeoffset(void)
114 114
115/* 115/*
116 * timer_interrupt() needs to keep up the real-time clock, 116 * timer_interrupt() needs to keep up the real-time clock,
117 * as well as call the "do_timer()" routine every clocktick 117 * as well as call the "xtime_update()" routine every clocktick
118 */ 118 */
119#ifdef CONFIG_CORE_TIMER_IRQ_L1 119#ifdef CONFIG_CORE_TIMER_IRQ_L1
120__attribute__((l1_text)) 120__attribute__((l1_text))
121#endif 121#endif
122irqreturn_t timer_interrupt(int irq, void *dummy) 122irqreturn_t timer_interrupt(int irq, void *dummy)
123{ 123{
124 write_seqlock(&xtime_lock); 124 xtime_update(1);
125 do_timer(1);
126 write_sequnlock(&xtime_lock);
127 125
128#ifdef CONFIG_IPIPE 126#ifdef CONFIG_IPIPE
129 update_root_process_times(get_irq_regs()); 127 update_root_process_times(get_irq_regs());
diff --git a/arch/cris/arch-v10/kernel/time.c b/arch/cris/arch-v10/kernel/time.c
index 00eb36f8debf..20c85b5dc7d0 100644
--- a/arch/cris/arch-v10/kernel/time.c
+++ b/arch/cris/arch-v10/kernel/time.c
@@ -140,7 +140,7 @@ stop_watchdog(void)
140 140
141/* 141/*
142 * timer_interrupt() needs to keep up the real-time clock, 142 * timer_interrupt() needs to keep up the real-time clock,
143 * as well as call the "do_timer()" routine every clocktick 143 * as well as call the "xtime_update()" routine every clocktick
144 */ 144 */
145 145
146//static unsigned short myjiff; /* used by our debug routine print_timestamp */ 146//static unsigned short myjiff; /* used by our debug routine print_timestamp */
@@ -176,7 +176,7 @@ timer_interrupt(int irq, void *dev_id)
176 176
177 /* call the real timer interrupt handler */ 177 /* call the real timer interrupt handler */
178 178
179 do_timer(1); 179 xtime_update(1);
180 180
181 cris_do_profile(regs); /* Save profiling information */ 181 cris_do_profile(regs); /* Save profiling information */
182 return IRQ_HANDLED; 182 return IRQ_HANDLED;
diff --git a/arch/cris/arch-v32/kernel/smp.c b/arch/cris/arch-v32/kernel/smp.c
index 84fed3b4b079..4c9e3e1ba5d1 100644
--- a/arch/cris/arch-v32/kernel/smp.c
+++ b/arch/cris/arch-v32/kernel/smp.c
@@ -26,7 +26,9 @@
26#define FLUSH_ALL (void*)0xffffffff 26#define FLUSH_ALL (void*)0xffffffff
27 27
28/* Vector of locks used for various atomic operations */ 28/* Vector of locks used for various atomic operations */
29spinlock_t cris_atomic_locks[] = { [0 ... LOCK_COUNT - 1] = SPIN_LOCK_UNLOCKED}; 29spinlock_t cris_atomic_locks[] = {
30 [0 ... LOCK_COUNT - 1] = __SPIN_LOCK_UNLOCKED(cris_atomic_locks)
31};
30 32
31/* CPU masks */ 33/* CPU masks */
32cpumask_t phys_cpu_present_map = CPU_MASK_NONE; 34cpumask_t phys_cpu_present_map = CPU_MASK_NONE;
diff --git a/arch/cris/arch-v32/kernel/time.c b/arch/cris/arch-v32/kernel/time.c
index a545211e999d..bb978ede8985 100644
--- a/arch/cris/arch-v32/kernel/time.c
+++ b/arch/cris/arch-v32/kernel/time.c
@@ -183,7 +183,7 @@ void handle_watchdog_bite(struct pt_regs *regs)
183 183
184/* 184/*
185 * timer_interrupt() needs to keep up the real-time clock, 185 * timer_interrupt() needs to keep up the real-time clock,
186 * as well as call the "do_timer()" routine every clocktick. 186 * as well as call the "xtime_update()" routine every clocktick.
187 */ 187 */
188extern void cris_do_profile(struct pt_regs *regs); 188extern void cris_do_profile(struct pt_regs *regs);
189 189
@@ -216,9 +216,7 @@ static inline irqreturn_t timer_interrupt(int irq, void *dev_id)
216 return IRQ_HANDLED; 216 return IRQ_HANDLED;
217 217
218 /* Call the real timer interrupt handler */ 218 /* Call the real timer interrupt handler */
219 write_seqlock(&xtime_lock); 219 xtime_update(1);
220 do_timer(1);
221 write_sequnlock(&xtime_lock);
222 return IRQ_HANDLED; 220 return IRQ_HANDLED;
223} 221}
224 222
diff --git a/arch/frv/include/asm/futex.h b/arch/frv/include/asm/futex.h
index 08b3d1da3583..4bea27f50a7a 100644
--- a/arch/frv/include/asm/futex.h
+++ b/arch/frv/include/asm/futex.h
@@ -7,10 +7,11 @@
7#include <asm/errno.h> 7#include <asm/errno.h>
8#include <asm/uaccess.h> 8#include <asm/uaccess.h>
9 9
10extern int futex_atomic_op_inuser(int encoded_op, int __user *uaddr); 10extern int futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr);
11 11
12static inline int 12static inline int
13futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval) 13futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
14 u32 oldval, u32 newval)
14{ 15{
15 return -ENOSYS; 16 return -ENOSYS;
16} 17}
diff --git a/arch/frv/kernel/futex.c b/arch/frv/kernel/futex.c
index 14f64b054c7e..d155ca9e5098 100644
--- a/arch/frv/kernel/futex.c
+++ b/arch/frv/kernel/futex.c
@@ -18,7 +18,7 @@
18 * the various futex operations; MMU fault checking is ignored under no-MMU 18 * the various futex operations; MMU fault checking is ignored under no-MMU
19 * conditions 19 * conditions
20 */ 20 */
21static inline int atomic_futex_op_xchg_set(int oparg, int __user *uaddr, int *_oldval) 21static inline int atomic_futex_op_xchg_set(int oparg, u32 __user *uaddr, int *_oldval)
22{ 22{
23 int oldval, ret; 23 int oldval, ret;
24 24
@@ -50,7 +50,7 @@ static inline int atomic_futex_op_xchg_set(int oparg, int __user *uaddr, int *_o
50 return ret; 50 return ret;
51} 51}
52 52
53static inline int atomic_futex_op_xchg_add(int oparg, int __user *uaddr, int *_oldval) 53static inline int atomic_futex_op_xchg_add(int oparg, u32 __user *uaddr, int *_oldval)
54{ 54{
55 int oldval, ret; 55 int oldval, ret;
56 56
@@ -83,7 +83,7 @@ static inline int atomic_futex_op_xchg_add(int oparg, int __user *uaddr, int *_o
83 return ret; 83 return ret;
84} 84}
85 85
86static inline int atomic_futex_op_xchg_or(int oparg, int __user *uaddr, int *_oldval) 86static inline int atomic_futex_op_xchg_or(int oparg, u32 __user *uaddr, int *_oldval)
87{ 87{
88 int oldval, ret; 88 int oldval, ret;
89 89
@@ -116,7 +116,7 @@ static inline int atomic_futex_op_xchg_or(int oparg, int __user *uaddr, int *_ol
116 return ret; 116 return ret;
117} 117}
118 118
119static inline int atomic_futex_op_xchg_and(int oparg, int __user *uaddr, int *_oldval) 119static inline int atomic_futex_op_xchg_and(int oparg, u32 __user *uaddr, int *_oldval)
120{ 120{
121 int oldval, ret; 121 int oldval, ret;
122 122
@@ -149,7 +149,7 @@ static inline int atomic_futex_op_xchg_and(int oparg, int __user *uaddr, int *_o
149 return ret; 149 return ret;
150} 150}
151 151
152static inline int atomic_futex_op_xchg_xor(int oparg, int __user *uaddr, int *_oldval) 152static inline int atomic_futex_op_xchg_xor(int oparg, u32 __user *uaddr, int *_oldval)
153{ 153{
154 int oldval, ret; 154 int oldval, ret;
155 155
@@ -186,7 +186,7 @@ static inline int atomic_futex_op_xchg_xor(int oparg, int __user *uaddr, int *_o
186/* 186/*
187 * do the futex operations 187 * do the futex operations
188 */ 188 */
189int futex_atomic_op_inuser(int encoded_op, int __user *uaddr) 189int futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr)
190{ 190{
191 int op = (encoded_op >> 28) & 7; 191 int op = (encoded_op >> 28) & 7;
192 int cmp = (encoded_op >> 24) & 15; 192 int cmp = (encoded_op >> 24) & 15;
@@ -197,7 +197,7 @@ int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
197 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28)) 197 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
198 oparg = 1 << oparg; 198 oparg = 1 << oparg;
199 199
200 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int))) 200 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
201 return -EFAULT; 201 return -EFAULT;
202 202
203 pagefault_disable(); 203 pagefault_disable();
diff --git a/arch/frv/kernel/time.c b/arch/frv/kernel/time.c
index 0ddbbae83cb2..b457de496b70 100644
--- a/arch/frv/kernel/time.c
+++ b/arch/frv/kernel/time.c
@@ -50,21 +50,13 @@ static struct irqaction timer_irq = {
50 50
51/* 51/*
52 * timer_interrupt() needs to keep up the real-time clock, 52 * timer_interrupt() needs to keep up the real-time clock,
53 * as well as call the "do_timer()" routine every clocktick 53 * as well as call the "xtime_update()" routine every clocktick
54 */ 54 */
55static irqreturn_t timer_interrupt(int irq, void *dummy) 55static irqreturn_t timer_interrupt(int irq, void *dummy)
56{ 56{
57 profile_tick(CPU_PROFILING); 57 profile_tick(CPU_PROFILING);
58 /*
59 * Here we are in the timer irq handler. We just have irqs locally
60 * disabled but we don't know if the timer_bh is running on the other
61 * CPU. We need to avoid to SMP race with it. NOTE: we don't need
62 * the irq version of write_lock because as just said we have irq
63 * locally disabled. -arca
64 */
65 write_seqlock(&xtime_lock);
66 58
67 do_timer(1); 59 xtime_update(1);
68 60
69#ifdef CONFIG_HEARTBEAT 61#ifdef CONFIG_HEARTBEAT
70 static unsigned short n; 62 static unsigned short n;
@@ -72,8 +64,6 @@ static irqreturn_t timer_interrupt(int irq, void *dummy)
72 __set_LEDS(n); 64 __set_LEDS(n);
73#endif /* CONFIG_HEARTBEAT */ 65#endif /* CONFIG_HEARTBEAT */
74 66
75 write_sequnlock(&xtime_lock);
76
77 update_process_times(user_mode(get_irq_regs())); 67 update_process_times(user_mode(get_irq_regs()));
78 68
79 return IRQ_HANDLED; 69 return IRQ_HANDLED;
diff --git a/arch/h8300/kernel/time.c b/arch/h8300/kernel/time.c
index 165005aff9df..32263a138aa6 100644
--- a/arch/h8300/kernel/time.c
+++ b/arch/h8300/kernel/time.c
@@ -35,9 +35,7 @@ void h8300_timer_tick(void)
35{ 35{
36 if (current->pid) 36 if (current->pid)
37 profile_tick(CPU_PROFILING); 37 profile_tick(CPU_PROFILING);
38 write_seqlock(&xtime_lock); 38 xtime_update(1);
39 do_timer(1);
40 write_sequnlock(&xtime_lock);
41 update_process_times(user_mode(get_irq_regs())); 39 update_process_times(user_mode(get_irq_regs()));
42} 40}
43 41
diff --git a/arch/h8300/kernel/timer/timer8.c b/arch/h8300/kernel/timer/timer8.c
index 3946c0fa8374..7a1533fad47d 100644
--- a/arch/h8300/kernel/timer/timer8.c
+++ b/arch/h8300/kernel/timer/timer8.c
@@ -61,7 +61,7 @@
61 61
62/* 62/*
63 * timer_interrupt() needs to keep up the real-time clock, 63 * timer_interrupt() needs to keep up the real-time clock,
64 * as well as call the "do_timer()" routine every clocktick 64 * as well as call the "xtime_update()" routine every clocktick
65 */ 65 */
66 66
67static irqreturn_t timer_interrupt(int irq, void *dev_id) 67static irqreturn_t timer_interrupt(int irq, void *dev_id)
diff --git a/arch/ia64/include/asm/futex.h b/arch/ia64/include/asm/futex.h
index c7f0f062239c..8428525ddb22 100644
--- a/arch/ia64/include/asm/futex.h
+++ b/arch/ia64/include/asm/futex.h
@@ -46,7 +46,7 @@ do { \
46} while (0) 46} while (0)
47 47
48static inline int 48static inline int
49futex_atomic_op_inuser (int encoded_op, int __user *uaddr) 49futex_atomic_op_inuser (int encoded_op, u32 __user *uaddr)
50{ 50{
51 int op = (encoded_op >> 28) & 7; 51 int op = (encoded_op >> 28) & 7;
52 int cmp = (encoded_op >> 24) & 15; 52 int cmp = (encoded_op >> 24) & 15;
@@ -56,7 +56,7 @@ futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
56 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28)) 56 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
57 oparg = 1 << oparg; 57 oparg = 1 << oparg;
58 58
59 if (! access_ok (VERIFY_WRITE, uaddr, sizeof(int))) 59 if (! access_ok (VERIFY_WRITE, uaddr, sizeof(u32)))
60 return -EFAULT; 60 return -EFAULT;
61 61
62 pagefault_disable(); 62 pagefault_disable();
@@ -100,23 +100,26 @@ futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
100} 100}
101 101
102static inline int 102static inline int
103futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval) 103futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
104 u32 oldval, u32 newval)
104{ 105{
105 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int))) 106 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
106 return -EFAULT; 107 return -EFAULT;
107 108
108 { 109 {
109 register unsigned long r8 __asm ("r8"); 110 register unsigned long r8 __asm ("r8") = 0;
111 unsigned long prev;
110 __asm__ __volatile__( 112 __asm__ __volatile__(
111 " mf;; \n" 113 " mf;; \n"
112 " mov ar.ccv=%3;; \n" 114 " mov ar.ccv=%3;; \n"
113 "[1:] cmpxchg4.acq %0=[%1],%2,ar.ccv \n" 115 "[1:] cmpxchg4.acq %0=[%1],%2,ar.ccv \n"
114 " .xdata4 \"__ex_table\", 1b-., 2f-. \n" 116 " .xdata4 \"__ex_table\", 1b-., 2f-. \n"
115 "[2:]" 117 "[2:]"
116 : "=r" (r8) 118 : "=r" (prev)
117 : "r" (uaddr), "r" (newval), 119 : "r" (uaddr), "r" (newval),
118 "rO" ((long) (unsigned) oldval) 120 "rO" ((long) (unsigned) oldval)
119 : "memory"); 121 : "memory");
122 *uval = prev;
120 return r8; 123 return r8;
121 } 124 }
122} 125}
diff --git a/arch/ia64/include/asm/rwsem.h b/arch/ia64/include/asm/rwsem.h
index 215d5454c7d3..3027e7516d85 100644
--- a/arch/ia64/include/asm/rwsem.h
+++ b/arch/ia64/include/asm/rwsem.h
@@ -25,20 +25,8 @@
25#error "Please don't include <asm/rwsem.h> directly, use <linux/rwsem.h> instead." 25#error "Please don't include <asm/rwsem.h> directly, use <linux/rwsem.h> instead."
26#endif 26#endif
27 27
28#include <linux/list.h>
29#include <linux/spinlock.h>
30
31#include <asm/intrinsics.h> 28#include <asm/intrinsics.h>
32 29
33/*
34 * the semaphore definition
35 */
36struct rw_semaphore {
37 signed long count;
38 spinlock_t wait_lock;
39 struct list_head wait_list;
40};
41
42#define RWSEM_UNLOCKED_VALUE __IA64_UL_CONST(0x0000000000000000) 30#define RWSEM_UNLOCKED_VALUE __IA64_UL_CONST(0x0000000000000000)
43#define RWSEM_ACTIVE_BIAS (1L) 31#define RWSEM_ACTIVE_BIAS (1L)
44#define RWSEM_ACTIVE_MASK (0xffffffffL) 32#define RWSEM_ACTIVE_MASK (0xffffffffL)
@@ -46,26 +34,6 @@ struct rw_semaphore {
46#define RWSEM_ACTIVE_READ_BIAS RWSEM_ACTIVE_BIAS 34#define RWSEM_ACTIVE_READ_BIAS RWSEM_ACTIVE_BIAS
47#define RWSEM_ACTIVE_WRITE_BIAS (RWSEM_WAITING_BIAS + RWSEM_ACTIVE_BIAS) 35#define RWSEM_ACTIVE_WRITE_BIAS (RWSEM_WAITING_BIAS + RWSEM_ACTIVE_BIAS)
48 36
49#define __RWSEM_INITIALIZER(name) \
50 { RWSEM_UNLOCKED_VALUE, __SPIN_LOCK_UNLOCKED((name).wait_lock), \
51 LIST_HEAD_INIT((name).wait_list) }
52
53#define DECLARE_RWSEM(name) \
54 struct rw_semaphore name = __RWSEM_INITIALIZER(name)
55
56extern struct rw_semaphore *rwsem_down_read_failed(struct rw_semaphore *sem);
57extern struct rw_semaphore *rwsem_down_write_failed(struct rw_semaphore *sem);
58extern struct rw_semaphore *rwsem_wake(struct rw_semaphore *sem);
59extern struct rw_semaphore *rwsem_downgrade_wake(struct rw_semaphore *sem);
60
61static inline void
62init_rwsem (struct rw_semaphore *sem)
63{
64 sem->count = RWSEM_UNLOCKED_VALUE;
65 spin_lock_init(&sem->wait_lock);
66 INIT_LIST_HEAD(&sem->wait_list);
67}
68
69/* 37/*
70 * lock for reading 38 * lock for reading
71 */ 39 */
@@ -174,9 +142,4 @@ __downgrade_write (struct rw_semaphore *sem)
174#define rwsem_atomic_add(delta, sem) atomic64_add(delta, (atomic64_t *)(&(sem)->count)) 142#define rwsem_atomic_add(delta, sem) atomic64_add(delta, (atomic64_t *)(&(sem)->count))
175#define rwsem_atomic_update(delta, sem) atomic64_add_return(delta, (atomic64_t *)(&(sem)->count)) 143#define rwsem_atomic_update(delta, sem) atomic64_add_return(delta, (atomic64_t *)(&(sem)->count))
176 144
177static inline int rwsem_is_locked(struct rw_semaphore *sem)
178{
179 return (sem->count != 0);
180}
181
182#endif /* _ASM_IA64_RWSEM_H */ 145#endif /* _ASM_IA64_RWSEM_H */
diff --git a/arch/ia64/include/asm/xen/hypercall.h b/arch/ia64/include/asm/xen/hypercall.h
index 96fc62366aa4..ed28bcd5bb85 100644
--- a/arch/ia64/include/asm/xen/hypercall.h
+++ b/arch/ia64/include/asm/xen/hypercall.h
@@ -107,7 +107,7 @@ extern unsigned long __hypercall(unsigned long a1, unsigned long a2,
107static inline int 107static inline int
108xencomm_arch_hypercall_sched_op(int cmd, struct xencomm_handle *arg) 108xencomm_arch_hypercall_sched_op(int cmd, struct xencomm_handle *arg)
109{ 109{
110 return _hypercall2(int, sched_op_new, cmd, arg); 110 return _hypercall2(int, sched_op, cmd, arg);
111} 111}
112 112
113static inline long 113static inline long
diff --git a/arch/ia64/kernel/time.c b/arch/ia64/kernel/time.c
index 9702fa92489e..156ad803d5b7 100644
--- a/arch/ia64/kernel/time.c
+++ b/arch/ia64/kernel/time.c
@@ -190,19 +190,10 @@ timer_interrupt (int irq, void *dev_id)
190 190
191 new_itm += local_cpu_data->itm_delta; 191 new_itm += local_cpu_data->itm_delta;
192 192
193 if (smp_processor_id() == time_keeper_id) { 193 if (smp_processor_id() == time_keeper_id)
194 /* 194 xtime_update(1);
195 * Here we are in the timer irq handler. We have irqs locally 195
196 * disabled, but we don't know if the timer_bh is running on 196 local_cpu_data->itm_next = new_itm;
197 * another CPU. We need to avoid to SMP race by acquiring the
198 * xtime_lock.
199 */
200 write_seqlock(&xtime_lock);
201 do_timer(1);
202 local_cpu_data->itm_next = new_itm;
203 write_sequnlock(&xtime_lock);
204 } else
205 local_cpu_data->itm_next = new_itm;
206 197
207 if (time_after(new_itm, ia64_get_itc())) 198 if (time_after(new_itm, ia64_get_itc()))
208 break; 199 break;
@@ -222,7 +213,7 @@ skip_process_time_accounting:
222 * comfort, we increase the safety margin by 213 * comfort, we increase the safety margin by
223 * intentionally dropping the next tick(s). We do NOT 214 * intentionally dropping the next tick(s). We do NOT
224 * update itm.next because that would force us to call 215 * update itm.next because that would force us to call
225 * do_timer() which in turn would let our clock run 216 * xtime_update() which in turn would let our clock run
226 * too fast (with the potentially devastating effect 217 * too fast (with the potentially devastating effect
227 * of losing monotony of time). 218 * of losing monotony of time).
228 */ 219 */
diff --git a/arch/ia64/xen/suspend.c b/arch/ia64/xen/suspend.c
index fd66b048c6fa..419c8620945a 100644
--- a/arch/ia64/xen/suspend.c
+++ b/arch/ia64/xen/suspend.c
@@ -37,19 +37,14 @@ xen_mm_unpin_all(void)
37 /* nothing */ 37 /* nothing */
38} 38}
39 39
40void xen_pre_device_suspend(void)
41{
42 /* nothing */
43}
44
45void 40void
46xen_pre_suspend() 41xen_arch_pre_suspend()
47{ 42{
48 /* nothing */ 43 /* nothing */
49} 44}
50 45
51void 46void
52xen_post_suspend(int suspend_cancelled) 47xen_arch_post_suspend(int suspend_cancelled)
53{ 48{
54 if (suspend_cancelled) 49 if (suspend_cancelled)
55 return; 50 return;
diff --git a/arch/ia64/xen/time.c b/arch/ia64/xen/time.c
index c1c544513e8d..1f8244a78bee 100644
--- a/arch/ia64/xen/time.c
+++ b/arch/ia64/xen/time.c
@@ -139,14 +139,11 @@ consider_steal_time(unsigned long new_itm)
139 run_posix_cpu_timers(p); 139 run_posix_cpu_timers(p);
140 delta_itm += local_cpu_data->itm_delta * (stolen + blocked); 140 delta_itm += local_cpu_data->itm_delta * (stolen + blocked);
141 141
142 if (cpu == time_keeper_id) { 142 if (cpu == time_keeper_id)
143 write_seqlock(&xtime_lock); 143 xtime_update(stolen + blocked);
144 do_timer(stolen + blocked); 144
145 local_cpu_data->itm_next = delta_itm + new_itm; 145 local_cpu_data->itm_next = delta_itm + new_itm;
146 write_sequnlock(&xtime_lock); 146
147 } else {
148 local_cpu_data->itm_next = delta_itm + new_itm;
149 }
150 per_cpu(xen_stolen_time, cpu) += NS_PER_TICK * stolen; 147 per_cpu(xen_stolen_time, cpu) += NS_PER_TICK * stolen;
151 per_cpu(xen_blocked_time, cpu) += NS_PER_TICK * blocked; 148 per_cpu(xen_blocked_time, cpu) += NS_PER_TICK * blocked;
152 } 149 }
diff --git a/arch/m32r/kernel/time.c b/arch/m32r/kernel/time.c
index bda86820bffd..84dd04048db9 100644
--- a/arch/m32r/kernel/time.c
+++ b/arch/m32r/kernel/time.c
@@ -107,15 +107,14 @@ u32 arch_gettimeoffset(void)
107 107
108/* 108/*
109 * timer_interrupt() needs to keep up the real-time clock, 109 * timer_interrupt() needs to keep up the real-time clock,
110 * as well as call the "do_timer()" routine every clocktick 110 * as well as call the "xtime_update()" routine every clocktick
111 */ 111 */
112static irqreturn_t timer_interrupt(int irq, void *dev_id) 112static irqreturn_t timer_interrupt(int irq, void *dev_id)
113{ 113{
114#ifndef CONFIG_SMP 114#ifndef CONFIG_SMP
115 profile_tick(CPU_PROFILING); 115 profile_tick(CPU_PROFILING);
116#endif 116#endif
117 /* XXX FIXME. Uh, the xtime_lock should be held here, no? */ 117 xtime_update(1);
118 do_timer(1);
119 118
120#ifndef CONFIG_SMP 119#ifndef CONFIG_SMP
121 update_process_times(user_mode(get_irq_regs())); 120 update_process_times(user_mode(get_irq_regs()));
diff --git a/arch/m68k/bvme6000/config.c b/arch/m68k/bvme6000/config.c
index 9fe6fefb5e14..1edd95095cb4 100644
--- a/arch/m68k/bvme6000/config.c
+++ b/arch/m68k/bvme6000/config.c
@@ -45,8 +45,8 @@ extern int bvme6000_set_clock_mmss (unsigned long);
45extern void bvme6000_reset (void); 45extern void bvme6000_reset (void);
46void bvme6000_set_vectors (void); 46void bvme6000_set_vectors (void);
47 47
48/* Save tick handler routine pointer, will point to do_timer() in 48/* Save tick handler routine pointer, will point to xtime_update() in
49 * kernel/sched.c, called via bvme6000_process_int() */ 49 * kernel/timer/timekeeping.c, called via bvme6000_process_int() */
50 50
51static irq_handler_t tick_handler; 51static irq_handler_t tick_handler;
52 52
diff --git a/arch/m68k/kernel/time.c b/arch/m68k/kernel/time.c
index 06438dac08ff..18b34ee5db3b 100644
--- a/arch/m68k/kernel/time.c
+++ b/arch/m68k/kernel/time.c
@@ -37,11 +37,11 @@ static inline int set_rtc_mmss(unsigned long nowtime)
37 37
38/* 38/*
39 * timer_interrupt() needs to keep up the real-time clock, 39 * timer_interrupt() needs to keep up the real-time clock,
40 * as well as call the "do_timer()" routine every clocktick 40 * as well as call the "xtime_update()" routine every clocktick
41 */ 41 */
42static irqreturn_t timer_interrupt(int irq, void *dummy) 42static irqreturn_t timer_interrupt(int irq, void *dummy)
43{ 43{
44 do_timer(1); 44 xtime_update(1);
45 update_process_times(user_mode(get_irq_regs())); 45 update_process_times(user_mode(get_irq_regs()));
46 profile_tick(CPU_PROFILING); 46 profile_tick(CPU_PROFILING);
47 47
diff --git a/arch/m68k/mvme147/config.c b/arch/m68k/mvme147/config.c
index 100baaa692a1..6cb9c3a9b6c9 100644
--- a/arch/m68k/mvme147/config.c
+++ b/arch/m68k/mvme147/config.c
@@ -46,8 +46,8 @@ extern void mvme147_reset (void);
46 46
47static int bcd2int (unsigned char b); 47static int bcd2int (unsigned char b);
48 48
49/* Save tick handler routine pointer, will point to do_timer() in 49/* Save tick handler routine pointer, will point to xtime_update() in
50 * kernel/sched.c, called via mvme147_process_int() */ 50 * kernel/time/timekeeping.c, called via mvme147_process_int() */
51 51
52irq_handler_t tick_handler; 52irq_handler_t tick_handler;
53 53
diff --git a/arch/m68k/mvme16x/config.c b/arch/m68k/mvme16x/config.c
index 11edf61cc2c4..0b28e2621653 100644
--- a/arch/m68k/mvme16x/config.c
+++ b/arch/m68k/mvme16x/config.c
@@ -51,8 +51,8 @@ extern void mvme16x_reset (void);
51 51
52int bcd2int (unsigned char b); 52int bcd2int (unsigned char b);
53 53
54/* Save tick handler routine pointer, will point to do_timer() in 54/* Save tick handler routine pointer, will point to xtime_update() in
55 * kernel/sched.c, called via mvme16x_process_int() */ 55 * kernel/time/timekeeping.c, called via mvme16x_process_int() */
56 56
57static irq_handler_t tick_handler; 57static irq_handler_t tick_handler;
58 58
diff --git a/arch/m68k/sun3/sun3ints.c b/arch/m68k/sun3/sun3ints.c
index 2d9e21bd313a..6464ad3ae3e6 100644
--- a/arch/m68k/sun3/sun3ints.c
+++ b/arch/m68k/sun3/sun3ints.c
@@ -66,7 +66,7 @@ static irqreturn_t sun3_int5(int irq, void *dev_id)
66#ifdef CONFIG_SUN3 66#ifdef CONFIG_SUN3
67 intersil_clear(); 67 intersil_clear();
68#endif 68#endif
69 do_timer(1); 69 xtime_update(1);
70 update_process_times(user_mode(get_irq_regs())); 70 update_process_times(user_mode(get_irq_regs()));
71 if (!(kstat_cpu(0).irqs[irq] % 20)) 71 if (!(kstat_cpu(0).irqs[irq] % 20))
72 sun3_leds(led_pattern[(kstat_cpu(0).irqs[irq] % 160) / 20]); 72 sun3_leds(led_pattern[(kstat_cpu(0).irqs[irq] % 160) / 20]);
diff --git a/arch/m68knommu/kernel/time.c b/arch/m68knommu/kernel/time.c
index d6ac2a43453c..6623909f70e6 100644
--- a/arch/m68knommu/kernel/time.c
+++ b/arch/m68knommu/kernel/time.c
@@ -36,7 +36,7 @@ static inline int set_rtc_mmss(unsigned long nowtime)
36#ifndef CONFIG_GENERIC_CLOCKEVENTS 36#ifndef CONFIG_GENERIC_CLOCKEVENTS
37/* 37/*
38 * timer_interrupt() needs to keep up the real-time clock, 38 * timer_interrupt() needs to keep up the real-time clock,
39 * as well as call the "do_timer()" routine every clocktick 39 * as well as call the "xtime_update()" routine every clocktick
40 */ 40 */
41irqreturn_t arch_timer_interrupt(int irq, void *dummy) 41irqreturn_t arch_timer_interrupt(int irq, void *dummy)
42{ 42{
@@ -44,11 +44,7 @@ irqreturn_t arch_timer_interrupt(int irq, void *dummy)
44 if (current->pid) 44 if (current->pid)
45 profile_tick(CPU_PROFILING); 45 profile_tick(CPU_PROFILING);
46 46
47 write_seqlock(&xtime_lock); 47 xtime_update(1);
48
49 do_timer(1);
50
51 write_sequnlock(&xtime_lock);
52 48
53 update_process_times(user_mode(get_irq_regs())); 49 update_process_times(user_mode(get_irq_regs()));
54 50
diff --git a/arch/microblaze/include/asm/futex.h b/arch/microblaze/include/asm/futex.h
index ad3fd61b2fe7..b0526d2716fa 100644
--- a/arch/microblaze/include/asm/futex.h
+++ b/arch/microblaze/include/asm/futex.h
@@ -29,7 +29,7 @@
29}) 29})
30 30
31static inline int 31static inline int
32futex_atomic_op_inuser(int encoded_op, int __user *uaddr) 32futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr)
33{ 33{
34 int op = (encoded_op >> 28) & 7; 34 int op = (encoded_op >> 28) & 7;
35 int cmp = (encoded_op >> 24) & 15; 35 int cmp = (encoded_op >> 24) & 15;
@@ -39,7 +39,7 @@ futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
39 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28)) 39 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
40 oparg = 1 << oparg; 40 oparg = 1 << oparg;
41 41
42 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int))) 42 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
43 return -EFAULT; 43 return -EFAULT;
44 44
45 pagefault_disable(); 45 pagefault_disable();
@@ -94,31 +94,34 @@ futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
94} 94}
95 95
96static inline int 96static inline int
97futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval) 97futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
98 u32 oldval, u32 newval)
98{ 99{
99 int prev, cmp; 100 int ret = 0, cmp;
101 u32 prev;
100 102
101 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int))) 103 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
102 return -EFAULT; 104 return -EFAULT;
103 105
104 __asm__ __volatile__ ("1: lwx %0, %2, r0; \ 106 __asm__ __volatile__ ("1: lwx %1, %3, r0; \
105 cmp %1, %0, %3; \ 107 cmp %2, %1, %4; \
106 beqi %1, 3f; \ 108 beqi %2, 3f; \
107 2: swx %4, %2, r0; \ 109 2: swx %5, %3, r0; \
108 addic %1, r0, 0; \ 110 addic %2, r0, 0; \
109 bnei %1, 1b; \ 111 bnei %2, 1b; \
110 3: \ 112 3: \
111 .section .fixup,\"ax\"; \ 113 .section .fixup,\"ax\"; \
112 4: brid 3b; \ 114 4: brid 3b; \
113 addik %0, r0, %5; \ 115 addik %0, r0, %6; \
114 .previous; \ 116 .previous; \
115 .section __ex_table,\"a\"; \ 117 .section __ex_table,\"a\"; \
116 .word 1b,4b,2b,4b; \ 118 .word 1b,4b,2b,4b; \
117 .previous;" \ 119 .previous;" \
118 : "=&r" (prev), "=&r"(cmp) \ 120 : "+r" (ret), "=&r" (prev), "=&r"(cmp) \
119 : "r" (uaddr), "r" (oldval), "r" (newval), "i" (-EFAULT)); 121 : "r" (uaddr), "r" (oldval), "r" (newval), "i" (-EFAULT));
120 122
121 return prev; 123 *uval = prev;
124 return ret;
122} 125}
123 126
124#endif /* __KERNEL__ */ 127#endif /* __KERNEL__ */
diff --git a/arch/microblaze/include/asm/pci-bridge.h b/arch/microblaze/include/asm/pci-bridge.h
index 0c68764ab547..10717669e0c2 100644
--- a/arch/microblaze/include/asm/pci-bridge.h
+++ b/arch/microblaze/include/asm/pci-bridge.h
@@ -104,11 +104,22 @@ struct pci_controller {
104 int global_number; /* PCI domain number */ 104 int global_number; /* PCI domain number */
105}; 105};
106 106
107#ifdef CONFIG_PCI
107static inline struct pci_controller *pci_bus_to_host(const struct pci_bus *bus) 108static inline struct pci_controller *pci_bus_to_host(const struct pci_bus *bus)
108{ 109{
109 return bus->sysdata; 110 return bus->sysdata;
110} 111}
111 112
113static inline struct device_node *pci_bus_to_OF_node(struct pci_bus *bus)
114{
115 struct pci_controller *host;
116
117 if (bus->self)
118 return pci_device_to_OF_node(bus->self);
119 host = pci_bus_to_host(bus);
120 return host ? host->dn : NULL;
121}
122
112static inline int isa_vaddr_is_ioport(void __iomem *address) 123static inline int isa_vaddr_is_ioport(void __iomem *address)
113{ 124{
114 /* No specific ISA handling on ppc32 at this stage, it 125 /* No specific ISA handling on ppc32 at this stage, it
@@ -116,6 +127,7 @@ static inline int isa_vaddr_is_ioport(void __iomem *address)
116 */ 127 */
117 return 0; 128 return 0;
118} 129}
130#endif /* CONFIG_PCI */
119 131
120/* These are used for config access before all the PCI probing 132/* These are used for config access before all the PCI probing
121 has been done. */ 133 has been done. */
diff --git a/arch/microblaze/include/asm/prom.h b/arch/microblaze/include/asm/prom.h
index 2e72af078b05..d0890d36ef61 100644
--- a/arch/microblaze/include/asm/prom.h
+++ b/arch/microblaze/include/asm/prom.h
@@ -64,21 +64,6 @@ extern void kdump_move_device_tree(void);
64/* CPU OF node matching */ 64/* CPU OF node matching */
65struct device_node *of_get_cpu_node(int cpu, unsigned int *thread); 65struct device_node *of_get_cpu_node(int cpu, unsigned int *thread);
66 66
67/**
68 * of_irq_map_pci - Resolve the interrupt for a PCI device
69 * @pdev: the device whose interrupt is to be resolved
70 * @out_irq: structure of_irq filled by this function
71 *
72 * This function resolves the PCI interrupt for a given PCI device. If a
73 * device-node exists for a given pci_dev, it will use normal OF tree
74 * walking. If not, it will implement standard swizzling and walk up the
75 * PCI tree until an device-node is found, at which point it will finish
76 * resolving using the OF tree walking.
77 */
78struct pci_dev;
79struct of_irq;
80extern int of_irq_map_pci(struct pci_dev *pdev, struct of_irq *out_irq);
81
82#endif /* __ASSEMBLY__ */ 67#endif /* __ASSEMBLY__ */
83#endif /* __KERNEL__ */ 68#endif /* __KERNEL__ */
84 69
diff --git a/arch/microblaze/kernel/prom_parse.c b/arch/microblaze/kernel/prom_parse.c
index 9ae24f4b882b..47187cc2cf00 100644
--- a/arch/microblaze/kernel/prom_parse.c
+++ b/arch/microblaze/kernel/prom_parse.c
@@ -2,88 +2,11 @@
2 2
3#include <linux/kernel.h> 3#include <linux/kernel.h>
4#include <linux/string.h> 4#include <linux/string.h>
5#include <linux/pci_regs.h>
6#include <linux/module.h> 5#include <linux/module.h>
7#include <linux/ioport.h> 6#include <linux/ioport.h>
8#include <linux/etherdevice.h> 7#include <linux/etherdevice.h>
9#include <linux/of_address.h> 8#include <linux/of_address.h>
10#include <asm/prom.h> 9#include <asm/prom.h>
11#include <asm/pci-bridge.h>
12
13#ifdef CONFIG_PCI
14int of_irq_map_pci(struct pci_dev *pdev, struct of_irq *out_irq)
15{
16 struct device_node *dn, *ppnode;
17 struct pci_dev *ppdev;
18 u32 lspec;
19 u32 laddr[3];
20 u8 pin;
21 int rc;
22
23 /* Check if we have a device node, if yes, fallback to standard OF
24 * parsing
25 */
26 dn = pci_device_to_OF_node(pdev);
27 if (dn)
28 return of_irq_map_one(dn, 0, out_irq);
29
30 /* Ok, we don't, time to have fun. Let's start by building up an
31 * interrupt spec. we assume #interrupt-cells is 1, which is standard
32 * for PCI. If you do different, then don't use that routine.
33 */
34 rc = pci_read_config_byte(pdev, PCI_INTERRUPT_PIN, &pin);
35 if (rc != 0)
36 return rc;
37 /* No pin, exit */
38 if (pin == 0)
39 return -ENODEV;
40
41 /* Now we walk up the PCI tree */
42 lspec = pin;
43 for (;;) {
44 /* Get the pci_dev of our parent */
45 ppdev = pdev->bus->self;
46
47 /* Ouch, it's a host bridge... */
48 if (ppdev == NULL) {
49 struct pci_controller *host;
50 host = pci_bus_to_host(pdev->bus);
51 ppnode = host ? host->dn : NULL;
52 /* No node for host bridge ? give up */
53 if (ppnode == NULL)
54 return -EINVAL;
55 } else
56 /* We found a P2P bridge, check if it has a node */
57 ppnode = pci_device_to_OF_node(ppdev);
58
59 /* Ok, we have found a parent with a device-node, hand over to
60 * the OF parsing code.
61 * We build a unit address from the linux device to be used for
62 * resolution. Note that we use the linux bus number which may
63 * not match your firmware bus numbering.
64 * Fortunately, in most cases, interrupt-map-mask doesn't
65 * include the bus number as part of the matching.
66 * You should still be careful about that though if you intend
67 * to rely on this function (you ship a firmware that doesn't
68 * create device nodes for all PCI devices).
69 */
70 if (ppnode)
71 break;
72
73 /* We can only get here if we hit a P2P bridge with no node,
74 * let's do standard swizzling and try again
75 */
76 lspec = pci_swizzle_interrupt_pin(pdev, lspec);
77 pdev = ppdev;
78 }
79
80 laddr[0] = (pdev->bus->number << 16)
81 | (pdev->devfn << 8);
82 laddr[1] = laddr[2] = 0;
83 return of_irq_map_raw(ppnode, &lspec, 1, laddr, out_irq);
84}
85EXPORT_SYMBOL_GPL(of_irq_map_pci);
86#endif /* CONFIG_PCI */
87 10
88void of_parse_dma_window(struct device_node *dn, const void *dma_window_prop, 11void of_parse_dma_window(struct device_node *dn, const void *dma_window_prop,
89 unsigned long *busno, unsigned long *phys, unsigned long *size) 12 unsigned long *busno, unsigned long *phys, unsigned long *size)
diff --git a/arch/microblaze/pci/pci-common.c b/arch/microblaze/pci/pci-common.c
index e363615d6798..1e01a1253631 100644
--- a/arch/microblaze/pci/pci-common.c
+++ b/arch/microblaze/pci/pci-common.c
@@ -29,6 +29,7 @@
29#include <linux/slab.h> 29#include <linux/slab.h>
30#include <linux/of.h> 30#include <linux/of.h>
31#include <linux/of_address.h> 31#include <linux/of_address.h>
32#include <linux/of_pci.h>
32 33
33#include <asm/processor.h> 34#include <asm/processor.h>
34#include <asm/io.h> 35#include <asm/io.h>
diff --git a/arch/mips/include/asm/futex.h b/arch/mips/include/asm/futex.h
index b9cce90346cf..6ebf1734b411 100644
--- a/arch/mips/include/asm/futex.h
+++ b/arch/mips/include/asm/futex.h
@@ -75,7 +75,7 @@
75} 75}
76 76
77static inline int 77static inline int
78futex_atomic_op_inuser(int encoded_op, int __user *uaddr) 78futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr)
79{ 79{
80 int op = (encoded_op >> 28) & 7; 80 int op = (encoded_op >> 28) & 7;
81 int cmp = (encoded_op >> 24) & 15; 81 int cmp = (encoded_op >> 24) & 15;
@@ -85,7 +85,7 @@ futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
85 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28)) 85 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
86 oparg = 1 << oparg; 86 oparg = 1 << oparg;
87 87
88 if (! access_ok (VERIFY_WRITE, uaddr, sizeof(int))) 88 if (! access_ok (VERIFY_WRITE, uaddr, sizeof(u32)))
89 return -EFAULT; 89 return -EFAULT;
90 90
91 pagefault_disable(); 91 pagefault_disable();
@@ -132,11 +132,13 @@ futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
132} 132}
133 133
134static inline int 134static inline int
135futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval) 135futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
136 u32 oldval, u32 newval)
136{ 137{
137 int retval; 138 int ret = 0;
139 u32 val;
138 140
139 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int))) 141 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
140 return -EFAULT; 142 return -EFAULT;
141 143
142 if (cpu_has_llsc && R10000_LLSC_WAR) { 144 if (cpu_has_llsc && R10000_LLSC_WAR) {
@@ -145,25 +147,25 @@ futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval)
145 " .set push \n" 147 " .set push \n"
146 " .set noat \n" 148 " .set noat \n"
147 " .set mips3 \n" 149 " .set mips3 \n"
148 "1: ll %0, %2 \n" 150 "1: ll %1, %3 \n"
149 " bne %0, %z3, 3f \n" 151 " bne %1, %z4, 3f \n"
150 " .set mips0 \n" 152 " .set mips0 \n"
151 " move $1, %z4 \n" 153 " move $1, %z5 \n"
152 " .set mips3 \n" 154 " .set mips3 \n"
153 "2: sc $1, %1 \n" 155 "2: sc $1, %2 \n"
154 " beqzl $1, 1b \n" 156 " beqzl $1, 1b \n"
155 __WEAK_LLSC_MB 157 __WEAK_LLSC_MB
156 "3: \n" 158 "3: \n"
157 " .set pop \n" 159 " .set pop \n"
158 " .section .fixup,\"ax\" \n" 160 " .section .fixup,\"ax\" \n"
159 "4: li %0, %5 \n" 161 "4: li %0, %6 \n"
160 " j 3b \n" 162 " j 3b \n"
161 " .previous \n" 163 " .previous \n"
162 " .section __ex_table,\"a\" \n" 164 " .section __ex_table,\"a\" \n"
163 " "__UA_ADDR "\t1b, 4b \n" 165 " "__UA_ADDR "\t1b, 4b \n"
164 " "__UA_ADDR "\t2b, 4b \n" 166 " "__UA_ADDR "\t2b, 4b \n"
165 " .previous \n" 167 " .previous \n"
166 : "=&r" (retval), "=R" (*uaddr) 168 : "+r" (ret), "=&r" (val), "=R" (*uaddr)
167 : "R" (*uaddr), "Jr" (oldval), "Jr" (newval), "i" (-EFAULT) 169 : "R" (*uaddr), "Jr" (oldval), "Jr" (newval), "i" (-EFAULT)
168 : "memory"); 170 : "memory");
169 } else if (cpu_has_llsc) { 171 } else if (cpu_has_llsc) {
@@ -172,31 +174,32 @@ futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval)
172 " .set push \n" 174 " .set push \n"
173 " .set noat \n" 175 " .set noat \n"
174 " .set mips3 \n" 176 " .set mips3 \n"
175 "1: ll %0, %2 \n" 177 "1: ll %1, %3 \n"
176 " bne %0, %z3, 3f \n" 178 " bne %1, %z4, 3f \n"
177 " .set mips0 \n" 179 " .set mips0 \n"
178 " move $1, %z4 \n" 180 " move $1, %z5 \n"
179 " .set mips3 \n" 181 " .set mips3 \n"
180 "2: sc $1, %1 \n" 182 "2: sc $1, %2 \n"
181 " beqz $1, 1b \n" 183 " beqz $1, 1b \n"
182 __WEAK_LLSC_MB 184 __WEAK_LLSC_MB
183 "3: \n" 185 "3: \n"
184 " .set pop \n" 186 " .set pop \n"
185 " .section .fixup,\"ax\" \n" 187 " .section .fixup,\"ax\" \n"
186 "4: li %0, %5 \n" 188 "4: li %0, %6 \n"
187 " j 3b \n" 189 " j 3b \n"
188 " .previous \n" 190 " .previous \n"
189 " .section __ex_table,\"a\" \n" 191 " .section __ex_table,\"a\" \n"
190 " "__UA_ADDR "\t1b, 4b \n" 192 " "__UA_ADDR "\t1b, 4b \n"
191 " "__UA_ADDR "\t2b, 4b \n" 193 " "__UA_ADDR "\t2b, 4b \n"
192 " .previous \n" 194 " .previous \n"
193 : "=&r" (retval), "=R" (*uaddr) 195 : "+r" (ret), "=&r" (val), "=R" (*uaddr)
194 : "R" (*uaddr), "Jr" (oldval), "Jr" (newval), "i" (-EFAULT) 196 : "R" (*uaddr), "Jr" (oldval), "Jr" (newval), "i" (-EFAULT)
195 : "memory"); 197 : "memory");
196 } else 198 } else
197 return -ENOSYS; 199 return -ENOSYS;
198 200
199 return retval; 201 *uval = val;
202 return ret;
200} 203}
201 204
202#endif 205#endif
diff --git a/arch/mn10300/kernel/time.c b/arch/mn10300/kernel/time.c
index 75da468090b9..5b955000626d 100644
--- a/arch/mn10300/kernel/time.c
+++ b/arch/mn10300/kernel/time.c
@@ -104,8 +104,6 @@ static irqreturn_t timer_interrupt(int irq, void *dev_id)
104 unsigned tsc, elapse; 104 unsigned tsc, elapse;
105 irqreturn_t ret; 105 irqreturn_t ret;
106 106
107 write_seqlock(&xtime_lock);
108
109 while (tsc = get_cycles(), 107 while (tsc = get_cycles(),
110 elapse = tsc - mn10300_last_tsc, /* time elapsed since last 108 elapse = tsc - mn10300_last_tsc, /* time elapsed since last
111 * tick */ 109 * tick */
@@ -114,11 +112,9 @@ static irqreturn_t timer_interrupt(int irq, void *dev_id)
114 mn10300_last_tsc += MN10300_TSC_PER_HZ; 112 mn10300_last_tsc += MN10300_TSC_PER_HZ;
115 113
116 /* advance the kernel's time tracking system */ 114 /* advance the kernel's time tracking system */
117 do_timer(1); 115 xtime_update(1);
118 } 116 }
119 117
120 write_sequnlock(&xtime_lock);
121
122 ret = local_timer_interrupt(); 118 ret = local_timer_interrupt();
123#ifdef CONFIG_SMP 119#ifdef CONFIG_SMP
124 send_IPI_allbutself(LOCAL_TIMER_IPI); 120 send_IPI_allbutself(LOCAL_TIMER_IPI);
diff --git a/arch/parisc/hpux/sys_hpux.c b/arch/parisc/hpux/sys_hpux.c
index 30394081d9b6..6ab9580b0b00 100644
--- a/arch/parisc/hpux/sys_hpux.c
+++ b/arch/parisc/hpux/sys_hpux.c
@@ -185,26 +185,21 @@ struct hpux_statfs {
185 int16_t f_pad; 185 int16_t f_pad;
186}; 186};
187 187
188static int do_statfs_hpux(struct path *path, struct hpux_statfs *buf) 188static int do_statfs_hpux(struct kstatfs *st, struct hpux_statfs __user *p)
189{ 189{
190 struct kstatfs st; 190 struct hpux_statfs buf;
191 int retval; 191 memset(&buf, 0, sizeof(buf));
192 192 buf.f_type = st->f_type;
193 retval = vfs_statfs(path, &st); 193 buf.f_bsize = st->f_bsize;
194 if (retval) 194 buf.f_blocks = st->f_blocks;
195 return retval; 195 buf.f_bfree = st->f_bfree;
196 196 buf.f_bavail = st->f_bavail;
197 memset(buf, 0, sizeof(*buf)); 197 buf.f_files = st->f_files;
198 buf->f_type = st.f_type; 198 buf.f_ffree = st->f_ffree;
199 buf->f_bsize = st.f_bsize; 199 buf.f_fsid[0] = st->f_fsid.val[0];
200 buf->f_blocks = st.f_blocks; 200 buf.f_fsid[1] = st->f_fsid.val[1];
201 buf->f_bfree = st.f_bfree; 201 if (copy_to_user(p, &buf, sizeof(buf)))
202 buf->f_bavail = st.f_bavail; 202 return -EFAULT;
203 buf->f_files = st.f_files;
204 buf->f_ffree = st.f_ffree;
205 buf->f_fsid[0] = st.f_fsid.val[0];
206 buf->f_fsid[1] = st.f_fsid.val[1];
207
208 return 0; 203 return 0;
209} 204}
210 205
@@ -212,35 +207,19 @@ static int do_statfs_hpux(struct path *path, struct hpux_statfs *buf)
212asmlinkage long hpux_statfs(const char __user *pathname, 207asmlinkage long hpux_statfs(const char __user *pathname,
213 struct hpux_statfs __user *buf) 208 struct hpux_statfs __user *buf)
214{ 209{
215 struct path path; 210 struct kstatfs st;
216 int error; 211 int error = user_statfs(pathname, &st);
217 212 if (!error)
218 error = user_path(pathname, &path); 213 error = do_statfs_hpux(&st, buf);
219 if (!error) {
220 struct hpux_statfs tmp;
221 error = do_statfs_hpux(&path, &tmp);
222 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
223 error = -EFAULT;
224 path_put(&path);
225 }
226 return error; 214 return error;
227} 215}
228 216
229asmlinkage long hpux_fstatfs(unsigned int fd, struct hpux_statfs __user * buf) 217asmlinkage long hpux_fstatfs(unsigned int fd, struct hpux_statfs __user * buf)
230{ 218{
231 struct file *file; 219 struct kstatfs st;
232 struct hpux_statfs tmp; 220 int error = fd_statfs(fd, &st);
233 int error; 221 if (!error)
234 222 error = do_statfs_hpux(&st, buf);
235 error = -EBADF;
236 file = fget(fd);
237 if (!file)
238 goto out;
239 error = do_statfs_hpux(&file->f_path, &tmp);
240 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
241 error = -EFAULT;
242 fput(file);
243 out:
244 return error; 223 return error;
245} 224}
246 225
diff --git a/arch/parisc/include/asm/fcntl.h b/arch/parisc/include/asm/fcntl.h
index f357fc693c89..0304b92ccfea 100644
--- a/arch/parisc/include/asm/fcntl.h
+++ b/arch/parisc/include/asm/fcntl.h
@@ -19,6 +19,8 @@
19#define O_NOFOLLOW 000000200 /* don't follow links */ 19#define O_NOFOLLOW 000000200 /* don't follow links */
20#define O_INVISIBLE 004000000 /* invisible I/O, for DMAPI/XDSM */ 20#define O_INVISIBLE 004000000 /* invisible I/O, for DMAPI/XDSM */
21 21
22#define O_PATH 020000000
23
22#define F_GETLK64 8 24#define F_GETLK64 8
23#define F_SETLK64 9 25#define F_SETLK64 9
24#define F_SETLKW64 10 26#define F_SETLKW64 10
diff --git a/arch/parisc/include/asm/futex.h b/arch/parisc/include/asm/futex.h
index 0c705c3a55ef..67a33cc27ef2 100644
--- a/arch/parisc/include/asm/futex.h
+++ b/arch/parisc/include/asm/futex.h
@@ -8,7 +8,7 @@
8#include <asm/errno.h> 8#include <asm/errno.h>
9 9
10static inline int 10static inline int
11futex_atomic_op_inuser (int encoded_op, int __user *uaddr) 11futex_atomic_op_inuser (int encoded_op, u32 __user *uaddr)
12{ 12{
13 int op = (encoded_op >> 28) & 7; 13 int op = (encoded_op >> 28) & 7;
14 int cmp = (encoded_op >> 24) & 15; 14 int cmp = (encoded_op >> 24) & 15;
@@ -18,7 +18,7 @@ futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
18 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28)) 18 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
19 oparg = 1 << oparg; 19 oparg = 1 << oparg;
20 20
21 if (! access_ok (VERIFY_WRITE, uaddr, sizeof(int))) 21 if (! access_ok (VERIFY_WRITE, uaddr, sizeof(u32)))
22 return -EFAULT; 22 return -EFAULT;
23 23
24 pagefault_disable(); 24 pagefault_disable();
@@ -51,10 +51,10 @@ futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
51 51
52/* Non-atomic version */ 52/* Non-atomic version */
53static inline int 53static inline int
54futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval) 54futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
55 u32 oldval, u32 newval)
55{ 56{
56 int err = 0; 57 u32 val;
57 int uval;
58 58
59 /* futex.c wants to do a cmpxchg_inatomic on kernel NULL, which is 59 /* futex.c wants to do a cmpxchg_inatomic on kernel NULL, which is
60 * our gateway page, and causes no end of trouble... 60 * our gateway page, and causes no end of trouble...
@@ -62,15 +62,15 @@ futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval)
62 if (segment_eq(KERNEL_DS, get_fs()) && !uaddr) 62 if (segment_eq(KERNEL_DS, get_fs()) && !uaddr)
63 return -EFAULT; 63 return -EFAULT;
64 64
65 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int))) 65 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
66 return -EFAULT; 66 return -EFAULT;
67 67
68 err = get_user(uval, uaddr); 68 if (get_user(val, uaddr))
69 if (err) return -EFAULT; 69 return -EFAULT;
70 if (uval == oldval) 70 if (val == oldval && put_user(newval, uaddr))
71 err = put_user(newval, uaddr); 71 return -EFAULT;
72 if (err) return -EFAULT; 72 *uval = val;
73 return uval; 73 return 0;
74} 74}
75 75
76#endif /*__KERNEL__*/ 76#endif /*__KERNEL__*/
diff --git a/arch/parisc/kernel/time.c b/arch/parisc/kernel/time.c
index 05511ccb61d2..45b7389d77aa 100644
--- a/arch/parisc/kernel/time.c
+++ b/arch/parisc/kernel/time.c
@@ -162,11 +162,8 @@ irqreturn_t __irq_entry timer_interrupt(int irq, void *dev_id)
162 update_process_times(user_mode(get_irq_regs())); 162 update_process_times(user_mode(get_irq_regs()));
163 } 163 }
164 164
165 if (cpu == 0) { 165 if (cpu == 0)
166 write_seqlock(&xtime_lock); 166 xtime_update(ticks_elapsed);
167 do_timer(ticks_elapsed);
168 write_sequnlock(&xtime_lock);
169 }
170 167
171 return IRQ_HANDLED; 168 return IRQ_HANDLED;
172} 169}
diff --git a/arch/powerpc/include/asm/futex.h b/arch/powerpc/include/asm/futex.h
index 7c589ef81fb0..c94e4a3fe2ef 100644
--- a/arch/powerpc/include/asm/futex.h
+++ b/arch/powerpc/include/asm/futex.h
@@ -30,7 +30,7 @@
30 : "b" (uaddr), "i" (-EFAULT), "r" (oparg) \ 30 : "b" (uaddr), "i" (-EFAULT), "r" (oparg) \
31 : "cr0", "memory") 31 : "cr0", "memory")
32 32
33static inline int futex_atomic_op_inuser (int encoded_op, int __user *uaddr) 33static inline int futex_atomic_op_inuser (int encoded_op, u32 __user *uaddr)
34{ 34{
35 int op = (encoded_op >> 28) & 7; 35 int op = (encoded_op >> 28) & 7;
36 int cmp = (encoded_op >> 24) & 15; 36 int cmp = (encoded_op >> 24) & 15;
@@ -40,7 +40,7 @@ static inline int futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
40 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28)) 40 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
41 oparg = 1 << oparg; 41 oparg = 1 << oparg;
42 42
43 if (! access_ok (VERIFY_WRITE, uaddr, sizeof(int))) 43 if (! access_ok (VERIFY_WRITE, uaddr, sizeof(u32)))
44 return -EFAULT; 44 return -EFAULT;
45 45
46 pagefault_disable(); 46 pagefault_disable();
@@ -82,35 +82,38 @@ static inline int futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
82} 82}
83 83
84static inline int 84static inline int
85futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval) 85futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
86 u32 oldval, u32 newval)
86{ 87{
87 int prev; 88 int ret = 0;
89 u32 prev;
88 90
89 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int))) 91 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
90 return -EFAULT; 92 return -EFAULT;
91 93
92 __asm__ __volatile__ ( 94 __asm__ __volatile__ (
93 PPC_RELEASE_BARRIER 95 PPC_RELEASE_BARRIER
94"1: lwarx %0,0,%2 # futex_atomic_cmpxchg_inatomic\n\ 96"1: lwarx %1,0,%3 # futex_atomic_cmpxchg_inatomic\n\
95 cmpw 0,%0,%3\n\ 97 cmpw 0,%1,%4\n\
96 bne- 3f\n" 98 bne- 3f\n"
97 PPC405_ERR77(0,%2) 99 PPC405_ERR77(0,%3)
98"2: stwcx. %4,0,%2\n\ 100"2: stwcx. %5,0,%3\n\
99 bne- 1b\n" 101 bne- 1b\n"
100 PPC_ACQUIRE_BARRIER 102 PPC_ACQUIRE_BARRIER
101"3: .section .fixup,\"ax\"\n\ 103"3: .section .fixup,\"ax\"\n\
1024: li %0,%5\n\ 1044: li %0,%6\n\
103 b 3b\n\ 105 b 3b\n\
104 .previous\n\ 106 .previous\n\
105 .section __ex_table,\"a\"\n\ 107 .section __ex_table,\"a\"\n\
106 .align 3\n\ 108 .align 3\n\
107 " PPC_LONG "1b,4b,2b,4b\n\ 109 " PPC_LONG "1b,4b,2b,4b\n\
108 .previous" \ 110 .previous" \
109 : "=&r" (prev), "+m" (*uaddr) 111 : "+r" (ret), "=&r" (prev), "+m" (*uaddr)
110 : "r" (uaddr), "r" (oldval), "r" (newval), "i" (-EFAULT) 112 : "r" (uaddr), "r" (oldval), "r" (newval), "i" (-EFAULT)
111 : "cc", "memory"); 113 : "cc", "memory");
112 114
113 return prev; 115 *uval = prev;
116 return ret;
114} 117}
115 118
116#endif /* __KERNEL__ */ 119#endif /* __KERNEL__ */
diff --git a/arch/powerpc/include/asm/pci-bridge.h b/arch/powerpc/include/asm/pci-bridge.h
index 51e9e6f90d12..edeb80fdd2c3 100644
--- a/arch/powerpc/include/asm/pci-bridge.h
+++ b/arch/powerpc/include/asm/pci-bridge.h
@@ -171,6 +171,16 @@ static inline struct pci_controller *pci_bus_to_host(const struct pci_bus *bus)
171 return bus->sysdata; 171 return bus->sysdata;
172} 172}
173 173
174static inline struct device_node *pci_bus_to_OF_node(struct pci_bus *bus)
175{
176 struct pci_controller *host;
177
178 if (bus->self)
179 return pci_device_to_OF_node(bus->self);
180 host = pci_bus_to_host(bus);
181 return host ? host->dn : NULL;
182}
183
174static inline int isa_vaddr_is_ioport(void __iomem *address) 184static inline int isa_vaddr_is_ioport(void __iomem *address)
175{ 185{
176 /* No specific ISA handling on ppc32 at this stage, it 186 /* No specific ISA handling on ppc32 at this stage, it
diff --git a/arch/powerpc/include/asm/prom.h b/arch/powerpc/include/asm/prom.h
index d72757585595..c189aa5fe1f4 100644
--- a/arch/powerpc/include/asm/prom.h
+++ b/arch/powerpc/include/asm/prom.h
@@ -70,21 +70,6 @@ static inline int of_node_to_nid(struct device_node *device) { return 0; }
70#endif 70#endif
71#define of_node_to_nid of_node_to_nid 71#define of_node_to_nid of_node_to_nid
72 72
73/**
74 * of_irq_map_pci - Resolve the interrupt for a PCI device
75 * @pdev: the device whose interrupt is to be resolved
76 * @out_irq: structure of_irq filled by this function
77 *
78 * This function resolves the PCI interrupt for a given PCI device. If a
79 * device-node exists for a given pci_dev, it will use normal OF tree
80 * walking. If not, it will implement standard swizzling and walk up the
81 * PCI tree until an device-node is found, at which point it will finish
82 * resolving using the OF tree walking.
83 */
84struct pci_dev;
85struct of_irq;
86extern int of_irq_map_pci(struct pci_dev *pdev, struct of_irq *out_irq);
87
88extern void of_instantiate_rtc(void); 73extern void of_instantiate_rtc(void);
89 74
90/* These includes are put at the bottom because they may contain things 75/* These includes are put at the bottom because they may contain things
diff --git a/arch/powerpc/include/asm/rwsem.h b/arch/powerpc/include/asm/rwsem.h
index 8447d89fbe72..bb1e2cdeb9bf 100644
--- a/arch/powerpc/include/asm/rwsem.h
+++ b/arch/powerpc/include/asm/rwsem.h
@@ -13,11 +13,6 @@
13 * by Paul Mackerras <paulus@samba.org>. 13 * by Paul Mackerras <paulus@samba.org>.
14 */ 14 */
15 15
16#include <linux/list.h>
17#include <linux/spinlock.h>
18#include <asm/atomic.h>
19#include <asm/system.h>
20
21/* 16/*
22 * the semaphore definition 17 * the semaphore definition
23 */ 18 */
@@ -33,47 +28,6 @@
33#define RWSEM_ACTIVE_READ_BIAS RWSEM_ACTIVE_BIAS 28#define RWSEM_ACTIVE_READ_BIAS RWSEM_ACTIVE_BIAS
34#define RWSEM_ACTIVE_WRITE_BIAS (RWSEM_WAITING_BIAS + RWSEM_ACTIVE_BIAS) 29#define RWSEM_ACTIVE_WRITE_BIAS (RWSEM_WAITING_BIAS + RWSEM_ACTIVE_BIAS)
35 30
36struct rw_semaphore {
37 long count;
38 spinlock_t wait_lock;
39 struct list_head wait_list;
40#ifdef CONFIG_DEBUG_LOCK_ALLOC
41 struct lockdep_map dep_map;
42#endif
43};
44
45#ifdef CONFIG_DEBUG_LOCK_ALLOC
46# define __RWSEM_DEP_MAP_INIT(lockname) , .dep_map = { .name = #lockname }
47#else
48# define __RWSEM_DEP_MAP_INIT(lockname)
49#endif
50
51#define __RWSEM_INITIALIZER(name) \
52{ \
53 RWSEM_UNLOCKED_VALUE, \
54 __SPIN_LOCK_UNLOCKED((name).wait_lock), \
55 LIST_HEAD_INIT((name).wait_list) \
56 __RWSEM_DEP_MAP_INIT(name) \
57}
58
59#define DECLARE_RWSEM(name) \
60 struct rw_semaphore name = __RWSEM_INITIALIZER(name)
61
62extern struct rw_semaphore *rwsem_down_read_failed(struct rw_semaphore *sem);
63extern struct rw_semaphore *rwsem_down_write_failed(struct rw_semaphore *sem);
64extern struct rw_semaphore *rwsem_wake(struct rw_semaphore *sem);
65extern struct rw_semaphore *rwsem_downgrade_wake(struct rw_semaphore *sem);
66
67extern void __init_rwsem(struct rw_semaphore *sem, const char *name,
68 struct lock_class_key *key);
69
70#define init_rwsem(sem) \
71 do { \
72 static struct lock_class_key __key; \
73 \
74 __init_rwsem((sem), #sem, &__key); \
75 } while (0)
76
77/* 31/*
78 * lock for reading 32 * lock for reading
79 */ 33 */
@@ -174,10 +128,5 @@ static inline long rwsem_atomic_update(long delta, struct rw_semaphore *sem)
174 return atomic_long_add_return(delta, (atomic_long_t *)&sem->count); 128 return atomic_long_add_return(delta, (atomic_long_t *)&sem->count);
175} 129}
176 130
177static inline int rwsem_is_locked(struct rw_semaphore *sem)
178{
179 return sem->count != 0;
180}
181
182#endif /* __KERNEL__ */ 131#endif /* __KERNEL__ */
183#endif /* _ASM_POWERPC_RWSEM_H */ 132#endif /* _ASM_POWERPC_RWSEM_H */
diff --git a/arch/powerpc/kernel/pci-common.c b/arch/powerpc/kernel/pci-common.c
index 10a44e68ef11..eb341be9a4d9 100644
--- a/arch/powerpc/kernel/pci-common.c
+++ b/arch/powerpc/kernel/pci-common.c
@@ -22,6 +22,7 @@
22#include <linux/init.h> 22#include <linux/init.h>
23#include <linux/bootmem.h> 23#include <linux/bootmem.h>
24#include <linux/of_address.h> 24#include <linux/of_address.h>
25#include <linux/of_pci.h>
25#include <linux/mm.h> 26#include <linux/mm.h>
26#include <linux/list.h> 27#include <linux/list.h>
27#include <linux/syscalls.h> 28#include <linux/syscalls.h>
diff --git a/arch/powerpc/kernel/prom_parse.c b/arch/powerpc/kernel/prom_parse.c
index c2b7a07cc3d3..47187cc2cf00 100644
--- a/arch/powerpc/kernel/prom_parse.c
+++ b/arch/powerpc/kernel/prom_parse.c
@@ -2,95 +2,11 @@
2 2
3#include <linux/kernel.h> 3#include <linux/kernel.h>
4#include <linux/string.h> 4#include <linux/string.h>
5#include <linux/pci_regs.h>
6#include <linux/module.h> 5#include <linux/module.h>
7#include <linux/ioport.h> 6#include <linux/ioport.h>
8#include <linux/etherdevice.h> 7#include <linux/etherdevice.h>
9#include <linux/of_address.h> 8#include <linux/of_address.h>
10#include <asm/prom.h> 9#include <asm/prom.h>
11#include <asm/pci-bridge.h>
12
13#ifdef CONFIG_PCI
14int of_irq_map_pci(struct pci_dev *pdev, struct of_irq *out_irq)
15{
16 struct device_node *dn, *ppnode;
17 struct pci_dev *ppdev;
18 u32 lspec;
19 u32 laddr[3];
20 u8 pin;
21 int rc;
22
23 /* Check if we have a device node, if yes, fallback to standard OF
24 * parsing
25 */
26 dn = pci_device_to_OF_node(pdev);
27 if (dn) {
28 rc = of_irq_map_one(dn, 0, out_irq);
29 if (!rc)
30 return rc;
31 }
32
33 /* Ok, we don't, time to have fun. Let's start by building up an
34 * interrupt spec. we assume #interrupt-cells is 1, which is standard
35 * for PCI. If you do different, then don't use that routine.
36 */
37 rc = pci_read_config_byte(pdev, PCI_INTERRUPT_PIN, &pin);
38 if (rc != 0)
39 return rc;
40 /* No pin, exit */
41 if (pin == 0)
42 return -ENODEV;
43
44 /* Now we walk up the PCI tree */
45 lspec = pin;
46 for (;;) {
47 /* Get the pci_dev of our parent */
48 ppdev = pdev->bus->self;
49
50 /* Ouch, it's a host bridge... */
51 if (ppdev == NULL) {
52#ifdef CONFIG_PPC64
53 ppnode = pci_bus_to_OF_node(pdev->bus);
54#else
55 struct pci_controller *host;
56 host = pci_bus_to_host(pdev->bus);
57 ppnode = host ? host->dn : NULL;
58#endif
59 /* No node for host bridge ? give up */
60 if (ppnode == NULL)
61 return -EINVAL;
62 } else
63 /* We found a P2P bridge, check if it has a node */
64 ppnode = pci_device_to_OF_node(ppdev);
65
66 /* Ok, we have found a parent with a device-node, hand over to
67 * the OF parsing code.
68 * We build a unit address from the linux device to be used for
69 * resolution. Note that we use the linux bus number which may
70 * not match your firmware bus numbering.
71 * Fortunately, in most cases, interrupt-map-mask doesn't include
72 * the bus number as part of the matching.
73 * You should still be careful about that though if you intend
74 * to rely on this function (you ship a firmware that doesn't
75 * create device nodes for all PCI devices).
76 */
77 if (ppnode)
78 break;
79
80 /* We can only get here if we hit a P2P bridge with no node,
81 * let's do standard swizzling and try again
82 */
83 lspec = pci_swizzle_interrupt_pin(pdev, lspec);
84 pdev = ppdev;
85 }
86
87 laddr[0] = (pdev->bus->number << 16)
88 | (pdev->devfn << 8);
89 laddr[1] = laddr[2] = 0;
90 return of_irq_map_raw(ppnode, &lspec, 1, laddr, out_irq);
91}
92EXPORT_SYMBOL_GPL(of_irq_map_pci);
93#endif /* CONFIG_PCI */
94 10
95void of_parse_dma_window(struct device_node *dn, const void *dma_window_prop, 11void of_parse_dma_window(struct device_node *dn, const void *dma_window_prop,
96 unsigned long *busno, unsigned long *phys, unsigned long *size) 12 unsigned long *busno, unsigned long *phys, unsigned long *size)
diff --git a/arch/powerpc/platforms/cell/spufs/syscalls.c b/arch/powerpc/platforms/cell/spufs/syscalls.c
index 187a7d32f86a..a3d2ce54ea2e 100644
--- a/arch/powerpc/platforms/cell/spufs/syscalls.c
+++ b/arch/powerpc/platforms/cell/spufs/syscalls.c
@@ -70,7 +70,7 @@ static long do_spu_create(const char __user *pathname, unsigned int flags,
70 if (!IS_ERR(tmp)) { 70 if (!IS_ERR(tmp)) {
71 struct nameidata nd; 71 struct nameidata nd;
72 72
73 ret = path_lookup(tmp, LOOKUP_PARENT, &nd); 73 ret = kern_path_parent(tmp, &nd);
74 if (!ret) { 74 if (!ret) {
75 nd.flags |= LOOKUP_OPEN | LOOKUP_CREATE; 75 nd.flags |= LOOKUP_OPEN | LOOKUP_CREATE;
76 ret = spufs_create(&nd, flags, mode, neighbor); 76 ret = spufs_create(&nd, flags, mode, neighbor);
diff --git a/arch/s390/include/asm/futex.h b/arch/s390/include/asm/futex.h
index 5c5d02de49e9..81cf36b691f1 100644
--- a/arch/s390/include/asm/futex.h
+++ b/arch/s390/include/asm/futex.h
@@ -7,7 +7,7 @@
7#include <linux/uaccess.h> 7#include <linux/uaccess.h>
8#include <asm/errno.h> 8#include <asm/errno.h>
9 9
10static inline int futex_atomic_op_inuser (int encoded_op, int __user *uaddr) 10static inline int futex_atomic_op_inuser (int encoded_op, u32 __user *uaddr)
11{ 11{
12 int op = (encoded_op >> 28) & 7; 12 int op = (encoded_op >> 28) & 7;
13 int cmp = (encoded_op >> 24) & 15; 13 int cmp = (encoded_op >> 24) & 15;
@@ -18,7 +18,7 @@ static inline int futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
18 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28)) 18 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
19 oparg = 1 << oparg; 19 oparg = 1 << oparg;
20 20
21 if (! access_ok (VERIFY_WRITE, uaddr, sizeof(int))) 21 if (! access_ok (VERIFY_WRITE, uaddr, sizeof(u32)))
22 return -EFAULT; 22 return -EFAULT;
23 23
24 pagefault_disable(); 24 pagefault_disable();
@@ -39,13 +39,13 @@ static inline int futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
39 return ret; 39 return ret;
40} 40}
41 41
42static inline int futex_atomic_cmpxchg_inatomic(int __user *uaddr, 42static inline int futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
43 int oldval, int newval) 43 u32 oldval, u32 newval)
44{ 44{
45 if (! access_ok (VERIFY_WRITE, uaddr, sizeof(int))) 45 if (! access_ok (VERIFY_WRITE, uaddr, sizeof(u32)))
46 return -EFAULT; 46 return -EFAULT;
47 47
48 return uaccess.futex_atomic_cmpxchg(uaddr, oldval, newval); 48 return uaccess.futex_atomic_cmpxchg(uval, uaddr, oldval, newval);
49} 49}
50 50
51#endif /* __KERNEL__ */ 51#endif /* __KERNEL__ */
diff --git a/arch/s390/include/asm/rwsem.h b/arch/s390/include/asm/rwsem.h
index 423fdda2322d..d0eb4653cebd 100644
--- a/arch/s390/include/asm/rwsem.h
+++ b/arch/s390/include/asm/rwsem.h
@@ -43,29 +43,6 @@
43 43
44#ifdef __KERNEL__ 44#ifdef __KERNEL__
45 45
46#include <linux/list.h>
47#include <linux/spinlock.h>
48
49struct rwsem_waiter;
50
51extern struct rw_semaphore *rwsem_down_read_failed(struct rw_semaphore *);
52extern struct rw_semaphore *rwsem_down_write_failed(struct rw_semaphore *);
53extern struct rw_semaphore *rwsem_wake(struct rw_semaphore *);
54extern struct rw_semaphore *rwsem_downgrade_wake(struct rw_semaphore *);
55extern struct rw_semaphore *rwsem_downgrade_write(struct rw_semaphore *);
56
57/*
58 * the semaphore definition
59 */
60struct rw_semaphore {
61 signed long count;
62 spinlock_t wait_lock;
63 struct list_head wait_list;
64#ifdef CONFIG_DEBUG_LOCK_ALLOC
65 struct lockdep_map dep_map;
66#endif
67};
68
69#ifndef __s390x__ 46#ifndef __s390x__
70#define RWSEM_UNLOCKED_VALUE 0x00000000 47#define RWSEM_UNLOCKED_VALUE 0x00000000
71#define RWSEM_ACTIVE_BIAS 0x00000001 48#define RWSEM_ACTIVE_BIAS 0x00000001
@@ -81,41 +58,6 @@ struct rw_semaphore {
81#define RWSEM_ACTIVE_WRITE_BIAS (RWSEM_WAITING_BIAS + RWSEM_ACTIVE_BIAS) 58#define RWSEM_ACTIVE_WRITE_BIAS (RWSEM_WAITING_BIAS + RWSEM_ACTIVE_BIAS)
82 59
83/* 60/*
84 * initialisation
85 */
86
87#ifdef CONFIG_DEBUG_LOCK_ALLOC
88# define __RWSEM_DEP_MAP_INIT(lockname) , .dep_map = { .name = #lockname }
89#else
90# define __RWSEM_DEP_MAP_INIT(lockname)
91#endif
92
93#define __RWSEM_INITIALIZER(name) \
94 { RWSEM_UNLOCKED_VALUE, __SPIN_LOCK_UNLOCKED((name).wait.lock), \
95 LIST_HEAD_INIT((name).wait_list) __RWSEM_DEP_MAP_INIT(name) }
96
97#define DECLARE_RWSEM(name) \
98 struct rw_semaphore name = __RWSEM_INITIALIZER(name)
99
100static inline void init_rwsem(struct rw_semaphore *sem)
101{
102 sem->count = RWSEM_UNLOCKED_VALUE;
103 spin_lock_init(&sem->wait_lock);
104 INIT_LIST_HEAD(&sem->wait_list);
105}
106
107extern void __init_rwsem(struct rw_semaphore *sem, const char *name,
108 struct lock_class_key *key);
109
110#define init_rwsem(sem) \
111do { \
112 static struct lock_class_key __key; \
113 \
114 __init_rwsem((sem), #sem, &__key); \
115} while (0)
116
117
118/*
119 * lock for reading 61 * lock for reading
120 */ 62 */
121static inline void __down_read(struct rw_semaphore *sem) 63static inline void __down_read(struct rw_semaphore *sem)
@@ -377,10 +319,5 @@ static inline long rwsem_atomic_update(long delta, struct rw_semaphore *sem)
377 return new; 319 return new;
378} 320}
379 321
380static inline int rwsem_is_locked(struct rw_semaphore *sem)
381{
382 return (sem->count != 0);
383}
384
385#endif /* __KERNEL__ */ 322#endif /* __KERNEL__ */
386#endif /* _S390_RWSEM_H */ 323#endif /* _S390_RWSEM_H */
diff --git a/arch/s390/include/asm/uaccess.h b/arch/s390/include/asm/uaccess.h
index d6b1ed0ec52b..2d9ea11f919a 100644
--- a/arch/s390/include/asm/uaccess.h
+++ b/arch/s390/include/asm/uaccess.h
@@ -83,8 +83,8 @@ struct uaccess_ops {
83 size_t (*clear_user)(size_t, void __user *); 83 size_t (*clear_user)(size_t, void __user *);
84 size_t (*strnlen_user)(size_t, const char __user *); 84 size_t (*strnlen_user)(size_t, const char __user *);
85 size_t (*strncpy_from_user)(size_t, const char __user *, char *); 85 size_t (*strncpy_from_user)(size_t, const char __user *, char *);
86 int (*futex_atomic_op)(int op, int __user *, int oparg, int *old); 86 int (*futex_atomic_op)(int op, u32 __user *, int oparg, int *old);
87 int (*futex_atomic_cmpxchg)(int __user *, int old, int new); 87 int (*futex_atomic_cmpxchg)(u32 *, u32 __user *, u32 old, u32 new);
88}; 88};
89 89
90extern struct uaccess_ops uaccess; 90extern struct uaccess_ops uaccess;
diff --git a/arch/s390/lib/uaccess.h b/arch/s390/lib/uaccess.h
index 126011df14f1..1d2536cb630b 100644
--- a/arch/s390/lib/uaccess.h
+++ b/arch/s390/lib/uaccess.h
@@ -12,12 +12,12 @@ extern size_t copy_from_user_std(size_t, const void __user *, void *);
12extern size_t copy_to_user_std(size_t, void __user *, const void *); 12extern size_t copy_to_user_std(size_t, void __user *, const void *);
13extern size_t strnlen_user_std(size_t, const char __user *); 13extern size_t strnlen_user_std(size_t, const char __user *);
14extern size_t strncpy_from_user_std(size_t, const char __user *, char *); 14extern size_t strncpy_from_user_std(size_t, const char __user *, char *);
15extern int futex_atomic_cmpxchg_std(int __user *, int, int); 15extern int futex_atomic_cmpxchg_std(u32 *, u32 __user *, u32, u32);
16extern int futex_atomic_op_std(int, int __user *, int, int *); 16extern int futex_atomic_op_std(int, u32 __user *, int, int *);
17 17
18extern size_t copy_from_user_pt(size_t, const void __user *, void *); 18extern size_t copy_from_user_pt(size_t, const void __user *, void *);
19extern size_t copy_to_user_pt(size_t, void __user *, const void *); 19extern size_t copy_to_user_pt(size_t, void __user *, const void *);
20extern int futex_atomic_op_pt(int, int __user *, int, int *); 20extern int futex_atomic_op_pt(int, u32 __user *, int, int *);
21extern int futex_atomic_cmpxchg_pt(int __user *, int, int); 21extern int futex_atomic_cmpxchg_pt(u32 *, u32 __user *, u32, u32);
22 22
23#endif /* __ARCH_S390_LIB_UACCESS_H */ 23#endif /* __ARCH_S390_LIB_UACCESS_H */
diff --git a/arch/s390/lib/uaccess_pt.c b/arch/s390/lib/uaccess_pt.c
index 404f2de296dc..74833831417f 100644
--- a/arch/s390/lib/uaccess_pt.c
+++ b/arch/s390/lib/uaccess_pt.c
@@ -302,7 +302,7 @@ fault:
302 : "0" (-EFAULT), "d" (oparg), "a" (uaddr), \ 302 : "0" (-EFAULT), "d" (oparg), "a" (uaddr), \
303 "m" (*uaddr) : "cc" ); 303 "m" (*uaddr) : "cc" );
304 304
305static int __futex_atomic_op_pt(int op, int __user *uaddr, int oparg, int *old) 305static int __futex_atomic_op_pt(int op, u32 __user *uaddr, int oparg, int *old)
306{ 306{
307 int oldval = 0, newval, ret; 307 int oldval = 0, newval, ret;
308 308
@@ -335,7 +335,7 @@ static int __futex_atomic_op_pt(int op, int __user *uaddr, int oparg, int *old)
335 return ret; 335 return ret;
336} 336}
337 337
338int futex_atomic_op_pt(int op, int __user *uaddr, int oparg, int *old) 338int futex_atomic_op_pt(int op, u32 __user *uaddr, int oparg, int *old)
339{ 339{
340 int ret; 340 int ret;
341 341
@@ -354,26 +354,29 @@ int futex_atomic_op_pt(int op, int __user *uaddr, int oparg, int *old)
354 return ret; 354 return ret;
355} 355}
356 356
357static int __futex_atomic_cmpxchg_pt(int __user *uaddr, int oldval, int newval) 357static int __futex_atomic_cmpxchg_pt(u32 *uval, u32 __user *uaddr,
358 u32 oldval, u32 newval)
358{ 359{
359 int ret; 360 int ret;
360 361
361 asm volatile("0: cs %1,%4,0(%5)\n" 362 asm volatile("0: cs %1,%4,0(%5)\n"
362 "1: lr %0,%1\n" 363 "1: la %0,0\n"
363 "2:\n" 364 "2:\n"
364 EX_TABLE(0b,2b) EX_TABLE(1b,2b) 365 EX_TABLE(0b,2b) EX_TABLE(1b,2b)
365 : "=d" (ret), "+d" (oldval), "=m" (*uaddr) 366 : "=d" (ret), "+d" (oldval), "=m" (*uaddr)
366 : "0" (-EFAULT), "d" (newval), "a" (uaddr), "m" (*uaddr) 367 : "0" (-EFAULT), "d" (newval), "a" (uaddr), "m" (*uaddr)
367 : "cc", "memory" ); 368 : "cc", "memory" );
369 *uval = oldval;
368 return ret; 370 return ret;
369} 371}
370 372
371int futex_atomic_cmpxchg_pt(int __user *uaddr, int oldval, int newval) 373int futex_atomic_cmpxchg_pt(u32 *uval, u32 __user *uaddr,
374 u32 oldval, u32 newval)
372{ 375{
373 int ret; 376 int ret;
374 377
375 if (segment_eq(get_fs(), KERNEL_DS)) 378 if (segment_eq(get_fs(), KERNEL_DS))
376 return __futex_atomic_cmpxchg_pt(uaddr, oldval, newval); 379 return __futex_atomic_cmpxchg_pt(uval, uaddr, oldval, newval);
377 spin_lock(&current->mm->page_table_lock); 380 spin_lock(&current->mm->page_table_lock);
378 uaddr = (int __user *) __dat_user_addr((unsigned long) uaddr); 381 uaddr = (int __user *) __dat_user_addr((unsigned long) uaddr);
379 if (!uaddr) { 382 if (!uaddr) {
@@ -382,7 +385,7 @@ int futex_atomic_cmpxchg_pt(int __user *uaddr, int oldval, int newval)
382 } 385 }
383 get_page(virt_to_page(uaddr)); 386 get_page(virt_to_page(uaddr));
384 spin_unlock(&current->mm->page_table_lock); 387 spin_unlock(&current->mm->page_table_lock);
385 ret = __futex_atomic_cmpxchg_pt(uaddr, oldval, newval); 388 ret = __futex_atomic_cmpxchg_pt(uval, uaddr, oldval, newval);
386 put_page(virt_to_page(uaddr)); 389 put_page(virt_to_page(uaddr));
387 return ret; 390 return ret;
388} 391}
diff --git a/arch/s390/lib/uaccess_std.c b/arch/s390/lib/uaccess_std.c
index a6c4f7ed24a4..bb1a7eed42ce 100644
--- a/arch/s390/lib/uaccess_std.c
+++ b/arch/s390/lib/uaccess_std.c
@@ -255,7 +255,7 @@ size_t strncpy_from_user_std(size_t size, const char __user *src, char *dst)
255 : "0" (-EFAULT), "d" (oparg), "a" (uaddr), \ 255 : "0" (-EFAULT), "d" (oparg), "a" (uaddr), \
256 "m" (*uaddr) : "cc"); 256 "m" (*uaddr) : "cc");
257 257
258int futex_atomic_op_std(int op, int __user *uaddr, int oparg, int *old) 258int futex_atomic_op_std(int op, u32 __user *uaddr, int oparg, int *old)
259{ 259{
260 int oldval = 0, newval, ret; 260 int oldval = 0, newval, ret;
261 261
@@ -287,19 +287,21 @@ int futex_atomic_op_std(int op, int __user *uaddr, int oparg, int *old)
287 return ret; 287 return ret;
288} 288}
289 289
290int futex_atomic_cmpxchg_std(int __user *uaddr, int oldval, int newval) 290int futex_atomic_cmpxchg_std(u32 *uval, u32 __user *uaddr,
291 u32 oldval, u32 newval)
291{ 292{
292 int ret; 293 int ret;
293 294
294 asm volatile( 295 asm volatile(
295 " sacf 256\n" 296 " sacf 256\n"
296 "0: cs %1,%4,0(%5)\n" 297 "0: cs %1,%4,0(%5)\n"
297 "1: lr %0,%1\n" 298 "1: la %0,0\n"
298 "2: sacf 0\n" 299 "2: sacf 0\n"
299 EX_TABLE(0b,2b) EX_TABLE(1b,2b) 300 EX_TABLE(0b,2b) EX_TABLE(1b,2b)
300 : "=d" (ret), "+d" (oldval), "=m" (*uaddr) 301 : "=d" (ret), "+d" (oldval), "=m" (*uaddr)
301 : "0" (-EFAULT), "d" (newval), "a" (uaddr), "m" (*uaddr) 302 : "0" (-EFAULT), "d" (newval), "a" (uaddr), "m" (*uaddr)
302 : "cc", "memory" ); 303 : "cc", "memory" );
304 *uval = oldval;
303 return ret; 305 return ret;
304} 306}
305 307
diff --git a/arch/sh/include/asm/futex-irq.h b/arch/sh/include/asm/futex-irq.h
index a9f16a7f9aea..6cb9f193a95e 100644
--- a/arch/sh/include/asm/futex-irq.h
+++ b/arch/sh/include/asm/futex-irq.h
@@ -3,7 +3,7 @@
3 3
4#include <asm/system.h> 4#include <asm/system.h>
5 5
6static inline int atomic_futex_op_xchg_set(int oparg, int __user *uaddr, 6static inline int atomic_futex_op_xchg_set(int oparg, u32 __user *uaddr,
7 int *oldval) 7 int *oldval)
8{ 8{
9 unsigned long flags; 9 unsigned long flags;
@@ -20,7 +20,7 @@ static inline int atomic_futex_op_xchg_set(int oparg, int __user *uaddr,
20 return ret; 20 return ret;
21} 21}
22 22
23static inline int atomic_futex_op_xchg_add(int oparg, int __user *uaddr, 23static inline int atomic_futex_op_xchg_add(int oparg, u32 __user *uaddr,
24 int *oldval) 24 int *oldval)
25{ 25{
26 unsigned long flags; 26 unsigned long flags;
@@ -37,7 +37,7 @@ static inline int atomic_futex_op_xchg_add(int oparg, int __user *uaddr,
37 return ret; 37 return ret;
38} 38}
39 39
40static inline int atomic_futex_op_xchg_or(int oparg, int __user *uaddr, 40static inline int atomic_futex_op_xchg_or(int oparg, u32 __user *uaddr,
41 int *oldval) 41 int *oldval)
42{ 42{
43 unsigned long flags; 43 unsigned long flags;
@@ -54,7 +54,7 @@ static inline int atomic_futex_op_xchg_or(int oparg, int __user *uaddr,
54 return ret; 54 return ret;
55} 55}
56 56
57static inline int atomic_futex_op_xchg_and(int oparg, int __user *uaddr, 57static inline int atomic_futex_op_xchg_and(int oparg, u32 __user *uaddr,
58 int *oldval) 58 int *oldval)
59{ 59{
60 unsigned long flags; 60 unsigned long flags;
@@ -71,7 +71,7 @@ static inline int atomic_futex_op_xchg_and(int oparg, int __user *uaddr,
71 return ret; 71 return ret;
72} 72}
73 73
74static inline int atomic_futex_op_xchg_xor(int oparg, int __user *uaddr, 74static inline int atomic_futex_op_xchg_xor(int oparg, u32 __user *uaddr,
75 int *oldval) 75 int *oldval)
76{ 76{
77 unsigned long flags; 77 unsigned long flags;
@@ -88,11 +88,13 @@ static inline int atomic_futex_op_xchg_xor(int oparg, int __user *uaddr,
88 return ret; 88 return ret;
89} 89}
90 90
91static inline int atomic_futex_op_cmpxchg_inatomic(int __user *uaddr, 91static inline int atomic_futex_op_cmpxchg_inatomic(u32 *uval,
92 int oldval, int newval) 92 u32 __user *uaddr,
93 u32 oldval, u32 newval)
93{ 94{
94 unsigned long flags; 95 unsigned long flags;
95 int ret, prev = 0; 96 int ret;
97 u32 prev = 0;
96 98
97 local_irq_save(flags); 99 local_irq_save(flags);
98 100
@@ -102,10 +104,8 @@ static inline int atomic_futex_op_cmpxchg_inatomic(int __user *uaddr,
102 104
103 local_irq_restore(flags); 105 local_irq_restore(flags);
104 106
105 if (ret) 107 *uval = prev;
106 return ret; 108 return ret;
107
108 return prev;
109} 109}
110 110
111#endif /* __ASM_SH_FUTEX_IRQ_H */ 111#endif /* __ASM_SH_FUTEX_IRQ_H */
diff --git a/arch/sh/include/asm/futex.h b/arch/sh/include/asm/futex.h
index 68256ec5fa35..7be39a646fbd 100644
--- a/arch/sh/include/asm/futex.h
+++ b/arch/sh/include/asm/futex.h
@@ -10,7 +10,7 @@
10/* XXX: UP variants, fix for SH-4A and SMP.. */ 10/* XXX: UP variants, fix for SH-4A and SMP.. */
11#include <asm/futex-irq.h> 11#include <asm/futex-irq.h>
12 12
13static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr) 13static inline int futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr)
14{ 14{
15 int op = (encoded_op >> 28) & 7; 15 int op = (encoded_op >> 28) & 7;
16 int cmp = (encoded_op >> 24) & 15; 16 int cmp = (encoded_op >> 24) & 15;
@@ -21,7 +21,7 @@ static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
21 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28)) 21 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
22 oparg = 1 << oparg; 22 oparg = 1 << oparg;
23 23
24 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int))) 24 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
25 return -EFAULT; 25 return -EFAULT;
26 26
27 pagefault_disable(); 27 pagefault_disable();
@@ -65,12 +65,13 @@ static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
65} 65}
66 66
67static inline int 67static inline int
68futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval) 68futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
69 u32 oldval, u32 newval)
69{ 70{
70 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int))) 71 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
71 return -EFAULT; 72 return -EFAULT;
72 73
73 return atomic_futex_op_cmpxchg_inatomic(uaddr, oldval, newval); 74 return atomic_futex_op_cmpxchg_inatomic(uval, uaddr, oldval, newval);
74} 75}
75 76
76#endif /* __KERNEL__ */ 77#endif /* __KERNEL__ */
diff --git a/arch/sh/include/asm/rwsem.h b/arch/sh/include/asm/rwsem.h
index 06e2251a5e48..edab57265293 100644
--- a/arch/sh/include/asm/rwsem.h
+++ b/arch/sh/include/asm/rwsem.h
@@ -11,64 +11,13 @@
11#endif 11#endif
12 12
13#ifdef __KERNEL__ 13#ifdef __KERNEL__
14#include <linux/list.h>
15#include <linux/spinlock.h>
16#include <asm/atomic.h>
17#include <asm/system.h>
18 14
19/*
20 * the semaphore definition
21 */
22struct rw_semaphore {
23 long count;
24#define RWSEM_UNLOCKED_VALUE 0x00000000 15#define RWSEM_UNLOCKED_VALUE 0x00000000
25#define RWSEM_ACTIVE_BIAS 0x00000001 16#define RWSEM_ACTIVE_BIAS 0x00000001
26#define RWSEM_ACTIVE_MASK 0x0000ffff 17#define RWSEM_ACTIVE_MASK 0x0000ffff
27#define RWSEM_WAITING_BIAS (-0x00010000) 18#define RWSEM_WAITING_BIAS (-0x00010000)
28#define RWSEM_ACTIVE_READ_BIAS RWSEM_ACTIVE_BIAS 19#define RWSEM_ACTIVE_READ_BIAS RWSEM_ACTIVE_BIAS
29#define RWSEM_ACTIVE_WRITE_BIAS (RWSEM_WAITING_BIAS + RWSEM_ACTIVE_BIAS) 20#define RWSEM_ACTIVE_WRITE_BIAS (RWSEM_WAITING_BIAS + RWSEM_ACTIVE_BIAS)
30 spinlock_t wait_lock;
31 struct list_head wait_list;
32#ifdef CONFIG_DEBUG_LOCK_ALLOC
33 struct lockdep_map dep_map;
34#endif
35};
36
37#ifdef CONFIG_DEBUG_LOCK_ALLOC
38# define __RWSEM_DEP_MAP_INIT(lockname) , .dep_map = { .name = #lockname }
39#else
40# define __RWSEM_DEP_MAP_INIT(lockname)
41#endif
42
43#define __RWSEM_INITIALIZER(name) \
44 { RWSEM_UNLOCKED_VALUE, __SPIN_LOCK_UNLOCKED((name).wait_lock), \
45 LIST_HEAD_INIT((name).wait_list) \
46 __RWSEM_DEP_MAP_INIT(name) }
47
48#define DECLARE_RWSEM(name) \
49 struct rw_semaphore name = __RWSEM_INITIALIZER(name)
50
51extern struct rw_semaphore *rwsem_down_read_failed(struct rw_semaphore *sem);
52extern struct rw_semaphore *rwsem_down_write_failed(struct rw_semaphore *sem);
53extern struct rw_semaphore *rwsem_wake(struct rw_semaphore *sem);
54extern struct rw_semaphore *rwsem_downgrade_wake(struct rw_semaphore *sem);
55
56extern void __init_rwsem(struct rw_semaphore *sem, const char *name,
57 struct lock_class_key *key);
58
59#define init_rwsem(sem) \
60do { \
61 static struct lock_class_key __key; \
62 \
63 __init_rwsem((sem), #sem, &__key); \
64} while (0)
65
66static inline void init_rwsem(struct rw_semaphore *sem)
67{
68 sem->count = RWSEM_UNLOCKED_VALUE;
69 spin_lock_init(&sem->wait_lock);
70 INIT_LIST_HEAD(&sem->wait_list);
71}
72 21
73/* 22/*
74 * lock for reading 23 * lock for reading
@@ -179,10 +128,5 @@ static inline int rwsem_atomic_update(int delta, struct rw_semaphore *sem)
179 return atomic_add_return(delta, (atomic_t *)(&sem->count)); 128 return atomic_add_return(delta, (atomic_t *)(&sem->count));
180} 129}
181 130
182static inline int rwsem_is_locked(struct rw_semaphore *sem)
183{
184 return (sem->count != 0);
185}
186
187#endif /* __KERNEL__ */ 131#endif /* __KERNEL__ */
188#endif /* _ASM_SH_RWSEM_H */ 132#endif /* _ASM_SH_RWSEM_H */
diff --git a/arch/sparc/include/asm/fcntl.h b/arch/sparc/include/asm/fcntl.h
index 38f37b333cc7..d0b83f66f356 100644
--- a/arch/sparc/include/asm/fcntl.h
+++ b/arch/sparc/include/asm/fcntl.h
@@ -34,6 +34,8 @@
34#define __O_SYNC 0x800000 34#define __O_SYNC 0x800000
35#define O_SYNC (__O_SYNC|O_DSYNC) 35#define O_SYNC (__O_SYNC|O_DSYNC)
36 36
37#define O_PATH 0x1000000
38
37#define F_GETOWN 5 /* for sockets. */ 39#define F_GETOWN 5 /* for sockets. */
38#define F_SETOWN 6 /* for sockets. */ 40#define F_SETOWN 6 /* for sockets. */
39#define F_GETLK 7 41#define F_GETLK 7
diff --git a/arch/sparc/include/asm/futex_64.h b/arch/sparc/include/asm/futex_64.h
index 47f95839dc69..444e7bea23bc 100644
--- a/arch/sparc/include/asm/futex_64.h
+++ b/arch/sparc/include/asm/futex_64.h
@@ -30,7 +30,7 @@
30 : "r" (uaddr), "r" (oparg), "i" (-EFAULT) \ 30 : "r" (uaddr), "r" (oparg), "i" (-EFAULT) \
31 : "memory") 31 : "memory")
32 32
33static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr) 33static inline int futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr)
34{ 34{
35 int op = (encoded_op >> 28) & 7; 35 int op = (encoded_op >> 28) & 7;
36 int cmp = (encoded_op >> 24) & 15; 36 int cmp = (encoded_op >> 24) & 15;
@@ -38,7 +38,7 @@ static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
38 int cmparg = (encoded_op << 20) >> 20; 38 int cmparg = (encoded_op << 20) >> 20;
39 int oldval = 0, ret, tem; 39 int oldval = 0, ret, tem;
40 40
41 if (unlikely(!access_ok(VERIFY_WRITE, uaddr, sizeof(int)))) 41 if (unlikely(!access_ok(VERIFY_WRITE, uaddr, sizeof(u32))))
42 return -EFAULT; 42 return -EFAULT;
43 if (unlikely((((unsigned long) uaddr) & 0x3UL))) 43 if (unlikely((((unsigned long) uaddr) & 0x3UL)))
44 return -EINVAL; 44 return -EINVAL;
@@ -85,26 +85,30 @@ static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
85} 85}
86 86
87static inline int 87static inline int
88futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval) 88futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
89 u32 oldval, u32 newval)
89{ 90{
91 int ret = 0;
92
90 __asm__ __volatile__( 93 __asm__ __volatile__(
91 "\n1: casa [%3] %%asi, %2, %0\n" 94 "\n1: casa [%4] %%asi, %3, %1\n"
92 "2:\n" 95 "2:\n"
93 " .section .fixup,#alloc,#execinstr\n" 96 " .section .fixup,#alloc,#execinstr\n"
94 " .align 4\n" 97 " .align 4\n"
95 "3: sethi %%hi(2b), %0\n" 98 "3: sethi %%hi(2b), %0\n"
96 " jmpl %0 + %%lo(2b), %%g0\n" 99 " jmpl %0 + %%lo(2b), %%g0\n"
97 " mov %4, %0\n" 100 " mov %5, %0\n"
98 " .previous\n" 101 " .previous\n"
99 " .section __ex_table,\"a\"\n" 102 " .section __ex_table,\"a\"\n"
100 " .align 4\n" 103 " .align 4\n"
101 " .word 1b, 3b\n" 104 " .word 1b, 3b\n"
102 " .previous\n" 105 " .previous\n"
103 : "=r" (newval) 106 : "+r" (ret), "=r" (newval)
104 : "0" (newval), "r" (oldval), "r" (uaddr), "i" (-EFAULT) 107 : "1" (newval), "r" (oldval), "r" (uaddr), "i" (-EFAULT)
105 : "memory"); 108 : "memory");
106 109
107 return newval; 110 *uval = newval;
111 return ret;
108} 112}
109 113
110#endif /* !(_SPARC64_FUTEX_H) */ 114#endif /* !(_SPARC64_FUTEX_H) */
diff --git a/arch/sparc/include/asm/rwsem.h b/arch/sparc/include/asm/rwsem.h
index a2b4302869bc..069bf4d663a1 100644
--- a/arch/sparc/include/asm/rwsem.h
+++ b/arch/sparc/include/asm/rwsem.h
@@ -13,53 +13,12 @@
13 13
14#ifdef __KERNEL__ 14#ifdef __KERNEL__
15 15
16#include <linux/list.h>
17#include <linux/spinlock.h>
18
19struct rwsem_waiter;
20
21struct rw_semaphore {
22 signed long count;
23#define RWSEM_UNLOCKED_VALUE 0x00000000L 16#define RWSEM_UNLOCKED_VALUE 0x00000000L
24#define RWSEM_ACTIVE_BIAS 0x00000001L 17#define RWSEM_ACTIVE_BIAS 0x00000001L
25#define RWSEM_ACTIVE_MASK 0xffffffffL 18#define RWSEM_ACTIVE_MASK 0xffffffffL
26#define RWSEM_WAITING_BIAS (-RWSEM_ACTIVE_MASK-1) 19#define RWSEM_WAITING_BIAS (-RWSEM_ACTIVE_MASK-1)
27#define RWSEM_ACTIVE_READ_BIAS RWSEM_ACTIVE_BIAS 20#define RWSEM_ACTIVE_READ_BIAS RWSEM_ACTIVE_BIAS
28#define RWSEM_ACTIVE_WRITE_BIAS (RWSEM_WAITING_BIAS + RWSEM_ACTIVE_BIAS) 21#define RWSEM_ACTIVE_WRITE_BIAS (RWSEM_WAITING_BIAS + RWSEM_ACTIVE_BIAS)
29 spinlock_t wait_lock;
30 struct list_head wait_list;
31#ifdef CONFIG_DEBUG_LOCK_ALLOC
32 struct lockdep_map dep_map;
33#endif
34};
35
36#ifdef CONFIG_DEBUG_LOCK_ALLOC
37# define __RWSEM_DEP_MAP_INIT(lockname) , .dep_map = { .name = #lockname }
38#else
39# define __RWSEM_DEP_MAP_INIT(lockname)
40#endif
41
42#define __RWSEM_INITIALIZER(name) \
43{ RWSEM_UNLOCKED_VALUE, __SPIN_LOCK_UNLOCKED((name).wait_lock), \
44 LIST_HEAD_INIT((name).wait_list) __RWSEM_DEP_MAP_INIT(name) }
45
46#define DECLARE_RWSEM(name) \
47 struct rw_semaphore name = __RWSEM_INITIALIZER(name)
48
49extern struct rw_semaphore *rwsem_down_read_failed(struct rw_semaphore *sem);
50extern struct rw_semaphore *rwsem_down_write_failed(struct rw_semaphore *sem);
51extern struct rw_semaphore *rwsem_wake(struct rw_semaphore *sem);
52extern struct rw_semaphore *rwsem_downgrade_wake(struct rw_semaphore *sem);
53
54extern void __init_rwsem(struct rw_semaphore *sem, const char *name,
55 struct lock_class_key *key);
56
57#define init_rwsem(sem) \
58do { \
59 static struct lock_class_key __key; \
60 \
61 __init_rwsem((sem), #sem, &__key); \
62} while (0)
63 22
64/* 23/*
65 * lock for reading 24 * lock for reading
@@ -160,11 +119,6 @@ static inline long rwsem_atomic_update(long delta, struct rw_semaphore *sem)
160 return atomic64_add_return(delta, (atomic64_t *)(&sem->count)); 119 return atomic64_add_return(delta, (atomic64_t *)(&sem->count));
161} 120}
162 121
163static inline int rwsem_is_locked(struct rw_semaphore *sem)
164{
165 return (sem->count != 0);
166}
167
168#endif /* __KERNEL__ */ 122#endif /* __KERNEL__ */
169 123
170#endif /* _SPARC64_RWSEM_H */ 124#endif /* _SPARC64_RWSEM_H */
diff --git a/arch/sparc/kernel/pcic.c b/arch/sparc/kernel/pcic.c
index aeaa09a3c655..2cdc131b50ac 100644
--- a/arch/sparc/kernel/pcic.c
+++ b/arch/sparc/kernel/pcic.c
@@ -700,10 +700,8 @@ static void pcic_clear_clock_irq(void)
700 700
701static irqreturn_t pcic_timer_handler (int irq, void *h) 701static irqreturn_t pcic_timer_handler (int irq, void *h)
702{ 702{
703 write_seqlock(&xtime_lock); /* Dummy, to show that we remember */
704 pcic_clear_clock_irq(); 703 pcic_clear_clock_irq();
705 do_timer(1); 704 xtime_update(1);
706 write_sequnlock(&xtime_lock);
707#ifndef CONFIG_SMP 705#ifndef CONFIG_SMP
708 update_process_times(user_mode(get_irq_regs())); 706 update_process_times(user_mode(get_irq_regs()));
709#endif 707#endif
diff --git a/arch/sparc/kernel/time_32.c b/arch/sparc/kernel/time_32.c
index 9c743b1886ff..4211bfc9bcad 100644
--- a/arch/sparc/kernel/time_32.c
+++ b/arch/sparc/kernel/time_32.c
@@ -85,7 +85,7 @@ int update_persistent_clock(struct timespec now)
85 85
86/* 86/*
87 * timer_interrupt() needs to keep up the real-time clock, 87 * timer_interrupt() needs to keep up the real-time clock,
88 * as well as call the "do_timer()" routine every clocktick 88 * as well as call the "xtime_update()" routine every clocktick
89 */ 89 */
90 90
91#define TICK_SIZE (tick_nsec / 1000) 91#define TICK_SIZE (tick_nsec / 1000)
@@ -96,14 +96,9 @@ static irqreturn_t timer_interrupt(int dummy, void *dev_id)
96 profile_tick(CPU_PROFILING); 96 profile_tick(CPU_PROFILING);
97#endif 97#endif
98 98
99 /* Protect counter clear so that do_gettimeoffset works */
100 write_seqlock(&xtime_lock);
101
102 clear_clock_irq(); 99 clear_clock_irq();
103 100
104 do_timer(1); 101 xtime_update(1);
105
106 write_sequnlock(&xtime_lock);
107 102
108#ifndef CONFIG_SMP 103#ifndef CONFIG_SMP
109 update_process_times(user_mode(get_irq_regs())); 104 update_process_times(user_mode(get_irq_regs()));
diff --git a/arch/sparc/lib/atomic32.c b/arch/sparc/lib/atomic32.c
index cbddeb38ffda..d3c7a12ad879 100644
--- a/arch/sparc/lib/atomic32.c
+++ b/arch/sparc/lib/atomic32.c
@@ -16,7 +16,7 @@
16#define ATOMIC_HASH(a) (&__atomic_hash[(((unsigned long)a)>>8) & (ATOMIC_HASH_SIZE-1)]) 16#define ATOMIC_HASH(a) (&__atomic_hash[(((unsigned long)a)>>8) & (ATOMIC_HASH_SIZE-1)])
17 17
18spinlock_t __atomic_hash[ATOMIC_HASH_SIZE] = { 18spinlock_t __atomic_hash[ATOMIC_HASH_SIZE] = {
19 [0 ... (ATOMIC_HASH_SIZE-1)] = SPIN_LOCK_UNLOCKED 19 [0 ... (ATOMIC_HASH_SIZE-1)] = __SPIN_LOCK_UNLOCKED(__atomic_hash)
20}; 20};
21 21
22#else /* SMP */ 22#else /* SMP */
diff --git a/arch/tile/include/asm/futex.h b/arch/tile/include/asm/futex.h
index fe0d10dcae57..d03ec124a598 100644
--- a/arch/tile/include/asm/futex.h
+++ b/arch/tile/include/asm/futex.h
@@ -29,16 +29,16 @@
29#include <linux/uaccess.h> 29#include <linux/uaccess.h>
30#include <linux/errno.h> 30#include <linux/errno.h>
31 31
32extern struct __get_user futex_set(int __user *v, int i); 32extern struct __get_user futex_set(u32 __user *v, int i);
33extern struct __get_user futex_add(int __user *v, int n); 33extern struct __get_user futex_add(u32 __user *v, int n);
34extern struct __get_user futex_or(int __user *v, int n); 34extern struct __get_user futex_or(u32 __user *v, int n);
35extern struct __get_user futex_andn(int __user *v, int n); 35extern struct __get_user futex_andn(u32 __user *v, int n);
36extern struct __get_user futex_cmpxchg(int __user *v, int o, int n); 36extern struct __get_user futex_cmpxchg(u32 __user *v, int o, int n);
37 37
38#ifndef __tilegx__ 38#ifndef __tilegx__
39extern struct __get_user futex_xor(int __user *v, int n); 39extern struct __get_user futex_xor(u32 __user *v, int n);
40#else 40#else
41static inline struct __get_user futex_xor(int __user *uaddr, int n) 41static inline struct __get_user futex_xor(u32 __user *uaddr, int n)
42{ 42{
43 struct __get_user asm_ret = __get_user_4(uaddr); 43 struct __get_user asm_ret = __get_user_4(uaddr);
44 if (!asm_ret.err) { 44 if (!asm_ret.err) {
@@ -53,7 +53,7 @@ static inline struct __get_user futex_xor(int __user *uaddr, int n)
53} 53}
54#endif 54#endif
55 55
56static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr) 56static inline int futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr)
57{ 57{
58 int op = (encoded_op >> 28) & 7; 58 int op = (encoded_op >> 28) & 7;
59 int cmp = (encoded_op >> 24) & 15; 59 int cmp = (encoded_op >> 24) & 15;
@@ -65,7 +65,7 @@ static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
65 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28)) 65 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
66 oparg = 1 << oparg; 66 oparg = 1 << oparg;
67 67
68 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int))) 68 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
69 return -EFAULT; 69 return -EFAULT;
70 70
71 pagefault_disable(); 71 pagefault_disable();
@@ -119,16 +119,17 @@ static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
119 return ret; 119 return ret;
120} 120}
121 121
122static inline int futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, 122static inline int futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
123 int newval) 123 u32 oldval, u32 newval)
124{ 124{
125 struct __get_user asm_ret; 125 struct __get_user asm_ret;
126 126
127 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int))) 127 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
128 return -EFAULT; 128 return -EFAULT;
129 129
130 asm_ret = futex_cmpxchg(uaddr, oldval, newval); 130 asm_ret = futex_cmpxchg(uaddr, oldval, newval);
131 return asm_ret.err ? asm_ret.err : asm_ret.val; 131 *uval = asm_ret.val;
132 return asm_ret.err;
132} 133}
133 134
134#ifndef __tilegx__ 135#ifndef __tilegx__
diff --git a/arch/um/Kconfig.common b/arch/um/Kconfig.common
index e351e14b4339..1e78940218c0 100644
--- a/arch/um/Kconfig.common
+++ b/arch/um/Kconfig.common
@@ -7,6 +7,7 @@ config UML
7 bool 7 bool
8 default y 8 default y
9 select HAVE_GENERIC_HARDIRQS 9 select HAVE_GENERIC_HARDIRQS
10 select GENERIC_HARDIRQS_NO_DEPRECATED
10 11
11config MMU 12config MMU
12 bool 13 bool
diff --git a/arch/um/Kconfig.x86 b/arch/um/Kconfig.x86
index 5ee328099c63..02fb017fed47 100644
--- a/arch/um/Kconfig.x86
+++ b/arch/um/Kconfig.x86
@@ -10,6 +10,8 @@ endmenu
10 10
11config UML_X86 11config UML_X86
12 def_bool y 12 def_bool y
13 select GENERIC_FIND_FIRST_BIT
14 select GENERIC_FIND_NEXT_BIT
13 15
14config 64BIT 16config 64BIT
15 bool 17 bool
@@ -19,6 +21,9 @@ config X86_32
19 def_bool !64BIT 21 def_bool !64BIT
20 select HAVE_AOUT 22 select HAVE_AOUT
21 23
24config X86_64
25 def_bool 64BIT
26
22config RWSEM_XCHGADD_ALGORITHM 27config RWSEM_XCHGADD_ALGORITHM
23 def_bool X86_XADD 28 def_bool X86_XADD
24 29
diff --git a/arch/um/drivers/mconsole_kern.c b/arch/um/drivers/mconsole_kern.c
index 975613b23dcf..c70e047eed72 100644
--- a/arch/um/drivers/mconsole_kern.c
+++ b/arch/um/drivers/mconsole_kern.c
@@ -124,35 +124,18 @@ void mconsole_log(struct mc_request *req)
124#if 0 124#if 0
125void mconsole_proc(struct mc_request *req) 125void mconsole_proc(struct mc_request *req)
126{ 126{
127 struct nameidata nd;
128 struct vfsmount *mnt = current->nsproxy->pid_ns->proc_mnt; 127 struct vfsmount *mnt = current->nsproxy->pid_ns->proc_mnt;
129 struct file *file; 128 struct file *file;
130 int n, err; 129 int n;
131 char *ptr = req->request.data, *buf; 130 char *ptr = req->request.data, *buf;
132 mm_segment_t old_fs = get_fs(); 131 mm_segment_t old_fs = get_fs();
133 132
134 ptr += strlen("proc"); 133 ptr += strlen("proc");
135 ptr = skip_spaces(ptr); 134 ptr = skip_spaces(ptr);
136 135
137 err = vfs_path_lookup(mnt->mnt_root, mnt, ptr, LOOKUP_FOLLOW, &nd); 136 file = file_open_root(mnt->mnt_root, mnt, ptr, O_RDONLY);
138 if (err) {
139 mconsole_reply(req, "Failed to look up file", 1, 0);
140 goto out;
141 }
142
143 err = may_open(&nd.path, MAY_READ, O_RDONLY);
144 if (result) {
145 mconsole_reply(req, "Failed to open file", 1, 0);
146 path_put(&nd.path);
147 goto out;
148 }
149
150 file = dentry_open(nd.path.dentry, nd.path.mnt, O_RDONLY,
151 current_cred());
152 err = PTR_ERR(file);
153 if (IS_ERR(file)) { 137 if (IS_ERR(file)) {
154 mconsole_reply(req, "Failed to open file", 1, 0); 138 mconsole_reply(req, "Failed to open file", 1, 0);
155 path_put(&nd.path);
156 goto out; 139 goto out;
157 } 140 }
158 141
diff --git a/arch/um/drivers/ubd_kern.c b/arch/um/drivers/ubd_kern.c
index ba4a98ba39c0..620f5b70957d 100644
--- a/arch/um/drivers/ubd_kern.c
+++ b/arch/um/drivers/ubd_kern.c
@@ -185,7 +185,7 @@ struct ubd {
185 .no_cow = 0, \ 185 .no_cow = 0, \
186 .shared = 0, \ 186 .shared = 0, \
187 .cow = DEFAULT_COW, \ 187 .cow = DEFAULT_COW, \
188 .lock = SPIN_LOCK_UNLOCKED, \ 188 .lock = __SPIN_LOCK_UNLOCKED(ubd_devs.lock), \
189 .request = NULL, \ 189 .request = NULL, \
190 .start_sg = 0, \ 190 .start_sg = 0, \
191 .end_sg = 0, \ 191 .end_sg = 0, \
diff --git a/arch/um/kernel/irq.c b/arch/um/kernel/irq.c
index 3f0ac9e0c966..64cfea80cfe2 100644
--- a/arch/um/kernel/irq.c
+++ b/arch/um/kernel/irq.c
@@ -35,8 +35,10 @@ int show_interrupts(struct seq_file *p, void *v)
35 } 35 }
36 36
37 if (i < NR_IRQS) { 37 if (i < NR_IRQS) {
38 raw_spin_lock_irqsave(&irq_desc[i].lock, flags); 38 struct irq_desc *desc = irq_to_desc(i);
39 action = irq_desc[i].action; 39
40 raw_spin_lock_irqsave(&desc->lock, flags);
41 action = desc->action;
40 if (!action) 42 if (!action)
41 goto skip; 43 goto skip;
42 seq_printf(p, "%3d: ",i); 44 seq_printf(p, "%3d: ",i);
@@ -46,7 +48,7 @@ int show_interrupts(struct seq_file *p, void *v)
46 for_each_online_cpu(j) 48 for_each_online_cpu(j)
47 seq_printf(p, "%10u ", kstat_irqs_cpu(i, j)); 49 seq_printf(p, "%10u ", kstat_irqs_cpu(i, j));
48#endif 50#endif
49 seq_printf(p, " %14s", irq_desc[i].chip->name); 51 seq_printf(p, " %14s", get_irq_desc_chip(desc)->name);
50 seq_printf(p, " %s", action->name); 52 seq_printf(p, " %s", action->name);
51 53
52 for (action=action->next; action; action = action->next) 54 for (action=action->next; action; action = action->next)
@@ -54,7 +56,7 @@ int show_interrupts(struct seq_file *p, void *v)
54 56
55 seq_putc(p, '\n'); 57 seq_putc(p, '\n');
56skip: 58skip:
57 raw_spin_unlock_irqrestore(&irq_desc[i].lock, flags); 59 raw_spin_unlock_irqrestore(&desc->lock, flags);
58 } else if (i == NR_IRQS) 60 } else if (i == NR_IRQS)
59 seq_putc(p, '\n'); 61 seq_putc(p, '\n');
60 62
@@ -360,10 +362,10 @@ EXPORT_SYMBOL(um_request_irq);
360EXPORT_SYMBOL(reactivate_fd); 362EXPORT_SYMBOL(reactivate_fd);
361 363
362/* 364/*
363 * irq_chip must define (startup || enable) && 365 * irq_chip must define at least enable/disable and ack when
364 * (shutdown || disable) && end 366 * the edge handler is used.
365 */ 367 */
366static void dummy(unsigned int irq) 368static void dummy(struct irq_data *d)
367{ 369{
368} 370}
369 371
@@ -371,20 +373,17 @@ static void dummy(unsigned int irq)
371static struct irq_chip normal_irq_type = { 373static struct irq_chip normal_irq_type = {
372 .name = "SIGIO", 374 .name = "SIGIO",
373 .release = free_irq_by_irq_and_dev, 375 .release = free_irq_by_irq_and_dev,
374 .disable = dummy, 376 .irq_disable = dummy,
375 .enable = dummy, 377 .irq_enable = dummy,
376 .ack = dummy, 378 .irq_ack = dummy,
377 .end = dummy
378}; 379};
379 380
380static struct irq_chip SIGVTALRM_irq_type = { 381static struct irq_chip SIGVTALRM_irq_type = {
381 .name = "SIGVTALRM", 382 .name = "SIGVTALRM",
382 .release = free_irq_by_irq_and_dev, 383 .release = free_irq_by_irq_and_dev,
383 .shutdown = dummy, /* never called */ 384 .irq_disable = dummy,
384 .disable = dummy, 385 .irq_enable = dummy,
385 .enable = dummy, 386 .irq_ack = dummy,
386 .ack = dummy,
387 .end = dummy
388}; 387};
389 388
390void __init init_IRQ(void) 389void __init init_IRQ(void)
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index d5ed94d30aad..f8958b01b975 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -64,8 +64,12 @@ config X86
64 select HAVE_TEXT_POKE_SMP 64 select HAVE_TEXT_POKE_SMP
65 select HAVE_GENERIC_HARDIRQS 65 select HAVE_GENERIC_HARDIRQS
66 select HAVE_SPARSE_IRQ 66 select HAVE_SPARSE_IRQ
67 select GENERIC_FIND_FIRST_BIT
68 select GENERIC_FIND_NEXT_BIT
67 select GENERIC_IRQ_PROBE 69 select GENERIC_IRQ_PROBE
68 select GENERIC_PENDING_IRQ if SMP 70 select GENERIC_PENDING_IRQ if SMP
71 select GENERIC_IRQ_SHOW
72 select IRQ_FORCED_THREADING
69 select USE_GENERIC_SMP_HELPERS if SMP 73 select USE_GENERIC_SMP_HELPERS if SMP
70 74
71config INSTRUCTION_DECODER 75config INSTRUCTION_DECODER
@@ -382,6 +386,8 @@ config X86_INTEL_CE
382 depends on X86_32 386 depends on X86_32
383 depends on X86_EXTENDED_PLATFORM 387 depends on X86_EXTENDED_PLATFORM
384 select X86_REBOOTFIXUPS 388 select X86_REBOOTFIXUPS
389 select OF
390 select OF_EARLY_FLATTREE
385 ---help--- 391 ---help---
386 Select for the Intel CE media processor (CE4100) SOC. 392 Select for the Intel CE media processor (CE4100) SOC.
387 This option compiles in support for the CE4100 SOC for settop 393 This option compiles in support for the CE4100 SOC for settop
@@ -811,7 +817,7 @@ config X86_LOCAL_APIC
811 817
812config X86_IO_APIC 818config X86_IO_APIC
813 def_bool y 819 def_bool y
814 depends on X86_64 || SMP || X86_32_NON_STANDARD || X86_UP_APIC 820 depends on X86_64 || SMP || X86_32_NON_STANDARD || X86_UP_IOAPIC
815 821
816config X86_VISWS_APIC 822config X86_VISWS_APIC
817 def_bool y 823 def_bool y
@@ -1705,7 +1711,7 @@ config HAVE_ARCH_EARLY_PFN_TO_NID
1705 depends on NUMA 1711 depends on NUMA
1706 1712
1707config USE_PERCPU_NUMA_NODE_ID 1713config USE_PERCPU_NUMA_NODE_ID
1708 def_bool X86_64 1714 def_bool y
1709 depends on NUMA 1715 depends on NUMA
1710 1716
1711menu "Power management and ACPI options" 1717menu "Power management and ACPI options"
@@ -2066,9 +2072,10 @@ config SCx200HR_TIMER
2066 2072
2067config OLPC 2073config OLPC
2068 bool "One Laptop Per Child support" 2074 bool "One Laptop Per Child support"
2075 depends on !X86_PAE
2069 select GPIOLIB 2076 select GPIOLIB
2070 select OLPC_OPENFIRMWARE 2077 select OF
2071 depends on !X86_64 && !X86_PAE 2078 select OF_PROMTREE if PROC_DEVICETREE
2072 ---help--- 2079 ---help---
2073 Add support for detecting the unique features of the OLPC 2080 Add support for detecting the unique features of the OLPC
2074 XO hardware. 2081 XO hardware.
@@ -2079,21 +2086,6 @@ config OLPC_XO1
2079 ---help--- 2086 ---help---
2080 Add support for non-essential features of the OLPC XO-1 laptop. 2087 Add support for non-essential features of the OLPC XO-1 laptop.
2081 2088
2082config OLPC_OPENFIRMWARE
2083 bool "Support for OLPC's Open Firmware"
2084 depends on !X86_64 && !X86_PAE
2085 default n
2086 select OF
2087 help
2088 This option adds support for the implementation of Open Firmware
2089 that is used on the OLPC XO-1 Children's Machine.
2090 If unsure, say N here.
2091
2092config OLPC_OPENFIRMWARE_DT
2093 bool
2094 default y if OLPC_OPENFIRMWARE && PROC_DEVICETREE
2095 select OF_PROMTREE
2096
2097endif # X86_32 2089endif # X86_32
2098 2090
2099config AMD_NB 2091config AMD_NB
diff --git a/arch/x86/Kconfig.cpu b/arch/x86/Kconfig.cpu
index e6d3013f7ec3..75d89ac58d28 100644
--- a/arch/x86/Kconfig.cpu
+++ b/arch/x86/Kconfig.cpu
@@ -294,11 +294,6 @@ config X86_GENERIC
294 294
295endif 295endif
296 296
297config X86_CPU
298 def_bool y
299 select GENERIC_FIND_FIRST_BIT
300 select GENERIC_FIND_NEXT_BIT
301
302# 297#
303# Define implied options from the CPU selection here 298# Define implied options from the CPU selection here
304config X86_INTERNODE_CACHE_SHIFT 299config X86_INTERNODE_CACHE_SHIFT
diff --git a/arch/x86/ia32/ia32entry.S b/arch/x86/ia32/ia32entry.S
index 518bb99c3394..430312ba6e3f 100644
--- a/arch/x86/ia32/ia32entry.S
+++ b/arch/x86/ia32/ia32entry.S
@@ -25,6 +25,8 @@
25#define sysretl_audit ia32_ret_from_sys_call 25#define sysretl_audit ia32_ret_from_sys_call
26#endif 26#endif
27 27
28 .section .entry.text, "ax"
29
28#define IA32_NR_syscalls ((ia32_syscall_end - ia32_sys_call_table)/8) 30#define IA32_NR_syscalls ((ia32_syscall_end - ia32_sys_call_table)/8)
29 31
30 .macro IA32_ARG_FIXUP noebp=0 32 .macro IA32_ARG_FIXUP noebp=0
@@ -126,26 +128,20 @@ ENTRY(ia32_sysenter_target)
126 */ 128 */
127 ENABLE_INTERRUPTS(CLBR_NONE) 129 ENABLE_INTERRUPTS(CLBR_NONE)
128 movl %ebp,%ebp /* zero extension */ 130 movl %ebp,%ebp /* zero extension */
129 pushq $__USER32_DS 131 pushq_cfi $__USER32_DS
130 CFI_ADJUST_CFA_OFFSET 8
131 /*CFI_REL_OFFSET ss,0*/ 132 /*CFI_REL_OFFSET ss,0*/
132 pushq %rbp 133 pushq_cfi %rbp
133 CFI_ADJUST_CFA_OFFSET 8
134 CFI_REL_OFFSET rsp,0 134 CFI_REL_OFFSET rsp,0
135 pushfq 135 pushfq_cfi
136 CFI_ADJUST_CFA_OFFSET 8
137 /*CFI_REL_OFFSET rflags,0*/ 136 /*CFI_REL_OFFSET rflags,0*/
138 movl 8*3-THREAD_SIZE+TI_sysenter_return(%rsp), %r10d 137 movl 8*3-THREAD_SIZE+TI_sysenter_return(%rsp), %r10d
139 CFI_REGISTER rip,r10 138 CFI_REGISTER rip,r10
140 pushq $__USER32_CS 139 pushq_cfi $__USER32_CS
141 CFI_ADJUST_CFA_OFFSET 8
142 /*CFI_REL_OFFSET cs,0*/ 140 /*CFI_REL_OFFSET cs,0*/
143 movl %eax, %eax 141 movl %eax, %eax
144 pushq %r10 142 pushq_cfi %r10
145 CFI_ADJUST_CFA_OFFSET 8
146 CFI_REL_OFFSET rip,0 143 CFI_REL_OFFSET rip,0
147 pushq %rax 144 pushq_cfi %rax
148 CFI_ADJUST_CFA_OFFSET 8
149 cld 145 cld
150 SAVE_ARGS 0,0,1 146 SAVE_ARGS 0,0,1
151 /* no need to do an access_ok check here because rbp has been 147 /* no need to do an access_ok check here because rbp has been
@@ -182,11 +178,9 @@ sysexit_from_sys_call:
182 xorq %r9,%r9 178 xorq %r9,%r9
183 xorq %r10,%r10 179 xorq %r10,%r10
184 xorq %r11,%r11 180 xorq %r11,%r11
185 popfq 181 popfq_cfi
186 CFI_ADJUST_CFA_OFFSET -8
187 /*CFI_RESTORE rflags*/ 182 /*CFI_RESTORE rflags*/
188 popq %rcx /* User %esp */ 183 popq_cfi %rcx /* User %esp */
189 CFI_ADJUST_CFA_OFFSET -8
190 CFI_REGISTER rsp,rcx 184 CFI_REGISTER rsp,rcx
191 TRACE_IRQS_ON 185 TRACE_IRQS_ON
192 ENABLE_INTERRUPTS_SYSEXIT32 186 ENABLE_INTERRUPTS_SYSEXIT32
@@ -421,8 +415,7 @@ ENTRY(ia32_syscall)
421 */ 415 */
422 ENABLE_INTERRUPTS(CLBR_NONE) 416 ENABLE_INTERRUPTS(CLBR_NONE)
423 movl %eax,%eax 417 movl %eax,%eax
424 pushq %rax 418 pushq_cfi %rax
425 CFI_ADJUST_CFA_OFFSET 8
426 cld 419 cld
427 /* note the registers are not zero extended to the sf. 420 /* note the registers are not zero extended to the sf.
428 this could be a problem. */ 421 this could be a problem. */
@@ -851,4 +844,7 @@ ia32_sys_call_table:
851 .quad sys_fanotify_init 844 .quad sys_fanotify_init
852 .quad sys32_fanotify_mark 845 .quad sys32_fanotify_mark
853 .quad sys_prlimit64 /* 340 */ 846 .quad sys_prlimit64 /* 340 */
847 .quad sys_name_to_handle_at
848 .quad compat_sys_open_by_handle_at
849 .quad compat_sys_clock_adjtime
854ia32_syscall_end: 850ia32_syscall_end:
diff --git a/arch/x86/include/asm/acpi.h b/arch/x86/include/asm/acpi.h
index 4ea15ca89b2b..b964ec457546 100644
--- a/arch/x86/include/asm/acpi.h
+++ b/arch/x86/include/asm/acpi.h
@@ -186,15 +186,7 @@ struct bootnode;
186 186
187#ifdef CONFIG_ACPI_NUMA 187#ifdef CONFIG_ACPI_NUMA
188extern int acpi_numa; 188extern int acpi_numa;
189extern void acpi_get_nodes(struct bootnode *physnodes, unsigned long start, 189extern int x86_acpi_numa_init(void);
190 unsigned long end);
191extern int acpi_scan_nodes(unsigned long start, unsigned long end);
192#define NR_NODE_MEMBLKS (MAX_NUMNODES*2)
193
194#ifdef CONFIG_NUMA_EMU
195extern void acpi_fake_nodes(const struct bootnode *fake_nodes,
196 int num_nodes);
197#endif
198#endif /* CONFIG_ACPI_NUMA */ 190#endif /* CONFIG_ACPI_NUMA */
199 191
200#define acpi_unlazy_tlb(x) leave_mm(x) 192#define acpi_unlazy_tlb(x) leave_mm(x)
diff --git a/arch/x86/include/asm/amd_nb.h b/arch/x86/include/asm/amd_nb.h
index 527fb966ab5c..331682231bb4 100644
--- a/arch/x86/include/asm/amd_nb.h
+++ b/arch/x86/include/asm/amd_nb.h
@@ -16,16 +16,10 @@ struct bootnode;
16extern bool early_is_amd_nb(u32 value); 16extern bool early_is_amd_nb(u32 value);
17extern int amd_cache_northbridges(void); 17extern int amd_cache_northbridges(void);
18extern void amd_flush_garts(void); 18extern void amd_flush_garts(void);
19extern int amd_numa_init(unsigned long start_pfn, unsigned long end_pfn); 19extern int amd_numa_init(void);
20extern int amd_scan_nodes(void);
21extern int amd_get_subcaches(int); 20extern int amd_get_subcaches(int);
22extern int amd_set_subcaches(int, int); 21extern int amd_set_subcaches(int, int);
23 22
24#ifdef CONFIG_NUMA_EMU
25extern void amd_fake_nodes(const struct bootnode *nodes, int nr_nodes);
26extern void amd_get_nodes(struct bootnode *nodes);
27#endif
28
29struct amd_northbridge { 23struct amd_northbridge {
30 struct pci_dev *misc; 24 struct pci_dev *misc;
31 struct pci_dev *link; 25 struct pci_dev *link;
diff --git a/arch/x86/include/asm/apic.h b/arch/x86/include/asm/apic.h
index 3c896946f4cc..a279d98ea95e 100644
--- a/arch/x86/include/asm/apic.h
+++ b/arch/x86/include/asm/apic.h
@@ -220,7 +220,6 @@ extern void enable_IR_x2apic(void);
220 220
221extern int get_physical_broadcast(void); 221extern int get_physical_broadcast(void);
222 222
223extern void apic_disable(void);
224extern int lapic_get_maxlvt(void); 223extern int lapic_get_maxlvt(void);
225extern void clear_local_APIC(void); 224extern void clear_local_APIC(void);
226extern void connect_bsp_APIC(void); 225extern void connect_bsp_APIC(void);
@@ -228,7 +227,6 @@ extern void disconnect_bsp_APIC(int virt_wire_setup);
228extern void disable_local_APIC(void); 227extern void disable_local_APIC(void);
229extern void lapic_shutdown(void); 228extern void lapic_shutdown(void);
230extern int verify_local_APIC(void); 229extern int verify_local_APIC(void);
231extern void cache_APIC_registers(void);
232extern void sync_Arb_IDs(void); 230extern void sync_Arb_IDs(void);
233extern void init_bsp_APIC(void); 231extern void init_bsp_APIC(void);
234extern void setup_local_APIC(void); 232extern void setup_local_APIC(void);
@@ -239,8 +237,7 @@ void register_lapic_address(unsigned long address);
239extern void setup_boot_APIC_clock(void); 237extern void setup_boot_APIC_clock(void);
240extern void setup_secondary_APIC_clock(void); 238extern void setup_secondary_APIC_clock(void);
241extern int APIC_init_uniprocessor(void); 239extern int APIC_init_uniprocessor(void);
242extern void enable_NMI_through_LVT0(void); 240extern int apic_force_enable(unsigned long addr);
243extern int apic_force_enable(void);
244 241
245/* 242/*
246 * On 32bit this is mach-xxx local 243 * On 32bit this is mach-xxx local
@@ -261,7 +258,6 @@ static inline void lapic_shutdown(void) { }
261#define local_apic_timer_c2_ok 1 258#define local_apic_timer_c2_ok 1
262static inline void init_apic_mappings(void) { } 259static inline void init_apic_mappings(void) { }
263static inline void disable_local_APIC(void) { } 260static inline void disable_local_APIC(void) { }
264static inline void apic_disable(void) { }
265# define setup_boot_APIC_clock x86_init_noop 261# define setup_boot_APIC_clock x86_init_noop
266# define setup_secondary_APIC_clock x86_init_noop 262# define setup_secondary_APIC_clock x86_init_noop
267#endif /* !CONFIG_X86_LOCAL_APIC */ 263#endif /* !CONFIG_X86_LOCAL_APIC */
@@ -307,8 +303,6 @@ struct apic {
307 303
308 void (*setup_apic_routing)(void); 304 void (*setup_apic_routing)(void);
309 int (*multi_timer_check)(int apic, int irq); 305 int (*multi_timer_check)(int apic, int irq);
310 int (*apicid_to_node)(int logical_apicid);
311 int (*cpu_to_logical_apicid)(int cpu);
312 int (*cpu_present_to_apicid)(int mps_cpu); 306 int (*cpu_present_to_apicid)(int mps_cpu);
313 void (*apicid_to_cpu_present)(int phys_apicid, physid_mask_t *retmap); 307 void (*apicid_to_cpu_present)(int phys_apicid, physid_mask_t *retmap);
314 void (*setup_portio_remap)(void); 308 void (*setup_portio_remap)(void);
@@ -356,6 +350,23 @@ struct apic {
356 void (*icr_write)(u32 low, u32 high); 350 void (*icr_write)(u32 low, u32 high);
357 void (*wait_icr_idle)(void); 351 void (*wait_icr_idle)(void);
358 u32 (*safe_wait_icr_idle)(void); 352 u32 (*safe_wait_icr_idle)(void);
353
354#ifdef CONFIG_X86_32
355 /*
356 * Called very early during boot from get_smp_config(). It should
357 * return the logical apicid. x86_[bios]_cpu_to_apicid is
358 * initialized before this function is called.
359 *
360 * If logical apicid can't be determined that early, the function
361 * may return BAD_APICID. Logical apicid will be configured after
362 * init_apic_ldr() while bringing up CPUs. Note that NUMA affinity
363 * won't be applied properly during early boot in this case.
364 */
365 int (*x86_32_early_logical_apicid)(int cpu);
366
367 /* determine CPU -> NUMA node mapping */
368 int (*x86_32_numa_cpu_node)(int cpu);
369#endif
359}; 370};
360 371
361/* 372/*
@@ -503,6 +514,11 @@ extern struct apic apic_noop;
503 514
504extern struct apic apic_default; 515extern struct apic apic_default;
505 516
517static inline int noop_x86_32_early_logical_apicid(int cpu)
518{
519 return BAD_APICID;
520}
521
506/* 522/*
507 * Set up the logical destination ID. 523 * Set up the logical destination ID.
508 * 524 *
@@ -522,7 +538,7 @@ static inline int default_phys_pkg_id(int cpuid_apic, int index_msb)
522 return cpuid_apic >> index_msb; 538 return cpuid_apic >> index_msb;
523} 539}
524 540
525extern int default_apicid_to_node(int logical_apicid); 541extern int default_x86_32_numa_cpu_node(int cpu);
526 542
527#endif 543#endif
528 544
@@ -558,12 +574,6 @@ static inline void default_ioapic_phys_id_map(physid_mask_t *phys_map, physid_ma
558 *retmap = *phys_map; 574 *retmap = *phys_map;
559} 575}
560 576
561/* Mapping from cpu number to logical apicid */
562static inline int default_cpu_to_logical_apicid(int cpu)
563{
564 return 1 << cpu;
565}
566
567static inline int __default_cpu_present_to_apicid(int mps_cpu) 577static inline int __default_cpu_present_to_apicid(int mps_cpu)
568{ 578{
569 if (mps_cpu < nr_cpu_ids && cpu_present(mps_cpu)) 579 if (mps_cpu < nr_cpu_ids && cpu_present(mps_cpu))
@@ -596,8 +606,4 @@ extern int default_check_phys_apicid_present(int phys_apicid);
596 606
597#endif /* CONFIG_X86_LOCAL_APIC */ 607#endif /* CONFIG_X86_LOCAL_APIC */
598 608
599#ifdef CONFIG_X86_32
600extern u8 cpu_2_logical_apicid[NR_CPUS];
601#endif
602
603#endif /* _ASM_X86_APIC_H */ 609#endif /* _ASM_X86_APIC_H */
diff --git a/arch/x86/include/asm/apicdef.h b/arch/x86/include/asm/apicdef.h
index 47a30ff8e517..d87988bacf3e 100644
--- a/arch/x86/include/asm/apicdef.h
+++ b/arch/x86/include/asm/apicdef.h
@@ -426,4 +426,16 @@ struct local_apic {
426#else 426#else
427 #define BAD_APICID 0xFFFFu 427 #define BAD_APICID 0xFFFFu
428#endif 428#endif
429
430enum ioapic_irq_destination_types {
431 dest_Fixed = 0,
432 dest_LowestPrio = 1,
433 dest_SMI = 2,
434 dest__reserved_1 = 3,
435 dest_NMI = 4,
436 dest_INIT = 5,
437 dest__reserved_2 = 6,
438 dest_ExtINT = 7
439};
440
429#endif /* _ASM_X86_APICDEF_H */ 441#endif /* _ASM_X86_APICDEF_H */
diff --git a/arch/x86/include/asm/bootparam.h b/arch/x86/include/asm/bootparam.h
index c8bfe63a06de..e020d88ec02d 100644
--- a/arch/x86/include/asm/bootparam.h
+++ b/arch/x86/include/asm/bootparam.h
@@ -12,6 +12,7 @@
12/* setup data types */ 12/* setup data types */
13#define SETUP_NONE 0 13#define SETUP_NONE 0
14#define SETUP_E820_EXT 1 14#define SETUP_E820_EXT 1
15#define SETUP_DTB 2
15 16
16/* extensible setup data list node */ 17/* extensible setup data list node */
17struct setup_data { 18struct setup_data {
diff --git a/arch/x86/include/asm/cpufeature.h b/arch/x86/include/asm/cpufeature.h
index 220e2ea08e80..91f3e087cf21 100644
--- a/arch/x86/include/asm/cpufeature.h
+++ b/arch/x86/include/asm/cpufeature.h
@@ -160,6 +160,7 @@
160#define X86_FEATURE_NODEID_MSR (6*32+19) /* NodeId MSR */ 160#define X86_FEATURE_NODEID_MSR (6*32+19) /* NodeId MSR */
161#define X86_FEATURE_TBM (6*32+21) /* trailing bit manipulations */ 161#define X86_FEATURE_TBM (6*32+21) /* trailing bit manipulations */
162#define X86_FEATURE_TOPOEXT (6*32+22) /* topology extensions CPUID leafs */ 162#define X86_FEATURE_TOPOEXT (6*32+22) /* topology extensions CPUID leafs */
163#define X86_FEATURE_PERFCTR_CORE (6*32+23) /* core performance counter extensions */
163 164
164/* 165/*
165 * Auxiliary flags: Linux defined - For features scattered in various 166 * Auxiliary flags: Linux defined - For features scattered in various
@@ -279,6 +280,7 @@ extern const char * const x86_power_flags[32];
279#define cpu_has_xsave boot_cpu_has(X86_FEATURE_XSAVE) 280#define cpu_has_xsave boot_cpu_has(X86_FEATURE_XSAVE)
280#define cpu_has_hypervisor boot_cpu_has(X86_FEATURE_HYPERVISOR) 281#define cpu_has_hypervisor boot_cpu_has(X86_FEATURE_HYPERVISOR)
281#define cpu_has_pclmulqdq boot_cpu_has(X86_FEATURE_PCLMULQDQ) 282#define cpu_has_pclmulqdq boot_cpu_has(X86_FEATURE_PCLMULQDQ)
283#define cpu_has_perfctr_core boot_cpu_has(X86_FEATURE_PERFCTR_CORE)
282 284
283#if defined(CONFIG_X86_INVLPG) || defined(CONFIG_X86_64) 285#if defined(CONFIG_X86_INVLPG) || defined(CONFIG_X86_64)
284# define cpu_has_invlpg 1 286# define cpu_has_invlpg 1
diff --git a/arch/x86/include/asm/e820.h b/arch/x86/include/asm/e820.h
index e99d55d74df5..908b96957d88 100644
--- a/arch/x86/include/asm/e820.h
+++ b/arch/x86/include/asm/e820.h
@@ -96,7 +96,7 @@ extern void e820_setup_gap(void);
96extern int e820_search_gap(unsigned long *gapstart, unsigned long *gapsize, 96extern int e820_search_gap(unsigned long *gapstart, unsigned long *gapsize,
97 unsigned long start_addr, unsigned long long end_addr); 97 unsigned long start_addr, unsigned long long end_addr);
98struct setup_data; 98struct setup_data;
99extern void parse_e820_ext(struct setup_data *data, unsigned long pa_data); 99extern void parse_e820_ext(struct setup_data *data);
100 100
101#if defined(CONFIG_X86_64) || \ 101#if defined(CONFIG_X86_64) || \
102 (defined(CONFIG_X86_32) && defined(CONFIG_HIBERNATION)) 102 (defined(CONFIG_X86_32) && defined(CONFIG_HIBERNATION))
diff --git a/arch/x86/include/asm/entry_arch.h b/arch/x86/include/asm/entry_arch.h
index 57650ab4a5f5..1cd6d26a0a8d 100644
--- a/arch/x86/include/asm/entry_arch.h
+++ b/arch/x86/include/asm/entry_arch.h
@@ -16,10 +16,13 @@ BUILD_INTERRUPT(call_function_single_interrupt,CALL_FUNCTION_SINGLE_VECTOR)
16BUILD_INTERRUPT(irq_move_cleanup_interrupt,IRQ_MOVE_CLEANUP_VECTOR) 16BUILD_INTERRUPT(irq_move_cleanup_interrupt,IRQ_MOVE_CLEANUP_VECTOR)
17BUILD_INTERRUPT(reboot_interrupt,REBOOT_VECTOR) 17BUILD_INTERRUPT(reboot_interrupt,REBOOT_VECTOR)
18 18
19.irpc idx, "01234567" 19.irp idx,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15, \
20 16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31
21.if NUM_INVALIDATE_TLB_VECTORS > \idx
20BUILD_INTERRUPT3(invalidate_interrupt\idx, 22BUILD_INTERRUPT3(invalidate_interrupt\idx,
21 (INVALIDATE_TLB_VECTOR_START)+\idx, 23 (INVALIDATE_TLB_VECTOR_START)+\idx,
22 smp_invalidate_interrupt) 24 smp_invalidate_interrupt)
25.endif
23.endr 26.endr
24#endif 27#endif
25 28
diff --git a/arch/x86/include/asm/frame.h b/arch/x86/include/asm/frame.h
index 06850a7194e1..2c6fc9e62812 100644
--- a/arch/x86/include/asm/frame.h
+++ b/arch/x86/include/asm/frame.h
@@ -7,14 +7,12 @@
7 frame pointer later */ 7 frame pointer later */
8#ifdef CONFIG_FRAME_POINTER 8#ifdef CONFIG_FRAME_POINTER
9 .macro FRAME 9 .macro FRAME
10 pushl %ebp 10 pushl_cfi %ebp
11 CFI_ADJUST_CFA_OFFSET 4
12 CFI_REL_OFFSET ebp,0 11 CFI_REL_OFFSET ebp,0
13 movl %esp,%ebp 12 movl %esp,%ebp
14 .endm 13 .endm
15 .macro ENDFRAME 14 .macro ENDFRAME
16 popl %ebp 15 popl_cfi %ebp
17 CFI_ADJUST_CFA_OFFSET -4
18 CFI_RESTORE ebp 16 CFI_RESTORE ebp
19 .endm 17 .endm
20#else 18#else
diff --git a/arch/x86/include/asm/futex.h b/arch/x86/include/asm/futex.h
index 1f11ce44e956..d09bb03653f0 100644
--- a/arch/x86/include/asm/futex.h
+++ b/arch/x86/include/asm/futex.h
@@ -37,7 +37,7 @@
37 "+m" (*uaddr), "=&r" (tem) \ 37 "+m" (*uaddr), "=&r" (tem) \
38 : "r" (oparg), "i" (-EFAULT), "1" (0)) 38 : "r" (oparg), "i" (-EFAULT), "1" (0))
39 39
40static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr) 40static inline int futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr)
41{ 41{
42 int op = (encoded_op >> 28) & 7; 42 int op = (encoded_op >> 28) & 7;
43 int cmp = (encoded_op >> 24) & 15; 43 int cmp = (encoded_op >> 24) & 15;
@@ -48,7 +48,7 @@ static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
48 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28)) 48 if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
49 oparg = 1 << oparg; 49 oparg = 1 << oparg;
50 50
51 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int))) 51 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
52 return -EFAULT; 52 return -EFAULT;
53 53
54#if defined(CONFIG_X86_32) && !defined(CONFIG_X86_BSWAP) 54#if defined(CONFIG_X86_32) && !defined(CONFIG_X86_BSWAP)
@@ -109,9 +109,10 @@ static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
109 return ret; 109 return ret;
110} 110}
111 111
112static inline int futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, 112static inline int futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
113 int newval) 113 u32 oldval, u32 newval)
114{ 114{
115 int ret = 0;
115 116
116#if defined(CONFIG_X86_32) && !defined(CONFIG_X86_BSWAP) 117#if defined(CONFIG_X86_32) && !defined(CONFIG_X86_BSWAP)
117 /* Real i386 machines have no cmpxchg instruction */ 118 /* Real i386 machines have no cmpxchg instruction */
@@ -119,21 +120,22 @@ static inline int futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval,
119 return -ENOSYS; 120 return -ENOSYS;
120#endif 121#endif
121 122
122 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int))) 123 if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
123 return -EFAULT; 124 return -EFAULT;
124 125
125 asm volatile("1:\t" LOCK_PREFIX "cmpxchgl %3, %1\n" 126 asm volatile("1:\t" LOCK_PREFIX "cmpxchgl %4, %2\n"
126 "2:\t.section .fixup, \"ax\"\n" 127 "2:\t.section .fixup, \"ax\"\n"
127 "3:\tmov %2, %0\n" 128 "3:\tmov %3, %0\n"
128 "\tjmp 2b\n" 129 "\tjmp 2b\n"
129 "\t.previous\n" 130 "\t.previous\n"
130 _ASM_EXTABLE(1b, 3b) 131 _ASM_EXTABLE(1b, 3b)
131 : "=a" (oldval), "+m" (*uaddr) 132 : "+r" (ret), "=a" (oldval), "+m" (*uaddr)
132 : "i" (-EFAULT), "r" (newval), "0" (oldval) 133 : "i" (-EFAULT), "r" (newval), "1" (oldval)
133 : "memory" 134 : "memory"
134 ); 135 );
135 136
136 return oldval; 137 *uval = oldval;
138 return ret;
137} 139}
138 140
139#endif 141#endif
diff --git a/arch/x86/include/asm/hw_irq.h b/arch/x86/include/asm/hw_irq.h
index 0274ec5a7e62..bb9efe8706e2 100644
--- a/arch/x86/include/asm/hw_irq.h
+++ b/arch/x86/include/asm/hw_irq.h
@@ -45,6 +45,30 @@ extern void invalidate_interrupt4(void);
45extern void invalidate_interrupt5(void); 45extern void invalidate_interrupt5(void);
46extern void invalidate_interrupt6(void); 46extern void invalidate_interrupt6(void);
47extern void invalidate_interrupt7(void); 47extern void invalidate_interrupt7(void);
48extern void invalidate_interrupt8(void);
49extern void invalidate_interrupt9(void);
50extern void invalidate_interrupt10(void);
51extern void invalidate_interrupt11(void);
52extern void invalidate_interrupt12(void);
53extern void invalidate_interrupt13(void);
54extern void invalidate_interrupt14(void);
55extern void invalidate_interrupt15(void);
56extern void invalidate_interrupt16(void);
57extern void invalidate_interrupt17(void);
58extern void invalidate_interrupt18(void);
59extern void invalidate_interrupt19(void);
60extern void invalidate_interrupt20(void);
61extern void invalidate_interrupt21(void);
62extern void invalidate_interrupt22(void);
63extern void invalidate_interrupt23(void);
64extern void invalidate_interrupt24(void);
65extern void invalidate_interrupt25(void);
66extern void invalidate_interrupt26(void);
67extern void invalidate_interrupt27(void);
68extern void invalidate_interrupt28(void);
69extern void invalidate_interrupt29(void);
70extern void invalidate_interrupt30(void);
71extern void invalidate_interrupt31(void);
48 72
49extern void irq_move_cleanup_interrupt(void); 73extern void irq_move_cleanup_interrupt(void);
50extern void reboot_interrupt(void); 74extern void reboot_interrupt(void);
diff --git a/arch/x86/include/asm/init.h b/arch/x86/include/asm/init.h
index 36fb1a6a5109..8dbe353e41e1 100644
--- a/arch/x86/include/asm/init.h
+++ b/arch/x86/include/asm/init.h
@@ -11,8 +11,8 @@ kernel_physical_mapping_init(unsigned long start,
11 unsigned long page_size_mask); 11 unsigned long page_size_mask);
12 12
13 13
14extern unsigned long __initdata e820_table_start; 14extern unsigned long __initdata pgt_buf_start;
15extern unsigned long __meminitdata e820_table_end; 15extern unsigned long __meminitdata pgt_buf_end;
16extern unsigned long __meminitdata e820_table_top; 16extern unsigned long __meminitdata pgt_buf_top;
17 17
18#endif /* _ASM_X86_INIT_32_H */ 18#endif /* _ASM_X86_INIT_32_H */
diff --git a/arch/x86/include/asm/io_apic.h b/arch/x86/include/asm/io_apic.h
index f327d386d6cc..c4bd267dfc50 100644
--- a/arch/x86/include/asm/io_apic.h
+++ b/arch/x86/include/asm/io_apic.h
@@ -63,17 +63,6 @@ union IO_APIC_reg_03 {
63 } __attribute__ ((packed)) bits; 63 } __attribute__ ((packed)) bits;
64}; 64};
65 65
66enum ioapic_irq_destination_types {
67 dest_Fixed = 0,
68 dest_LowestPrio = 1,
69 dest_SMI = 2,
70 dest__reserved_1 = 3,
71 dest_NMI = 4,
72 dest_INIT = 5,
73 dest__reserved_2 = 6,
74 dest_ExtINT = 7
75};
76
77struct IO_APIC_route_entry { 66struct IO_APIC_route_entry {
78 __u32 vector : 8, 67 __u32 vector : 8,
79 delivery_mode : 3, /* 000: FIXED 68 delivery_mode : 3, /* 000: FIXED
@@ -106,6 +95,10 @@ struct IR_IO_APIC_route_entry {
106 index : 15; 95 index : 15;
107} __attribute__ ((packed)); 96} __attribute__ ((packed));
108 97
98#define IOAPIC_AUTO -1
99#define IOAPIC_EDGE 0
100#define IOAPIC_LEVEL 1
101
109#ifdef CONFIG_X86_IO_APIC 102#ifdef CONFIG_X86_IO_APIC
110 103
111/* 104/*
@@ -150,11 +143,6 @@ extern int timer_through_8259;
150#define io_apic_assign_pci_irqs \ 143#define io_apic_assign_pci_irqs \
151 (mp_irq_entries && !skip_ioapic_setup && io_apic_irqs) 144 (mp_irq_entries && !skip_ioapic_setup && io_apic_irqs)
152 145
153extern u8 io_apic_unique_id(u8 id);
154extern int io_apic_get_unique_id(int ioapic, int apic_id);
155extern int io_apic_get_version(int ioapic);
156extern int io_apic_get_redir_entries(int ioapic);
157
158struct io_apic_irq_attr; 146struct io_apic_irq_attr;
159extern int io_apic_set_pci_routing(struct device *dev, int irq, 147extern int io_apic_set_pci_routing(struct device *dev, int irq,
160 struct io_apic_irq_attr *irq_attr); 148 struct io_apic_irq_attr *irq_attr);
@@ -162,6 +150,8 @@ void setup_IO_APIC_irq_extra(u32 gsi);
162extern void ioapic_and_gsi_init(void); 150extern void ioapic_and_gsi_init(void);
163extern void ioapic_insert_resources(void); 151extern void ioapic_insert_resources(void);
164 152
153int io_apic_setup_irq_pin(unsigned int irq, int node, struct io_apic_irq_attr *attr);
154
165extern struct IO_APIC_route_entry **alloc_ioapic_entries(void); 155extern struct IO_APIC_route_entry **alloc_ioapic_entries(void);
166extern void free_ioapic_entries(struct IO_APIC_route_entry **ioapic_entries); 156extern void free_ioapic_entries(struct IO_APIC_route_entry **ioapic_entries);
167extern int save_IO_APIC_setup(struct IO_APIC_route_entry **ioapic_entries); 157extern int save_IO_APIC_setup(struct IO_APIC_route_entry **ioapic_entries);
@@ -186,6 +176,8 @@ extern void __init pre_init_apic_IRQ0(void);
186 176
187extern void mp_save_irq(struct mpc_intsrc *m); 177extern void mp_save_irq(struct mpc_intsrc *m);
188 178
179extern void disable_ioapic_support(void);
180
189#else /* !CONFIG_X86_IO_APIC */ 181#else /* !CONFIG_X86_IO_APIC */
190 182
191#define io_apic_assign_pci_irqs 0 183#define io_apic_assign_pci_irqs 0
@@ -199,6 +191,26 @@ static inline int mp_find_ioapic(u32 gsi) { return 0; }
199struct io_apic_irq_attr; 191struct io_apic_irq_attr;
200static inline int io_apic_set_pci_routing(struct device *dev, int irq, 192static inline int io_apic_set_pci_routing(struct device *dev, int irq,
201 struct io_apic_irq_attr *irq_attr) { return 0; } 193 struct io_apic_irq_attr *irq_attr) { return 0; }
194
195static inline struct IO_APIC_route_entry **alloc_ioapic_entries(void)
196{
197 return NULL;
198}
199
200static inline void free_ioapic_entries(struct IO_APIC_route_entry **ent) { }
201static inline int save_IO_APIC_setup(struct IO_APIC_route_entry **ent)
202{
203 return -ENOMEM;
204}
205
206static inline void mask_IO_APIC_setup(struct IO_APIC_route_entry **ent) { }
207static inline int restore_IO_APIC_setup(struct IO_APIC_route_entry **ent)
208{
209 return -ENOMEM;
210}
211
212static inline void mp_save_irq(struct mpc_intsrc *m) { };
213static inline void disable_ioapic_support(void) { }
202#endif 214#endif
203 215
204#endif /* _ASM_X86_IO_APIC_H */ 216#endif /* _ASM_X86_IO_APIC_H */
diff --git a/arch/x86/include/asm/ipi.h b/arch/x86/include/asm/ipi.h
index 0b7228268a63..615fa9061b57 100644
--- a/arch/x86/include/asm/ipi.h
+++ b/arch/x86/include/asm/ipi.h
@@ -123,10 +123,6 @@ extern void default_send_IPI_mask_sequence_phys(const struct cpumask *mask,
123 int vector); 123 int vector);
124extern void default_send_IPI_mask_allbutself_phys(const struct cpumask *mask, 124extern void default_send_IPI_mask_allbutself_phys(const struct cpumask *mask,
125 int vector); 125 int vector);
126extern void default_send_IPI_mask_sequence_logical(const struct cpumask *mask,
127 int vector);
128extern void default_send_IPI_mask_allbutself_logical(const struct cpumask *mask,
129 int vector);
130 126
131/* Avoid include hell */ 127/* Avoid include hell */
132#define NMI_VECTOR 0x02 128#define NMI_VECTOR 0x02
@@ -150,6 +146,10 @@ static inline void __default_local_send_IPI_all(int vector)
150} 146}
151 147
152#ifdef CONFIG_X86_32 148#ifdef CONFIG_X86_32
149extern void default_send_IPI_mask_sequence_logical(const struct cpumask *mask,
150 int vector);
151extern void default_send_IPI_mask_allbutself_logical(const struct cpumask *mask,
152 int vector);
153extern void default_send_IPI_mask_logical(const struct cpumask *mask, 153extern void default_send_IPI_mask_logical(const struct cpumask *mask,
154 int vector); 154 int vector);
155extern void default_send_IPI_allbutself(int vector); 155extern void default_send_IPI_allbutself(int vector);
diff --git a/arch/x86/include/asm/irq.h b/arch/x86/include/asm/irq.h
index c704b38c57a2..ba870bb6dd8e 100644
--- a/arch/x86/include/asm/irq.h
+++ b/arch/x86/include/asm/irq.h
@@ -10,9 +10,6 @@
10#include <asm/apicdef.h> 10#include <asm/apicdef.h>
11#include <asm/irq_vectors.h> 11#include <asm/irq_vectors.h>
12 12
13/* Even though we don't support this, supply it to appease OF */
14static inline void irq_dispose_mapping(unsigned int virq) { }
15
16static inline int irq_canonicalize(int irq) 13static inline int irq_canonicalize(int irq)
17{ 14{
18 return ((irq == 2) ? 9 : irq); 15 return ((irq == 2) ? 9 : irq);
diff --git a/arch/x86/include/asm/irq_controller.h b/arch/x86/include/asm/irq_controller.h
new file mode 100644
index 000000000000..423bbbddf36d
--- /dev/null
+++ b/arch/x86/include/asm/irq_controller.h
@@ -0,0 +1,12 @@
1#ifndef __IRQ_CONTROLLER__
2#define __IRQ_CONTROLLER__
3
4struct irq_domain {
5 int (*xlate)(struct irq_domain *h, const u32 *intspec, u32 intsize,
6 u32 *out_hwirq, u32 *out_type);
7 void *priv;
8 struct device_node *controller;
9 struct list_head l;
10};
11
12#endif
diff --git a/arch/x86/include/asm/irq_vectors.h b/arch/x86/include/asm/irq_vectors.h
index 6af0894dafb4..6e976ee3b3ef 100644
--- a/arch/x86/include/asm/irq_vectors.h
+++ b/arch/x86/include/asm/irq_vectors.h
@@ -1,6 +1,7 @@
1#ifndef _ASM_X86_IRQ_VECTORS_H 1#ifndef _ASM_X86_IRQ_VECTORS_H
2#define _ASM_X86_IRQ_VECTORS_H 2#define _ASM_X86_IRQ_VECTORS_H
3 3
4#include <linux/threads.h>
4/* 5/*
5 * Linux IRQ vector layout. 6 * Linux IRQ vector layout.
6 * 7 *
@@ -16,8 +17,8 @@
16 * Vectors 0 ... 31 : system traps and exceptions - hardcoded events 17 * Vectors 0 ... 31 : system traps and exceptions - hardcoded events
17 * Vectors 32 ... 127 : device interrupts 18 * Vectors 32 ... 127 : device interrupts
18 * Vector 128 : legacy int80 syscall interface 19 * Vector 128 : legacy int80 syscall interface
19 * Vectors 129 ... 237 : device interrupts 20 * Vectors 129 ... INVALIDATE_TLB_VECTOR_START-1 : device interrupts
20 * Vectors 238 ... 255 : special interrupts 21 * Vectors INVALIDATE_TLB_VECTOR_START ... 255 : special interrupts
21 * 22 *
22 * 64-bit x86 has per CPU IDT tables, 32-bit has one shared IDT table. 23 * 64-bit x86 has per CPU IDT tables, 32-bit has one shared IDT table.
23 * 24 *
@@ -96,37 +97,43 @@
96#define THRESHOLD_APIC_VECTOR 0xf9 97#define THRESHOLD_APIC_VECTOR 0xf9
97#define REBOOT_VECTOR 0xf8 98#define REBOOT_VECTOR 0xf8
98 99
99/* f0-f7 used for spreading out TLB flushes: */
100#define INVALIDATE_TLB_VECTOR_END 0xf7
101#define INVALIDATE_TLB_VECTOR_START 0xf0
102#define NUM_INVALIDATE_TLB_VECTORS 8
103
104/*
105 * Local APIC timer IRQ vector is on a different priority level,
106 * to work around the 'lost local interrupt if more than 2 IRQ
107 * sources per level' errata.
108 */
109#define LOCAL_TIMER_VECTOR 0xef
110
111/* 100/*
112 * Generic system vector for platform specific use 101 * Generic system vector for platform specific use
113 */ 102 */
114#define X86_PLATFORM_IPI_VECTOR 0xed 103#define X86_PLATFORM_IPI_VECTOR 0xf7
115 104
116/* 105/*
117 * IRQ work vector: 106 * IRQ work vector:
118 */ 107 */
119#define IRQ_WORK_VECTOR 0xec 108#define IRQ_WORK_VECTOR 0xf6
120 109
121#define UV_BAU_MESSAGE 0xea 110#define UV_BAU_MESSAGE 0xf5
122 111
123/* 112/*
124 * Self IPI vector for machine checks 113 * Self IPI vector for machine checks
125 */ 114 */
126#define MCE_SELF_VECTOR 0xeb 115#define MCE_SELF_VECTOR 0xf4
127 116
128/* Xen vector callback to receive events in a HVM domain */ 117/* Xen vector callback to receive events in a HVM domain */
129#define XEN_HVM_EVTCHN_CALLBACK 0xe9 118#define XEN_HVM_EVTCHN_CALLBACK 0xf3
119
120/*
121 * Local APIC timer IRQ vector is on a different priority level,
122 * to work around the 'lost local interrupt if more than 2 IRQ
123 * sources per level' errata.
124 */
125#define LOCAL_TIMER_VECTOR 0xef
126
127/* up to 32 vectors used for spreading out TLB flushes: */
128#if NR_CPUS <= 32
129# define NUM_INVALIDATE_TLB_VECTORS (NR_CPUS)
130#else
131# define NUM_INVALIDATE_TLB_VECTORS (32)
132#endif
133
134#define INVALIDATE_TLB_VECTOR_END (0xee)
135#define INVALIDATE_TLB_VECTOR_START \
136 (INVALIDATE_TLB_VECTOR_END-NUM_INVALIDATE_TLB_VECTORS+1)
130 137
131#define NR_VECTORS 256 138#define NR_VECTORS 256
132 139
diff --git a/arch/x86/include/asm/kdebug.h b/arch/x86/include/asm/kdebug.h
index ca242d35e873..518bbbb9ee59 100644
--- a/arch/x86/include/asm/kdebug.h
+++ b/arch/x86/include/asm/kdebug.h
@@ -13,7 +13,6 @@ enum die_val {
13 DIE_PANIC, 13 DIE_PANIC,
14 DIE_NMI, 14 DIE_NMI,
15 DIE_DIE, 15 DIE_DIE,
16 DIE_NMIWATCHDOG,
17 DIE_KERNELDEBUG, 16 DIE_KERNELDEBUG,
18 DIE_TRAP, 17 DIE_TRAP,
19 DIE_GPF, 18 DIE_GPF,
diff --git a/arch/x86/include/asm/mpspec.h b/arch/x86/include/asm/mpspec.h
index 0c90dd9f0505..9c7d95f6174b 100644
--- a/arch/x86/include/asm/mpspec.h
+++ b/arch/x86/include/asm/mpspec.h
@@ -25,7 +25,6 @@ extern int pic_mode;
25#define MAX_IRQ_SOURCES 256 25#define MAX_IRQ_SOURCES 256
26 26
27extern unsigned int def_to_bigsmp; 27extern unsigned int def_to_bigsmp;
28extern u8 apicid_2_node[];
29 28
30#ifdef CONFIG_X86_NUMAQ 29#ifdef CONFIG_X86_NUMAQ
31extern int mp_bus_id_to_node[MAX_MP_BUSSES]; 30extern int mp_bus_id_to_node[MAX_MP_BUSSES];
@@ -33,8 +32,6 @@ extern int mp_bus_id_to_local[MAX_MP_BUSSES];
33extern int quad_local_to_mp_bus_id [NR_CPUS/4][4]; 32extern int quad_local_to_mp_bus_id [NR_CPUS/4][4];
34#endif 33#endif
35 34
36#define MAX_APICID 256
37
38#else /* CONFIG_X86_64: */ 35#else /* CONFIG_X86_64: */
39 36
40#define MAX_MP_BUSSES 256 37#define MAX_MP_BUSSES 256
diff --git a/arch/x86/include/asm/msr-index.h b/arch/x86/include/asm/msr-index.h
index 43a18c77676d..823d48223400 100644
--- a/arch/x86/include/asm/msr-index.h
+++ b/arch/x86/include/asm/msr-index.h
@@ -52,6 +52,9 @@
52#define MSR_IA32_MCG_STATUS 0x0000017a 52#define MSR_IA32_MCG_STATUS 0x0000017a
53#define MSR_IA32_MCG_CTL 0x0000017b 53#define MSR_IA32_MCG_CTL 0x0000017b
54 54
55#define MSR_OFFCORE_RSP_0 0x000001a6
56#define MSR_OFFCORE_RSP_1 0x000001a7
57
55#define MSR_IA32_PEBS_ENABLE 0x000003f1 58#define MSR_IA32_PEBS_ENABLE 0x000003f1
56#define MSR_IA32_DS_AREA 0x00000600 59#define MSR_IA32_DS_AREA 0x00000600
57#define MSR_IA32_PERF_CAPABILITIES 0x00000345 60#define MSR_IA32_PERF_CAPABILITIES 0x00000345
diff --git a/arch/x86/include/asm/nmi.h b/arch/x86/include/asm/nmi.h
index c76f5b92b840..07f46016d3ff 100644
--- a/arch/x86/include/asm/nmi.h
+++ b/arch/x86/include/asm/nmi.h
@@ -7,7 +7,6 @@
7 7
8#ifdef CONFIG_X86_LOCAL_APIC 8#ifdef CONFIG_X86_LOCAL_APIC
9 9
10extern void die_nmi(char *str, struct pt_regs *regs, int do_panic);
11extern int avail_to_resrv_perfctr_nmi_bit(unsigned int); 10extern int avail_to_resrv_perfctr_nmi_bit(unsigned int);
12extern int reserve_perfctr_nmi(unsigned int); 11extern int reserve_perfctr_nmi(unsigned int);
13extern void release_perfctr_nmi(unsigned int); 12extern void release_perfctr_nmi(unsigned int);
diff --git a/arch/x86/include/asm/numa.h b/arch/x86/include/asm/numa.h
index 27da400d3138..3d4dab43c994 100644
--- a/arch/x86/include/asm/numa.h
+++ b/arch/x86/include/asm/numa.h
@@ -1,5 +1,57 @@
1#ifndef _ASM_X86_NUMA_H
2#define _ASM_X86_NUMA_H
3
4#include <asm/topology.h>
5#include <asm/apicdef.h>
6
7#ifdef CONFIG_NUMA
8
9#define NR_NODE_MEMBLKS (MAX_NUMNODES*2)
10
11/*
12 * __apicid_to_node[] stores the raw mapping between physical apicid and
13 * node and is used to initialize cpu_to_node mapping.
14 *
15 * The mapping may be overridden by apic->numa_cpu_node() on 32bit and thus
16 * should be accessed by the accessors - set_apicid_to_node() and
17 * numa_cpu_node().
18 */
19extern s16 __apicid_to_node[MAX_LOCAL_APIC];
20
21static inline void set_apicid_to_node(int apicid, s16 node)
22{
23 __apicid_to_node[apicid] = node;
24}
25#else /* CONFIG_NUMA */
26static inline void set_apicid_to_node(int apicid, s16 node)
27{
28}
29#endif /* CONFIG_NUMA */
30
1#ifdef CONFIG_X86_32 31#ifdef CONFIG_X86_32
2# include "numa_32.h" 32# include "numa_32.h"
3#else 33#else
4# include "numa_64.h" 34# include "numa_64.h"
5#endif 35#endif
36
37#ifdef CONFIG_NUMA
38extern void __cpuinit numa_set_node(int cpu, int node);
39extern void __cpuinit numa_clear_node(int cpu);
40extern void __init numa_init_array(void);
41extern void __init init_cpu_to_node(void);
42extern void __cpuinit numa_add_cpu(int cpu);
43extern void __cpuinit numa_remove_cpu(int cpu);
44#else /* CONFIG_NUMA */
45static inline void numa_set_node(int cpu, int node) { }
46static inline void numa_clear_node(int cpu) { }
47static inline void numa_init_array(void) { }
48static inline void init_cpu_to_node(void) { }
49static inline void numa_add_cpu(int cpu) { }
50static inline void numa_remove_cpu(int cpu) { }
51#endif /* CONFIG_NUMA */
52
53#ifdef CONFIG_DEBUG_PER_CPU_MAPS
54struct cpumask __cpuinit *debug_cpumask_set_cpu(int cpu, int enable);
55#endif
56
57#endif /* _ASM_X86_NUMA_H */
diff --git a/arch/x86/include/asm/numa_32.h b/arch/x86/include/asm/numa_32.h
index b0ef2b449a9d..c6beed1ef103 100644
--- a/arch/x86/include/asm/numa_32.h
+++ b/arch/x86/include/asm/numa_32.h
@@ -4,7 +4,12 @@
4extern int numa_off; 4extern int numa_off;
5 5
6extern int pxm_to_nid(int pxm); 6extern int pxm_to_nid(int pxm);
7extern void numa_remove_cpu(int cpu); 7
8#ifdef CONFIG_NUMA
9extern int __cpuinit numa_cpu_node(int cpu);
10#else /* CONFIG_NUMA */
11static inline int numa_cpu_node(int cpu) { return NUMA_NO_NODE; }
12#endif /* CONFIG_NUMA */
8 13
9#ifdef CONFIG_HIGHMEM 14#ifdef CONFIG_HIGHMEM
10extern void set_highmem_pages_init(void); 15extern void set_highmem_pages_init(void);
diff --git a/arch/x86/include/asm/numa_64.h b/arch/x86/include/asm/numa_64.h
index 0493be39607c..344eb1790b46 100644
--- a/arch/x86/include/asm/numa_64.h
+++ b/arch/x86/include/asm/numa_64.h
@@ -2,23 +2,16 @@
2#define _ASM_X86_NUMA_64_H 2#define _ASM_X86_NUMA_64_H
3 3
4#include <linux/nodemask.h> 4#include <linux/nodemask.h>
5#include <asm/apicdef.h>
6 5
7struct bootnode { 6struct bootnode {
8 u64 start; 7 u64 start;
9 u64 end; 8 u64 end;
10}; 9};
11 10
12extern int compute_hash_shift(struct bootnode *nodes, int numblks,
13 int *nodeids);
14
15#define ZONE_ALIGN (1UL << (MAX_ORDER+PAGE_SHIFT)) 11#define ZONE_ALIGN (1UL << (MAX_ORDER+PAGE_SHIFT))
16 12
17extern void numa_init_array(void);
18extern int numa_off; 13extern int numa_off;
19 14
20extern s16 apicid_to_node[MAX_LOCAL_APIC];
21
22extern unsigned long numa_free_all_bootmem(void); 15extern unsigned long numa_free_all_bootmem(void);
23extern void setup_node_bootmem(int nodeid, unsigned long start, 16extern void setup_node_bootmem(int nodeid, unsigned long start,
24 unsigned long end); 17 unsigned long end);
@@ -31,11 +24,11 @@ extern void setup_node_bootmem(int nodeid, unsigned long start,
31 */ 24 */
32#define NODE_MIN_SIZE (4*1024*1024) 25#define NODE_MIN_SIZE (4*1024*1024)
33 26
34extern void __init init_cpu_to_node(void); 27extern nodemask_t numa_nodes_parsed __initdata;
35extern void __cpuinit numa_set_node(int cpu, int node); 28
36extern void __cpuinit numa_clear_node(int cpu); 29extern int __cpuinit numa_cpu_node(int cpu);
37extern void __cpuinit numa_add_cpu(int cpu); 30extern int __init numa_add_memblk(int nodeid, u64 start, u64 end);
38extern void __cpuinit numa_remove_cpu(int cpu); 31extern void __init numa_set_distance(int from, int to, int distance);
39 32
40#ifdef CONFIG_NUMA_EMU 33#ifdef CONFIG_NUMA_EMU
41#define FAKE_NODE_MIN_SIZE ((u64)32 << 20) 34#define FAKE_NODE_MIN_SIZE ((u64)32 << 20)
@@ -43,11 +36,7 @@ extern void __cpuinit numa_remove_cpu(int cpu);
43void numa_emu_cmdline(char *); 36void numa_emu_cmdline(char *);
44#endif /* CONFIG_NUMA_EMU */ 37#endif /* CONFIG_NUMA_EMU */
45#else 38#else
46static inline void init_cpu_to_node(void) { } 39static inline int numa_cpu_node(int cpu) { return NUMA_NO_NODE; }
47static inline void numa_set_node(int cpu, int node) { }
48static inline void numa_clear_node(int cpu) { }
49static inline void numa_add_cpu(int cpu, int node) { }
50static inline void numa_remove_cpu(int cpu) { }
51#endif 40#endif
52 41
53#endif /* _ASM_X86_NUMA_64_H */ 42#endif /* _ASM_X86_NUMA_64_H */
diff --git a/arch/x86/include/asm/olpc_ofw.h b/arch/x86/include/asm/olpc_ofw.h
index 641988efe063..c5d3a5abbb9f 100644
--- a/arch/x86/include/asm/olpc_ofw.h
+++ b/arch/x86/include/asm/olpc_ofw.h
@@ -6,7 +6,7 @@
6 6
7#define OLPC_OFW_SIG 0x2057464F /* aka "OFW " */ 7#define OLPC_OFW_SIG 0x2057464F /* aka "OFW " */
8 8
9#ifdef CONFIG_OLPC_OPENFIRMWARE 9#ifdef CONFIG_OLPC
10 10
11extern bool olpc_ofw_is_installed(void); 11extern bool olpc_ofw_is_installed(void);
12 12
@@ -26,19 +26,15 @@ extern void setup_olpc_ofw_pgd(void);
26/* check if OFW was detected during boot */ 26/* check if OFW was detected during boot */
27extern bool olpc_ofw_present(void); 27extern bool olpc_ofw_present(void);
28 28
29#else /* !CONFIG_OLPC_OPENFIRMWARE */ 29#else /* !CONFIG_OLPC */
30
31static inline bool olpc_ofw_is_installed(void) { return false; }
32static inline void olpc_ofw_detect(void) { } 30static inline void olpc_ofw_detect(void) { }
33static inline void setup_olpc_ofw_pgd(void) { } 31static inline void setup_olpc_ofw_pgd(void) { }
34static inline bool olpc_ofw_present(void) { return false; } 32#endif /* !CONFIG_OLPC */
35
36#endif /* !CONFIG_OLPC_OPENFIRMWARE */
37 33
38#ifdef CONFIG_OLPC_OPENFIRMWARE_DT 34#ifdef CONFIG_OF_PROMTREE
39extern void olpc_dt_build_devicetree(void); 35extern void olpc_dt_build_devicetree(void);
40#else 36#else
41static inline void olpc_dt_build_devicetree(void) { } 37static inline void olpc_dt_build_devicetree(void) { }
42#endif /* CONFIG_OLPC_OPENFIRMWARE_DT */ 38#endif
43 39
44#endif /* _ASM_X86_OLPC_OFW_H */ 40#endif /* _ASM_X86_OLPC_OFW_H */
diff --git a/arch/x86/include/asm/page_types.h b/arch/x86/include/asm/page_types.h
index 1df66211fd1b..bce688d54c12 100644
--- a/arch/x86/include/asm/page_types.h
+++ b/arch/x86/include/asm/page_types.h
@@ -2,6 +2,7 @@
2#define _ASM_X86_PAGE_DEFS_H 2#define _ASM_X86_PAGE_DEFS_H
3 3
4#include <linux/const.h> 4#include <linux/const.h>
5#include <linux/types.h>
5 6
6/* PAGE_SHIFT determines the page size */ 7/* PAGE_SHIFT determines the page size */
7#define PAGE_SHIFT 12 8#define PAGE_SHIFT 12
@@ -45,11 +46,15 @@ extern int devmem_is_allowed(unsigned long pagenr);
45extern unsigned long max_low_pfn_mapped; 46extern unsigned long max_low_pfn_mapped;
46extern unsigned long max_pfn_mapped; 47extern unsigned long max_pfn_mapped;
47 48
49static inline phys_addr_t get_max_mapped(void)
50{
51 return (phys_addr_t)max_pfn_mapped << PAGE_SHIFT;
52}
53
48extern unsigned long init_memory_mapping(unsigned long start, 54extern unsigned long init_memory_mapping(unsigned long start,
49 unsigned long end); 55 unsigned long end);
50 56
51extern void initmem_init(unsigned long start_pfn, unsigned long end_pfn, 57extern void initmem_init(void);
52 int acpi, int k8);
53extern void free_initmem(void); 58extern void free_initmem(void);
54 59
55#endif /* !__ASSEMBLY__ */ 60#endif /* !__ASSEMBLY__ */
diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h
index 45636cefa186..4c25ab48257b 100644
--- a/arch/x86/include/asm/processor.h
+++ b/arch/x86/include/asm/processor.h
@@ -94,10 +94,6 @@ struct cpuinfo_x86 {
94 int x86_cache_alignment; /* In bytes */ 94 int x86_cache_alignment; /* In bytes */
95 int x86_power; 95 int x86_power;
96 unsigned long loops_per_jiffy; 96 unsigned long loops_per_jiffy;
97#ifdef CONFIG_SMP
98 /* cpus sharing the last level cache: */
99 cpumask_var_t llc_shared_map;
100#endif
101 /* cpuid returned max cores value: */ 97 /* cpuid returned max cores value: */
102 u16 x86_max_cores; 98 u16 x86_max_cores;
103 u16 apicid; 99 u16 apicid;
diff --git a/arch/x86/include/asm/prom.h b/arch/x86/include/asm/prom.h
index b4ec95f07518..971e0b46446e 100644
--- a/arch/x86/include/asm/prom.h
+++ b/arch/x86/include/asm/prom.h
@@ -1 +1,69 @@
1/* dummy prom.h; here to make linux/of.h's #includes happy */ 1/*
2 * Definitions for Device tree / OpenFirmware handling on X86
3 *
4 * based on arch/powerpc/include/asm/prom.h which is
5 * Copyright (C) 1996-2005 Paul Mackerras.
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#ifndef _ASM_X86_PROM_H
14#define _ASM_X86_PROM_H
15#ifndef __ASSEMBLY__
16
17#include <linux/of.h>
18#include <linux/types.h>
19#include <linux/pci.h>
20
21#include <asm/irq.h>
22#include <asm/atomic.h>
23#include <asm/setup.h>
24#include <asm/irq_controller.h>
25
26#ifdef CONFIG_OF
27extern int of_ioapic;
28extern u64 initial_dtb;
29extern void add_dtb(u64 data);
30extern void x86_add_irq_domains(void);
31void __cpuinit x86_of_pci_init(void);
32void x86_dtb_init(void);
33
34static inline struct device_node *pci_device_to_OF_node(struct pci_dev *pdev)
35{
36 return pdev ? pdev->dev.of_node : NULL;
37}
38
39static inline struct device_node *pci_bus_to_OF_node(struct pci_bus *bus)
40{
41 return pci_device_to_OF_node(bus->self);
42}
43
44#else
45static inline void add_dtb(u64 data) { }
46static inline void x86_add_irq_domains(void) { }
47static inline void x86_of_pci_init(void) { }
48static inline void x86_dtb_init(void) { }
49#define of_ioapic 0
50#endif
51
52extern char cmd_line[COMMAND_LINE_SIZE];
53
54#define pci_address_to_pio pci_address_to_pio
55unsigned long pci_address_to_pio(phys_addr_t addr);
56
57/**
58 * irq_dispose_mapping - Unmap an interrupt
59 * @virq: linux virq number of the interrupt to unmap
60 *
61 * FIXME: We really should implement proper virq handling like power,
62 * but that's going to be major surgery.
63 */
64static inline void irq_dispose_mapping(unsigned int virq) { }
65
66#define HAVE_ARCH_DEVTREE_FIXUPS
67
68#endif /* __ASSEMBLY__ */
69#endif
diff --git a/arch/x86/include/asm/rwsem.h b/arch/x86/include/asm/rwsem.h
index d1e41b0f9b60..df4cd32b4cc6 100644
--- a/arch/x86/include/asm/rwsem.h
+++ b/arch/x86/include/asm/rwsem.h
@@ -37,26 +37,9 @@
37#endif 37#endif
38 38
39#ifdef __KERNEL__ 39#ifdef __KERNEL__
40
41#include <linux/list.h>
42#include <linux/spinlock.h>
43#include <linux/lockdep.h>
44#include <asm/asm.h> 40#include <asm/asm.h>
45 41
46struct rwsem_waiter;
47
48extern asmregparm struct rw_semaphore *
49 rwsem_down_read_failed(struct rw_semaphore *sem);
50extern asmregparm struct rw_semaphore *
51 rwsem_down_write_failed(struct rw_semaphore *sem);
52extern asmregparm struct rw_semaphore *
53 rwsem_wake(struct rw_semaphore *);
54extern asmregparm struct rw_semaphore *
55 rwsem_downgrade_wake(struct rw_semaphore *sem);
56
57/* 42/*
58 * the semaphore definition
59 *
60 * The bias values and the counter type limits the number of 43 * The bias values and the counter type limits the number of
61 * potential readers/writers to 32767 for 32 bits and 2147483647 44 * potential readers/writers to 32767 for 32 bits and 2147483647
62 * for 64 bits. 45 * for 64 bits.
@@ -74,43 +57,6 @@ extern asmregparm struct rw_semaphore *
74#define RWSEM_ACTIVE_READ_BIAS RWSEM_ACTIVE_BIAS 57#define RWSEM_ACTIVE_READ_BIAS RWSEM_ACTIVE_BIAS
75#define RWSEM_ACTIVE_WRITE_BIAS (RWSEM_WAITING_BIAS + RWSEM_ACTIVE_BIAS) 58#define RWSEM_ACTIVE_WRITE_BIAS (RWSEM_WAITING_BIAS + RWSEM_ACTIVE_BIAS)
76 59
77typedef signed long rwsem_count_t;
78
79struct rw_semaphore {
80 rwsem_count_t count;
81 spinlock_t wait_lock;
82 struct list_head wait_list;
83#ifdef CONFIG_DEBUG_LOCK_ALLOC
84 struct lockdep_map dep_map;
85#endif
86};
87
88#ifdef CONFIG_DEBUG_LOCK_ALLOC
89# define __RWSEM_DEP_MAP_INIT(lockname) , .dep_map = { .name = #lockname }
90#else
91# define __RWSEM_DEP_MAP_INIT(lockname)
92#endif
93
94
95#define __RWSEM_INITIALIZER(name) \
96{ \
97 RWSEM_UNLOCKED_VALUE, __SPIN_LOCK_UNLOCKED((name).wait_lock), \
98 LIST_HEAD_INIT((name).wait_list) __RWSEM_DEP_MAP_INIT(name) \
99}
100
101#define DECLARE_RWSEM(name) \
102 struct rw_semaphore name = __RWSEM_INITIALIZER(name)
103
104extern void __init_rwsem(struct rw_semaphore *sem, const char *name,
105 struct lock_class_key *key);
106
107#define init_rwsem(sem) \
108do { \
109 static struct lock_class_key __key; \
110 \
111 __init_rwsem((sem), #sem, &__key); \
112} while (0)
113
114/* 60/*
115 * lock for reading 61 * lock for reading
116 */ 62 */
@@ -133,7 +79,7 @@ static inline void __down_read(struct rw_semaphore *sem)
133 */ 79 */
134static inline int __down_read_trylock(struct rw_semaphore *sem) 80static inline int __down_read_trylock(struct rw_semaphore *sem)
135{ 81{
136 rwsem_count_t result, tmp; 82 long result, tmp;
137 asm volatile("# beginning __down_read_trylock\n\t" 83 asm volatile("# beginning __down_read_trylock\n\t"
138 " mov %0,%1\n\t" 84 " mov %0,%1\n\t"
139 "1:\n\t" 85 "1:\n\t"
@@ -155,7 +101,7 @@ static inline int __down_read_trylock(struct rw_semaphore *sem)
155 */ 101 */
156static inline void __down_write_nested(struct rw_semaphore *sem, int subclass) 102static inline void __down_write_nested(struct rw_semaphore *sem, int subclass)
157{ 103{
158 rwsem_count_t tmp; 104 long tmp;
159 asm volatile("# beginning down_write\n\t" 105 asm volatile("# beginning down_write\n\t"
160 LOCK_PREFIX " xadd %1,(%2)\n\t" 106 LOCK_PREFIX " xadd %1,(%2)\n\t"
161 /* adds 0xffff0001, returns the old value */ 107 /* adds 0xffff0001, returns the old value */
@@ -180,9 +126,8 @@ static inline void __down_write(struct rw_semaphore *sem)
180 */ 126 */
181static inline int __down_write_trylock(struct rw_semaphore *sem) 127static inline int __down_write_trylock(struct rw_semaphore *sem)
182{ 128{
183 rwsem_count_t ret = cmpxchg(&sem->count, 129 long ret = cmpxchg(&sem->count, RWSEM_UNLOCKED_VALUE,
184 RWSEM_UNLOCKED_VALUE, 130 RWSEM_ACTIVE_WRITE_BIAS);
185 RWSEM_ACTIVE_WRITE_BIAS);
186 if (ret == RWSEM_UNLOCKED_VALUE) 131 if (ret == RWSEM_UNLOCKED_VALUE)
187 return 1; 132 return 1;
188 return 0; 133 return 0;
@@ -193,7 +138,7 @@ static inline int __down_write_trylock(struct rw_semaphore *sem)
193 */ 138 */
194static inline void __up_read(struct rw_semaphore *sem) 139static inline void __up_read(struct rw_semaphore *sem)
195{ 140{
196 rwsem_count_t tmp; 141 long tmp;
197 asm volatile("# beginning __up_read\n\t" 142 asm volatile("# beginning __up_read\n\t"
198 LOCK_PREFIX " xadd %1,(%2)\n\t" 143 LOCK_PREFIX " xadd %1,(%2)\n\t"
199 /* subtracts 1, returns the old value */ 144 /* subtracts 1, returns the old value */
@@ -211,7 +156,7 @@ static inline void __up_read(struct rw_semaphore *sem)
211 */ 156 */
212static inline void __up_write(struct rw_semaphore *sem) 157static inline void __up_write(struct rw_semaphore *sem)
213{ 158{
214 rwsem_count_t tmp; 159 long tmp;
215 asm volatile("# beginning __up_write\n\t" 160 asm volatile("# beginning __up_write\n\t"
216 LOCK_PREFIX " xadd %1,(%2)\n\t" 161 LOCK_PREFIX " xadd %1,(%2)\n\t"
217 /* subtracts 0xffff0001, returns the old value */ 162 /* subtracts 0xffff0001, returns the old value */
@@ -247,8 +192,7 @@ static inline void __downgrade_write(struct rw_semaphore *sem)
247/* 192/*
248 * implement atomic add functionality 193 * implement atomic add functionality
249 */ 194 */
250static inline void rwsem_atomic_add(rwsem_count_t delta, 195static inline void rwsem_atomic_add(long delta, struct rw_semaphore *sem)
251 struct rw_semaphore *sem)
252{ 196{
253 asm volatile(LOCK_PREFIX _ASM_ADD "%1,%0" 197 asm volatile(LOCK_PREFIX _ASM_ADD "%1,%0"
254 : "+m" (sem->count) 198 : "+m" (sem->count)
@@ -258,10 +202,9 @@ static inline void rwsem_atomic_add(rwsem_count_t delta,
258/* 202/*
259 * implement exchange and add functionality 203 * implement exchange and add functionality
260 */ 204 */
261static inline rwsem_count_t rwsem_atomic_update(rwsem_count_t delta, 205static inline long rwsem_atomic_update(long delta, struct rw_semaphore *sem)
262 struct rw_semaphore *sem)
263{ 206{
264 rwsem_count_t tmp = delta; 207 long tmp = delta;
265 208
266 asm volatile(LOCK_PREFIX "xadd %0,%1" 209 asm volatile(LOCK_PREFIX "xadd %0,%1"
267 : "+r" (tmp), "+m" (sem->count) 210 : "+r" (tmp), "+m" (sem->count)
@@ -270,10 +213,5 @@ static inline rwsem_count_t rwsem_atomic_update(rwsem_count_t delta,
270 return tmp + delta; 213 return tmp + delta;
271} 214}
272 215
273static inline int rwsem_is_locked(struct rw_semaphore *sem)
274{
275 return (sem->count != 0);
276}
277
278#endif /* __KERNEL__ */ 216#endif /* __KERNEL__ */
279#endif /* _ASM_X86_RWSEM_H */ 217#endif /* _ASM_X86_RWSEM_H */
diff --git a/arch/x86/include/asm/smp.h b/arch/x86/include/asm/smp.h
index 1f4695136776..73b11bc0ae6f 100644
--- a/arch/x86/include/asm/smp.h
+++ b/arch/x86/include/asm/smp.h
@@ -17,12 +17,24 @@
17#endif 17#endif
18#include <asm/thread_info.h> 18#include <asm/thread_info.h>
19#include <asm/cpumask.h> 19#include <asm/cpumask.h>
20#include <asm/cpufeature.h>
20 21
21extern int smp_num_siblings; 22extern int smp_num_siblings;
22extern unsigned int num_processors; 23extern unsigned int num_processors;
23 24
25static inline bool cpu_has_ht_siblings(void)
26{
27 bool has_siblings = false;
28#ifdef CONFIG_SMP
29 has_siblings = cpu_has_ht && smp_num_siblings > 1;
30#endif
31 return has_siblings;
32}
33
24DECLARE_PER_CPU(cpumask_var_t, cpu_sibling_map); 34DECLARE_PER_CPU(cpumask_var_t, cpu_sibling_map);
25DECLARE_PER_CPU(cpumask_var_t, cpu_core_map); 35DECLARE_PER_CPU(cpumask_var_t, cpu_core_map);
36/* cpus sharing the last level cache: */
37DECLARE_PER_CPU(cpumask_var_t, cpu_llc_shared_map);
26DECLARE_PER_CPU(u16, cpu_llc_id); 38DECLARE_PER_CPU(u16, cpu_llc_id);
27DECLARE_PER_CPU(int, cpu_number); 39DECLARE_PER_CPU(int, cpu_number);
28 40
@@ -36,8 +48,16 @@ static inline struct cpumask *cpu_core_mask(int cpu)
36 return per_cpu(cpu_core_map, cpu); 48 return per_cpu(cpu_core_map, cpu);
37} 49}
38 50
51static inline struct cpumask *cpu_llc_shared_mask(int cpu)
52{
53 return per_cpu(cpu_llc_shared_map, cpu);
54}
55
39DECLARE_EARLY_PER_CPU(u16, x86_cpu_to_apicid); 56DECLARE_EARLY_PER_CPU(u16, x86_cpu_to_apicid);
40DECLARE_EARLY_PER_CPU(u16, x86_bios_cpu_apicid); 57DECLARE_EARLY_PER_CPU(u16, x86_bios_cpu_apicid);
58#if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86_32)
59DECLARE_EARLY_PER_CPU(int, x86_cpu_to_logical_apicid);
60#endif
41 61
42/* Static state in head.S used to set up a CPU */ 62/* Static state in head.S used to set up a CPU */
43extern unsigned long stack_start; /* Initial stack pointer address */ 63extern unsigned long stack_start; /* Initial stack pointer address */
diff --git a/arch/x86/include/asm/system.h b/arch/x86/include/asm/system.h
index 33ecc3ea8782..12569e691ce3 100644
--- a/arch/x86/include/asm/system.h
+++ b/arch/x86/include/asm/system.h
@@ -98,8 +98,6 @@ do { \
98 */ 98 */
99#define HAVE_DISABLE_HLT 99#define HAVE_DISABLE_HLT
100#else 100#else
101#define __SAVE(reg, offset) "movq %%" #reg ",(14-" #offset ")*8(%%rsp)\n\t"
102#define __RESTORE(reg, offset) "movq (14-" #offset ")*8(%%rsp),%%" #reg "\n\t"
103 101
104/* frame pointer must be last for get_wchan */ 102/* frame pointer must be last for get_wchan */
105#define SAVE_CONTEXT "pushf ; pushq %%rbp ; movq %%rsi,%%rbp\n\t" 103#define SAVE_CONTEXT "pushf ; pushq %%rbp ; movq %%rsi,%%rbp\n\t"
diff --git a/arch/x86/include/asm/topology.h b/arch/x86/include/asm/topology.h
index 21899cc31e52..910a7084f7f2 100644
--- a/arch/x86/include/asm/topology.h
+++ b/arch/x86/include/asm/topology.h
@@ -47,21 +47,6 @@
47 47
48#include <asm/mpspec.h> 48#include <asm/mpspec.h>
49 49
50#ifdef CONFIG_X86_32
51
52/* Mappings between logical cpu number and node number */
53extern int cpu_to_node_map[];
54
55/* Returns the number of the node containing CPU 'cpu' */
56static inline int __cpu_to_node(int cpu)
57{
58 return cpu_to_node_map[cpu];
59}
60#define early_cpu_to_node __cpu_to_node
61#define cpu_to_node __cpu_to_node
62
63#else /* CONFIG_X86_64 */
64
65/* Mappings between logical cpu number and node number */ 50/* Mappings between logical cpu number and node number */
66DECLARE_EARLY_PER_CPU(int, x86_cpu_to_node_map); 51DECLARE_EARLY_PER_CPU(int, x86_cpu_to_node_map);
67 52
@@ -84,8 +69,6 @@ static inline int early_cpu_to_node(int cpu)
84 69
85#endif /* !CONFIG_DEBUG_PER_CPU_MAPS */ 70#endif /* !CONFIG_DEBUG_PER_CPU_MAPS */
86 71
87#endif /* CONFIG_X86_64 */
88
89/* Mappings between node number and cpus on that node. */ 72/* Mappings between node number and cpus on that node. */
90extern cpumask_var_t node_to_cpumask_map[MAX_NUMNODES]; 73extern cpumask_var_t node_to_cpumask_map[MAX_NUMNODES];
91 74
@@ -155,7 +138,7 @@ extern unsigned long node_remap_size[];
155 .balance_interval = 1, \ 138 .balance_interval = 1, \
156} 139}
157 140
158#ifdef CONFIG_X86_64_ACPI_NUMA 141#ifdef CONFIG_X86_64
159extern int __node_distance(int, int); 142extern int __node_distance(int, int);
160#define node_distance(a, b) __node_distance(a, b) 143#define node_distance(a, b) __node_distance(a, b)
161#endif 144#endif
diff --git a/arch/x86/include/asm/unistd_32.h b/arch/x86/include/asm/unistd_32.h
index b766a5e8ba0e..ffaf183c619a 100644
--- a/arch/x86/include/asm/unistd_32.h
+++ b/arch/x86/include/asm/unistd_32.h
@@ -346,10 +346,13 @@
346#define __NR_fanotify_init 338 346#define __NR_fanotify_init 338
347#define __NR_fanotify_mark 339 347#define __NR_fanotify_mark 339
348#define __NR_prlimit64 340 348#define __NR_prlimit64 340
349#define __NR_name_to_handle_at 341
350#define __NR_open_by_handle_at 342
351#define __NR_clock_adjtime 343
349 352
350#ifdef __KERNEL__ 353#ifdef __KERNEL__
351 354
352#define NR_syscalls 341 355#define NR_syscalls 344
353 356
354#define __ARCH_WANT_IPC_PARSE_VERSION 357#define __ARCH_WANT_IPC_PARSE_VERSION
355#define __ARCH_WANT_OLD_READDIR 358#define __ARCH_WANT_OLD_READDIR
diff --git a/arch/x86/include/asm/unistd_64.h b/arch/x86/include/asm/unistd_64.h
index 363e9b8a715b..5466bea670e7 100644
--- a/arch/x86/include/asm/unistd_64.h
+++ b/arch/x86/include/asm/unistd_64.h
@@ -669,6 +669,12 @@ __SYSCALL(__NR_fanotify_init, sys_fanotify_init)
669__SYSCALL(__NR_fanotify_mark, sys_fanotify_mark) 669__SYSCALL(__NR_fanotify_mark, sys_fanotify_mark)
670#define __NR_prlimit64 302 670#define __NR_prlimit64 302
671__SYSCALL(__NR_prlimit64, sys_prlimit64) 671__SYSCALL(__NR_prlimit64, sys_prlimit64)
672#define __NR_name_to_handle_at 303
673__SYSCALL(__NR_name_to_handle_at, sys_name_to_handle_at)
674#define __NR_open_by_handle_at 304
675__SYSCALL(__NR_open_by_handle_at, sys_open_by_handle_at)
676#define __NR_clock_adjtime 305
677__SYSCALL(__NR_clock_adjtime, sys_clock_adjtime)
672 678
673#ifndef __NO_STUBS 679#ifndef __NO_STUBS
674#define __ARCH_WANT_OLD_READDIR 680#define __ARCH_WANT_OLD_READDIR
diff --git a/arch/x86/include/asm/x86_init.h b/arch/x86/include/asm/x86_init.h
index 64642ad019fb..643ebf2e2ad8 100644
--- a/arch/x86/include/asm/x86_init.h
+++ b/arch/x86/include/asm/x86_init.h
@@ -83,11 +83,13 @@ struct x86_init_paging {
83 * boot cpu 83 * boot cpu
84 * @tsc_pre_init: platform function called before TSC init 84 * @tsc_pre_init: platform function called before TSC init
85 * @timer_init: initialize the platform timer (default PIT/HPET) 85 * @timer_init: initialize the platform timer (default PIT/HPET)
86 * @wallclock_init: init the wallclock device
86 */ 87 */
87struct x86_init_timers { 88struct x86_init_timers {
88 void (*setup_percpu_clockev)(void); 89 void (*setup_percpu_clockev)(void);
89 void (*tsc_pre_init)(void); 90 void (*tsc_pre_init)(void);
90 void (*timer_init)(void); 91 void (*timer_init)(void);
92 void (*wallclock_init)(void);
91}; 93};
92 94
93/** 95/**
diff --git a/arch/x86/include/asm/xen/hypercall.h b/arch/x86/include/asm/xen/hypercall.h
index a3c28ae4025b..8508bfe52296 100644
--- a/arch/x86/include/asm/xen/hypercall.h
+++ b/arch/x86/include/asm/xen/hypercall.h
@@ -287,7 +287,7 @@ HYPERVISOR_fpu_taskswitch(int set)
287static inline int 287static inline int
288HYPERVISOR_sched_op(int cmd, void *arg) 288HYPERVISOR_sched_op(int cmd, void *arg)
289{ 289{
290 return _hypercall2(int, sched_op_new, cmd, arg); 290 return _hypercall2(int, sched_op, cmd, arg);
291} 291}
292 292
293static inline long 293static inline long
@@ -422,10 +422,17 @@ HYPERVISOR_set_segment_base(int reg, unsigned long value)
422#endif 422#endif
423 423
424static inline int 424static inline int
425HYPERVISOR_suspend(unsigned long srec) 425HYPERVISOR_suspend(unsigned long start_info_mfn)
426{ 426{
427 return _hypercall3(int, sched_op, SCHEDOP_shutdown, 427 struct sched_shutdown r = { .reason = SHUTDOWN_suspend };
428 SHUTDOWN_suspend, srec); 428
429 /*
430 * For a PV guest the tools require that the start_info mfn be
431 * present in rdx/edx when the hypercall is made. Per the
432 * hypercall calling convention this is the third hypercall
433 * argument, which is start_info_mfn here.
434 */
435 return _hypercall3(int, sched_op, SCHEDOP_shutdown, &r, start_info_mfn);
429} 436}
430 437
431static inline int 438static inline int
diff --git a/arch/x86/include/asm/xen/page.h b/arch/x86/include/asm/xen/page.h
index f25bdf238a33..c61934fbf22a 100644
--- a/arch/x86/include/asm/xen/page.h
+++ b/arch/x86/include/asm/xen/page.h
@@ -29,8 +29,10 @@ typedef struct xpaddr {
29 29
30/**** MACHINE <-> PHYSICAL CONVERSION MACROS ****/ 30/**** MACHINE <-> PHYSICAL CONVERSION MACROS ****/
31#define INVALID_P2M_ENTRY (~0UL) 31#define INVALID_P2M_ENTRY (~0UL)
32#define FOREIGN_FRAME_BIT (1UL<<31) 32#define FOREIGN_FRAME_BIT (1UL<<(BITS_PER_LONG-1))
33#define IDENTITY_FRAME_BIT (1UL<<(BITS_PER_LONG-2))
33#define FOREIGN_FRAME(m) ((m) | FOREIGN_FRAME_BIT) 34#define FOREIGN_FRAME(m) ((m) | FOREIGN_FRAME_BIT)
35#define IDENTITY_FRAME(m) ((m) | IDENTITY_FRAME_BIT)
34 36
35/* Maximum amount of memory we can handle in a domain in pages */ 37/* Maximum amount of memory we can handle in a domain in pages */
36#define MAX_DOMAIN_PAGES \ 38#define MAX_DOMAIN_PAGES \
@@ -41,12 +43,18 @@ extern unsigned int machine_to_phys_order;
41 43
42extern unsigned long get_phys_to_machine(unsigned long pfn); 44extern unsigned long get_phys_to_machine(unsigned long pfn);
43extern bool set_phys_to_machine(unsigned long pfn, unsigned long mfn); 45extern bool set_phys_to_machine(unsigned long pfn, unsigned long mfn);
46extern bool __set_phys_to_machine(unsigned long pfn, unsigned long mfn);
47extern unsigned long set_phys_range_identity(unsigned long pfn_s,
48 unsigned long pfn_e);
44 49
45extern int m2p_add_override(unsigned long mfn, struct page *page); 50extern int m2p_add_override(unsigned long mfn, struct page *page);
46extern int m2p_remove_override(struct page *page); 51extern int m2p_remove_override(struct page *page);
47extern struct page *m2p_find_override(unsigned long mfn); 52extern struct page *m2p_find_override(unsigned long mfn);
48extern unsigned long m2p_find_override_pfn(unsigned long mfn, unsigned long pfn); 53extern unsigned long m2p_find_override_pfn(unsigned long mfn, unsigned long pfn);
49 54
55#ifdef CONFIG_XEN_DEBUG_FS
56extern int p2m_dump_show(struct seq_file *m, void *v);
57#endif
50static inline unsigned long pfn_to_mfn(unsigned long pfn) 58static inline unsigned long pfn_to_mfn(unsigned long pfn)
51{ 59{
52 unsigned long mfn; 60 unsigned long mfn;
@@ -57,7 +65,7 @@ static inline unsigned long pfn_to_mfn(unsigned long pfn)
57 mfn = get_phys_to_machine(pfn); 65 mfn = get_phys_to_machine(pfn);
58 66
59 if (mfn != INVALID_P2M_ENTRY) 67 if (mfn != INVALID_P2M_ENTRY)
60 mfn &= ~FOREIGN_FRAME_BIT; 68 mfn &= ~(FOREIGN_FRAME_BIT | IDENTITY_FRAME_BIT);
61 69
62 return mfn; 70 return mfn;
63} 71}
@@ -73,25 +81,44 @@ static inline int phys_to_machine_mapping_valid(unsigned long pfn)
73static inline unsigned long mfn_to_pfn(unsigned long mfn) 81static inline unsigned long mfn_to_pfn(unsigned long mfn)
74{ 82{
75 unsigned long pfn; 83 unsigned long pfn;
84 int ret = 0;
76 85
77 if (xen_feature(XENFEAT_auto_translated_physmap)) 86 if (xen_feature(XENFEAT_auto_translated_physmap))
78 return mfn; 87 return mfn;
79 88
89 if (unlikely((mfn >> machine_to_phys_order) != 0)) {
90 pfn = ~0;
91 goto try_override;
92 }
80 pfn = 0; 93 pfn = 0;
81 /* 94 /*
82 * The array access can fail (e.g., device space beyond end of RAM). 95 * The array access can fail (e.g., device space beyond end of RAM).
83 * In such cases it doesn't matter what we return (we return garbage), 96 * In such cases it doesn't matter what we return (we return garbage),
84 * but we must handle the fault without crashing! 97 * but we must handle the fault without crashing!
85 */ 98 */
86 __get_user(pfn, &machine_to_phys_mapping[mfn]); 99 ret = __get_user(pfn, &machine_to_phys_mapping[mfn]);
87 100try_override:
88 /* 101 /* ret might be < 0 if there are no entries in the m2p for mfn */
89 * If this appears to be a foreign mfn (because the pfn 102 if (ret < 0)
90 * doesn't map back to the mfn), then check the local override 103 pfn = ~0;
91 * table to see if there's a better pfn to use. 104 else if (get_phys_to_machine(pfn) != mfn)
105 /*
106 * If this appears to be a foreign mfn (because the pfn
107 * doesn't map back to the mfn), then check the local override
108 * table to see if there's a better pfn to use.
109 *
110 * m2p_find_override_pfn returns ~0 if it doesn't find anything.
111 */
112 pfn = m2p_find_override_pfn(mfn, ~0);
113
114 /*
115 * pfn is ~0 if there are no entries in the m2p for mfn or if the
116 * entry doesn't map back to the mfn and m2p_override doesn't have a
117 * valid entry for it.
92 */ 118 */
93 if (get_phys_to_machine(pfn) != mfn) 119 if (pfn == ~0 &&
94 pfn = m2p_find_override_pfn(mfn, pfn); 120 get_phys_to_machine(mfn) == IDENTITY_FRAME(mfn))
121 pfn = mfn;
95 122
96 return pfn; 123 return pfn;
97} 124}
diff --git a/arch/x86/include/asm/xen/pci.h b/arch/x86/include/asm/xen/pci.h
index 2329b3eaf8d3..aa8620989162 100644
--- a/arch/x86/include/asm/xen/pci.h
+++ b/arch/x86/include/asm/xen/pci.h
@@ -27,16 +27,16 @@ static inline void __init xen_setup_pirqs(void)
27 * its own functions. 27 * its own functions.
28 */ 28 */
29struct xen_pci_frontend_ops { 29struct xen_pci_frontend_ops {
30 int (*enable_msi)(struct pci_dev *dev, int **vectors); 30 int (*enable_msi)(struct pci_dev *dev, int vectors[]);
31 void (*disable_msi)(struct pci_dev *dev); 31 void (*disable_msi)(struct pci_dev *dev);
32 int (*enable_msix)(struct pci_dev *dev, int **vectors, int nvec); 32 int (*enable_msix)(struct pci_dev *dev, int vectors[], int nvec);
33 void (*disable_msix)(struct pci_dev *dev); 33 void (*disable_msix)(struct pci_dev *dev);
34}; 34};
35 35
36extern struct xen_pci_frontend_ops *xen_pci_frontend; 36extern struct xen_pci_frontend_ops *xen_pci_frontend;
37 37
38static inline int xen_pci_frontend_enable_msi(struct pci_dev *dev, 38static inline int xen_pci_frontend_enable_msi(struct pci_dev *dev,
39 int **vectors) 39 int vectors[])
40{ 40{
41 if (xen_pci_frontend && xen_pci_frontend->enable_msi) 41 if (xen_pci_frontend && xen_pci_frontend->enable_msi)
42 return xen_pci_frontend->enable_msi(dev, vectors); 42 return xen_pci_frontend->enable_msi(dev, vectors);
@@ -48,7 +48,7 @@ static inline void xen_pci_frontend_disable_msi(struct pci_dev *dev)
48 xen_pci_frontend->disable_msi(dev); 48 xen_pci_frontend->disable_msi(dev);
49} 49}
50static inline int xen_pci_frontend_enable_msix(struct pci_dev *dev, 50static inline int xen_pci_frontend_enable_msix(struct pci_dev *dev,
51 int **vectors, int nvec) 51 int vectors[], int nvec)
52{ 52{
53 if (xen_pci_frontend && xen_pci_frontend->enable_msix) 53 if (xen_pci_frontend && xen_pci_frontend->enable_msix)
54 return xen_pci_frontend->enable_msix(dev, vectors, nvec); 54 return xen_pci_frontend->enable_msix(dev, vectors, nvec);
diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile
index 34244b2cd880..62445ba2f8a8 100644
--- a/arch/x86/kernel/Makefile
+++ b/arch/x86/kernel/Makefile
@@ -66,9 +66,9 @@ obj-$(CONFIG_PCI) += early-quirks.o
66apm-y := apm_32.o 66apm-y := apm_32.o
67obj-$(CONFIG_APM) += apm.o 67obj-$(CONFIG_APM) += apm.o
68obj-$(CONFIG_SMP) += smp.o 68obj-$(CONFIG_SMP) += smp.o
69obj-$(CONFIG_SMP) += smpboot.o tsc_sync.o 69obj-$(CONFIG_SMP) += smpboot.o
70obj-$(CONFIG_SMP) += tsc_sync.o
70obj-$(CONFIG_SMP) += setup_percpu.o 71obj-$(CONFIG_SMP) += setup_percpu.o
71obj-$(CONFIG_X86_64_SMP) += tsc_sync.o
72obj-$(CONFIG_X86_TRAMPOLINE) += trampoline_$(BITS).o 72obj-$(CONFIG_X86_TRAMPOLINE) += trampoline_$(BITS).o
73obj-$(CONFIG_X86_MPPARSE) += mpparse.o 73obj-$(CONFIG_X86_MPPARSE) += mpparse.o
74obj-y += apic/ 74obj-y += apic/
@@ -109,6 +109,7 @@ obj-$(CONFIG_MICROCODE) += microcode.o
109obj-$(CONFIG_X86_CHECK_BIOS_CORRUPTION) += check.o 109obj-$(CONFIG_X86_CHECK_BIOS_CORRUPTION) += check.o
110 110
111obj-$(CONFIG_SWIOTLB) += pci-swiotlb.o 111obj-$(CONFIG_SWIOTLB) += pci-swiotlb.o
112obj-$(CONFIG_OF) += devicetree.o
112 113
113### 114###
114# 64 bit specific files 115# 64 bit specific files
diff --git a/arch/x86/kernel/acpi/boot.c b/arch/x86/kernel/acpi/boot.c
index 3e6e2d68f761..9a966c579af5 100644
--- a/arch/x86/kernel/acpi/boot.c
+++ b/arch/x86/kernel/acpi/boot.c
@@ -595,14 +595,8 @@ static void acpi_map_cpu2node(acpi_handle handle, int cpu, int physid)
595 nid = acpi_get_node(handle); 595 nid = acpi_get_node(handle);
596 if (nid == -1 || !node_online(nid)) 596 if (nid == -1 || !node_online(nid))
597 return; 597 return;
598#ifdef CONFIG_X86_64 598 set_apicid_to_node(physid, nid);
599 apicid_to_node[physid] = nid;
600 numa_set_node(cpu, nid); 599 numa_set_node(cpu, nid);
601#else /* CONFIG_X86_32 */
602 apicid_2_node[physid] = nid;
603 cpu_to_node_map[cpu] = nid;
604#endif
605
606#endif 600#endif
607} 601}
608 602
diff --git a/arch/x86/kernel/apb_timer.c b/arch/x86/kernel/apb_timer.c
index 51d4e1663066..1293c709ee85 100644
--- a/arch/x86/kernel/apb_timer.c
+++ b/arch/x86/kernel/apb_timer.c
@@ -508,64 +508,12 @@ static int apbt_next_event(unsigned long delta,
508 return 0; 508 return 0;
509} 509}
510 510
511/*
512 * APB timer clock is not in sync with pclk on Langwell, which translates to
513 * unreliable read value caused by sampling error. the error does not add up
514 * overtime and only happens when sampling a 0 as a 1 by mistake. so the time
515 * would go backwards. the following code is trying to prevent time traveling
516 * backwards. little bit paranoid.
517 */
518static cycle_t apbt_read_clocksource(struct clocksource *cs) 511static cycle_t apbt_read_clocksource(struct clocksource *cs)
519{ 512{
520 unsigned long t0, t1, t2; 513 unsigned long current_count;
521 static unsigned long last_read; 514
522 515 current_count = apbt_readl(phy_cs_timer_id, APBTMR_N_CURRENT_VALUE);
523bad_count: 516 return (cycle_t)~current_count;
524 t1 = apbt_readl(phy_cs_timer_id,
525 APBTMR_N_CURRENT_VALUE);
526 t2 = apbt_readl(phy_cs_timer_id,
527 APBTMR_N_CURRENT_VALUE);
528 if (unlikely(t1 < t2)) {
529 pr_debug("APBT: read current count error %lx:%lx:%lx\n",
530 t1, t2, t2 - t1);
531 goto bad_count;
532 }
533 /*
534 * check against cached last read, makes sure time does not go back.
535 * it could be a normal rollover but we will do tripple check anyway
536 */
537 if (unlikely(t2 > last_read)) {
538 /* check if we have a normal rollover */
539 unsigned long raw_intr_status =
540 apbt_readl_reg(APBTMRS_RAW_INT_STATUS);
541 /*
542 * cs timer interrupt is masked but raw intr bit is set if
543 * rollover occurs. then we read EOI reg to clear it.
544 */
545 if (raw_intr_status & (1 << phy_cs_timer_id)) {
546 apbt_readl(phy_cs_timer_id, APBTMR_N_EOI);
547 goto out;
548 }
549 pr_debug("APB CS going back %lx:%lx:%lx ",
550 t2, last_read, t2 - last_read);
551bad_count_x3:
552 pr_debug("triple check enforced\n");
553 t0 = apbt_readl(phy_cs_timer_id,
554 APBTMR_N_CURRENT_VALUE);
555 udelay(1);
556 t1 = apbt_readl(phy_cs_timer_id,
557 APBTMR_N_CURRENT_VALUE);
558 udelay(1);
559 t2 = apbt_readl(phy_cs_timer_id,
560 APBTMR_N_CURRENT_VALUE);
561 if ((t2 > t1) || (t1 > t0)) {
562 printk(KERN_ERR "Error: APB CS tripple check failed\n");
563 goto bad_count_x3;
564 }
565 }
566out:
567 last_read = t2;
568 return (cycle_t)~t2;
569} 517}
570 518
571static int apbt_clocksource_register(void) 519static int apbt_clocksource_register(void)
diff --git a/arch/x86/kernel/aperture_64.c b/arch/x86/kernel/aperture_64.c
index 5955a7800a96..7b1e8e10b89c 100644
--- a/arch/x86/kernel/aperture_64.c
+++ b/arch/x86/kernel/aperture_64.c
@@ -13,7 +13,7 @@
13#include <linux/kernel.h> 13#include <linux/kernel.h>
14#include <linux/types.h> 14#include <linux/types.h>
15#include <linux/init.h> 15#include <linux/init.h>
16#include <linux/bootmem.h> 16#include <linux/memblock.h>
17#include <linux/mmzone.h> 17#include <linux/mmzone.h>
18#include <linux/pci_ids.h> 18#include <linux/pci_ids.h>
19#include <linux/pci.h> 19#include <linux/pci.h>
@@ -57,7 +57,7 @@ static void __init insert_aperture_resource(u32 aper_base, u32 aper_size)
57static u32 __init allocate_aperture(void) 57static u32 __init allocate_aperture(void)
58{ 58{
59 u32 aper_size; 59 u32 aper_size;
60 void *p; 60 unsigned long addr;
61 61
62 /* aper_size should <= 1G */ 62 /* aper_size should <= 1G */
63 if (fallback_aper_order > 5) 63 if (fallback_aper_order > 5)
@@ -83,27 +83,26 @@ static u32 __init allocate_aperture(void)
83 * so don't use 512M below as gart iommu, leave the space for kernel 83 * so don't use 512M below as gart iommu, leave the space for kernel
84 * code for safe 84 * code for safe
85 */ 85 */
86 p = __alloc_bootmem_nopanic(aper_size, aper_size, 512ULL<<20); 86 addr = memblock_find_in_range(0, 1ULL<<32, aper_size, 512ULL<<20);
87 if (addr == MEMBLOCK_ERROR || addr + aper_size > 0xffffffff) {
88 printk(KERN_ERR
89 "Cannot allocate aperture memory hole (%lx,%uK)\n",
90 addr, aper_size>>10);
91 return 0;
92 }
93 memblock_x86_reserve_range(addr, addr + aper_size, "aperture64");
87 /* 94 /*
88 * Kmemleak should not scan this block as it may not be mapped via the 95 * Kmemleak should not scan this block as it may not be mapped via the
89 * kernel direct mapping. 96 * kernel direct mapping.
90 */ 97 */
91 kmemleak_ignore(p); 98 kmemleak_ignore(phys_to_virt(addr));
92 if (!p || __pa(p)+aper_size > 0xffffffff) {
93 printk(KERN_ERR
94 "Cannot allocate aperture memory hole (%p,%uK)\n",
95 p, aper_size>>10);
96 if (p)
97 free_bootmem(__pa(p), aper_size);
98 return 0;
99 }
100 printk(KERN_INFO "Mapping aperture over %d KB of RAM @ %lx\n", 99 printk(KERN_INFO "Mapping aperture over %d KB of RAM @ %lx\n",
101 aper_size >> 10, __pa(p)); 100 aper_size >> 10, addr);
102 insert_aperture_resource((u32)__pa(p), aper_size); 101 insert_aperture_resource((u32)addr, aper_size);
103 register_nosave_region((u32)__pa(p) >> PAGE_SHIFT, 102 register_nosave_region(addr >> PAGE_SHIFT,
104 (u32)__pa(p+aper_size) >> PAGE_SHIFT); 103 (addr+aper_size) >> PAGE_SHIFT);
105 104
106 return (u32)__pa(p); 105 return (u32)addr;
107} 106}
108 107
109 108
diff --git a/arch/x86/kernel/apic/apic.c b/arch/x86/kernel/apic/apic.c
index 76b96d74978a..966673f44141 100644
--- a/arch/x86/kernel/apic/apic.c
+++ b/arch/x86/kernel/apic/apic.c
@@ -43,6 +43,7 @@
43#include <asm/i8259.h> 43#include <asm/i8259.h>
44#include <asm/proto.h> 44#include <asm/proto.h>
45#include <asm/apic.h> 45#include <asm/apic.h>
46#include <asm/io_apic.h>
46#include <asm/desc.h> 47#include <asm/desc.h>
47#include <asm/hpet.h> 48#include <asm/hpet.h>
48#include <asm/idle.h> 49#include <asm/idle.h>
@@ -78,12 +79,21 @@ EXPORT_EARLY_PER_CPU_SYMBOL(x86_cpu_to_apicid);
78EXPORT_EARLY_PER_CPU_SYMBOL(x86_bios_cpu_apicid); 79EXPORT_EARLY_PER_CPU_SYMBOL(x86_bios_cpu_apicid);
79 80
80#ifdef CONFIG_X86_32 81#ifdef CONFIG_X86_32
82
83/*
84 * On x86_32, the mapping between cpu and logical apicid may vary
85 * depending on apic in use. The following early percpu variable is
86 * used for the mapping. This is where the behaviors of x86_64 and 32
87 * actually diverge. Let's keep it ugly for now.
88 */
89DEFINE_EARLY_PER_CPU(int, x86_cpu_to_logical_apicid, BAD_APICID);
90
81/* 91/*
82 * Knob to control our willingness to enable the local APIC. 92 * Knob to control our willingness to enable the local APIC.
83 * 93 *
84 * +1=force-enable 94 * +1=force-enable
85 */ 95 */
86static int force_enable_local_apic; 96static int force_enable_local_apic __initdata;
87/* 97/*
88 * APIC command line parameters 98 * APIC command line parameters
89 */ 99 */
@@ -153,7 +163,7 @@ early_param("nox2apic", setup_nox2apic);
153unsigned long mp_lapic_addr; 163unsigned long mp_lapic_addr;
154int disable_apic; 164int disable_apic;
155/* Disable local APIC timer from the kernel commandline or via dmi quirk */ 165/* Disable local APIC timer from the kernel commandline or via dmi quirk */
156static int disable_apic_timer __cpuinitdata; 166static int disable_apic_timer __initdata;
157/* Local APIC timer works in C2 */ 167/* Local APIC timer works in C2 */
158int local_apic_timer_c2_ok; 168int local_apic_timer_c2_ok;
159EXPORT_SYMBOL_GPL(local_apic_timer_c2_ok); 169EXPORT_SYMBOL_GPL(local_apic_timer_c2_ok);
@@ -177,29 +187,8 @@ static struct resource lapic_resource = {
177 187
178static unsigned int calibration_result; 188static unsigned int calibration_result;
179 189
180static int lapic_next_event(unsigned long delta,
181 struct clock_event_device *evt);
182static void lapic_timer_setup(enum clock_event_mode mode,
183 struct clock_event_device *evt);
184static void lapic_timer_broadcast(const struct cpumask *mask);
185static void apic_pm_activate(void); 190static void apic_pm_activate(void);
186 191
187/*
188 * The local apic timer can be used for any function which is CPU local.
189 */
190static struct clock_event_device lapic_clockevent = {
191 .name = "lapic",
192 .features = CLOCK_EVT_FEAT_PERIODIC | CLOCK_EVT_FEAT_ONESHOT
193 | CLOCK_EVT_FEAT_C3STOP | CLOCK_EVT_FEAT_DUMMY,
194 .shift = 32,
195 .set_mode = lapic_timer_setup,
196 .set_next_event = lapic_next_event,
197 .broadcast = lapic_timer_broadcast,
198 .rating = 100,
199 .irq = -1,
200};
201static DEFINE_PER_CPU(struct clock_event_device, lapic_events);
202
203static unsigned long apic_phys; 192static unsigned long apic_phys;
204 193
205/* 194/*
@@ -238,7 +227,7 @@ static int modern_apic(void)
238 * right after this call apic become NOOP driven 227 * right after this call apic become NOOP driven
239 * so apic->write/read doesn't do anything 228 * so apic->write/read doesn't do anything
240 */ 229 */
241void apic_disable(void) 230static void __init apic_disable(void)
242{ 231{
243 pr_info("APIC: switched to apic NOOP\n"); 232 pr_info("APIC: switched to apic NOOP\n");
244 apic = &apic_noop; 233 apic = &apic_noop;
@@ -282,23 +271,6 @@ u64 native_apic_icr_read(void)
282 return icr1 | ((u64)icr2 << 32); 271 return icr1 | ((u64)icr2 << 32);
283} 272}
284 273
285/**
286 * enable_NMI_through_LVT0 - enable NMI through local vector table 0
287 */
288void __cpuinit enable_NMI_through_LVT0(void)
289{
290 unsigned int v;
291
292 /* unmask and set to NMI */
293 v = APIC_DM_NMI;
294
295 /* Level triggered for 82489DX (32bit mode) */
296 if (!lapic_is_integrated())
297 v |= APIC_LVT_LEVEL_TRIGGER;
298
299 apic_write(APIC_LVT0, v);
300}
301
302#ifdef CONFIG_X86_32 274#ifdef CONFIG_X86_32
303/** 275/**
304 * get_physical_broadcast - Get number of physical broadcast IDs 276 * get_physical_broadcast - Get number of physical broadcast IDs
@@ -508,6 +480,23 @@ static void lapic_timer_broadcast(const struct cpumask *mask)
508#endif 480#endif
509} 481}
510 482
483
484/*
485 * The local apic timer can be used for any function which is CPU local.
486 */
487static struct clock_event_device lapic_clockevent = {
488 .name = "lapic",
489 .features = CLOCK_EVT_FEAT_PERIODIC | CLOCK_EVT_FEAT_ONESHOT
490 | CLOCK_EVT_FEAT_C3STOP | CLOCK_EVT_FEAT_DUMMY,
491 .shift = 32,
492 .set_mode = lapic_timer_setup,
493 .set_next_event = lapic_next_event,
494 .broadcast = lapic_timer_broadcast,
495 .rating = 100,
496 .irq = -1,
497};
498static DEFINE_PER_CPU(struct clock_event_device, lapic_events);
499
511/* 500/*
512 * Setup the local APIC timer for this CPU. Copy the initialized values 501 * Setup the local APIC timer for this CPU. Copy the initialized values
513 * of the boot CPU and register the clock event in the framework. 502 * of the boot CPU and register the clock event in the framework.
@@ -1209,7 +1198,7 @@ void __cpuinit setup_local_APIC(void)
1209 rdtscll(tsc); 1198 rdtscll(tsc);
1210 1199
1211 if (disable_apic) { 1200 if (disable_apic) {
1212 arch_disable_smp_support(); 1201 disable_ioapic_support();
1213 return; 1202 return;
1214 } 1203 }
1215 1204
@@ -1237,6 +1226,19 @@ void __cpuinit setup_local_APIC(void)
1237 */ 1226 */
1238 apic->init_apic_ldr(); 1227 apic->init_apic_ldr();
1239 1228
1229#ifdef CONFIG_X86_32
1230 /*
1231 * APIC LDR is initialized. If logical_apicid mapping was
1232 * initialized during get_smp_config(), make sure it matches the
1233 * actual value.
1234 */
1235 i = early_per_cpu(x86_cpu_to_logical_apicid, cpu);
1236 WARN_ON(i != BAD_APICID && i != logical_smp_processor_id());
1237 /* always use the value from LDR */
1238 early_per_cpu(x86_cpu_to_logical_apicid, cpu) =
1239 logical_smp_processor_id();
1240#endif
1241
1240 /* 1242 /*
1241 * Set Task Priority to 'accept all'. We never change this 1243 * Set Task Priority to 'accept all'. We never change this
1242 * later on. 1244 * later on.
@@ -1448,7 +1450,7 @@ int __init enable_IR(void)
1448void __init enable_IR_x2apic(void) 1450void __init enable_IR_x2apic(void)
1449{ 1451{
1450 unsigned long flags; 1452 unsigned long flags;
1451 struct IO_APIC_route_entry **ioapic_entries = NULL; 1453 struct IO_APIC_route_entry **ioapic_entries;
1452 int ret, x2apic_enabled = 0; 1454 int ret, x2apic_enabled = 0;
1453 int dmar_table_init_ret; 1455 int dmar_table_init_ret;
1454 1456
@@ -1537,7 +1539,7 @@ static int __init detect_init_APIC(void)
1537} 1539}
1538#else 1540#else
1539 1541
1540static int apic_verify(void) 1542static int __init apic_verify(void)
1541{ 1543{
1542 u32 features, h, l; 1544 u32 features, h, l;
1543 1545
@@ -1562,7 +1564,7 @@ static int apic_verify(void)
1562 return 0; 1564 return 0;
1563} 1565}
1564 1566
1565int apic_force_enable(void) 1567int __init apic_force_enable(unsigned long addr)
1566{ 1568{
1567 u32 h, l; 1569 u32 h, l;
1568 1570
@@ -1578,7 +1580,7 @@ int apic_force_enable(void)
1578 if (!(l & MSR_IA32_APICBASE_ENABLE)) { 1580 if (!(l & MSR_IA32_APICBASE_ENABLE)) {
1579 pr_info("Local APIC disabled by BIOS -- reenabling.\n"); 1581 pr_info("Local APIC disabled by BIOS -- reenabling.\n");
1580 l &= ~MSR_IA32_APICBASE_BASE; 1582 l &= ~MSR_IA32_APICBASE_BASE;
1581 l |= MSR_IA32_APICBASE_ENABLE | APIC_DEFAULT_PHYS_BASE; 1583 l |= MSR_IA32_APICBASE_ENABLE | addr;
1582 wrmsr(MSR_IA32_APICBASE, l, h); 1584 wrmsr(MSR_IA32_APICBASE, l, h);
1583 enabled_via_apicbase = 1; 1585 enabled_via_apicbase = 1;
1584 } 1586 }
@@ -1619,7 +1621,7 @@ static int __init detect_init_APIC(void)
1619 "you can enable it with \"lapic\"\n"); 1621 "you can enable it with \"lapic\"\n");
1620 return -1; 1622 return -1;
1621 } 1623 }
1622 if (apic_force_enable()) 1624 if (apic_force_enable(APIC_DEFAULT_PHYS_BASE))
1623 return -1; 1625 return -1;
1624 } else { 1626 } else {
1625 if (apic_verify()) 1627 if (apic_verify())
@@ -1930,17 +1932,6 @@ void __cpuinit generic_processor_info(int apicid, int version)
1930{ 1932{
1931 int cpu; 1933 int cpu;
1932 1934
1933 /*
1934 * Validate version
1935 */
1936 if (version == 0x0) {
1937 pr_warning("BIOS bug, APIC version is 0 for CPU#%d! "
1938 "fixing up to 0x10. (tell your hw vendor)\n",
1939 version);
1940 version = 0x10;
1941 }
1942 apic_version[apicid] = version;
1943
1944 if (num_processors >= nr_cpu_ids) { 1935 if (num_processors >= nr_cpu_ids) {
1945 int max = nr_cpu_ids; 1936 int max = nr_cpu_ids;
1946 int thiscpu = max + disabled_cpus; 1937 int thiscpu = max + disabled_cpus;
@@ -1954,22 +1945,34 @@ void __cpuinit generic_processor_info(int apicid, int version)
1954 } 1945 }
1955 1946
1956 num_processors++; 1947 num_processors++;
1957 cpu = cpumask_next_zero(-1, cpu_present_mask);
1958
1959 if (version != apic_version[boot_cpu_physical_apicid])
1960 WARN_ONCE(1,
1961 "ACPI: apic version mismatch, bootcpu: %x cpu %d: %x\n",
1962 apic_version[boot_cpu_physical_apicid], cpu, version);
1963
1964 physid_set(apicid, phys_cpu_present_map);
1965 if (apicid == boot_cpu_physical_apicid) { 1948 if (apicid == boot_cpu_physical_apicid) {
1966 /* 1949 /*
1967 * x86_bios_cpu_apicid is required to have processors listed 1950 * x86_bios_cpu_apicid is required to have processors listed
1968 * in same order as logical cpu numbers. Hence the first 1951 * in same order as logical cpu numbers. Hence the first
1969 * entry is BSP, and so on. 1952 * entry is BSP, and so on.
1953 * boot_cpu_init() already hold bit 0 in cpu_present_mask
1954 * for BSP.
1970 */ 1955 */
1971 cpu = 0; 1956 cpu = 0;
1957 } else
1958 cpu = cpumask_next_zero(-1, cpu_present_mask);
1959
1960 /*
1961 * Validate version
1962 */
1963 if (version == 0x0) {
1964 pr_warning("BIOS bug: APIC version is 0 for CPU %d/0x%x, fixing up to 0x10\n",
1965 cpu, apicid);
1966 version = 0x10;
1972 } 1967 }
1968 apic_version[apicid] = version;
1969
1970 if (version != apic_version[boot_cpu_physical_apicid]) {
1971 pr_warning("BIOS bug: APIC version mismatch, boot CPU: %x, CPU %d: version %x\n",
1972 apic_version[boot_cpu_physical_apicid], cpu, version);
1973 }
1974
1975 physid_set(apicid, phys_cpu_present_map);
1973 if (apicid > max_physical_apicid) 1976 if (apicid > max_physical_apicid)
1974 max_physical_apicid = apicid; 1977 max_physical_apicid = apicid;
1975 1978
@@ -1977,7 +1980,10 @@ void __cpuinit generic_processor_info(int apicid, int version)
1977 early_per_cpu(x86_cpu_to_apicid, cpu) = apicid; 1980 early_per_cpu(x86_cpu_to_apicid, cpu) = apicid;
1978 early_per_cpu(x86_bios_cpu_apicid, cpu) = apicid; 1981 early_per_cpu(x86_bios_cpu_apicid, cpu) = apicid;
1979#endif 1982#endif
1980 1983#ifdef CONFIG_X86_32
1984 early_per_cpu(x86_cpu_to_logical_apicid, cpu) =
1985 apic->x86_32_early_logical_apicid(cpu);
1986#endif
1981 set_cpu_possible(cpu, true); 1987 set_cpu_possible(cpu, true);
1982 set_cpu_present(cpu, true); 1988 set_cpu_present(cpu, true);
1983} 1989}
@@ -1998,10 +2004,14 @@ void default_init_apic_ldr(void)
1998} 2004}
1999 2005
2000#ifdef CONFIG_X86_32 2006#ifdef CONFIG_X86_32
2001int default_apicid_to_node(int logical_apicid) 2007int default_x86_32_numa_cpu_node(int cpu)
2002{ 2008{
2003#ifdef CONFIG_SMP 2009#ifdef CONFIG_NUMA
2004 return apicid_2_node[hard_smp_processor_id()]; 2010 int apicid = early_per_cpu(x86_cpu_to_apicid, cpu);
2011
2012 if (apicid != BAD_APICID)
2013 return __apicid_to_node[apicid];
2014 return NUMA_NO_NODE;
2005#else 2015#else
2006 return 0; 2016 return 0;
2007#endif 2017#endif
diff --git a/arch/x86/kernel/apic/apic_flat_64.c b/arch/x86/kernel/apic/apic_flat_64.c
index 09d3b17ce0c2..5652d31fe108 100644
--- a/arch/x86/kernel/apic/apic_flat_64.c
+++ b/arch/x86/kernel/apic/apic_flat_64.c
@@ -185,8 +185,6 @@ struct apic apic_flat = {
185 .ioapic_phys_id_map = NULL, 185 .ioapic_phys_id_map = NULL,
186 .setup_apic_routing = NULL, 186 .setup_apic_routing = NULL,
187 .multi_timer_check = NULL, 187 .multi_timer_check = NULL,
188 .apicid_to_node = NULL,
189 .cpu_to_logical_apicid = NULL,
190 .cpu_present_to_apicid = default_cpu_present_to_apicid, 188 .cpu_present_to_apicid = default_cpu_present_to_apicid,
191 .apicid_to_cpu_present = NULL, 189 .apicid_to_cpu_present = NULL,
192 .setup_portio_remap = NULL, 190 .setup_portio_remap = NULL,
@@ -337,8 +335,6 @@ struct apic apic_physflat = {
337 .ioapic_phys_id_map = NULL, 335 .ioapic_phys_id_map = NULL,
338 .setup_apic_routing = NULL, 336 .setup_apic_routing = NULL,
339 .multi_timer_check = NULL, 337 .multi_timer_check = NULL,
340 .apicid_to_node = NULL,
341 .cpu_to_logical_apicid = NULL,
342 .cpu_present_to_apicid = default_cpu_present_to_apicid, 338 .cpu_present_to_apicid = default_cpu_present_to_apicid,
343 .apicid_to_cpu_present = NULL, 339 .apicid_to_cpu_present = NULL,
344 .setup_portio_remap = NULL, 340 .setup_portio_remap = NULL,
diff --git a/arch/x86/kernel/apic/apic_noop.c b/arch/x86/kernel/apic/apic_noop.c
index e31b9ffe25f5..f1baa2dc087a 100644
--- a/arch/x86/kernel/apic/apic_noop.c
+++ b/arch/x86/kernel/apic/apic_noop.c
@@ -54,11 +54,6 @@ static u64 noop_apic_icr_read(void)
54 return 0; 54 return 0;
55} 55}
56 56
57static int noop_cpu_to_logical_apicid(int cpu)
58{
59 return 0;
60}
61
62static int noop_phys_pkg_id(int cpuid_apic, int index_msb) 57static int noop_phys_pkg_id(int cpuid_apic, int index_msb)
63{ 58{
64 return 0; 59 return 0;
@@ -113,12 +108,6 @@ static void noop_vector_allocation_domain(int cpu, struct cpumask *retmask)
113 cpumask_set_cpu(cpu, retmask); 108 cpumask_set_cpu(cpu, retmask);
114} 109}
115 110
116int noop_apicid_to_node(int logical_apicid)
117{
118 /* we're always on node 0 */
119 return 0;
120}
121
122static u32 noop_apic_read(u32 reg) 111static u32 noop_apic_read(u32 reg)
123{ 112{
124 WARN_ON_ONCE((cpu_has_apic && !disable_apic)); 113 WARN_ON_ONCE((cpu_has_apic && !disable_apic));
@@ -130,6 +119,14 @@ static void noop_apic_write(u32 reg, u32 v)
130 WARN_ON_ONCE(cpu_has_apic && !disable_apic); 119 WARN_ON_ONCE(cpu_has_apic && !disable_apic);
131} 120}
132 121
122#ifdef CONFIG_X86_32
123static int noop_x86_32_numa_cpu_node(int cpu)
124{
125 /* we're always on node 0 */
126 return 0;
127}
128#endif
129
133struct apic apic_noop = { 130struct apic apic_noop = {
134 .name = "noop", 131 .name = "noop",
135 .probe = noop_probe, 132 .probe = noop_probe,
@@ -153,9 +150,7 @@ struct apic apic_noop = {
153 .ioapic_phys_id_map = default_ioapic_phys_id_map, 150 .ioapic_phys_id_map = default_ioapic_phys_id_map,
154 .setup_apic_routing = NULL, 151 .setup_apic_routing = NULL,
155 .multi_timer_check = NULL, 152 .multi_timer_check = NULL,
156 .apicid_to_node = noop_apicid_to_node,
157 153
158 .cpu_to_logical_apicid = noop_cpu_to_logical_apicid,
159 .cpu_present_to_apicid = default_cpu_present_to_apicid, 154 .cpu_present_to_apicid = default_cpu_present_to_apicid,
160 .apicid_to_cpu_present = physid_set_mask_of_physid, 155 .apicid_to_cpu_present = physid_set_mask_of_physid,
161 156
@@ -197,4 +192,9 @@ struct apic apic_noop = {
197 .icr_write = noop_apic_icr_write, 192 .icr_write = noop_apic_icr_write,
198 .wait_icr_idle = noop_apic_wait_icr_idle, 193 .wait_icr_idle = noop_apic_wait_icr_idle,
199 .safe_wait_icr_idle = noop_safe_apic_wait_icr_idle, 194 .safe_wait_icr_idle = noop_safe_apic_wait_icr_idle,
195
196#ifdef CONFIG_X86_32
197 .x86_32_early_logical_apicid = noop_x86_32_early_logical_apicid,
198 .x86_32_numa_cpu_node = noop_x86_32_numa_cpu_node,
199#endif
200}; 200};
diff --git a/arch/x86/kernel/apic/bigsmp_32.c b/arch/x86/kernel/apic/bigsmp_32.c
index cb804c5091b9..541a2e431659 100644
--- a/arch/x86/kernel/apic/bigsmp_32.c
+++ b/arch/x86/kernel/apic/bigsmp_32.c
@@ -45,6 +45,12 @@ static unsigned long bigsmp_check_apicid_present(int bit)
45 return 1; 45 return 1;
46} 46}
47 47
48static int bigsmp_early_logical_apicid(int cpu)
49{
50 /* on bigsmp, logical apicid is the same as physical */
51 return early_per_cpu(x86_cpu_to_apicid, cpu);
52}
53
48static inline unsigned long calculate_ldr(int cpu) 54static inline unsigned long calculate_ldr(int cpu)
49{ 55{
50 unsigned long val, id; 56 unsigned long val, id;
@@ -80,11 +86,6 @@ static void bigsmp_setup_apic_routing(void)
80 nr_ioapics); 86 nr_ioapics);
81} 87}
82 88
83static int bigsmp_apicid_to_node(int logical_apicid)
84{
85 return apicid_2_node[hard_smp_processor_id()];
86}
87
88static int bigsmp_cpu_present_to_apicid(int mps_cpu) 89static int bigsmp_cpu_present_to_apicid(int mps_cpu)
89{ 90{
90 if (mps_cpu < nr_cpu_ids) 91 if (mps_cpu < nr_cpu_ids)
@@ -93,14 +94,6 @@ static int bigsmp_cpu_present_to_apicid(int mps_cpu)
93 return BAD_APICID; 94 return BAD_APICID;
94} 95}
95 96
96/* Mapping from cpu number to logical apicid */
97static inline int bigsmp_cpu_to_logical_apicid(int cpu)
98{
99 if (cpu >= nr_cpu_ids)
100 return BAD_APICID;
101 return cpu_physical_id(cpu);
102}
103
104static void bigsmp_ioapic_phys_id_map(physid_mask_t *phys_map, physid_mask_t *retmap) 97static void bigsmp_ioapic_phys_id_map(physid_mask_t *phys_map, physid_mask_t *retmap)
105{ 98{
106 /* For clustered we don't have a good way to do this yet - hack */ 99 /* For clustered we don't have a good way to do this yet - hack */
@@ -115,7 +108,11 @@ static int bigsmp_check_phys_apicid_present(int phys_apicid)
115/* As we are using single CPU as destination, pick only one CPU here */ 108/* As we are using single CPU as destination, pick only one CPU here */
116static unsigned int bigsmp_cpu_mask_to_apicid(const struct cpumask *cpumask) 109static unsigned int bigsmp_cpu_mask_to_apicid(const struct cpumask *cpumask)
117{ 110{
118 return bigsmp_cpu_to_logical_apicid(cpumask_first(cpumask)); 111 int cpu = cpumask_first(cpumask);
112
113 if (cpu < nr_cpu_ids)
114 return cpu_physical_id(cpu);
115 return BAD_APICID;
119} 116}
120 117
121static unsigned int bigsmp_cpu_mask_to_apicid_and(const struct cpumask *cpumask, 118static unsigned int bigsmp_cpu_mask_to_apicid_and(const struct cpumask *cpumask,
@@ -129,9 +126,9 @@ static unsigned int bigsmp_cpu_mask_to_apicid_and(const struct cpumask *cpumask,
129 */ 126 */
130 for_each_cpu_and(cpu, cpumask, andmask) { 127 for_each_cpu_and(cpu, cpumask, andmask) {
131 if (cpumask_test_cpu(cpu, cpu_online_mask)) 128 if (cpumask_test_cpu(cpu, cpu_online_mask))
132 break; 129 return cpu_physical_id(cpu);
133 } 130 }
134 return bigsmp_cpu_to_logical_apicid(cpu); 131 return BAD_APICID;
135} 132}
136 133
137static int bigsmp_phys_pkg_id(int cpuid_apic, int index_msb) 134static int bigsmp_phys_pkg_id(int cpuid_apic, int index_msb)
@@ -219,8 +216,6 @@ struct apic apic_bigsmp = {
219 .ioapic_phys_id_map = bigsmp_ioapic_phys_id_map, 216 .ioapic_phys_id_map = bigsmp_ioapic_phys_id_map,
220 .setup_apic_routing = bigsmp_setup_apic_routing, 217 .setup_apic_routing = bigsmp_setup_apic_routing,
221 .multi_timer_check = NULL, 218 .multi_timer_check = NULL,
222 .apicid_to_node = bigsmp_apicid_to_node,
223 .cpu_to_logical_apicid = bigsmp_cpu_to_logical_apicid,
224 .cpu_present_to_apicid = bigsmp_cpu_present_to_apicid, 219 .cpu_present_to_apicid = bigsmp_cpu_present_to_apicid,
225 .apicid_to_cpu_present = physid_set_mask_of_physid, 220 .apicid_to_cpu_present = physid_set_mask_of_physid,
226 .setup_portio_remap = NULL, 221 .setup_portio_remap = NULL,
@@ -256,4 +251,7 @@ struct apic apic_bigsmp = {
256 .icr_write = native_apic_icr_write, 251 .icr_write = native_apic_icr_write,
257 .wait_icr_idle = native_apic_wait_icr_idle, 252 .wait_icr_idle = native_apic_wait_icr_idle,
258 .safe_wait_icr_idle = native_safe_apic_wait_icr_idle, 253 .safe_wait_icr_idle = native_safe_apic_wait_icr_idle,
254
255 .x86_32_early_logical_apicid = bigsmp_early_logical_apicid,
256 .x86_32_numa_cpu_node = default_x86_32_numa_cpu_node,
259}; 257};
diff --git a/arch/x86/kernel/apic/es7000_32.c b/arch/x86/kernel/apic/es7000_32.c
index 8593582d8022..3e9de4854c5b 100644
--- a/arch/x86/kernel/apic/es7000_32.c
+++ b/arch/x86/kernel/apic/es7000_32.c
@@ -460,6 +460,12 @@ static unsigned long es7000_check_apicid_present(int bit)
460 return physid_isset(bit, phys_cpu_present_map); 460 return physid_isset(bit, phys_cpu_present_map);
461} 461}
462 462
463static int es7000_early_logical_apicid(int cpu)
464{
465 /* on es7000, logical apicid is the same as physical */
466 return early_per_cpu(x86_bios_cpu_apicid, cpu);
467}
468
463static unsigned long calculate_ldr(int cpu) 469static unsigned long calculate_ldr(int cpu)
464{ 470{
465 unsigned long id = per_cpu(x86_bios_cpu_apicid, cpu); 471 unsigned long id = per_cpu(x86_bios_cpu_apicid, cpu);
@@ -504,12 +510,11 @@ static void es7000_setup_apic_routing(void)
504 nr_ioapics, cpumask_bits(es7000_target_cpus())[0]); 510 nr_ioapics, cpumask_bits(es7000_target_cpus())[0]);
505} 511}
506 512
507static int es7000_apicid_to_node(int logical_apicid) 513static int es7000_numa_cpu_node(int cpu)
508{ 514{
509 return 0; 515 return 0;
510} 516}
511 517
512
513static int es7000_cpu_present_to_apicid(int mps_cpu) 518static int es7000_cpu_present_to_apicid(int mps_cpu)
514{ 519{
515 if (!mps_cpu) 520 if (!mps_cpu)
@@ -528,18 +533,6 @@ static void es7000_apicid_to_cpu_present(int phys_apicid, physid_mask_t *retmap)
528 ++cpu_id; 533 ++cpu_id;
529} 534}
530 535
531/* Mapping from cpu number to logical apicid */
532static int es7000_cpu_to_logical_apicid(int cpu)
533{
534#ifdef CONFIG_SMP
535 if (cpu >= nr_cpu_ids)
536 return BAD_APICID;
537 return cpu_2_logical_apicid[cpu];
538#else
539 return logical_smp_processor_id();
540#endif
541}
542
543static void es7000_ioapic_phys_id_map(physid_mask_t *phys_map, physid_mask_t *retmap) 536static void es7000_ioapic_phys_id_map(physid_mask_t *phys_map, physid_mask_t *retmap)
544{ 537{
545 /* For clustered we don't have a good way to do this yet - hack */ 538 /* For clustered we don't have a good way to do this yet - hack */
@@ -561,7 +554,7 @@ static unsigned int es7000_cpu_mask_to_apicid(const struct cpumask *cpumask)
561 * The cpus in the mask must all be on the apic cluster. 554 * The cpus in the mask must all be on the apic cluster.
562 */ 555 */
563 for_each_cpu(cpu, cpumask) { 556 for_each_cpu(cpu, cpumask) {
564 int new_apicid = es7000_cpu_to_logical_apicid(cpu); 557 int new_apicid = early_per_cpu(x86_cpu_to_logical_apicid, cpu);
565 558
566 if (round && APIC_CLUSTER(apicid) != APIC_CLUSTER(new_apicid)) { 559 if (round && APIC_CLUSTER(apicid) != APIC_CLUSTER(new_apicid)) {
567 WARN(1, "Not a valid mask!"); 560 WARN(1, "Not a valid mask!");
@@ -578,7 +571,7 @@ static unsigned int
578es7000_cpu_mask_to_apicid_and(const struct cpumask *inmask, 571es7000_cpu_mask_to_apicid_and(const struct cpumask *inmask,
579 const struct cpumask *andmask) 572 const struct cpumask *andmask)
580{ 573{
581 int apicid = es7000_cpu_to_logical_apicid(0); 574 int apicid = early_per_cpu(x86_cpu_to_logical_apicid, 0);
582 cpumask_var_t cpumask; 575 cpumask_var_t cpumask;
583 576
584 if (!alloc_cpumask_var(&cpumask, GFP_ATOMIC)) 577 if (!alloc_cpumask_var(&cpumask, GFP_ATOMIC))
@@ -655,8 +648,6 @@ struct apic __refdata apic_es7000_cluster = {
655 .ioapic_phys_id_map = es7000_ioapic_phys_id_map, 648 .ioapic_phys_id_map = es7000_ioapic_phys_id_map,
656 .setup_apic_routing = es7000_setup_apic_routing, 649 .setup_apic_routing = es7000_setup_apic_routing,
657 .multi_timer_check = NULL, 650 .multi_timer_check = NULL,
658 .apicid_to_node = es7000_apicid_to_node,
659 .cpu_to_logical_apicid = es7000_cpu_to_logical_apicid,
660 .cpu_present_to_apicid = es7000_cpu_present_to_apicid, 651 .cpu_present_to_apicid = es7000_cpu_present_to_apicid,
661 .apicid_to_cpu_present = es7000_apicid_to_cpu_present, 652 .apicid_to_cpu_present = es7000_apicid_to_cpu_present,
662 .setup_portio_remap = NULL, 653 .setup_portio_remap = NULL,
@@ -695,6 +686,9 @@ struct apic __refdata apic_es7000_cluster = {
695 .icr_write = native_apic_icr_write, 686 .icr_write = native_apic_icr_write,
696 .wait_icr_idle = native_apic_wait_icr_idle, 687 .wait_icr_idle = native_apic_wait_icr_idle,
697 .safe_wait_icr_idle = native_safe_apic_wait_icr_idle, 688 .safe_wait_icr_idle = native_safe_apic_wait_icr_idle,
689
690 .x86_32_early_logical_apicid = es7000_early_logical_apicid,
691 .x86_32_numa_cpu_node = es7000_numa_cpu_node,
698}; 692};
699 693
700struct apic __refdata apic_es7000 = { 694struct apic __refdata apic_es7000 = {
@@ -720,8 +714,6 @@ struct apic __refdata apic_es7000 = {
720 .ioapic_phys_id_map = es7000_ioapic_phys_id_map, 714 .ioapic_phys_id_map = es7000_ioapic_phys_id_map,
721 .setup_apic_routing = es7000_setup_apic_routing, 715 .setup_apic_routing = es7000_setup_apic_routing,
722 .multi_timer_check = NULL, 716 .multi_timer_check = NULL,
723 .apicid_to_node = es7000_apicid_to_node,
724 .cpu_to_logical_apicid = es7000_cpu_to_logical_apicid,
725 .cpu_present_to_apicid = es7000_cpu_present_to_apicid, 717 .cpu_present_to_apicid = es7000_cpu_present_to_apicid,
726 .apicid_to_cpu_present = es7000_apicid_to_cpu_present, 718 .apicid_to_cpu_present = es7000_apicid_to_cpu_present,
727 .setup_portio_remap = NULL, 719 .setup_portio_remap = NULL,
@@ -758,4 +750,7 @@ struct apic __refdata apic_es7000 = {
758 .icr_write = native_apic_icr_write, 750 .icr_write = native_apic_icr_write,
759 .wait_icr_idle = native_apic_wait_icr_idle, 751 .wait_icr_idle = native_apic_wait_icr_idle,
760 .safe_wait_icr_idle = native_safe_apic_wait_icr_idle, 752 .safe_wait_icr_idle = native_safe_apic_wait_icr_idle,
753
754 .x86_32_early_logical_apicid = es7000_early_logical_apicid,
755 .x86_32_numa_cpu_node = es7000_numa_cpu_node,
761}; 756};
diff --git a/arch/x86/kernel/apic/hw_nmi.c b/arch/x86/kernel/apic/hw_nmi.c
index 79fd43ca6f96..c4e557a1ebb6 100644
--- a/arch/x86/kernel/apic/hw_nmi.c
+++ b/arch/x86/kernel/apic/hw_nmi.c
@@ -83,7 +83,6 @@ arch_trigger_all_cpu_backtrace_handler(struct notifier_block *self,
83 arch_spin_lock(&lock); 83 arch_spin_lock(&lock);
84 printk(KERN_WARNING "NMI backtrace for cpu %d\n", cpu); 84 printk(KERN_WARNING "NMI backtrace for cpu %d\n", cpu);
85 show_regs(regs); 85 show_regs(regs);
86 dump_stack();
87 arch_spin_unlock(&lock); 86 arch_spin_unlock(&lock);
88 cpumask_clear_cpu(cpu, to_cpumask(backtrace_mask)); 87 cpumask_clear_cpu(cpu, to_cpumask(backtrace_mask));
89 return NOTIFY_STOP; 88 return NOTIFY_STOP;
diff --git a/arch/x86/kernel/apic/io_apic.c b/arch/x86/kernel/apic/io_apic.c
index ca9e2a3545a9..4b5ebd26f565 100644
--- a/arch/x86/kernel/apic/io_apic.c
+++ b/arch/x86/kernel/apic/io_apic.c
@@ -108,7 +108,10 @@ DECLARE_BITMAP(mp_bus_not_pci, MAX_MP_BUSSES);
108 108
109int skip_ioapic_setup; 109int skip_ioapic_setup;
110 110
111void arch_disable_smp_support(void) 111/**
112 * disable_ioapic_support() - disables ioapic support at runtime
113 */
114void disable_ioapic_support(void)
112{ 115{
113#ifdef CONFIG_PCI 116#ifdef CONFIG_PCI
114 noioapicquirk = 1; 117 noioapicquirk = 1;
@@ -120,11 +123,14 @@ void arch_disable_smp_support(void)
120static int __init parse_noapic(char *str) 123static int __init parse_noapic(char *str)
121{ 124{
122 /* disable IO-APIC */ 125 /* disable IO-APIC */
123 arch_disable_smp_support(); 126 disable_ioapic_support();
124 return 0; 127 return 0;
125} 128}
126early_param("noapic", parse_noapic); 129early_param("noapic", parse_noapic);
127 130
131static int io_apic_setup_irq_pin_once(unsigned int irq, int node,
132 struct io_apic_irq_attr *attr);
133
128/* Will be called in mpparse/acpi/sfi codes for saving IRQ info */ 134/* Will be called in mpparse/acpi/sfi codes for saving IRQ info */
129void mp_save_irq(struct mpc_intsrc *m) 135void mp_save_irq(struct mpc_intsrc *m)
130{ 136{
@@ -181,7 +187,7 @@ int __init arch_early_irq_init(void)
181 irq_reserve_irqs(0, legacy_pic->nr_legacy_irqs); 187 irq_reserve_irqs(0, legacy_pic->nr_legacy_irqs);
182 188
183 for (i = 0; i < count; i++) { 189 for (i = 0; i < count; i++) {
184 set_irq_chip_data(i, &cfg[i]); 190 irq_set_chip_data(i, &cfg[i]);
185 zalloc_cpumask_var_node(&cfg[i].domain, GFP_KERNEL, node); 191 zalloc_cpumask_var_node(&cfg[i].domain, GFP_KERNEL, node);
186 zalloc_cpumask_var_node(&cfg[i].old_domain, GFP_KERNEL, node); 192 zalloc_cpumask_var_node(&cfg[i].old_domain, GFP_KERNEL, node);
187 /* 193 /*
@@ -200,7 +206,7 @@ int __init arch_early_irq_init(void)
200#ifdef CONFIG_SPARSE_IRQ 206#ifdef CONFIG_SPARSE_IRQ
201static struct irq_cfg *irq_cfg(unsigned int irq) 207static struct irq_cfg *irq_cfg(unsigned int irq)
202{ 208{
203 return get_irq_chip_data(irq); 209 return irq_get_chip_data(irq);
204} 210}
205 211
206static struct irq_cfg *alloc_irq_cfg(unsigned int irq, int node) 212static struct irq_cfg *alloc_irq_cfg(unsigned int irq, int node)
@@ -226,7 +232,7 @@ static void free_irq_cfg(unsigned int at, struct irq_cfg *cfg)
226{ 232{
227 if (!cfg) 233 if (!cfg)
228 return; 234 return;
229 set_irq_chip_data(at, NULL); 235 irq_set_chip_data(at, NULL);
230 free_cpumask_var(cfg->domain); 236 free_cpumask_var(cfg->domain);
231 free_cpumask_var(cfg->old_domain); 237 free_cpumask_var(cfg->old_domain);
232 kfree(cfg); 238 kfree(cfg);
@@ -256,14 +262,14 @@ static struct irq_cfg *alloc_irq_and_cfg_at(unsigned int at, int node)
256 if (res < 0) { 262 if (res < 0) {
257 if (res != -EEXIST) 263 if (res != -EEXIST)
258 return NULL; 264 return NULL;
259 cfg = get_irq_chip_data(at); 265 cfg = irq_get_chip_data(at);
260 if (cfg) 266 if (cfg)
261 return cfg; 267 return cfg;
262 } 268 }
263 269
264 cfg = alloc_irq_cfg(at, node); 270 cfg = alloc_irq_cfg(at, node);
265 if (cfg) 271 if (cfg)
266 set_irq_chip_data(at, cfg); 272 irq_set_chip_data(at, cfg);
267 else 273 else
268 irq_free_desc(at); 274 irq_free_desc(at);
269 return cfg; 275 return cfg;
@@ -818,7 +824,7 @@ static int EISA_ELCR(unsigned int irq)
818#define default_MCA_trigger(idx) (1) 824#define default_MCA_trigger(idx) (1)
819#define default_MCA_polarity(idx) default_ISA_polarity(idx) 825#define default_MCA_polarity(idx) default_ISA_polarity(idx)
820 826
821static int MPBIOS_polarity(int idx) 827static int irq_polarity(int idx)
822{ 828{
823 int bus = mp_irqs[idx].srcbus; 829 int bus = mp_irqs[idx].srcbus;
824 int polarity; 830 int polarity;
@@ -860,7 +866,7 @@ static int MPBIOS_polarity(int idx)
860 return polarity; 866 return polarity;
861} 867}
862 868
863static int MPBIOS_trigger(int idx) 869static int irq_trigger(int idx)
864{ 870{
865 int bus = mp_irqs[idx].srcbus; 871 int bus = mp_irqs[idx].srcbus;
866 int trigger; 872 int trigger;
@@ -932,16 +938,6 @@ static int MPBIOS_trigger(int idx)
932 return trigger; 938 return trigger;
933} 939}
934 940
935static inline int irq_polarity(int idx)
936{
937 return MPBIOS_polarity(idx);
938}
939
940static inline int irq_trigger(int idx)
941{
942 return MPBIOS_trigger(idx);
943}
944
945static int pin_2_irq(int idx, int apic, int pin) 941static int pin_2_irq(int idx, int apic, int pin)
946{ 942{
947 int irq; 943 int irq;
@@ -1189,7 +1185,7 @@ void __setup_vector_irq(int cpu)
1189 raw_spin_lock(&vector_lock); 1185 raw_spin_lock(&vector_lock);
1190 /* Mark the inuse vectors */ 1186 /* Mark the inuse vectors */
1191 for_each_active_irq(irq) { 1187 for_each_active_irq(irq) {
1192 cfg = get_irq_chip_data(irq); 1188 cfg = irq_get_chip_data(irq);
1193 if (!cfg) 1189 if (!cfg)
1194 continue; 1190 continue;
1195 /* 1191 /*
@@ -1220,10 +1216,6 @@ void __setup_vector_irq(int cpu)
1220static struct irq_chip ioapic_chip; 1216static struct irq_chip ioapic_chip;
1221static struct irq_chip ir_ioapic_chip; 1217static struct irq_chip ir_ioapic_chip;
1222 1218
1223#define IOAPIC_AUTO -1
1224#define IOAPIC_EDGE 0
1225#define IOAPIC_LEVEL 1
1226
1227#ifdef CONFIG_X86_32 1219#ifdef CONFIG_X86_32
1228static inline int IO_APIC_irq_trigger(int irq) 1220static inline int IO_APIC_irq_trigger(int irq)
1229{ 1221{
@@ -1248,35 +1240,31 @@ static inline int IO_APIC_irq_trigger(int irq)
1248} 1240}
1249#endif 1241#endif
1250 1242
1251static void ioapic_register_intr(unsigned int irq, unsigned long trigger) 1243static void ioapic_register_intr(unsigned int irq, struct irq_cfg *cfg,
1244 unsigned long trigger)
1252{ 1245{
1246 struct irq_chip *chip = &ioapic_chip;
1247 irq_flow_handler_t hdl;
1248 bool fasteoi;
1253 1249
1254 if ((trigger == IOAPIC_AUTO && IO_APIC_irq_trigger(irq)) || 1250 if ((trigger == IOAPIC_AUTO && IO_APIC_irq_trigger(irq)) ||
1255 trigger == IOAPIC_LEVEL) 1251 trigger == IOAPIC_LEVEL) {
1256 irq_set_status_flags(irq, IRQ_LEVEL); 1252 irq_set_status_flags(irq, IRQ_LEVEL);
1257 else 1253 fasteoi = true;
1254 } else {
1258 irq_clear_status_flags(irq, IRQ_LEVEL); 1255 irq_clear_status_flags(irq, IRQ_LEVEL);
1256 fasteoi = false;
1257 }
1259 1258
1260 if (irq_remapped(get_irq_chip_data(irq))) { 1259 if (irq_remapped(cfg)) {
1261 irq_set_status_flags(irq, IRQ_MOVE_PCNTXT); 1260 irq_set_status_flags(irq, IRQ_MOVE_PCNTXT);
1262 if (trigger) 1261 chip = &ir_ioapic_chip;
1263 set_irq_chip_and_handler_name(irq, &ir_ioapic_chip, 1262 fasteoi = trigger != 0;
1264 handle_fasteoi_irq,
1265 "fasteoi");
1266 else
1267 set_irq_chip_and_handler_name(irq, &ir_ioapic_chip,
1268 handle_edge_irq, "edge");
1269 return;
1270 } 1263 }
1271 1264
1272 if ((trigger == IOAPIC_AUTO && IO_APIC_irq_trigger(irq)) || 1265 hdl = fasteoi ? handle_fasteoi_irq : handle_edge_irq;
1273 trigger == IOAPIC_LEVEL) 1266 irq_set_chip_and_handler_name(irq, chip, hdl,
1274 set_irq_chip_and_handler_name(irq, &ioapic_chip, 1267 fasteoi ? "fasteoi" : "edge");
1275 handle_fasteoi_irq,
1276 "fasteoi");
1277 else
1278 set_irq_chip_and_handler_name(irq, &ioapic_chip,
1279 handle_edge_irq, "edge");
1280} 1268}
1281 1269
1282static int setup_ioapic_entry(int apic_id, int irq, 1270static int setup_ioapic_entry(int apic_id, int irq,
@@ -1374,7 +1362,7 @@ static void setup_ioapic_irq(int apic_id, int pin, unsigned int irq,
1374 return; 1362 return;
1375 } 1363 }
1376 1364
1377 ioapic_register_intr(irq, trigger); 1365 ioapic_register_intr(irq, cfg, trigger);
1378 if (irq < legacy_pic->nr_legacy_irqs) 1366 if (irq < legacy_pic->nr_legacy_irqs)
1379 legacy_pic->mask(irq); 1367 legacy_pic->mask(irq);
1380 1368
@@ -1385,33 +1373,26 @@ static struct {
1385 DECLARE_BITMAP(pin_programmed, MP_MAX_IOAPIC_PIN + 1); 1373 DECLARE_BITMAP(pin_programmed, MP_MAX_IOAPIC_PIN + 1);
1386} mp_ioapic_routing[MAX_IO_APICS]; 1374} mp_ioapic_routing[MAX_IO_APICS];
1387 1375
1388static void __init setup_IO_APIC_irqs(void) 1376static bool __init io_apic_pin_not_connected(int idx, int apic_id, int pin)
1389{ 1377{
1390 int apic_id, pin, idx, irq, notcon = 0; 1378 if (idx != -1)
1391 int node = cpu_to_node(0); 1379 return false;
1392 struct irq_cfg *cfg;
1393 1380
1394 apic_printk(APIC_VERBOSE, KERN_DEBUG "init IO_APIC IRQs\n"); 1381 apic_printk(APIC_VERBOSE, KERN_DEBUG " apic %d pin %d not connected\n",
1382 mp_ioapics[apic_id].apicid, pin);
1383 return true;
1384}
1385
1386static void __init __io_apic_setup_irqs(unsigned int apic_id)
1387{
1388 int idx, node = cpu_to_node(0);
1389 struct io_apic_irq_attr attr;
1390 unsigned int pin, irq;
1395 1391
1396 for (apic_id = 0; apic_id < nr_ioapics; apic_id++)
1397 for (pin = 0; pin < nr_ioapic_registers[apic_id]; pin++) { 1392 for (pin = 0; pin < nr_ioapic_registers[apic_id]; pin++) {
1398 idx = find_irq_entry(apic_id, pin, mp_INT); 1393 idx = find_irq_entry(apic_id, pin, mp_INT);
1399 if (idx == -1) { 1394 if (io_apic_pin_not_connected(idx, apic_id, pin))
1400 if (!notcon) {
1401 notcon = 1;
1402 apic_printk(APIC_VERBOSE,
1403 KERN_DEBUG " %d-%d",
1404 mp_ioapics[apic_id].apicid, pin);
1405 } else
1406 apic_printk(APIC_VERBOSE, " %d-%d",
1407 mp_ioapics[apic_id].apicid, pin);
1408 continue; 1395 continue;
1409 }
1410 if (notcon) {
1411 apic_printk(APIC_VERBOSE,
1412 " (apicid-pin) not connected\n");
1413 notcon = 0;
1414 }
1415 1396
1416 irq = pin_2_irq(idx, apic_id, pin); 1397 irq = pin_2_irq(idx, apic_id, pin);
1417 1398
@@ -1423,25 +1404,24 @@ static void __init setup_IO_APIC_irqs(void)
1423 * installed and if it returns 1: 1404 * installed and if it returns 1:
1424 */ 1405 */
1425 if (apic->multi_timer_check && 1406 if (apic->multi_timer_check &&
1426 apic->multi_timer_check(apic_id, irq)) 1407 apic->multi_timer_check(apic_id, irq))
1427 continue; 1408 continue;
1428 1409
1429 cfg = alloc_irq_and_cfg_at(irq, node); 1410 set_io_apic_irq_attr(&attr, apic_id, pin, irq_trigger(idx),
1430 if (!cfg) 1411 irq_polarity(idx));
1431 continue;
1432 1412
1433 add_pin_to_irq_node(cfg, node, apic_id, pin); 1413 io_apic_setup_irq_pin(irq, node, &attr);
1434 /*
1435 * don't mark it in pin_programmed, so later acpi could
1436 * set it correctly when irq < 16
1437 */
1438 setup_ioapic_irq(apic_id, pin, irq, cfg, irq_trigger(idx),
1439 irq_polarity(idx));
1440 } 1414 }
1415}
1441 1416
1442 if (notcon) 1417static void __init setup_IO_APIC_irqs(void)
1443 apic_printk(APIC_VERBOSE, 1418{
1444 " (apicid-pin) not connected\n"); 1419 unsigned int apic_id;
1420
1421 apic_printk(APIC_VERBOSE, KERN_DEBUG "init IO_APIC IRQs\n");
1422
1423 for (apic_id = 0; apic_id < nr_ioapics; apic_id++)
1424 __io_apic_setup_irqs(apic_id);
1445} 1425}
1446 1426
1447/* 1427/*
@@ -1452,7 +1432,7 @@ static void __init setup_IO_APIC_irqs(void)
1452void setup_IO_APIC_irq_extra(u32 gsi) 1432void setup_IO_APIC_irq_extra(u32 gsi)
1453{ 1433{
1454 int apic_id = 0, pin, idx, irq, node = cpu_to_node(0); 1434 int apic_id = 0, pin, idx, irq, node = cpu_to_node(0);
1455 struct irq_cfg *cfg; 1435 struct io_apic_irq_attr attr;
1456 1436
1457 /* 1437 /*
1458 * Convert 'gsi' to 'ioapic.pin'. 1438 * Convert 'gsi' to 'ioapic.pin'.
@@ -1472,21 +1452,10 @@ void setup_IO_APIC_irq_extra(u32 gsi)
1472 if (apic_id == 0 || irq < NR_IRQS_LEGACY) 1452 if (apic_id == 0 || irq < NR_IRQS_LEGACY)
1473 return; 1453 return;
1474 1454
1475 cfg = alloc_irq_and_cfg_at(irq, node); 1455 set_io_apic_irq_attr(&attr, apic_id, pin, irq_trigger(idx),
1476 if (!cfg) 1456 irq_polarity(idx));
1477 return;
1478
1479 add_pin_to_irq_node(cfg, node, apic_id, pin);
1480
1481 if (test_bit(pin, mp_ioapic_routing[apic_id].pin_programmed)) {
1482 pr_debug("Pin %d-%d already programmed\n",
1483 mp_ioapics[apic_id].apicid, pin);
1484 return;
1485 }
1486 set_bit(pin, mp_ioapic_routing[apic_id].pin_programmed);
1487 1457
1488 setup_ioapic_irq(apic_id, pin, irq, cfg, 1458 io_apic_setup_irq_pin_once(irq, node, &attr);
1489 irq_trigger(idx), irq_polarity(idx));
1490} 1459}
1491 1460
1492/* 1461/*
@@ -1518,7 +1487,8 @@ static void __init setup_timer_IRQ0_pin(unsigned int apic_id, unsigned int pin,
1518 * The timer IRQ doesn't have to know that behind the 1487 * The timer IRQ doesn't have to know that behind the
1519 * scene we may have a 8259A-master in AEOI mode ... 1488 * scene we may have a 8259A-master in AEOI mode ...
1520 */ 1489 */
1521 set_irq_chip_and_handler_name(0, &ioapic_chip, handle_edge_irq, "edge"); 1490 irq_set_chip_and_handler_name(0, &ioapic_chip, handle_edge_irq,
1491 "edge");
1522 1492
1523 /* 1493 /*
1524 * Add it to the IO-APIC irq-routing table: 1494 * Add it to the IO-APIC irq-routing table:
@@ -1625,7 +1595,7 @@ __apicdebuginit(void) print_IO_APIC(void)
1625 for_each_active_irq(irq) { 1595 for_each_active_irq(irq) {
1626 struct irq_pin_list *entry; 1596 struct irq_pin_list *entry;
1627 1597
1628 cfg = get_irq_chip_data(irq); 1598 cfg = irq_get_chip_data(irq);
1629 if (!cfg) 1599 if (!cfg)
1630 continue; 1600 continue;
1631 entry = cfg->irq_2_pin; 1601 entry = cfg->irq_2_pin;
@@ -2391,7 +2361,7 @@ static void irq_complete_move(struct irq_cfg *cfg)
2391 2361
2392void irq_force_complete_move(int irq) 2362void irq_force_complete_move(int irq)
2393{ 2363{
2394 struct irq_cfg *cfg = get_irq_chip_data(irq); 2364 struct irq_cfg *cfg = irq_get_chip_data(irq);
2395 2365
2396 if (!cfg) 2366 if (!cfg)
2397 return; 2367 return;
@@ -2405,7 +2375,7 @@ static inline void irq_complete_move(struct irq_cfg *cfg) { }
2405static void ack_apic_edge(struct irq_data *data) 2375static void ack_apic_edge(struct irq_data *data)
2406{ 2376{
2407 irq_complete_move(data->chip_data); 2377 irq_complete_move(data->chip_data);
2408 move_native_irq(data->irq); 2378 irq_move_irq(data);
2409 ack_APIC_irq(); 2379 ack_APIC_irq();
2410} 2380}
2411 2381
@@ -2462,7 +2432,7 @@ static void ack_apic_level(struct irq_data *data)
2462 irq_complete_move(cfg); 2432 irq_complete_move(cfg);
2463#ifdef CONFIG_GENERIC_PENDING_IRQ 2433#ifdef CONFIG_GENERIC_PENDING_IRQ
2464 /* If we are moving the irq we need to mask it */ 2434 /* If we are moving the irq we need to mask it */
2465 if (unlikely(irq_to_desc(irq)->status & IRQ_MOVE_PENDING)) { 2435 if (unlikely(irqd_is_setaffinity_pending(data))) {
2466 do_unmask_irq = 1; 2436 do_unmask_irq = 1;
2467 mask_ioapic(cfg); 2437 mask_ioapic(cfg);
2468 } 2438 }
@@ -2551,7 +2521,7 @@ static void ack_apic_level(struct irq_data *data)
2551 * and you can go talk to the chipset vendor about it. 2521 * and you can go talk to the chipset vendor about it.
2552 */ 2522 */
2553 if (!io_apic_level_ack_pending(cfg)) 2523 if (!io_apic_level_ack_pending(cfg))
2554 move_masked_irq(irq); 2524 irq_move_masked_irq(data);
2555 unmask_ioapic(cfg); 2525 unmask_ioapic(cfg);
2556 } 2526 }
2557} 2527}
@@ -2614,7 +2584,7 @@ static inline void init_IO_APIC_traps(void)
2614 * 0x80, because int 0x80 is hm, kind of importantish. ;) 2584 * 0x80, because int 0x80 is hm, kind of importantish. ;)
2615 */ 2585 */
2616 for_each_active_irq(irq) { 2586 for_each_active_irq(irq) {
2617 cfg = get_irq_chip_data(irq); 2587 cfg = irq_get_chip_data(irq);
2618 if (IO_APIC_IRQ(irq) && cfg && !cfg->vector) { 2588 if (IO_APIC_IRQ(irq) && cfg && !cfg->vector) {
2619 /* 2589 /*
2620 * Hmm.. We don't have an entry for this, 2590 * Hmm.. We don't have an entry for this,
@@ -2625,7 +2595,7 @@ static inline void init_IO_APIC_traps(void)
2625 legacy_pic->make_irq(irq); 2595 legacy_pic->make_irq(irq);
2626 else 2596 else
2627 /* Strange. Oh, well.. */ 2597 /* Strange. Oh, well.. */
2628 set_irq_chip(irq, &no_irq_chip); 2598 irq_set_chip(irq, &no_irq_chip);
2629 } 2599 }
2630 } 2600 }
2631} 2601}
@@ -2665,7 +2635,7 @@ static struct irq_chip lapic_chip __read_mostly = {
2665static void lapic_register_intr(int irq) 2635static void lapic_register_intr(int irq)
2666{ 2636{
2667 irq_clear_status_flags(irq, IRQ_LEVEL); 2637 irq_clear_status_flags(irq, IRQ_LEVEL);
2668 set_irq_chip_and_handler_name(irq, &lapic_chip, handle_edge_irq, 2638 irq_set_chip_and_handler_name(irq, &lapic_chip, handle_edge_irq,
2669 "edge"); 2639 "edge");
2670} 2640}
2671 2641
@@ -2749,7 +2719,7 @@ int timer_through_8259 __initdata;
2749 */ 2719 */
2750static inline void __init check_timer(void) 2720static inline void __init check_timer(void)
2751{ 2721{
2752 struct irq_cfg *cfg = get_irq_chip_data(0); 2722 struct irq_cfg *cfg = irq_get_chip_data(0);
2753 int node = cpu_to_node(0); 2723 int node = cpu_to_node(0);
2754 int apic1, pin1, apic2, pin2; 2724 int apic1, pin1, apic2, pin2;
2755 unsigned long flags; 2725 unsigned long flags;
@@ -3060,7 +3030,7 @@ unsigned int create_irq_nr(unsigned int from, int node)
3060 raw_spin_unlock_irqrestore(&vector_lock, flags); 3030 raw_spin_unlock_irqrestore(&vector_lock, flags);
3061 3031
3062 if (ret) { 3032 if (ret) {
3063 set_irq_chip_data(irq, cfg); 3033 irq_set_chip_data(irq, cfg);
3064 irq_clear_status_flags(irq, IRQ_NOREQUEST); 3034 irq_clear_status_flags(irq, IRQ_NOREQUEST);
3065 } else { 3035 } else {
3066 free_irq_at(irq, cfg); 3036 free_irq_at(irq, cfg);
@@ -3085,7 +3055,7 @@ int create_irq(void)
3085 3055
3086void destroy_irq(unsigned int irq) 3056void destroy_irq(unsigned int irq)
3087{ 3057{
3088 struct irq_cfg *cfg = get_irq_chip_data(irq); 3058 struct irq_cfg *cfg = irq_get_chip_data(irq);
3089 unsigned long flags; 3059 unsigned long flags;
3090 3060
3091 irq_set_status_flags(irq, IRQ_NOREQUEST|IRQ_NOPROBE); 3061 irq_set_status_flags(irq, IRQ_NOREQUEST|IRQ_NOPROBE);
@@ -3119,7 +3089,7 @@ static int msi_compose_msg(struct pci_dev *pdev, unsigned int irq,
3119 3089
3120 dest = apic->cpu_mask_to_apicid_and(cfg->domain, apic->target_cpus()); 3090 dest = apic->cpu_mask_to_apicid_and(cfg->domain, apic->target_cpus());
3121 3091
3122 if (irq_remapped(get_irq_chip_data(irq))) { 3092 if (irq_remapped(cfg)) {
3123 struct irte irte; 3093 struct irte irte;
3124 int ir_index; 3094 int ir_index;
3125 u16 sub_handle; 3095 u16 sub_handle;
@@ -3291,6 +3261,7 @@ static int msi_alloc_irte(struct pci_dev *dev, int irq, int nvec)
3291 3261
3292static int setup_msi_irq(struct pci_dev *dev, struct msi_desc *msidesc, int irq) 3262static int setup_msi_irq(struct pci_dev *dev, struct msi_desc *msidesc, int irq)
3293{ 3263{
3264 struct irq_chip *chip = &msi_chip;
3294 struct msi_msg msg; 3265 struct msi_msg msg;
3295 int ret; 3266 int ret;
3296 3267
@@ -3298,14 +3269,15 @@ static int setup_msi_irq(struct pci_dev *dev, struct msi_desc *msidesc, int irq)
3298 if (ret < 0) 3269 if (ret < 0)
3299 return ret; 3270 return ret;
3300 3271
3301 set_irq_msi(irq, msidesc); 3272 irq_set_msi_desc(irq, msidesc);
3302 write_msi_msg(irq, &msg); 3273 write_msi_msg(irq, &msg);
3303 3274
3304 if (irq_remapped(get_irq_chip_data(irq))) { 3275 if (irq_remapped(irq_get_chip_data(irq))) {
3305 irq_set_status_flags(irq, IRQ_MOVE_PCNTXT); 3276 irq_set_status_flags(irq, IRQ_MOVE_PCNTXT);
3306 set_irq_chip_and_handler_name(irq, &msi_ir_chip, handle_edge_irq, "edge"); 3277 chip = &msi_ir_chip;
3307 } else 3278 }
3308 set_irq_chip_and_handler_name(irq, &msi_chip, handle_edge_irq, "edge"); 3279
3280 irq_set_chip_and_handler_name(irq, chip, handle_edge_irq, "edge");
3309 3281
3310 dev_printk(KERN_DEBUG, &dev->dev, "irq %d for MSI/MSI-X\n", irq); 3282 dev_printk(KERN_DEBUG, &dev->dev, "irq %d for MSI/MSI-X\n", irq);
3311 3283
@@ -3423,8 +3395,8 @@ int arch_setup_dmar_msi(unsigned int irq)
3423 if (ret < 0) 3395 if (ret < 0)
3424 return ret; 3396 return ret;
3425 dmar_msi_write(irq, &msg); 3397 dmar_msi_write(irq, &msg);
3426 set_irq_chip_and_handler_name(irq, &dmar_msi_type, handle_edge_irq, 3398 irq_set_chip_and_handler_name(irq, &dmar_msi_type, handle_edge_irq,
3427 "edge"); 3399 "edge");
3428 return 0; 3400 return 0;
3429} 3401}
3430#endif 3402#endif
@@ -3482,6 +3454,7 @@ static struct irq_chip hpet_msi_type = {
3482 3454
3483int arch_setup_hpet_msi(unsigned int irq, unsigned int id) 3455int arch_setup_hpet_msi(unsigned int irq, unsigned int id)
3484{ 3456{
3457 struct irq_chip *chip = &hpet_msi_type;
3485 struct msi_msg msg; 3458 struct msi_msg msg;
3486 int ret; 3459 int ret;
3487 3460
@@ -3501,15 +3474,12 @@ int arch_setup_hpet_msi(unsigned int irq, unsigned int id)
3501 if (ret < 0) 3474 if (ret < 0)
3502 return ret; 3475 return ret;
3503 3476
3504 hpet_msi_write(get_irq_data(irq), &msg); 3477 hpet_msi_write(irq_get_handler_data(irq), &msg);
3505 irq_set_status_flags(irq, IRQ_MOVE_PCNTXT); 3478 irq_set_status_flags(irq, IRQ_MOVE_PCNTXT);
3506 if (irq_remapped(get_irq_chip_data(irq))) 3479 if (irq_remapped(irq_get_chip_data(irq)))
3507 set_irq_chip_and_handler_name(irq, &ir_hpet_msi_type, 3480 chip = &ir_hpet_msi_type;
3508 handle_edge_irq, "edge");
3509 else
3510 set_irq_chip_and_handler_name(irq, &hpet_msi_type,
3511 handle_edge_irq, "edge");
3512 3481
3482 irq_set_chip_and_handler_name(irq, chip, handle_edge_irq, "edge");
3513 return 0; 3483 return 0;
3514} 3484}
3515#endif 3485#endif
@@ -3596,7 +3566,7 @@ int arch_setup_ht_irq(unsigned int irq, struct pci_dev *dev)
3596 3566
3597 write_ht_irq_msg(irq, &msg); 3567 write_ht_irq_msg(irq, &msg);
3598 3568
3599 set_irq_chip_and_handler_name(irq, &ht_irq_chip, 3569 irq_set_chip_and_handler_name(irq, &ht_irq_chip,
3600 handle_edge_irq, "edge"); 3570 handle_edge_irq, "edge");
3601 3571
3602 dev_printk(KERN_DEBUG, &dev->dev, "irq %d for HT\n", irq); 3572 dev_printk(KERN_DEBUG, &dev->dev, "irq %d for HT\n", irq);
@@ -3605,7 +3575,40 @@ int arch_setup_ht_irq(unsigned int irq, struct pci_dev *dev)
3605} 3575}
3606#endif /* CONFIG_HT_IRQ */ 3576#endif /* CONFIG_HT_IRQ */
3607 3577
3608int __init io_apic_get_redir_entries (int ioapic) 3578int
3579io_apic_setup_irq_pin(unsigned int irq, int node, struct io_apic_irq_attr *attr)
3580{
3581 struct irq_cfg *cfg = alloc_irq_and_cfg_at(irq, node);
3582 int ret;
3583
3584 if (!cfg)
3585 return -EINVAL;
3586 ret = __add_pin_to_irq_node(cfg, node, attr->ioapic, attr->ioapic_pin);
3587 if (!ret)
3588 setup_ioapic_irq(attr->ioapic, attr->ioapic_pin, irq, cfg,
3589 attr->trigger, attr->polarity);
3590 return ret;
3591}
3592
3593static int io_apic_setup_irq_pin_once(unsigned int irq, int node,
3594 struct io_apic_irq_attr *attr)
3595{
3596 unsigned int id = attr->ioapic, pin = attr->ioapic_pin;
3597 int ret;
3598
3599 /* Avoid redundant programming */
3600 if (test_bit(pin, mp_ioapic_routing[id].pin_programmed)) {
3601 pr_debug("Pin %d-%d already programmed\n",
3602 mp_ioapics[id].apicid, pin);
3603 return 0;
3604 }
3605 ret = io_apic_setup_irq_pin(irq, node, attr);
3606 if (!ret)
3607 set_bit(pin, mp_ioapic_routing[id].pin_programmed);
3608 return ret;
3609}
3610
3611static int __init io_apic_get_redir_entries(int ioapic)
3609{ 3612{
3610 union IO_APIC_reg_01 reg_01; 3613 union IO_APIC_reg_01 reg_01;
3611 unsigned long flags; 3614 unsigned long flags;
@@ -3659,96 +3662,24 @@ int __init arch_probe_nr_irqs(void)
3659} 3662}
3660#endif 3663#endif
3661 3664
3662static int __io_apic_set_pci_routing(struct device *dev, int irq, 3665int io_apic_set_pci_routing(struct device *dev, int irq,
3663 struct io_apic_irq_attr *irq_attr) 3666 struct io_apic_irq_attr *irq_attr)
3664{ 3667{
3665 struct irq_cfg *cfg;
3666 int node; 3668 int node;
3667 int ioapic, pin;
3668 int trigger, polarity;
3669 3669
3670 ioapic = irq_attr->ioapic;
3671 if (!IO_APIC_IRQ(irq)) { 3670 if (!IO_APIC_IRQ(irq)) {
3672 apic_printk(APIC_QUIET,KERN_ERR "IOAPIC[%d]: Invalid reference to IRQ 0\n", 3671 apic_printk(APIC_QUIET,KERN_ERR "IOAPIC[%d]: Invalid reference to IRQ 0\n",
3673 ioapic); 3672 irq_attr->ioapic);
3674 return -EINVAL; 3673 return -EINVAL;
3675 } 3674 }
3676 3675
3677 if (dev) 3676 node = dev ? dev_to_node(dev) : cpu_to_node(0);
3678 node = dev_to_node(dev);
3679 else
3680 node = cpu_to_node(0);
3681
3682 cfg = alloc_irq_and_cfg_at(irq, node);
3683 if (!cfg)
3684 return 0;
3685
3686 pin = irq_attr->ioapic_pin;
3687 trigger = irq_attr->trigger;
3688 polarity = irq_attr->polarity;
3689 3677
3690 /* 3678 return io_apic_setup_irq_pin_once(irq, node, irq_attr);
3691 * IRQs < 16 are already in the irq_2_pin[] map
3692 */
3693 if (irq >= legacy_pic->nr_legacy_irqs) {
3694 if (__add_pin_to_irq_node(cfg, node, ioapic, pin)) {
3695 printk(KERN_INFO "can not add pin %d for irq %d\n",
3696 pin, irq);
3697 return 0;
3698 }
3699 }
3700
3701 setup_ioapic_irq(ioapic, pin, irq, cfg, trigger, polarity);
3702
3703 return 0;
3704} 3679}
3705 3680
3706int io_apic_set_pci_routing(struct device *dev, int irq,
3707 struct io_apic_irq_attr *irq_attr)
3708{
3709 int ioapic, pin;
3710 /*
3711 * Avoid pin reprogramming. PRTs typically include entries
3712 * with redundant pin->gsi mappings (but unique PCI devices);
3713 * we only program the IOAPIC on the first.
3714 */
3715 ioapic = irq_attr->ioapic;
3716 pin = irq_attr->ioapic_pin;
3717 if (test_bit(pin, mp_ioapic_routing[ioapic].pin_programmed)) {
3718 pr_debug("Pin %d-%d already programmed\n",
3719 mp_ioapics[ioapic].apicid, pin);
3720 return 0;
3721 }
3722 set_bit(pin, mp_ioapic_routing[ioapic].pin_programmed);
3723
3724 return __io_apic_set_pci_routing(dev, irq, irq_attr);
3725}
3726
3727u8 __init io_apic_unique_id(u8 id)
3728{
3729#ifdef CONFIG_X86_32 3681#ifdef CONFIG_X86_32
3730 if ((boot_cpu_data.x86_vendor == X86_VENDOR_INTEL) && 3682static int __init io_apic_get_unique_id(int ioapic, int apic_id)
3731 !APIC_XAPIC(apic_version[boot_cpu_physical_apicid]))
3732 return io_apic_get_unique_id(nr_ioapics, id);
3733 else
3734 return id;
3735#else
3736 int i;
3737 DECLARE_BITMAP(used, 256);
3738
3739 bitmap_zero(used, 256);
3740 for (i = 0; i < nr_ioapics; i++) {
3741 struct mpc_ioapic *ia = &mp_ioapics[i];
3742 __set_bit(ia->apicid, used);
3743 }
3744 if (!test_bit(id, used))
3745 return id;
3746 return find_first_zero_bit(used, 256);
3747#endif
3748}
3749
3750#ifdef CONFIG_X86_32
3751int __init io_apic_get_unique_id(int ioapic, int apic_id)
3752{ 3683{
3753 union IO_APIC_reg_00 reg_00; 3684 union IO_APIC_reg_00 reg_00;
3754 static physid_mask_t apic_id_map = PHYSID_MASK_NONE; 3685 static physid_mask_t apic_id_map = PHYSID_MASK_NONE;
@@ -3821,9 +3752,33 @@ int __init io_apic_get_unique_id(int ioapic, int apic_id)
3821 3752
3822 return apic_id; 3753 return apic_id;
3823} 3754}
3755
3756static u8 __init io_apic_unique_id(u8 id)
3757{
3758 if ((boot_cpu_data.x86_vendor == X86_VENDOR_INTEL) &&
3759 !APIC_XAPIC(apic_version[boot_cpu_physical_apicid]))
3760 return io_apic_get_unique_id(nr_ioapics, id);
3761 else
3762 return id;
3763}
3764#else
3765static u8 __init io_apic_unique_id(u8 id)
3766{
3767 int i;
3768 DECLARE_BITMAP(used, 256);
3769
3770 bitmap_zero(used, 256);
3771 for (i = 0; i < nr_ioapics; i++) {
3772 struct mpc_ioapic *ia = &mp_ioapics[i];
3773 __set_bit(ia->apicid, used);
3774 }
3775 if (!test_bit(id, used))
3776 return id;
3777 return find_first_zero_bit(used, 256);
3778}
3824#endif 3779#endif
3825 3780
3826int __init io_apic_get_version(int ioapic) 3781static int __init io_apic_get_version(int ioapic)
3827{ 3782{
3828 union IO_APIC_reg_01 reg_01; 3783 union IO_APIC_reg_01 reg_01;
3829 unsigned long flags; 3784 unsigned long flags;
@@ -3868,8 +3823,8 @@ int acpi_get_override_irq(u32 gsi, int *trigger, int *polarity)
3868void __init setup_ioapic_dest(void) 3823void __init setup_ioapic_dest(void)
3869{ 3824{
3870 int pin, ioapic, irq, irq_entry; 3825 int pin, ioapic, irq, irq_entry;
3871 struct irq_desc *desc;
3872 const struct cpumask *mask; 3826 const struct cpumask *mask;
3827 struct irq_data *idata;
3873 3828
3874 if (skip_ioapic_setup == 1) 3829 if (skip_ioapic_setup == 1)
3875 return; 3830 return;
@@ -3884,21 +3839,20 @@ void __init setup_ioapic_dest(void)
3884 if ((ioapic > 0) && (irq > 16)) 3839 if ((ioapic > 0) && (irq > 16))
3885 continue; 3840 continue;
3886 3841
3887 desc = irq_to_desc(irq); 3842 idata = irq_get_irq_data(irq);
3888 3843
3889 /* 3844 /*
3890 * Honour affinities which have been set in early boot 3845 * Honour affinities which have been set in early boot
3891 */ 3846 */
3892 if (desc->status & 3847 if (!irqd_can_balance(idata) || irqd_affinity_was_set(idata))
3893 (IRQ_NO_BALANCING | IRQ_AFFINITY_SET)) 3848 mask = idata->affinity;
3894 mask = desc->irq_data.affinity;
3895 else 3849 else
3896 mask = apic->target_cpus(); 3850 mask = apic->target_cpus();
3897 3851
3898 if (intr_remapping_enabled) 3852 if (intr_remapping_enabled)
3899 ir_ioapic_set_affinity(&desc->irq_data, mask, false); 3853 ir_ioapic_set_affinity(idata, mask, false);
3900 else 3854 else
3901 ioapic_set_affinity(&desc->irq_data, mask, false); 3855 ioapic_set_affinity(idata, mask, false);
3902 } 3856 }
3903 3857
3904} 3858}
@@ -4026,7 +3980,7 @@ int mp_find_ioapic_pin(int ioapic, u32 gsi)
4026 return gsi - mp_gsi_routing[ioapic].gsi_base; 3980 return gsi - mp_gsi_routing[ioapic].gsi_base;
4027} 3981}
4028 3982
4029static int bad_ioapic(unsigned long address) 3983static __init int bad_ioapic(unsigned long address)
4030{ 3984{
4031 if (nr_ioapics >= MAX_IO_APICS) { 3985 if (nr_ioapics >= MAX_IO_APICS) {
4032 printk(KERN_WARNING "WARING: Max # of I/O APICs (%d) exceeded " 3986 printk(KERN_WARNING "WARING: Max # of I/O APICs (%d) exceeded "
@@ -4086,20 +4040,16 @@ void __init mp_register_ioapic(int id, u32 address, u32 gsi_base)
4086/* Enable IOAPIC early just for system timer */ 4040/* Enable IOAPIC early just for system timer */
4087void __init pre_init_apic_IRQ0(void) 4041void __init pre_init_apic_IRQ0(void)
4088{ 4042{
4089 struct irq_cfg *cfg; 4043 struct io_apic_irq_attr attr = { 0, 0, 0, 0 };
4090 4044
4091 printk(KERN_INFO "Early APIC setup for system timer0\n"); 4045 printk(KERN_INFO "Early APIC setup for system timer0\n");
4092#ifndef CONFIG_SMP 4046#ifndef CONFIG_SMP
4093 physid_set_mask_of_physid(boot_cpu_physical_apicid, 4047 physid_set_mask_of_physid(boot_cpu_physical_apicid,
4094 &phys_cpu_present_map); 4048 &phys_cpu_present_map);
4095#endif 4049#endif
4096 /* Make sure the irq descriptor is set up */
4097 cfg = alloc_irq_and_cfg_at(0, 0);
4098
4099 setup_local_APIC(); 4050 setup_local_APIC();
4100 4051
4101 add_pin_to_irq_node(cfg, 0, 0, 0); 4052 io_apic_setup_irq_pin(0, 0, &attr);
4102 set_irq_chip_and_handler_name(0, &ioapic_chip, handle_edge_irq, "edge"); 4053 irq_set_chip_and_handler_name(0, &ioapic_chip, handle_edge_irq,
4103 4054 "edge");
4104 setup_ioapic_irq(0, 0, 0, cfg, 0, 0);
4105} 4055}
diff --git a/arch/x86/kernel/apic/ipi.c b/arch/x86/kernel/apic/ipi.c
index 08385e090a6f..cce91bf26676 100644
--- a/arch/x86/kernel/apic/ipi.c
+++ b/arch/x86/kernel/apic/ipi.c
@@ -56,6 +56,8 @@ void default_send_IPI_mask_allbutself_phys(const struct cpumask *mask,
56 local_irq_restore(flags); 56 local_irq_restore(flags);
57} 57}
58 58
59#ifdef CONFIG_X86_32
60
59void default_send_IPI_mask_sequence_logical(const struct cpumask *mask, 61void default_send_IPI_mask_sequence_logical(const struct cpumask *mask,
60 int vector) 62 int vector)
61{ 63{
@@ -71,8 +73,8 @@ void default_send_IPI_mask_sequence_logical(const struct cpumask *mask,
71 local_irq_save(flags); 73 local_irq_save(flags);
72 for_each_cpu(query_cpu, mask) 74 for_each_cpu(query_cpu, mask)
73 __default_send_IPI_dest_field( 75 __default_send_IPI_dest_field(
74 apic->cpu_to_logical_apicid(query_cpu), vector, 76 early_per_cpu(x86_cpu_to_logical_apicid, query_cpu),
75 apic->dest_logical); 77 vector, apic->dest_logical);
76 local_irq_restore(flags); 78 local_irq_restore(flags);
77} 79}
78 80
@@ -90,14 +92,12 @@ void default_send_IPI_mask_allbutself_logical(const struct cpumask *mask,
90 if (query_cpu == this_cpu) 92 if (query_cpu == this_cpu)
91 continue; 93 continue;
92 __default_send_IPI_dest_field( 94 __default_send_IPI_dest_field(
93 apic->cpu_to_logical_apicid(query_cpu), vector, 95 early_per_cpu(x86_cpu_to_logical_apicid, query_cpu),
94 apic->dest_logical); 96 vector, apic->dest_logical);
95 } 97 }
96 local_irq_restore(flags); 98 local_irq_restore(flags);
97} 99}
98 100
99#ifdef CONFIG_X86_32
100
101/* 101/*
102 * This is only used on smaller machines. 102 * This is only used on smaller machines.
103 */ 103 */
diff --git a/arch/x86/kernel/apic/numaq_32.c b/arch/x86/kernel/apic/numaq_32.c
index 960f26ab5c9f..6273eee5134b 100644
--- a/arch/x86/kernel/apic/numaq_32.c
+++ b/arch/x86/kernel/apic/numaq_32.c
@@ -373,13 +373,6 @@ static inline void numaq_ioapic_phys_id_map(physid_mask_t *phys_map, physid_mask
373 return physids_promote(0xFUL, retmap); 373 return physids_promote(0xFUL, retmap);
374} 374}
375 375
376static inline int numaq_cpu_to_logical_apicid(int cpu)
377{
378 if (cpu >= nr_cpu_ids)
379 return BAD_APICID;
380 return cpu_2_logical_apicid[cpu];
381}
382
383/* 376/*
384 * Supporting over 60 cpus on NUMA-Q requires a locality-dependent 377 * Supporting over 60 cpus on NUMA-Q requires a locality-dependent
385 * cpu to APIC ID relation to properly interact with the intelligent 378 * cpu to APIC ID relation to properly interact with the intelligent
@@ -398,6 +391,15 @@ static inline int numaq_apicid_to_node(int logical_apicid)
398 return logical_apicid >> 4; 391 return logical_apicid >> 4;
399} 392}
400 393
394static int numaq_numa_cpu_node(int cpu)
395{
396 int logical_apicid = early_per_cpu(x86_cpu_to_logical_apicid, cpu);
397
398 if (logical_apicid != BAD_APICID)
399 return numaq_apicid_to_node(logical_apicid);
400 return NUMA_NO_NODE;
401}
402
401static void numaq_apicid_to_cpu_present(int logical_apicid, physid_mask_t *retmap) 403static void numaq_apicid_to_cpu_present(int logical_apicid, physid_mask_t *retmap)
402{ 404{
403 int node = numaq_apicid_to_node(logical_apicid); 405 int node = numaq_apicid_to_node(logical_apicid);
@@ -508,8 +510,6 @@ struct apic __refdata apic_numaq = {
508 .ioapic_phys_id_map = numaq_ioapic_phys_id_map, 510 .ioapic_phys_id_map = numaq_ioapic_phys_id_map,
509 .setup_apic_routing = numaq_setup_apic_routing, 511 .setup_apic_routing = numaq_setup_apic_routing,
510 .multi_timer_check = numaq_multi_timer_check, 512 .multi_timer_check = numaq_multi_timer_check,
511 .apicid_to_node = numaq_apicid_to_node,
512 .cpu_to_logical_apicid = numaq_cpu_to_logical_apicid,
513 .cpu_present_to_apicid = numaq_cpu_present_to_apicid, 513 .cpu_present_to_apicid = numaq_cpu_present_to_apicid,
514 .apicid_to_cpu_present = numaq_apicid_to_cpu_present, 514 .apicid_to_cpu_present = numaq_apicid_to_cpu_present,
515 .setup_portio_remap = numaq_setup_portio_remap, 515 .setup_portio_remap = numaq_setup_portio_remap,
@@ -547,4 +547,7 @@ struct apic __refdata apic_numaq = {
547 .icr_write = native_apic_icr_write, 547 .icr_write = native_apic_icr_write,
548 .wait_icr_idle = native_apic_wait_icr_idle, 548 .wait_icr_idle = native_apic_wait_icr_idle,
549 .safe_wait_icr_idle = native_safe_apic_wait_icr_idle, 549 .safe_wait_icr_idle = native_safe_apic_wait_icr_idle,
550
551 .x86_32_early_logical_apicid = noop_x86_32_early_logical_apicid,
552 .x86_32_numa_cpu_node = numaq_numa_cpu_node,
550}; 553};
diff --git a/arch/x86/kernel/apic/probe_32.c b/arch/x86/kernel/apic/probe_32.c
index 99d2fe016084..fc84c7b61108 100644
--- a/arch/x86/kernel/apic/probe_32.c
+++ b/arch/x86/kernel/apic/probe_32.c
@@ -77,6 +77,11 @@ void __init default_setup_apic_routing(void)
77 apic->setup_apic_routing(); 77 apic->setup_apic_routing();
78} 78}
79 79
80static int default_x86_32_early_logical_apicid(int cpu)
81{
82 return 1 << cpu;
83}
84
80static void setup_apic_flat_routing(void) 85static void setup_apic_flat_routing(void)
81{ 86{
82#ifdef CONFIG_X86_IO_APIC 87#ifdef CONFIG_X86_IO_APIC
@@ -130,8 +135,6 @@ struct apic apic_default = {
130 .ioapic_phys_id_map = default_ioapic_phys_id_map, 135 .ioapic_phys_id_map = default_ioapic_phys_id_map,
131 .setup_apic_routing = setup_apic_flat_routing, 136 .setup_apic_routing = setup_apic_flat_routing,
132 .multi_timer_check = NULL, 137 .multi_timer_check = NULL,
133 .apicid_to_node = default_apicid_to_node,
134 .cpu_to_logical_apicid = default_cpu_to_logical_apicid,
135 .cpu_present_to_apicid = default_cpu_present_to_apicid, 138 .cpu_present_to_apicid = default_cpu_present_to_apicid,
136 .apicid_to_cpu_present = physid_set_mask_of_physid, 139 .apicid_to_cpu_present = physid_set_mask_of_physid,
137 .setup_portio_remap = NULL, 140 .setup_portio_remap = NULL,
@@ -167,6 +170,9 @@ struct apic apic_default = {
167 .icr_write = native_apic_icr_write, 170 .icr_write = native_apic_icr_write,
168 .wait_icr_idle = native_apic_wait_icr_idle, 171 .wait_icr_idle = native_apic_wait_icr_idle,
169 .safe_wait_icr_idle = native_safe_apic_wait_icr_idle, 172 .safe_wait_icr_idle = native_safe_apic_wait_icr_idle,
173
174 .x86_32_early_logical_apicid = default_x86_32_early_logical_apicid,
175 .x86_32_numa_cpu_node = default_x86_32_numa_cpu_node,
170}; 176};
171 177
172extern struct apic apic_numaq; 178extern struct apic apic_numaq;
diff --git a/arch/x86/kernel/apic/summit_32.c b/arch/x86/kernel/apic/summit_32.c
index 9b419263d90d..e4b8059b414a 100644
--- a/arch/x86/kernel/apic/summit_32.c
+++ b/arch/x86/kernel/apic/summit_32.c
@@ -194,11 +194,10 @@ static unsigned long summit_check_apicid_present(int bit)
194 return 1; 194 return 1;
195} 195}
196 196
197static void summit_init_apic_ldr(void) 197static int summit_early_logical_apicid(int cpu)
198{ 198{
199 unsigned long val, id;
200 int count = 0; 199 int count = 0;
201 u8 my_id = (u8)hard_smp_processor_id(); 200 u8 my_id = early_per_cpu(x86_cpu_to_apicid, cpu);
202 u8 my_cluster = APIC_CLUSTER(my_id); 201 u8 my_cluster = APIC_CLUSTER(my_id);
203#ifdef CONFIG_SMP 202#ifdef CONFIG_SMP
204 u8 lid; 203 u8 lid;
@@ -206,7 +205,7 @@ static void summit_init_apic_ldr(void)
206 205
207 /* Create logical APIC IDs by counting CPUs already in cluster. */ 206 /* Create logical APIC IDs by counting CPUs already in cluster. */
208 for (count = 0, i = nr_cpu_ids; --i >= 0; ) { 207 for (count = 0, i = nr_cpu_ids; --i >= 0; ) {
209 lid = cpu_2_logical_apicid[i]; 208 lid = early_per_cpu(x86_cpu_to_logical_apicid, i);
210 if (lid != BAD_APICID && APIC_CLUSTER(lid) == my_cluster) 209 if (lid != BAD_APICID && APIC_CLUSTER(lid) == my_cluster)
211 ++count; 210 ++count;
212 } 211 }
@@ -214,7 +213,15 @@ static void summit_init_apic_ldr(void)
214 /* We only have a 4 wide bitmap in cluster mode. If a deranged 213 /* We only have a 4 wide bitmap in cluster mode. If a deranged
215 * BIOS puts 5 CPUs in one APIC cluster, we're hosed. */ 214 * BIOS puts 5 CPUs in one APIC cluster, we're hosed. */
216 BUG_ON(count >= XAPIC_DEST_CPUS_SHIFT); 215 BUG_ON(count >= XAPIC_DEST_CPUS_SHIFT);
217 id = my_cluster | (1UL << count); 216 return my_cluster | (1UL << count);
217}
218
219static void summit_init_apic_ldr(void)
220{
221 int cpu = smp_processor_id();
222 unsigned long id = early_per_cpu(x86_cpu_to_logical_apicid, cpu);
223 unsigned long val;
224
218 apic_write(APIC_DFR, SUMMIT_APIC_DFR_VALUE); 225 apic_write(APIC_DFR, SUMMIT_APIC_DFR_VALUE);
219 val = apic_read(APIC_LDR) & ~APIC_LDR_MASK; 226 val = apic_read(APIC_LDR) & ~APIC_LDR_MASK;
220 val |= SET_APIC_LOGICAL_ID(id); 227 val |= SET_APIC_LOGICAL_ID(id);
@@ -232,27 +239,6 @@ static void summit_setup_apic_routing(void)
232 nr_ioapics); 239 nr_ioapics);
233} 240}
234 241
235static int summit_apicid_to_node(int logical_apicid)
236{
237#ifdef CONFIG_SMP
238 return apicid_2_node[hard_smp_processor_id()];
239#else
240 return 0;
241#endif
242}
243
244/* Mapping from cpu number to logical apicid */
245static inline int summit_cpu_to_logical_apicid(int cpu)
246{
247#ifdef CONFIG_SMP
248 if (cpu >= nr_cpu_ids)
249 return BAD_APICID;
250 return cpu_2_logical_apicid[cpu];
251#else
252 return logical_smp_processor_id();
253#endif
254}
255
256static int summit_cpu_present_to_apicid(int mps_cpu) 242static int summit_cpu_present_to_apicid(int mps_cpu)
257{ 243{
258 if (mps_cpu < nr_cpu_ids) 244 if (mps_cpu < nr_cpu_ids)
@@ -286,7 +272,7 @@ static unsigned int summit_cpu_mask_to_apicid(const struct cpumask *cpumask)
286 * The cpus in the mask must all be on the apic cluster. 272 * The cpus in the mask must all be on the apic cluster.
287 */ 273 */
288 for_each_cpu(cpu, cpumask) { 274 for_each_cpu(cpu, cpumask) {
289 int new_apicid = summit_cpu_to_logical_apicid(cpu); 275 int new_apicid = early_per_cpu(x86_cpu_to_logical_apicid, cpu);
290 276
291 if (round && APIC_CLUSTER(apicid) != APIC_CLUSTER(new_apicid)) { 277 if (round && APIC_CLUSTER(apicid) != APIC_CLUSTER(new_apicid)) {
292 printk("%s: Not a valid mask!\n", __func__); 278 printk("%s: Not a valid mask!\n", __func__);
@@ -301,7 +287,7 @@ static unsigned int summit_cpu_mask_to_apicid(const struct cpumask *cpumask)
301static unsigned int summit_cpu_mask_to_apicid_and(const struct cpumask *inmask, 287static unsigned int summit_cpu_mask_to_apicid_and(const struct cpumask *inmask,
302 const struct cpumask *andmask) 288 const struct cpumask *andmask)
303{ 289{
304 int apicid = summit_cpu_to_logical_apicid(0); 290 int apicid = early_per_cpu(x86_cpu_to_logical_apicid, 0);
305 cpumask_var_t cpumask; 291 cpumask_var_t cpumask;
306 292
307 if (!alloc_cpumask_var(&cpumask, GFP_ATOMIC)) 293 if (!alloc_cpumask_var(&cpumask, GFP_ATOMIC))
@@ -528,8 +514,6 @@ struct apic apic_summit = {
528 .ioapic_phys_id_map = summit_ioapic_phys_id_map, 514 .ioapic_phys_id_map = summit_ioapic_phys_id_map,
529 .setup_apic_routing = summit_setup_apic_routing, 515 .setup_apic_routing = summit_setup_apic_routing,
530 .multi_timer_check = NULL, 516 .multi_timer_check = NULL,
531 .apicid_to_node = summit_apicid_to_node,
532 .cpu_to_logical_apicid = summit_cpu_to_logical_apicid,
533 .cpu_present_to_apicid = summit_cpu_present_to_apicid, 517 .cpu_present_to_apicid = summit_cpu_present_to_apicid,
534 .apicid_to_cpu_present = summit_apicid_to_cpu_present, 518 .apicid_to_cpu_present = summit_apicid_to_cpu_present,
535 .setup_portio_remap = NULL, 519 .setup_portio_remap = NULL,
@@ -565,4 +549,7 @@ struct apic apic_summit = {
565 .icr_write = native_apic_icr_write, 549 .icr_write = native_apic_icr_write,
566 .wait_icr_idle = native_apic_wait_icr_idle, 550 .wait_icr_idle = native_apic_wait_icr_idle,
567 .safe_wait_icr_idle = native_safe_apic_wait_icr_idle, 551 .safe_wait_icr_idle = native_safe_apic_wait_icr_idle,
552
553 .x86_32_early_logical_apicid = summit_early_logical_apicid,
554 .x86_32_numa_cpu_node = default_x86_32_numa_cpu_node,
568}; 555};
diff --git a/arch/x86/kernel/apic/x2apic_cluster.c b/arch/x86/kernel/apic/x2apic_cluster.c
index cf69c59f4910..90949bbd566d 100644
--- a/arch/x86/kernel/apic/x2apic_cluster.c
+++ b/arch/x86/kernel/apic/x2apic_cluster.c
@@ -206,8 +206,6 @@ struct apic apic_x2apic_cluster = {
206 .ioapic_phys_id_map = NULL, 206 .ioapic_phys_id_map = NULL,
207 .setup_apic_routing = NULL, 207 .setup_apic_routing = NULL,
208 .multi_timer_check = NULL, 208 .multi_timer_check = NULL,
209 .apicid_to_node = NULL,
210 .cpu_to_logical_apicid = NULL,
211 .cpu_present_to_apicid = default_cpu_present_to_apicid, 209 .cpu_present_to_apicid = default_cpu_present_to_apicid,
212 .apicid_to_cpu_present = NULL, 210 .apicid_to_cpu_present = NULL,
213 .setup_portio_remap = NULL, 211 .setup_portio_remap = NULL,
diff --git a/arch/x86/kernel/apic/x2apic_phys.c b/arch/x86/kernel/apic/x2apic_phys.c
index 8972f38c5ced..c7e6d6645bf4 100644
--- a/arch/x86/kernel/apic/x2apic_phys.c
+++ b/arch/x86/kernel/apic/x2apic_phys.c
@@ -195,8 +195,6 @@ struct apic apic_x2apic_phys = {
195 .ioapic_phys_id_map = NULL, 195 .ioapic_phys_id_map = NULL,
196 .setup_apic_routing = NULL, 196 .setup_apic_routing = NULL,
197 .multi_timer_check = NULL, 197 .multi_timer_check = NULL,
198 .apicid_to_node = NULL,
199 .cpu_to_logical_apicid = NULL,
200 .cpu_present_to_apicid = default_cpu_present_to_apicid, 198 .cpu_present_to_apicid = default_cpu_present_to_apicid,
201 .apicid_to_cpu_present = NULL, 199 .apicid_to_cpu_present = NULL,
202 .setup_portio_remap = NULL, 200 .setup_portio_remap = NULL,
diff --git a/arch/x86/kernel/apic/x2apic_uv_x.c b/arch/x86/kernel/apic/x2apic_uv_x.c
index bd16b58b8850..3c289281394c 100644
--- a/arch/x86/kernel/apic/x2apic_uv_x.c
+++ b/arch/x86/kernel/apic/x2apic_uv_x.c
@@ -338,8 +338,6 @@ struct apic __refdata apic_x2apic_uv_x = {
338 .ioapic_phys_id_map = NULL, 338 .ioapic_phys_id_map = NULL,
339 .setup_apic_routing = NULL, 339 .setup_apic_routing = NULL,
340 .multi_timer_check = NULL, 340 .multi_timer_check = NULL,
341 .apicid_to_node = NULL,
342 .cpu_to_logical_apicid = NULL,
343 .cpu_present_to_apicid = default_cpu_present_to_apicid, 341 .cpu_present_to_apicid = default_cpu_present_to_apicid,
344 .apicid_to_cpu_present = NULL, 342 .apicid_to_cpu_present = NULL,
345 .setup_portio_remap = NULL, 343 .setup_portio_remap = NULL,
diff --git a/arch/x86/kernel/asm-offsets.c b/arch/x86/kernel/asm-offsets.c
index cfa82c899f47..4f13fafc5264 100644
--- a/arch/x86/kernel/asm-offsets.c
+++ b/arch/x86/kernel/asm-offsets.c
@@ -1,5 +1,70 @@
1/*
2 * Generate definitions needed by assembly language modules.
3 * This code generates raw asm output which is post-processed to extract
4 * and format the required data.
5 */
6#define COMPILE_OFFSETS
7
8#include <linux/crypto.h>
9#include <linux/sched.h>
10#include <linux/stddef.h>
11#include <linux/hardirq.h>
12#include <linux/suspend.h>
13#include <linux/kbuild.h>
14#include <asm/processor.h>
15#include <asm/thread_info.h>
16#include <asm/sigframe.h>
17#include <asm/bootparam.h>
18#include <asm/suspend.h>
19
20#ifdef CONFIG_XEN
21#include <xen/interface/xen.h>
22#endif
23
1#ifdef CONFIG_X86_32 24#ifdef CONFIG_X86_32
2# include "asm-offsets_32.c" 25# include "asm-offsets_32.c"
3#else 26#else
4# include "asm-offsets_64.c" 27# include "asm-offsets_64.c"
5#endif 28#endif
29
30void common(void) {
31 BLANK();
32 OFFSET(TI_flags, thread_info, flags);
33 OFFSET(TI_status, thread_info, status);
34 OFFSET(TI_addr_limit, thread_info, addr_limit);
35 OFFSET(TI_preempt_count, thread_info, preempt_count);
36
37 BLANK();
38 OFFSET(crypto_tfm_ctx_offset, crypto_tfm, __crt_ctx);
39
40 BLANK();
41 OFFSET(pbe_address, pbe, address);
42 OFFSET(pbe_orig_address, pbe, orig_address);
43 OFFSET(pbe_next, pbe, next);
44
45#ifdef CONFIG_PARAVIRT
46 BLANK();
47 OFFSET(PARAVIRT_enabled, pv_info, paravirt_enabled);
48 OFFSET(PARAVIRT_PATCH_pv_cpu_ops, paravirt_patch_template, pv_cpu_ops);
49 OFFSET(PARAVIRT_PATCH_pv_irq_ops, paravirt_patch_template, pv_irq_ops);
50 OFFSET(PV_IRQ_irq_disable, pv_irq_ops, irq_disable);
51 OFFSET(PV_IRQ_irq_enable, pv_irq_ops, irq_enable);
52 OFFSET(PV_CPU_iret, pv_cpu_ops, iret);
53 OFFSET(PV_CPU_irq_enable_sysexit, pv_cpu_ops, irq_enable_sysexit);
54 OFFSET(PV_CPU_read_cr0, pv_cpu_ops, read_cr0);
55 OFFSET(PV_MMU_read_cr2, pv_mmu_ops, read_cr2);
56#endif
57
58#ifdef CONFIG_XEN
59 BLANK();
60 OFFSET(XEN_vcpu_info_mask, vcpu_info, evtchn_upcall_mask);
61 OFFSET(XEN_vcpu_info_pending, vcpu_info, evtchn_upcall_pending);
62#endif
63
64 BLANK();
65 OFFSET(BP_scratch, boot_params, scratch);
66 OFFSET(BP_loadflags, boot_params, hdr.loadflags);
67 OFFSET(BP_hardware_subarch, boot_params, hdr.hardware_subarch);
68 OFFSET(BP_version, boot_params, hdr.version);
69 OFFSET(BP_kernel_alignment, boot_params, hdr.kernel_alignment);
70}
diff --git a/arch/x86/kernel/asm-offsets_32.c b/arch/x86/kernel/asm-offsets_32.c
index 1a4088dda37a..c29d631af6fc 100644
--- a/arch/x86/kernel/asm-offsets_32.c
+++ b/arch/x86/kernel/asm-offsets_32.c
@@ -1,26 +1,4 @@
1/*
2 * Generate definitions needed by assembly language modules.
3 * This code generates raw asm output which is post-processed
4 * to extract and format the required data.
5 */
6
7#include <linux/crypto.h>
8#include <linux/sched.h>
9#include <linux/signal.h>
10#include <linux/personality.h>
11#include <linux/suspend.h>
12#include <linux/kbuild.h>
13#include <asm/ucontext.h> 1#include <asm/ucontext.h>
14#include <asm/sigframe.h>
15#include <asm/pgtable.h>
16#include <asm/fixmap.h>
17#include <asm/processor.h>
18#include <asm/thread_info.h>
19#include <asm/bootparam.h>
20#include <asm/elf.h>
21#include <asm/suspend.h>
22
23#include <xen/interface/xen.h>
24 2
25#include <linux/lguest.h> 3#include <linux/lguest.h>
26#include "../../../drivers/lguest/lg.h" 4#include "../../../drivers/lguest/lg.h"
@@ -51,21 +29,10 @@ void foo(void)
51 OFFSET(CPUINFO_x86_vendor_id, cpuinfo_x86, x86_vendor_id); 29 OFFSET(CPUINFO_x86_vendor_id, cpuinfo_x86, x86_vendor_id);
52 BLANK(); 30 BLANK();
53 31
54 OFFSET(TI_task, thread_info, task);
55 OFFSET(TI_exec_domain, thread_info, exec_domain);
56 OFFSET(TI_flags, thread_info, flags);
57 OFFSET(TI_status, thread_info, status);
58 OFFSET(TI_preempt_count, thread_info, preempt_count);
59 OFFSET(TI_addr_limit, thread_info, addr_limit);
60 OFFSET(TI_restart_block, thread_info, restart_block);
61 OFFSET(TI_sysenter_return, thread_info, sysenter_return); 32 OFFSET(TI_sysenter_return, thread_info, sysenter_return);
62 OFFSET(TI_cpu, thread_info, cpu); 33 OFFSET(TI_cpu, thread_info, cpu);
63 BLANK(); 34 BLANK();
64 35
65 OFFSET(GDS_size, desc_ptr, size);
66 OFFSET(GDS_address, desc_ptr, address);
67 BLANK();
68
69 OFFSET(PT_EBX, pt_regs, bx); 36 OFFSET(PT_EBX, pt_regs, bx);
70 OFFSET(PT_ECX, pt_regs, cx); 37 OFFSET(PT_ECX, pt_regs, cx);
71 OFFSET(PT_EDX, pt_regs, dx); 38 OFFSET(PT_EDX, pt_regs, dx);
@@ -85,42 +52,13 @@ void foo(void)
85 OFFSET(PT_OLDSS, pt_regs, ss); 52 OFFSET(PT_OLDSS, pt_regs, ss);
86 BLANK(); 53 BLANK();
87 54
88 OFFSET(EXEC_DOMAIN_handler, exec_domain, handler);
89 OFFSET(IA32_RT_SIGFRAME_sigcontext, rt_sigframe, uc.uc_mcontext); 55 OFFSET(IA32_RT_SIGFRAME_sigcontext, rt_sigframe, uc.uc_mcontext);
90 BLANK(); 56 BLANK();
91 57
92 OFFSET(pbe_address, pbe, address);
93 OFFSET(pbe_orig_address, pbe, orig_address);
94 OFFSET(pbe_next, pbe, next);
95
96 /* Offset from the sysenter stack to tss.sp0 */ 58 /* Offset from the sysenter stack to tss.sp0 */
97 DEFINE(TSS_sysenter_sp0, offsetof(struct tss_struct, x86_tss.sp0) - 59 DEFINE(TSS_sysenter_sp0, offsetof(struct tss_struct, x86_tss.sp0) -
98 sizeof(struct tss_struct)); 60 sizeof(struct tss_struct));
99 61
100 DEFINE(PAGE_SIZE_asm, PAGE_SIZE);
101 DEFINE(PAGE_SHIFT_asm, PAGE_SHIFT);
102 DEFINE(THREAD_SIZE_asm, THREAD_SIZE);
103
104 OFFSET(crypto_tfm_ctx_offset, crypto_tfm, __crt_ctx);
105
106#ifdef CONFIG_PARAVIRT
107 BLANK();
108 OFFSET(PARAVIRT_enabled, pv_info, paravirt_enabled);
109 OFFSET(PARAVIRT_PATCH_pv_cpu_ops, paravirt_patch_template, pv_cpu_ops);
110 OFFSET(PARAVIRT_PATCH_pv_irq_ops, paravirt_patch_template, pv_irq_ops);
111 OFFSET(PV_IRQ_irq_disable, pv_irq_ops, irq_disable);
112 OFFSET(PV_IRQ_irq_enable, pv_irq_ops, irq_enable);
113 OFFSET(PV_CPU_iret, pv_cpu_ops, iret);
114 OFFSET(PV_CPU_irq_enable_sysexit, pv_cpu_ops, irq_enable_sysexit);
115 OFFSET(PV_CPU_read_cr0, pv_cpu_ops, read_cr0);
116#endif
117
118#ifdef CONFIG_XEN
119 BLANK();
120 OFFSET(XEN_vcpu_info_mask, vcpu_info, evtchn_upcall_mask);
121 OFFSET(XEN_vcpu_info_pending, vcpu_info, evtchn_upcall_pending);
122#endif
123
124#if defined(CONFIG_LGUEST) || defined(CONFIG_LGUEST_GUEST) || defined(CONFIG_LGUEST_MODULE) 62#if defined(CONFIG_LGUEST) || defined(CONFIG_LGUEST_GUEST) || defined(CONFIG_LGUEST_MODULE)
125 BLANK(); 63 BLANK();
126 OFFSET(LGUEST_DATA_irq_enabled, lguest_data, irq_enabled); 64 OFFSET(LGUEST_DATA_irq_enabled, lguest_data, irq_enabled);
@@ -139,11 +77,4 @@ void foo(void)
139 OFFSET(LGUEST_PAGES_regs_errcode, lguest_pages, regs.errcode); 77 OFFSET(LGUEST_PAGES_regs_errcode, lguest_pages, regs.errcode);
140 OFFSET(LGUEST_PAGES_regs, lguest_pages, regs); 78 OFFSET(LGUEST_PAGES_regs, lguest_pages, regs);
141#endif 79#endif
142
143 BLANK();
144 OFFSET(BP_scratch, boot_params, scratch);
145 OFFSET(BP_loadflags, boot_params, hdr.loadflags);
146 OFFSET(BP_hardware_subarch, boot_params, hdr.hardware_subarch);
147 OFFSET(BP_version, boot_params, hdr.version);
148 OFFSET(BP_kernel_alignment, boot_params, hdr.kernel_alignment);
149} 80}
diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c
index 4a6aeedcd965..e72a1194af22 100644
--- a/arch/x86/kernel/asm-offsets_64.c
+++ b/arch/x86/kernel/asm-offsets_64.c
@@ -1,27 +1,4 @@
1/*
2 * Generate definitions needed by assembly language modules.
3 * This code generates raw asm output which is post-processed to extract
4 * and format the required data.
5 */
6#define COMPILE_OFFSETS
7
8#include <linux/crypto.h>
9#include <linux/sched.h>
10#include <linux/stddef.h>
11#include <linux/errno.h>
12#include <linux/hardirq.h>
13#include <linux/suspend.h>
14#include <linux/kbuild.h>
15#include <asm/processor.h>
16#include <asm/segment.h>
17#include <asm/thread_info.h>
18#include <asm/ia32.h> 1#include <asm/ia32.h>
19#include <asm/bootparam.h>
20#include <asm/suspend.h>
21
22#include <xen/interface/xen.h>
23
24#include <asm/sigframe.h>
25 2
26#define __NO_STUBS 1 3#define __NO_STUBS 1
27#undef __SYSCALL 4#undef __SYSCALL
@@ -33,41 +10,19 @@ static char syscalls[] = {
33 10
34int main(void) 11int main(void)
35{ 12{
36#define ENTRY(entry) DEFINE(tsk_ ## entry, offsetof(struct task_struct, entry))
37 ENTRY(state);
38 ENTRY(flags);
39 ENTRY(pid);
40 BLANK();
41#undef ENTRY
42#define ENTRY(entry) DEFINE(TI_ ## entry, offsetof(struct thread_info, entry))
43 ENTRY(flags);
44 ENTRY(addr_limit);
45 ENTRY(preempt_count);
46 ENTRY(status);
47#ifdef CONFIG_IA32_EMULATION
48 ENTRY(sysenter_return);
49#endif
50 BLANK();
51#undef ENTRY
52#ifdef CONFIG_PARAVIRT 13#ifdef CONFIG_PARAVIRT
53 BLANK();
54 OFFSET(PARAVIRT_enabled, pv_info, paravirt_enabled);
55 OFFSET(PARAVIRT_PATCH_pv_cpu_ops, paravirt_patch_template, pv_cpu_ops);
56 OFFSET(PARAVIRT_PATCH_pv_irq_ops, paravirt_patch_template, pv_irq_ops);
57 OFFSET(PV_IRQ_irq_disable, pv_irq_ops, irq_disable);
58 OFFSET(PV_IRQ_irq_enable, pv_irq_ops, irq_enable);
59 OFFSET(PV_IRQ_adjust_exception_frame, pv_irq_ops, adjust_exception_frame); 14 OFFSET(PV_IRQ_adjust_exception_frame, pv_irq_ops, adjust_exception_frame);
60 OFFSET(PV_CPU_iret, pv_cpu_ops, iret);
61 OFFSET(PV_CPU_usergs_sysret32, pv_cpu_ops, usergs_sysret32); 15 OFFSET(PV_CPU_usergs_sysret32, pv_cpu_ops, usergs_sysret32);
62 OFFSET(PV_CPU_usergs_sysret64, pv_cpu_ops, usergs_sysret64); 16 OFFSET(PV_CPU_usergs_sysret64, pv_cpu_ops, usergs_sysret64);
63 OFFSET(PV_CPU_irq_enable_sysexit, pv_cpu_ops, irq_enable_sysexit);
64 OFFSET(PV_CPU_swapgs, pv_cpu_ops, swapgs); 17 OFFSET(PV_CPU_swapgs, pv_cpu_ops, swapgs);
65 OFFSET(PV_MMU_read_cr2, pv_mmu_ops, read_cr2); 18 BLANK();
66#endif 19#endif
67 20
68
69#ifdef CONFIG_IA32_EMULATION 21#ifdef CONFIG_IA32_EMULATION
70#define ENTRY(entry) DEFINE(IA32_SIGCONTEXT_ ## entry, offsetof(struct sigcontext_ia32, entry)) 22 OFFSET(TI_sysenter_return, thread_info, sysenter_return);
23 BLANK();
24
25#define ENTRY(entry) OFFSET(IA32_SIGCONTEXT_ ## entry, sigcontext_ia32, entry)
71 ENTRY(ax); 26 ENTRY(ax);
72 ENTRY(bx); 27 ENTRY(bx);
73 ENTRY(cx); 28 ENTRY(cx);
@@ -79,15 +34,12 @@ int main(void)
79 ENTRY(ip); 34 ENTRY(ip);
80 BLANK(); 35 BLANK();
81#undef ENTRY 36#undef ENTRY
82 DEFINE(IA32_RT_SIGFRAME_sigcontext, 37
83 offsetof (struct rt_sigframe_ia32, uc.uc_mcontext)); 38 OFFSET(IA32_RT_SIGFRAME_sigcontext, rt_sigframe_ia32, uc.uc_mcontext);
84 BLANK(); 39 BLANK();
85#endif 40#endif
86 DEFINE(pbe_address, offsetof(struct pbe, address)); 41
87 DEFINE(pbe_orig_address, offsetof(struct pbe, orig_address)); 42#define ENTRY(entry) OFFSET(pt_regs_ ## entry, pt_regs, entry)
88 DEFINE(pbe_next, offsetof(struct pbe, next));
89 BLANK();
90#define ENTRY(entry) DEFINE(pt_regs_ ## entry, offsetof(struct pt_regs, entry))
91 ENTRY(bx); 43 ENTRY(bx);
92 ENTRY(bx); 44 ENTRY(bx);
93 ENTRY(cx); 45 ENTRY(cx);
@@ -107,7 +59,8 @@ int main(void)
107 ENTRY(flags); 59 ENTRY(flags);
108 BLANK(); 60 BLANK();
109#undef ENTRY 61#undef ENTRY
110#define ENTRY(entry) DEFINE(saved_context_ ## entry, offsetof(struct saved_context, entry)) 62
63#define ENTRY(entry) OFFSET(saved_context_ ## entry, saved_context, entry)
111 ENTRY(cr0); 64 ENTRY(cr0);
112 ENTRY(cr2); 65 ENTRY(cr2);
113 ENTRY(cr3); 66 ENTRY(cr3);
@@ -115,26 +68,11 @@ int main(void)
115 ENTRY(cr8); 68 ENTRY(cr8);
116 BLANK(); 69 BLANK();
117#undef ENTRY 70#undef ENTRY
118 DEFINE(TSS_ist, offsetof(struct tss_struct, x86_tss.ist));
119 BLANK();
120 DEFINE(crypto_tfm_ctx_offset, offsetof(struct crypto_tfm, __crt_ctx));
121 BLANK();
122 DEFINE(__NR_syscall_max, sizeof(syscalls) - 1);
123 71
72 OFFSET(TSS_ist, tss_struct, x86_tss.ist);
124 BLANK(); 73 BLANK();
125 OFFSET(BP_scratch, boot_params, scratch);
126 OFFSET(BP_loadflags, boot_params, hdr.loadflags);
127 OFFSET(BP_hardware_subarch, boot_params, hdr.hardware_subarch);
128 OFFSET(BP_version, boot_params, hdr.version);
129 OFFSET(BP_kernel_alignment, boot_params, hdr.kernel_alignment);
130 74
131 BLANK(); 75 DEFINE(__NR_syscall_max, sizeof(syscalls) - 1);
132 DEFINE(PAGE_SIZE_asm, PAGE_SIZE); 76
133#ifdef CONFIG_XEN
134 BLANK();
135 OFFSET(XEN_vcpu_info_mask, vcpu_info, evtchn_upcall_mask);
136 OFFSET(XEN_vcpu_info_pending, vcpu_info, evtchn_upcall_pending);
137#undef ENTRY
138#endif
139 return 0; 77 return 0;
140} 78}
diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c
index 58f1b012e1c8..3ecece0217ef 100644
--- a/arch/x86/kernel/cpu/amd.c
+++ b/arch/x86/kernel/cpu/amd.c
@@ -233,18 +233,22 @@ static void __cpuinit init_amd_k7(struct cpuinfo_x86 *c)
233} 233}
234#endif 234#endif
235 235
236#if defined(CONFIG_NUMA) && defined(CONFIG_X86_64) 236#ifdef CONFIG_NUMA
237/*
238 * To workaround broken NUMA config. Read the comment in
239 * srat_detect_node().
240 */
237static int __cpuinit nearby_node(int apicid) 241static int __cpuinit nearby_node(int apicid)
238{ 242{
239 int i, node; 243 int i, node;
240 244
241 for (i = apicid - 1; i >= 0; i--) { 245 for (i = apicid - 1; i >= 0; i--) {
242 node = apicid_to_node[i]; 246 node = __apicid_to_node[i];
243 if (node != NUMA_NO_NODE && node_online(node)) 247 if (node != NUMA_NO_NODE && node_online(node))
244 return node; 248 return node;
245 } 249 }
246 for (i = apicid + 1; i < MAX_LOCAL_APIC; i++) { 250 for (i = apicid + 1; i < MAX_LOCAL_APIC; i++) {
247 node = apicid_to_node[i]; 251 node = __apicid_to_node[i];
248 if (node != NUMA_NO_NODE && node_online(node)) 252 if (node != NUMA_NO_NODE && node_online(node))
249 return node; 253 return node;
250 } 254 }
@@ -338,31 +342,40 @@ EXPORT_SYMBOL_GPL(amd_get_nb_id);
338 342
339static void __cpuinit srat_detect_node(struct cpuinfo_x86 *c) 343static void __cpuinit srat_detect_node(struct cpuinfo_x86 *c)
340{ 344{
341#if defined(CONFIG_NUMA) && defined(CONFIG_X86_64) 345#ifdef CONFIG_NUMA
342 int cpu = smp_processor_id(); 346 int cpu = smp_processor_id();
343 int node; 347 int node;
344 unsigned apicid = c->apicid; 348 unsigned apicid = c->apicid;
345 349
346 node = per_cpu(cpu_llc_id, cpu); 350 node = numa_cpu_node(cpu);
351 if (node == NUMA_NO_NODE)
352 node = per_cpu(cpu_llc_id, cpu);
347 353
348 if (apicid_to_node[apicid] != NUMA_NO_NODE)
349 node = apicid_to_node[apicid];
350 if (!node_online(node)) { 354 if (!node_online(node)) {
351 /* Two possibilities here: 355 /*
352 - The CPU is missing memory and no node was created. 356 * Two possibilities here:
353 In that case try picking one from a nearby CPU 357 *
354 - The APIC IDs differ from the HyperTransport node IDs 358 * - The CPU is missing memory and no node was created. In
355 which the K8 northbridge parsing fills in. 359 * that case try picking one from a nearby CPU.
356 Assume they are all increased by a constant offset, 360 *
357 but in the same order as the HT nodeids. 361 * - The APIC IDs differ from the HyperTransport node IDs
358 If that doesn't result in a usable node fall back to the 362 * which the K8 northbridge parsing fills in. Assume
359 path for the previous case. */ 363 * they are all increased by a constant offset, but in
360 364 * the same order as the HT nodeids. If that doesn't
365 * result in a usable node fall back to the path for the
366 * previous case.
367 *
368 * This workaround operates directly on the mapping between
369 * APIC ID and NUMA node, assuming certain relationship
370 * between APIC ID, HT node ID and NUMA topology. As going
371 * through CPU mapping may alter the outcome, directly
372 * access __apicid_to_node[].
373 */
361 int ht_nodeid = c->initial_apicid; 374 int ht_nodeid = c->initial_apicid;
362 375
363 if (ht_nodeid >= 0 && 376 if (ht_nodeid >= 0 &&
364 apicid_to_node[ht_nodeid] != NUMA_NO_NODE) 377 __apicid_to_node[ht_nodeid] != NUMA_NO_NODE)
365 node = apicid_to_node[ht_nodeid]; 378 node = __apicid_to_node[ht_nodeid];
366 /* Pick a nearby node */ 379 /* Pick a nearby node */
367 if (!node_online(node)) 380 if (!node_online(node))
368 node = nearby_node(apicid); 381 node = nearby_node(apicid);
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index 1d59834396bd..e2ced0074a45 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -675,7 +675,7 @@ void __init early_cpu_init(void)
675 const struct cpu_dev *const *cdev; 675 const struct cpu_dev *const *cdev;
676 int count = 0; 676 int count = 0;
677 677
678#ifdef PROCESSOR_SELECT 678#ifdef CONFIG_PROCESSOR_SELECT
679 printk(KERN_INFO "KERNEL supported cpus:\n"); 679 printk(KERN_INFO "KERNEL supported cpus:\n");
680#endif 680#endif
681 681
@@ -687,7 +687,7 @@ void __init early_cpu_init(void)
687 cpu_devs[count] = cpudev; 687 cpu_devs[count] = cpudev;
688 count++; 688 count++;
689 689
690#ifdef PROCESSOR_SELECT 690#ifdef CONFIG_PROCESSOR_SELECT
691 { 691 {
692 unsigned int j; 692 unsigned int j;
693 693
@@ -869,7 +869,7 @@ static void __cpuinit identify_cpu(struct cpuinfo_x86 *c)
869 869
870 select_idle_routine(c); 870 select_idle_routine(c);
871 871
872#if defined(CONFIG_NUMA) && defined(CONFIG_X86_64) 872#ifdef CONFIG_NUMA
873 numa_add_cpu(smp_processor_id()); 873 numa_add_cpu(smp_processor_id());
874#endif 874#endif
875} 875}
diff --git a/arch/x86/kernel/cpu/intel.c b/arch/x86/kernel/cpu/intel.c
index d16c2c53d6bf..df86bc8c859d 100644
--- a/arch/x86/kernel/cpu/intel.c
+++ b/arch/x86/kernel/cpu/intel.c
@@ -276,14 +276,13 @@ static void __cpuinit intel_workarounds(struct cpuinfo_x86 *c)
276 276
277static void __cpuinit srat_detect_node(struct cpuinfo_x86 *c) 277static void __cpuinit srat_detect_node(struct cpuinfo_x86 *c)
278{ 278{
279#if defined(CONFIG_NUMA) && defined(CONFIG_X86_64) 279#ifdef CONFIG_NUMA
280 unsigned node; 280 unsigned node;
281 int cpu = smp_processor_id(); 281 int cpu = smp_processor_id();
282 int apicid = cpu_has_apic ? hard_smp_processor_id() : c->apicid;
283 282
284 /* Don't do the funky fallback heuristics the AMD version employs 283 /* Don't do the funky fallback heuristics the AMD version employs
285 for now. */ 284 for now. */
286 node = apicid_to_node[apicid]; 285 node = numa_cpu_node(cpu);
287 if (node == NUMA_NO_NODE || !node_online(node)) { 286 if (node == NUMA_NO_NODE || !node_online(node)) {
288 /* reuse the value from init_cpu_to_node() */ 287 /* reuse the value from init_cpu_to_node() */
289 node = cpu_to_node(cpu); 288 node = cpu_to_node(cpu);
diff --git a/arch/x86/kernel/cpu/intel_cacheinfo.c b/arch/x86/kernel/cpu/intel_cacheinfo.c
index 90cc675ac746..1ce1af2899df 100644
--- a/arch/x86/kernel/cpu/intel_cacheinfo.c
+++ b/arch/x86/kernel/cpu/intel_cacheinfo.c
@@ -768,11 +768,11 @@ static void __cpuinit cache_shared_cpu_map_setup(unsigned int cpu, int index)
768 struct cpuinfo_x86 *c = &cpu_data(cpu); 768 struct cpuinfo_x86 *c = &cpu_data(cpu);
769 769
770 if ((index == 3) && (c->x86_vendor == X86_VENDOR_AMD)) { 770 if ((index == 3) && (c->x86_vendor == X86_VENDOR_AMD)) {
771 for_each_cpu(i, c->llc_shared_map) { 771 for_each_cpu(i, cpu_llc_shared_mask(cpu)) {
772 if (!per_cpu(ici_cpuid4_info, i)) 772 if (!per_cpu(ici_cpuid4_info, i))
773 continue; 773 continue;
774 this_leaf = CPUID4_INFO_IDX(i, index); 774 this_leaf = CPUID4_INFO_IDX(i, index);
775 for_each_cpu(sibling, c->llc_shared_map) { 775 for_each_cpu(sibling, cpu_llc_shared_mask(cpu)) {
776 if (!cpu_online(sibling)) 776 if (!cpu_online(sibling))
777 continue; 777 continue;
778 set_bit(sibling, this_leaf->shared_cpu_map); 778 set_bit(sibling, this_leaf->shared_cpu_map);
diff --git a/arch/x86/kernel/cpu/mcheck/mce_amd.c b/arch/x86/kernel/cpu/mcheck/mce_amd.c
index 5bf2fac52aca..167f97b5596e 100644
--- a/arch/x86/kernel/cpu/mcheck/mce_amd.c
+++ b/arch/x86/kernel/cpu/mcheck/mce_amd.c
@@ -527,15 +527,12 @@ static __cpuinit int threshold_create_bank(unsigned int cpu, unsigned int bank)
527 int i, err = 0; 527 int i, err = 0;
528 struct threshold_bank *b = NULL; 528 struct threshold_bank *b = NULL;
529 char name[32]; 529 char name[32];
530#ifdef CONFIG_SMP
531 struct cpuinfo_x86 *c = &cpu_data(cpu);
532#endif
533 530
534 sprintf(name, "threshold_bank%i", bank); 531 sprintf(name, "threshold_bank%i", bank);
535 532
536#ifdef CONFIG_SMP 533#ifdef CONFIG_SMP
537 if (cpu_data(cpu).cpu_core_id && shared_bank[bank]) { /* symlink */ 534 if (cpu_data(cpu).cpu_core_id && shared_bank[bank]) { /* symlink */
538 i = cpumask_first(c->llc_shared_map); 535 i = cpumask_first(cpu_llc_shared_mask(cpu));
539 536
540 /* first core not up yet */ 537 /* first core not up yet */
541 if (cpu_data(i).cpu_core_id) 538 if (cpu_data(i).cpu_core_id)
@@ -555,7 +552,7 @@ static __cpuinit int threshold_create_bank(unsigned int cpu, unsigned int bank)
555 if (err) 552 if (err)
556 goto out; 553 goto out;
557 554
558 cpumask_copy(b->cpus, c->llc_shared_map); 555 cpumask_copy(b->cpus, cpu_llc_shared_mask(cpu));
559 per_cpu(threshold_banks, cpu)[bank] = b; 556 per_cpu(threshold_banks, cpu)[bank] = b;
560 557
561 goto out; 558 goto out;
diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index 9d977a2ea693..26604188aa49 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -30,6 +30,7 @@
30#include <asm/stacktrace.h> 30#include <asm/stacktrace.h>
31#include <asm/nmi.h> 31#include <asm/nmi.h>
32#include <asm/compat.h> 32#include <asm/compat.h>
33#include <asm/smp.h>
33 34
34#if 0 35#if 0
35#undef wrmsrl 36#undef wrmsrl
@@ -93,6 +94,8 @@ struct amd_nb {
93 struct event_constraint event_constraints[X86_PMC_IDX_MAX]; 94 struct event_constraint event_constraints[X86_PMC_IDX_MAX];
94}; 95};
95 96
97struct intel_percore;
98
96#define MAX_LBR_ENTRIES 16 99#define MAX_LBR_ENTRIES 16
97 100
98struct cpu_hw_events { 101struct cpu_hw_events {
@@ -128,6 +131,13 @@ struct cpu_hw_events {
128 struct perf_branch_entry lbr_entries[MAX_LBR_ENTRIES]; 131 struct perf_branch_entry lbr_entries[MAX_LBR_ENTRIES];
129 132
130 /* 133 /*
134 * Intel percore register state.
135 * Coordinate shared resources between HT threads.
136 */
137 int percore_used; /* Used by this CPU? */
138 struct intel_percore *per_core;
139
140 /*
131 * AMD specific bits 141 * AMD specific bits
132 */ 142 */
133 struct amd_nb *amd_nb; 143 struct amd_nb *amd_nb;
@@ -166,8 +176,10 @@ struct cpu_hw_events {
166/* 176/*
167 * Constraint on the Event code + UMask 177 * Constraint on the Event code + UMask
168 */ 178 */
169#define PEBS_EVENT_CONSTRAINT(c, n) \ 179#define INTEL_UEVENT_CONSTRAINT(c, n) \
170 EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK) 180 EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK)
181#define PEBS_EVENT_CONSTRAINT(c, n) \
182 INTEL_UEVENT_CONSTRAINT(c, n)
171 183
172#define EVENT_CONSTRAINT_END \ 184#define EVENT_CONSTRAINT_END \
173 EVENT_CONSTRAINT(0, 0, 0) 185 EVENT_CONSTRAINT(0, 0, 0)
@@ -175,6 +187,28 @@ struct cpu_hw_events {
175#define for_each_event_constraint(e, c) \ 187#define for_each_event_constraint(e, c) \
176 for ((e) = (c); (e)->weight; (e)++) 188 for ((e) = (c); (e)->weight; (e)++)
177 189
190/*
191 * Extra registers for specific events.
192 * Some events need large masks and require external MSRs.
193 * Define a mapping to these extra registers.
194 */
195struct extra_reg {
196 unsigned int event;
197 unsigned int msr;
198 u64 config_mask;
199 u64 valid_mask;
200};
201
202#define EVENT_EXTRA_REG(e, ms, m, vm) { \
203 .event = (e), \
204 .msr = (ms), \
205 .config_mask = (m), \
206 .valid_mask = (vm), \
207 }
208#define INTEL_EVENT_EXTRA_REG(event, msr, vm) \
209 EVENT_EXTRA_REG(event, msr, ARCH_PERFMON_EVENTSEL_EVENT, vm)
210#define EVENT_EXTRA_END EVENT_EXTRA_REG(0, 0, 0, 0)
211
178union perf_capabilities { 212union perf_capabilities {
179 struct { 213 struct {
180 u64 lbr_format : 6; 214 u64 lbr_format : 6;
@@ -219,6 +253,7 @@ struct x86_pmu {
219 void (*put_event_constraints)(struct cpu_hw_events *cpuc, 253 void (*put_event_constraints)(struct cpu_hw_events *cpuc,
220 struct perf_event *event); 254 struct perf_event *event);
221 struct event_constraint *event_constraints; 255 struct event_constraint *event_constraints;
256 struct event_constraint *percore_constraints;
222 void (*quirks)(void); 257 void (*quirks)(void);
223 int perfctr_second_write; 258 int perfctr_second_write;
224 259
@@ -247,6 +282,11 @@ struct x86_pmu {
247 */ 282 */
248 unsigned long lbr_tos, lbr_from, lbr_to; /* MSR base regs */ 283 unsigned long lbr_tos, lbr_from, lbr_to; /* MSR base regs */
249 int lbr_nr; /* hardware stack size */ 284 int lbr_nr; /* hardware stack size */
285
286 /*
287 * Extra registers for events
288 */
289 struct extra_reg *extra_regs;
250}; 290};
251 291
252static struct x86_pmu x86_pmu __read_mostly; 292static struct x86_pmu x86_pmu __read_mostly;
@@ -271,6 +311,10 @@ static u64 __read_mostly hw_cache_event_ids
271 [PERF_COUNT_HW_CACHE_MAX] 311 [PERF_COUNT_HW_CACHE_MAX]
272 [PERF_COUNT_HW_CACHE_OP_MAX] 312 [PERF_COUNT_HW_CACHE_OP_MAX]
273 [PERF_COUNT_HW_CACHE_RESULT_MAX]; 313 [PERF_COUNT_HW_CACHE_RESULT_MAX];
314static u64 __read_mostly hw_cache_extra_regs
315 [PERF_COUNT_HW_CACHE_MAX]
316 [PERF_COUNT_HW_CACHE_OP_MAX]
317 [PERF_COUNT_HW_CACHE_RESULT_MAX];
274 318
275/* 319/*
276 * Propagate event elapsed time into the generic event. 320 * Propagate event elapsed time into the generic event.
@@ -298,7 +342,7 @@ x86_perf_event_update(struct perf_event *event)
298 */ 342 */
299again: 343again:
300 prev_raw_count = local64_read(&hwc->prev_count); 344 prev_raw_count = local64_read(&hwc->prev_count);
301 rdmsrl(hwc->event_base + idx, new_raw_count); 345 rdmsrl(hwc->event_base, new_raw_count);
302 346
303 if (local64_cmpxchg(&hwc->prev_count, prev_raw_count, 347 if (local64_cmpxchg(&hwc->prev_count, prev_raw_count,
304 new_raw_count) != prev_raw_count) 348 new_raw_count) != prev_raw_count)
@@ -321,6 +365,49 @@ again:
321 return new_raw_count; 365 return new_raw_count;
322} 366}
323 367
368/* using X86_FEATURE_PERFCTR_CORE to later implement ALTERNATIVE() here */
369static inline int x86_pmu_addr_offset(int index)
370{
371 if (boot_cpu_has(X86_FEATURE_PERFCTR_CORE))
372 return index << 1;
373 return index;
374}
375
376static inline unsigned int x86_pmu_config_addr(int index)
377{
378 return x86_pmu.eventsel + x86_pmu_addr_offset(index);
379}
380
381static inline unsigned int x86_pmu_event_addr(int index)
382{
383 return x86_pmu.perfctr + x86_pmu_addr_offset(index);
384}
385
386/*
387 * Find and validate any extra registers to set up.
388 */
389static int x86_pmu_extra_regs(u64 config, struct perf_event *event)
390{
391 struct extra_reg *er;
392
393 event->hw.extra_reg = 0;
394 event->hw.extra_config = 0;
395
396 if (!x86_pmu.extra_regs)
397 return 0;
398
399 for (er = x86_pmu.extra_regs; er->msr; er++) {
400 if (er->event != (config & er->config_mask))
401 continue;
402 if (event->attr.config1 & ~er->valid_mask)
403 return -EINVAL;
404 event->hw.extra_reg = er->msr;
405 event->hw.extra_config = event->attr.config1;
406 break;
407 }
408 return 0;
409}
410
324static atomic_t active_events; 411static atomic_t active_events;
325static DEFINE_MUTEX(pmc_reserve_mutex); 412static DEFINE_MUTEX(pmc_reserve_mutex);
326 413
@@ -331,12 +418,12 @@ static bool reserve_pmc_hardware(void)
331 int i; 418 int i;
332 419
333 for (i = 0; i < x86_pmu.num_counters; i++) { 420 for (i = 0; i < x86_pmu.num_counters; i++) {
334 if (!reserve_perfctr_nmi(x86_pmu.perfctr + i)) 421 if (!reserve_perfctr_nmi(x86_pmu_event_addr(i)))
335 goto perfctr_fail; 422 goto perfctr_fail;
336 } 423 }
337 424
338 for (i = 0; i < x86_pmu.num_counters; i++) { 425 for (i = 0; i < x86_pmu.num_counters; i++) {
339 if (!reserve_evntsel_nmi(x86_pmu.eventsel + i)) 426 if (!reserve_evntsel_nmi(x86_pmu_config_addr(i)))
340 goto eventsel_fail; 427 goto eventsel_fail;
341 } 428 }
342 429
@@ -344,13 +431,13 @@ static bool reserve_pmc_hardware(void)
344 431
345eventsel_fail: 432eventsel_fail:
346 for (i--; i >= 0; i--) 433 for (i--; i >= 0; i--)
347 release_evntsel_nmi(x86_pmu.eventsel + i); 434 release_evntsel_nmi(x86_pmu_config_addr(i));
348 435
349 i = x86_pmu.num_counters; 436 i = x86_pmu.num_counters;
350 437
351perfctr_fail: 438perfctr_fail:
352 for (i--; i >= 0; i--) 439 for (i--; i >= 0; i--)
353 release_perfctr_nmi(x86_pmu.perfctr + i); 440 release_perfctr_nmi(x86_pmu_event_addr(i));
354 441
355 return false; 442 return false;
356} 443}
@@ -360,8 +447,8 @@ static void release_pmc_hardware(void)
360 int i; 447 int i;
361 448
362 for (i = 0; i < x86_pmu.num_counters; i++) { 449 for (i = 0; i < x86_pmu.num_counters; i++) {
363 release_perfctr_nmi(x86_pmu.perfctr + i); 450 release_perfctr_nmi(x86_pmu_event_addr(i));
364 release_evntsel_nmi(x86_pmu.eventsel + i); 451 release_evntsel_nmi(x86_pmu_config_addr(i));
365 } 452 }
366} 453}
367 454
@@ -382,7 +469,7 @@ static bool check_hw_exists(void)
382 * complain and bail. 469 * complain and bail.
383 */ 470 */
384 for (i = 0; i < x86_pmu.num_counters; i++) { 471 for (i = 0; i < x86_pmu.num_counters; i++) {
385 reg = x86_pmu.eventsel + i; 472 reg = x86_pmu_config_addr(i);
386 ret = rdmsrl_safe(reg, &val); 473 ret = rdmsrl_safe(reg, &val);
387 if (ret) 474 if (ret)
388 goto msr_fail; 475 goto msr_fail;
@@ -407,8 +494,8 @@ static bool check_hw_exists(void)
407 * that don't trap on the MSR access and always return 0s. 494 * that don't trap on the MSR access and always return 0s.
408 */ 495 */
409 val = 0xabcdUL; 496 val = 0xabcdUL;
410 ret = checking_wrmsrl(x86_pmu.perfctr, val); 497 ret = checking_wrmsrl(x86_pmu_event_addr(0), val);
411 ret |= rdmsrl_safe(x86_pmu.perfctr, &val_new); 498 ret |= rdmsrl_safe(x86_pmu_event_addr(0), &val_new);
412 if (ret || val != val_new) 499 if (ret || val != val_new)
413 goto msr_fail; 500 goto msr_fail;
414 501
@@ -442,8 +529,9 @@ static inline int x86_pmu_initialized(void)
442} 529}
443 530
444static inline int 531static inline int
445set_ext_hw_attr(struct hw_perf_event *hwc, struct perf_event_attr *attr) 532set_ext_hw_attr(struct hw_perf_event *hwc, struct perf_event *event)
446{ 533{
534 struct perf_event_attr *attr = &event->attr;
447 unsigned int cache_type, cache_op, cache_result; 535 unsigned int cache_type, cache_op, cache_result;
448 u64 config, val; 536 u64 config, val;
449 537
@@ -470,8 +558,8 @@ set_ext_hw_attr(struct hw_perf_event *hwc, struct perf_event_attr *attr)
470 return -EINVAL; 558 return -EINVAL;
471 559
472 hwc->config |= val; 560 hwc->config |= val;
473 561 attr->config1 = hw_cache_extra_regs[cache_type][cache_op][cache_result];
474 return 0; 562 return x86_pmu_extra_regs(val, event);
475} 563}
476 564
477static int x86_setup_perfctr(struct perf_event *event) 565static int x86_setup_perfctr(struct perf_event *event)
@@ -496,10 +584,10 @@ static int x86_setup_perfctr(struct perf_event *event)
496 } 584 }
497 585
498 if (attr->type == PERF_TYPE_RAW) 586 if (attr->type == PERF_TYPE_RAW)
499 return 0; 587 return x86_pmu_extra_regs(event->attr.config, event);
500 588
501 if (attr->type == PERF_TYPE_HW_CACHE) 589 if (attr->type == PERF_TYPE_HW_CACHE)
502 return set_ext_hw_attr(hwc, attr); 590 return set_ext_hw_attr(hwc, event);
503 591
504 if (attr->config >= x86_pmu.max_events) 592 if (attr->config >= x86_pmu.max_events)
505 return -EINVAL; 593 return -EINVAL;
@@ -617,11 +705,11 @@ static void x86_pmu_disable_all(void)
617 705
618 if (!test_bit(idx, cpuc->active_mask)) 706 if (!test_bit(idx, cpuc->active_mask))
619 continue; 707 continue;
620 rdmsrl(x86_pmu.eventsel + idx, val); 708 rdmsrl(x86_pmu_config_addr(idx), val);
621 if (!(val & ARCH_PERFMON_EVENTSEL_ENABLE)) 709 if (!(val & ARCH_PERFMON_EVENTSEL_ENABLE))
622 continue; 710 continue;
623 val &= ~ARCH_PERFMON_EVENTSEL_ENABLE; 711 val &= ~ARCH_PERFMON_EVENTSEL_ENABLE;
624 wrmsrl(x86_pmu.eventsel + idx, val); 712 wrmsrl(x86_pmu_config_addr(idx), val);
625 } 713 }
626} 714}
627 715
@@ -642,21 +730,26 @@ static void x86_pmu_disable(struct pmu *pmu)
642 x86_pmu.disable_all(); 730 x86_pmu.disable_all();
643} 731}
644 732
733static inline void __x86_pmu_enable_event(struct hw_perf_event *hwc,
734 u64 enable_mask)
735{
736 if (hwc->extra_reg)
737 wrmsrl(hwc->extra_reg, hwc->extra_config);
738 wrmsrl(hwc->config_base, hwc->config | enable_mask);
739}
740
645static void x86_pmu_enable_all(int added) 741static void x86_pmu_enable_all(int added)
646{ 742{
647 struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); 743 struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
648 int idx; 744 int idx;
649 745
650 for (idx = 0; idx < x86_pmu.num_counters; idx++) { 746 for (idx = 0; idx < x86_pmu.num_counters; idx++) {
651 struct perf_event *event = cpuc->events[idx]; 747 struct hw_perf_event *hwc = &cpuc->events[idx]->hw;
652 u64 val;
653 748
654 if (!test_bit(idx, cpuc->active_mask)) 749 if (!test_bit(idx, cpuc->active_mask))
655 continue; 750 continue;
656 751
657 val = event->hw.config; 752 __x86_pmu_enable_event(hwc, ARCH_PERFMON_EVENTSEL_ENABLE);
658 val |= ARCH_PERFMON_EVENTSEL_ENABLE;
659 wrmsrl(x86_pmu.eventsel + idx, val);
660 } 753 }
661} 754}
662 755
@@ -821,15 +914,10 @@ static inline void x86_assign_hw_event(struct perf_event *event,
821 hwc->event_base = 0; 914 hwc->event_base = 0;
822 } else if (hwc->idx >= X86_PMC_IDX_FIXED) { 915 } else if (hwc->idx >= X86_PMC_IDX_FIXED) {
823 hwc->config_base = MSR_ARCH_PERFMON_FIXED_CTR_CTRL; 916 hwc->config_base = MSR_ARCH_PERFMON_FIXED_CTR_CTRL;
824 /* 917 hwc->event_base = MSR_ARCH_PERFMON_FIXED_CTR0;
825 * We set it so that event_base + idx in wrmsr/rdmsr maps to
826 * MSR_ARCH_PERFMON_FIXED_CTR0 ... CTR2:
827 */
828 hwc->event_base =
829 MSR_ARCH_PERFMON_FIXED_CTR0 - X86_PMC_IDX_FIXED;
830 } else { 918 } else {
831 hwc->config_base = x86_pmu.eventsel; 919 hwc->config_base = x86_pmu_config_addr(hwc->idx);
832 hwc->event_base = x86_pmu.perfctr; 920 hwc->event_base = x86_pmu_event_addr(hwc->idx);
833 } 921 }
834} 922}
835 923
@@ -915,17 +1003,11 @@ static void x86_pmu_enable(struct pmu *pmu)
915 x86_pmu.enable_all(added); 1003 x86_pmu.enable_all(added);
916} 1004}
917 1005
918static inline void __x86_pmu_enable_event(struct hw_perf_event *hwc,
919 u64 enable_mask)
920{
921 wrmsrl(hwc->config_base + hwc->idx, hwc->config | enable_mask);
922}
923
924static inline void x86_pmu_disable_event(struct perf_event *event) 1006static inline void x86_pmu_disable_event(struct perf_event *event)
925{ 1007{
926 struct hw_perf_event *hwc = &event->hw; 1008 struct hw_perf_event *hwc = &event->hw;
927 1009
928 wrmsrl(hwc->config_base + hwc->idx, hwc->config); 1010 wrmsrl(hwc->config_base, hwc->config);
929} 1011}
930 1012
931static DEFINE_PER_CPU(u64 [X86_PMC_IDX_MAX], pmc_prev_left); 1013static DEFINE_PER_CPU(u64 [X86_PMC_IDX_MAX], pmc_prev_left);
@@ -978,7 +1060,7 @@ x86_perf_event_set_period(struct perf_event *event)
978 */ 1060 */
979 local64_set(&hwc->prev_count, (u64)-left); 1061 local64_set(&hwc->prev_count, (u64)-left);
980 1062
981 wrmsrl(hwc->event_base + idx, (u64)(-left) & x86_pmu.cntval_mask); 1063 wrmsrl(hwc->event_base, (u64)(-left) & x86_pmu.cntval_mask);
982 1064
983 /* 1065 /*
984 * Due to erratum on certan cpu we need 1066 * Due to erratum on certan cpu we need
@@ -986,7 +1068,7 @@ x86_perf_event_set_period(struct perf_event *event)
986 * is updated properly 1068 * is updated properly
987 */ 1069 */
988 if (x86_pmu.perfctr_second_write) { 1070 if (x86_pmu.perfctr_second_write) {
989 wrmsrl(hwc->event_base + idx, 1071 wrmsrl(hwc->event_base,
990 (u64)(-left) & x86_pmu.cntval_mask); 1072 (u64)(-left) & x86_pmu.cntval_mask);
991 } 1073 }
992 1074
@@ -1113,8 +1195,8 @@ void perf_event_print_debug(void)
1113 pr_info("CPU#%d: active: %016llx\n", cpu, *(u64 *)cpuc->active_mask); 1195 pr_info("CPU#%d: active: %016llx\n", cpu, *(u64 *)cpuc->active_mask);
1114 1196
1115 for (idx = 0; idx < x86_pmu.num_counters; idx++) { 1197 for (idx = 0; idx < x86_pmu.num_counters; idx++) {
1116 rdmsrl(x86_pmu.eventsel + idx, pmc_ctrl); 1198 rdmsrl(x86_pmu_config_addr(idx), pmc_ctrl);
1117 rdmsrl(x86_pmu.perfctr + idx, pmc_count); 1199 rdmsrl(x86_pmu_event_addr(idx), pmc_count);
1118 1200
1119 prev_left = per_cpu(pmc_prev_left[idx], cpu); 1201 prev_left = per_cpu(pmc_prev_left[idx], cpu);
1120 1202
@@ -1389,7 +1471,7 @@ static void __init pmu_check_apic(void)
1389 pr_info("no hardware sampling interrupt available.\n"); 1471 pr_info("no hardware sampling interrupt available.\n");
1390} 1472}
1391 1473
1392int __init init_hw_perf_events(void) 1474static int __init init_hw_perf_events(void)
1393{ 1475{
1394 struct event_constraint *c; 1476 struct event_constraint *c;
1395 int err; 1477 int err;
@@ -1608,7 +1690,7 @@ out:
1608 return ret; 1690 return ret;
1609} 1691}
1610 1692
1611int x86_pmu_event_init(struct perf_event *event) 1693static int x86_pmu_event_init(struct perf_event *event)
1612{ 1694{
1613 struct pmu *tmp; 1695 struct pmu *tmp;
1614 int err; 1696 int err;
diff --git a/arch/x86/kernel/cpu/perf_event_amd.c b/arch/x86/kernel/cpu/perf_event_amd.c
index 67e2202a6039..461f62bbd774 100644
--- a/arch/x86/kernel/cpu/perf_event_amd.c
+++ b/arch/x86/kernel/cpu/perf_event_amd.c
@@ -127,6 +127,11 @@ static int amd_pmu_hw_config(struct perf_event *event)
127/* 127/*
128 * AMD64 events are detected based on their event codes. 128 * AMD64 events are detected based on their event codes.
129 */ 129 */
130static inline unsigned int amd_get_event_code(struct hw_perf_event *hwc)
131{
132 return ((hwc->config >> 24) & 0x0f00) | (hwc->config & 0x00ff);
133}
134
130static inline int amd_is_nb_event(struct hw_perf_event *hwc) 135static inline int amd_is_nb_event(struct hw_perf_event *hwc)
131{ 136{
132 return (hwc->config & 0xe0) == 0xe0; 137 return (hwc->config & 0xe0) == 0xe0;
@@ -385,13 +390,181 @@ static __initconst const struct x86_pmu amd_pmu = {
385 .cpu_dead = amd_pmu_cpu_dead, 390 .cpu_dead = amd_pmu_cpu_dead,
386}; 391};
387 392
393/* AMD Family 15h */
394
395#define AMD_EVENT_TYPE_MASK 0x000000F0ULL
396
397#define AMD_EVENT_FP 0x00000000ULL ... 0x00000010ULL
398#define AMD_EVENT_LS 0x00000020ULL ... 0x00000030ULL
399#define AMD_EVENT_DC 0x00000040ULL ... 0x00000050ULL
400#define AMD_EVENT_CU 0x00000060ULL ... 0x00000070ULL
401#define AMD_EVENT_IC_DE 0x00000080ULL ... 0x00000090ULL
402#define AMD_EVENT_EX_LS 0x000000C0ULL
403#define AMD_EVENT_DE 0x000000D0ULL
404#define AMD_EVENT_NB 0x000000E0ULL ... 0x000000F0ULL
405
406/*
407 * AMD family 15h event code/PMC mappings:
408 *
409 * type = event_code & 0x0F0:
410 *
411 * 0x000 FP PERF_CTL[5:3]
412 * 0x010 FP PERF_CTL[5:3]
413 * 0x020 LS PERF_CTL[5:0]
414 * 0x030 LS PERF_CTL[5:0]
415 * 0x040 DC PERF_CTL[5:0]
416 * 0x050 DC PERF_CTL[5:0]
417 * 0x060 CU PERF_CTL[2:0]
418 * 0x070 CU PERF_CTL[2:0]
419 * 0x080 IC/DE PERF_CTL[2:0]
420 * 0x090 IC/DE PERF_CTL[2:0]
421 * 0x0A0 ---
422 * 0x0B0 ---
423 * 0x0C0 EX/LS PERF_CTL[5:0]
424 * 0x0D0 DE PERF_CTL[2:0]
425 * 0x0E0 NB NB_PERF_CTL[3:0]
426 * 0x0F0 NB NB_PERF_CTL[3:0]
427 *
428 * Exceptions:
429 *
430 * 0x003 FP PERF_CTL[3]
431 * 0x00B FP PERF_CTL[3]
432 * 0x00D FP PERF_CTL[3]
433 * 0x023 DE PERF_CTL[2:0]
434 * 0x02D LS PERF_CTL[3]
435 * 0x02E LS PERF_CTL[3,0]
436 * 0x043 CU PERF_CTL[2:0]
437 * 0x045 CU PERF_CTL[2:0]
438 * 0x046 CU PERF_CTL[2:0]
439 * 0x054 CU PERF_CTL[2:0]
440 * 0x055 CU PERF_CTL[2:0]
441 * 0x08F IC PERF_CTL[0]
442 * 0x187 DE PERF_CTL[0]
443 * 0x188 DE PERF_CTL[0]
444 * 0x0DB EX PERF_CTL[5:0]
445 * 0x0DC LS PERF_CTL[5:0]
446 * 0x0DD LS PERF_CTL[5:0]
447 * 0x0DE LS PERF_CTL[5:0]
448 * 0x0DF LS PERF_CTL[5:0]
449 * 0x1D6 EX PERF_CTL[5:0]
450 * 0x1D8 EX PERF_CTL[5:0]
451 */
452
453static struct event_constraint amd_f15_PMC0 = EVENT_CONSTRAINT(0, 0x01, 0);
454static struct event_constraint amd_f15_PMC20 = EVENT_CONSTRAINT(0, 0x07, 0);
455static struct event_constraint amd_f15_PMC3 = EVENT_CONSTRAINT(0, 0x08, 0);
456static struct event_constraint amd_f15_PMC30 = EVENT_CONSTRAINT(0, 0x09, 0);
457static struct event_constraint amd_f15_PMC50 = EVENT_CONSTRAINT(0, 0x3F, 0);
458static struct event_constraint amd_f15_PMC53 = EVENT_CONSTRAINT(0, 0x38, 0);
459
460static struct event_constraint *
461amd_get_event_constraints_f15h(struct cpu_hw_events *cpuc, struct perf_event *event)
462{
463 unsigned int event_code = amd_get_event_code(&event->hw);
464
465 switch (event_code & AMD_EVENT_TYPE_MASK) {
466 case AMD_EVENT_FP:
467 switch (event_code) {
468 case 0x003:
469 case 0x00B:
470 case 0x00D:
471 return &amd_f15_PMC3;
472 default:
473 return &amd_f15_PMC53;
474 }
475 case AMD_EVENT_LS:
476 case AMD_EVENT_DC:
477 case AMD_EVENT_EX_LS:
478 switch (event_code) {
479 case 0x023:
480 case 0x043:
481 case 0x045:
482 case 0x046:
483 case 0x054:
484 case 0x055:
485 return &amd_f15_PMC20;
486 case 0x02D:
487 return &amd_f15_PMC3;
488 case 0x02E:
489 return &amd_f15_PMC30;
490 default:
491 return &amd_f15_PMC50;
492 }
493 case AMD_EVENT_CU:
494 case AMD_EVENT_IC_DE:
495 case AMD_EVENT_DE:
496 switch (event_code) {
497 case 0x08F:
498 case 0x187:
499 case 0x188:
500 return &amd_f15_PMC0;
501 case 0x0DB ... 0x0DF:
502 case 0x1D6:
503 case 0x1D8:
504 return &amd_f15_PMC50;
505 default:
506 return &amd_f15_PMC20;
507 }
508 case AMD_EVENT_NB:
509 /* not yet implemented */
510 return &emptyconstraint;
511 default:
512 return &emptyconstraint;
513 }
514}
515
516static __initconst const struct x86_pmu amd_pmu_f15h = {
517 .name = "AMD Family 15h",
518 .handle_irq = x86_pmu_handle_irq,
519 .disable_all = x86_pmu_disable_all,
520 .enable_all = x86_pmu_enable_all,
521 .enable = x86_pmu_enable_event,
522 .disable = x86_pmu_disable_event,
523 .hw_config = amd_pmu_hw_config,
524 .schedule_events = x86_schedule_events,
525 .eventsel = MSR_F15H_PERF_CTL,
526 .perfctr = MSR_F15H_PERF_CTR,
527 .event_map = amd_pmu_event_map,
528 .max_events = ARRAY_SIZE(amd_perfmon_event_map),
529 .num_counters = 6,
530 .cntval_bits = 48,
531 .cntval_mask = (1ULL << 48) - 1,
532 .apic = 1,
533 /* use highest bit to detect overflow */
534 .max_period = (1ULL << 47) - 1,
535 .get_event_constraints = amd_get_event_constraints_f15h,
536 /* nortbridge counters not yet implemented: */
537#if 0
538 .put_event_constraints = amd_put_event_constraints,
539
540 .cpu_prepare = amd_pmu_cpu_prepare,
541 .cpu_starting = amd_pmu_cpu_starting,
542 .cpu_dead = amd_pmu_cpu_dead,
543#endif
544};
545
388static __init int amd_pmu_init(void) 546static __init int amd_pmu_init(void)
389{ 547{
390 /* Performance-monitoring supported from K7 and later: */ 548 /* Performance-monitoring supported from K7 and later: */
391 if (boot_cpu_data.x86 < 6) 549 if (boot_cpu_data.x86 < 6)
392 return -ENODEV; 550 return -ENODEV;
393 551
394 x86_pmu = amd_pmu; 552 /*
553 * If core performance counter extensions exists, it must be
554 * family 15h, otherwise fail. See x86_pmu_addr_offset().
555 */
556 switch (boot_cpu_data.x86) {
557 case 0x15:
558 if (!cpu_has_perfctr_core)
559 return -ENODEV;
560 x86_pmu = amd_pmu_f15h;
561 break;
562 default:
563 if (cpu_has_perfctr_core)
564 return -ENODEV;
565 x86_pmu = amd_pmu;
566 break;
567 }
395 568
396 /* Events are common for all AMDs */ 569 /* Events are common for all AMDs */
397 memcpy(hw_cache_event_ids, amd_hw_cache_event_ids, 570 memcpy(hw_cache_event_ids, amd_hw_cache_event_ids,
diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c
index 008835c1d79c..8fc2b2cee1da 100644
--- a/arch/x86/kernel/cpu/perf_event_intel.c
+++ b/arch/x86/kernel/cpu/perf_event_intel.c
@@ -1,5 +1,27 @@
1#ifdef CONFIG_CPU_SUP_INTEL 1#ifdef CONFIG_CPU_SUP_INTEL
2 2
3#define MAX_EXTRA_REGS 2
4
5/*
6 * Per register state.
7 */
8struct er_account {
9 int ref; /* reference count */
10 unsigned int extra_reg; /* extra MSR number */
11 u64 extra_config; /* extra MSR config */
12};
13
14/*
15 * Per core state
16 * This used to coordinate shared registers for HT threads.
17 */
18struct intel_percore {
19 raw_spinlock_t lock; /* protect structure */
20 struct er_account regs[MAX_EXTRA_REGS];
21 int refcnt; /* number of threads */
22 unsigned core_id;
23};
24
3/* 25/*
4 * Intel PerfMon, used on Core and later. 26 * Intel PerfMon, used on Core and later.
5 */ 27 */
@@ -64,6 +86,18 @@ static struct event_constraint intel_nehalem_event_constraints[] =
64 EVENT_CONSTRAINT_END 86 EVENT_CONSTRAINT_END
65}; 87};
66 88
89static struct extra_reg intel_nehalem_extra_regs[] =
90{
91 INTEL_EVENT_EXTRA_REG(0xb7, MSR_OFFCORE_RSP_0, 0xffff),
92 EVENT_EXTRA_END
93};
94
95static struct event_constraint intel_nehalem_percore_constraints[] =
96{
97 INTEL_EVENT_CONSTRAINT(0xb7, 0),
98 EVENT_CONSTRAINT_END
99};
100
67static struct event_constraint intel_westmere_event_constraints[] = 101static struct event_constraint intel_westmere_event_constraints[] =
68{ 102{
69 FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */ 103 FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */
@@ -76,6 +110,33 @@ static struct event_constraint intel_westmere_event_constraints[] =
76 EVENT_CONSTRAINT_END 110 EVENT_CONSTRAINT_END
77}; 111};
78 112
113static struct event_constraint intel_snb_event_constraints[] =
114{
115 FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */
116 FIXED_EVENT_CONSTRAINT(0x003c, 1), /* CPU_CLK_UNHALTED.CORE */
117 /* FIXED_EVENT_CONSTRAINT(0x013c, 2), CPU_CLK_UNHALTED.REF */
118 INTEL_EVENT_CONSTRAINT(0x48, 0x4), /* L1D_PEND_MISS.PENDING */
119 INTEL_EVENT_CONSTRAINT(0xb7, 0x1), /* OFF_CORE_RESPONSE_0 */
120 INTEL_EVENT_CONSTRAINT(0xbb, 0x8), /* OFF_CORE_RESPONSE_1 */
121 INTEL_UEVENT_CONSTRAINT(0x01c0, 0x2), /* INST_RETIRED.PREC_DIST */
122 INTEL_EVENT_CONSTRAINT(0xcd, 0x8), /* MEM_TRANS_RETIRED.LOAD_LATENCY */
123 EVENT_CONSTRAINT_END
124};
125
126static struct extra_reg intel_westmere_extra_regs[] =
127{
128 INTEL_EVENT_EXTRA_REG(0xb7, MSR_OFFCORE_RSP_0, 0xffff),
129 INTEL_EVENT_EXTRA_REG(0xbb, MSR_OFFCORE_RSP_1, 0xffff),
130 EVENT_EXTRA_END
131};
132
133static struct event_constraint intel_westmere_percore_constraints[] =
134{
135 INTEL_EVENT_CONSTRAINT(0xb7, 0),
136 INTEL_EVENT_CONSTRAINT(0xbb, 0),
137 EVENT_CONSTRAINT_END
138};
139
79static struct event_constraint intel_gen_event_constraints[] = 140static struct event_constraint intel_gen_event_constraints[] =
80{ 141{
81 FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */ 142 FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */
@@ -89,6 +150,106 @@ static u64 intel_pmu_event_map(int hw_event)
89 return intel_perfmon_event_map[hw_event]; 150 return intel_perfmon_event_map[hw_event];
90} 151}
91 152
153static __initconst const u64 snb_hw_cache_event_ids
154 [PERF_COUNT_HW_CACHE_MAX]
155 [PERF_COUNT_HW_CACHE_OP_MAX]
156 [PERF_COUNT_HW_CACHE_RESULT_MAX] =
157{
158 [ C(L1D) ] = {
159 [ C(OP_READ) ] = {
160 [ C(RESULT_ACCESS) ] = 0xf1d0, /* MEM_UOP_RETIRED.LOADS */
161 [ C(RESULT_MISS) ] = 0x0151, /* L1D.REPLACEMENT */
162 },
163 [ C(OP_WRITE) ] = {
164 [ C(RESULT_ACCESS) ] = 0xf2d0, /* MEM_UOP_RETIRED.STORES */
165 [ C(RESULT_MISS) ] = 0x0851, /* L1D.ALL_M_REPLACEMENT */
166 },
167 [ C(OP_PREFETCH) ] = {
168 [ C(RESULT_ACCESS) ] = 0x0,
169 [ C(RESULT_MISS) ] = 0x024e, /* HW_PRE_REQ.DL1_MISS */
170 },
171 },
172 [ C(L1I ) ] = {
173 [ C(OP_READ) ] = {
174 [ C(RESULT_ACCESS) ] = 0x0,
175 [ C(RESULT_MISS) ] = 0x0280, /* ICACHE.MISSES */
176 },
177 [ C(OP_WRITE) ] = {
178 [ C(RESULT_ACCESS) ] = -1,
179 [ C(RESULT_MISS) ] = -1,
180 },
181 [ C(OP_PREFETCH) ] = {
182 [ C(RESULT_ACCESS) ] = 0x0,
183 [ C(RESULT_MISS) ] = 0x0,
184 },
185 },
186 [ C(LL ) ] = {
187 /*
188 * TBD: Need Off-core Response Performance Monitoring support
189 */
190 [ C(OP_READ) ] = {
191 /* OFFCORE_RESPONSE_0.ANY_DATA.LOCAL_CACHE */
192 [ C(RESULT_ACCESS) ] = 0x01b7,
193 /* OFFCORE_RESPONSE_1.ANY_DATA.ANY_LLC_MISS */
194 [ C(RESULT_MISS) ] = 0x01bb,
195 },
196 [ C(OP_WRITE) ] = {
197 /* OFFCORE_RESPONSE_0.ANY_RFO.LOCAL_CACHE */
198 [ C(RESULT_ACCESS) ] = 0x01b7,
199 /* OFFCORE_RESPONSE_1.ANY_RFO.ANY_LLC_MISS */
200 [ C(RESULT_MISS) ] = 0x01bb,
201 },
202 [ C(OP_PREFETCH) ] = {
203 /* OFFCORE_RESPONSE_0.PREFETCH.LOCAL_CACHE */
204 [ C(RESULT_ACCESS) ] = 0x01b7,
205 /* OFFCORE_RESPONSE_1.PREFETCH.ANY_LLC_MISS */
206 [ C(RESULT_MISS) ] = 0x01bb,
207 },
208 },
209 [ C(DTLB) ] = {
210 [ C(OP_READ) ] = {
211 [ C(RESULT_ACCESS) ] = 0x81d0, /* MEM_UOP_RETIRED.ALL_LOADS */
212 [ C(RESULT_MISS) ] = 0x0108, /* DTLB_LOAD_MISSES.CAUSES_A_WALK */
213 },
214 [ C(OP_WRITE) ] = {
215 [ C(RESULT_ACCESS) ] = 0x82d0, /* MEM_UOP_RETIRED.ALL_STORES */
216 [ C(RESULT_MISS) ] = 0x0149, /* DTLB_STORE_MISSES.MISS_CAUSES_A_WALK */
217 },
218 [ C(OP_PREFETCH) ] = {
219 [ C(RESULT_ACCESS) ] = 0x0,
220 [ C(RESULT_MISS) ] = 0x0,
221 },
222 },
223 [ C(ITLB) ] = {
224 [ C(OP_READ) ] = {
225 [ C(RESULT_ACCESS) ] = 0x1085, /* ITLB_MISSES.STLB_HIT */
226 [ C(RESULT_MISS) ] = 0x0185, /* ITLB_MISSES.CAUSES_A_WALK */
227 },
228 [ C(OP_WRITE) ] = {
229 [ C(RESULT_ACCESS) ] = -1,
230 [ C(RESULT_MISS) ] = -1,
231 },
232 [ C(OP_PREFETCH) ] = {
233 [ C(RESULT_ACCESS) ] = -1,
234 [ C(RESULT_MISS) ] = -1,
235 },
236 },
237 [ C(BPU ) ] = {
238 [ C(OP_READ) ] = {
239 [ C(RESULT_ACCESS) ] = 0x00c4, /* BR_INST_RETIRED.ALL_BRANCHES */
240 [ C(RESULT_MISS) ] = 0x00c5, /* BR_MISP_RETIRED.ALL_BRANCHES */
241 },
242 [ C(OP_WRITE) ] = {
243 [ C(RESULT_ACCESS) ] = -1,
244 [ C(RESULT_MISS) ] = -1,
245 },
246 [ C(OP_PREFETCH) ] = {
247 [ C(RESULT_ACCESS) ] = -1,
248 [ C(RESULT_MISS) ] = -1,
249 },
250 },
251};
252
92static __initconst const u64 westmere_hw_cache_event_ids 253static __initconst const u64 westmere_hw_cache_event_ids
93 [PERF_COUNT_HW_CACHE_MAX] 254 [PERF_COUNT_HW_CACHE_MAX]
94 [PERF_COUNT_HW_CACHE_OP_MAX] 255 [PERF_COUNT_HW_CACHE_OP_MAX]
@@ -124,16 +285,26 @@ static __initconst const u64 westmere_hw_cache_event_ids
124 }, 285 },
125 [ C(LL ) ] = { 286 [ C(LL ) ] = {
126 [ C(OP_READ) ] = { 287 [ C(OP_READ) ] = {
127 [ C(RESULT_ACCESS) ] = 0x0324, /* L2_RQSTS.LOADS */ 288 /* OFFCORE_RESPONSE_0.ANY_DATA.LOCAL_CACHE */
128 [ C(RESULT_MISS) ] = 0x0224, /* L2_RQSTS.LD_MISS */ 289 [ C(RESULT_ACCESS) ] = 0x01b7,
290 /* OFFCORE_RESPONSE_1.ANY_DATA.ANY_LLC_MISS */
291 [ C(RESULT_MISS) ] = 0x01bb,
129 }, 292 },
293 /*
294 * Use RFO, not WRITEBACK, because a write miss would typically occur
295 * on RFO.
296 */
130 [ C(OP_WRITE) ] = { 297 [ C(OP_WRITE) ] = {
131 [ C(RESULT_ACCESS) ] = 0x0c24, /* L2_RQSTS.RFOS */ 298 /* OFFCORE_RESPONSE_1.ANY_RFO.LOCAL_CACHE */
132 [ C(RESULT_MISS) ] = 0x0824, /* L2_RQSTS.RFO_MISS */ 299 [ C(RESULT_ACCESS) ] = 0x01bb,
300 /* OFFCORE_RESPONSE_0.ANY_RFO.ANY_LLC_MISS */
301 [ C(RESULT_MISS) ] = 0x01b7,
133 }, 302 },
134 [ C(OP_PREFETCH) ] = { 303 [ C(OP_PREFETCH) ] = {
135 [ C(RESULT_ACCESS) ] = 0x4f2e, /* LLC Reference */ 304 /* OFFCORE_RESPONSE_0.PREFETCH.LOCAL_CACHE */
136 [ C(RESULT_MISS) ] = 0x412e, /* LLC Misses */ 305 [ C(RESULT_ACCESS) ] = 0x01b7,
306 /* OFFCORE_RESPONSE_1.PREFETCH.ANY_LLC_MISS */
307 [ C(RESULT_MISS) ] = 0x01bb,
137 }, 308 },
138 }, 309 },
139 [ C(DTLB) ] = { 310 [ C(DTLB) ] = {
@@ -180,6 +351,39 @@ static __initconst const u64 westmere_hw_cache_event_ids
180 }, 351 },
181}; 352};
182 353
354/*
355 * OFFCORE_RESPONSE MSR bits (subset), See IA32 SDM Vol 3 30.6.1.3
356 */
357
358#define DMND_DATA_RD (1 << 0)
359#define DMND_RFO (1 << 1)
360#define DMND_WB (1 << 3)
361#define PF_DATA_RD (1 << 4)
362#define PF_DATA_RFO (1 << 5)
363#define RESP_UNCORE_HIT (1 << 8)
364#define RESP_MISS (0xf600) /* non uncore hit */
365
366static __initconst const u64 nehalem_hw_cache_extra_regs
367 [PERF_COUNT_HW_CACHE_MAX]
368 [PERF_COUNT_HW_CACHE_OP_MAX]
369 [PERF_COUNT_HW_CACHE_RESULT_MAX] =
370{
371 [ C(LL ) ] = {
372 [ C(OP_READ) ] = {
373 [ C(RESULT_ACCESS) ] = DMND_DATA_RD|RESP_UNCORE_HIT,
374 [ C(RESULT_MISS) ] = DMND_DATA_RD|RESP_MISS,
375 },
376 [ C(OP_WRITE) ] = {
377 [ C(RESULT_ACCESS) ] = DMND_RFO|DMND_WB|RESP_UNCORE_HIT,
378 [ C(RESULT_MISS) ] = DMND_RFO|DMND_WB|RESP_MISS,
379 },
380 [ C(OP_PREFETCH) ] = {
381 [ C(RESULT_ACCESS) ] = PF_DATA_RD|PF_DATA_RFO|RESP_UNCORE_HIT,
382 [ C(RESULT_MISS) ] = PF_DATA_RD|PF_DATA_RFO|RESP_MISS,
383 },
384 }
385};
386
183static __initconst const u64 nehalem_hw_cache_event_ids 387static __initconst const u64 nehalem_hw_cache_event_ids
184 [PERF_COUNT_HW_CACHE_MAX] 388 [PERF_COUNT_HW_CACHE_MAX]
185 [PERF_COUNT_HW_CACHE_OP_MAX] 389 [PERF_COUNT_HW_CACHE_OP_MAX]
@@ -215,16 +419,26 @@ static __initconst const u64 nehalem_hw_cache_event_ids
215 }, 419 },
216 [ C(LL ) ] = { 420 [ C(LL ) ] = {
217 [ C(OP_READ) ] = { 421 [ C(OP_READ) ] = {
218 [ C(RESULT_ACCESS) ] = 0x0324, /* L2_RQSTS.LOADS */ 422 /* OFFCORE_RESPONSE.ANY_DATA.LOCAL_CACHE */
219 [ C(RESULT_MISS) ] = 0x0224, /* L2_RQSTS.LD_MISS */ 423 [ C(RESULT_ACCESS) ] = 0x01b7,
424 /* OFFCORE_RESPONSE.ANY_DATA.ANY_LLC_MISS */
425 [ C(RESULT_MISS) ] = 0x01b7,
220 }, 426 },
427 /*
428 * Use RFO, not WRITEBACK, because a write miss would typically occur
429 * on RFO.
430 */
221 [ C(OP_WRITE) ] = { 431 [ C(OP_WRITE) ] = {
222 [ C(RESULT_ACCESS) ] = 0x0c24, /* L2_RQSTS.RFOS */ 432 /* OFFCORE_RESPONSE.ANY_RFO.LOCAL_CACHE */
223 [ C(RESULT_MISS) ] = 0x0824, /* L2_RQSTS.RFO_MISS */ 433 [ C(RESULT_ACCESS) ] = 0x01b7,
434 /* OFFCORE_RESPONSE.ANY_RFO.ANY_LLC_MISS */
435 [ C(RESULT_MISS) ] = 0x01b7,
224 }, 436 },
225 [ C(OP_PREFETCH) ] = { 437 [ C(OP_PREFETCH) ] = {
226 [ C(RESULT_ACCESS) ] = 0x4f2e, /* LLC Reference */ 438 /* OFFCORE_RESPONSE.PREFETCH.LOCAL_CACHE */
227 [ C(RESULT_MISS) ] = 0x412e, /* LLC Misses */ 439 [ C(RESULT_ACCESS) ] = 0x01b7,
440 /* OFFCORE_RESPONSE.PREFETCH.ANY_LLC_MISS */
441 [ C(RESULT_MISS) ] = 0x01b7,
228 }, 442 },
229 }, 443 },
230 [ C(DTLB) ] = { 444 [ C(DTLB) ] = {
@@ -691,8 +905,8 @@ static void intel_pmu_reset(void)
691 printk("clearing PMU state on CPU#%d\n", smp_processor_id()); 905 printk("clearing PMU state on CPU#%d\n", smp_processor_id());
692 906
693 for (idx = 0; idx < x86_pmu.num_counters; idx++) { 907 for (idx = 0; idx < x86_pmu.num_counters; idx++) {
694 checking_wrmsrl(x86_pmu.eventsel + idx, 0ull); 908 checking_wrmsrl(x86_pmu_config_addr(idx), 0ull);
695 checking_wrmsrl(x86_pmu.perfctr + idx, 0ull); 909 checking_wrmsrl(x86_pmu_event_addr(idx), 0ull);
696 } 910 }
697 for (idx = 0; idx < x86_pmu.num_counters_fixed; idx++) 911 for (idx = 0; idx < x86_pmu.num_counters_fixed; idx++)
698 checking_wrmsrl(MSR_ARCH_PERFMON_FIXED_CTR0 + idx, 0ull); 912 checking_wrmsrl(MSR_ARCH_PERFMON_FIXED_CTR0 + idx, 0ull);
@@ -794,6 +1008,67 @@ intel_bts_constraints(struct perf_event *event)
794} 1008}
795 1009
796static struct event_constraint * 1010static struct event_constraint *
1011intel_percore_constraints(struct cpu_hw_events *cpuc, struct perf_event *event)
1012{
1013 struct hw_perf_event *hwc = &event->hw;
1014 unsigned int e = hwc->config & ARCH_PERFMON_EVENTSEL_EVENT;
1015 struct event_constraint *c;
1016 struct intel_percore *pc;
1017 struct er_account *era;
1018 int i;
1019 int free_slot;
1020 int found;
1021
1022 if (!x86_pmu.percore_constraints || hwc->extra_alloc)
1023 return NULL;
1024
1025 for (c = x86_pmu.percore_constraints; c->cmask; c++) {
1026 if (e != c->code)
1027 continue;
1028
1029 /*
1030 * Allocate resource per core.
1031 */
1032 pc = cpuc->per_core;
1033 if (!pc)
1034 break;
1035 c = &emptyconstraint;
1036 raw_spin_lock(&pc->lock);
1037 free_slot = -1;
1038 found = 0;
1039 for (i = 0; i < MAX_EXTRA_REGS; i++) {
1040 era = &pc->regs[i];
1041 if (era->ref > 0 && hwc->extra_reg == era->extra_reg) {
1042 /* Allow sharing same config */
1043 if (hwc->extra_config == era->extra_config) {
1044 era->ref++;
1045 cpuc->percore_used = 1;
1046 hwc->extra_alloc = 1;
1047 c = NULL;
1048 }
1049 /* else conflict */
1050 found = 1;
1051 break;
1052 } else if (era->ref == 0 && free_slot == -1)
1053 free_slot = i;
1054 }
1055 if (!found && free_slot != -1) {
1056 era = &pc->regs[free_slot];
1057 era->ref = 1;
1058 era->extra_reg = hwc->extra_reg;
1059 era->extra_config = hwc->extra_config;
1060 cpuc->percore_used = 1;
1061 hwc->extra_alloc = 1;
1062 c = NULL;
1063 }
1064 raw_spin_unlock(&pc->lock);
1065 return c;
1066 }
1067
1068 return NULL;
1069}
1070
1071static struct event_constraint *
797intel_get_event_constraints(struct cpu_hw_events *cpuc, struct perf_event *event) 1072intel_get_event_constraints(struct cpu_hw_events *cpuc, struct perf_event *event)
798{ 1073{
799 struct event_constraint *c; 1074 struct event_constraint *c;
@@ -806,9 +1081,51 @@ intel_get_event_constraints(struct cpu_hw_events *cpuc, struct perf_event *event
806 if (c) 1081 if (c)
807 return c; 1082 return c;
808 1083
1084 c = intel_percore_constraints(cpuc, event);
1085 if (c)
1086 return c;
1087
809 return x86_get_event_constraints(cpuc, event); 1088 return x86_get_event_constraints(cpuc, event);
810} 1089}
811 1090
1091static void intel_put_event_constraints(struct cpu_hw_events *cpuc,
1092 struct perf_event *event)
1093{
1094 struct extra_reg *er;
1095 struct intel_percore *pc;
1096 struct er_account *era;
1097 struct hw_perf_event *hwc = &event->hw;
1098 int i, allref;
1099
1100 if (!cpuc->percore_used)
1101 return;
1102
1103 for (er = x86_pmu.extra_regs; er->msr; er++) {
1104 if (er->event != (hwc->config & er->config_mask))
1105 continue;
1106
1107 pc = cpuc->per_core;
1108 raw_spin_lock(&pc->lock);
1109 for (i = 0; i < MAX_EXTRA_REGS; i++) {
1110 era = &pc->regs[i];
1111 if (era->ref > 0 &&
1112 era->extra_config == hwc->extra_config &&
1113 era->extra_reg == er->msr) {
1114 era->ref--;
1115 hwc->extra_alloc = 0;
1116 break;
1117 }
1118 }
1119 allref = 0;
1120 for (i = 0; i < MAX_EXTRA_REGS; i++)
1121 allref += pc->regs[i].ref;
1122 if (allref == 0)
1123 cpuc->percore_used = 0;
1124 raw_spin_unlock(&pc->lock);
1125 break;
1126 }
1127}
1128
812static int intel_pmu_hw_config(struct perf_event *event) 1129static int intel_pmu_hw_config(struct perf_event *event)
813{ 1130{
814 int ret = x86_pmu_hw_config(event); 1131 int ret = x86_pmu_hw_config(event);
@@ -880,20 +1197,67 @@ static __initconst const struct x86_pmu core_pmu = {
880 */ 1197 */
881 .max_period = (1ULL << 31) - 1, 1198 .max_period = (1ULL << 31) - 1,
882 .get_event_constraints = intel_get_event_constraints, 1199 .get_event_constraints = intel_get_event_constraints,
1200 .put_event_constraints = intel_put_event_constraints,
883 .event_constraints = intel_core_event_constraints, 1201 .event_constraints = intel_core_event_constraints,
884}; 1202};
885 1203
1204static int intel_pmu_cpu_prepare(int cpu)
1205{
1206 struct cpu_hw_events *cpuc = &per_cpu(cpu_hw_events, cpu);
1207
1208 if (!cpu_has_ht_siblings())
1209 return NOTIFY_OK;
1210
1211 cpuc->per_core = kzalloc_node(sizeof(struct intel_percore),
1212 GFP_KERNEL, cpu_to_node(cpu));
1213 if (!cpuc->per_core)
1214 return NOTIFY_BAD;
1215
1216 raw_spin_lock_init(&cpuc->per_core->lock);
1217 cpuc->per_core->core_id = -1;
1218 return NOTIFY_OK;
1219}
1220
886static void intel_pmu_cpu_starting(int cpu) 1221static void intel_pmu_cpu_starting(int cpu)
887{ 1222{
1223 struct cpu_hw_events *cpuc = &per_cpu(cpu_hw_events, cpu);
1224 int core_id = topology_core_id(cpu);
1225 int i;
1226
888 init_debug_store_on_cpu(cpu); 1227 init_debug_store_on_cpu(cpu);
889 /* 1228 /*
890 * Deal with CPUs that don't clear their LBRs on power-up. 1229 * Deal with CPUs that don't clear their LBRs on power-up.
891 */ 1230 */
892 intel_pmu_lbr_reset(); 1231 intel_pmu_lbr_reset();
1232
1233 if (!cpu_has_ht_siblings())
1234 return;
1235
1236 for_each_cpu(i, topology_thread_cpumask(cpu)) {
1237 struct intel_percore *pc = per_cpu(cpu_hw_events, i).per_core;
1238
1239 if (pc && pc->core_id == core_id) {
1240 kfree(cpuc->per_core);
1241 cpuc->per_core = pc;
1242 break;
1243 }
1244 }
1245
1246 cpuc->per_core->core_id = core_id;
1247 cpuc->per_core->refcnt++;
893} 1248}
894 1249
895static void intel_pmu_cpu_dying(int cpu) 1250static void intel_pmu_cpu_dying(int cpu)
896{ 1251{
1252 struct cpu_hw_events *cpuc = &per_cpu(cpu_hw_events, cpu);
1253 struct intel_percore *pc = cpuc->per_core;
1254
1255 if (pc) {
1256 if (pc->core_id == -1 || --pc->refcnt == 0)
1257 kfree(pc);
1258 cpuc->per_core = NULL;
1259 }
1260
897 fini_debug_store_on_cpu(cpu); 1261 fini_debug_store_on_cpu(cpu);
898} 1262}
899 1263
@@ -918,7 +1282,9 @@ static __initconst const struct x86_pmu intel_pmu = {
918 */ 1282 */
919 .max_period = (1ULL << 31) - 1, 1283 .max_period = (1ULL << 31) - 1,
920 .get_event_constraints = intel_get_event_constraints, 1284 .get_event_constraints = intel_get_event_constraints,
1285 .put_event_constraints = intel_put_event_constraints,
921 1286
1287 .cpu_prepare = intel_pmu_cpu_prepare,
922 .cpu_starting = intel_pmu_cpu_starting, 1288 .cpu_starting = intel_pmu_cpu_starting,
923 .cpu_dying = intel_pmu_cpu_dying, 1289 .cpu_dying = intel_pmu_cpu_dying,
924}; 1290};
@@ -1024,6 +1390,7 @@ static __init int intel_pmu_init(void)
1024 intel_pmu_lbr_init_core(); 1390 intel_pmu_lbr_init_core();
1025 1391
1026 x86_pmu.event_constraints = intel_core2_event_constraints; 1392 x86_pmu.event_constraints = intel_core2_event_constraints;
1393 x86_pmu.pebs_constraints = intel_core2_pebs_event_constraints;
1027 pr_cont("Core2 events, "); 1394 pr_cont("Core2 events, ");
1028 break; 1395 break;
1029 1396
@@ -1032,11 +1399,16 @@ static __init int intel_pmu_init(void)
1032 case 46: /* 45 nm nehalem-ex, "Beckton" */ 1399 case 46: /* 45 nm nehalem-ex, "Beckton" */
1033 memcpy(hw_cache_event_ids, nehalem_hw_cache_event_ids, 1400 memcpy(hw_cache_event_ids, nehalem_hw_cache_event_ids,
1034 sizeof(hw_cache_event_ids)); 1401 sizeof(hw_cache_event_ids));
1402 memcpy(hw_cache_extra_regs, nehalem_hw_cache_extra_regs,
1403 sizeof(hw_cache_extra_regs));
1035 1404
1036 intel_pmu_lbr_init_nhm(); 1405 intel_pmu_lbr_init_nhm();
1037 1406
1038 x86_pmu.event_constraints = intel_nehalem_event_constraints; 1407 x86_pmu.event_constraints = intel_nehalem_event_constraints;
1408 x86_pmu.pebs_constraints = intel_nehalem_pebs_event_constraints;
1409 x86_pmu.percore_constraints = intel_nehalem_percore_constraints;
1039 x86_pmu.enable_all = intel_pmu_nhm_enable_all; 1410 x86_pmu.enable_all = intel_pmu_nhm_enable_all;
1411 x86_pmu.extra_regs = intel_nehalem_extra_regs;
1040 pr_cont("Nehalem events, "); 1412 pr_cont("Nehalem events, ");
1041 break; 1413 break;
1042 1414
@@ -1047,6 +1419,7 @@ static __init int intel_pmu_init(void)
1047 intel_pmu_lbr_init_atom(); 1419 intel_pmu_lbr_init_atom();
1048 1420
1049 x86_pmu.event_constraints = intel_gen_event_constraints; 1421 x86_pmu.event_constraints = intel_gen_event_constraints;
1422 x86_pmu.pebs_constraints = intel_atom_pebs_event_constraints;
1050 pr_cont("Atom events, "); 1423 pr_cont("Atom events, ");
1051 break; 1424 break;
1052 1425
@@ -1054,14 +1427,30 @@ static __init int intel_pmu_init(void)
1054 case 44: /* 32 nm nehalem, "Gulftown" */ 1427 case 44: /* 32 nm nehalem, "Gulftown" */
1055 memcpy(hw_cache_event_ids, westmere_hw_cache_event_ids, 1428 memcpy(hw_cache_event_ids, westmere_hw_cache_event_ids,
1056 sizeof(hw_cache_event_ids)); 1429 sizeof(hw_cache_event_ids));
1430 memcpy(hw_cache_extra_regs, nehalem_hw_cache_extra_regs,
1431 sizeof(hw_cache_extra_regs));
1057 1432
1058 intel_pmu_lbr_init_nhm(); 1433 intel_pmu_lbr_init_nhm();
1059 1434
1060 x86_pmu.event_constraints = intel_westmere_event_constraints; 1435 x86_pmu.event_constraints = intel_westmere_event_constraints;
1436 x86_pmu.percore_constraints = intel_westmere_percore_constraints;
1061 x86_pmu.enable_all = intel_pmu_nhm_enable_all; 1437 x86_pmu.enable_all = intel_pmu_nhm_enable_all;
1438 x86_pmu.pebs_constraints = intel_westmere_pebs_event_constraints;
1439 x86_pmu.extra_regs = intel_westmere_extra_regs;
1062 pr_cont("Westmere events, "); 1440 pr_cont("Westmere events, ");
1063 break; 1441 break;
1064 1442
1443 case 42: /* SandyBridge */
1444 memcpy(hw_cache_event_ids, snb_hw_cache_event_ids,
1445 sizeof(hw_cache_event_ids));
1446
1447 intel_pmu_lbr_init_nhm();
1448
1449 x86_pmu.event_constraints = intel_snb_event_constraints;
1450 x86_pmu.pebs_constraints = intel_snb_pebs_events;
1451 pr_cont("SandyBridge events, ");
1452 break;
1453
1065 default: 1454 default:
1066 /* 1455 /*
1067 * default constraints for v2 and up 1456 * default constraints for v2 and up
diff --git a/arch/x86/kernel/cpu/perf_event_intel_ds.c b/arch/x86/kernel/cpu/perf_event_intel_ds.c
index b7dcd9f2b8a0..b95c66ae4a2a 100644
--- a/arch/x86/kernel/cpu/perf_event_intel_ds.c
+++ b/arch/x86/kernel/cpu/perf_event_intel_ds.c
@@ -361,30 +361,88 @@ static int intel_pmu_drain_bts_buffer(void)
361/* 361/*
362 * PEBS 362 * PEBS
363 */ 363 */
364 364static struct event_constraint intel_core2_pebs_event_constraints[] = {
365static struct event_constraint intel_core_pebs_events[] = { 365 PEBS_EVENT_CONSTRAINT(0x00c0, 0x1), /* INST_RETIRED.ANY */
366 PEBS_EVENT_CONSTRAINT(0x00c0, 0x1), /* INSTR_RETIRED.ANY */
367 PEBS_EVENT_CONSTRAINT(0xfec1, 0x1), /* X87_OPS_RETIRED.ANY */ 366 PEBS_EVENT_CONSTRAINT(0xfec1, 0x1), /* X87_OPS_RETIRED.ANY */
368 PEBS_EVENT_CONSTRAINT(0x00c5, 0x1), /* BR_INST_RETIRED.MISPRED */ 367 PEBS_EVENT_CONSTRAINT(0x00c5, 0x1), /* BR_INST_RETIRED.MISPRED */
369 PEBS_EVENT_CONSTRAINT(0x1fc7, 0x1), /* SIMD_INST_RETURED.ANY */ 368 PEBS_EVENT_CONSTRAINT(0x1fc7, 0x1), /* SIMD_INST_RETURED.ANY */
370 PEBS_EVENT_CONSTRAINT(0x01cb, 0x1), /* MEM_LOAD_RETIRED.L1D_MISS */ 369 INTEL_EVENT_CONSTRAINT(0xcb, 0x1), /* MEM_LOAD_RETIRED.* */
371 PEBS_EVENT_CONSTRAINT(0x02cb, 0x1), /* MEM_LOAD_RETIRED.L1D_LINE_MISS */ 370 EVENT_CONSTRAINT_END
372 PEBS_EVENT_CONSTRAINT(0x04cb, 0x1), /* MEM_LOAD_RETIRED.L2_MISS */ 371};
373 PEBS_EVENT_CONSTRAINT(0x08cb, 0x1), /* MEM_LOAD_RETIRED.L2_LINE_MISS */ 372
374 PEBS_EVENT_CONSTRAINT(0x10cb, 0x1), /* MEM_LOAD_RETIRED.DTLB_MISS */ 373static struct event_constraint intel_atom_pebs_event_constraints[] = {
374 PEBS_EVENT_CONSTRAINT(0x00c0, 0x1), /* INST_RETIRED.ANY */
375 PEBS_EVENT_CONSTRAINT(0x00c5, 0x1), /* MISPREDICTED_BRANCH_RETIRED */
376 INTEL_EVENT_CONSTRAINT(0xcb, 0x1), /* MEM_LOAD_RETIRED.* */
375 EVENT_CONSTRAINT_END 377 EVENT_CONSTRAINT_END
376}; 378};
377 379
378static struct event_constraint intel_nehalem_pebs_events[] = { 380static struct event_constraint intel_nehalem_pebs_event_constraints[] = {
379 PEBS_EVENT_CONSTRAINT(0x00c0, 0xf), /* INSTR_RETIRED.ANY */ 381 INTEL_EVENT_CONSTRAINT(0x0b, 0xf), /* MEM_INST_RETIRED.* */
380 PEBS_EVENT_CONSTRAINT(0xfec1, 0xf), /* X87_OPS_RETIRED.ANY */ 382 INTEL_EVENT_CONSTRAINT(0x0f, 0xf), /* MEM_UNCORE_RETIRED.* */
381 PEBS_EVENT_CONSTRAINT(0x00c5, 0xf), /* BR_INST_RETIRED.MISPRED */ 383 PEBS_EVENT_CONSTRAINT(0x010c, 0xf), /* MEM_STORE_RETIRED.DTLB_MISS */
382 PEBS_EVENT_CONSTRAINT(0x1fc7, 0xf), /* SIMD_INST_RETURED.ANY */ 384 INTEL_EVENT_CONSTRAINT(0xc0, 0xf), /* INST_RETIRED.ANY */
383 PEBS_EVENT_CONSTRAINT(0x01cb, 0xf), /* MEM_LOAD_RETIRED.L1D_MISS */ 385 INTEL_EVENT_CONSTRAINT(0xc2, 0xf), /* UOPS_RETIRED.* */
384 PEBS_EVENT_CONSTRAINT(0x02cb, 0xf), /* MEM_LOAD_RETIRED.L1D_LINE_MISS */ 386 INTEL_EVENT_CONSTRAINT(0xc4, 0xf), /* BR_INST_RETIRED.* */
385 PEBS_EVENT_CONSTRAINT(0x04cb, 0xf), /* MEM_LOAD_RETIRED.L2_MISS */ 387 PEBS_EVENT_CONSTRAINT(0x02c5, 0xf), /* BR_MISP_RETIRED.NEAR_CALL */
386 PEBS_EVENT_CONSTRAINT(0x08cb, 0xf), /* MEM_LOAD_RETIRED.L2_LINE_MISS */ 388 INTEL_EVENT_CONSTRAINT(0xc7, 0xf), /* SSEX_UOPS_RETIRED.* */
387 PEBS_EVENT_CONSTRAINT(0x10cb, 0xf), /* MEM_LOAD_RETIRED.DTLB_MISS */ 389 PEBS_EVENT_CONSTRAINT(0x20c8, 0xf), /* ITLB_MISS_RETIRED */
390 INTEL_EVENT_CONSTRAINT(0xcb, 0xf), /* MEM_LOAD_RETIRED.* */
391 INTEL_EVENT_CONSTRAINT(0xf7, 0xf), /* FP_ASSIST.* */
392 EVENT_CONSTRAINT_END
393};
394
395static struct event_constraint intel_westmere_pebs_event_constraints[] = {
396 INTEL_EVENT_CONSTRAINT(0x0b, 0xf), /* MEM_INST_RETIRED.* */
397 INTEL_EVENT_CONSTRAINT(0x0f, 0xf), /* MEM_UNCORE_RETIRED.* */
398 PEBS_EVENT_CONSTRAINT(0x010c, 0xf), /* MEM_STORE_RETIRED.DTLB_MISS */
399 INTEL_EVENT_CONSTRAINT(0xc0, 0xf), /* INSTR_RETIRED.* */
400 INTEL_EVENT_CONSTRAINT(0xc2, 0xf), /* UOPS_RETIRED.* */
401
402 INTEL_EVENT_CONSTRAINT(0xc4, 0xf), /* BR_INST_RETIRED.* */
403 INTEL_EVENT_CONSTRAINT(0xc5, 0xf), /* BR_MISP_RETIRED.* */
404 INTEL_EVENT_CONSTRAINT(0xc7, 0xf), /* SSEX_UOPS_RETIRED.* */
405 PEBS_EVENT_CONSTRAINT(0x20c8, 0xf), /* ITLB_MISS_RETIRED */
406 INTEL_EVENT_CONSTRAINT(0xcb, 0xf), /* MEM_LOAD_RETIRED.* */
407 INTEL_EVENT_CONSTRAINT(0xf7, 0xf), /* FP_ASSIST.* */
408 EVENT_CONSTRAINT_END
409};
410
411static struct event_constraint intel_snb_pebs_events[] = {
412 PEBS_EVENT_CONSTRAINT(0x01c0, 0x2), /* INST_RETIRED.PRECDIST */
413 PEBS_EVENT_CONSTRAINT(0x01c2, 0xf), /* UOPS_RETIRED.ALL */
414 PEBS_EVENT_CONSTRAINT(0x02c2, 0xf), /* UOPS_RETIRED.RETIRE_SLOTS */
415 PEBS_EVENT_CONSTRAINT(0x01c4, 0xf), /* BR_INST_RETIRED.CONDITIONAL */
416 PEBS_EVENT_CONSTRAINT(0x02c4, 0xf), /* BR_INST_RETIRED.NEAR_CALL */
417 PEBS_EVENT_CONSTRAINT(0x04c4, 0xf), /* BR_INST_RETIRED.ALL_BRANCHES */
418 PEBS_EVENT_CONSTRAINT(0x08c4, 0xf), /* BR_INST_RETIRED.NEAR_RETURN */
419 PEBS_EVENT_CONSTRAINT(0x10c4, 0xf), /* BR_INST_RETIRED.NOT_TAKEN */
420 PEBS_EVENT_CONSTRAINT(0x20c4, 0xf), /* BR_INST_RETIRED.NEAR_TAKEN */
421 PEBS_EVENT_CONSTRAINT(0x40c4, 0xf), /* BR_INST_RETIRED.FAR_BRANCH */
422 PEBS_EVENT_CONSTRAINT(0x01c5, 0xf), /* BR_MISP_RETIRED.CONDITIONAL */
423 PEBS_EVENT_CONSTRAINT(0x02c5, 0xf), /* BR_MISP_RETIRED.NEAR_CALL */
424 PEBS_EVENT_CONSTRAINT(0x04c5, 0xf), /* BR_MISP_RETIRED.ALL_BRANCHES */
425 PEBS_EVENT_CONSTRAINT(0x10c5, 0xf), /* BR_MISP_RETIRED.NOT_TAKEN */
426 PEBS_EVENT_CONSTRAINT(0x20c5, 0xf), /* BR_MISP_RETIRED.TAKEN */
427 PEBS_EVENT_CONSTRAINT(0x01cd, 0x8), /* MEM_TRANS_RETIRED.LOAD_LATENCY */
428 PEBS_EVENT_CONSTRAINT(0x02cd, 0x8), /* MEM_TRANS_RETIRED.PRECISE_STORE */
429 PEBS_EVENT_CONSTRAINT(0x11d0, 0xf), /* MEM_UOP_RETIRED.STLB_MISS_LOADS */
430 PEBS_EVENT_CONSTRAINT(0x12d0, 0xf), /* MEM_UOP_RETIRED.STLB_MISS_STORES */
431 PEBS_EVENT_CONSTRAINT(0x21d0, 0xf), /* MEM_UOP_RETIRED.LOCK_LOADS */
432 PEBS_EVENT_CONSTRAINT(0x22d0, 0xf), /* MEM_UOP_RETIRED.LOCK_STORES */
433 PEBS_EVENT_CONSTRAINT(0x41d0, 0xf), /* MEM_UOP_RETIRED.SPLIT_LOADS */
434 PEBS_EVENT_CONSTRAINT(0x42d0, 0xf), /* MEM_UOP_RETIRED.SPLIT_STORES */
435 PEBS_EVENT_CONSTRAINT(0x81d0, 0xf), /* MEM_UOP_RETIRED.ANY_LOADS */
436 PEBS_EVENT_CONSTRAINT(0x82d0, 0xf), /* MEM_UOP_RETIRED.ANY_STORES */
437 PEBS_EVENT_CONSTRAINT(0x01d1, 0xf), /* MEM_LOAD_UOPS_RETIRED.L1_HIT */
438 PEBS_EVENT_CONSTRAINT(0x02d1, 0xf), /* MEM_LOAD_UOPS_RETIRED.L2_HIT */
439 PEBS_EVENT_CONSTRAINT(0x04d1, 0xf), /* MEM_LOAD_UOPS_RETIRED.LLC_HIT */
440 PEBS_EVENT_CONSTRAINT(0x40d1, 0xf), /* MEM_LOAD_UOPS_RETIRED.HIT_LFB */
441 PEBS_EVENT_CONSTRAINT(0x01d2, 0xf), /* MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_MISS */
442 PEBS_EVENT_CONSTRAINT(0x02d2, 0xf), /* MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HIT */
443 PEBS_EVENT_CONSTRAINT(0x04d2, 0xf), /* MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HITM */
444 PEBS_EVENT_CONSTRAINT(0x08d2, 0xf), /* MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_NONE */
445 PEBS_EVENT_CONSTRAINT(0x02d4, 0xf), /* MEM_LOAD_UOPS_MISC_RETIRED.LLC_MISS */
388 EVENT_CONSTRAINT_END 446 EVENT_CONSTRAINT_END
389}; 447};
390 448
@@ -695,20 +753,17 @@ static void intel_ds_init(void)
695 printk(KERN_CONT "PEBS fmt0%c, ", pebs_type); 753 printk(KERN_CONT "PEBS fmt0%c, ", pebs_type);
696 x86_pmu.pebs_record_size = sizeof(struct pebs_record_core); 754 x86_pmu.pebs_record_size = sizeof(struct pebs_record_core);
697 x86_pmu.drain_pebs = intel_pmu_drain_pebs_core; 755 x86_pmu.drain_pebs = intel_pmu_drain_pebs_core;
698 x86_pmu.pebs_constraints = intel_core_pebs_events;
699 break; 756 break;
700 757
701 case 1: 758 case 1:
702 printk(KERN_CONT "PEBS fmt1%c, ", pebs_type); 759 printk(KERN_CONT "PEBS fmt1%c, ", pebs_type);
703 x86_pmu.pebs_record_size = sizeof(struct pebs_record_nhm); 760 x86_pmu.pebs_record_size = sizeof(struct pebs_record_nhm);
704 x86_pmu.drain_pebs = intel_pmu_drain_pebs_nhm; 761 x86_pmu.drain_pebs = intel_pmu_drain_pebs_nhm;
705 x86_pmu.pebs_constraints = intel_nehalem_pebs_events;
706 break; 762 break;
707 763
708 default: 764 default:
709 printk(KERN_CONT "no PEBS fmt%d%c, ", format, pebs_type); 765 printk(KERN_CONT "no PEBS fmt%d%c, ", format, pebs_type);
710 x86_pmu.pebs = 0; 766 x86_pmu.pebs = 0;
711 break;
712 } 767 }
713 } 768 }
714} 769}
diff --git a/arch/x86/kernel/cpu/perf_event_p4.c b/arch/x86/kernel/cpu/perf_event_p4.c
index ff751a9f182b..3769ac822f96 100644
--- a/arch/x86/kernel/cpu/perf_event_p4.c
+++ b/arch/x86/kernel/cpu/perf_event_p4.c
@@ -764,9 +764,9 @@ static inline int p4_pmu_clear_cccr_ovf(struct hw_perf_event *hwc)
764 u64 v; 764 u64 v;
765 765
766 /* an official way for overflow indication */ 766 /* an official way for overflow indication */
767 rdmsrl(hwc->config_base + hwc->idx, v); 767 rdmsrl(hwc->config_base, v);
768 if (v & P4_CCCR_OVF) { 768 if (v & P4_CCCR_OVF) {
769 wrmsrl(hwc->config_base + hwc->idx, v & ~P4_CCCR_OVF); 769 wrmsrl(hwc->config_base, v & ~P4_CCCR_OVF);
770 return 1; 770 return 1;
771 } 771 }
772 772
@@ -815,7 +815,7 @@ static inline void p4_pmu_disable_event(struct perf_event *event)
815 * state we need to clear P4_CCCR_OVF, otherwise interrupt get 815 * state we need to clear P4_CCCR_OVF, otherwise interrupt get
816 * asserted again and again 816 * asserted again and again
817 */ 817 */
818 (void)checking_wrmsrl(hwc->config_base + hwc->idx, 818 (void)checking_wrmsrl(hwc->config_base,
819 (u64)(p4_config_unpack_cccr(hwc->config)) & 819 (u64)(p4_config_unpack_cccr(hwc->config)) &
820 ~P4_CCCR_ENABLE & ~P4_CCCR_OVF & ~P4_CCCR_RESERVED); 820 ~P4_CCCR_ENABLE & ~P4_CCCR_OVF & ~P4_CCCR_RESERVED);
821} 821}
@@ -885,7 +885,7 @@ static void p4_pmu_enable_event(struct perf_event *event)
885 p4_pmu_enable_pebs(hwc->config); 885 p4_pmu_enable_pebs(hwc->config);
886 886
887 (void)checking_wrmsrl(escr_addr, escr_conf); 887 (void)checking_wrmsrl(escr_addr, escr_conf);
888 (void)checking_wrmsrl(hwc->config_base + hwc->idx, 888 (void)checking_wrmsrl(hwc->config_base,
889 (cccr & ~P4_CCCR_RESERVED) | P4_CCCR_ENABLE); 889 (cccr & ~P4_CCCR_RESERVED) | P4_CCCR_ENABLE);
890} 890}
891 891
diff --git a/arch/x86/kernel/cpu/perf_event_p6.c b/arch/x86/kernel/cpu/perf_event_p6.c
index 34ba07be2cda..20c097e33860 100644
--- a/arch/x86/kernel/cpu/perf_event_p6.c
+++ b/arch/x86/kernel/cpu/perf_event_p6.c
@@ -68,7 +68,7 @@ p6_pmu_disable_event(struct perf_event *event)
68 if (cpuc->enabled) 68 if (cpuc->enabled)
69 val |= ARCH_PERFMON_EVENTSEL_ENABLE; 69 val |= ARCH_PERFMON_EVENTSEL_ENABLE;
70 70
71 (void)checking_wrmsrl(hwc->config_base + hwc->idx, val); 71 (void)checking_wrmsrl(hwc->config_base, val);
72} 72}
73 73
74static void p6_pmu_enable_event(struct perf_event *event) 74static void p6_pmu_enable_event(struct perf_event *event)
@@ -81,7 +81,7 @@ static void p6_pmu_enable_event(struct perf_event *event)
81 if (cpuc->enabled) 81 if (cpuc->enabled)
82 val |= ARCH_PERFMON_EVENTSEL_ENABLE; 82 val |= ARCH_PERFMON_EVENTSEL_ENABLE;
83 83
84 (void)checking_wrmsrl(hwc->config_base + hwc->idx, val); 84 (void)checking_wrmsrl(hwc->config_base, val);
85} 85}
86 86
87static __initconst const struct x86_pmu p6_pmu = { 87static __initconst const struct x86_pmu p6_pmu = {
diff --git a/arch/x86/kernel/cpu/perfctr-watchdog.c b/arch/x86/kernel/cpu/perfctr-watchdog.c
index d5a236615501..966512b2cacf 100644
--- a/arch/x86/kernel/cpu/perfctr-watchdog.c
+++ b/arch/x86/kernel/cpu/perfctr-watchdog.c
@@ -46,6 +46,8 @@ static inline unsigned int nmi_perfctr_msr_to_bit(unsigned int msr)
46 /* returns the bit offset of the performance counter register */ 46 /* returns the bit offset of the performance counter register */
47 switch (boot_cpu_data.x86_vendor) { 47 switch (boot_cpu_data.x86_vendor) {
48 case X86_VENDOR_AMD: 48 case X86_VENDOR_AMD:
49 if (msr >= MSR_F15H_PERF_CTR)
50 return (msr - MSR_F15H_PERF_CTR) >> 1;
49 return msr - MSR_K7_PERFCTR0; 51 return msr - MSR_K7_PERFCTR0;
50 case X86_VENDOR_INTEL: 52 case X86_VENDOR_INTEL:
51 if (cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON)) 53 if (cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON))
@@ -70,6 +72,8 @@ static inline unsigned int nmi_evntsel_msr_to_bit(unsigned int msr)
70 /* returns the bit offset of the event selection register */ 72 /* returns the bit offset of the event selection register */
71 switch (boot_cpu_data.x86_vendor) { 73 switch (boot_cpu_data.x86_vendor) {
72 case X86_VENDOR_AMD: 74 case X86_VENDOR_AMD:
75 if (msr >= MSR_F15H_PERF_CTL)
76 return (msr - MSR_F15H_PERF_CTL) >> 1;
73 return msr - MSR_K7_EVNTSEL0; 77 return msr - MSR_K7_EVNTSEL0;
74 case X86_VENDOR_INTEL: 78 case X86_VENDOR_INTEL:
75 if (cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON)) 79 if (cpu_has(&boot_cpu_data, X86_FEATURE_ARCH_PERFMON))
diff --git a/arch/x86/kernel/devicetree.c b/arch/x86/kernel/devicetree.c
new file mode 100644
index 000000000000..7a8cebc9ff29
--- /dev/null
+++ b/arch/x86/kernel/devicetree.c
@@ -0,0 +1,441 @@
1/*
2 * Architecture specific OF callbacks.
3 */
4#include <linux/bootmem.h>
5#include <linux/io.h>
6#include <linux/interrupt.h>
7#include <linux/list.h>
8#include <linux/of.h>
9#include <linux/of_fdt.h>
10#include <linux/of_address.h>
11#include <linux/of_platform.h>
12#include <linux/of_irq.h>
13#include <linux/slab.h>
14#include <linux/pci.h>
15#include <linux/of_pci.h>
16
17#include <asm/hpet.h>
18#include <asm/irq_controller.h>
19#include <asm/apic.h>
20#include <asm/pci_x86.h>
21
22__initdata u64 initial_dtb;
23char __initdata cmd_line[COMMAND_LINE_SIZE];
24static LIST_HEAD(irq_domains);
25static DEFINE_RAW_SPINLOCK(big_irq_lock);
26
27int __initdata of_ioapic;
28
29#ifdef CONFIG_X86_IO_APIC
30static void add_interrupt_host(struct irq_domain *ih)
31{
32 unsigned long flags;
33
34 raw_spin_lock_irqsave(&big_irq_lock, flags);
35 list_add(&ih->l, &irq_domains);
36 raw_spin_unlock_irqrestore(&big_irq_lock, flags);
37}
38#endif
39
40static struct irq_domain *get_ih_from_node(struct device_node *controller)
41{
42 struct irq_domain *ih, *found = NULL;
43 unsigned long flags;
44
45 raw_spin_lock_irqsave(&big_irq_lock, flags);
46 list_for_each_entry(ih, &irq_domains, l) {
47 if (ih->controller == controller) {
48 found = ih;
49 break;
50 }
51 }
52 raw_spin_unlock_irqrestore(&big_irq_lock, flags);
53 return found;
54}
55
56unsigned int irq_create_of_mapping(struct device_node *controller,
57 const u32 *intspec, unsigned int intsize)
58{
59 struct irq_domain *ih;
60 u32 virq, type;
61 int ret;
62
63 ih = get_ih_from_node(controller);
64 if (!ih)
65 return 0;
66 ret = ih->xlate(ih, intspec, intsize, &virq, &type);
67 if (ret)
68 return ret;
69 if (type == IRQ_TYPE_NONE)
70 return virq;
71 /* set the mask if it is different from current */
72 if (type == (irq_to_desc(virq)->status & IRQF_TRIGGER_MASK))
73 set_irq_type(virq, type);
74 return virq;
75}
76EXPORT_SYMBOL_GPL(irq_create_of_mapping);
77
78unsigned long pci_address_to_pio(phys_addr_t address)
79{
80 /*
81 * The ioport address can be directly used by inX / outX
82 */
83 BUG_ON(address >= (1 << 16));
84 return (unsigned long)address;
85}
86EXPORT_SYMBOL_GPL(pci_address_to_pio);
87
88void __init early_init_dt_scan_chosen_arch(unsigned long node)
89{
90 BUG();
91}
92
93void __init early_init_dt_add_memory_arch(u64 base, u64 size)
94{
95 BUG();
96}
97
98void * __init early_init_dt_alloc_memory_arch(u64 size, u64 align)
99{
100 return __alloc_bootmem(size, align, __pa(MAX_DMA_ADDRESS));
101}
102
103void __init add_dtb(u64 data)
104{
105 initial_dtb = data + offsetof(struct setup_data, data);
106}
107
108/*
109 * CE4100 ids. Will be moved to machine_device_initcall() once we have it.
110 */
111static struct of_device_id __initdata ce4100_ids[] = {
112 { .compatible = "intel,ce4100-cp", },
113 { .compatible = "isa", },
114 { .compatible = "pci", },
115 {},
116};
117
118static int __init add_bus_probe(void)
119{
120 if (!of_have_populated_dt())
121 return 0;
122
123 return of_platform_bus_probe(NULL, ce4100_ids, NULL);
124}
125module_init(add_bus_probe);
126
127#ifdef CONFIG_PCI
128static int x86_of_pci_irq_enable(struct pci_dev *dev)
129{
130 struct of_irq oirq;
131 u32 virq;
132 int ret;
133 u8 pin;
134
135 ret = pci_read_config_byte(dev, PCI_INTERRUPT_PIN, &pin);
136 if (ret)
137 return ret;
138 if (!pin)
139 return 0;
140
141 ret = of_irq_map_pci(dev, &oirq);
142 if (ret)
143 return ret;
144
145 virq = irq_create_of_mapping(oirq.controller, oirq.specifier,
146 oirq.size);
147 if (virq == 0)
148 return -EINVAL;
149 dev->irq = virq;
150 return 0;
151}
152
153static void x86_of_pci_irq_disable(struct pci_dev *dev)
154{
155}
156
157void __cpuinit x86_of_pci_init(void)
158{
159 struct device_node *np;
160
161 pcibios_enable_irq = x86_of_pci_irq_enable;
162 pcibios_disable_irq = x86_of_pci_irq_disable;
163
164 for_each_node_by_type(np, "pci") {
165 const void *prop;
166 struct pci_bus *bus;
167 unsigned int bus_min;
168 struct device_node *child;
169
170 prop = of_get_property(np, "bus-range", NULL);
171 if (!prop)
172 continue;
173 bus_min = be32_to_cpup(prop);
174
175 bus = pci_find_bus(0, bus_min);
176 if (!bus) {
177 printk(KERN_ERR "Can't find a node for bus %s.\n",
178 np->full_name);
179 continue;
180 }
181
182 if (bus->self)
183 bus->self->dev.of_node = np;
184 else
185 bus->dev.of_node = np;
186
187 for_each_child_of_node(np, child) {
188 struct pci_dev *dev;
189 u32 devfn;
190
191 prop = of_get_property(child, "reg", NULL);
192 if (!prop)
193 continue;
194
195 devfn = (be32_to_cpup(prop) >> 8) & 0xff;
196 dev = pci_get_slot(bus, devfn);
197 if (!dev)
198 continue;
199 dev->dev.of_node = child;
200 pci_dev_put(dev);
201 }
202 }
203}
204#endif
205
206static void __init dtb_setup_hpet(void)
207{
208#ifdef CONFIG_HPET_TIMER
209 struct device_node *dn;
210 struct resource r;
211 int ret;
212
213 dn = of_find_compatible_node(NULL, NULL, "intel,ce4100-hpet");
214 if (!dn)
215 return;
216 ret = of_address_to_resource(dn, 0, &r);
217 if (ret) {
218 WARN_ON(1);
219 return;
220 }
221 hpet_address = r.start;
222#endif
223}
224
225static void __init dtb_lapic_setup(void)
226{
227#ifdef CONFIG_X86_LOCAL_APIC
228 struct device_node *dn;
229 struct resource r;
230 int ret;
231
232 dn = of_find_compatible_node(NULL, NULL, "intel,ce4100-lapic");
233 if (!dn)
234 return;
235
236 ret = of_address_to_resource(dn, 0, &r);
237 if (WARN_ON(ret))
238 return;
239
240 /* Did the boot loader setup the local APIC ? */
241 if (!cpu_has_apic) {
242 if (apic_force_enable(r.start))
243 return;
244 }
245 smp_found_config = 1;
246 pic_mode = 1;
247 register_lapic_address(r.start);
248 generic_processor_info(boot_cpu_physical_apicid,
249 GET_APIC_VERSION(apic_read(APIC_LVR)));
250#endif
251}
252
253#ifdef CONFIG_X86_IO_APIC
254static unsigned int ioapic_id;
255
256static void __init dtb_add_ioapic(struct device_node *dn)
257{
258 struct resource r;
259 int ret;
260
261 ret = of_address_to_resource(dn, 0, &r);
262 if (ret) {
263 printk(KERN_ERR "Can't obtain address from node %s.\n",
264 dn->full_name);
265 return;
266 }
267 mp_register_ioapic(++ioapic_id, r.start, gsi_top);
268}
269
270static void __init dtb_ioapic_setup(void)
271{
272 struct device_node *dn;
273
274 for_each_compatible_node(dn, NULL, "intel,ce4100-ioapic")
275 dtb_add_ioapic(dn);
276
277 if (nr_ioapics) {
278 of_ioapic = 1;
279 return;
280 }
281 printk(KERN_ERR "Error: No information about IO-APIC in OF.\n");
282}
283#else
284static void __init dtb_ioapic_setup(void) {}
285#endif
286
287static void __init dtb_apic_setup(void)
288{
289 dtb_lapic_setup();
290 dtb_ioapic_setup();
291}
292
293#ifdef CONFIG_OF_FLATTREE
294static void __init x86_flattree_get_config(void)
295{
296 u32 size, map_len;
297 void *new_dtb;
298
299 if (!initial_dtb)
300 return;
301
302 map_len = max(PAGE_SIZE - (initial_dtb & ~PAGE_MASK),
303 (u64)sizeof(struct boot_param_header));
304
305 initial_boot_params = early_memremap(initial_dtb, map_len);
306 size = be32_to_cpu(initial_boot_params->totalsize);
307 if (map_len < size) {
308 early_iounmap(initial_boot_params, map_len);
309 initial_boot_params = early_memremap(initial_dtb, size);
310 map_len = size;
311 }
312
313 new_dtb = alloc_bootmem(size);
314 memcpy(new_dtb, initial_boot_params, size);
315 early_iounmap(initial_boot_params, map_len);
316
317 initial_boot_params = new_dtb;
318
319 /* root level address cells */
320 of_scan_flat_dt(early_init_dt_scan_root, NULL);
321
322 unflatten_device_tree();
323}
324#else
325static inline void x86_flattree_get_config(void) { }
326#endif
327
328void __init x86_dtb_init(void)
329{
330 x86_flattree_get_config();
331
332 if (!of_have_populated_dt())
333 return;
334
335 dtb_setup_hpet();
336 dtb_apic_setup();
337}
338
339#ifdef CONFIG_X86_IO_APIC
340
341struct of_ioapic_type {
342 u32 out_type;
343 u32 trigger;
344 u32 polarity;
345};
346
347static struct of_ioapic_type of_ioapic_type[] =
348{
349 {
350 .out_type = IRQ_TYPE_EDGE_RISING,
351 .trigger = IOAPIC_EDGE,
352 .polarity = 1,
353 },
354 {
355 .out_type = IRQ_TYPE_LEVEL_LOW,
356 .trigger = IOAPIC_LEVEL,
357 .polarity = 0,
358 },
359 {
360 .out_type = IRQ_TYPE_LEVEL_HIGH,
361 .trigger = IOAPIC_LEVEL,
362 .polarity = 1,
363 },
364 {
365 .out_type = IRQ_TYPE_EDGE_FALLING,
366 .trigger = IOAPIC_EDGE,
367 .polarity = 0,
368 },
369};
370
371static int ioapic_xlate(struct irq_domain *id, const u32 *intspec, u32 intsize,
372 u32 *out_hwirq, u32 *out_type)
373{
374 struct io_apic_irq_attr attr;
375 struct of_ioapic_type *it;
376 u32 line, idx, type;
377
378 if (intsize < 2)
379 return -EINVAL;
380
381 line = *intspec;
382 idx = (u32) id->priv;
383 *out_hwirq = line + mp_gsi_routing[idx].gsi_base;
384
385 intspec++;
386 type = *intspec;
387
388 if (type >= ARRAY_SIZE(of_ioapic_type))
389 return -EINVAL;
390
391 it = of_ioapic_type + type;
392 *out_type = it->out_type;
393
394 set_io_apic_irq_attr(&attr, idx, line, it->trigger, it->polarity);
395
396 return io_apic_setup_irq_pin(*out_hwirq, cpu_to_node(0), &attr);
397}
398
399static void __init ioapic_add_ofnode(struct device_node *np)
400{
401 struct resource r;
402 int i, ret;
403
404 ret = of_address_to_resource(np, 0, &r);
405 if (ret) {
406 printk(KERN_ERR "Failed to obtain address for %s\n",
407 np->full_name);
408 return;
409 }
410
411 for (i = 0; i < nr_ioapics; i++) {
412 if (r.start == mp_ioapics[i].apicaddr) {
413 struct irq_domain *id;
414
415 id = kzalloc(sizeof(*id), GFP_KERNEL);
416 BUG_ON(!id);
417 id->controller = np;
418 id->xlate = ioapic_xlate;
419 id->priv = (void *)i;
420 add_interrupt_host(id);
421 return;
422 }
423 }
424 printk(KERN_ERR "IOxAPIC at %s is not registered.\n", np->full_name);
425}
426
427void __init x86_add_irq_domains(void)
428{
429 struct device_node *dp;
430
431 if (!of_have_populated_dt())
432 return;
433
434 for_each_node_with_property(dp, "interrupt-controller") {
435 if (of_device_is_compatible(dp, "intel,ce4100-ioapic"))
436 ioapic_add_ofnode(dp);
437 }
438}
439#else
440void __init x86_add_irq_domains(void) { }
441#endif
diff --git a/arch/x86/kernel/dumpstack.c b/arch/x86/kernel/dumpstack.c
index df20723a6a1b..220a1c11cfde 100644
--- a/arch/x86/kernel/dumpstack.c
+++ b/arch/x86/kernel/dumpstack.c
@@ -320,31 +320,6 @@ void die(const char *str, struct pt_regs *regs, long err)
320 oops_end(flags, regs, sig); 320 oops_end(flags, regs, sig);
321} 321}
322 322
323void notrace __kprobes
324die_nmi(char *str, struct pt_regs *regs, int do_panic)
325{
326 unsigned long flags;
327
328 if (notify_die(DIE_NMIWATCHDOG, str, regs, 0, 2, SIGINT) == NOTIFY_STOP)
329 return;
330
331 /*
332 * We are in trouble anyway, lets at least try
333 * to get a message out.
334 */
335 flags = oops_begin();
336 printk(KERN_EMERG "%s", str);
337 printk(" on CPU%d, ip %08lx, registers:\n",
338 smp_processor_id(), regs->ip);
339 show_registers(regs);
340 oops_end(flags, regs, 0);
341 if (do_panic || panic_on_oops)
342 panic("Non maskable interrupt");
343 nmi_exit();
344 local_irq_enable();
345 do_exit(SIGBUS);
346}
347
348static int __init oops_setup(char *s) 323static int __init oops_setup(char *s)
349{ 324{
350 if (!s) 325 if (!s)
diff --git a/arch/x86/kernel/e820.c b/arch/x86/kernel/e820.c
index 294f26da0c0c..cdf5bfd9d4d5 100644
--- a/arch/x86/kernel/e820.c
+++ b/arch/x86/kernel/e820.c
@@ -667,21 +667,15 @@ __init void e820_setup_gap(void)
667 * boot_params.e820_map, others are passed via SETUP_E820_EXT node of 667 * boot_params.e820_map, others are passed via SETUP_E820_EXT node of
668 * linked list of struct setup_data, which is parsed here. 668 * linked list of struct setup_data, which is parsed here.
669 */ 669 */
670void __init parse_e820_ext(struct setup_data *sdata, unsigned long pa_data) 670void __init parse_e820_ext(struct setup_data *sdata)
671{ 671{
672 u32 map_len;
673 int entries; 672 int entries;
674 struct e820entry *extmap; 673 struct e820entry *extmap;
675 674
676 entries = sdata->len / sizeof(struct e820entry); 675 entries = sdata->len / sizeof(struct e820entry);
677 map_len = sdata->len + sizeof(struct setup_data);
678 if (map_len > PAGE_SIZE)
679 sdata = early_ioremap(pa_data, map_len);
680 extmap = (struct e820entry *)(sdata->data); 676 extmap = (struct e820entry *)(sdata->data);
681 __append_e820_map(extmap, entries); 677 __append_e820_map(extmap, entries);
682 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map); 678 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
683 if (map_len > PAGE_SIZE)
684 early_iounmap(sdata, map_len);
685 printk(KERN_INFO "extended physical RAM map:\n"); 679 printk(KERN_INFO "extended physical RAM map:\n");
686 e820_print_map("extended"); 680 e820_print_map("extended");
687} 681}
@@ -847,15 +841,21 @@ static int __init parse_memopt(char *p)
847 if (!p) 841 if (!p)
848 return -EINVAL; 842 return -EINVAL;
849 843
850#ifdef CONFIG_X86_32
851 if (!strcmp(p, "nopentium")) { 844 if (!strcmp(p, "nopentium")) {
845#ifdef CONFIG_X86_32
852 setup_clear_cpu_cap(X86_FEATURE_PSE); 846 setup_clear_cpu_cap(X86_FEATURE_PSE);
853 return 0; 847 return 0;
854 } 848#else
849 printk(KERN_WARNING "mem=nopentium ignored! (only supported on x86_32)\n");
850 return -EINVAL;
855#endif 851#endif
852 }
856 853
857 userdef = 1; 854 userdef = 1;
858 mem_size = memparse(p, &p); 855 mem_size = memparse(p, &p);
856 /* don't remove all of memory when handling "mem={invalid}" param */
857 if (mem_size == 0)
858 return -EINVAL;
859 e820_remove_range(mem_size, ULLONG_MAX - mem_size, E820_RAM, 1); 859 e820_remove_range(mem_size, ULLONG_MAX - mem_size, E820_RAM, 1);
860 860
861 return 0; 861 return 0;
diff --git a/arch/x86/kernel/entry_32.S b/arch/x86/kernel/entry_32.S
index c8b4efad7ebb..fa41f7298c84 100644
--- a/arch/x86/kernel/entry_32.S
+++ b/arch/x86/kernel/entry_32.S
@@ -65,6 +65,8 @@
65#define sysexit_audit syscall_exit_work 65#define sysexit_audit syscall_exit_work
66#endif 66#endif
67 67
68 .section .entry.text, "ax"
69
68/* 70/*
69 * We use macros for low-level operations which need to be overridden 71 * We use macros for low-level operations which need to be overridden
70 * for paravirtualization. The following will never clobber any registers: 72 * for paravirtualization. The following will never clobber any registers:
@@ -395,7 +397,7 @@ sysenter_past_esp:
395 * A tiny bit of offset fixup is necessary - 4*4 means the 4 words 397 * A tiny bit of offset fixup is necessary - 4*4 means the 4 words
396 * pushed above; +8 corresponds to copy_thread's esp0 setting. 398 * pushed above; +8 corresponds to copy_thread's esp0 setting.
397 */ 399 */
398 pushl_cfi ((TI_sysenter_return)-THREAD_SIZE_asm+8+4*4)(%esp) 400 pushl_cfi ((TI_sysenter_return)-THREAD_SIZE+8+4*4)(%esp)
399 CFI_REL_OFFSET eip, 0 401 CFI_REL_OFFSET eip, 0
400 402
401 pushl_cfi %eax 403 pushl_cfi %eax
@@ -788,7 +790,7 @@ ENDPROC(ptregs_clone)
788 */ 790 */
789.section .init.rodata,"a" 791.section .init.rodata,"a"
790ENTRY(interrupt) 792ENTRY(interrupt)
791.text 793.section .entry.text, "ax"
792 .p2align 5 794 .p2align 5
793 .p2align CONFIG_X86_L1_CACHE_SHIFT 795 .p2align CONFIG_X86_L1_CACHE_SHIFT
794ENTRY(irq_entries_start) 796ENTRY(irq_entries_start)
@@ -807,7 +809,7 @@ vector=FIRST_EXTERNAL_VECTOR
807 .endif 809 .endif
808 .previous 810 .previous
809 .long 1b 811 .long 1b
810 .text 812 .section .entry.text, "ax"
811vector=vector+1 813vector=vector+1
812 .endif 814 .endif
813 .endr 815 .endr
@@ -1409,8 +1411,7 @@ END(general_protection)
1409#ifdef CONFIG_KVM_GUEST 1411#ifdef CONFIG_KVM_GUEST
1410ENTRY(async_page_fault) 1412ENTRY(async_page_fault)
1411 RING0_EC_FRAME 1413 RING0_EC_FRAME
1412 pushl $do_async_page_fault 1414 pushl_cfi $do_async_page_fault
1413 CFI_ADJUST_CFA_OFFSET 4
1414 jmp error_code 1415 jmp error_code
1415 CFI_ENDPROC 1416 CFI_ENDPROC
1416END(apf_page_fault) 1417END(apf_page_fault)
diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S
index aed1ffbeb0c9..b72b4a6466a9 100644
--- a/arch/x86/kernel/entry_64.S
+++ b/arch/x86/kernel/entry_64.S
@@ -61,6 +61,8 @@
61#define __AUDIT_ARCH_LE 0x40000000 61#define __AUDIT_ARCH_LE 0x40000000
62 62
63 .code64 63 .code64
64 .section .entry.text, "ax"
65
64#ifdef CONFIG_FUNCTION_TRACER 66#ifdef CONFIG_FUNCTION_TRACER
65#ifdef CONFIG_DYNAMIC_FTRACE 67#ifdef CONFIG_DYNAMIC_FTRACE
66ENTRY(mcount) 68ENTRY(mcount)
@@ -744,7 +746,7 @@ END(stub_rt_sigreturn)
744 */ 746 */
745 .section .init.rodata,"a" 747 .section .init.rodata,"a"
746ENTRY(interrupt) 748ENTRY(interrupt)
747 .text 749 .section .entry.text
748 .p2align 5 750 .p2align 5
749 .p2align CONFIG_X86_L1_CACHE_SHIFT 751 .p2align CONFIG_X86_L1_CACHE_SHIFT
750ENTRY(irq_entries_start) 752ENTRY(irq_entries_start)
@@ -763,7 +765,7 @@ vector=FIRST_EXTERNAL_VECTOR
763 .endif 765 .endif
764 .previous 766 .previous
765 .quad 1b 767 .quad 1b
766 .text 768 .section .entry.text
767vector=vector+1 769vector=vector+1
768 .endif 770 .endif
769 .endr 771 .endr
@@ -975,9 +977,12 @@ apicinterrupt X86_PLATFORM_IPI_VECTOR \
975 x86_platform_ipi smp_x86_platform_ipi 977 x86_platform_ipi smp_x86_platform_ipi
976 978
977#ifdef CONFIG_SMP 979#ifdef CONFIG_SMP
978.irpc idx, "01234567" 980.irp idx,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15, \
981 16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31
982.if NUM_INVALIDATE_TLB_VECTORS > \idx
979apicinterrupt (INVALIDATE_TLB_VECTOR_START)+\idx \ 983apicinterrupt (INVALIDATE_TLB_VECTOR_START)+\idx \
980 invalidate_interrupt\idx smp_invalidate_interrupt 984 invalidate_interrupt\idx smp_invalidate_interrupt
985.endif
981.endr 986.endr
982#endif 987#endif
983 988
@@ -1248,7 +1253,7 @@ ENTRY(xen_do_hypervisor_callback) # do_hypervisor_callback(struct *pt_regs)
1248 decl PER_CPU_VAR(irq_count) 1253 decl PER_CPU_VAR(irq_count)
1249 jmp error_exit 1254 jmp error_exit
1250 CFI_ENDPROC 1255 CFI_ENDPROC
1251END(do_hypervisor_callback) 1256END(xen_do_hypervisor_callback)
1252 1257
1253/* 1258/*
1254 * Hypervisor uses this for application faults while it executes. 1259 * Hypervisor uses this for application faults while it executes.
diff --git a/arch/x86/kernel/ftrace.c b/arch/x86/kernel/ftrace.c
index 382eb2936d4d..a93742a57468 100644
--- a/arch/x86/kernel/ftrace.c
+++ b/arch/x86/kernel/ftrace.c
@@ -437,18 +437,19 @@ void prepare_ftrace_return(unsigned long *parent, unsigned long self_addr,
437 return; 437 return;
438 } 438 }
439 439
440 if (ftrace_push_return_trace(old, self_addr, &trace.depth,
441 frame_pointer) == -EBUSY) {
442 *parent = old;
443 return;
444 }
445
446 trace.func = self_addr; 440 trace.func = self_addr;
441 trace.depth = current->curr_ret_stack + 1;
447 442
448 /* Only trace if the calling function expects to */ 443 /* Only trace if the calling function expects to */
449 if (!ftrace_graph_entry(&trace)) { 444 if (!ftrace_graph_entry(&trace)) {
450 current->curr_ret_stack--;
451 *parent = old; 445 *parent = old;
446 return;
447 }
448
449 if (ftrace_push_return_trace(old, self_addr, &trace.depth,
450 frame_pointer) == -EBUSY) {
451 *parent = old;
452 return;
452 } 453 }
453} 454}
454#endif /* CONFIG_FUNCTION_GRAPH_TRACER */ 455#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
diff --git a/arch/x86/kernel/head_32.S b/arch/x86/kernel/head_32.S
index 767d6c43de37..ce0be7cd085e 100644
--- a/arch/x86/kernel/head_32.S
+++ b/arch/x86/kernel/head_32.S
@@ -73,7 +73,7 @@ MAPPING_BEYOND_END = PAGE_TABLE_SIZE(LOWMEM_PAGES) << PAGE_SHIFT
73 */ 73 */
74KERNEL_PAGES = LOWMEM_PAGES 74KERNEL_PAGES = LOWMEM_PAGES
75 75
76INIT_MAP_SIZE = PAGE_TABLE_SIZE(KERNEL_PAGES) * PAGE_SIZE_asm 76INIT_MAP_SIZE = PAGE_TABLE_SIZE(KERNEL_PAGES) * PAGE_SIZE
77RESERVE_BRK(pagetables, INIT_MAP_SIZE) 77RESERVE_BRK(pagetables, INIT_MAP_SIZE)
78 78
79/* 79/*
@@ -137,7 +137,7 @@ ENTRY(startup_32)
137 movsl 137 movsl
1381: 1381:
139 139
140#ifdef CONFIG_OLPC_OPENFIRMWARE 140#ifdef CONFIG_OLPC
141 /* save OFW's pgdir table for later use when calling into OFW */ 141 /* save OFW's pgdir table for later use when calling into OFW */
142 movl %cr3, %eax 142 movl %cr3, %eax
143 movl %eax, pa(olpc_ofw_pgd) 143 movl %eax, pa(olpc_ofw_pgd)
@@ -623,7 +623,7 @@ ENTRY(initial_code)
623 * BSS section 623 * BSS section
624 */ 624 */
625__PAGE_ALIGNED_BSS 625__PAGE_ALIGNED_BSS
626 .align PAGE_SIZE_asm 626 .align PAGE_SIZE
627#ifdef CONFIG_X86_PAE 627#ifdef CONFIG_X86_PAE
628initial_pg_pmd: 628initial_pg_pmd:
629 .fill 1024*KPMDS,4,0 629 .fill 1024*KPMDS,4,0
@@ -644,7 +644,7 @@ ENTRY(swapper_pg_dir)
644#ifdef CONFIG_X86_PAE 644#ifdef CONFIG_X86_PAE
645__PAGE_ALIGNED_DATA 645__PAGE_ALIGNED_DATA
646 /* Page-aligned for the benefit of paravirt? */ 646 /* Page-aligned for the benefit of paravirt? */
647 .align PAGE_SIZE_asm 647 .align PAGE_SIZE
648ENTRY(initial_page_table) 648ENTRY(initial_page_table)
649 .long pa(initial_pg_pmd+PGD_IDENT_ATTR),0 /* low identity map */ 649 .long pa(initial_pg_pmd+PGD_IDENT_ATTR),0 /* low identity map */
650# if KPMDS == 3 650# if KPMDS == 3
@@ -662,7 +662,7 @@ ENTRY(initial_page_table)
662# else 662# else
663# error "Kernel PMDs should be 1, 2 or 3" 663# error "Kernel PMDs should be 1, 2 or 3"
664# endif 664# endif
665 .align PAGE_SIZE_asm /* needs to be page-sized too */ 665 .align PAGE_SIZE /* needs to be page-sized too */
666#endif 666#endif
667 667
668.data 668.data
diff --git a/arch/x86/kernel/hpet.c b/arch/x86/kernel/hpet.c
index 4ff5968f12d2..bfe8f729e086 100644
--- a/arch/x86/kernel/hpet.c
+++ b/arch/x86/kernel/hpet.c
@@ -503,7 +503,7 @@ static int hpet_assign_irq(struct hpet_dev *dev)
503 if (!irq) 503 if (!irq)
504 return -EINVAL; 504 return -EINVAL;
505 505
506 set_irq_data(irq, dev); 506 irq_set_handler_data(irq, dev);
507 507
508 if (hpet_setup_msi_irq(irq)) 508 if (hpet_setup_msi_irq(irq))
509 return -EINVAL; 509 return -EINVAL;
diff --git a/arch/x86/kernel/i8259.c b/arch/x86/kernel/i8259.c
index 20757cb2efa3..d9ca749c123b 100644
--- a/arch/x86/kernel/i8259.c
+++ b/arch/x86/kernel/i8259.c
@@ -112,7 +112,7 @@ static void make_8259A_irq(unsigned int irq)
112{ 112{
113 disable_irq_nosync(irq); 113 disable_irq_nosync(irq);
114 io_apic_irqs &= ~(1<<irq); 114 io_apic_irqs &= ~(1<<irq);
115 set_irq_chip_and_handler_name(irq, &i8259A_chip, handle_level_irq, 115 irq_set_chip_and_handler_name(irq, &i8259A_chip, handle_level_irq,
116 i8259A_chip.name); 116 i8259A_chip.name);
117 enable_irq(irq); 117 enable_irq(irq);
118} 118}
diff --git a/arch/x86/kernel/ioport.c b/arch/x86/kernel/ioport.c
index 8eec0ec59af2..8c968974253d 100644
--- a/arch/x86/kernel/ioport.c
+++ b/arch/x86/kernel/ioport.c
@@ -14,22 +14,9 @@
14#include <linux/slab.h> 14#include <linux/slab.h>
15#include <linux/thread_info.h> 15#include <linux/thread_info.h>
16#include <linux/syscalls.h> 16#include <linux/syscalls.h>
17#include <linux/bitmap.h>
17#include <asm/syscalls.h> 18#include <asm/syscalls.h>
18 19
19/* Set EXTENT bits starting at BASE in BITMAP to value TURN_ON. */
20static void set_bitmap(unsigned long *bitmap, unsigned int base,
21 unsigned int extent, int new_value)
22{
23 unsigned int i;
24
25 for (i = base; i < base + extent; i++) {
26 if (new_value)
27 __set_bit(i, bitmap);
28 else
29 __clear_bit(i, bitmap);
30 }
31}
32
33/* 20/*
34 * this changes the io permissions bitmap in the current task. 21 * this changes the io permissions bitmap in the current task.
35 */ 22 */
@@ -69,7 +56,10 @@ asmlinkage long sys_ioperm(unsigned long from, unsigned long num, int turn_on)
69 */ 56 */
70 tss = &per_cpu(init_tss, get_cpu()); 57 tss = &per_cpu(init_tss, get_cpu());
71 58
72 set_bitmap(t->io_bitmap_ptr, from, num, !turn_on); 59 if (turn_on)
60 bitmap_clear(t->io_bitmap_ptr, from, num);
61 else
62 bitmap_set(t->io_bitmap_ptr, from, num);
73 63
74 /* 64 /*
75 * Search for a (possibly new) maximum. This is simple and stupid, 65 * Search for a (possibly new) maximum. This is simple and stupid,
diff --git a/arch/x86/kernel/irq.c b/arch/x86/kernel/irq.c
index 387b6a0c9e81..948a31eae75f 100644
--- a/arch/x86/kernel/irq.c
+++ b/arch/x86/kernel/irq.c
@@ -44,9 +44,9 @@ void ack_bad_irq(unsigned int irq)
44 44
45#define irq_stats(x) (&per_cpu(irq_stat, x)) 45#define irq_stats(x) (&per_cpu(irq_stat, x))
46/* 46/*
47 * /proc/interrupts printing: 47 * /proc/interrupts printing for arch specific interrupts
48 */ 48 */
49static int show_other_interrupts(struct seq_file *p, int prec) 49int arch_show_interrupts(struct seq_file *p, int prec)
50{ 50{
51 int j; 51 int j;
52 52
@@ -122,59 +122,6 @@ static int show_other_interrupts(struct seq_file *p, int prec)
122 return 0; 122 return 0;
123} 123}
124 124
125int show_interrupts(struct seq_file *p, void *v)
126{
127 unsigned long flags, any_count = 0;
128 int i = *(loff_t *) v, j, prec;
129 struct irqaction *action;
130 struct irq_desc *desc;
131
132 if (i > nr_irqs)
133 return 0;
134
135 for (prec = 3, j = 1000; prec < 10 && j <= nr_irqs; ++prec)
136 j *= 10;
137
138 if (i == nr_irqs)
139 return show_other_interrupts(p, prec);
140
141 /* print header */
142 if (i == 0) {
143 seq_printf(p, "%*s", prec + 8, "");
144 for_each_online_cpu(j)
145 seq_printf(p, "CPU%-8d", j);
146 seq_putc(p, '\n');
147 }
148
149 desc = irq_to_desc(i);
150 if (!desc)
151 return 0;
152
153 raw_spin_lock_irqsave(&desc->lock, flags);
154 for_each_online_cpu(j)
155 any_count |= kstat_irqs_cpu(i, j);
156 action = desc->action;
157 if (!action && !any_count)
158 goto out;
159
160 seq_printf(p, "%*d: ", prec, i);
161 for_each_online_cpu(j)
162 seq_printf(p, "%10u ", kstat_irqs_cpu(i, j));
163 seq_printf(p, " %8s", desc->irq_data.chip->name);
164 seq_printf(p, "-%-8s", desc->name);
165
166 if (action) {
167 seq_printf(p, " %s", action->name);
168 while ((action = action->next) != NULL)
169 seq_printf(p, ", %s", action->name);
170 }
171
172 seq_putc(p, '\n');
173out:
174 raw_spin_unlock_irqrestore(&desc->lock, flags);
175 return 0;
176}
177
178/* 125/*
179 * /proc/stat helpers 126 * /proc/stat helpers
180 */ 127 */
@@ -276,15 +223,6 @@ void smp_x86_platform_ipi(struct pt_regs *regs)
276 223
277EXPORT_SYMBOL_GPL(vector_used_by_percpu_irq); 224EXPORT_SYMBOL_GPL(vector_used_by_percpu_irq);
278 225
279#ifdef CONFIG_OF
280unsigned int irq_create_of_mapping(struct device_node *controller,
281 const u32 *intspec, unsigned int intsize)
282{
283 return intspec[0];
284}
285EXPORT_SYMBOL_GPL(irq_create_of_mapping);
286#endif
287
288#ifdef CONFIG_HOTPLUG_CPU 226#ifdef CONFIG_HOTPLUG_CPU
289/* A cpu has been removed from cpu_online_mask. Reset irq affinities. */ 227/* A cpu has been removed from cpu_online_mask. Reset irq affinities. */
290void fixup_irqs(void) 228void fixup_irqs(void)
@@ -293,6 +231,7 @@ void fixup_irqs(void)
293 static int warned; 231 static int warned;
294 struct irq_desc *desc; 232 struct irq_desc *desc;
295 struct irq_data *data; 233 struct irq_data *data;
234 struct irq_chip *chip;
296 235
297 for_each_irq_desc(irq, desc) { 236 for_each_irq_desc(irq, desc) {
298 int break_affinity = 0; 237 int break_affinity = 0;
@@ -307,10 +246,10 @@ void fixup_irqs(void)
307 /* interrupt's are disabled at this point */ 246 /* interrupt's are disabled at this point */
308 raw_spin_lock(&desc->lock); 247 raw_spin_lock(&desc->lock);
309 248
310 data = &desc->irq_data; 249 data = irq_desc_get_irq_data(desc);
311 affinity = data->affinity; 250 affinity = data->affinity;
312 if (!irq_has_action(irq) || 251 if (!irq_has_action(irq) ||
313 cpumask_equal(affinity, cpu_online_mask)) { 252 cpumask_subset(affinity, cpu_online_mask)) {
314 raw_spin_unlock(&desc->lock); 253 raw_spin_unlock(&desc->lock);
315 continue; 254 continue;
316 } 255 }
@@ -327,16 +266,17 @@ void fixup_irqs(void)
327 affinity = cpu_all_mask; 266 affinity = cpu_all_mask;
328 } 267 }
329 268
330 if (!(desc->status & IRQ_MOVE_PCNTXT) && data->chip->irq_mask) 269 chip = irq_data_get_irq_chip(data);
331 data->chip->irq_mask(data); 270 if (!irqd_can_move_in_process_context(data) && chip->irq_mask)
271 chip->irq_mask(data);
332 272
333 if (data->chip->irq_set_affinity) 273 if (chip->irq_set_affinity)
334 data->chip->irq_set_affinity(data, affinity, true); 274 chip->irq_set_affinity(data, affinity, true);
335 else if (!(warned++)) 275 else if (!(warned++))
336 set_affinity = 0; 276 set_affinity = 0;
337 277
338 if (!(desc->status & IRQ_MOVE_PCNTXT) && data->chip->irq_unmask) 278 if (!irqd_can_move_in_process_context(data) && chip->irq_unmask)
339 data->chip->irq_unmask(data); 279 chip->irq_unmask(data);
340 280
341 raw_spin_unlock(&desc->lock); 281 raw_spin_unlock(&desc->lock);
342 282
@@ -368,10 +308,11 @@ void fixup_irqs(void)
368 irq = __this_cpu_read(vector_irq[vector]); 308 irq = __this_cpu_read(vector_irq[vector]);
369 309
370 desc = irq_to_desc(irq); 310 desc = irq_to_desc(irq);
371 data = &desc->irq_data; 311 data = irq_desc_get_irq_data(desc);
312 chip = irq_data_get_irq_chip(data);
372 raw_spin_lock(&desc->lock); 313 raw_spin_lock(&desc->lock);
373 if (data->chip->irq_retrigger) 314 if (chip->irq_retrigger)
374 data->chip->irq_retrigger(data); 315 chip->irq_retrigger(data);
375 raw_spin_unlock(&desc->lock); 316 raw_spin_unlock(&desc->lock);
376 } 317 }
377 } 318 }
diff --git a/arch/x86/kernel/irqinit.c b/arch/x86/kernel/irqinit.c
index c752e973958d..f470e4ef993e 100644
--- a/arch/x86/kernel/irqinit.c
+++ b/arch/x86/kernel/irqinit.c
@@ -25,6 +25,7 @@
25#include <asm/setup.h> 25#include <asm/setup.h>
26#include <asm/i8259.h> 26#include <asm/i8259.h>
27#include <asm/traps.h> 27#include <asm/traps.h>
28#include <asm/prom.h>
28 29
29/* 30/*
30 * ISA PIC or low IO-APIC triggered (INTA-cycle or APIC) interrupts: 31 * ISA PIC or low IO-APIC triggered (INTA-cycle or APIC) interrupts:
@@ -71,6 +72,7 @@ static irqreturn_t math_error_irq(int cpl, void *dev_id)
71static struct irqaction fpu_irq = { 72static struct irqaction fpu_irq = {
72 .handler = math_error_irq, 73 .handler = math_error_irq,
73 .name = "fpu", 74 .name = "fpu",
75 .flags = IRQF_NO_THREAD,
74}; 76};
75#endif 77#endif
76 78
@@ -80,6 +82,7 @@ static struct irqaction fpu_irq = {
80static struct irqaction irq2 = { 82static struct irqaction irq2 = {
81 .handler = no_action, 83 .handler = no_action,
82 .name = "cascade", 84 .name = "cascade",
85 .flags = IRQF_NO_THREAD,
83}; 86};
84 87
85DEFINE_PER_CPU(vector_irq_t, vector_irq) = { 88DEFINE_PER_CPU(vector_irq_t, vector_irq) = {
@@ -110,7 +113,7 @@ void __init init_ISA_irqs(void)
110 legacy_pic->init(0); 113 legacy_pic->init(0);
111 114
112 for (i = 0; i < legacy_pic->nr_legacy_irqs; i++) 115 for (i = 0; i < legacy_pic->nr_legacy_irqs; i++)
113 set_irq_chip_and_handler_name(i, chip, handle_level_irq, name); 116 irq_set_chip_and_handler_name(i, chip, handle_level_irq, name);
114} 117}
115 118
116void __init init_IRQ(void) 119void __init init_IRQ(void)
@@ -118,6 +121,12 @@ void __init init_IRQ(void)
118 int i; 121 int i;
119 122
120 /* 123 /*
124 * We probably need a better place for this, but it works for
125 * now ...
126 */
127 x86_add_irq_domains();
128
129 /*
121 * On cpu 0, Assign IRQ0_VECTOR..IRQ15_VECTOR's to IRQ 0..15. 130 * On cpu 0, Assign IRQ0_VECTOR..IRQ15_VECTOR's to IRQ 0..15.
122 * If these IRQ's are handled by legacy interrupt-controllers like PIC, 131 * If these IRQ's are handled by legacy interrupt-controllers like PIC,
123 * then this configuration will likely be static after the boot. If 132 * then this configuration will likely be static after the boot. If
@@ -164,14 +173,77 @@ static void __init smp_intr_init(void)
164 alloc_intr_gate(RESCHEDULE_VECTOR, reschedule_interrupt); 173 alloc_intr_gate(RESCHEDULE_VECTOR, reschedule_interrupt);
165 174
166 /* IPIs for invalidation */ 175 /* IPIs for invalidation */
167 alloc_intr_gate(INVALIDATE_TLB_VECTOR_START+0, invalidate_interrupt0); 176#define ALLOC_INVTLB_VEC(NR) \
168 alloc_intr_gate(INVALIDATE_TLB_VECTOR_START+1, invalidate_interrupt1); 177 alloc_intr_gate(INVALIDATE_TLB_VECTOR_START+NR, \
169 alloc_intr_gate(INVALIDATE_TLB_VECTOR_START+2, invalidate_interrupt2); 178 invalidate_interrupt##NR)
170 alloc_intr_gate(INVALIDATE_TLB_VECTOR_START+3, invalidate_interrupt3); 179
171 alloc_intr_gate(INVALIDATE_TLB_VECTOR_START+4, invalidate_interrupt4); 180 switch (NUM_INVALIDATE_TLB_VECTORS) {
172 alloc_intr_gate(INVALIDATE_TLB_VECTOR_START+5, invalidate_interrupt5); 181 default:
173 alloc_intr_gate(INVALIDATE_TLB_VECTOR_START+6, invalidate_interrupt6); 182 ALLOC_INVTLB_VEC(31);
174 alloc_intr_gate(INVALIDATE_TLB_VECTOR_START+7, invalidate_interrupt7); 183 case 31:
184 ALLOC_INVTLB_VEC(30);
185 case 30:
186 ALLOC_INVTLB_VEC(29);
187 case 29:
188 ALLOC_INVTLB_VEC(28);
189 case 28:
190 ALLOC_INVTLB_VEC(27);
191 case 27:
192 ALLOC_INVTLB_VEC(26);
193 case 26:
194 ALLOC_INVTLB_VEC(25);
195 case 25:
196 ALLOC_INVTLB_VEC(24);
197 case 24:
198 ALLOC_INVTLB_VEC(23);
199 case 23:
200 ALLOC_INVTLB_VEC(22);
201 case 22:
202 ALLOC_INVTLB_VEC(21);
203 case 21:
204 ALLOC_INVTLB_VEC(20);
205 case 20:
206 ALLOC_INVTLB_VEC(19);
207 case 19:
208 ALLOC_INVTLB_VEC(18);
209 case 18:
210 ALLOC_INVTLB_VEC(17);
211 case 17:
212 ALLOC_INVTLB_VEC(16);
213 case 16:
214 ALLOC_INVTLB_VEC(15);
215 case 15:
216 ALLOC_INVTLB_VEC(14);
217 case 14:
218 ALLOC_INVTLB_VEC(13);
219 case 13:
220 ALLOC_INVTLB_VEC(12);
221 case 12:
222 ALLOC_INVTLB_VEC(11);
223 case 11:
224 ALLOC_INVTLB_VEC(10);
225 case 10:
226 ALLOC_INVTLB_VEC(9);
227 case 9:
228 ALLOC_INVTLB_VEC(8);
229 case 8:
230 ALLOC_INVTLB_VEC(7);
231 case 7:
232 ALLOC_INVTLB_VEC(6);
233 case 6:
234 ALLOC_INVTLB_VEC(5);
235 case 5:
236 ALLOC_INVTLB_VEC(4);
237 case 4:
238 ALLOC_INVTLB_VEC(3);
239 case 3:
240 ALLOC_INVTLB_VEC(2);
241 case 2:
242 ALLOC_INVTLB_VEC(1);
243 case 1:
244 ALLOC_INVTLB_VEC(0);
245 break;
246 }
175 247
176 /* IPI for generic function call */ 248 /* IPI for generic function call */
177 alloc_intr_gate(CALL_FUNCTION_VECTOR, call_function_interrupt); 249 alloc_intr_gate(CALL_FUNCTION_VECTOR, call_function_interrupt);
@@ -243,7 +315,7 @@ void __init native_init_IRQ(void)
243 set_intr_gate(i, interrupt[i-FIRST_EXTERNAL_VECTOR]); 315 set_intr_gate(i, interrupt[i-FIRST_EXTERNAL_VECTOR]);
244 } 316 }
245 317
246 if (!acpi_ioapic) 318 if (!acpi_ioapic && !of_ioapic)
247 setup_irq(2, &irq2); 319 setup_irq(2, &irq2);
248 320
249#ifdef CONFIG_X86_32 321#ifdef CONFIG_X86_32
diff --git a/arch/x86/kernel/kgdb.c b/arch/x86/kernel/kgdb.c
index a4130005028a..7c64c420a9f6 100644
--- a/arch/x86/kernel/kgdb.c
+++ b/arch/x86/kernel/kgdb.c
@@ -533,15 +533,6 @@ static int __kgdb_notify(struct die_args *args, unsigned long cmd)
533 } 533 }
534 return NOTIFY_DONE; 534 return NOTIFY_DONE;
535 535
536 case DIE_NMIWATCHDOG:
537 if (atomic_read(&kgdb_active) != -1) {
538 /* KGDB CPU roundup: */
539 kgdb_nmicallback(raw_smp_processor_id(), regs);
540 return NOTIFY_STOP;
541 }
542 /* Enter debugger: */
543 break;
544
545 case DIE_DEBUG: 536 case DIE_DEBUG:
546 if (atomic_read(&kgdb_cpu_doing_single_step) != -1) { 537 if (atomic_read(&kgdb_cpu_doing_single_step) != -1) {
547 if (user_mode(regs)) 538 if (user_mode(regs))
diff --git a/arch/x86/kernel/kprobes.c b/arch/x86/kernel/kprobes.c
index d91c477b3f62..c969fd9d1566 100644
--- a/arch/x86/kernel/kprobes.c
+++ b/arch/x86/kernel/kprobes.c
@@ -1276,6 +1276,14 @@ static int __kprobes can_optimize(unsigned long paddr)
1276 if (!kallsyms_lookup_size_offset(paddr, &size, &offset)) 1276 if (!kallsyms_lookup_size_offset(paddr, &size, &offset))
1277 return 0; 1277 return 0;
1278 1278
1279 /*
1280 * Do not optimize in the entry code due to the unstable
1281 * stack handling.
1282 */
1283 if ((paddr >= (unsigned long )__entry_text_start) &&
1284 (paddr < (unsigned long )__entry_text_end))
1285 return 0;
1286
1279 /* Check there is enough space for a relative jump. */ 1287 /* Check there is enough space for a relative jump. */
1280 if (size - offset < RELATIVEJUMP_SIZE) 1288 if (size - offset < RELATIVEJUMP_SIZE)
1281 return 0; 1289 return 0;
diff --git a/arch/x86/kernel/microcode_amd.c b/arch/x86/kernel/microcode_amd.c
index 0fe6d1a66c38..c5610384ab16 100644
--- a/arch/x86/kernel/microcode_amd.c
+++ b/arch/x86/kernel/microcode_amd.c
@@ -66,7 +66,6 @@ struct microcode_amd {
66 unsigned int mpb[0]; 66 unsigned int mpb[0];
67}; 67};
68 68
69#define UCODE_MAX_SIZE 2048
70#define UCODE_CONTAINER_SECTION_HDR 8 69#define UCODE_CONTAINER_SECTION_HDR 8
71#define UCODE_CONTAINER_HEADER_SIZE 12 70#define UCODE_CONTAINER_HEADER_SIZE 12
72 71
@@ -77,20 +76,20 @@ static int collect_cpu_info_amd(int cpu, struct cpu_signature *csig)
77 struct cpuinfo_x86 *c = &cpu_data(cpu); 76 struct cpuinfo_x86 *c = &cpu_data(cpu);
78 u32 dummy; 77 u32 dummy;
79 78
80 memset(csig, 0, sizeof(*csig));
81 if (c->x86_vendor != X86_VENDOR_AMD || c->x86 < 0x10) { 79 if (c->x86_vendor != X86_VENDOR_AMD || c->x86 < 0x10) {
82 pr_warning("microcode: CPU%d: AMD CPU family 0x%x not " 80 pr_warning("CPU%d: family %d not supported\n", cpu, c->x86);
83 "supported\n", cpu, c->x86);
84 return -1; 81 return -1;
85 } 82 }
83
86 rdmsr(MSR_AMD64_PATCH_LEVEL, csig->rev, dummy); 84 rdmsr(MSR_AMD64_PATCH_LEVEL, csig->rev, dummy);
87 pr_info("CPU%d: patch_level=0x%x\n", cpu, csig->rev); 85 pr_info("CPU%d: patch_level=0x%08x\n", cpu, csig->rev);
86
88 return 0; 87 return 0;
89} 88}
90 89
91static int get_matching_microcode(int cpu, void *mc, int rev) 90static int get_matching_microcode(int cpu, struct microcode_header_amd *mc_hdr,
91 int rev)
92{ 92{
93 struct microcode_header_amd *mc_header = mc;
94 unsigned int current_cpu_id; 93 unsigned int current_cpu_id;
95 u16 equiv_cpu_id = 0; 94 u16 equiv_cpu_id = 0;
96 unsigned int i = 0; 95 unsigned int i = 0;
@@ -109,17 +108,17 @@ static int get_matching_microcode(int cpu, void *mc, int rev)
109 if (!equiv_cpu_id) 108 if (!equiv_cpu_id)
110 return 0; 109 return 0;
111 110
112 if (mc_header->processor_rev_id != equiv_cpu_id) 111 if (mc_hdr->processor_rev_id != equiv_cpu_id)
113 return 0; 112 return 0;
114 113
115 /* ucode might be chipset specific -- currently we don't support this */ 114 /* ucode might be chipset specific -- currently we don't support this */
116 if (mc_header->nb_dev_id || mc_header->sb_dev_id) { 115 if (mc_hdr->nb_dev_id || mc_hdr->sb_dev_id) {
117 pr_err("CPU%d: loading of chipset specific code not yet supported\n", 116 pr_err("CPU%d: chipset specific code not yet supported\n",
118 cpu); 117 cpu);
119 return 0; 118 return 0;
120 } 119 }
121 120
122 if (mc_header->patch_id <= rev) 121 if (mc_hdr->patch_id <= rev)
123 return 0; 122 return 0;
124 123
125 return 1; 124 return 1;
@@ -144,71 +143,93 @@ static int apply_microcode_amd(int cpu)
144 143
145 /* check current patch id and patch's id for match */ 144 /* check current patch id and patch's id for match */
146 if (rev != mc_amd->hdr.patch_id) { 145 if (rev != mc_amd->hdr.patch_id) {
147 pr_err("CPU%d: update failed (for patch_level=0x%x)\n", 146 pr_err("CPU%d: update failed for patch_level=0x%08x\n",
148 cpu, mc_amd->hdr.patch_id); 147 cpu, mc_amd->hdr.patch_id);
149 return -1; 148 return -1;
150 } 149 }
151 150
152 pr_info("CPU%d: updated (new patch_level=0x%x)\n", cpu, rev); 151 pr_info("CPU%d: new patch_level=0x%08x\n", cpu, rev);
153 uci->cpu_sig.rev = rev; 152 uci->cpu_sig.rev = rev;
154 153
155 return 0; 154 return 0;
156} 155}
157 156
158static void * 157static unsigned int verify_ucode_size(int cpu, const u8 *buf, unsigned int size)
159get_next_ucode(const u8 *buf, unsigned int size, unsigned int *mc_size)
160{ 158{
161 unsigned int total_size; 159 struct cpuinfo_x86 *c = &cpu_data(cpu);
162 u8 section_hdr[UCODE_CONTAINER_SECTION_HDR]; 160 unsigned int max_size, actual_size;
163 void *mc; 161
162#define F1XH_MPB_MAX_SIZE 2048
163#define F14H_MPB_MAX_SIZE 1824
164#define F15H_MPB_MAX_SIZE 4096
165
166 switch (c->x86) {
167 case 0x14:
168 max_size = F14H_MPB_MAX_SIZE;
169 break;
170 case 0x15:
171 max_size = F15H_MPB_MAX_SIZE;
172 break;
173 default:
174 max_size = F1XH_MPB_MAX_SIZE;
175 break;
176 }
164 177
165 get_ucode_data(section_hdr, buf, UCODE_CONTAINER_SECTION_HDR); 178 actual_size = buf[4] + (buf[5] << 8);
166 179
167 if (section_hdr[0] != UCODE_UCODE_TYPE) { 180 if (actual_size > size || actual_size > max_size) {
168 pr_err("error: invalid type field in container file section header\n"); 181 pr_err("section size mismatch\n");
169 return NULL; 182 return 0;
170 } 183 }
171 184
172 total_size = (unsigned long) (section_hdr[4] + (section_hdr[5] << 8)); 185 return actual_size;
186}
173 187
174 if (total_size > size || total_size > UCODE_MAX_SIZE) { 188static struct microcode_header_amd *
175 pr_err("error: size mismatch\n"); 189get_next_ucode(int cpu, const u8 *buf, unsigned int size, unsigned int *mc_size)
176 return NULL; 190{
191 struct microcode_header_amd *mc = NULL;
192 unsigned int actual_size = 0;
193
194 if (buf[0] != UCODE_UCODE_TYPE) {
195 pr_err("invalid type field in container file section header\n");
196 goto out;
177 } 197 }
178 198
179 mc = vzalloc(UCODE_MAX_SIZE); 199 actual_size = verify_ucode_size(cpu, buf, size);
200 if (!actual_size)
201 goto out;
202
203 mc = vzalloc(actual_size);
180 if (!mc) 204 if (!mc)
181 return NULL; 205 goto out;
182 206
183 get_ucode_data(mc, buf + UCODE_CONTAINER_SECTION_HDR, total_size); 207 get_ucode_data(mc, buf + UCODE_CONTAINER_SECTION_HDR, actual_size);
184 *mc_size = total_size + UCODE_CONTAINER_SECTION_HDR; 208 *mc_size = actual_size + UCODE_CONTAINER_SECTION_HDR;
185 209
210out:
186 return mc; 211 return mc;
187} 212}
188 213
189static int install_equiv_cpu_table(const u8 *buf) 214static int install_equiv_cpu_table(const u8 *buf)
190{ 215{
191 u8 *container_hdr[UCODE_CONTAINER_HEADER_SIZE]; 216 unsigned int *ibuf = (unsigned int *)buf;
192 unsigned int *buf_pos = (unsigned int *)container_hdr; 217 unsigned int type = ibuf[1];
193 unsigned long size; 218 unsigned int size = ibuf[2];
194 219
195 get_ucode_data(&container_hdr, buf, UCODE_CONTAINER_HEADER_SIZE); 220 if (type != UCODE_EQUIV_CPU_TABLE_TYPE || !size) {
196 221 pr_err("empty section/"
197 size = buf_pos[2]; 222 "invalid type field in container file section header\n");
198 223 return -EINVAL;
199 if (buf_pos[1] != UCODE_EQUIV_CPU_TABLE_TYPE || !size) {
200 pr_err("error: invalid type field in container file section header\n");
201 return 0;
202 } 224 }
203 225
204 equiv_cpu_table = vmalloc(size); 226 equiv_cpu_table = vmalloc(size);
205 if (!equiv_cpu_table) { 227 if (!equiv_cpu_table) {
206 pr_err("failed to allocate equivalent CPU table\n"); 228 pr_err("failed to allocate equivalent CPU table\n");
207 return 0; 229 return -ENOMEM;
208 } 230 }
209 231
210 buf += UCODE_CONTAINER_HEADER_SIZE; 232 get_ucode_data(equiv_cpu_table, buf + UCODE_CONTAINER_HEADER_SIZE, size);
211 get_ucode_data(equiv_cpu_table, buf, size);
212 233
213 return size + UCODE_CONTAINER_HEADER_SIZE; /* add header length */ 234 return size + UCODE_CONTAINER_HEADER_SIZE; /* add header length */
214} 235}
@@ -223,16 +244,16 @@ static enum ucode_state
223generic_load_microcode(int cpu, const u8 *data, size_t size) 244generic_load_microcode(int cpu, const u8 *data, size_t size)
224{ 245{
225 struct ucode_cpu_info *uci = ucode_cpu_info + cpu; 246 struct ucode_cpu_info *uci = ucode_cpu_info + cpu;
247 struct microcode_header_amd *mc_hdr = NULL;
248 unsigned int mc_size, leftover;
249 int offset;
226 const u8 *ucode_ptr = data; 250 const u8 *ucode_ptr = data;
227 void *new_mc = NULL; 251 void *new_mc = NULL;
228 void *mc; 252 unsigned int new_rev = uci->cpu_sig.rev;
229 int new_rev = uci->cpu_sig.rev;
230 unsigned int leftover;
231 unsigned long offset;
232 enum ucode_state state = UCODE_OK; 253 enum ucode_state state = UCODE_OK;
233 254
234 offset = install_equiv_cpu_table(ucode_ptr); 255 offset = install_equiv_cpu_table(ucode_ptr);
235 if (!offset) { 256 if (offset < 0) {
236 pr_err("failed to create equivalent cpu table\n"); 257 pr_err("failed to create equivalent cpu table\n");
237 return UCODE_ERROR; 258 return UCODE_ERROR;
238 } 259 }
@@ -241,64 +262,65 @@ generic_load_microcode(int cpu, const u8 *data, size_t size)
241 leftover = size - offset; 262 leftover = size - offset;
242 263
243 while (leftover) { 264 while (leftover) {
244 unsigned int uninitialized_var(mc_size); 265 mc_hdr = get_next_ucode(cpu, ucode_ptr, leftover, &mc_size);
245 struct microcode_header_amd *mc_header; 266 if (!mc_hdr)
246
247 mc = get_next_ucode(ucode_ptr, leftover, &mc_size);
248 if (!mc)
249 break; 267 break;
250 268
251 mc_header = (struct microcode_header_amd *)mc; 269 if (get_matching_microcode(cpu, mc_hdr, new_rev)) {
252 if (get_matching_microcode(cpu, mc, new_rev)) {
253 vfree(new_mc); 270 vfree(new_mc);
254 new_rev = mc_header->patch_id; 271 new_rev = mc_hdr->patch_id;
255 new_mc = mc; 272 new_mc = mc_hdr;
256 } else 273 } else
257 vfree(mc); 274 vfree(mc_hdr);
258 275
259 ucode_ptr += mc_size; 276 ucode_ptr += mc_size;
260 leftover -= mc_size; 277 leftover -= mc_size;
261 } 278 }
262 279
263 if (new_mc) { 280 if (!new_mc) {
264 if (!leftover) {
265 vfree(uci->mc);
266 uci->mc = new_mc;
267 pr_debug("CPU%d found a matching microcode update with version 0x%x (current=0x%x)\n",
268 cpu, new_rev, uci->cpu_sig.rev);
269 } else {
270 vfree(new_mc);
271 state = UCODE_ERROR;
272 }
273 } else
274 state = UCODE_NFOUND; 281 state = UCODE_NFOUND;
282 goto free_table;
283 }
275 284
285 if (!leftover) {
286 vfree(uci->mc);
287 uci->mc = new_mc;
288 pr_debug("CPU%d update ucode (0x%08x -> 0x%08x)\n",
289 cpu, uci->cpu_sig.rev, new_rev);
290 } else {
291 vfree(new_mc);
292 state = UCODE_ERROR;
293 }
294
295free_table:
276 free_equiv_cpu_table(); 296 free_equiv_cpu_table();
277 297
278 return state; 298 return state;
279} 299}
280 300
281static enum ucode_state request_microcode_fw(int cpu, struct device *device) 301static enum ucode_state request_microcode_amd(int cpu, struct device *device)
282{ 302{
283 const char *fw_name = "amd-ucode/microcode_amd.bin"; 303 const char *fw_name = "amd-ucode/microcode_amd.bin";
284 const struct firmware *firmware; 304 const struct firmware *fw;
285 enum ucode_state ret; 305 enum ucode_state ret = UCODE_NFOUND;
286 306
287 if (request_firmware(&firmware, fw_name, device)) { 307 if (request_firmware(&fw, fw_name, device)) {
288 printk(KERN_ERR "microcode: failed to load file %s\n", fw_name); 308 pr_err("failed to load file %s\n", fw_name);
289 return UCODE_NFOUND; 309 goto out;
290 } 310 }
291 311
292 if (*(u32 *)firmware->data != UCODE_MAGIC) { 312 ret = UCODE_ERROR;
293 pr_err("invalid UCODE_MAGIC (0x%08x)\n", 313 if (*(u32 *)fw->data != UCODE_MAGIC) {
294 *(u32 *)firmware->data); 314 pr_err("invalid magic value (0x%08x)\n", *(u32 *)fw->data);
295 return UCODE_ERROR; 315 goto fw_release;
296 } 316 }
297 317
298 ret = generic_load_microcode(cpu, firmware->data, firmware->size); 318 ret = generic_load_microcode(cpu, fw->data, fw->size);
299 319
300 release_firmware(firmware); 320fw_release:
321 release_firmware(fw);
301 322
323out:
302 return ret; 324 return ret;
303} 325}
304 326
@@ -319,7 +341,7 @@ static void microcode_fini_cpu_amd(int cpu)
319 341
320static struct microcode_ops microcode_amd_ops = { 342static struct microcode_ops microcode_amd_ops = {
321 .request_microcode_user = request_microcode_user, 343 .request_microcode_user = request_microcode_user,
322 .request_microcode_fw = request_microcode_fw, 344 .request_microcode_fw = request_microcode_amd,
323 .collect_cpu_info = collect_cpu_info_amd, 345 .collect_cpu_info = collect_cpu_info_amd,
324 .apply_microcode = apply_microcode_amd, 346 .apply_microcode = apply_microcode_amd,
325 .microcode_fini_cpu = microcode_fini_cpu_amd, 347 .microcode_fini_cpu = microcode_fini_cpu_amd,
diff --git a/arch/x86/kernel/microcode_core.c b/arch/x86/kernel/microcode_core.c
index 1cca374a2bac..87af68e0e1e1 100644
--- a/arch/x86/kernel/microcode_core.c
+++ b/arch/x86/kernel/microcode_core.c
@@ -417,8 +417,10 @@ static int mc_sysdev_add(struct sys_device *sys_dev)
417 if (err) 417 if (err)
418 return err; 418 return err;
419 419
420 if (microcode_init_cpu(cpu) == UCODE_ERROR) 420 if (microcode_init_cpu(cpu) == UCODE_ERROR) {
421 err = -EINVAL; 421 sysfs_remove_group(&sys_dev->kobj, &mc_attr_group);
422 return -EINVAL;
423 }
422 424
423 return err; 425 return err;
424} 426}
diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c
index ff4554198981..99fa3adf0141 100644
--- a/arch/x86/kernel/process.c
+++ b/arch/x86/kernel/process.c
@@ -110,12 +110,9 @@ void show_regs_common(void)
110 init_utsname()->release, 110 init_utsname()->release,
111 (int)strcspn(init_utsname()->version, " "), 111 (int)strcspn(init_utsname()->version, " "),
112 init_utsname()->version); 112 init_utsname()->version);
113 printk(KERN_CONT " "); 113 printk(KERN_CONT " %s %s", vendor, product);
114 printk(KERN_CONT "%s %s", vendor, product); 114 if (board)
115 if (board) { 115 printk(KERN_CONT "/%s", board);
116 printk(KERN_CONT "/");
117 printk(KERN_CONT "%s", board);
118 }
119 printk(KERN_CONT "\n"); 116 printk(KERN_CONT "\n");
120} 117}
121 118
diff --git a/arch/x86/kernel/rtc.c b/arch/x86/kernel/rtc.c
index 6f39cab052d5..3f2ad2640d85 100644
--- a/arch/x86/kernel/rtc.c
+++ b/arch/x86/kernel/rtc.c
@@ -6,6 +6,7 @@
6#include <linux/acpi.h> 6#include <linux/acpi.h>
7#include <linux/bcd.h> 7#include <linux/bcd.h>
8#include <linux/pnp.h> 8#include <linux/pnp.h>
9#include <linux/of.h>
9 10
10#include <asm/vsyscall.h> 11#include <asm/vsyscall.h>
11#include <asm/x86_init.h> 12#include <asm/x86_init.h>
@@ -236,6 +237,8 @@ static __init int add_rtc_cmos(void)
236 } 237 }
237 } 238 }
238#endif 239#endif
240 if (of_have_populated_dt())
241 return 0;
239 242
240 platform_device_register(&rtc_device); 243 platform_device_register(&rtc_device);
241 dev_info(&rtc_device.dev, 244 dev_info(&rtc_device.dev,
diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c
index d3cfe26c0252..b176f2b1f45d 100644
--- a/arch/x86/kernel/setup.c
+++ b/arch/x86/kernel/setup.c
@@ -113,6 +113,7 @@
113#endif 113#endif
114#include <asm/mce.h> 114#include <asm/mce.h>
115#include <asm/alternative.h> 115#include <asm/alternative.h>
116#include <asm/prom.h>
116 117
117/* 118/*
118 * end_pfn only includes RAM, while max_pfn_mapped includes all e820 entries. 119 * end_pfn only includes RAM, while max_pfn_mapped includes all e820 entries.
@@ -293,10 +294,32 @@ static void __init init_gbpages(void)
293 else 294 else
294 direct_gbpages = 0; 295 direct_gbpages = 0;
295} 296}
297
298static void __init cleanup_highmap_brk_end(void)
299{
300 pud_t *pud;
301 pmd_t *pmd;
302
303 mmu_cr4_features = read_cr4();
304
305 /*
306 * _brk_end cannot change anymore, but it and _end may be
307 * located on different 2M pages. cleanup_highmap(), however,
308 * can only consider _end when it runs, so destroy any
309 * mappings beyond _brk_end here.
310 */
311 pud = pud_offset(pgd_offset_k(_brk_end), _brk_end);
312 pmd = pmd_offset(pud, _brk_end - 1);
313 while (++pmd <= pmd_offset(pud, (unsigned long)_end - 1))
314 pmd_clear(pmd);
315}
296#else 316#else
297static inline void init_gbpages(void) 317static inline void init_gbpages(void)
298{ 318{
299} 319}
320static inline void cleanup_highmap_brk_end(void)
321{
322}
300#endif 323#endif
301 324
302static void __init reserve_brk(void) 325static void __init reserve_brk(void)
@@ -307,6 +330,8 @@ static void __init reserve_brk(void)
307 /* Mark brk area as locked down and no longer taking any 330 /* Mark brk area as locked down and no longer taking any
308 new allocations */ 331 new allocations */
309 _brk_start = 0; 332 _brk_start = 0;
333
334 cleanup_highmap_brk_end();
310} 335}
311 336
312#ifdef CONFIG_BLK_DEV_INITRD 337#ifdef CONFIG_BLK_DEV_INITRD
@@ -429,16 +454,30 @@ static void __init parse_setup_data(void)
429 return; 454 return;
430 pa_data = boot_params.hdr.setup_data; 455 pa_data = boot_params.hdr.setup_data;
431 while (pa_data) { 456 while (pa_data) {
432 data = early_memremap(pa_data, PAGE_SIZE); 457 u32 data_len, map_len;
458
459 map_len = max(PAGE_SIZE - (pa_data & ~PAGE_MASK),
460 (u64)sizeof(struct setup_data));
461 data = early_memremap(pa_data, map_len);
462 data_len = data->len + sizeof(struct setup_data);
463 if (data_len > map_len) {
464 early_iounmap(data, map_len);
465 data = early_memremap(pa_data, data_len);
466 map_len = data_len;
467 }
468
433 switch (data->type) { 469 switch (data->type) {
434 case SETUP_E820_EXT: 470 case SETUP_E820_EXT:
435 parse_e820_ext(data, pa_data); 471 parse_e820_ext(data);
472 break;
473 case SETUP_DTB:
474 add_dtb(pa_data);
436 break; 475 break;
437 default: 476 default:
438 break; 477 break;
439 } 478 }
440 pa_data = data->next; 479 pa_data = data->next;
441 early_iounmap(data, PAGE_SIZE); 480 early_iounmap(data, map_len);
442 } 481 }
443} 482}
444 483
@@ -680,15 +719,6 @@ static int __init parse_reservelow(char *p)
680 719
681early_param("reservelow", parse_reservelow); 720early_param("reservelow", parse_reservelow);
682 721
683static u64 __init get_max_mapped(void)
684{
685 u64 end = max_pfn_mapped;
686
687 end <<= PAGE_SHIFT;
688
689 return end;
690}
691
692/* 722/*
693 * Determine if we were loaded by an EFI loader. If so, then we have also been 723 * Determine if we were loaded by an EFI loader. If so, then we have also been
694 * passed the efi memmap, systab, etc., so we should use these data structures 724 * passed the efi memmap, systab, etc., so we should use these data structures
@@ -704,8 +734,6 @@ static u64 __init get_max_mapped(void)
704 734
705void __init setup_arch(char **cmdline_p) 735void __init setup_arch(char **cmdline_p)
706{ 736{
707 int acpi = 0;
708 int amd = 0;
709 unsigned long flags; 737 unsigned long flags;
710 738
711#ifdef CONFIG_X86_32 739#ifdef CONFIG_X86_32
@@ -984,19 +1012,7 @@ void __init setup_arch(char **cmdline_p)
984 1012
985 early_acpi_boot_init(); 1013 early_acpi_boot_init();
986 1014
987#ifdef CONFIG_ACPI_NUMA 1015 initmem_init();
988 /*
989 * Parse SRAT to discover nodes.
990 */
991 acpi = acpi_numa_init();
992#endif
993
994#ifdef CONFIG_AMD_NUMA
995 if (!acpi)
996 amd = !amd_numa_init(0, max_pfn);
997#endif
998
999 initmem_init(0, max_pfn, acpi, amd);
1000 memblock_find_dma_reserve(); 1016 memblock_find_dma_reserve();
1001 dma32_reserve_bootmem(); 1017 dma32_reserve_bootmem();
1002 1018
@@ -1029,8 +1045,8 @@ void __init setup_arch(char **cmdline_p)
1029 * Read APIC and some other early information from ACPI tables. 1045 * Read APIC and some other early information from ACPI tables.
1030 */ 1046 */
1031 acpi_boot_init(); 1047 acpi_boot_init();
1032
1033 sfi_init(); 1048 sfi_init();
1049 x86_dtb_init();
1034 1050
1035 /* 1051 /*
1036 * get boot-time SMP configuration: 1052 * get boot-time SMP configuration:
@@ -1040,9 +1056,7 @@ void __init setup_arch(char **cmdline_p)
1040 1056
1041 prefill_possible_map(); 1057 prefill_possible_map();
1042 1058
1043#ifdef CONFIG_X86_64
1044 init_cpu_to_node(); 1059 init_cpu_to_node();
1045#endif
1046 1060
1047 init_apic_mappings(); 1061 init_apic_mappings();
1048 ioapic_and_gsi_init(); 1062 ioapic_and_gsi_init();
@@ -1066,6 +1080,8 @@ void __init setup_arch(char **cmdline_p)
1066#endif 1080#endif
1067 x86_init.oem.banner(); 1081 x86_init.oem.banner();
1068 1082
1083 x86_init.timers.wallclock_init();
1084
1069 mcheck_init(); 1085 mcheck_init();
1070 1086
1071 local_irq_save(flags); 1087 local_irq_save(flags);
diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c
index 002b79685f73..71f4727da373 100644
--- a/arch/x86/kernel/setup_percpu.c
+++ b/arch/x86/kernel/setup_percpu.c
@@ -225,10 +225,15 @@ void __init setup_per_cpu_areas(void)
225 per_cpu(x86_bios_cpu_apicid, cpu) = 225 per_cpu(x86_bios_cpu_apicid, cpu) =
226 early_per_cpu_map(x86_bios_cpu_apicid, cpu); 226 early_per_cpu_map(x86_bios_cpu_apicid, cpu);
227#endif 227#endif
228#ifdef CONFIG_X86_32
229 per_cpu(x86_cpu_to_logical_apicid, cpu) =
230 early_per_cpu_map(x86_cpu_to_logical_apicid, cpu);
231#endif
228#ifdef CONFIG_X86_64 232#ifdef CONFIG_X86_64
229 per_cpu(irq_stack_ptr, cpu) = 233 per_cpu(irq_stack_ptr, cpu) =
230 per_cpu(irq_stack_union.irq_stack, cpu) + 234 per_cpu(irq_stack_union.irq_stack, cpu) +
231 IRQ_STACK_SIZE - 64; 235 IRQ_STACK_SIZE - 64;
236#endif
232#ifdef CONFIG_NUMA 237#ifdef CONFIG_NUMA
233 per_cpu(x86_cpu_to_node_map, cpu) = 238 per_cpu(x86_cpu_to_node_map, cpu) =
234 early_per_cpu_map(x86_cpu_to_node_map, cpu); 239 early_per_cpu_map(x86_cpu_to_node_map, cpu);
@@ -242,7 +247,6 @@ void __init setup_per_cpu_areas(void)
242 */ 247 */
243 set_cpu_numa_node(cpu, early_cpu_to_node(cpu)); 248 set_cpu_numa_node(cpu, early_cpu_to_node(cpu));
244#endif 249#endif
245#endif
246 /* 250 /*
247 * Up to this point, the boot CPU has been using .init.data 251 * Up to this point, the boot CPU has been using .init.data
248 * area. Reload any changed state for the boot CPU. 252 * area. Reload any changed state for the boot CPU.
@@ -256,7 +260,10 @@ void __init setup_per_cpu_areas(void)
256 early_per_cpu_ptr(x86_cpu_to_apicid) = NULL; 260 early_per_cpu_ptr(x86_cpu_to_apicid) = NULL;
257 early_per_cpu_ptr(x86_bios_cpu_apicid) = NULL; 261 early_per_cpu_ptr(x86_bios_cpu_apicid) = NULL;
258#endif 262#endif
259#if defined(CONFIG_X86_64) && defined(CONFIG_NUMA) 263#ifdef CONFIG_X86_32
264 early_per_cpu_ptr(x86_cpu_to_logical_apicid) = NULL;
265#endif
266#ifdef CONFIG_NUMA
260 early_per_cpu_ptr(x86_cpu_to_node_map) = NULL; 267 early_per_cpu_ptr(x86_cpu_to_node_map) = NULL;
261#endif 268#endif
262 269
diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index 1bfb1c615a62..e9efdfd51c8d 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -64,6 +64,7 @@
64#include <asm/mtrr.h> 64#include <asm/mtrr.h>
65#include <asm/mwait.h> 65#include <asm/mwait.h>
66#include <asm/apic.h> 66#include <asm/apic.h>
67#include <asm/io_apic.h>
67#include <asm/setup.h> 68#include <asm/setup.h>
68#include <asm/uv/uv.h> 69#include <asm/uv/uv.h>
69#include <linux/mc146818rtc.h> 70#include <linux/mc146818rtc.h>
@@ -71,10 +72,6 @@
71#include <asm/smpboot_hooks.h> 72#include <asm/smpboot_hooks.h>
72#include <asm/i8259.h> 73#include <asm/i8259.h>
73 74
74#ifdef CONFIG_X86_32
75u8 apicid_2_node[MAX_APICID];
76#endif
77
78/* State of each CPU */ 75/* State of each CPU */
79DEFINE_PER_CPU(int, cpu_state) = { 0 }; 76DEFINE_PER_CPU(int, cpu_state) = { 0 };
80 77
@@ -130,68 +127,14 @@ EXPORT_PER_CPU_SYMBOL(cpu_sibling_map);
130DEFINE_PER_CPU(cpumask_var_t, cpu_core_map); 127DEFINE_PER_CPU(cpumask_var_t, cpu_core_map);
131EXPORT_PER_CPU_SYMBOL(cpu_core_map); 128EXPORT_PER_CPU_SYMBOL(cpu_core_map);
132 129
130DEFINE_PER_CPU(cpumask_var_t, cpu_llc_shared_map);
131
133/* Per CPU bogomips and other parameters */ 132/* Per CPU bogomips and other parameters */
134DEFINE_PER_CPU_SHARED_ALIGNED(struct cpuinfo_x86, cpu_info); 133DEFINE_PER_CPU_SHARED_ALIGNED(struct cpuinfo_x86, cpu_info);
135EXPORT_PER_CPU_SYMBOL(cpu_info); 134EXPORT_PER_CPU_SYMBOL(cpu_info);
136 135
137atomic_t init_deasserted; 136atomic_t init_deasserted;
138 137
139#if defined(CONFIG_NUMA) && defined(CONFIG_X86_32)
140/* which node each logical CPU is on */
141int cpu_to_node_map[NR_CPUS] __read_mostly = { [0 ... NR_CPUS-1] = 0 };
142EXPORT_SYMBOL(cpu_to_node_map);
143
144/* set up a mapping between cpu and node. */
145static void map_cpu_to_node(int cpu, int node)
146{
147 printk(KERN_INFO "Mapping cpu %d to node %d\n", cpu, node);
148 cpumask_set_cpu(cpu, node_to_cpumask_map[node]);
149 cpu_to_node_map[cpu] = node;
150}
151
152/* undo a mapping between cpu and node. */
153static void unmap_cpu_to_node(int cpu)
154{
155 int node;
156
157 printk(KERN_INFO "Unmapping cpu %d from all nodes\n", cpu);
158 for (node = 0; node < MAX_NUMNODES; node++)
159 cpumask_clear_cpu(cpu, node_to_cpumask_map[node]);
160 cpu_to_node_map[cpu] = 0;
161}
162#else /* !(CONFIG_NUMA && CONFIG_X86_32) */
163#define map_cpu_to_node(cpu, node) ({})
164#define unmap_cpu_to_node(cpu) ({})
165#endif
166
167#ifdef CONFIG_X86_32
168static int boot_cpu_logical_apicid;
169
170u8 cpu_2_logical_apicid[NR_CPUS] __read_mostly =
171 { [0 ... NR_CPUS-1] = BAD_APICID };
172
173static void map_cpu_to_logical_apicid(void)
174{
175 int cpu = smp_processor_id();
176 int apicid = logical_smp_processor_id();
177 int node = apic->apicid_to_node(apicid);
178
179 if (!node_online(node))
180 node = first_online_node;
181
182 cpu_2_logical_apicid[cpu] = apicid;
183 map_cpu_to_node(cpu, node);
184}
185
186void numa_remove_cpu(int cpu)
187{
188 cpu_2_logical_apicid[cpu] = BAD_APICID;
189 unmap_cpu_to_node(cpu);
190}
191#else
192#define map_cpu_to_logical_apicid() do {} while (0)
193#endif
194
195/* 138/*
196 * Report back to the Boot Processor. 139 * Report back to the Boot Processor.
197 * Running on AP. 140 * Running on AP.
@@ -259,7 +202,6 @@ static void __cpuinit smp_callin(void)
259 apic->smp_callin_clear_local_apic(); 202 apic->smp_callin_clear_local_apic();
260 setup_local_APIC(); 203 setup_local_APIC();
261 end_local_APIC_setup(); 204 end_local_APIC_setup();
262 map_cpu_to_logical_apicid();
263 205
264 /* 206 /*
265 * Need to setup vector mappings before we enable interrupts. 207 * Need to setup vector mappings before we enable interrupts.
@@ -355,23 +297,6 @@ notrace static void __cpuinit start_secondary(void *unused)
355 cpu_idle(); 297 cpu_idle();
356} 298}
357 299
358#ifdef CONFIG_CPUMASK_OFFSTACK
359/* In this case, llc_shared_map is a pointer to a cpumask. */
360static inline void copy_cpuinfo_x86(struct cpuinfo_x86 *dst,
361 const struct cpuinfo_x86 *src)
362{
363 struct cpumask *llc = dst->llc_shared_map;
364 *dst = *src;
365 dst->llc_shared_map = llc;
366}
367#else
368static inline void copy_cpuinfo_x86(struct cpuinfo_x86 *dst,
369 const struct cpuinfo_x86 *src)
370{
371 *dst = *src;
372}
373#endif /* CONFIG_CPUMASK_OFFSTACK */
374
375/* 300/*
376 * The bootstrap kernel entry code has set these up. Save them for 301 * The bootstrap kernel entry code has set these up. Save them for
377 * a given CPU 302 * a given CPU
@@ -381,7 +306,7 @@ void __cpuinit smp_store_cpu_info(int id)
381{ 306{
382 struct cpuinfo_x86 *c = &cpu_data(id); 307 struct cpuinfo_x86 *c = &cpu_data(id);
383 308
384 copy_cpuinfo_x86(c, &boot_cpu_data); 309 *c = boot_cpu_data;
385 c->cpu_index = id; 310 c->cpu_index = id;
386 if (id != 0) 311 if (id != 0)
387 identify_secondary_cpu(c); 312 identify_secondary_cpu(c);
@@ -389,15 +314,12 @@ void __cpuinit smp_store_cpu_info(int id)
389 314
390static void __cpuinit link_thread_siblings(int cpu1, int cpu2) 315static void __cpuinit link_thread_siblings(int cpu1, int cpu2)
391{ 316{
392 struct cpuinfo_x86 *c1 = &cpu_data(cpu1);
393 struct cpuinfo_x86 *c2 = &cpu_data(cpu2);
394
395 cpumask_set_cpu(cpu1, cpu_sibling_mask(cpu2)); 317 cpumask_set_cpu(cpu1, cpu_sibling_mask(cpu2));
396 cpumask_set_cpu(cpu2, cpu_sibling_mask(cpu1)); 318 cpumask_set_cpu(cpu2, cpu_sibling_mask(cpu1));
397 cpumask_set_cpu(cpu1, cpu_core_mask(cpu2)); 319 cpumask_set_cpu(cpu1, cpu_core_mask(cpu2));
398 cpumask_set_cpu(cpu2, cpu_core_mask(cpu1)); 320 cpumask_set_cpu(cpu2, cpu_core_mask(cpu1));
399 cpumask_set_cpu(cpu1, c2->llc_shared_map); 321 cpumask_set_cpu(cpu1, cpu_llc_shared_mask(cpu2));
400 cpumask_set_cpu(cpu2, c1->llc_shared_map); 322 cpumask_set_cpu(cpu2, cpu_llc_shared_mask(cpu1));
401} 323}
402 324
403 325
@@ -426,7 +348,7 @@ void __cpuinit set_cpu_sibling_map(int cpu)
426 cpumask_set_cpu(cpu, cpu_sibling_mask(cpu)); 348 cpumask_set_cpu(cpu, cpu_sibling_mask(cpu));
427 } 349 }
428 350
429 cpumask_set_cpu(cpu, c->llc_shared_map); 351 cpumask_set_cpu(cpu, cpu_llc_shared_mask(cpu));
430 352
431 if (__this_cpu_read(cpu_info.x86_max_cores) == 1) { 353 if (__this_cpu_read(cpu_info.x86_max_cores) == 1) {
432 cpumask_copy(cpu_core_mask(cpu), cpu_sibling_mask(cpu)); 354 cpumask_copy(cpu_core_mask(cpu), cpu_sibling_mask(cpu));
@@ -437,8 +359,8 @@ void __cpuinit set_cpu_sibling_map(int cpu)
437 for_each_cpu(i, cpu_sibling_setup_mask) { 359 for_each_cpu(i, cpu_sibling_setup_mask) {
438 if (per_cpu(cpu_llc_id, cpu) != BAD_APICID && 360 if (per_cpu(cpu_llc_id, cpu) != BAD_APICID &&
439 per_cpu(cpu_llc_id, cpu) == per_cpu(cpu_llc_id, i)) { 361 per_cpu(cpu_llc_id, cpu) == per_cpu(cpu_llc_id, i)) {
440 cpumask_set_cpu(i, c->llc_shared_map); 362 cpumask_set_cpu(i, cpu_llc_shared_mask(cpu));
441 cpumask_set_cpu(cpu, cpu_data(i).llc_shared_map); 363 cpumask_set_cpu(cpu, cpu_llc_shared_mask(i));
442 } 364 }
443 if (c->phys_proc_id == cpu_data(i).phys_proc_id) { 365 if (c->phys_proc_id == cpu_data(i).phys_proc_id) {
444 cpumask_set_cpu(i, cpu_core_mask(cpu)); 366 cpumask_set_cpu(i, cpu_core_mask(cpu));
@@ -477,7 +399,7 @@ const struct cpumask *cpu_coregroup_mask(int cpu)
477 !(cpu_has(c, X86_FEATURE_AMD_DCM))) 399 !(cpu_has(c, X86_FEATURE_AMD_DCM)))
478 return cpu_core_mask(cpu); 400 return cpu_core_mask(cpu);
479 else 401 else
480 return c->llc_shared_map; 402 return cpu_llc_shared_mask(cpu);
481} 403}
482 404
483static void impress_friends(void) 405static void impress_friends(void)
@@ -946,6 +868,14 @@ int __cpuinit native_cpu_up(unsigned int cpu)
946 return 0; 868 return 0;
947} 869}
948 870
871/**
872 * arch_disable_smp_support() - disables SMP support for x86 at runtime
873 */
874void arch_disable_smp_support(void)
875{
876 disable_ioapic_support();
877}
878
949/* 879/*
950 * Fall back to non SMP mode after errors. 880 * Fall back to non SMP mode after errors.
951 * 881 *
@@ -961,7 +891,6 @@ static __init void disable_smp(void)
961 physid_set_mask_of_physid(boot_cpu_physical_apicid, &phys_cpu_present_map); 891 physid_set_mask_of_physid(boot_cpu_physical_apicid, &phys_cpu_present_map);
962 else 892 else
963 physid_set_mask_of_physid(0, &phys_cpu_present_map); 893 physid_set_mask_of_physid(0, &phys_cpu_present_map);
964 map_cpu_to_logical_apicid();
965 cpumask_set_cpu(0, cpu_sibling_mask(0)); 894 cpumask_set_cpu(0, cpu_sibling_mask(0));
966 cpumask_set_cpu(0, cpu_core_mask(0)); 895 cpumask_set_cpu(0, cpu_core_mask(0));
967} 896}
@@ -1046,7 +975,7 @@ static int __init smp_sanity_check(unsigned max_cpus)
1046 "(tell your hw vendor)\n"); 975 "(tell your hw vendor)\n");
1047 } 976 }
1048 smpboot_clear_io_apic(); 977 smpboot_clear_io_apic();
1049 arch_disable_smp_support(); 978 disable_ioapic_support();
1050 return -1; 979 return -1;
1051 } 980 }
1052 981
@@ -1090,21 +1019,19 @@ void __init native_smp_prepare_cpus(unsigned int max_cpus)
1090 1019
1091 preempt_disable(); 1020 preempt_disable();
1092 smp_cpu_index_default(); 1021 smp_cpu_index_default();
1093 memcpy(__this_cpu_ptr(&cpu_info), &boot_cpu_data, sizeof(cpu_info)); 1022
1094 cpumask_copy(cpu_callin_mask, cpumask_of(0));
1095 mb();
1096 /* 1023 /*
1097 * Setup boot CPU information 1024 * Setup boot CPU information
1098 */ 1025 */
1099 smp_store_cpu_info(0); /* Final full version of the data */ 1026 smp_store_cpu_info(0); /* Final full version of the data */
1100#ifdef CONFIG_X86_32 1027 cpumask_copy(cpu_callin_mask, cpumask_of(0));
1101 boot_cpu_logical_apicid = logical_smp_processor_id(); 1028 mb();
1102#endif 1029
1103 current_thread_info()->cpu = 0; /* needed? */ 1030 current_thread_info()->cpu = 0; /* needed? */
1104 for_each_possible_cpu(i) { 1031 for_each_possible_cpu(i) {
1105 zalloc_cpumask_var(&per_cpu(cpu_sibling_map, i), GFP_KERNEL); 1032 zalloc_cpumask_var(&per_cpu(cpu_sibling_map, i), GFP_KERNEL);
1106 zalloc_cpumask_var(&per_cpu(cpu_core_map, i), GFP_KERNEL); 1033 zalloc_cpumask_var(&per_cpu(cpu_core_map, i), GFP_KERNEL);
1107 zalloc_cpumask_var(&cpu_data(i).llc_shared_map, GFP_KERNEL); 1034 zalloc_cpumask_var(&per_cpu(cpu_llc_shared_map, i), GFP_KERNEL);
1108 } 1035 }
1109 set_cpu_sibling_map(0); 1036 set_cpu_sibling_map(0);
1110 1037
@@ -1140,8 +1067,6 @@ void __init native_smp_prepare_cpus(unsigned int max_cpus)
1140 1067
1141 bsp_end_local_APIC_setup(); 1068 bsp_end_local_APIC_setup();
1142 1069
1143 map_cpu_to_logical_apicid();
1144
1145 if (apic->setup_portio_remap) 1070 if (apic->setup_portio_remap)
1146 apic->setup_portio_remap(); 1071 apic->setup_portio_remap();
1147 1072
diff --git a/arch/x86/kernel/syscall_table_32.S b/arch/x86/kernel/syscall_table_32.S
index b35786dc9b8f..5f181742e8f9 100644
--- a/arch/x86/kernel/syscall_table_32.S
+++ b/arch/x86/kernel/syscall_table_32.S
@@ -340,3 +340,6 @@ ENTRY(sys_call_table)
340 .long sys_fanotify_init 340 .long sys_fanotify_init
341 .long sys_fanotify_mark 341 .long sys_fanotify_mark
342 .long sys_prlimit64 /* 340 */ 342 .long sys_prlimit64 /* 340 */
343 .long sys_name_to_handle_at
344 .long sys_open_by_handle_at
345 .long sys_clock_adjtime
diff --git a/arch/x86/kernel/vmlinux.lds.S b/arch/x86/kernel/vmlinux.lds.S
index bf4700755184..0381e1f3baed 100644
--- a/arch/x86/kernel/vmlinux.lds.S
+++ b/arch/x86/kernel/vmlinux.lds.S
@@ -105,6 +105,7 @@ SECTIONS
105 SCHED_TEXT 105 SCHED_TEXT
106 LOCK_TEXT 106 LOCK_TEXT
107 KPROBES_TEXT 107 KPROBES_TEXT
108 ENTRY_TEXT
108 IRQENTRY_TEXT 109 IRQENTRY_TEXT
109 *(.fixup) 110 *(.fixup)
110 *(.gnu.warning) 111 *(.gnu.warning)
@@ -305,7 +306,7 @@ SECTIONS
305 } 306 }
306 307
307#if !defined(CONFIG_X86_64) || !defined(CONFIG_SMP) 308#if !defined(CONFIG_X86_64) || !defined(CONFIG_SMP)
308 PERCPU(THREAD_SIZE) 309 PERCPU(PAGE_SIZE)
309#endif 310#endif
310 311
311 . = ALIGN(PAGE_SIZE); 312 . = ALIGN(PAGE_SIZE);
diff --git a/arch/x86/kernel/x8664_ksyms_64.c b/arch/x86/kernel/x8664_ksyms_64.c
index 1b950d151e58..9796c2f3d074 100644
--- a/arch/x86/kernel/x8664_ksyms_64.c
+++ b/arch/x86/kernel/x8664_ksyms_64.c
@@ -52,6 +52,7 @@ extern void *__memcpy(void *, const void *, __kernel_size_t);
52EXPORT_SYMBOL(memset); 52EXPORT_SYMBOL(memset);
53EXPORT_SYMBOL(memcpy); 53EXPORT_SYMBOL(memcpy);
54EXPORT_SYMBOL(__memcpy); 54EXPORT_SYMBOL(__memcpy);
55EXPORT_SYMBOL(memmove);
55 56
56EXPORT_SYMBOL(empty_zero_page); 57EXPORT_SYMBOL(empty_zero_page);
57#ifndef CONFIG_PARAVIRT 58#ifndef CONFIG_PARAVIRT
diff --git a/arch/x86/kernel/x86_init.c b/arch/x86/kernel/x86_init.c
index ceb2911aa439..c11514e9128b 100644
--- a/arch/x86/kernel/x86_init.c
+++ b/arch/x86/kernel/x86_init.c
@@ -70,6 +70,7 @@ struct x86_init_ops x86_init __initdata = {
70 .setup_percpu_clockev = setup_boot_APIC_clock, 70 .setup_percpu_clockev = setup_boot_APIC_clock,
71 .tsc_pre_init = x86_init_noop, 71 .tsc_pre_init = x86_init_noop,
72 .timer_init = hpet_time_init, 72 .timer_init = hpet_time_init,
73 .wallclock_init = x86_init_noop,
73 }, 74 },
74 75
75 .iommu = { 76 .iommu = {
diff --git a/arch/x86/kvm/trace.h b/arch/x86/kvm/trace.h
index 1357d7cf4ec8..db932760ea82 100644
--- a/arch/x86/kvm/trace.h
+++ b/arch/x86/kvm/trace.h
@@ -62,21 +62,21 @@ TRACE_EVENT(kvm_hv_hypercall,
62 TP_ARGS(code, fast, rep_cnt, rep_idx, ingpa, outgpa), 62 TP_ARGS(code, fast, rep_cnt, rep_idx, ingpa, outgpa),
63 63
64 TP_STRUCT__entry( 64 TP_STRUCT__entry(
65 __field( __u16, code )
66 __field( bool, fast )
67 __field( __u16, rep_cnt ) 65 __field( __u16, rep_cnt )
68 __field( __u16, rep_idx ) 66 __field( __u16, rep_idx )
69 __field( __u64, ingpa ) 67 __field( __u64, ingpa )
70 __field( __u64, outgpa ) 68 __field( __u64, outgpa )
69 __field( __u16, code )
70 __field( bool, fast )
71 ), 71 ),
72 72
73 TP_fast_assign( 73 TP_fast_assign(
74 __entry->code = code;
75 __entry->fast = fast;
76 __entry->rep_cnt = rep_cnt; 74 __entry->rep_cnt = rep_cnt;
77 __entry->rep_idx = rep_idx; 75 __entry->rep_idx = rep_idx;
78 __entry->ingpa = ingpa; 76 __entry->ingpa = ingpa;
79 __entry->outgpa = outgpa; 77 __entry->outgpa = outgpa;
78 __entry->code = code;
79 __entry->fast = fast;
80 ), 80 ),
81 81
82 TP_printk("code 0x%x %s cnt 0x%x idx 0x%x in 0x%llx out 0x%llx", 82 TP_printk("code 0x%x %s cnt 0x%x idx 0x%x in 0x%llx out 0x%llx",
diff --git a/arch/x86/lguest/boot.c b/arch/x86/lguest/boot.c
index eba687f0cc0c..b9ec1c74943c 100644
--- a/arch/x86/lguest/boot.c
+++ b/arch/x86/lguest/boot.c
@@ -847,7 +847,7 @@ static void __init lguest_init_IRQ(void)
847void lguest_setup_irq(unsigned int irq) 847void lguest_setup_irq(unsigned int irq)
848{ 848{
849 irq_alloc_desc_at(irq, 0); 849 irq_alloc_desc_at(irq, 0);
850 set_irq_chip_and_handler_name(irq, &lguest_irq_controller, 850 irq_set_chip_and_handler_name(irq, &lguest_irq_controller,
851 handle_level_irq, "level"); 851 handle_level_irq, "level");
852} 852}
853 853
@@ -995,7 +995,7 @@ static void lguest_time_irq(unsigned int irq, struct irq_desc *desc)
995static void lguest_time_init(void) 995static void lguest_time_init(void)
996{ 996{
997 /* Set up the timer interrupt (0) to go to our simple timer routine */ 997 /* Set up the timer interrupt (0) to go to our simple timer routine */
998 set_irq_handler(0, lguest_time_irq); 998 irq_set_handler(0, lguest_time_irq);
999 999
1000 clocksource_register(&lguest_clock); 1000 clocksource_register(&lguest_clock);
1001 1001
diff --git a/arch/x86/lib/atomic64_386_32.S b/arch/x86/lib/atomic64_386_32.S
index 2cda60a06e65..e8e7e0d06f42 100644
--- a/arch/x86/lib/atomic64_386_32.S
+++ b/arch/x86/lib/atomic64_386_32.S
@@ -15,14 +15,12 @@
15 15
16/* if you want SMP support, implement these with real spinlocks */ 16/* if you want SMP support, implement these with real spinlocks */
17.macro LOCK reg 17.macro LOCK reg
18 pushfl 18 pushfl_cfi
19 CFI_ADJUST_CFA_OFFSET 4
20 cli 19 cli
21.endm 20.endm
22 21
23.macro UNLOCK reg 22.macro UNLOCK reg
24 popfl 23 popfl_cfi
25 CFI_ADJUST_CFA_OFFSET -4
26.endm 24.endm
27 25
28#define BEGIN(op) \ 26#define BEGIN(op) \
diff --git a/arch/x86/lib/atomic64_cx8_32.S b/arch/x86/lib/atomic64_cx8_32.S
index 71e080de3352..391a083674b4 100644
--- a/arch/x86/lib/atomic64_cx8_32.S
+++ b/arch/x86/lib/atomic64_cx8_32.S
@@ -14,14 +14,12 @@
14#include <asm/dwarf2.h> 14#include <asm/dwarf2.h>
15 15
16.macro SAVE reg 16.macro SAVE reg
17 pushl %\reg 17 pushl_cfi %\reg
18 CFI_ADJUST_CFA_OFFSET 4
19 CFI_REL_OFFSET \reg, 0 18 CFI_REL_OFFSET \reg, 0
20.endm 19.endm
21 20
22.macro RESTORE reg 21.macro RESTORE reg
23 popl %\reg 22 popl_cfi %\reg
24 CFI_ADJUST_CFA_OFFSET -4
25 CFI_RESTORE \reg 23 CFI_RESTORE \reg
26.endm 24.endm
27 25
diff --git a/arch/x86/lib/checksum_32.S b/arch/x86/lib/checksum_32.S
index adbccd0bbb78..78d16a554db0 100644
--- a/arch/x86/lib/checksum_32.S
+++ b/arch/x86/lib/checksum_32.S
@@ -50,11 +50,9 @@ unsigned int csum_partial(const unsigned char * buff, int len, unsigned int sum)
50 */ 50 */
51ENTRY(csum_partial) 51ENTRY(csum_partial)
52 CFI_STARTPROC 52 CFI_STARTPROC
53 pushl %esi 53 pushl_cfi %esi
54 CFI_ADJUST_CFA_OFFSET 4
55 CFI_REL_OFFSET esi, 0 54 CFI_REL_OFFSET esi, 0
56 pushl %ebx 55 pushl_cfi %ebx
57 CFI_ADJUST_CFA_OFFSET 4
58 CFI_REL_OFFSET ebx, 0 56 CFI_REL_OFFSET ebx, 0
59 movl 20(%esp),%eax # Function arg: unsigned int sum 57 movl 20(%esp),%eax # Function arg: unsigned int sum
60 movl 16(%esp),%ecx # Function arg: int len 58 movl 16(%esp),%ecx # Function arg: int len
@@ -132,11 +130,9 @@ ENTRY(csum_partial)
132 jz 8f 130 jz 8f
133 roll $8, %eax 131 roll $8, %eax
1348: 1328:
135 popl %ebx 133 popl_cfi %ebx
136 CFI_ADJUST_CFA_OFFSET -4
137 CFI_RESTORE ebx 134 CFI_RESTORE ebx
138 popl %esi 135 popl_cfi %esi
139 CFI_ADJUST_CFA_OFFSET -4
140 CFI_RESTORE esi 136 CFI_RESTORE esi
141 ret 137 ret
142 CFI_ENDPROC 138 CFI_ENDPROC
@@ -148,11 +144,9 @@ ENDPROC(csum_partial)
148 144
149ENTRY(csum_partial) 145ENTRY(csum_partial)
150 CFI_STARTPROC 146 CFI_STARTPROC
151 pushl %esi 147 pushl_cfi %esi
152 CFI_ADJUST_CFA_OFFSET 4
153 CFI_REL_OFFSET esi, 0 148 CFI_REL_OFFSET esi, 0
154 pushl %ebx 149 pushl_cfi %ebx
155 CFI_ADJUST_CFA_OFFSET 4
156 CFI_REL_OFFSET ebx, 0 150 CFI_REL_OFFSET ebx, 0
157 movl 20(%esp),%eax # Function arg: unsigned int sum 151 movl 20(%esp),%eax # Function arg: unsigned int sum
158 movl 16(%esp),%ecx # Function arg: int len 152 movl 16(%esp),%ecx # Function arg: int len
@@ -260,11 +254,9 @@ ENTRY(csum_partial)
260 jz 90f 254 jz 90f
261 roll $8, %eax 255 roll $8, %eax
26290: 25690:
263 popl %ebx 257 popl_cfi %ebx
264 CFI_ADJUST_CFA_OFFSET -4
265 CFI_RESTORE ebx 258 CFI_RESTORE ebx
266 popl %esi 259 popl_cfi %esi
267 CFI_ADJUST_CFA_OFFSET -4
268 CFI_RESTORE esi 260 CFI_RESTORE esi
269 ret 261 ret
270 CFI_ENDPROC 262 CFI_ENDPROC
@@ -309,14 +301,11 @@ ENTRY(csum_partial_copy_generic)
309 CFI_STARTPROC 301 CFI_STARTPROC
310 subl $4,%esp 302 subl $4,%esp
311 CFI_ADJUST_CFA_OFFSET 4 303 CFI_ADJUST_CFA_OFFSET 4
312 pushl %edi 304 pushl_cfi %edi
313 CFI_ADJUST_CFA_OFFSET 4
314 CFI_REL_OFFSET edi, 0 305 CFI_REL_OFFSET edi, 0
315 pushl %esi 306 pushl_cfi %esi
316 CFI_ADJUST_CFA_OFFSET 4
317 CFI_REL_OFFSET esi, 0 307 CFI_REL_OFFSET esi, 0
318 pushl %ebx 308 pushl_cfi %ebx
319 CFI_ADJUST_CFA_OFFSET 4
320 CFI_REL_OFFSET ebx, 0 309 CFI_REL_OFFSET ebx, 0
321 movl ARGBASE+16(%esp),%eax # sum 310 movl ARGBASE+16(%esp),%eax # sum
322 movl ARGBASE+12(%esp),%ecx # len 311 movl ARGBASE+12(%esp),%ecx # len
@@ -426,17 +415,13 @@ DST( movb %cl, (%edi) )
426 415
427.previous 416.previous
428 417
429 popl %ebx 418 popl_cfi %ebx
430 CFI_ADJUST_CFA_OFFSET -4
431 CFI_RESTORE ebx 419 CFI_RESTORE ebx
432 popl %esi 420 popl_cfi %esi
433 CFI_ADJUST_CFA_OFFSET -4
434 CFI_RESTORE esi 421 CFI_RESTORE esi
435 popl %edi 422 popl_cfi %edi
436 CFI_ADJUST_CFA_OFFSET -4
437 CFI_RESTORE edi 423 CFI_RESTORE edi
438 popl %ecx # equivalent to addl $4,%esp 424 popl_cfi %ecx # equivalent to addl $4,%esp
439 CFI_ADJUST_CFA_OFFSET -4
440 ret 425 ret
441 CFI_ENDPROC 426 CFI_ENDPROC
442ENDPROC(csum_partial_copy_generic) 427ENDPROC(csum_partial_copy_generic)
@@ -459,14 +444,11 @@ ENDPROC(csum_partial_copy_generic)
459 444
460ENTRY(csum_partial_copy_generic) 445ENTRY(csum_partial_copy_generic)
461 CFI_STARTPROC 446 CFI_STARTPROC
462 pushl %ebx 447 pushl_cfi %ebx
463 CFI_ADJUST_CFA_OFFSET 4
464 CFI_REL_OFFSET ebx, 0 448 CFI_REL_OFFSET ebx, 0
465 pushl %edi 449 pushl_cfi %edi
466 CFI_ADJUST_CFA_OFFSET 4
467 CFI_REL_OFFSET edi, 0 450 CFI_REL_OFFSET edi, 0
468 pushl %esi 451 pushl_cfi %esi
469 CFI_ADJUST_CFA_OFFSET 4
470 CFI_REL_OFFSET esi, 0 452 CFI_REL_OFFSET esi, 0
471 movl ARGBASE+4(%esp),%esi #src 453 movl ARGBASE+4(%esp),%esi #src
472 movl ARGBASE+8(%esp),%edi #dst 454 movl ARGBASE+8(%esp),%edi #dst
@@ -527,14 +509,11 @@ DST( movb %dl, (%edi) )
527 jmp 7b 509 jmp 7b
528.previous 510.previous
529 511
530 popl %esi 512 popl_cfi %esi
531 CFI_ADJUST_CFA_OFFSET -4
532 CFI_RESTORE esi 513 CFI_RESTORE esi
533 popl %edi 514 popl_cfi %edi
534 CFI_ADJUST_CFA_OFFSET -4
535 CFI_RESTORE edi 515 CFI_RESTORE edi
536 popl %ebx 516 popl_cfi %ebx
537 CFI_ADJUST_CFA_OFFSET -4
538 CFI_RESTORE ebx 517 CFI_RESTORE ebx
539 ret 518 ret
540 CFI_ENDPROC 519 CFI_ENDPROC
diff --git a/arch/x86/lib/memmove_64.S b/arch/x86/lib/memmove_64.S
new file mode 100644
index 000000000000..0ecb8433e5a8
--- /dev/null
+++ b/arch/x86/lib/memmove_64.S
@@ -0,0 +1,197 @@
1/*
2 * Normally compiler builtins are used, but sometimes the compiler calls out
3 * of line code. Based on asm-i386/string.h.
4 *
5 * This assembly file is re-written from memmove_64.c file.
6 * - Copyright 2011 Fenghua Yu <fenghua.yu@intel.com>
7 */
8#define _STRING_C
9#include <linux/linkage.h>
10#include <asm/dwarf2.h>
11
12#undef memmove
13
14/*
15 * Implement memmove(). This can handle overlap between src and dst.
16 *
17 * Input:
18 * rdi: dest
19 * rsi: src
20 * rdx: count
21 *
22 * Output:
23 * rax: dest
24 */
25ENTRY(memmove)
26 CFI_STARTPROC
27 /* Handle more 32bytes in loop */
28 mov %rdi, %rax
29 cmp $0x20, %rdx
30 jb 1f
31
32 /* Decide forward/backward copy mode */
33 cmp %rdi, %rsi
34 jb 2f
35
36 /*
37 * movsq instruction have many startup latency
38 * so we handle small size by general register.
39 */
40 cmp $680, %rdx
41 jb 3f
42 /*
43 * movsq instruction is only good for aligned case.
44 */
45
46 cmpb %dil, %sil
47 je 4f
483:
49 sub $0x20, %rdx
50 /*
51 * We gobble 32byts forward in each loop.
52 */
535:
54 sub $0x20, %rdx
55 movq 0*8(%rsi), %r11
56 movq 1*8(%rsi), %r10
57 movq 2*8(%rsi), %r9
58 movq 3*8(%rsi), %r8
59 leaq 4*8(%rsi), %rsi
60
61 movq %r11, 0*8(%rdi)
62 movq %r10, 1*8(%rdi)
63 movq %r9, 2*8(%rdi)
64 movq %r8, 3*8(%rdi)
65 leaq 4*8(%rdi), %rdi
66 jae 5b
67 addq $0x20, %rdx
68 jmp 1f
69 /*
70 * Handle data forward by movsq.
71 */
72 .p2align 4
734:
74 movq %rdx, %rcx
75 movq -8(%rsi, %rdx), %r11
76 lea -8(%rdi, %rdx), %r10
77 shrq $3, %rcx
78 rep movsq
79 movq %r11, (%r10)
80 jmp 13f
81 /*
82 * Handle data backward by movsq.
83 */
84 .p2align 4
857:
86 movq %rdx, %rcx
87 movq (%rsi), %r11
88 movq %rdi, %r10
89 leaq -8(%rsi, %rdx), %rsi
90 leaq -8(%rdi, %rdx), %rdi
91 shrq $3, %rcx
92 std
93 rep movsq
94 cld
95 movq %r11, (%r10)
96 jmp 13f
97
98 /*
99 * Start to prepare for backward copy.
100 */
101 .p2align 4
1022:
103 cmp $680, %rdx
104 jb 6f
105 cmp %dil, %sil
106 je 7b
1076:
108 /*
109 * Calculate copy position to tail.
110 */
111 addq %rdx, %rsi
112 addq %rdx, %rdi
113 subq $0x20, %rdx
114 /*
115 * We gobble 32byts backward in each loop.
116 */
1178:
118 subq $0x20, %rdx
119 movq -1*8(%rsi), %r11
120 movq -2*8(%rsi), %r10
121 movq -3*8(%rsi), %r9
122 movq -4*8(%rsi), %r8
123 leaq -4*8(%rsi), %rsi
124
125 movq %r11, -1*8(%rdi)
126 movq %r10, -2*8(%rdi)
127 movq %r9, -3*8(%rdi)
128 movq %r8, -4*8(%rdi)
129 leaq -4*8(%rdi), %rdi
130 jae 8b
131 /*
132 * Calculate copy position to head.
133 */
134 addq $0x20, %rdx
135 subq %rdx, %rsi
136 subq %rdx, %rdi
1371:
138 cmpq $16, %rdx
139 jb 9f
140 /*
141 * Move data from 16 bytes to 31 bytes.
142 */
143 movq 0*8(%rsi), %r11
144 movq 1*8(%rsi), %r10
145 movq -2*8(%rsi, %rdx), %r9
146 movq -1*8(%rsi, %rdx), %r8
147 movq %r11, 0*8(%rdi)
148 movq %r10, 1*8(%rdi)
149 movq %r9, -2*8(%rdi, %rdx)
150 movq %r8, -1*8(%rdi, %rdx)
151 jmp 13f
152 .p2align 4
1539:
154 cmpq $8, %rdx
155 jb 10f
156 /*
157 * Move data from 8 bytes to 15 bytes.
158 */
159 movq 0*8(%rsi), %r11
160 movq -1*8(%rsi, %rdx), %r10
161 movq %r11, 0*8(%rdi)
162 movq %r10, -1*8(%rdi, %rdx)
163 jmp 13f
16410:
165 cmpq $4, %rdx
166 jb 11f
167 /*
168 * Move data from 4 bytes to 7 bytes.
169 */
170 movl (%rsi), %r11d
171 movl -4(%rsi, %rdx), %r10d
172 movl %r11d, (%rdi)
173 movl %r10d, -4(%rdi, %rdx)
174 jmp 13f
17511:
176 cmp $2, %rdx
177 jb 12f
178 /*
179 * Move data from 2 bytes to 3 bytes.
180 */
181 movw (%rsi), %r11w
182 movw -2(%rsi, %rdx), %r10w
183 movw %r11w, (%rdi)
184 movw %r10w, -2(%rdi, %rdx)
185 jmp 13f
18612:
187 cmp $1, %rdx
188 jb 13f
189 /*
190 * Move data for 1 byte.
191 */
192 movb (%rsi), %r11b
193 movb %r11b, (%rdi)
19413:
195 retq
196 CFI_ENDPROC
197ENDPROC(memmove)
diff --git a/arch/x86/lib/memmove_64.c b/arch/x86/lib/memmove_64.c
deleted file mode 100644
index 6d0f0ec41b34..000000000000
--- a/arch/x86/lib/memmove_64.c
+++ /dev/null
@@ -1,192 +0,0 @@
1/* Normally compiler builtins are used, but sometimes the compiler calls out
2 of line code. Based on asm-i386/string.h.
3 */
4#define _STRING_C
5#include <linux/string.h>
6#include <linux/module.h>
7
8#undef memmove
9void *memmove(void *dest, const void *src, size_t count)
10{
11 unsigned long d0,d1,d2,d3,d4,d5,d6,d7;
12 char *ret;
13
14 __asm__ __volatile__(
15 /* Handle more 32bytes in loop */
16 "mov %2, %3\n\t"
17 "cmp $0x20, %0\n\t"
18 "jb 1f\n\t"
19
20 /* Decide forward/backward copy mode */
21 "cmp %2, %1\n\t"
22 "jb 2f\n\t"
23
24 /*
25 * movsq instruction have many startup latency
26 * so we handle small size by general register.
27 */
28 "cmp $680, %0\n\t"
29 "jb 3f\n\t"
30 /*
31 * movsq instruction is only good for aligned case.
32 */
33 "cmpb %%dil, %%sil\n\t"
34 "je 4f\n\t"
35 "3:\n\t"
36 "sub $0x20, %0\n\t"
37 /*
38 * We gobble 32byts forward in each loop.
39 */
40 "5:\n\t"
41 "sub $0x20, %0\n\t"
42 "movq 0*8(%1), %4\n\t"
43 "movq 1*8(%1), %5\n\t"
44 "movq 2*8(%1), %6\n\t"
45 "movq 3*8(%1), %7\n\t"
46 "leaq 4*8(%1), %1\n\t"
47
48 "movq %4, 0*8(%2)\n\t"
49 "movq %5, 1*8(%2)\n\t"
50 "movq %6, 2*8(%2)\n\t"
51 "movq %7, 3*8(%2)\n\t"
52 "leaq 4*8(%2), %2\n\t"
53 "jae 5b\n\t"
54 "addq $0x20, %0\n\t"
55 "jmp 1f\n\t"
56 /*
57 * Handle data forward by movsq.
58 */
59 ".p2align 4\n\t"
60 "4:\n\t"
61 "movq %0, %8\n\t"
62 "movq -8(%1, %0), %4\n\t"
63 "lea -8(%2, %0), %5\n\t"
64 "shrq $3, %8\n\t"
65 "rep movsq\n\t"
66 "movq %4, (%5)\n\t"
67 "jmp 13f\n\t"
68 /*
69 * Handle data backward by movsq.
70 */
71 ".p2align 4\n\t"
72 "7:\n\t"
73 "movq %0, %8\n\t"
74 "movq (%1), %4\n\t"
75 "movq %2, %5\n\t"
76 "leaq -8(%1, %0), %1\n\t"
77 "leaq -8(%2, %0), %2\n\t"
78 "shrq $3, %8\n\t"
79 "std\n\t"
80 "rep movsq\n\t"
81 "cld\n\t"
82 "movq %4, (%5)\n\t"
83 "jmp 13f\n\t"
84
85 /*
86 * Start to prepare for backward copy.
87 */
88 ".p2align 4\n\t"
89 "2:\n\t"
90 "cmp $680, %0\n\t"
91 "jb 6f \n\t"
92 "cmp %%dil, %%sil\n\t"
93 "je 7b \n\t"
94 "6:\n\t"
95 /*
96 * Calculate copy position to tail.
97 */
98 "addq %0, %1\n\t"
99 "addq %0, %2\n\t"
100 "subq $0x20, %0\n\t"
101 /*
102 * We gobble 32byts backward in each loop.
103 */
104 "8:\n\t"
105 "subq $0x20, %0\n\t"
106 "movq -1*8(%1), %4\n\t"
107 "movq -2*8(%1), %5\n\t"
108 "movq -3*8(%1), %6\n\t"
109 "movq -4*8(%1), %7\n\t"
110 "leaq -4*8(%1), %1\n\t"
111
112 "movq %4, -1*8(%2)\n\t"
113 "movq %5, -2*8(%2)\n\t"
114 "movq %6, -3*8(%2)\n\t"
115 "movq %7, -4*8(%2)\n\t"
116 "leaq -4*8(%2), %2\n\t"
117 "jae 8b\n\t"
118 /*
119 * Calculate copy position to head.
120 */
121 "addq $0x20, %0\n\t"
122 "subq %0, %1\n\t"
123 "subq %0, %2\n\t"
124 "1:\n\t"
125 "cmpq $16, %0\n\t"
126 "jb 9f\n\t"
127 /*
128 * Move data from 16 bytes to 31 bytes.
129 */
130 "movq 0*8(%1), %4\n\t"
131 "movq 1*8(%1), %5\n\t"
132 "movq -2*8(%1, %0), %6\n\t"
133 "movq -1*8(%1, %0), %7\n\t"
134 "movq %4, 0*8(%2)\n\t"
135 "movq %5, 1*8(%2)\n\t"
136 "movq %6, -2*8(%2, %0)\n\t"
137 "movq %7, -1*8(%2, %0)\n\t"
138 "jmp 13f\n\t"
139 ".p2align 4\n\t"
140 "9:\n\t"
141 "cmpq $8, %0\n\t"
142 "jb 10f\n\t"
143 /*
144 * Move data from 8 bytes to 15 bytes.
145 */
146 "movq 0*8(%1), %4\n\t"
147 "movq -1*8(%1, %0), %5\n\t"
148 "movq %4, 0*8(%2)\n\t"
149 "movq %5, -1*8(%2, %0)\n\t"
150 "jmp 13f\n\t"
151 "10:\n\t"
152 "cmpq $4, %0\n\t"
153 "jb 11f\n\t"
154 /*
155 * Move data from 4 bytes to 7 bytes.
156 */
157 "movl (%1), %4d\n\t"
158 "movl -4(%1, %0), %5d\n\t"
159 "movl %4d, (%2)\n\t"
160 "movl %5d, -4(%2, %0)\n\t"
161 "jmp 13f\n\t"
162 "11:\n\t"
163 "cmp $2, %0\n\t"
164 "jb 12f\n\t"
165 /*
166 * Move data from 2 bytes to 3 bytes.
167 */
168 "movw (%1), %4w\n\t"
169 "movw -2(%1, %0), %5w\n\t"
170 "movw %4w, (%2)\n\t"
171 "movw %5w, -2(%2, %0)\n\t"
172 "jmp 13f\n\t"
173 "12:\n\t"
174 "cmp $1, %0\n\t"
175 "jb 13f\n\t"
176 /*
177 * Move data for 1 byte.
178 */
179 "movb (%1), %4b\n\t"
180 "movb %4b, (%2)\n\t"
181 "13:\n\t"
182 : "=&d" (d0), "=&S" (d1), "=&D" (d2), "=&a" (ret) ,
183 "=r"(d3), "=r"(d4), "=r"(d5), "=r"(d6), "=&c" (d7)
184 :"0" (count),
185 "1" (src),
186 "2" (dest)
187 :"memory");
188
189 return ret;
190
191}
192EXPORT_SYMBOL(memmove);
diff --git a/arch/x86/lib/rwsem_64.S b/arch/x86/lib/rwsem_64.S
index 41fcf00e49df..67743977398b 100644
--- a/arch/x86/lib/rwsem_64.S
+++ b/arch/x86/lib/rwsem_64.S
@@ -23,43 +23,50 @@
23#include <asm/dwarf2.h> 23#include <asm/dwarf2.h>
24 24
25#define save_common_regs \ 25#define save_common_regs \
26 pushq %rdi; \ 26 pushq_cfi %rdi; CFI_REL_OFFSET rdi, 0; \
27 pushq %rsi; \ 27 pushq_cfi %rsi; CFI_REL_OFFSET rsi, 0; \
28 pushq %rcx; \ 28 pushq_cfi %rcx; CFI_REL_OFFSET rcx, 0; \
29 pushq %r8; \ 29 pushq_cfi %r8; CFI_REL_OFFSET r8, 0; \
30 pushq %r9; \ 30 pushq_cfi %r9; CFI_REL_OFFSET r9, 0; \
31 pushq %r10; \ 31 pushq_cfi %r10; CFI_REL_OFFSET r10, 0; \
32 pushq %r11 32 pushq_cfi %r11; CFI_REL_OFFSET r11, 0
33 33
34#define restore_common_regs \ 34#define restore_common_regs \
35 popq %r11; \ 35 popq_cfi %r11; CFI_RESTORE r11; \
36 popq %r10; \ 36 popq_cfi %r10; CFI_RESTORE r10; \
37 popq %r9; \ 37 popq_cfi %r9; CFI_RESTORE r9; \
38 popq %r8; \ 38 popq_cfi %r8; CFI_RESTORE r8; \
39 popq %rcx; \ 39 popq_cfi %rcx; CFI_RESTORE rcx; \
40 popq %rsi; \ 40 popq_cfi %rsi; CFI_RESTORE rsi; \
41 popq %rdi 41 popq_cfi %rdi; CFI_RESTORE rdi
42 42
43/* Fix up special calling conventions */ 43/* Fix up special calling conventions */
44ENTRY(call_rwsem_down_read_failed) 44ENTRY(call_rwsem_down_read_failed)
45 CFI_STARTPROC
45 save_common_regs 46 save_common_regs
46 pushq %rdx 47 pushq_cfi %rdx
48 CFI_REL_OFFSET rdx, 0
47 movq %rax,%rdi 49 movq %rax,%rdi
48 call rwsem_down_read_failed 50 call rwsem_down_read_failed
49 popq %rdx 51 popq_cfi %rdx
52 CFI_RESTORE rdx
50 restore_common_regs 53 restore_common_regs
51 ret 54 ret
52 ENDPROC(call_rwsem_down_read_failed) 55 CFI_ENDPROC
56ENDPROC(call_rwsem_down_read_failed)
53 57
54ENTRY(call_rwsem_down_write_failed) 58ENTRY(call_rwsem_down_write_failed)
59 CFI_STARTPROC
55 save_common_regs 60 save_common_regs
56 movq %rax,%rdi 61 movq %rax,%rdi
57 call rwsem_down_write_failed 62 call rwsem_down_write_failed
58 restore_common_regs 63 restore_common_regs
59 ret 64 ret
60 ENDPROC(call_rwsem_down_write_failed) 65 CFI_ENDPROC
66ENDPROC(call_rwsem_down_write_failed)
61 67
62ENTRY(call_rwsem_wake) 68ENTRY(call_rwsem_wake)
69 CFI_STARTPROC
63 decl %edx /* do nothing if still outstanding active readers */ 70 decl %edx /* do nothing if still outstanding active readers */
64 jnz 1f 71 jnz 1f
65 save_common_regs 72 save_common_regs
@@ -67,15 +74,20 @@ ENTRY(call_rwsem_wake)
67 call rwsem_wake 74 call rwsem_wake
68 restore_common_regs 75 restore_common_regs
691: ret 761: ret
70 ENDPROC(call_rwsem_wake) 77 CFI_ENDPROC
78ENDPROC(call_rwsem_wake)
71 79
72/* Fix up special calling conventions */ 80/* Fix up special calling conventions */
73ENTRY(call_rwsem_downgrade_wake) 81ENTRY(call_rwsem_downgrade_wake)
82 CFI_STARTPROC
74 save_common_regs 83 save_common_regs
75 pushq %rdx 84 pushq_cfi %rdx
85 CFI_REL_OFFSET rdx, 0
76 movq %rax,%rdi 86 movq %rax,%rdi
77 call rwsem_downgrade_wake 87 call rwsem_downgrade_wake
78 popq %rdx 88 popq_cfi %rdx
89 CFI_RESTORE rdx
79 restore_common_regs 90 restore_common_regs
80 ret 91 ret
81 ENDPROC(call_rwsem_downgrade_wake) 92 CFI_ENDPROC
93ENDPROC(call_rwsem_downgrade_wake)
diff --git a/arch/x86/lib/semaphore_32.S b/arch/x86/lib/semaphore_32.S
index 648fe4741782..06691daa4108 100644
--- a/arch/x86/lib/semaphore_32.S
+++ b/arch/x86/lib/semaphore_32.S
@@ -36,7 +36,7 @@
36 */ 36 */
37#ifdef CONFIG_SMP 37#ifdef CONFIG_SMP
38ENTRY(__write_lock_failed) 38ENTRY(__write_lock_failed)
39 CFI_STARTPROC simple 39 CFI_STARTPROC
40 FRAME 40 FRAME
412: LOCK_PREFIX 412: LOCK_PREFIX
42 addl $ RW_LOCK_BIAS,(%eax) 42 addl $ RW_LOCK_BIAS,(%eax)
@@ -74,29 +74,23 @@ ENTRY(__read_lock_failed)
74/* Fix up special calling conventions */ 74/* Fix up special calling conventions */
75ENTRY(call_rwsem_down_read_failed) 75ENTRY(call_rwsem_down_read_failed)
76 CFI_STARTPROC 76 CFI_STARTPROC
77 push %ecx 77 pushl_cfi %ecx
78 CFI_ADJUST_CFA_OFFSET 4
79 CFI_REL_OFFSET ecx,0 78 CFI_REL_OFFSET ecx,0
80 push %edx 79 pushl_cfi %edx
81 CFI_ADJUST_CFA_OFFSET 4
82 CFI_REL_OFFSET edx,0 80 CFI_REL_OFFSET edx,0
83 call rwsem_down_read_failed 81 call rwsem_down_read_failed
84 pop %edx 82 popl_cfi %edx
85 CFI_ADJUST_CFA_OFFSET -4 83 popl_cfi %ecx
86 pop %ecx
87 CFI_ADJUST_CFA_OFFSET -4
88 ret 84 ret
89 CFI_ENDPROC 85 CFI_ENDPROC
90 ENDPROC(call_rwsem_down_read_failed) 86 ENDPROC(call_rwsem_down_read_failed)
91 87
92ENTRY(call_rwsem_down_write_failed) 88ENTRY(call_rwsem_down_write_failed)
93 CFI_STARTPROC 89 CFI_STARTPROC
94 push %ecx 90 pushl_cfi %ecx
95 CFI_ADJUST_CFA_OFFSET 4
96 CFI_REL_OFFSET ecx,0 91 CFI_REL_OFFSET ecx,0
97 calll rwsem_down_write_failed 92 calll rwsem_down_write_failed
98 pop %ecx 93 popl_cfi %ecx
99 CFI_ADJUST_CFA_OFFSET -4
100 ret 94 ret
101 CFI_ENDPROC 95 CFI_ENDPROC
102 ENDPROC(call_rwsem_down_write_failed) 96 ENDPROC(call_rwsem_down_write_failed)
@@ -105,12 +99,10 @@ ENTRY(call_rwsem_wake)
105 CFI_STARTPROC 99 CFI_STARTPROC
106 decw %dx /* do nothing if still outstanding active readers */ 100 decw %dx /* do nothing if still outstanding active readers */
107 jnz 1f 101 jnz 1f
108 push %ecx 102 pushl_cfi %ecx
109 CFI_ADJUST_CFA_OFFSET 4
110 CFI_REL_OFFSET ecx,0 103 CFI_REL_OFFSET ecx,0
111 call rwsem_wake 104 call rwsem_wake
112 pop %ecx 105 popl_cfi %ecx
113 CFI_ADJUST_CFA_OFFSET -4
1141: ret 1061: ret
115 CFI_ENDPROC 107 CFI_ENDPROC
116 ENDPROC(call_rwsem_wake) 108 ENDPROC(call_rwsem_wake)
@@ -118,17 +110,13 @@ ENTRY(call_rwsem_wake)
118/* Fix up special calling conventions */ 110/* Fix up special calling conventions */
119ENTRY(call_rwsem_downgrade_wake) 111ENTRY(call_rwsem_downgrade_wake)
120 CFI_STARTPROC 112 CFI_STARTPROC
121 push %ecx 113 pushl_cfi %ecx
122 CFI_ADJUST_CFA_OFFSET 4
123 CFI_REL_OFFSET ecx,0 114 CFI_REL_OFFSET ecx,0
124 push %edx 115 pushl_cfi %edx
125 CFI_ADJUST_CFA_OFFSET 4
126 CFI_REL_OFFSET edx,0 116 CFI_REL_OFFSET edx,0
127 call rwsem_downgrade_wake 117 call rwsem_downgrade_wake
128 pop %edx 118 popl_cfi %edx
129 CFI_ADJUST_CFA_OFFSET -4 119 popl_cfi %ecx
130 pop %ecx
131 CFI_ADJUST_CFA_OFFSET -4
132 ret 120 ret
133 CFI_ENDPROC 121 CFI_ENDPROC
134 ENDPROC(call_rwsem_downgrade_wake) 122 ENDPROC(call_rwsem_downgrade_wake)
diff --git a/arch/x86/lib/thunk_32.S b/arch/x86/lib/thunk_32.S
index 650b11e00ecc..2930ae05d773 100644
--- a/arch/x86/lib/thunk_32.S
+++ b/arch/x86/lib/thunk_32.S
@@ -7,24 +7,6 @@
7 7
8 #include <linux/linkage.h> 8 #include <linux/linkage.h>
9 9
10#define ARCH_TRACE_IRQS_ON \
11 pushl %eax; \
12 pushl %ecx; \
13 pushl %edx; \
14 call trace_hardirqs_on; \
15 popl %edx; \
16 popl %ecx; \
17 popl %eax;
18
19#define ARCH_TRACE_IRQS_OFF \
20 pushl %eax; \
21 pushl %ecx; \
22 pushl %edx; \
23 call trace_hardirqs_off; \
24 popl %edx; \
25 popl %ecx; \
26 popl %eax;
27
28#ifdef CONFIG_TRACE_IRQFLAGS 10#ifdef CONFIG_TRACE_IRQFLAGS
29 /* put return address in eax (arg1) */ 11 /* put return address in eax (arg1) */
30 .macro thunk_ra name,func 12 .macro thunk_ra name,func
diff --git a/arch/x86/lib/thunk_64.S b/arch/x86/lib/thunk_64.S
index bf9a7d5a5428..782b082c9ff7 100644
--- a/arch/x86/lib/thunk_64.S
+++ b/arch/x86/lib/thunk_64.S
@@ -22,26 +22,6 @@
22 CFI_ENDPROC 22 CFI_ENDPROC
23 .endm 23 .endm
24 24
25 /* rdi: arg1 ... normal C conventions. rax is passed from C. */
26 .macro thunk_retrax name,func
27 .globl \name
28\name:
29 CFI_STARTPROC
30 SAVE_ARGS
31 call \func
32 jmp restore_norax
33 CFI_ENDPROC
34 .endm
35
36
37 .section .sched.text, "ax"
38#ifdef CONFIG_RWSEM_XCHGADD_ALGORITHM
39 thunk rwsem_down_read_failed_thunk,rwsem_down_read_failed
40 thunk rwsem_down_write_failed_thunk,rwsem_down_write_failed
41 thunk rwsem_wake_thunk,rwsem_wake
42 thunk rwsem_downgrade_thunk,rwsem_downgrade_wake
43#endif
44
45#ifdef CONFIG_TRACE_IRQFLAGS 25#ifdef CONFIG_TRACE_IRQFLAGS
46 /* put return address in rdi (arg1) */ 26 /* put return address in rdi (arg1) */
47 .macro thunk_ra name,func 27 .macro thunk_ra name,func
@@ -72,10 +52,3 @@ restore:
72 RESTORE_ARGS 52 RESTORE_ARGS
73 ret 53 ret
74 CFI_ENDPROC 54 CFI_ENDPROC
75
76 CFI_STARTPROC
77 SAVE_ARGS
78restore_norax:
79 RESTORE_ARGS 1
80 ret
81 CFI_ENDPROC
diff --git a/arch/x86/mm/Makefile b/arch/x86/mm/Makefile
index 09df2f9a3d69..3e608edf9958 100644
--- a/arch/x86/mm/Makefile
+++ b/arch/x86/mm/Makefile
@@ -25,6 +25,7 @@ obj-$(CONFIG_MMIOTRACE_TEST) += testmmiotrace.o
25obj-$(CONFIG_NUMA) += numa.o numa_$(BITS).o 25obj-$(CONFIG_NUMA) += numa.o numa_$(BITS).o
26obj-$(CONFIG_AMD_NUMA) += amdtopology_64.o 26obj-$(CONFIG_AMD_NUMA) += amdtopology_64.o
27obj-$(CONFIG_ACPI_NUMA) += srat_$(BITS).o 27obj-$(CONFIG_ACPI_NUMA) += srat_$(BITS).o
28obj-$(CONFIG_NUMA_EMU) += numa_emulation.o
28 29
29obj-$(CONFIG_HAVE_MEMBLOCK) += memblock.o 30obj-$(CONFIG_HAVE_MEMBLOCK) += memblock.o
30 31
diff --git a/arch/x86/mm/amdtopology_64.c b/arch/x86/mm/amdtopology_64.c
index f21962c435ed..0919c26820d4 100644
--- a/arch/x86/mm/amdtopology_64.c
+++ b/arch/x86/mm/amdtopology_64.c
@@ -26,9 +26,7 @@
26#include <asm/apic.h> 26#include <asm/apic.h>
27#include <asm/amd_nb.h> 27#include <asm/amd_nb.h>
28 28
29static struct bootnode __initdata nodes[8];
30static unsigned char __initdata nodeids[8]; 29static unsigned char __initdata nodeids[8];
31static nodemask_t __initdata nodes_parsed = NODE_MASK_NONE;
32 30
33static __init int find_northbridge(void) 31static __init int find_northbridge(void)
34{ 32{
@@ -51,7 +49,7 @@ static __init int find_northbridge(void)
51 return num; 49 return num;
52 } 50 }
53 51
54 return -1; 52 return -ENOENT;
55} 53}
56 54
57static __init void early_get_boot_cpu_id(void) 55static __init void early_get_boot_cpu_id(void)
@@ -69,17 +67,18 @@ static __init void early_get_boot_cpu_id(void)
69#endif 67#endif
70} 68}
71 69
72int __init amd_numa_init(unsigned long start_pfn, unsigned long end_pfn) 70int __init amd_numa_init(void)
73{ 71{
74 unsigned long start = PFN_PHYS(start_pfn); 72 unsigned long start = PFN_PHYS(0);
75 unsigned long end = PFN_PHYS(end_pfn); 73 unsigned long end = PFN_PHYS(max_pfn);
76 unsigned numnodes; 74 unsigned numnodes;
77 unsigned long prevbase; 75 unsigned long prevbase;
78 int i, nb, found = 0; 76 int i, j, nb;
79 u32 nodeid, reg; 77 u32 nodeid, reg;
78 unsigned int bits, cores, apicid_base;
80 79
81 if (!early_pci_allowed()) 80 if (!early_pci_allowed())
82 return -1; 81 return -EINVAL;
83 82
84 nb = find_northbridge(); 83 nb = find_northbridge();
85 if (nb < 0) 84 if (nb < 0)
@@ -90,7 +89,7 @@ int __init amd_numa_init(unsigned long start_pfn, unsigned long end_pfn)
90 reg = read_pci_config(0, nb, 0, 0x60); 89 reg = read_pci_config(0, nb, 0, 0x60);
91 numnodes = ((reg >> 4) & 0xF) + 1; 90 numnodes = ((reg >> 4) & 0xF) + 1;
92 if (numnodes <= 1) 91 if (numnodes <= 1)
93 return -1; 92 return -ENOENT;
94 93
95 pr_info("Number of physical nodes %d\n", numnodes); 94 pr_info("Number of physical nodes %d\n", numnodes);
96 95
@@ -121,9 +120,9 @@ int __init amd_numa_init(unsigned long start_pfn, unsigned long end_pfn)
121 if ((base >> 8) & 3 || (limit >> 8) & 3) { 120 if ((base >> 8) & 3 || (limit >> 8) & 3) {
122 pr_err("Node %d using interleaving mode %lx/%lx\n", 121 pr_err("Node %d using interleaving mode %lx/%lx\n",
123 nodeid, (base >> 8) & 3, (limit >> 8) & 3); 122 nodeid, (base >> 8) & 3, (limit >> 8) & 3);
124 return -1; 123 return -EINVAL;
125 } 124 }
126 if (node_isset(nodeid, nodes_parsed)) { 125 if (node_isset(nodeid, numa_nodes_parsed)) {
127 pr_info("Node %d already present, skipping\n", 126 pr_info("Node %d already present, skipping\n",
128 nodeid); 127 nodeid);
129 continue; 128 continue;
@@ -160,117 +159,28 @@ int __init amd_numa_init(unsigned long start_pfn, unsigned long end_pfn)
160 if (prevbase > base) { 159 if (prevbase > base) {
161 pr_err("Node map not sorted %lx,%lx\n", 160 pr_err("Node map not sorted %lx,%lx\n",
162 prevbase, base); 161 prevbase, base);
163 return -1; 162 return -EINVAL;
164 } 163 }
165 164
166 pr_info("Node %d MemBase %016lx Limit %016lx\n", 165 pr_info("Node %d MemBase %016lx Limit %016lx\n",
167 nodeid, base, limit); 166 nodeid, base, limit);
168 167
169 found++;
170
171 nodes[nodeid].start = base;
172 nodes[nodeid].end = limit;
173
174 prevbase = base; 168 prevbase = base;
175 169 numa_add_memblk(nodeid, base, limit);
176 node_set(nodeid, nodes_parsed); 170 node_set(nodeid, numa_nodes_parsed);
177 }
178
179 if (!found)
180 return -1;
181 return 0;
182}
183
184#ifdef CONFIG_NUMA_EMU
185static s16 fake_apicid_to_node[MAX_LOCAL_APIC] __initdata = {
186 [0 ... MAX_LOCAL_APIC-1] = NUMA_NO_NODE
187};
188
189void __init amd_get_nodes(struct bootnode *physnodes)
190{
191 int i;
192
193 for_each_node_mask(i, nodes_parsed) {
194 physnodes[i].start = nodes[i].start;
195 physnodes[i].end = nodes[i].end;
196 } 171 }
197}
198
199static int __init find_node_by_addr(unsigned long addr)
200{
201 int ret = NUMA_NO_NODE;
202 int i;
203
204 for (i = 0; i < 8; i++)
205 if (addr >= nodes[i].start && addr < nodes[i].end) {
206 ret = i;
207 break;
208 }
209 return ret;
210}
211 172
212/* 173 if (!nodes_weight(numa_nodes_parsed))
213 * For NUMA emulation, fake proximity domain (_PXM) to node id mappings must be 174 return -ENOENT;
214 * setup to represent the physical topology but reflect the emulated
215 * environment. For each emulated node, the real node which it appears on is
216 * found and a fake pxm to nid mapping is created which mirrors the actual
217 * locality. node_distance() then represents the correct distances between
218 * emulated nodes by using the fake acpi mappings to pxms.
219 */
220void __init amd_fake_nodes(const struct bootnode *nodes, int nr_nodes)
221{
222 unsigned int bits;
223 unsigned int cores;
224 unsigned int apicid_base = 0;
225 int i;
226 175
176 /*
177 * We seem to have valid NUMA configuration. Map apicids to nodes
178 * using the coreid bits from early_identify_cpu.
179 */
227 bits = boot_cpu_data.x86_coreid_bits; 180 bits = boot_cpu_data.x86_coreid_bits;
228 cores = 1 << bits; 181 cores = 1 << bits;
229 early_get_boot_cpu_id();
230 if (boot_cpu_physical_apicid > 0)
231 apicid_base = boot_cpu_physical_apicid;
232
233 for (i = 0; i < nr_nodes; i++) {
234 int index;
235 int nid;
236 int j;
237
238 nid = find_node_by_addr(nodes[i].start);
239 if (nid == NUMA_NO_NODE)
240 continue;
241
242 index = nodeids[nid] << bits;
243 if (fake_apicid_to_node[index + apicid_base] == NUMA_NO_NODE)
244 for (j = apicid_base; j < cores + apicid_base; j++)
245 fake_apicid_to_node[index + j] = i;
246#ifdef CONFIG_ACPI_NUMA
247 __acpi_map_pxm_to_node(nid, i);
248#endif
249 }
250 memcpy(apicid_to_node, fake_apicid_to_node, sizeof(apicid_to_node));
251}
252#endif /* CONFIG_NUMA_EMU */
253
254int __init amd_scan_nodes(void)
255{
256 unsigned int bits;
257 unsigned int cores;
258 unsigned int apicid_base;
259 int i;
260
261 BUG_ON(nodes_empty(nodes_parsed));
262 node_possible_map = nodes_parsed;
263 memnode_shift = compute_hash_shift(nodes, 8, NULL);
264 if (memnode_shift < 0) {
265 pr_err("No NUMA node hash function found. Contact maintainer\n");
266 return -1;
267 }
268 pr_info("Using node hash shift of %d\n", memnode_shift);
269
270 /* use the coreid bits from early_identify_cpu */
271 bits = boot_cpu_data.x86_coreid_bits;
272 cores = (1<<bits);
273 apicid_base = 0; 182 apicid_base = 0;
183
274 /* get the APIC ID of the BSP early for systems with apicid lifting */ 184 /* get the APIC ID of the BSP early for systems with apicid lifting */
275 early_get_boot_cpu_id(); 185 early_get_boot_cpu_id();
276 if (boot_cpu_physical_apicid > 0) { 186 if (boot_cpu_physical_apicid > 0) {
@@ -278,17 +188,9 @@ int __init amd_scan_nodes(void)
278 apicid_base = boot_cpu_physical_apicid; 188 apicid_base = boot_cpu_physical_apicid;
279 } 189 }
280 190
281 for_each_node_mask(i, node_possible_map) { 191 for_each_node_mask(i, numa_nodes_parsed)
282 int j;
283
284 memblock_x86_register_active_regions(i,
285 nodes[i].start >> PAGE_SHIFT,
286 nodes[i].end >> PAGE_SHIFT);
287 for (j = apicid_base; j < cores + apicid_base; j++) 192 for (j = apicid_base; j < cores + apicid_base; j++)
288 apicid_to_node[(i << bits) + j] = i; 193 set_apicid_to_node((i << bits) + j, i);
289 setup_node_bootmem(i, nodes[i].start, nodes[i].end);
290 }
291 194
292 numa_init_array();
293 return 0; 195 return 0;
294} 196}
diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c
index 947f42abe820..286d289b039b 100644
--- a/arch/x86/mm/init.c
+++ b/arch/x86/mm/init.c
@@ -18,9 +18,9 @@
18 18
19DEFINE_PER_CPU(struct mmu_gather, mmu_gathers); 19DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
20 20
21unsigned long __initdata e820_table_start; 21unsigned long __initdata pgt_buf_start;
22unsigned long __meminitdata e820_table_end; 22unsigned long __meminitdata pgt_buf_end;
23unsigned long __meminitdata e820_table_top; 23unsigned long __meminitdata pgt_buf_top;
24 24
25int after_bootmem; 25int after_bootmem;
26 26
@@ -33,7 +33,7 @@ int direct_gbpages
33static void __init find_early_table_space(unsigned long end, int use_pse, 33static void __init find_early_table_space(unsigned long end, int use_pse,
34 int use_gbpages) 34 int use_gbpages)
35{ 35{
36 unsigned long puds, pmds, ptes, tables, start; 36 unsigned long puds, pmds, ptes, tables, start = 0, good_end = end;
37 phys_addr_t base; 37 phys_addr_t base;
38 38
39 puds = (end + PUD_SIZE - 1) >> PUD_SHIFT; 39 puds = (end + PUD_SIZE - 1) >> PUD_SHIFT;
@@ -65,29 +65,20 @@ static void __init find_early_table_space(unsigned long end, int use_pse,
65#ifdef CONFIG_X86_32 65#ifdef CONFIG_X86_32
66 /* for fixmap */ 66 /* for fixmap */
67 tables += roundup(__end_of_fixed_addresses * sizeof(pte_t), PAGE_SIZE); 67 tables += roundup(__end_of_fixed_addresses * sizeof(pte_t), PAGE_SIZE);
68#endif
69 68
70 /* 69 good_end = max_pfn_mapped << PAGE_SHIFT;
71 * RED-PEN putting page tables only on node 0 could
72 * cause a hotspot and fill up ZONE_DMA. The page tables
73 * need roughly 0.5KB per GB.
74 */
75#ifdef CONFIG_X86_32
76 start = 0x7000;
77#else
78 start = 0x8000;
79#endif 70#endif
80 base = memblock_find_in_range(start, max_pfn_mapped<<PAGE_SHIFT, 71
81 tables, PAGE_SIZE); 72 base = memblock_find_in_range(start, good_end, tables, PAGE_SIZE);
82 if (base == MEMBLOCK_ERROR) 73 if (base == MEMBLOCK_ERROR)
83 panic("Cannot find space for the kernel page tables"); 74 panic("Cannot find space for the kernel page tables");
84 75
85 e820_table_start = base >> PAGE_SHIFT; 76 pgt_buf_start = base >> PAGE_SHIFT;
86 e820_table_end = e820_table_start; 77 pgt_buf_end = pgt_buf_start;
87 e820_table_top = e820_table_start + (tables >> PAGE_SHIFT); 78 pgt_buf_top = pgt_buf_start + (tables >> PAGE_SHIFT);
88 79
89 printk(KERN_DEBUG "kernel direct mapping tables up to %lx @ %lx-%lx\n", 80 printk(KERN_DEBUG "kernel direct mapping tables up to %lx @ %lx-%lx\n",
90 end, e820_table_start << PAGE_SHIFT, e820_table_top << PAGE_SHIFT); 81 end, pgt_buf_start << PAGE_SHIFT, pgt_buf_top << PAGE_SHIFT);
91} 82}
92 83
93struct map_range { 84struct map_range {
@@ -279,30 +270,11 @@ unsigned long __init_refok init_memory_mapping(unsigned long start,
279 load_cr3(swapper_pg_dir); 270 load_cr3(swapper_pg_dir);
280#endif 271#endif
281 272
282#ifdef CONFIG_X86_64
283 if (!after_bootmem && !start) {
284 pud_t *pud;
285 pmd_t *pmd;
286
287 mmu_cr4_features = read_cr4();
288
289 /*
290 * _brk_end cannot change anymore, but it and _end may be
291 * located on different 2M pages. cleanup_highmap(), however,
292 * can only consider _end when it runs, so destroy any
293 * mappings beyond _brk_end here.
294 */
295 pud = pud_offset(pgd_offset_k(_brk_end), _brk_end);
296 pmd = pmd_offset(pud, _brk_end - 1);
297 while (++pmd <= pmd_offset(pud, (unsigned long)_end - 1))
298 pmd_clear(pmd);
299 }
300#endif
301 __flush_tlb_all(); 273 __flush_tlb_all();
302 274
303 if (!after_bootmem && e820_table_end > e820_table_start) 275 if (!after_bootmem && pgt_buf_end > pgt_buf_start)
304 memblock_x86_reserve_range(e820_table_start << PAGE_SHIFT, 276 memblock_x86_reserve_range(pgt_buf_start << PAGE_SHIFT,
305 e820_table_end << PAGE_SHIFT, "PGTABLE"); 277 pgt_buf_end << PAGE_SHIFT, "PGTABLE");
306 278
307 if (!after_bootmem) 279 if (!after_bootmem)
308 early_memtest(start, end); 280 early_memtest(start, end);
diff --git a/arch/x86/mm/init_32.c b/arch/x86/mm/init_32.c
index c821074b7f0b..73ad7ebd6e9c 100644
--- a/arch/x86/mm/init_32.c
+++ b/arch/x86/mm/init_32.c
@@ -62,10 +62,10 @@ bool __read_mostly __vmalloc_start_set = false;
62 62
63static __init void *alloc_low_page(void) 63static __init void *alloc_low_page(void)
64{ 64{
65 unsigned long pfn = e820_table_end++; 65 unsigned long pfn = pgt_buf_end++;
66 void *adr; 66 void *adr;
67 67
68 if (pfn >= e820_table_top) 68 if (pfn >= pgt_buf_top)
69 panic("alloc_low_page: ran out of memory"); 69 panic("alloc_low_page: ran out of memory");
70 70
71 adr = __va(pfn * PAGE_SIZE); 71 adr = __va(pfn * PAGE_SIZE);
@@ -163,8 +163,8 @@ static pte_t *__init page_table_kmap_check(pte_t *pte, pmd_t *pmd,
163 if (pmd_idx_kmap_begin != pmd_idx_kmap_end 163 if (pmd_idx_kmap_begin != pmd_idx_kmap_end
164 && (vaddr >> PMD_SHIFT) >= pmd_idx_kmap_begin 164 && (vaddr >> PMD_SHIFT) >= pmd_idx_kmap_begin
165 && (vaddr >> PMD_SHIFT) <= pmd_idx_kmap_end 165 && (vaddr >> PMD_SHIFT) <= pmd_idx_kmap_end
166 && ((__pa(pte) >> PAGE_SHIFT) < e820_table_start 166 && ((__pa(pte) >> PAGE_SHIFT) < pgt_buf_start
167 || (__pa(pte) >> PAGE_SHIFT) >= e820_table_end)) { 167 || (__pa(pte) >> PAGE_SHIFT) >= pgt_buf_end)) {
168 pte_t *newpte; 168 pte_t *newpte;
169 int i; 169 int i;
170 170
@@ -644,8 +644,7 @@ void __init find_low_pfn_range(void)
644} 644}
645 645
646#ifndef CONFIG_NEED_MULTIPLE_NODES 646#ifndef CONFIG_NEED_MULTIPLE_NODES
647void __init initmem_init(unsigned long start_pfn, unsigned long end_pfn, 647void __init initmem_init(void)
648 int acpi, int k8)
649{ 648{
650#ifdef CONFIG_HIGHMEM 649#ifdef CONFIG_HIGHMEM
651 highstart_pfn = highend_pfn = max_pfn; 650 highstart_pfn = highend_pfn = max_pfn;
diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c
index c14a5422e152..a08a62cb136e 100644
--- a/arch/x86/mm/init_64.c
+++ b/arch/x86/mm/init_64.c
@@ -314,7 +314,7 @@ void __init cleanup_highmap(void)
314 314
315static __ref void *alloc_low_page(unsigned long *phys) 315static __ref void *alloc_low_page(unsigned long *phys)
316{ 316{
317 unsigned long pfn = e820_table_end++; 317 unsigned long pfn = pgt_buf_end++;
318 void *adr; 318 void *adr;
319 319
320 if (after_bootmem) { 320 if (after_bootmem) {
@@ -324,7 +324,7 @@ static __ref void *alloc_low_page(unsigned long *phys)
324 return adr; 324 return adr;
325 } 325 }
326 326
327 if (pfn >= e820_table_top) 327 if (pfn >= pgt_buf_top)
328 panic("alloc_low_page: ran out of memory"); 328 panic("alloc_low_page: ran out of memory");
329 329
330 adr = early_memremap(pfn * PAGE_SIZE, PAGE_SIZE); 330 adr = early_memremap(pfn * PAGE_SIZE, PAGE_SIZE);
@@ -333,12 +333,28 @@ static __ref void *alloc_low_page(unsigned long *phys)
333 return adr; 333 return adr;
334} 334}
335 335
336static __ref void *map_low_page(void *virt)
337{
338 void *adr;
339 unsigned long phys, left;
340
341 if (after_bootmem)
342 return virt;
343
344 phys = __pa(virt);
345 left = phys & (PAGE_SIZE - 1);
346 adr = early_memremap(phys & PAGE_MASK, PAGE_SIZE);
347 adr = (void *)(((unsigned long)adr) | left);
348
349 return adr;
350}
351
336static __ref void unmap_low_page(void *adr) 352static __ref void unmap_low_page(void *adr)
337{ 353{
338 if (after_bootmem) 354 if (after_bootmem)
339 return; 355 return;
340 356
341 early_iounmap(adr, PAGE_SIZE); 357 early_iounmap((void *)((unsigned long)adr & PAGE_MASK), PAGE_SIZE);
342} 358}
343 359
344static unsigned long __meminit 360static unsigned long __meminit
@@ -386,15 +402,6 @@ phys_pte_init(pte_t *pte_page, unsigned long addr, unsigned long end,
386} 402}
387 403
388static unsigned long __meminit 404static unsigned long __meminit
389phys_pte_update(pmd_t *pmd, unsigned long address, unsigned long end,
390 pgprot_t prot)
391{
392 pte_t *pte = (pte_t *)pmd_page_vaddr(*pmd);
393
394 return phys_pte_init(pte, address, end, prot);
395}
396
397static unsigned long __meminit
398phys_pmd_init(pmd_t *pmd_page, unsigned long address, unsigned long end, 405phys_pmd_init(pmd_t *pmd_page, unsigned long address, unsigned long end,
399 unsigned long page_size_mask, pgprot_t prot) 406 unsigned long page_size_mask, pgprot_t prot)
400{ 407{
@@ -420,8 +427,10 @@ phys_pmd_init(pmd_t *pmd_page, unsigned long address, unsigned long end,
420 if (pmd_val(*pmd)) { 427 if (pmd_val(*pmd)) {
421 if (!pmd_large(*pmd)) { 428 if (!pmd_large(*pmd)) {
422 spin_lock(&init_mm.page_table_lock); 429 spin_lock(&init_mm.page_table_lock);
423 last_map_addr = phys_pte_update(pmd, address, 430 pte = map_low_page((pte_t *)pmd_page_vaddr(*pmd));
431 last_map_addr = phys_pte_init(pte, address,
424 end, prot); 432 end, prot);
433 unmap_low_page(pte);
425 spin_unlock(&init_mm.page_table_lock); 434 spin_unlock(&init_mm.page_table_lock);
426 continue; 435 continue;
427 } 436 }
@@ -468,18 +477,6 @@ phys_pmd_init(pmd_t *pmd_page, unsigned long address, unsigned long end,
468} 477}
469 478
470static unsigned long __meminit 479static unsigned long __meminit
471phys_pmd_update(pud_t *pud, unsigned long address, unsigned long end,
472 unsigned long page_size_mask, pgprot_t prot)
473{
474 pmd_t *pmd = pmd_offset(pud, 0);
475 unsigned long last_map_addr;
476
477 last_map_addr = phys_pmd_init(pmd, address, end, page_size_mask, prot);
478 __flush_tlb_all();
479 return last_map_addr;
480}
481
482static unsigned long __meminit
483phys_pud_init(pud_t *pud_page, unsigned long addr, unsigned long end, 480phys_pud_init(pud_t *pud_page, unsigned long addr, unsigned long end,
484 unsigned long page_size_mask) 481 unsigned long page_size_mask)
485{ 482{
@@ -504,8 +501,11 @@ phys_pud_init(pud_t *pud_page, unsigned long addr, unsigned long end,
504 501
505 if (pud_val(*pud)) { 502 if (pud_val(*pud)) {
506 if (!pud_large(*pud)) { 503 if (!pud_large(*pud)) {
507 last_map_addr = phys_pmd_update(pud, addr, end, 504 pmd = map_low_page(pmd_offset(pud, 0));
505 last_map_addr = phys_pmd_init(pmd, addr, end,
508 page_size_mask, prot); 506 page_size_mask, prot);
507 unmap_low_page(pmd);
508 __flush_tlb_all();
509 continue; 509 continue;
510 } 510 }
511 /* 511 /*
@@ -553,17 +553,6 @@ phys_pud_init(pud_t *pud_page, unsigned long addr, unsigned long end,
553 return last_map_addr; 553 return last_map_addr;
554} 554}
555 555
556static unsigned long __meminit
557phys_pud_update(pgd_t *pgd, unsigned long addr, unsigned long end,
558 unsigned long page_size_mask)
559{
560 pud_t *pud;
561
562 pud = (pud_t *)pgd_page_vaddr(*pgd);
563
564 return phys_pud_init(pud, addr, end, page_size_mask);
565}
566
567unsigned long __meminit 556unsigned long __meminit
568kernel_physical_mapping_init(unsigned long start, 557kernel_physical_mapping_init(unsigned long start,
569 unsigned long end, 558 unsigned long end,
@@ -587,8 +576,10 @@ kernel_physical_mapping_init(unsigned long start,
587 next = end; 576 next = end;
588 577
589 if (pgd_val(*pgd)) { 578 if (pgd_val(*pgd)) {
590 last_map_addr = phys_pud_update(pgd, __pa(start), 579 pud = map_low_page((pud_t *)pgd_page_vaddr(*pgd));
580 last_map_addr = phys_pud_init(pud, __pa(start),
591 __pa(end), page_size_mask); 581 __pa(end), page_size_mask);
582 unmap_low_page(pud);
592 continue; 583 continue;
593 } 584 }
594 585
@@ -612,10 +603,9 @@ kernel_physical_mapping_init(unsigned long start,
612} 603}
613 604
614#ifndef CONFIG_NUMA 605#ifndef CONFIG_NUMA
615void __init initmem_init(unsigned long start_pfn, unsigned long end_pfn, 606void __init initmem_init(void)
616 int acpi, int k8)
617{ 607{
618 memblock_x86_register_active_regions(0, start_pfn, end_pfn); 608 memblock_x86_register_active_regions(0, 0, max_pfn);
619} 609}
620#endif 610#endif
621 611
diff --git a/arch/x86/mm/numa.c b/arch/x86/mm/numa.c
index ebf6d7887a38..9559d360fde7 100644
--- a/arch/x86/mm/numa.c
+++ b/arch/x86/mm/numa.c
@@ -26,12 +26,50 @@ static __init int numa_setup(char *opt)
26early_param("numa", numa_setup); 26early_param("numa", numa_setup);
27 27
28/* 28/*
29 * Which logical CPUs are on which nodes 29 * apicid, cpu, node mappings
30 */ 30 */
31s16 __apicid_to_node[MAX_LOCAL_APIC] __cpuinitdata = {
32 [0 ... MAX_LOCAL_APIC-1] = NUMA_NO_NODE
33};
34
31cpumask_var_t node_to_cpumask_map[MAX_NUMNODES]; 35cpumask_var_t node_to_cpumask_map[MAX_NUMNODES];
32EXPORT_SYMBOL(node_to_cpumask_map); 36EXPORT_SYMBOL(node_to_cpumask_map);
33 37
34/* 38/*
39 * Map cpu index to node index
40 */
41DEFINE_EARLY_PER_CPU(int, x86_cpu_to_node_map, NUMA_NO_NODE);
42EXPORT_EARLY_PER_CPU_SYMBOL(x86_cpu_to_node_map);
43
44void __cpuinit numa_set_node(int cpu, int node)
45{
46 int *cpu_to_node_map = early_per_cpu_ptr(x86_cpu_to_node_map);
47
48 /* early setting, no percpu area yet */
49 if (cpu_to_node_map) {
50 cpu_to_node_map[cpu] = node;
51 return;
52 }
53
54#ifdef CONFIG_DEBUG_PER_CPU_MAPS
55 if (cpu >= nr_cpu_ids || !cpu_possible(cpu)) {
56 printk(KERN_ERR "numa_set_node: invalid cpu# (%d)\n", cpu);
57 dump_stack();
58 return;
59 }
60#endif
61 per_cpu(x86_cpu_to_node_map, cpu) = node;
62
63 if (node != NUMA_NO_NODE)
64 set_cpu_numa_node(cpu, node);
65}
66
67void __cpuinit numa_clear_node(int cpu)
68{
69 numa_set_node(cpu, NUMA_NO_NODE);
70}
71
72/*
35 * Allocate node_to_cpumask_map based on number of available nodes 73 * Allocate node_to_cpumask_map based on number of available nodes
36 * Requires node_possible_map to be valid. 74 * Requires node_possible_map to be valid.
37 * 75 *
@@ -57,7 +95,174 @@ void __init setup_node_to_cpumask_map(void)
57 pr_debug("Node to cpumask map for %d nodes\n", nr_node_ids); 95 pr_debug("Node to cpumask map for %d nodes\n", nr_node_ids);
58} 96}
59 97
60#ifdef CONFIG_DEBUG_PER_CPU_MAPS 98/*
99 * There are unfortunately some poorly designed mainboards around that
100 * only connect memory to a single CPU. This breaks the 1:1 cpu->node
101 * mapping. To avoid this fill in the mapping for all possible CPUs,
102 * as the number of CPUs is not known yet. We round robin the existing
103 * nodes.
104 */
105void __init numa_init_array(void)
106{
107 int rr, i;
108
109 rr = first_node(node_online_map);
110 for (i = 0; i < nr_cpu_ids; i++) {
111 if (early_cpu_to_node(i) != NUMA_NO_NODE)
112 continue;
113 numa_set_node(i, rr);
114 rr = next_node(rr, node_online_map);
115 if (rr == MAX_NUMNODES)
116 rr = first_node(node_online_map);
117 }
118}
119
120static __init int find_near_online_node(int node)
121{
122 int n, val;
123 int min_val = INT_MAX;
124 int best_node = -1;
125
126 for_each_online_node(n) {
127 val = node_distance(node, n);
128
129 if (val < min_val) {
130 min_val = val;
131 best_node = n;
132 }
133 }
134
135 return best_node;
136}
137
138/*
139 * Setup early cpu_to_node.
140 *
141 * Populate cpu_to_node[] only if x86_cpu_to_apicid[],
142 * and apicid_to_node[] tables have valid entries for a CPU.
143 * This means we skip cpu_to_node[] initialisation for NUMA
144 * emulation and faking node case (when running a kernel compiled
145 * for NUMA on a non NUMA box), which is OK as cpu_to_node[]
146 * is already initialized in a round robin manner at numa_init_array,
147 * prior to this call, and this initialization is good enough
148 * for the fake NUMA cases.
149 *
150 * Called before the per_cpu areas are setup.
151 */
152void __init init_cpu_to_node(void)
153{
154 int cpu;
155 u16 *cpu_to_apicid = early_per_cpu_ptr(x86_cpu_to_apicid);
156
157 BUG_ON(cpu_to_apicid == NULL);
158
159 for_each_possible_cpu(cpu) {
160 int node = numa_cpu_node(cpu);
161
162 if (node == NUMA_NO_NODE)
163 continue;
164 if (!node_online(node))
165 node = find_near_online_node(node);
166 numa_set_node(cpu, node);
167 }
168}
169
170#ifndef CONFIG_DEBUG_PER_CPU_MAPS
171
172# ifndef CONFIG_NUMA_EMU
173void __cpuinit numa_add_cpu(int cpu)
174{
175 cpumask_set_cpu(cpu, node_to_cpumask_map[early_cpu_to_node(cpu)]);
176}
177
178void __cpuinit numa_remove_cpu(int cpu)
179{
180 cpumask_clear_cpu(cpu, node_to_cpumask_map[early_cpu_to_node(cpu)]);
181}
182# endif /* !CONFIG_NUMA_EMU */
183
184#else /* !CONFIG_DEBUG_PER_CPU_MAPS */
185
186int __cpu_to_node(int cpu)
187{
188 if (early_per_cpu_ptr(x86_cpu_to_node_map)) {
189 printk(KERN_WARNING
190 "cpu_to_node(%d): usage too early!\n", cpu);
191 dump_stack();
192 return early_per_cpu_ptr(x86_cpu_to_node_map)[cpu];
193 }
194 return per_cpu(x86_cpu_to_node_map, cpu);
195}
196EXPORT_SYMBOL(__cpu_to_node);
197
198/*
199 * Same function as cpu_to_node() but used if called before the
200 * per_cpu areas are setup.
201 */
202int early_cpu_to_node(int cpu)
203{
204 if (early_per_cpu_ptr(x86_cpu_to_node_map))
205 return early_per_cpu_ptr(x86_cpu_to_node_map)[cpu];
206
207 if (!cpu_possible(cpu)) {
208 printk(KERN_WARNING
209 "early_cpu_to_node(%d): no per_cpu area!\n", cpu);
210 dump_stack();
211 return NUMA_NO_NODE;
212 }
213 return per_cpu(x86_cpu_to_node_map, cpu);
214}
215
216struct cpumask __cpuinit *debug_cpumask_set_cpu(int cpu, int enable)
217{
218 int node = early_cpu_to_node(cpu);
219 struct cpumask *mask;
220 char buf[64];
221
222 if (node == NUMA_NO_NODE) {
223 /* early_cpu_to_node() already emits a warning and trace */
224 return NULL;
225 }
226 mask = node_to_cpumask_map[node];
227 if (!mask) {
228 pr_err("node_to_cpumask_map[%i] NULL\n", node);
229 dump_stack();
230 return NULL;
231 }
232
233 cpulist_scnprintf(buf, sizeof(buf), mask);
234 printk(KERN_DEBUG "%s cpu %d node %d: mask now %s\n",
235 enable ? "numa_add_cpu" : "numa_remove_cpu",
236 cpu, node, buf);
237 return mask;
238}
239
240# ifndef CONFIG_NUMA_EMU
241static void __cpuinit numa_set_cpumask(int cpu, int enable)
242{
243 struct cpumask *mask;
244
245 mask = debug_cpumask_set_cpu(cpu, enable);
246 if (!mask)
247 return;
248
249 if (enable)
250 cpumask_set_cpu(cpu, mask);
251 else
252 cpumask_clear_cpu(cpu, mask);
253}
254
255void __cpuinit numa_add_cpu(int cpu)
256{
257 numa_set_cpumask(cpu, 1);
258}
259
260void __cpuinit numa_remove_cpu(int cpu)
261{
262 numa_set_cpumask(cpu, 0);
263}
264# endif /* !CONFIG_NUMA_EMU */
265
61/* 266/*
62 * Returns a pointer to the bitmask of CPUs on Node 'node'. 267 * Returns a pointer to the bitmask of CPUs on Node 'node'.
63 */ 268 */
@@ -80,4 +285,5 @@ const struct cpumask *cpumask_of_node(int node)
80 return node_to_cpumask_map[node]; 285 return node_to_cpumask_map[node];
81} 286}
82EXPORT_SYMBOL(cpumask_of_node); 287EXPORT_SYMBOL(cpumask_of_node);
83#endif 288
289#endif /* !CONFIG_DEBUG_PER_CPU_MAPS */
diff --git a/arch/x86/mm/numa_32.c b/arch/x86/mm/numa_32.c
index 84a3e4c9f277..bde3906420df 100644
--- a/arch/x86/mm/numa_32.c
+++ b/arch/x86/mm/numa_32.c
@@ -110,6 +110,12 @@ void set_pmd_pfn(unsigned long vaddr, unsigned long pfn, pgprot_t flags);
110 110
111static unsigned long kva_start_pfn; 111static unsigned long kva_start_pfn;
112static unsigned long kva_pages; 112static unsigned long kva_pages;
113
114int __cpuinit numa_cpu_node(int cpu)
115{
116 return apic->x86_32_numa_cpu_node(cpu);
117}
118
113/* 119/*
114 * FLAT - support for basic PC memory model with discontig enabled, essentially 120 * FLAT - support for basic PC memory model with discontig enabled, essentially
115 * a single node with all available processors in it with a flat 121 * a single node with all available processors in it with a flat
@@ -346,8 +352,7 @@ static void init_remap_allocator(int nid)
346 (ulong) node_remap_end_vaddr[nid]); 352 (ulong) node_remap_end_vaddr[nid]);
347} 353}
348 354
349void __init initmem_init(unsigned long start_pfn, unsigned long end_pfn, 355void __init initmem_init(void)
350 int acpi, int k8)
351{ 356{
352 int nid; 357 int nid;
353 long kva_target_pfn; 358 long kva_target_pfn;
@@ -361,6 +366,7 @@ void __init initmem_init(unsigned long start_pfn, unsigned long end_pfn,
361 */ 366 */
362 367
363 get_memcfg_numa(); 368 get_memcfg_numa();
369 numa_init_array();
364 370
365 kva_pages = roundup(calculate_numa_remap_pages(), PTRS_PER_PTE); 371 kva_pages = roundup(calculate_numa_remap_pages(), PTRS_PER_PTE);
366 372
diff --git a/arch/x86/mm/numa_64.c b/arch/x86/mm/numa_64.c
index 1337c51b07d7..9ec0f209a6a4 100644
--- a/arch/x86/mm/numa_64.c
+++ b/arch/x86/mm/numa_64.c
@@ -13,31 +13,30 @@
13#include <linux/module.h> 13#include <linux/module.h>
14#include <linux/nodemask.h> 14#include <linux/nodemask.h>
15#include <linux/sched.h> 15#include <linux/sched.h>
16#include <linux/acpi.h>
16 17
17#include <asm/e820.h> 18#include <asm/e820.h>
18#include <asm/proto.h> 19#include <asm/proto.h>
19#include <asm/dma.h> 20#include <asm/dma.h>
20#include <asm/numa.h>
21#include <asm/acpi.h> 21#include <asm/acpi.h>
22#include <asm/amd_nb.h> 22#include <asm/amd_nb.h>
23 23
24#include "numa_internal.h"
25
24struct pglist_data *node_data[MAX_NUMNODES] __read_mostly; 26struct pglist_data *node_data[MAX_NUMNODES] __read_mostly;
25EXPORT_SYMBOL(node_data); 27EXPORT_SYMBOL(node_data);
26 28
27struct memnode memnode; 29nodemask_t numa_nodes_parsed __initdata;
28 30
29s16 apicid_to_node[MAX_LOCAL_APIC] __cpuinitdata = { 31struct memnode memnode;
30 [0 ... MAX_LOCAL_APIC-1] = NUMA_NO_NODE
31};
32 32
33static unsigned long __initdata nodemap_addr; 33static unsigned long __initdata nodemap_addr;
34static unsigned long __initdata nodemap_size; 34static unsigned long __initdata nodemap_size;
35 35
36/* 36static struct numa_meminfo numa_meminfo __initdata;
37 * Map cpu index to node index 37
38 */ 38static int numa_distance_cnt;
39DEFINE_EARLY_PER_CPU(int, x86_cpu_to_node_map, NUMA_NO_NODE); 39static u8 *numa_distance;
40EXPORT_EARLY_PER_CPU_SYMBOL(x86_cpu_to_node_map);
41 40
42/* 41/*
43 * Given a shift value, try to populate memnodemap[] 42 * Given a shift value, try to populate memnodemap[]
@@ -46,16 +45,15 @@ EXPORT_EARLY_PER_CPU_SYMBOL(x86_cpu_to_node_map);
46 * 0 if memnodmap[] too small (of shift too small) 45 * 0 if memnodmap[] too small (of shift too small)
47 * -1 if node overlap or lost ram (shift too big) 46 * -1 if node overlap or lost ram (shift too big)
48 */ 47 */
49static int __init populate_memnodemap(const struct bootnode *nodes, 48static int __init populate_memnodemap(const struct numa_meminfo *mi, int shift)
50 int numnodes, int shift, int *nodeids)
51{ 49{
52 unsigned long addr, end; 50 unsigned long addr, end;
53 int i, res = -1; 51 int i, res = -1;
54 52
55 memset(memnodemap, 0xff, sizeof(s16)*memnodemapsize); 53 memset(memnodemap, 0xff, sizeof(s16)*memnodemapsize);
56 for (i = 0; i < numnodes; i++) { 54 for (i = 0; i < mi->nr_blks; i++) {
57 addr = nodes[i].start; 55 addr = mi->blk[i].start;
58 end = nodes[i].end; 56 end = mi->blk[i].end;
59 if (addr >= end) 57 if (addr >= end)
60 continue; 58 continue;
61 if ((end >> shift) >= memnodemapsize) 59 if ((end >> shift) >= memnodemapsize)
@@ -63,12 +61,7 @@ static int __init populate_memnodemap(const struct bootnode *nodes,
63 do { 61 do {
64 if (memnodemap[addr >> shift] != NUMA_NO_NODE) 62 if (memnodemap[addr >> shift] != NUMA_NO_NODE)
65 return -1; 63 return -1;
66 64 memnodemap[addr >> shift] = mi->blk[i].nid;
67 if (!nodeids)
68 memnodemap[addr >> shift] = i;
69 else
70 memnodemap[addr >> shift] = nodeids[i];
71
72 addr += (1UL << shift); 65 addr += (1UL << shift);
73 } while (addr < end); 66 } while (addr < end);
74 res = 1; 67 res = 1;
@@ -86,7 +79,7 @@ static int __init allocate_cachealigned_memnodemap(void)
86 79
87 addr = 0x8000; 80 addr = 0x8000;
88 nodemap_size = roundup(sizeof(s16) * memnodemapsize, L1_CACHE_BYTES); 81 nodemap_size = roundup(sizeof(s16) * memnodemapsize, L1_CACHE_BYTES);
89 nodemap_addr = memblock_find_in_range(addr, max_pfn<<PAGE_SHIFT, 82 nodemap_addr = memblock_find_in_range(addr, get_max_mapped(),
90 nodemap_size, L1_CACHE_BYTES); 83 nodemap_size, L1_CACHE_BYTES);
91 if (nodemap_addr == MEMBLOCK_ERROR) { 84 if (nodemap_addr == MEMBLOCK_ERROR) {
92 printk(KERN_ERR 85 printk(KERN_ERR
@@ -106,16 +99,15 @@ static int __init allocate_cachealigned_memnodemap(void)
106 * The LSB of all start and end addresses in the node map is the value of the 99 * The LSB of all start and end addresses in the node map is the value of the
107 * maximum possible shift. 100 * maximum possible shift.
108 */ 101 */
109static int __init extract_lsb_from_nodes(const struct bootnode *nodes, 102static int __init extract_lsb_from_nodes(const struct numa_meminfo *mi)
110 int numnodes)
111{ 103{
112 int i, nodes_used = 0; 104 int i, nodes_used = 0;
113 unsigned long start, end; 105 unsigned long start, end;
114 unsigned long bitfield = 0, memtop = 0; 106 unsigned long bitfield = 0, memtop = 0;
115 107
116 for (i = 0; i < numnodes; i++) { 108 for (i = 0; i < mi->nr_blks; i++) {
117 start = nodes[i].start; 109 start = mi->blk[i].start;
118 end = nodes[i].end; 110 end = mi->blk[i].end;
119 if (start >= end) 111 if (start >= end)
120 continue; 112 continue;
121 bitfield |= start; 113 bitfield |= start;
@@ -131,18 +123,17 @@ static int __init extract_lsb_from_nodes(const struct bootnode *nodes,
131 return i; 123 return i;
132} 124}
133 125
134int __init compute_hash_shift(struct bootnode *nodes, int numnodes, 126static int __init compute_hash_shift(const struct numa_meminfo *mi)
135 int *nodeids)
136{ 127{
137 int shift; 128 int shift;
138 129
139 shift = extract_lsb_from_nodes(nodes, numnodes); 130 shift = extract_lsb_from_nodes(mi);
140 if (allocate_cachealigned_memnodemap()) 131 if (allocate_cachealigned_memnodemap())
141 return -1; 132 return -1;
142 printk(KERN_DEBUG "NUMA: Using %d for the hash shift.\n", 133 printk(KERN_DEBUG "NUMA: Using %d for the hash shift.\n",
143 shift); 134 shift);
144 135
145 if (populate_memnodemap(nodes, numnodes, shift, nodeids) != 1) { 136 if (populate_memnodemap(mi, shift) != 1) {
146 printk(KERN_INFO "Your memory is not aligned you need to " 137 printk(KERN_INFO "Your memory is not aligned you need to "
147 "rebuild your kernel with a bigger NODEMAPSIZE " 138 "rebuild your kernel with a bigger NODEMAPSIZE "
148 "shift=%d\n", shift); 139 "shift=%d\n", shift);
@@ -188,6 +179,63 @@ static void * __init early_node_mem(int nodeid, unsigned long start,
188 return NULL; 179 return NULL;
189} 180}
190 181
182static int __init numa_add_memblk_to(int nid, u64 start, u64 end,
183 struct numa_meminfo *mi)
184{
185 /* ignore zero length blks */
186 if (start == end)
187 return 0;
188
189 /* whine about and ignore invalid blks */
190 if (start > end || nid < 0 || nid >= MAX_NUMNODES) {
191 pr_warning("NUMA: Warning: invalid memblk node %d (%Lx-%Lx)\n",
192 nid, start, end);
193 return 0;
194 }
195
196 if (mi->nr_blks >= NR_NODE_MEMBLKS) {
197 pr_err("NUMA: too many memblk ranges\n");
198 return -EINVAL;
199 }
200
201 mi->blk[mi->nr_blks].start = start;
202 mi->blk[mi->nr_blks].end = end;
203 mi->blk[mi->nr_blks].nid = nid;
204 mi->nr_blks++;
205 return 0;
206}
207
208/**
209 * numa_remove_memblk_from - Remove one numa_memblk from a numa_meminfo
210 * @idx: Index of memblk to remove
211 * @mi: numa_meminfo to remove memblk from
212 *
213 * Remove @idx'th numa_memblk from @mi by shifting @mi->blk[] and
214 * decrementing @mi->nr_blks.
215 */
216void __init numa_remove_memblk_from(int idx, struct numa_meminfo *mi)
217{
218 mi->nr_blks--;
219 memmove(&mi->blk[idx], &mi->blk[idx + 1],
220 (mi->nr_blks - idx) * sizeof(mi->blk[0]));
221}
222
223/**
224 * numa_add_memblk - Add one numa_memblk to numa_meminfo
225 * @nid: NUMA node ID of the new memblk
226 * @start: Start address of the new memblk
227 * @end: End address of the new memblk
228 *
229 * Add a new memblk to the default numa_meminfo.
230 *
231 * RETURNS:
232 * 0 on success, -errno on failure.
233 */
234int __init numa_add_memblk(int nid, u64 start, u64 end)
235{
236 return numa_add_memblk_to(nid, start, end, &numa_meminfo);
237}
238
191/* Initialize bootmem allocator for a node */ 239/* Initialize bootmem allocator for a node */
192void __init 240void __init
193setup_node_bootmem(int nodeid, unsigned long start, unsigned long end) 241setup_node_bootmem(int nodeid, unsigned long start, unsigned long end)
@@ -234,692 +282,386 @@ setup_node_bootmem(int nodeid, unsigned long start, unsigned long end)
234 node_set_online(nodeid); 282 node_set_online(nodeid);
235} 283}
236 284
237/* 285/**
238 * There are unfortunately some poorly designed mainboards around that 286 * numa_cleanup_meminfo - Cleanup a numa_meminfo
239 * only connect memory to a single CPU. This breaks the 1:1 cpu->node 287 * @mi: numa_meminfo to clean up
240 * mapping. To avoid this fill in the mapping for all possible CPUs, 288 *
241 * as the number of CPUs is not known yet. We round robin the existing 289 * Sanitize @mi by merging and removing unncessary memblks. Also check for
242 * nodes. 290 * conflicts and clear unused memblks.
291 *
292 * RETURNS:
293 * 0 on success, -errno on failure.
243 */ 294 */
244void __init numa_init_array(void) 295int __init numa_cleanup_meminfo(struct numa_meminfo *mi)
245{ 296{
246 int rr, i; 297 const u64 low = 0;
298 const u64 high = (u64)max_pfn << PAGE_SHIFT;
299 int i, j, k;
247 300
248 rr = first_node(node_online_map); 301 for (i = 0; i < mi->nr_blks; i++) {
249 for (i = 0; i < nr_cpu_ids; i++) { 302 struct numa_memblk *bi = &mi->blk[i];
250 if (early_cpu_to_node(i) != NUMA_NO_NODE)
251 continue;
252 numa_set_node(i, rr);
253 rr = next_node(rr, node_online_map);
254 if (rr == MAX_NUMNODES)
255 rr = first_node(node_online_map);
256 }
257}
258
259#ifdef CONFIG_NUMA_EMU
260/* Numa emulation */
261static struct bootnode nodes[MAX_NUMNODES] __initdata;
262static struct bootnode physnodes[MAX_NUMNODES] __cpuinitdata;
263static char *cmdline __initdata;
264 303
265void __init numa_emu_cmdline(char *str) 304 /* make sure all blocks are inside the limits */
266{ 305 bi->start = max(bi->start, low);
267 cmdline = str; 306 bi->end = min(bi->end, high);
268}
269 307
270static int __init setup_physnodes(unsigned long start, unsigned long end, 308 /* and there's no empty block */
271 int acpi, int amd) 309 if (bi->start == bi->end) {
272{ 310 numa_remove_memblk_from(i--, mi);
273 int ret = 0;
274 int i;
275
276 memset(physnodes, 0, sizeof(physnodes));
277#ifdef CONFIG_ACPI_NUMA
278 if (acpi)
279 acpi_get_nodes(physnodes, start, end);
280#endif
281#ifdef CONFIG_AMD_NUMA
282 if (amd)
283 amd_get_nodes(physnodes);
284#endif
285 /*
286 * Basic sanity checking on the physical node map: there may be errors
287 * if the SRAT or AMD code incorrectly reported the topology or the mem=
288 * kernel parameter is used.
289 */
290 for (i = 0; i < MAX_NUMNODES; i++) {
291 if (physnodes[i].start == physnodes[i].end)
292 continue;
293 if (physnodes[i].start > end) {
294 physnodes[i].end = physnodes[i].start;
295 continue;
296 }
297 if (physnodes[i].end < start) {
298 physnodes[i].start = physnodes[i].end;
299 continue; 311 continue;
300 } 312 }
301 if (physnodes[i].start < start)
302 physnodes[i].start = start;
303 if (physnodes[i].end > end)
304 physnodes[i].end = end;
305 ret++;
306 }
307 313
308 /* 314 for (j = i + 1; j < mi->nr_blks; j++) {
309 * If no physical topology was detected, a single node is faked to cover 315 struct numa_memblk *bj = &mi->blk[j];
310 * the entire address space. 316 unsigned long start, end;
311 */
312 if (!ret) {
313 physnodes[ret].start = start;
314 physnodes[ret].end = end;
315 ret = 1;
316 }
317 return ret;
318}
319
320static void __init fake_physnodes(int acpi, int amd, int nr_nodes)
321{
322 int i;
323
324 BUG_ON(acpi && amd);
325#ifdef CONFIG_ACPI_NUMA
326 if (acpi)
327 acpi_fake_nodes(nodes, nr_nodes);
328#endif
329#ifdef CONFIG_AMD_NUMA
330 if (amd)
331 amd_fake_nodes(nodes, nr_nodes);
332#endif
333 if (!acpi && !amd)
334 for (i = 0; i < nr_cpu_ids; i++)
335 numa_set_node(i, 0);
336}
337
338/*
339 * Setups up nid to range from addr to addr + size. If the end
340 * boundary is greater than max_addr, then max_addr is used instead.
341 * The return value is 0 if there is additional memory left for
342 * allocation past addr and -1 otherwise. addr is adjusted to be at
343 * the end of the node.
344 */
345static int __init setup_node_range(int nid, u64 *addr, u64 size, u64 max_addr)
346{
347 int ret = 0;
348 nodes[nid].start = *addr;
349 *addr += size;
350 if (*addr >= max_addr) {
351 *addr = max_addr;
352 ret = -1;
353 }
354 nodes[nid].end = *addr;
355 node_set(nid, node_possible_map);
356 printk(KERN_INFO "Faking node %d at %016Lx-%016Lx (%LuMB)\n", nid,
357 nodes[nid].start, nodes[nid].end,
358 (nodes[nid].end - nodes[nid].start) >> 20);
359 return ret;
360}
361
362/*
363 * Sets up nr_nodes fake nodes interleaved over physical nodes ranging from addr
364 * to max_addr. The return value is the number of nodes allocated.
365 */
366static int __init split_nodes_interleave(u64 addr, u64 max_addr, int nr_nodes)
367{
368 nodemask_t physnode_mask = NODE_MASK_NONE;
369 u64 size;
370 int big;
371 int ret = 0;
372 int i;
373
374 if (nr_nodes <= 0)
375 return -1;
376 if (nr_nodes > MAX_NUMNODES) {
377 pr_info("numa=fake=%d too large, reducing to %d\n",
378 nr_nodes, MAX_NUMNODES);
379 nr_nodes = MAX_NUMNODES;
380 }
381
382 size = (max_addr - addr - memblock_x86_hole_size(addr, max_addr)) / nr_nodes;
383 /*
384 * Calculate the number of big nodes that can be allocated as a result
385 * of consolidating the remainder.
386 */
387 big = ((size & ~FAKE_NODE_MIN_HASH_MASK) * nr_nodes) /
388 FAKE_NODE_MIN_SIZE;
389
390 size &= FAKE_NODE_MIN_HASH_MASK;
391 if (!size) {
392 pr_err("Not enough memory for each node. "
393 "NUMA emulation disabled.\n");
394 return -1;
395 }
396
397 for (i = 0; i < MAX_NUMNODES; i++)
398 if (physnodes[i].start != physnodes[i].end)
399 node_set(i, physnode_mask);
400
401 /*
402 * Continue to fill physical nodes with fake nodes until there is no
403 * memory left on any of them.
404 */
405 while (nodes_weight(physnode_mask)) {
406 for_each_node_mask(i, physnode_mask) {
407 u64 end = physnodes[i].start + size;
408 u64 dma32_end = PFN_PHYS(MAX_DMA32_PFN);
409
410 if (ret < big)
411 end += FAKE_NODE_MIN_SIZE;
412 317
413 /* 318 /*
414 * Continue to add memory to this fake node if its 319 * See whether there are overlapping blocks. Whine
415 * non-reserved memory is less than the per-node size. 320 * about but allow overlaps of the same nid. They
321 * will be merged below.
416 */ 322 */
417 while (end - physnodes[i].start - 323 if (bi->end > bj->start && bi->start < bj->end) {
418 memblock_x86_hole_size(physnodes[i].start, end) < size) { 324 if (bi->nid != bj->nid) {
419 end += FAKE_NODE_MIN_SIZE; 325 pr_err("NUMA: node %d (%Lx-%Lx) overlaps with node %d (%Lx-%Lx)\n",
420 if (end > physnodes[i].end) { 326 bi->nid, bi->start, bi->end,
421 end = physnodes[i].end; 327 bj->nid, bj->start, bj->end);
422 break; 328 return -EINVAL;
423 } 329 }
330 pr_warning("NUMA: Warning: node %d (%Lx-%Lx) overlaps with itself (%Lx-%Lx)\n",
331 bi->nid, bi->start, bi->end,
332 bj->start, bj->end);
424 } 333 }
425 334
426 /* 335 /*
427 * If there won't be at least FAKE_NODE_MIN_SIZE of 336 * Join together blocks on the same node, holes
428 * non-reserved memory in ZONE_DMA32 for the next node, 337 * between which don't overlap with memory on other
429 * this one must extend to the boundary. 338 * nodes.
430 */
431 if (end < dma32_end && dma32_end - end -
432 memblock_x86_hole_size(end, dma32_end) < FAKE_NODE_MIN_SIZE)
433 end = dma32_end;
434
435 /*
436 * If there won't be enough non-reserved memory for the
437 * next node, this one must extend to the end of the
438 * physical node.
439 */ 339 */
440 if (physnodes[i].end - end - 340 if (bi->nid != bj->nid)
441 memblock_x86_hole_size(end, physnodes[i].end) < size) 341 continue;
442 end = physnodes[i].end; 342 start = max(min(bi->start, bj->start), low);
443 343 end = min(max(bi->end, bj->end), high);
444 /* 344 for (k = 0; k < mi->nr_blks; k++) {
445 * Avoid allocating more nodes than requested, which can 345 struct numa_memblk *bk = &mi->blk[k];
446 * happen as a result of rounding down each node's size 346
447 * to FAKE_NODE_MIN_SIZE. 347 if (bi->nid == bk->nid)
448 */ 348 continue;
449 if (nodes_weight(physnode_mask) + ret >= nr_nodes) 349 if (start < bk->end && end > bk->start)
450 end = physnodes[i].end; 350 break;
451 351 }
452 if (setup_node_range(ret++, &physnodes[i].start, 352 if (k < mi->nr_blks)
453 end - physnodes[i].start, 353 continue;
454 physnodes[i].end) < 0) 354 printk(KERN_INFO "NUMA: Node %d [%Lx,%Lx) + [%Lx,%Lx) -> [%lx,%lx)\n",
455 node_clear(i, physnode_mask); 355 bi->nid, bi->start, bi->end, bj->start, bj->end,
356 start, end);
357 bi->start = start;
358 bi->end = end;
359 numa_remove_memblk_from(j--, mi);
456 } 360 }
457 } 361 }
458 return ret;
459}
460
461/*
462 * Returns the end address of a node so that there is at least `size' amount of
463 * non-reserved memory or `max_addr' is reached.
464 */
465static u64 __init find_end_of_node(u64 start, u64 max_addr, u64 size)
466{
467 u64 end = start + size;
468 362
469 while (end - start - memblock_x86_hole_size(start, end) < size) { 363 for (i = mi->nr_blks; i < ARRAY_SIZE(mi->blk); i++) {
470 end += FAKE_NODE_MIN_SIZE; 364 mi->blk[i].start = mi->blk[i].end = 0;
471 if (end > max_addr) { 365 mi->blk[i].nid = NUMA_NO_NODE;
472 end = max_addr;
473 break;
474 }
475 } 366 }
476 return end; 367
368 return 0;
477} 369}
478 370
479/* 371/*
480 * Sets up fake nodes of `size' interleaved over physical nodes ranging from 372 * Set nodes, which have memory in @mi, in *@nodemask.
481 * `addr' to `max_addr'. The return value is the number of nodes allocated.
482 */ 373 */
483static int __init split_nodes_size_interleave(u64 addr, u64 max_addr, u64 size) 374static void __init numa_nodemask_from_meminfo(nodemask_t *nodemask,
375 const struct numa_meminfo *mi)
484{ 376{
485 nodemask_t physnode_mask = NODE_MASK_NONE;
486 u64 min_size;
487 int ret = 0;
488 int i; 377 int i;
489 378
490 if (!size) 379 for (i = 0; i < ARRAY_SIZE(mi->blk); i++)
491 return -1; 380 if (mi->blk[i].start != mi->blk[i].end &&
492 /* 381 mi->blk[i].nid != NUMA_NO_NODE)
493 * The limit on emulated nodes is MAX_NUMNODES, so the size per node is 382 node_set(mi->blk[i].nid, *nodemask);
494 * increased accordingly if the requested size is too small. This 383}
495 * creates a uniform distribution of node sizes across the entire
496 * machine (but not necessarily over physical nodes).
497 */
498 min_size = (max_addr - addr - memblock_x86_hole_size(addr, max_addr)) /
499 MAX_NUMNODES;
500 min_size = max(min_size, FAKE_NODE_MIN_SIZE);
501 if ((min_size & FAKE_NODE_MIN_HASH_MASK) < min_size)
502 min_size = (min_size + FAKE_NODE_MIN_SIZE) &
503 FAKE_NODE_MIN_HASH_MASK;
504 if (size < min_size) {
505 pr_err("Fake node size %LuMB too small, increasing to %LuMB\n",
506 size >> 20, min_size >> 20);
507 size = min_size;
508 }
509 size &= FAKE_NODE_MIN_HASH_MASK;
510
511 for (i = 0; i < MAX_NUMNODES; i++)
512 if (physnodes[i].start != physnodes[i].end)
513 node_set(i, physnode_mask);
514 /*
515 * Fill physical nodes with fake nodes of size until there is no memory
516 * left on any of them.
517 */
518 while (nodes_weight(physnode_mask)) {
519 for_each_node_mask(i, physnode_mask) {
520 u64 dma32_end = MAX_DMA32_PFN << PAGE_SHIFT;
521 u64 end;
522
523 end = find_end_of_node(physnodes[i].start,
524 physnodes[i].end, size);
525 /*
526 * If there won't be at least FAKE_NODE_MIN_SIZE of
527 * non-reserved memory in ZONE_DMA32 for the next node,
528 * this one must extend to the boundary.
529 */
530 if (end < dma32_end && dma32_end - end -
531 memblock_x86_hole_size(end, dma32_end) < FAKE_NODE_MIN_SIZE)
532 end = dma32_end;
533 384
534 /* 385/**
535 * If there won't be enough non-reserved memory for the 386 * numa_reset_distance - Reset NUMA distance table
536 * next node, this one must extend to the end of the 387 *
537 * physical node. 388 * The current table is freed. The next numa_set_distance() call will
538 */ 389 * create a new one.
539 if (physnodes[i].end - end - 390 */
540 memblock_x86_hole_size(end, physnodes[i].end) < size) 391void __init numa_reset_distance(void)
541 end = physnodes[i].end; 392{
393 size_t size = numa_distance_cnt * numa_distance_cnt * sizeof(numa_distance[0]);
542 394
543 /* 395 /* numa_distance could be 1LU marking allocation failure, test cnt */
544 * Setup the fake node that will be allocated as bootmem 396 if (numa_distance_cnt)
545 * later. If setup_node_range() returns non-zero, there 397 memblock_x86_free_range(__pa(numa_distance),
546 * is no more memory available on this physical node. 398 __pa(numa_distance) + size);
547 */ 399 numa_distance_cnt = 0;
548 if (setup_node_range(ret++, &physnodes[i].start, 400 numa_distance = NULL; /* enable table creation */
549 end - physnodes[i].start,
550 physnodes[i].end) < 0)
551 node_clear(i, physnode_mask);
552 }
553 }
554 return ret;
555} 401}
556 402
557/* 403static int __init numa_alloc_distance(void)
558 * Sets up the system RAM area from start_pfn to last_pfn according to the
559 * numa=fake command-line option.
560 */
561static int __init numa_emulation(unsigned long start_pfn,
562 unsigned long last_pfn, int acpi, int amd)
563{ 404{
564 u64 addr = start_pfn << PAGE_SHIFT; 405 nodemask_t nodes_parsed;
565 u64 max_addr = last_pfn << PAGE_SHIFT; 406 size_t size;
566 int num_nodes; 407 int i, j, cnt = 0;
567 int i; 408 u64 phys;
568 409
569 /* 410 /* size the new table and allocate it */
570 * If the numa=fake command-line contains a 'M' or 'G', it represents 411 nodes_parsed = numa_nodes_parsed;
571 * the fixed node size. Otherwise, if it is just a single number N, 412 numa_nodemask_from_meminfo(&nodes_parsed, &numa_meminfo);
572 * split the system RAM into N fake nodes.
573 */
574 if (strchr(cmdline, 'M') || strchr(cmdline, 'G')) {
575 u64 size;
576 413
577 size = memparse(cmdline, &cmdline); 414 for_each_node_mask(i, nodes_parsed)
578 num_nodes = split_nodes_size_interleave(addr, max_addr, size); 415 cnt = i;
579 } else { 416 cnt++;
580 unsigned long n; 417 size = cnt * cnt * sizeof(numa_distance[0]);
581 418
582 n = simple_strtoul(cmdline, NULL, 0); 419 phys = memblock_find_in_range(0, (u64)max_pfn_mapped << PAGE_SHIFT,
583 num_nodes = split_nodes_interleave(addr, max_addr, n); 420 size, PAGE_SIZE);
421 if (phys == MEMBLOCK_ERROR) {
422 pr_warning("NUMA: Warning: can't allocate distance table!\n");
423 /* don't retry until explicitly reset */
424 numa_distance = (void *)1LU;
425 return -ENOMEM;
584 } 426 }
427 memblock_x86_reserve_range(phys, phys + size, "NUMA DIST");
585 428
586 if (num_nodes < 0) 429 numa_distance = __va(phys);
587 return num_nodes; 430 numa_distance_cnt = cnt;
588 memnode_shift = compute_hash_shift(nodes, num_nodes, NULL); 431
589 if (memnode_shift < 0) { 432 /* fill with the default distances */
590 memnode_shift = 0; 433 for (i = 0; i < cnt; i++)
591 printk(KERN_ERR "No NUMA hash function found. NUMA emulation " 434 for (j = 0; j < cnt; j++)
592 "disabled.\n"); 435 numa_distance[i * cnt + j] = i == j ?
593 return -1; 436 LOCAL_DISTANCE : REMOTE_DISTANCE;
594 } 437 printk(KERN_DEBUG "NUMA: Initialized distance table, cnt=%d\n", cnt);
595 438
596 /*
597 * We need to vacate all active ranges that may have been registered for
598 * the e820 memory map.
599 */
600 remove_all_active_ranges();
601 for_each_node_mask(i, node_possible_map) {
602 memblock_x86_register_active_regions(i, nodes[i].start >> PAGE_SHIFT,
603 nodes[i].end >> PAGE_SHIFT);
604 setup_node_bootmem(i, nodes[i].start, nodes[i].end);
605 }
606 setup_physnodes(addr, max_addr, acpi, amd);
607 fake_physnodes(acpi, amd, num_nodes);
608 numa_init_array();
609 return 0; 439 return 0;
610} 440}
611#endif /* CONFIG_NUMA_EMU */
612 441
613void __init initmem_init(unsigned long start_pfn, unsigned long last_pfn, 442/**
614 int acpi, int amd) 443 * numa_set_distance - Set NUMA distance from one NUMA to another
444 * @from: the 'from' node to set distance
445 * @to: the 'to' node to set distance
446 * @distance: NUMA distance
447 *
448 * Set the distance from node @from to @to to @distance. If distance table
449 * doesn't exist, one which is large enough to accomodate all the currently
450 * known nodes will be created.
451 *
452 * If such table cannot be allocated, a warning is printed and further
453 * calls are ignored until the distance table is reset with
454 * numa_reset_distance().
455 *
456 * If @from or @to is higher than the highest known node at the time of
457 * table creation or @distance doesn't make sense, the call is ignored.
458 * This is to allow simplification of specific NUMA config implementations.
459 */
460void __init numa_set_distance(int from, int to, int distance)
615{ 461{
616 int i; 462 if (!numa_distance && numa_alloc_distance() < 0)
617
618 nodes_clear(node_possible_map);
619 nodes_clear(node_online_map);
620
621#ifdef CONFIG_NUMA_EMU
622 setup_physnodes(start_pfn << PAGE_SHIFT, last_pfn << PAGE_SHIFT,
623 acpi, amd);
624 if (cmdline && !numa_emulation(start_pfn, last_pfn, acpi, amd))
625 return; 463 return;
626 setup_physnodes(start_pfn << PAGE_SHIFT, last_pfn << PAGE_SHIFT,
627 acpi, amd);
628 nodes_clear(node_possible_map);
629 nodes_clear(node_online_map);
630#endif
631 464
632#ifdef CONFIG_ACPI_NUMA 465 if (from >= numa_distance_cnt || to >= numa_distance_cnt) {
633 if (!numa_off && acpi && !acpi_scan_nodes(start_pfn << PAGE_SHIFT, 466 printk_once(KERN_DEBUG "NUMA: Debug: distance out of bound, from=%d to=%d distance=%d\n",
634 last_pfn << PAGE_SHIFT)) 467 from, to, distance);
635 return; 468 return;
636 nodes_clear(node_possible_map); 469 }
637 nodes_clear(node_online_map);
638#endif
639 470
640#ifdef CONFIG_AMD_NUMA 471 if ((u8)distance != distance ||
641 if (!numa_off && amd && !amd_scan_nodes()) 472 (from == to && distance != LOCAL_DISTANCE)) {
473 pr_warn_once("NUMA: Warning: invalid distance parameter, from=%d to=%d distance=%d\n",
474 from, to, distance);
642 return; 475 return;
643 nodes_clear(node_possible_map); 476 }
644 nodes_clear(node_online_map);
645#endif
646 printk(KERN_INFO "%s\n",
647 numa_off ? "NUMA turned off" : "No NUMA configuration found");
648 477
649 printk(KERN_INFO "Faking a node at %016lx-%016lx\n", 478 numa_distance[from * numa_distance_cnt + to] = distance;
650 start_pfn << PAGE_SHIFT,
651 last_pfn << PAGE_SHIFT);
652 /* setup dummy node covering all memory */
653 memnode_shift = 63;
654 memnodemap = memnode.embedded_map;
655 memnodemap[0] = 0;
656 node_set_online(0);
657 node_set(0, node_possible_map);
658 for (i = 0; i < nr_cpu_ids; i++)
659 numa_set_node(i, 0);
660 memblock_x86_register_active_regions(0, start_pfn, last_pfn);
661 setup_node_bootmem(0, start_pfn << PAGE_SHIFT, last_pfn << PAGE_SHIFT);
662} 479}
663 480
664unsigned long __init numa_free_all_bootmem(void) 481int __node_distance(int from, int to)
665{ 482{
666 unsigned long pages = 0; 483 if (from >= numa_distance_cnt || to >= numa_distance_cnt)
667 int i; 484 return from == to ? LOCAL_DISTANCE : REMOTE_DISTANCE;
485 return numa_distance[from * numa_distance_cnt + to];
486}
487EXPORT_SYMBOL(__node_distance);
668 488
669 for_each_online_node(i) 489/*
670 pages += free_all_bootmem_node(NODE_DATA(i)); 490 * Sanity check to catch more bad NUMA configurations (they are amazingly
491 * common). Make sure the nodes cover all memory.
492 */
493static bool __init numa_meminfo_cover_memory(const struct numa_meminfo *mi)
494{
495 unsigned long numaram, e820ram;
496 int i;
671 497
672 pages += free_all_memory_core_early(MAX_NUMNODES); 498 numaram = 0;
499 for (i = 0; i < mi->nr_blks; i++) {
500 unsigned long s = mi->blk[i].start >> PAGE_SHIFT;
501 unsigned long e = mi->blk[i].end >> PAGE_SHIFT;
502 numaram += e - s;
503 numaram -= __absent_pages_in_range(mi->blk[i].nid, s, e);
504 if ((long)numaram < 0)
505 numaram = 0;
506 }
673 507
674 return pages; 508 e820ram = max_pfn - (memblock_x86_hole_size(0,
509 max_pfn << PAGE_SHIFT) >> PAGE_SHIFT);
510 /* We seem to lose 3 pages somewhere. Allow 1M of slack. */
511 if ((long)(e820ram - numaram) >= (1 << (20 - PAGE_SHIFT))) {
512 printk(KERN_ERR "NUMA: nodes only cover %luMB of your %luMB e820 RAM. Not used.\n",
513 (numaram << PAGE_SHIFT) >> 20,
514 (e820ram << PAGE_SHIFT) >> 20);
515 return false;
516 }
517 return true;
675} 518}
676 519
677#ifdef CONFIG_NUMA 520static int __init numa_register_memblks(struct numa_meminfo *mi)
678
679static __init int find_near_online_node(int node)
680{ 521{
681 int n, val; 522 int i, nid;
682 int min_val = INT_MAX;
683 int best_node = -1;
684 523
685 for_each_online_node(n) { 524 /* Account for nodes with cpus and no memory */
686 val = node_distance(node, n); 525 node_possible_map = numa_nodes_parsed;
526 numa_nodemask_from_meminfo(&node_possible_map, mi);
527 if (WARN_ON(nodes_empty(node_possible_map)))
528 return -EINVAL;
687 529
688 if (val < min_val) { 530 memnode_shift = compute_hash_shift(mi);
689 min_val = val; 531 if (memnode_shift < 0) {
690 best_node = n; 532 printk(KERN_ERR "NUMA: No NUMA node hash function found. Contact maintainer\n");
533 return -EINVAL;
534 }
535
536 for (i = 0; i < mi->nr_blks; i++)
537 memblock_x86_register_active_regions(mi->blk[i].nid,
538 mi->blk[i].start >> PAGE_SHIFT,
539 mi->blk[i].end >> PAGE_SHIFT);
540
541 /* for out of order entries */
542 sort_node_map();
543 if (!numa_meminfo_cover_memory(mi))
544 return -EINVAL;
545
546 /* Finally register nodes. */
547 for_each_node_mask(nid, node_possible_map) {
548 u64 start = (u64)max_pfn << PAGE_SHIFT;
549 u64 end = 0;
550
551 for (i = 0; i < mi->nr_blks; i++) {
552 if (nid != mi->blk[i].nid)
553 continue;
554 start = min(mi->blk[i].start, start);
555 end = max(mi->blk[i].end, end);
691 } 556 }
557
558 if (start < end)
559 setup_node_bootmem(nid, start, end);
692 } 560 }
693 561
694 return best_node; 562 return 0;
695} 563}
696 564
697/* 565/**
698 * Setup early cpu_to_node. 566 * dummy_numma_init - Fallback dummy NUMA init
699 * 567 *
700 * Populate cpu_to_node[] only if x86_cpu_to_apicid[], 568 * Used if there's no underlying NUMA architecture, NUMA initialization
701 * and apicid_to_node[] tables have valid entries for a CPU. 569 * fails, or NUMA is disabled on the command line.
702 * This means we skip cpu_to_node[] initialisation for NUMA
703 * emulation and faking node case (when running a kernel compiled
704 * for NUMA on a non NUMA box), which is OK as cpu_to_node[]
705 * is already initialized in a round robin manner at numa_init_array,
706 * prior to this call, and this initialization is good enough
707 * for the fake NUMA cases.
708 * 570 *
709 * Called before the per_cpu areas are setup. 571 * Must online at least one node and add memory blocks that cover all
572 * allowed memory. This function must not fail.
710 */ 573 */
711void __init init_cpu_to_node(void) 574static int __init dummy_numa_init(void)
712{ 575{
713 int cpu; 576 printk(KERN_INFO "%s\n",
714 u16 *cpu_to_apicid = early_per_cpu_ptr(x86_cpu_to_apicid); 577 numa_off ? "NUMA turned off" : "No NUMA configuration found");
715 578 printk(KERN_INFO "Faking a node at %016lx-%016lx\n",
716 BUG_ON(cpu_to_apicid == NULL); 579 0LU, max_pfn << PAGE_SHIFT);
717 580
718 for_each_possible_cpu(cpu) { 581 node_set(0, numa_nodes_parsed);
719 int node; 582 numa_add_memblk(0, 0, (u64)max_pfn << PAGE_SHIFT);
720 u16 apicid = cpu_to_apicid[cpu];
721 583
722 if (apicid == BAD_APICID) 584 return 0;
723 continue;
724 node = apicid_to_node[apicid];
725 if (node == NUMA_NO_NODE)
726 continue;
727 if (!node_online(node))
728 node = find_near_online_node(node);
729 numa_set_node(cpu, node);
730 }
731} 585}
732#endif
733 586
734 587static int __init numa_init(int (*init_func)(void))
735void __cpuinit numa_set_node(int cpu, int node)
736{ 588{
737 int *cpu_to_node_map = early_per_cpu_ptr(x86_cpu_to_node_map); 589 int i;
738 590 int ret;
739 /* early setting, no percpu area yet */
740 if (cpu_to_node_map) {
741 cpu_to_node_map[cpu] = node;
742 return;
743 }
744
745#ifdef CONFIG_DEBUG_PER_CPU_MAPS
746 if (cpu >= nr_cpu_ids || !cpu_possible(cpu)) {
747 printk(KERN_ERR "numa_set_node: invalid cpu# (%d)\n", cpu);
748 dump_stack();
749 return;
750 }
751#endif
752 per_cpu(x86_cpu_to_node_map, cpu) = node;
753 591
754 if (node != NUMA_NO_NODE) 592 for (i = 0; i < MAX_LOCAL_APIC; i++)
755 set_cpu_numa_node(cpu, node); 593 set_apicid_to_node(i, NUMA_NO_NODE);
756}
757 594
758void __cpuinit numa_clear_node(int cpu) 595 nodes_clear(numa_nodes_parsed);
759{ 596 nodes_clear(node_possible_map);
760 numa_set_node(cpu, NUMA_NO_NODE); 597 nodes_clear(node_online_map);
761} 598 memset(&numa_meminfo, 0, sizeof(numa_meminfo));
762 599 remove_all_active_ranges();
763#ifndef CONFIG_DEBUG_PER_CPU_MAPS 600 numa_reset_distance();
764 601
765#ifndef CONFIG_NUMA_EMU 602 ret = init_func();
766void __cpuinit numa_add_cpu(int cpu) 603 if (ret < 0)
767{ 604 return ret;
768 cpumask_set_cpu(cpu, node_to_cpumask_map[early_cpu_to_node(cpu)]); 605 ret = numa_cleanup_meminfo(&numa_meminfo);
769} 606 if (ret < 0)
607 return ret;
770 608
771void __cpuinit numa_remove_cpu(int cpu) 609 numa_emulation(&numa_meminfo, numa_distance_cnt);
772{
773 cpumask_clear_cpu(cpu, node_to_cpumask_map[early_cpu_to_node(cpu)]);
774}
775#else
776void __cpuinit numa_add_cpu(int cpu)
777{
778 unsigned long addr;
779 u16 apicid;
780 int physnid;
781 int nid = NUMA_NO_NODE;
782 610
783 nid = early_cpu_to_node(cpu); 611 ret = numa_register_memblks(&numa_meminfo);
784 BUG_ON(nid == NUMA_NO_NODE || !node_online(nid)); 612 if (ret < 0)
613 return ret;
785 614
786 /* 615 for (i = 0; i < nr_cpu_ids; i++) {
787 * Use the starting address of the emulated node to find which physical 616 int nid = early_cpu_to_node(i);
788 * node it is allocated on.
789 */
790 addr = node_start_pfn(nid) << PAGE_SHIFT;
791 for (physnid = 0; physnid < MAX_NUMNODES; physnid++)
792 if (addr >= physnodes[physnid].start &&
793 addr < physnodes[physnid].end)
794 break;
795 617
796 /* 618 if (nid == NUMA_NO_NODE)
797 * Map the cpu to each emulated node that is allocated on the physical 619 continue;
798 * node of the cpu's apic id. 620 if (!node_online(nid))
799 */ 621 numa_clear_node(i);
800 for_each_online_node(nid) {
801 addr = node_start_pfn(nid) << PAGE_SHIFT;
802 if (addr >= physnodes[physnid].start &&
803 addr < physnodes[physnid].end)
804 cpumask_set_cpu(cpu, node_to_cpumask_map[nid]);
805 } 622 }
623 numa_init_array();
624 return 0;
806} 625}
807 626
808void __cpuinit numa_remove_cpu(int cpu) 627void __init initmem_init(void)
809{ 628{
810 int i; 629 int ret;
811 630
812 for_each_online_node(i) 631 if (!numa_off) {
813 cpumask_clear_cpu(cpu, node_to_cpumask_map[i]); 632#ifdef CONFIG_ACPI_NUMA
814} 633 ret = numa_init(x86_acpi_numa_init);
815#endif /* !CONFIG_NUMA_EMU */ 634 if (!ret)
816 635 return;
817#else /* CONFIG_DEBUG_PER_CPU_MAPS */ 636#endif
818static struct cpumask __cpuinit *debug_cpumask_set_cpu(int cpu, int enable) 637#ifdef CONFIG_AMD_NUMA
819{ 638 ret = numa_init(amd_numa_init);
820 int node = early_cpu_to_node(cpu); 639 if (!ret)
821 struct cpumask *mask; 640 return;
822 char buf[64]; 641#endif
823
824 mask = node_to_cpumask_map[node];
825 if (!mask) {
826 pr_err("node_to_cpumask_map[%i] NULL\n", node);
827 dump_stack();
828 return NULL;
829 } 642 }
830 643
831 cpulist_scnprintf(buf, sizeof(buf), mask); 644 numa_init(dummy_numa_init);
832 printk(KERN_DEBUG "%s cpu %d node %d: mask now %s\n",
833 enable ? "numa_add_cpu" : "numa_remove_cpu",
834 cpu, node, buf);
835 return mask;
836} 645}
837 646
838/* 647unsigned long __init numa_free_all_bootmem(void)
839 * --------- debug versions of the numa functions ---------
840 */
841#ifndef CONFIG_NUMA_EMU
842static void __cpuinit numa_set_cpumask(int cpu, int enable)
843{
844 struct cpumask *mask;
845
846 mask = debug_cpumask_set_cpu(cpu, enable);
847 if (!mask)
848 return;
849
850 if (enable)
851 cpumask_set_cpu(cpu, mask);
852 else
853 cpumask_clear_cpu(cpu, mask);
854}
855#else
856static void __cpuinit numa_set_cpumask(int cpu, int enable)
857{ 648{
858 int node = early_cpu_to_node(cpu); 649 unsigned long pages = 0;
859 struct cpumask *mask;
860 int i; 650 int i;
861 651
862 for_each_online_node(i) { 652 for_each_online_node(i)
863 unsigned long addr; 653 pages += free_all_bootmem_node(NODE_DATA(i));
864
865 addr = node_start_pfn(i) << PAGE_SHIFT;
866 if (addr < physnodes[node].start ||
867 addr >= physnodes[node].end)
868 continue;
869 mask = debug_cpumask_set_cpu(cpu, enable);
870 if (!mask)
871 return;
872
873 if (enable)
874 cpumask_set_cpu(cpu, mask);
875 else
876 cpumask_clear_cpu(cpu, mask);
877 }
878}
879#endif /* CONFIG_NUMA_EMU */
880 654
881void __cpuinit numa_add_cpu(int cpu) 655 pages += free_all_memory_core_early(MAX_NUMNODES);
882{
883 numa_set_cpumask(cpu, 1);
884}
885 656
886void __cpuinit numa_remove_cpu(int cpu) 657 return pages;
887{
888 numa_set_cpumask(cpu, 0);
889} 658}
890 659
891int __cpu_to_node(int cpu) 660int __cpuinit numa_cpu_node(int cpu)
892{ 661{
893 if (early_per_cpu_ptr(x86_cpu_to_node_map)) { 662 int apicid = early_per_cpu(x86_cpu_to_apicid, cpu);
894 printk(KERN_WARNING
895 "cpu_to_node(%d): usage too early!\n", cpu);
896 dump_stack();
897 return early_per_cpu_ptr(x86_cpu_to_node_map)[cpu];
898 }
899 return per_cpu(x86_cpu_to_node_map, cpu);
900}
901EXPORT_SYMBOL(__cpu_to_node);
902 663
903/* 664 if (apicid != BAD_APICID)
904 * Same function as cpu_to_node() but used if called before the 665 return __apicid_to_node[apicid];
905 * per_cpu areas are setup. 666 return NUMA_NO_NODE;
906 */
907int early_cpu_to_node(int cpu)
908{
909 if (early_per_cpu_ptr(x86_cpu_to_node_map))
910 return early_per_cpu_ptr(x86_cpu_to_node_map)[cpu];
911
912 if (!cpu_possible(cpu)) {
913 printk(KERN_WARNING
914 "early_cpu_to_node(%d): no per_cpu area!\n", cpu);
915 dump_stack();
916 return NUMA_NO_NODE;
917 }
918 return per_cpu(x86_cpu_to_node_map, cpu);
919} 667}
920
921/*
922 * --------- end of debug versions of the numa functions ---------
923 */
924
925#endif /* CONFIG_DEBUG_PER_CPU_MAPS */
diff --git a/arch/x86/mm/numa_emulation.c b/arch/x86/mm/numa_emulation.c
new file mode 100644
index 000000000000..ad091e4cff17
--- /dev/null
+++ b/arch/x86/mm/numa_emulation.c
@@ -0,0 +1,494 @@
1/*
2 * NUMA emulation
3 */
4#include <linux/kernel.h>
5#include <linux/errno.h>
6#include <linux/topology.h>
7#include <linux/memblock.h>
8#include <asm/dma.h>
9
10#include "numa_internal.h"
11
12static int emu_nid_to_phys[MAX_NUMNODES] __cpuinitdata;
13static char *emu_cmdline __initdata;
14
15void __init numa_emu_cmdline(char *str)
16{
17 emu_cmdline = str;
18}
19
20static int __init emu_find_memblk_by_nid(int nid, const struct numa_meminfo *mi)
21{
22 int i;
23
24 for (i = 0; i < mi->nr_blks; i++)
25 if (mi->blk[i].nid == nid)
26 return i;
27 return -ENOENT;
28}
29
30/*
31 * Sets up nid to range from @start to @end. The return value is -errno if
32 * something went wrong, 0 otherwise.
33 */
34static int __init emu_setup_memblk(struct numa_meminfo *ei,
35 struct numa_meminfo *pi,
36 int nid, int phys_blk, u64 size)
37{
38 struct numa_memblk *eb = &ei->blk[ei->nr_blks];
39 struct numa_memblk *pb = &pi->blk[phys_blk];
40
41 if (ei->nr_blks >= NR_NODE_MEMBLKS) {
42 pr_err("NUMA: Too many emulated memblks, failing emulation\n");
43 return -EINVAL;
44 }
45
46 ei->nr_blks++;
47 eb->start = pb->start;
48 eb->end = pb->start + size;
49 eb->nid = nid;
50
51 if (emu_nid_to_phys[nid] == NUMA_NO_NODE)
52 emu_nid_to_phys[nid] = pb->nid;
53
54 pb->start += size;
55 if (pb->start >= pb->end) {
56 WARN_ON_ONCE(pb->start > pb->end);
57 numa_remove_memblk_from(phys_blk, pi);
58 }
59
60 printk(KERN_INFO "Faking node %d at %016Lx-%016Lx (%LuMB)\n", nid,
61 eb->start, eb->end, (eb->end - eb->start) >> 20);
62 return 0;
63}
64
65/*
66 * Sets up nr_nodes fake nodes interleaved over physical nodes ranging from addr
67 * to max_addr. The return value is the number of nodes allocated.
68 */
69static int __init split_nodes_interleave(struct numa_meminfo *ei,
70 struct numa_meminfo *pi,
71 u64 addr, u64 max_addr, int nr_nodes)
72{
73 nodemask_t physnode_mask = NODE_MASK_NONE;
74 u64 size;
75 int big;
76 int nid = 0;
77 int i, ret;
78
79 if (nr_nodes <= 0)
80 return -1;
81 if (nr_nodes > MAX_NUMNODES) {
82 pr_info("numa=fake=%d too large, reducing to %d\n",
83 nr_nodes, MAX_NUMNODES);
84 nr_nodes = MAX_NUMNODES;
85 }
86
87 size = (max_addr - addr - memblock_x86_hole_size(addr, max_addr)) / nr_nodes;
88 /*
89 * Calculate the number of big nodes that can be allocated as a result
90 * of consolidating the remainder.
91 */
92 big = ((size & ~FAKE_NODE_MIN_HASH_MASK) * nr_nodes) /
93 FAKE_NODE_MIN_SIZE;
94
95 size &= FAKE_NODE_MIN_HASH_MASK;
96 if (!size) {
97 pr_err("Not enough memory for each node. "
98 "NUMA emulation disabled.\n");
99 return -1;
100 }
101
102 for (i = 0; i < pi->nr_blks; i++)
103 node_set(pi->blk[i].nid, physnode_mask);
104
105 /*
106 * Continue to fill physical nodes with fake nodes until there is no
107 * memory left on any of them.
108 */
109 while (nodes_weight(physnode_mask)) {
110 for_each_node_mask(i, physnode_mask) {
111 u64 dma32_end = PFN_PHYS(MAX_DMA32_PFN);
112 u64 start, limit, end;
113 int phys_blk;
114
115 phys_blk = emu_find_memblk_by_nid(i, pi);
116 if (phys_blk < 0) {
117 node_clear(i, physnode_mask);
118 continue;
119 }
120 start = pi->blk[phys_blk].start;
121 limit = pi->blk[phys_blk].end;
122 end = start + size;
123
124 if (nid < big)
125 end += FAKE_NODE_MIN_SIZE;
126
127 /*
128 * Continue to add memory to this fake node if its
129 * non-reserved memory is less than the per-node size.
130 */
131 while (end - start -
132 memblock_x86_hole_size(start, end) < size) {
133 end += FAKE_NODE_MIN_SIZE;
134 if (end > limit) {
135 end = limit;
136 break;
137 }
138 }
139
140 /*
141 * If there won't be at least FAKE_NODE_MIN_SIZE of
142 * non-reserved memory in ZONE_DMA32 for the next node,
143 * this one must extend to the boundary.
144 */
145 if (end < dma32_end && dma32_end - end -
146 memblock_x86_hole_size(end, dma32_end) < FAKE_NODE_MIN_SIZE)
147 end = dma32_end;
148
149 /*
150 * If there won't be enough non-reserved memory for the
151 * next node, this one must extend to the end of the
152 * physical node.
153 */
154 if (limit - end -
155 memblock_x86_hole_size(end, limit) < size)
156 end = limit;
157
158 ret = emu_setup_memblk(ei, pi, nid++ % nr_nodes,
159 phys_blk,
160 min(end, limit) - start);
161 if (ret < 0)
162 return ret;
163 }
164 }
165 return 0;
166}
167
168/*
169 * Returns the end address of a node so that there is at least `size' amount of
170 * non-reserved memory or `max_addr' is reached.
171 */
172static u64 __init find_end_of_node(u64 start, u64 max_addr, u64 size)
173{
174 u64 end = start + size;
175
176 while (end - start - memblock_x86_hole_size(start, end) < size) {
177 end += FAKE_NODE_MIN_SIZE;
178 if (end > max_addr) {
179 end = max_addr;
180 break;
181 }
182 }
183 return end;
184}
185
186/*
187 * Sets up fake nodes of `size' interleaved over physical nodes ranging from
188 * `addr' to `max_addr'. The return value is the number of nodes allocated.
189 */
190static int __init split_nodes_size_interleave(struct numa_meminfo *ei,
191 struct numa_meminfo *pi,
192 u64 addr, u64 max_addr, u64 size)
193{
194 nodemask_t physnode_mask = NODE_MASK_NONE;
195 u64 min_size;
196 int nid = 0;
197 int i, ret;
198
199 if (!size)
200 return -1;
201 /*
202 * The limit on emulated nodes is MAX_NUMNODES, so the size per node is
203 * increased accordingly if the requested size is too small. This
204 * creates a uniform distribution of node sizes across the entire
205 * machine (but not necessarily over physical nodes).
206 */
207 min_size = (max_addr - addr - memblock_x86_hole_size(addr, max_addr)) /
208 MAX_NUMNODES;
209 min_size = max(min_size, FAKE_NODE_MIN_SIZE);
210 if ((min_size & FAKE_NODE_MIN_HASH_MASK) < min_size)
211 min_size = (min_size + FAKE_NODE_MIN_SIZE) &
212 FAKE_NODE_MIN_HASH_MASK;
213 if (size < min_size) {
214 pr_err("Fake node size %LuMB too small, increasing to %LuMB\n",
215 size >> 20, min_size >> 20);
216 size = min_size;
217 }
218 size &= FAKE_NODE_MIN_HASH_MASK;
219
220 for (i = 0; i < pi->nr_blks; i++)
221 node_set(pi->blk[i].nid, physnode_mask);
222
223 /*
224 * Fill physical nodes with fake nodes of size until there is no memory
225 * left on any of them.
226 */
227 while (nodes_weight(physnode_mask)) {
228 for_each_node_mask(i, physnode_mask) {
229 u64 dma32_end = MAX_DMA32_PFN << PAGE_SHIFT;
230 u64 start, limit, end;
231 int phys_blk;
232
233 phys_blk = emu_find_memblk_by_nid(i, pi);
234 if (phys_blk < 0) {
235 node_clear(i, physnode_mask);
236 continue;
237 }
238 start = pi->blk[phys_blk].start;
239 limit = pi->blk[phys_blk].end;
240
241 end = find_end_of_node(start, limit, size);
242 /*
243 * If there won't be at least FAKE_NODE_MIN_SIZE of
244 * non-reserved memory in ZONE_DMA32 for the next node,
245 * this one must extend to the boundary.
246 */
247 if (end < dma32_end && dma32_end - end -
248 memblock_x86_hole_size(end, dma32_end) < FAKE_NODE_MIN_SIZE)
249 end = dma32_end;
250
251 /*
252 * If there won't be enough non-reserved memory for the
253 * next node, this one must extend to the end of the
254 * physical node.
255 */
256 if (limit - end -
257 memblock_x86_hole_size(end, limit) < size)
258 end = limit;
259
260 ret = emu_setup_memblk(ei, pi, nid++ % MAX_NUMNODES,
261 phys_blk,
262 min(end, limit) - start);
263 if (ret < 0)
264 return ret;
265 }
266 }
267 return 0;
268}
269
270/**
271 * numa_emulation - Emulate NUMA nodes
272 * @numa_meminfo: NUMA configuration to massage
273 * @numa_dist_cnt: The size of the physical NUMA distance table
274 *
275 * Emulate NUMA nodes according to the numa=fake kernel parameter.
276 * @numa_meminfo contains the physical memory configuration and is modified
277 * to reflect the emulated configuration on success. @numa_dist_cnt is
278 * used to determine the size of the physical distance table.
279 *
280 * On success, the following modifications are made.
281 *
282 * - @numa_meminfo is updated to reflect the emulated nodes.
283 *
284 * - __apicid_to_node[] is updated such that APIC IDs are mapped to the
285 * emulated nodes.
286 *
287 * - NUMA distance table is rebuilt to represent distances between emulated
288 * nodes. The distances are determined considering how emulated nodes
289 * are mapped to physical nodes and match the actual distances.
290 *
291 * - emu_nid_to_phys[] reflects how emulated nodes are mapped to physical
292 * nodes. This is used by numa_add_cpu() and numa_remove_cpu().
293 *
294 * If emulation is not enabled or fails, emu_nid_to_phys[] is filled with
295 * identity mapping and no other modification is made.
296 */
297void __init numa_emulation(struct numa_meminfo *numa_meminfo, int numa_dist_cnt)
298{
299 static struct numa_meminfo ei __initdata;
300 static struct numa_meminfo pi __initdata;
301 const u64 max_addr = max_pfn << PAGE_SHIFT;
302 u8 *phys_dist = NULL;
303 size_t phys_size = numa_dist_cnt * numa_dist_cnt * sizeof(phys_dist[0]);
304 int max_emu_nid, dfl_phys_nid;
305 int i, j, ret;
306
307 if (!emu_cmdline)
308 goto no_emu;
309
310 memset(&ei, 0, sizeof(ei));
311 pi = *numa_meminfo;
312
313 for (i = 0; i < MAX_NUMNODES; i++)
314 emu_nid_to_phys[i] = NUMA_NO_NODE;
315
316 /*
317 * If the numa=fake command-line contains a 'M' or 'G', it represents
318 * the fixed node size. Otherwise, if it is just a single number N,
319 * split the system RAM into N fake nodes.
320 */
321 if (strchr(emu_cmdline, 'M') || strchr(emu_cmdline, 'G')) {
322 u64 size;
323
324 size = memparse(emu_cmdline, &emu_cmdline);
325 ret = split_nodes_size_interleave(&ei, &pi, 0, max_addr, size);
326 } else {
327 unsigned long n;
328
329 n = simple_strtoul(emu_cmdline, NULL, 0);
330 ret = split_nodes_interleave(&ei, &pi, 0, max_addr, n);
331 }
332
333 if (ret < 0)
334 goto no_emu;
335
336 if (numa_cleanup_meminfo(&ei) < 0) {
337 pr_warning("NUMA: Warning: constructed meminfo invalid, disabling emulation\n");
338 goto no_emu;
339 }
340
341 /* copy the physical distance table */
342 if (numa_dist_cnt) {
343 u64 phys;
344
345 phys = memblock_find_in_range(0,
346 (u64)max_pfn_mapped << PAGE_SHIFT,
347 phys_size, PAGE_SIZE);
348 if (phys == MEMBLOCK_ERROR) {
349 pr_warning("NUMA: Warning: can't allocate copy of distance table, disabling emulation\n");
350 goto no_emu;
351 }
352 memblock_x86_reserve_range(phys, phys + phys_size, "TMP NUMA DIST");
353 phys_dist = __va(phys);
354
355 for (i = 0; i < numa_dist_cnt; i++)
356 for (j = 0; j < numa_dist_cnt; j++)
357 phys_dist[i * numa_dist_cnt + j] =
358 node_distance(i, j);
359 }
360
361 /*
362 * Determine the max emulated nid and the default phys nid to use
363 * for unmapped nodes.
364 */
365 max_emu_nid = 0;
366 dfl_phys_nid = NUMA_NO_NODE;
367 for (i = 0; i < ARRAY_SIZE(emu_nid_to_phys); i++) {
368 if (emu_nid_to_phys[i] != NUMA_NO_NODE) {
369 max_emu_nid = i;
370 if (dfl_phys_nid == NUMA_NO_NODE)
371 dfl_phys_nid = emu_nid_to_phys[i];
372 }
373 }
374 if (dfl_phys_nid == NUMA_NO_NODE) {
375 pr_warning("NUMA: Warning: can't determine default physical node, disabling emulation\n");
376 goto no_emu;
377 }
378
379 /* commit */
380 *numa_meminfo = ei;
381
382 /*
383 * Transform __apicid_to_node table to use emulated nids by
384 * reverse-mapping phys_nid. The maps should always exist but fall
385 * back to zero just in case.
386 */
387 for (i = 0; i < ARRAY_SIZE(__apicid_to_node); i++) {
388 if (__apicid_to_node[i] == NUMA_NO_NODE)
389 continue;
390 for (j = 0; j < ARRAY_SIZE(emu_nid_to_phys); j++)
391 if (__apicid_to_node[i] == emu_nid_to_phys[j])
392 break;
393 __apicid_to_node[i] = j < ARRAY_SIZE(emu_nid_to_phys) ? j : 0;
394 }
395
396 /* make sure all emulated nodes are mapped to a physical node */
397 for (i = 0; i < ARRAY_SIZE(emu_nid_to_phys); i++)
398 if (emu_nid_to_phys[i] == NUMA_NO_NODE)
399 emu_nid_to_phys[i] = dfl_phys_nid;
400
401 /* transform distance table */
402 numa_reset_distance();
403 for (i = 0; i < max_emu_nid + 1; i++) {
404 for (j = 0; j < max_emu_nid + 1; j++) {
405 int physi = emu_nid_to_phys[i];
406 int physj = emu_nid_to_phys[j];
407 int dist;
408
409 if (physi >= numa_dist_cnt || physj >= numa_dist_cnt)
410 dist = physi == physj ?
411 LOCAL_DISTANCE : REMOTE_DISTANCE;
412 else
413 dist = phys_dist[physi * numa_dist_cnt + physj];
414
415 numa_set_distance(i, j, dist);
416 }
417 }
418
419 /* free the copied physical distance table */
420 if (phys_dist)
421 memblock_x86_free_range(__pa(phys_dist), __pa(phys_dist) + phys_size);
422 return;
423
424no_emu:
425 /* No emulation. Build identity emu_nid_to_phys[] for numa_add_cpu() */
426 for (i = 0; i < ARRAY_SIZE(emu_nid_to_phys); i++)
427 emu_nid_to_phys[i] = i;
428}
429
430#ifndef CONFIG_DEBUG_PER_CPU_MAPS
431void __cpuinit numa_add_cpu(int cpu)
432{
433 int physnid, nid;
434
435 nid = early_cpu_to_node(cpu);
436 BUG_ON(nid == NUMA_NO_NODE || !node_online(nid));
437
438 physnid = emu_nid_to_phys[nid];
439
440 /*
441 * Map the cpu to each emulated node that is allocated on the physical
442 * node of the cpu's apic id.
443 */
444 for_each_online_node(nid)
445 if (emu_nid_to_phys[nid] == physnid)
446 cpumask_set_cpu(cpu, node_to_cpumask_map[nid]);
447}
448
449void __cpuinit numa_remove_cpu(int cpu)
450{
451 int i;
452
453 for_each_online_node(i)
454 cpumask_clear_cpu(cpu, node_to_cpumask_map[i]);
455}
456#else /* !CONFIG_DEBUG_PER_CPU_MAPS */
457static void __cpuinit numa_set_cpumask(int cpu, int enable)
458{
459 struct cpumask *mask;
460 int nid, physnid, i;
461
462 nid = early_cpu_to_node(cpu);
463 if (nid == NUMA_NO_NODE) {
464 /* early_cpu_to_node() already emits a warning and trace */
465 return;
466 }
467
468 physnid = emu_nid_to_phys[nid];
469
470 for_each_online_node(i) {
471 if (emu_nid_to_phys[nid] != physnid)
472 continue;
473
474 mask = debug_cpumask_set_cpu(cpu, enable);
475 if (!mask)
476 return;
477
478 if (enable)
479 cpumask_set_cpu(cpu, mask);
480 else
481 cpumask_clear_cpu(cpu, mask);
482 }
483}
484
485void __cpuinit numa_add_cpu(int cpu)
486{
487 numa_set_cpumask(cpu, 1);
488}
489
490void __cpuinit numa_remove_cpu(int cpu)
491{
492 numa_set_cpumask(cpu, 0);
493}
494#endif /* !CONFIG_DEBUG_PER_CPU_MAPS */
diff --git a/arch/x86/mm/numa_internal.h b/arch/x86/mm/numa_internal.h
new file mode 100644
index 000000000000..ef2d97377d7c
--- /dev/null
+++ b/arch/x86/mm/numa_internal.h
@@ -0,0 +1,31 @@
1#ifndef __X86_MM_NUMA_INTERNAL_H
2#define __X86_MM_NUMA_INTERNAL_H
3
4#include <linux/types.h>
5#include <asm/numa.h>
6
7struct numa_memblk {
8 u64 start;
9 u64 end;
10 int nid;
11};
12
13struct numa_meminfo {
14 int nr_blks;
15 struct numa_memblk blk[NR_NODE_MEMBLKS];
16};
17
18void __init numa_remove_memblk_from(int idx, struct numa_meminfo *mi);
19int __init numa_cleanup_meminfo(struct numa_meminfo *mi);
20void __init numa_reset_distance(void);
21
22#ifdef CONFIG_NUMA_EMU
23void __init numa_emulation(struct numa_meminfo *numa_meminfo,
24 int numa_dist_cnt);
25#else
26static inline void numa_emulation(struct numa_meminfo *numa_meminfo,
27 int numa_dist_cnt)
28{ }
29#endif
30
31#endif /* __X86_MM_NUMA_INTERNAL_H */
diff --git a/arch/x86/mm/srat_32.c b/arch/x86/mm/srat_32.c
index ae96e7b8051d..48651c6f657d 100644
--- a/arch/x86/mm/srat_32.c
+++ b/arch/x86/mm/srat_32.c
@@ -57,7 +57,7 @@ struct node_memory_chunk_s {
57static struct node_memory_chunk_s __initdata node_memory_chunk[MAXCHUNKS]; 57static struct node_memory_chunk_s __initdata node_memory_chunk[MAXCHUNKS];
58 58
59static int __initdata num_memory_chunks; /* total number of memory chunks */ 59static int __initdata num_memory_chunks; /* total number of memory chunks */
60static u8 __initdata apicid_to_pxm[MAX_APICID]; 60static u8 __initdata apicid_to_pxm[MAX_LOCAL_APIC];
61 61
62int acpi_numa __initdata; 62int acpi_numa __initdata;
63 63
@@ -254,8 +254,8 @@ int __init get_memcfg_from_srat(void)
254 printk(KERN_DEBUG "Number of memory chunks in system = %d\n", 254 printk(KERN_DEBUG "Number of memory chunks in system = %d\n",
255 num_memory_chunks); 255 num_memory_chunks);
256 256
257 for (i = 0; i < MAX_APICID; i++) 257 for (i = 0; i < MAX_LOCAL_APIC; i++)
258 apicid_2_node[i] = pxm_to_node(apicid_to_pxm[i]); 258 set_apicid_to_node(i, pxm_to_node(apicid_to_pxm[i]));
259 259
260 for (j = 0; j < num_memory_chunks; j++){ 260 for (j = 0; j < num_memory_chunks; j++){
261 struct node_memory_chunk_s * chunk = &node_memory_chunk[j]; 261 struct node_memory_chunk_s * chunk = &node_memory_chunk[j];
diff --git a/arch/x86/mm/srat_64.c b/arch/x86/mm/srat_64.c
index 603d285d1daa..8e9d3394f6d4 100644
--- a/arch/x86/mm/srat_64.c
+++ b/arch/x86/mm/srat_64.c
@@ -26,88 +26,34 @@
26 26
27int acpi_numa __initdata; 27int acpi_numa __initdata;
28 28
29static struct acpi_table_slit *acpi_slit;
30
31static nodemask_t nodes_parsed __initdata;
32static nodemask_t cpu_nodes_parsed __initdata;
33static struct bootnode nodes[MAX_NUMNODES] __initdata;
34static struct bootnode nodes_add[MAX_NUMNODES]; 29static struct bootnode nodes_add[MAX_NUMNODES];
35 30
36static int num_node_memblks __initdata;
37static struct bootnode node_memblk_range[NR_NODE_MEMBLKS] __initdata;
38static int memblk_nodeid[NR_NODE_MEMBLKS] __initdata;
39
40static __init int setup_node(int pxm) 31static __init int setup_node(int pxm)
41{ 32{
42 return acpi_map_pxm_to_node(pxm); 33 return acpi_map_pxm_to_node(pxm);
43} 34}
44 35
45static __init int conflicting_memblks(unsigned long start, unsigned long end)
46{
47 int i;
48 for (i = 0; i < num_node_memblks; i++) {
49 struct bootnode *nd = &node_memblk_range[i];
50 if (nd->start == nd->end)
51 continue;
52 if (nd->end > start && nd->start < end)
53 return memblk_nodeid[i];
54 if (nd->end == end && nd->start == start)
55 return memblk_nodeid[i];
56 }
57 return -1;
58}
59
60static __init void cutoff_node(int i, unsigned long start, unsigned long end)
61{
62 struct bootnode *nd = &nodes[i];
63
64 if (nd->start < start) {
65 nd->start = start;
66 if (nd->end < nd->start)
67 nd->start = nd->end;
68 }
69 if (nd->end > end) {
70 nd->end = end;
71 if (nd->start > nd->end)
72 nd->start = nd->end;
73 }
74}
75
76static __init void bad_srat(void) 36static __init void bad_srat(void)
77{ 37{
78 int i;
79 printk(KERN_ERR "SRAT: SRAT not used.\n"); 38 printk(KERN_ERR "SRAT: SRAT not used.\n");
80 acpi_numa = -1; 39 acpi_numa = -1;
81 for (i = 0; i < MAX_LOCAL_APIC; i++) 40 memset(nodes_add, 0, sizeof(nodes_add));
82 apicid_to_node[i] = NUMA_NO_NODE;
83 for (i = 0; i < MAX_NUMNODES; i++) {
84 nodes[i].start = nodes[i].end = 0;
85 nodes_add[i].start = nodes_add[i].end = 0;
86 }
87 remove_all_active_ranges();
88} 41}
89 42
90static __init inline int srat_disabled(void) 43static __init inline int srat_disabled(void)
91{ 44{
92 return numa_off || acpi_numa < 0; 45 return acpi_numa < 0;
93} 46}
94 47
95/* Callback for SLIT parsing */ 48/* Callback for SLIT parsing */
96void __init acpi_numa_slit_init(struct acpi_table_slit *slit) 49void __init acpi_numa_slit_init(struct acpi_table_slit *slit)
97{ 50{
98 unsigned length; 51 int i, j;
99 unsigned long phys;
100
101 length = slit->header.length;
102 phys = memblock_find_in_range(0, max_pfn_mapped<<PAGE_SHIFT, length,
103 PAGE_SIZE);
104
105 if (phys == MEMBLOCK_ERROR)
106 panic(" Can not save slit!\n");
107 52
108 acpi_slit = __va(phys); 53 for (i = 0; i < slit->locality_count; i++)
109 memcpy(acpi_slit, slit, length); 54 for (j = 0; j < slit->locality_count; j++)
110 memblock_x86_reserve_range(phys, phys + length, "ACPI SLIT"); 55 numa_set_distance(pxm_to_node(i), pxm_to_node(j),
56 slit->entry[slit->locality_count * i + j]);
111} 57}
112 58
113/* Callback for Proximity Domain -> x2APIC mapping */ 59/* Callback for Proximity Domain -> x2APIC mapping */
@@ -138,8 +84,8 @@ acpi_numa_x2apic_affinity_init(struct acpi_srat_x2apic_cpu_affinity *pa)
138 printk(KERN_INFO "SRAT: PXM %u -> APIC 0x%04x -> Node %u skipped apicid that is too big\n", pxm, apic_id, node); 84 printk(KERN_INFO "SRAT: PXM %u -> APIC 0x%04x -> Node %u skipped apicid that is too big\n", pxm, apic_id, node);
139 return; 85 return;
140 } 86 }
141 apicid_to_node[apic_id] = node; 87 set_apicid_to_node(apic_id, node);
142 node_set(node, cpu_nodes_parsed); 88 node_set(node, numa_nodes_parsed);
143 acpi_numa = 1; 89 acpi_numa = 1;
144 printk(KERN_INFO "SRAT: PXM %u -> APIC 0x%04x -> Node %u\n", 90 printk(KERN_INFO "SRAT: PXM %u -> APIC 0x%04x -> Node %u\n",
145 pxm, apic_id, node); 91 pxm, apic_id, node);
@@ -178,8 +124,8 @@ acpi_numa_processor_affinity_init(struct acpi_srat_cpu_affinity *pa)
178 return; 124 return;
179 } 125 }
180 126
181 apicid_to_node[apic_id] = node; 127 set_apicid_to_node(apic_id, node);
182 node_set(node, cpu_nodes_parsed); 128 node_set(node, numa_nodes_parsed);
183 acpi_numa = 1; 129 acpi_numa = 1;
184 printk(KERN_INFO "SRAT: PXM %u -> APIC 0x%02x -> Node %u\n", 130 printk(KERN_INFO "SRAT: PXM %u -> APIC 0x%02x -> Node %u\n",
185 pxm, apic_id, node); 131 pxm, apic_id, node);
@@ -241,7 +187,7 @@ update_nodes_add(int node, unsigned long start, unsigned long end)
241 } 187 }
242 188
243 if (changed) { 189 if (changed) {
244 node_set(node, cpu_nodes_parsed); 190 node_set(node, numa_nodes_parsed);
245 printk(KERN_INFO "SRAT: hot plug zone found %Lx - %Lx\n", 191 printk(KERN_INFO "SRAT: hot plug zone found %Lx - %Lx\n",
246 nd->start, nd->end); 192 nd->start, nd->end);
247 } 193 }
@@ -251,10 +197,8 @@ update_nodes_add(int node, unsigned long start, unsigned long end)
251void __init 197void __init
252acpi_numa_memory_affinity_init(struct acpi_srat_mem_affinity *ma) 198acpi_numa_memory_affinity_init(struct acpi_srat_mem_affinity *ma)
253{ 199{
254 struct bootnode *nd, oldnode;
255 unsigned long start, end; 200 unsigned long start, end;
256 int node, pxm; 201 int node, pxm;
257 int i;
258 202
259 if (srat_disabled()) 203 if (srat_disabled())
260 return; 204 return;
@@ -276,300 +220,31 @@ acpi_numa_memory_affinity_init(struct acpi_srat_mem_affinity *ma)
276 bad_srat(); 220 bad_srat();
277 return; 221 return;
278 } 222 }
279 i = conflicting_memblks(start, end); 223
280 if (i == node) { 224 if (numa_add_memblk(node, start, end) < 0) {
281 printk(KERN_WARNING
282 "SRAT: Warning: PXM %d (%lx-%lx) overlaps with itself (%Lx-%Lx)\n",
283 pxm, start, end, nodes[i].start, nodes[i].end);
284 } else if (i >= 0) {
285 printk(KERN_ERR
286 "SRAT: PXM %d (%lx-%lx) overlaps with PXM %d (%Lx-%Lx)\n",
287 pxm, start, end, node_to_pxm(i),
288 nodes[i].start, nodes[i].end);
289 bad_srat(); 225 bad_srat();
290 return; 226 return;
291 } 227 }
292 nd = &nodes[node];
293 oldnode = *nd;
294 if (!node_test_and_set(node, nodes_parsed)) {
295 nd->start = start;
296 nd->end = end;
297 } else {
298 if (start < nd->start)
299 nd->start = start;
300 if (nd->end < end)
301 nd->end = end;
302 }
303 228
304 printk(KERN_INFO "SRAT: Node %u PXM %u %lx-%lx\n", node, pxm, 229 printk(KERN_INFO "SRAT: Node %u PXM %u %lx-%lx\n", node, pxm,
305 start, end); 230 start, end);
306 231
307 if (ma->flags & ACPI_SRAT_MEM_HOT_PLUGGABLE) { 232 if (ma->flags & ACPI_SRAT_MEM_HOT_PLUGGABLE)
308 update_nodes_add(node, start, end); 233 update_nodes_add(node, start, end);
309 /* restore nodes[node] */
310 *nd = oldnode;
311 if ((nd->start | nd->end) == 0)
312 node_clear(node, nodes_parsed);
313 }
314
315 node_memblk_range[num_node_memblks].start = start;
316 node_memblk_range[num_node_memblks].end = end;
317 memblk_nodeid[num_node_memblks] = node;
318 num_node_memblks++;
319}
320
321/* Sanity check to catch more bad SRATs (they are amazingly common).
322 Make sure the PXMs cover all memory. */
323static int __init nodes_cover_memory(const struct bootnode *nodes)
324{
325 int i;
326 unsigned long pxmram, e820ram;
327
328 pxmram = 0;
329 for_each_node_mask(i, nodes_parsed) {
330 unsigned long s = nodes[i].start >> PAGE_SHIFT;
331 unsigned long e = nodes[i].end >> PAGE_SHIFT;
332 pxmram += e - s;
333 pxmram -= __absent_pages_in_range(i, s, e);
334 if ((long)pxmram < 0)
335 pxmram = 0;
336 }
337
338 e820ram = max_pfn - (memblock_x86_hole_size(0, max_pfn<<PAGE_SHIFT)>>PAGE_SHIFT);
339 /* We seem to lose 3 pages somewhere. Allow 1M of slack. */
340 if ((long)(e820ram - pxmram) >= (1<<(20 - PAGE_SHIFT))) {
341 printk(KERN_ERR
342 "SRAT: PXMs only cover %luMB of your %luMB e820 RAM. Not used.\n",
343 (pxmram << PAGE_SHIFT) >> 20,
344 (e820ram << PAGE_SHIFT) >> 20);
345 return 0;
346 }
347 return 1;
348} 234}
349 235
350void __init acpi_numa_arch_fixup(void) {} 236void __init acpi_numa_arch_fixup(void) {}
351 237
352#ifdef CONFIG_NUMA_EMU 238int __init x86_acpi_numa_init(void)
353void __init acpi_get_nodes(struct bootnode *physnodes, unsigned long start,
354 unsigned long end)
355{
356 int i;
357
358 for_each_node_mask(i, nodes_parsed) {
359 cutoff_node(i, start, end);
360 physnodes[i].start = nodes[i].start;
361 physnodes[i].end = nodes[i].end;
362 }
363}
364#endif /* CONFIG_NUMA_EMU */
365
366/* Use the information discovered above to actually set up the nodes. */
367int __init acpi_scan_nodes(unsigned long start, unsigned long end)
368{ 239{
369 int i; 240 int ret;
370
371 if (acpi_numa <= 0)
372 return -1;
373
374 /* First clean up the node list */
375 for (i = 0; i < MAX_NUMNODES; i++)
376 cutoff_node(i, start, end);
377
378 /*
379 * Join together blocks on the same node, holes between
380 * which don't overlap with memory on other nodes.
381 */
382 for (i = 0; i < num_node_memblks; ++i) {
383 int j, k;
384
385 for (j = i + 1; j < num_node_memblks; ++j) {
386 unsigned long start, end;
387
388 if (memblk_nodeid[i] != memblk_nodeid[j])
389 continue;
390 start = min(node_memblk_range[i].end,
391 node_memblk_range[j].end);
392 end = max(node_memblk_range[i].start,
393 node_memblk_range[j].start);
394 for (k = 0; k < num_node_memblks; ++k) {
395 if (memblk_nodeid[i] == memblk_nodeid[k])
396 continue;
397 if (start < node_memblk_range[k].end &&
398 end > node_memblk_range[k].start)
399 break;
400 }
401 if (k < num_node_memblks)
402 continue;
403 start = min(node_memblk_range[i].start,
404 node_memblk_range[j].start);
405 end = max(node_memblk_range[i].end,
406 node_memblk_range[j].end);
407 printk(KERN_INFO "SRAT: Node %d "
408 "[%Lx,%Lx) + [%Lx,%Lx) -> [%lx,%lx)\n",
409 memblk_nodeid[i],
410 node_memblk_range[i].start,
411 node_memblk_range[i].end,
412 node_memblk_range[j].start,
413 node_memblk_range[j].end,
414 start, end);
415 node_memblk_range[i].start = start;
416 node_memblk_range[i].end = end;
417 k = --num_node_memblks - j;
418 memmove(memblk_nodeid + j, memblk_nodeid + j+1,
419 k * sizeof(*memblk_nodeid));
420 memmove(node_memblk_range + j, node_memblk_range + j+1,
421 k * sizeof(*node_memblk_range));
422 --j;
423 }
424 }
425
426 memnode_shift = compute_hash_shift(node_memblk_range, num_node_memblks,
427 memblk_nodeid);
428 if (memnode_shift < 0) {
429 printk(KERN_ERR
430 "SRAT: No NUMA node hash function found. Contact maintainer\n");
431 bad_srat();
432 return -1;
433 }
434
435 for (i = 0; i < num_node_memblks; i++)
436 memblock_x86_register_active_regions(memblk_nodeid[i],
437 node_memblk_range[i].start >> PAGE_SHIFT,
438 node_memblk_range[i].end >> PAGE_SHIFT);
439
440 /* for out of order entries in SRAT */
441 sort_node_map();
442 if (!nodes_cover_memory(nodes)) {
443 bad_srat();
444 return -1;
445 }
446 241
447 /* Account for nodes with cpus and no memory */ 242 ret = acpi_numa_init();
448 nodes_or(node_possible_map, nodes_parsed, cpu_nodes_parsed); 243 if (ret < 0)
449 244 return ret;
450 /* Finally register nodes */ 245 return srat_disabled() ? -EINVAL : 0;
451 for_each_node_mask(i, node_possible_map)
452 setup_node_bootmem(i, nodes[i].start, nodes[i].end);
453 /* Try again in case setup_node_bootmem missed one due
454 to missing bootmem */
455 for_each_node_mask(i, node_possible_map)
456 if (!node_online(i))
457 setup_node_bootmem(i, nodes[i].start, nodes[i].end);
458
459 for (i = 0; i < nr_cpu_ids; i++) {
460 int node = early_cpu_to_node(i);
461
462 if (node == NUMA_NO_NODE)
463 continue;
464 if (!node_online(node))
465 numa_clear_node(i);
466 }
467 numa_init_array();
468 return 0;
469}
470
471#ifdef CONFIG_NUMA_EMU
472static int fake_node_to_pxm_map[MAX_NUMNODES] __initdata = {
473 [0 ... MAX_NUMNODES-1] = PXM_INVAL
474};
475static s16 fake_apicid_to_node[MAX_LOCAL_APIC] __initdata = {
476 [0 ... MAX_LOCAL_APIC-1] = NUMA_NO_NODE
477};
478static int __init find_node_by_addr(unsigned long addr)
479{
480 int ret = NUMA_NO_NODE;
481 int i;
482
483 for_each_node_mask(i, nodes_parsed) {
484 /*
485 * Find the real node that this emulated node appears on. For
486 * the sake of simplicity, we only use a real node's starting
487 * address to determine which emulated node it appears on.
488 */
489 if (addr >= nodes[i].start && addr < nodes[i].end) {
490 ret = i;
491 break;
492 }
493 }
494 return ret;
495} 246}
496 247
497/*
498 * In NUMA emulation, we need to setup proximity domain (_PXM) to node ID
499 * mappings that respect the real ACPI topology but reflect our emulated
500 * environment. For each emulated node, we find which real node it appears on
501 * and create PXM to NID mappings for those fake nodes which mirror that
502 * locality. SLIT will now represent the correct distances between emulated
503 * nodes as a result of the real topology.
504 */
505void __init acpi_fake_nodes(const struct bootnode *fake_nodes, int num_nodes)
506{
507 int i, j;
508
509 for (i = 0; i < num_nodes; i++) {
510 int nid, pxm;
511
512 nid = find_node_by_addr(fake_nodes[i].start);
513 if (nid == NUMA_NO_NODE)
514 continue;
515 pxm = node_to_pxm(nid);
516 if (pxm == PXM_INVAL)
517 continue;
518 fake_node_to_pxm_map[i] = pxm;
519 /*
520 * For each apicid_to_node mapping that exists for this real
521 * node, it must now point to the fake node ID.
522 */
523 for (j = 0; j < MAX_LOCAL_APIC; j++)
524 if (apicid_to_node[j] == nid &&
525 fake_apicid_to_node[j] == NUMA_NO_NODE)
526 fake_apicid_to_node[j] = i;
527 }
528
529 /*
530 * If there are apicid-to-node mappings for physical nodes that do not
531 * have a corresponding emulated node, it should default to a guaranteed
532 * value.
533 */
534 for (i = 0; i < MAX_LOCAL_APIC; i++)
535 if (apicid_to_node[i] != NUMA_NO_NODE &&
536 fake_apicid_to_node[i] == NUMA_NO_NODE)
537 fake_apicid_to_node[i] = 0;
538
539 for (i = 0; i < num_nodes; i++)
540 __acpi_map_pxm_to_node(fake_node_to_pxm_map[i], i);
541 memcpy(apicid_to_node, fake_apicid_to_node, sizeof(apicid_to_node));
542
543 nodes_clear(nodes_parsed);
544 for (i = 0; i < num_nodes; i++)
545 if (fake_nodes[i].start != fake_nodes[i].end)
546 node_set(i, nodes_parsed);
547}
548
549static int null_slit_node_compare(int a, int b)
550{
551 return node_to_pxm(a) == node_to_pxm(b);
552}
553#else
554static int null_slit_node_compare(int a, int b)
555{
556 return a == b;
557}
558#endif /* CONFIG_NUMA_EMU */
559
560int __node_distance(int a, int b)
561{
562 int index;
563
564 if (!acpi_slit)
565 return null_slit_node_compare(a, b) ? LOCAL_DISTANCE :
566 REMOTE_DISTANCE;
567 index = acpi_slit->locality_count * node_to_pxm(a);
568 return acpi_slit->entry[index + node_to_pxm(b)];
569}
570
571EXPORT_SYMBOL(__node_distance);
572
573#if defined(CONFIG_MEMORY_HOTPLUG_SPARSE) || defined(CONFIG_ACPI_HOTPLUG_MEMORY) 248#if defined(CONFIG_MEMORY_HOTPLUG_SPARSE) || defined(CONFIG_ACPI_HOTPLUG_MEMORY)
574int memory_add_physaddr_to_nid(u64 start) 249int memory_add_physaddr_to_nid(u64 start)
575{ 250{
diff --git a/arch/x86/mm/tlb.c b/arch/x86/mm/tlb.c
index 6acc724d5d8f..d6c0418c3e47 100644
--- a/arch/x86/mm/tlb.c
+++ b/arch/x86/mm/tlb.c
@@ -179,12 +179,8 @@ static void flush_tlb_others_ipi(const struct cpumask *cpumask,
179 sender = this_cpu_read(tlb_vector_offset); 179 sender = this_cpu_read(tlb_vector_offset);
180 f = &flush_state[sender]; 180 f = &flush_state[sender];
181 181
182 /* 182 if (nr_cpu_ids > NUM_INVALIDATE_TLB_VECTORS)
183 * Could avoid this lock when 183 raw_spin_lock(&f->tlbstate_lock);
184 * num_online_cpus() <= NUM_INVALIDATE_TLB_VECTORS, but it is
185 * probably not worth checking this for a cache-hot lock.
186 */
187 raw_spin_lock(&f->tlbstate_lock);
188 184
189 f->flush_mm = mm; 185 f->flush_mm = mm;
190 f->flush_va = va; 186 f->flush_va = va;
@@ -202,7 +198,8 @@ static void flush_tlb_others_ipi(const struct cpumask *cpumask,
202 198
203 f->flush_mm = NULL; 199 f->flush_mm = NULL;
204 f->flush_va = 0; 200 f->flush_va = 0;
205 raw_spin_unlock(&f->tlbstate_lock); 201 if (nr_cpu_ids > NUM_INVALIDATE_TLB_VECTORS)
202 raw_spin_unlock(&f->tlbstate_lock);
206} 203}
207 204
208void native_flush_tlb_others(const struct cpumask *cpumask, 205void native_flush_tlb_others(const struct cpumask *cpumask,
@@ -211,11 +208,10 @@ void native_flush_tlb_others(const struct cpumask *cpumask,
211 if (is_uv_system()) { 208 if (is_uv_system()) {
212 unsigned int cpu; 209 unsigned int cpu;
213 210
214 cpu = get_cpu(); 211 cpu = smp_processor_id();
215 cpumask = uv_flush_tlb_others(cpumask, mm, va, cpu); 212 cpumask = uv_flush_tlb_others(cpumask, mm, va, cpu);
216 if (cpumask) 213 if (cpumask)
217 flush_tlb_others_ipi(cpumask, mm, va); 214 flush_tlb_others_ipi(cpumask, mm, va);
218 put_cpu();
219 return; 215 return;
220 } 216 }
221 flush_tlb_others_ipi(cpumask, mm, va); 217 flush_tlb_others_ipi(cpumask, mm, va);
diff --git a/arch/x86/pci/ce4100.c b/arch/x86/pci/ce4100.c
index 9260b3eb18d4..67858be4b52b 100644
--- a/arch/x86/pci/ce4100.c
+++ b/arch/x86/pci/ce4100.c
@@ -255,7 +255,7 @@ int bridge_read(unsigned int devfn, int reg, int len, u32 *value)
255static int ce4100_conf_read(unsigned int seg, unsigned int bus, 255static int ce4100_conf_read(unsigned int seg, unsigned int bus,
256 unsigned int devfn, int reg, int len, u32 *value) 256 unsigned int devfn, int reg, int len, u32 *value)
257{ 257{
258 int i, retval = 1; 258 int i;
259 259
260 if (bus == 1) { 260 if (bus == 1) {
261 for (i = 0; i < ARRAY_SIZE(bus1_fixups); i++) { 261 for (i = 0; i < ARRAY_SIZE(bus1_fixups); i++) {
diff --git a/arch/x86/pci/xen.c b/arch/x86/pci/xen.c
index 25cd4a07d09f..8c4085a95ef1 100644
--- a/arch/x86/pci/xen.c
+++ b/arch/x86/pci/xen.c
@@ -20,7 +20,8 @@
20#include <asm/xen/pci.h> 20#include <asm/xen/pci.h>
21 21
22#ifdef CONFIG_ACPI 22#ifdef CONFIG_ACPI
23static int xen_hvm_register_pirq(u32 gsi, int triggering) 23static int acpi_register_gsi_xen_hvm(struct device *dev, u32 gsi,
24 int trigger, int polarity)
24{ 25{
25 int rc, irq; 26 int rc, irq;
26 struct physdev_map_pirq map_irq; 27 struct physdev_map_pirq map_irq;
@@ -41,7 +42,7 @@ static int xen_hvm_register_pirq(u32 gsi, int triggering)
41 return -1; 42 return -1;
42 } 43 }
43 44
44 if (triggering == ACPI_EDGE_SENSITIVE) { 45 if (trigger == ACPI_EDGE_SENSITIVE) {
45 shareable = 0; 46 shareable = 0;
46 name = "ioapic-edge"; 47 name = "ioapic-edge";
47 } else { 48 } else {
@@ -55,12 +56,6 @@ static int xen_hvm_register_pirq(u32 gsi, int triggering)
55 56
56 return irq; 57 return irq;
57} 58}
58
59static int acpi_register_gsi_xen_hvm(struct device *dev, u32 gsi,
60 int trigger, int polarity)
61{
62 return xen_hvm_register_pirq(gsi, trigger);
63}
64#endif 59#endif
65 60
66#if defined(CONFIG_PCI_MSI) 61#if defined(CONFIG_PCI_MSI)
@@ -91,7 +86,7 @@ static void xen_msi_compose_msg(struct pci_dev *pdev, unsigned int pirq,
91 86
92static int xen_hvm_setup_msi_irqs(struct pci_dev *dev, int nvec, int type) 87static int xen_hvm_setup_msi_irqs(struct pci_dev *dev, int nvec, int type)
93{ 88{
94 int irq, pirq, ret = 0; 89 int irq, pirq;
95 struct msi_desc *msidesc; 90 struct msi_desc *msidesc;
96 struct msi_msg msg; 91 struct msi_msg msg;
97 92
@@ -99,39 +94,32 @@ static int xen_hvm_setup_msi_irqs(struct pci_dev *dev, int nvec, int type)
99 __read_msi_msg(msidesc, &msg); 94 __read_msi_msg(msidesc, &msg);
100 pirq = MSI_ADDR_EXT_DEST_ID(msg.address_hi) | 95 pirq = MSI_ADDR_EXT_DEST_ID(msg.address_hi) |
101 ((msg.address_lo >> MSI_ADDR_DEST_ID_SHIFT) & 0xff); 96 ((msg.address_lo >> MSI_ADDR_DEST_ID_SHIFT) & 0xff);
102 if (xen_irq_from_pirq(pirq) >= 0 && msg.data == XEN_PIRQ_MSI_DATA) { 97 if (msg.data != XEN_PIRQ_MSI_DATA ||
103 xen_allocate_pirq_msi((type == PCI_CAP_ID_MSIX) ? 98 xen_irq_from_pirq(pirq) < 0) {
104 "msi-x" : "msi", &irq, &pirq, XEN_ALLOC_IRQ); 99 pirq = xen_allocate_pirq_msi(dev, msidesc);
105 if (irq < 0) 100 if (pirq < 0)
106 goto error; 101 goto error;
107 ret = set_irq_msi(irq, msidesc); 102 xen_msi_compose_msg(dev, pirq, &msg);
108 if (ret < 0) 103 __write_msi_msg(msidesc, &msg);
109 goto error_while; 104 dev_dbg(&dev->dev, "xen: msi bound to pirq=%d\n", pirq);
110 printk(KERN_DEBUG "xen: msi already setup: msi --> irq=%d" 105 } else {
111 " pirq=%d\n", irq, pirq); 106 dev_dbg(&dev->dev,
112 return 0; 107 "xen: msi already bound to pirq=%d\n", pirq);
113 } 108 }
114 xen_allocate_pirq_msi((type == PCI_CAP_ID_MSIX) ? 109 irq = xen_bind_pirq_msi_to_irq(dev, msidesc, pirq, 0,
115 "msi-x" : "msi", &irq, &pirq, (XEN_ALLOC_IRQ | XEN_ALLOC_PIRQ)); 110 (type == PCI_CAP_ID_MSIX) ?
116 if (irq < 0 || pirq < 0) 111 "msi-x" : "msi");
112 if (irq < 0)
117 goto error; 113 goto error;
118 printk(KERN_DEBUG "xen: msi --> irq=%d, pirq=%d\n", irq, pirq); 114 dev_dbg(&dev->dev,
119 xen_msi_compose_msg(dev, pirq, &msg); 115 "xen: msi --> pirq=%d --> irq=%d\n", pirq, irq);
120 ret = set_irq_msi(irq, msidesc);
121 if (ret < 0)
122 goto error_while;
123 write_msi_msg(irq, &msg);
124 } 116 }
125 return 0; 117 return 0;
126 118
127error_while:
128 unbind_from_irqhandler(irq, NULL);
129error: 119error:
130 if (ret == -ENODEV) 120 dev_err(&dev->dev,
131 dev_err(&dev->dev, "Xen PCI frontend has not registered" \ 121 "Xen PCI frontend has not registered MSI/MSI-X support!\n");
132 " MSI/MSI-X support!\n"); 122 return -ENODEV;
133
134 return ret;
135} 123}
136 124
137/* 125/*
@@ -150,35 +138,26 @@ static int xen_setup_msi_irqs(struct pci_dev *dev, int nvec, int type)
150 return -ENOMEM; 138 return -ENOMEM;
151 139
152 if (type == PCI_CAP_ID_MSIX) 140 if (type == PCI_CAP_ID_MSIX)
153 ret = xen_pci_frontend_enable_msix(dev, &v, nvec); 141 ret = xen_pci_frontend_enable_msix(dev, v, nvec);
154 else 142 else
155 ret = xen_pci_frontend_enable_msi(dev, &v); 143 ret = xen_pci_frontend_enable_msi(dev, v);
156 if (ret) 144 if (ret)
157 goto error; 145 goto error;
158 i = 0; 146 i = 0;
159 list_for_each_entry(msidesc, &dev->msi_list, list) { 147 list_for_each_entry(msidesc, &dev->msi_list, list) {
160 irq = xen_allocate_pirq(v[i], 0, /* not sharable */ 148 irq = xen_bind_pirq_msi_to_irq(dev, msidesc, v[i], 0,
161 (type == PCI_CAP_ID_MSIX) ? 149 (type == PCI_CAP_ID_MSIX) ?
162 "pcifront-msi-x" : "pcifront-msi"); 150 "pcifront-msi-x" :
163 if (irq < 0) { 151 "pcifront-msi");
164 ret = -1; 152 if (irq < 0)
165 goto free; 153 goto free;
166 }
167
168 ret = set_irq_msi(irq, msidesc);
169 if (ret)
170 goto error_while;
171 i++; 154 i++;
172 } 155 }
173 kfree(v); 156 kfree(v);
174 return 0; 157 return 0;
175 158
176error_while:
177 unbind_from_irqhandler(irq, NULL);
178error: 159error:
179 if (ret == -ENODEV) 160 dev_err(&dev->dev, "Xen PCI frontend has not registered MSI/MSI-X support!\n");
180 dev_err(&dev->dev, "Xen PCI frontend has not registered" \
181 " MSI/MSI-X support!\n");
182free: 161free:
183 kfree(v); 162 kfree(v);
184 return ret; 163 return ret;
@@ -193,6 +172,9 @@ static void xen_teardown_msi_irqs(struct pci_dev *dev)
193 xen_pci_frontend_disable_msix(dev); 172 xen_pci_frontend_disable_msix(dev);
194 else 173 else
195 xen_pci_frontend_disable_msi(dev); 174 xen_pci_frontend_disable_msi(dev);
175
176 /* Free the IRQ's and the msidesc using the generic code. */
177 default_teardown_msi_irqs(dev);
196} 178}
197 179
198static void xen_teardown_msi_irq(unsigned int irq) 180static void xen_teardown_msi_irq(unsigned int irq)
@@ -200,47 +182,82 @@ static void xen_teardown_msi_irq(unsigned int irq)
200 xen_destroy_irq(irq); 182 xen_destroy_irq(irq);
201} 183}
202 184
185#ifdef CONFIG_XEN_DOM0
203static int xen_initdom_setup_msi_irqs(struct pci_dev *dev, int nvec, int type) 186static int xen_initdom_setup_msi_irqs(struct pci_dev *dev, int nvec, int type)
204{ 187{
205 int irq, ret; 188 int ret = 0;
206 struct msi_desc *msidesc; 189 struct msi_desc *msidesc;
207 190
208 list_for_each_entry(msidesc, &dev->msi_list, list) { 191 list_for_each_entry(msidesc, &dev->msi_list, list) {
209 irq = xen_create_msi_irq(dev, msidesc, type); 192 struct physdev_map_pirq map_irq;
210 if (irq < 0)
211 return -1;
212 193
213 ret = set_irq_msi(irq, msidesc); 194 memset(&map_irq, 0, sizeof(map_irq));
214 if (ret) 195 map_irq.domid = DOMID_SELF;
215 goto error; 196 map_irq.type = MAP_PIRQ_TYPE_MSI;
216 } 197 map_irq.index = -1;
217 return 0; 198 map_irq.pirq = -1;
199 map_irq.bus = dev->bus->number;
200 map_irq.devfn = dev->devfn;
218 201
219error: 202 if (type == PCI_CAP_ID_MSIX) {
220 xen_destroy_irq(irq); 203 int pos;
204 u32 table_offset, bir;
205
206 pos = pci_find_capability(dev, PCI_CAP_ID_MSIX);
207
208 pci_read_config_dword(dev, pos + PCI_MSIX_TABLE,
209 &table_offset);
210 bir = (u8)(table_offset & PCI_MSIX_FLAGS_BIRMASK);
211
212 map_irq.table_base = pci_resource_start(dev, bir);
213 map_irq.entry_nr = msidesc->msi_attrib.entry_nr;
214 }
215
216 ret = HYPERVISOR_physdev_op(PHYSDEVOP_map_pirq, &map_irq);
217 if (ret) {
218 dev_warn(&dev->dev, "xen map irq failed %d\n", ret);
219 goto out;
220 }
221
222 ret = xen_bind_pirq_msi_to_irq(dev, msidesc,
223 map_irq.pirq, map_irq.index,
224 (type == PCI_CAP_ID_MSIX) ?
225 "msi-x" : "msi");
226 if (ret < 0)
227 goto out;
228 }
229 ret = 0;
230out:
221 return ret; 231 return ret;
222} 232}
223#endif 233#endif
234#endif
224 235
225static int xen_pcifront_enable_irq(struct pci_dev *dev) 236static int xen_pcifront_enable_irq(struct pci_dev *dev)
226{ 237{
227 int rc; 238 int rc;
228 int share = 1; 239 int share = 1;
240 u8 gsi;
229 241
230 dev_info(&dev->dev, "Xen PCI enabling IRQ: %d\n", dev->irq); 242 rc = pci_read_config_byte(dev, PCI_INTERRUPT_LINE, &gsi);
231 243 if (rc < 0) {
232 if (dev->irq < 0) 244 dev_warn(&dev->dev, "Xen PCI: failed to read interrupt line: %d\n",
233 return -EINVAL; 245 rc);
246 return rc;
247 }
234 248
235 if (dev->irq < NR_IRQS_LEGACY) 249 if (gsi < NR_IRQS_LEGACY)
236 share = 0; 250 share = 0;
237 251
238 rc = xen_allocate_pirq(dev->irq, share, "pcifront"); 252 rc = xen_allocate_pirq(gsi, share, "pcifront");
239 if (rc < 0) { 253 if (rc < 0) {
240 dev_warn(&dev->dev, "Xen PCI IRQ: %d, failed to register:%d\n", 254 dev_warn(&dev->dev, "Xen PCI: failed to register GSI%d: %d\n",
241 dev->irq, rc); 255 gsi, rc);
242 return rc; 256 return rc;
243 } 257 }
258
259 dev->irq = rc;
260 dev_info(&dev->dev, "Xen PCI mapped GSI%d to IRQ%d\n", gsi, dev->irq);
244 return 0; 261 return 0;
245} 262}
246 263
diff --git a/arch/x86/platform/ce4100/ce4100.c b/arch/x86/platform/ce4100/ce4100.c
index cd6f184c3b3f..28071bb31db7 100644
--- a/arch/x86/platform/ce4100/ce4100.c
+++ b/arch/x86/platform/ce4100/ce4100.c
@@ -16,21 +16,19 @@
16#include <linux/serial_8250.h> 16#include <linux/serial_8250.h>
17 17
18#include <asm/ce4100.h> 18#include <asm/ce4100.h>
19#include <asm/prom.h>
19#include <asm/setup.h> 20#include <asm/setup.h>
21#include <asm/i8259.h>
20#include <asm/io.h> 22#include <asm/io.h>
23#include <asm/io_apic.h>
21 24
22static int ce4100_i8042_detect(void) 25static int ce4100_i8042_detect(void)
23{ 26{
24 return 0; 27 return 0;
25} 28}
26 29
27static void __init sdv_find_smp_config(void)
28{
29}
30
31#ifdef CONFIG_SERIAL_8250 30#ifdef CONFIG_SERIAL_8250
32 31
33
34static unsigned int mem_serial_in(struct uart_port *p, int offset) 32static unsigned int mem_serial_in(struct uart_port *p, int offset)
35{ 33{
36 offset = offset << p->regshift; 34 offset = offset << p->regshift;
@@ -119,6 +117,15 @@ static void __init sdv_arch_setup(void)
119 sdv_serial_fixup(); 117 sdv_serial_fixup();
120} 118}
121 119
120#ifdef CONFIG_X86_IO_APIC
121static void __cpuinit sdv_pci_init(void)
122{
123 x86_of_pci_init();
124 /* We can't set this earlier, because we need to calibrate the timer */
125 legacy_pic = &null_legacy_pic;
126}
127#endif
128
122/* 129/*
123 * CE4100 specific x86_init function overrides and early setup 130 * CE4100 specific x86_init function overrides and early setup
124 * calls. 131 * calls.
@@ -129,6 +136,11 @@ void __init x86_ce4100_early_setup(void)
129 x86_platform.i8042_detect = ce4100_i8042_detect; 136 x86_platform.i8042_detect = ce4100_i8042_detect;
130 x86_init.resources.probe_roms = x86_init_noop; 137 x86_init.resources.probe_roms = x86_init_noop;
131 x86_init.mpparse.get_smp_config = x86_init_uint_noop; 138 x86_init.mpparse.get_smp_config = x86_init_uint_noop;
132 x86_init.mpparse.find_smp_config = sdv_find_smp_config; 139 x86_init.mpparse.find_smp_config = x86_init_noop;
133 x86_init.pci.init = ce4100_pci_init; 140 x86_init.pci.init = ce4100_pci_init;
141
142#ifdef CONFIG_X86_IO_APIC
143 x86_init.pci.init_irq = sdv_pci_init;
144 x86_init.mpparse.setup_ioapic_ids = setup_ioapic_ids_from_mpc_nocheck;
145#endif
134} 146}
diff --git a/arch/x86/platform/ce4100/falconfalls.dts b/arch/x86/platform/ce4100/falconfalls.dts
new file mode 100644
index 000000000000..dc701ea58546
--- /dev/null
+++ b/arch/x86/platform/ce4100/falconfalls.dts
@@ -0,0 +1,428 @@
1/*
2 * CE4100 on Falcon Falls
3 *
4 * (c) Copyright 2010 Intel Corporation
5 *
6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms of the GNU General Public License as published by the
8 * Free Software Foundation; version 2 of the License.
9 */
10/dts-v1/;
11/ {
12 model = "intel,falconfalls";
13 compatible = "intel,falconfalls";
14 #address-cells = <1>;
15 #size-cells = <1>;
16
17 cpus {
18 #address-cells = <1>;
19 #size-cells = <0>;
20
21 cpu@0 {
22 device_type = "cpu";
23 compatible = "intel,ce4100";
24 reg = <0>;
25 lapic = <&lapic0>;
26 };
27 };
28
29 soc@0 {
30 #address-cells = <1>;
31 #size-cells = <1>;
32 compatible = "intel,ce4100-cp";
33 ranges;
34
35 ioapic1: interrupt-controller@fec00000 {
36 #interrupt-cells = <2>;
37 compatible = "intel,ce4100-ioapic";
38 interrupt-controller;
39 reg = <0xfec00000 0x1000>;
40 };
41
42 timer@fed00000 {
43 compatible = "intel,ce4100-hpet";
44 reg = <0xfed00000 0x200>;
45 };
46
47 lapic0: interrupt-controller@fee00000 {
48 compatible = "intel,ce4100-lapic";
49 reg = <0xfee00000 0x1000>;
50 };
51
52 pci@3fc {
53 #address-cells = <3>;
54 #size-cells = <2>;
55 compatible = "intel,ce4100-pci", "pci";
56 device_type = "pci";
57 bus-range = <0 0>;
58 ranges = <0x2000000 0 0xbffff000 0xbffff000 0 0x1000
59 0x2000000 0 0xdffe0000 0xdffe0000 0 0x1000
60 0x0000000 0 0x0 0x0 0 0x100>;
61
62 /* Secondary IO-APIC */
63 ioapic2: interrupt-controller@0,1 {
64 #interrupt-cells = <2>;
65 compatible = "intel,ce4100-ioapic";
66 interrupt-controller;
67 reg = <0x100 0x0 0x0 0x0 0x0>;
68 assigned-addresses = <0x02000000 0x0 0xbffff000 0x0 0x1000>;
69 };
70
71 pci@1,0 {
72 #address-cells = <3>;
73 #size-cells = <2>;
74 compatible = "intel,ce4100-pci", "pci";
75 device_type = "pci";
76 bus-range = <1 1>;
77 ranges = <0x2000000 0 0xdffe0000 0x2000000 0 0xdffe0000 0 0x1000>;
78
79 interrupt-parent = <&ioapic2>;
80
81 display@2,0 {
82 compatible = "pci8086,2e5b.2",
83 "pci8086,2e5b",
84 "pciclass038000",
85 "pciclass0380";
86
87 reg = <0x11000 0x0 0x0 0x0 0x0>;
88 interrupts = <0 1>;
89 };
90
91 multimedia@3,0 {
92 compatible = "pci8086,2e5c.2",
93 "pci8086,2e5c",
94 "pciclass048000",
95 "pciclass0480";
96
97 reg = <0x11800 0x0 0x0 0x0 0x0>;
98 interrupts = <2 1>;
99 };
100
101 multimedia@4,0 {
102 compatible = "pci8086,2e5d.2",
103 "pci8086,2e5d",
104 "pciclass048000",
105 "pciclass0480";
106
107 reg = <0x12000 0x0 0x0 0x0 0x0>;
108 interrupts = <4 1>;
109 };
110
111 multimedia@4,1 {
112 compatible = "pci8086,2e5e.2",
113 "pci8086,2e5e",
114 "pciclass048000",
115 "pciclass0480";
116
117 reg = <0x12100 0x0 0x0 0x0 0x0>;
118 interrupts = <5 1>;
119 };
120
121 sound@6,0 {
122 compatible = "pci8086,2e5f.2",
123 "pci8086,2e5f",
124 "pciclass040100",
125 "pciclass0401";
126
127 reg = <0x13000 0x0 0x0 0x0 0x0>;
128 interrupts = <6 1>;
129 };
130
131 sound@6,1 {
132 compatible = "pci8086,2e5f.2",
133 "pci8086,2e5f",
134 "pciclass040100",
135 "pciclass0401";
136
137 reg = <0x13100 0x0 0x0 0x0 0x0>;
138 interrupts = <7 1>;
139 };
140
141 sound@6,2 {
142 compatible = "pci8086,2e60.2",
143 "pci8086,2e60",
144 "pciclass040100",
145 "pciclass0401";
146
147 reg = <0x13200 0x0 0x0 0x0 0x0>;
148 interrupts = <8 1>;
149 };
150
151 display@8,0 {
152 compatible = "pci8086,2e61.2",
153 "pci8086,2e61",
154 "pciclass038000",
155 "pciclass0380";
156
157 reg = <0x14000 0x0 0x0 0x0 0x0>;
158 interrupts = <9 1>;
159 };
160
161 display@8,1 {
162 compatible = "pci8086,2e62.2",
163 "pci8086,2e62",
164 "pciclass038000",
165 "pciclass0380";
166
167 reg = <0x14100 0x0 0x0 0x0 0x0>;
168 interrupts = <10 1>;
169 };
170
171 multimedia@8,2 {
172 compatible = "pci8086,2e63.2",
173 "pci8086,2e63",
174 "pciclass048000",
175 "pciclass0480";
176
177 reg = <0x14200 0x0 0x0 0x0 0x0>;
178 interrupts = <11 1>;
179 };
180
181 entertainment-encryption@9,0 {
182 compatible = "pci8086,2e64.2",
183 "pci8086,2e64",
184 "pciclass101000",
185 "pciclass1010";
186
187 reg = <0x14800 0x0 0x0 0x0 0x0>;
188 interrupts = <12 1>;
189 };
190
191 localbus@a,0 {
192 compatible = "pci8086,2e65.2",
193 "pci8086,2e65",
194 "pciclassff0000",
195 "pciclassff00";
196
197 reg = <0x15000 0x0 0x0 0x0 0x0>;
198 };
199
200 serial@b,0 {
201 compatible = "pci8086,2e66.2",
202 "pci8086,2e66",
203 "pciclass070003",
204 "pciclass0700";
205
206 reg = <0x15800 0x0 0x0 0x0 0x0>;
207 interrupts = <14 1>;
208 };
209
210 gpio@b,1 {
211 compatible = "pci8086,2e67.2",
212 "pci8086,2e67",
213 "pciclassff0000",
214 "pciclassff00";
215
216 #gpio-cells = <2>;
217 reg = <0x15900 0x0 0x0 0x0 0x0>;
218 interrupts = <15 1>;
219 gpio-controller;
220 };
221
222 i2c-controller@b,2 {
223 #address-cells = <2>;
224 #size-cells = <1>;
225 compatible = "pci8086,2e68.2",
226 "pci8086,2e68",
227 "pciclass,ff0000",
228 "pciclass,ff00";
229
230 reg = <0x15a00 0x0 0x0 0x0 0x0>;
231 interrupts = <16 1>;
232 ranges = <0 0 0x02000000 0 0xdffe0500 0x100
233 1 0 0x02000000 0 0xdffe0600 0x100
234 2 0 0x02000000 0 0xdffe0700 0x100>;
235
236 i2c@0 {
237 #address-cells = <1>;
238 #size-cells = <0>;
239 compatible = "intel,ce4100-i2c-controller";
240 reg = <0 0 0x100>;
241 };
242
243 i2c@1 {
244 #address-cells = <1>;
245 #size-cells = <0>;
246 compatible = "intel,ce4100-i2c-controller";
247 reg = <1 0 0x100>;
248
249 gpio@26 {
250 #gpio-cells = <2>;
251 compatible = "ti,pcf8575";
252 reg = <0x26>;
253 gpio-controller;
254 };
255 };
256
257 i2c@2 {
258 #address-cells = <1>;
259 #size-cells = <0>;
260 compatible = "intel,ce4100-i2c-controller";
261 reg = <2 0 0x100>;
262
263 gpio@26 {
264 #gpio-cells = <2>;
265 compatible = "ti,pcf8575";
266 reg = <0x26>;
267 gpio-controller;
268 };
269 };
270 };
271
272 smard-card@b,3 {
273 compatible = "pci8086,2e69.2",
274 "pci8086,2e69",
275 "pciclass070500",
276 "pciclass0705";
277
278 reg = <0x15b00 0x0 0x0 0x0 0x0>;
279 interrupts = <15 1>;
280 };
281
282 spi-controller@b,4 {
283 #address-cells = <1>;
284 #size-cells = <0>;
285 compatible =
286 "pci8086,2e6a.2",
287 "pci8086,2e6a",
288 "pciclass,ff0000",
289 "pciclass,ff00";
290
291 reg = <0x15c00 0x0 0x0 0x0 0x0>;
292 interrupts = <15 1>;
293
294 dac@0 {
295 compatible = "ti,pcm1755";
296 reg = <0>;
297 spi-max-frequency = <115200>;
298 };
299
300 dac@1 {
301 compatible = "ti,pcm1609a";
302 reg = <1>;
303 spi-max-frequency = <115200>;
304 };
305
306 eeprom@2 {
307 compatible = "atmel,at93c46";
308 reg = <2>;
309 spi-max-frequency = <115200>;
310 };
311 };
312
313 multimedia@b,7 {
314 compatible = "pci8086,2e6d.2",
315 "pci8086,2e6d",
316 "pciclassff0000",
317 "pciclassff00";
318
319 reg = <0x15f00 0x0 0x0 0x0 0x0>;
320 };
321
322 ethernet@c,0 {
323 compatible = "pci8086,2e6e.2",
324 "pci8086,2e6e",
325 "pciclass020000",
326 "pciclass0200";
327
328 reg = <0x16000 0x0 0x0 0x0 0x0>;
329 interrupts = <21 1>;
330 };
331
332 clock@c,1 {
333 compatible = "pci8086,2e6f.2",
334 "pci8086,2e6f",
335 "pciclassff0000",
336 "pciclassff00";
337
338 reg = <0x16100 0x0 0x0 0x0 0x0>;
339 interrupts = <3 1>;
340 };
341
342 usb@d,0 {
343 compatible = "pci8086,2e70.2",
344 "pci8086,2e70",
345 "pciclass0c0320",
346 "pciclass0c03";
347
348 reg = <0x16800 0x0 0x0 0x0 0x0>;
349 interrupts = <22 3>;
350 };
351
352 usb@d,1 {
353 compatible = "pci8086,2e70.2",
354 "pci8086,2e70",
355 "pciclass0c0320",
356 "pciclass0c03";
357
358 reg = <0x16900 0x0 0x0 0x0 0x0>;
359 interrupts = <22 3>;
360 };
361
362 sata@e,0 {
363 compatible = "pci8086,2e71.0",
364 "pci8086,2e71",
365 "pciclass010601",
366 "pciclass0106";
367
368 reg = <0x17000 0x0 0x0 0x0 0x0>;
369 interrupts = <23 3>;
370 };
371
372 flash@f,0 {
373 compatible = "pci8086,701.1",
374 "pci8086,701",
375 "pciclass050100",
376 "pciclass0501";
377
378 reg = <0x17800 0x0 0x0 0x0 0x0>;
379 interrupts = <13 1>;
380 };
381
382 entertainment-encryption@10,0 {
383 compatible = "pci8086,702.1",
384 "pci8086,702",
385 "pciclass101000",
386 "pciclass1010";
387
388 reg = <0x18000 0x0 0x0 0x0 0x0>;
389 };
390
391 co-processor@11,0 {
392 compatible = "pci8086,703.1",
393 "pci8086,703",
394 "pciclass0b4000",
395 "pciclass0b40";
396
397 reg = <0x18800 0x0 0x0 0x0 0x0>;
398 interrupts = <1 1>;
399 };
400
401 multimedia@12,0 {
402 compatible = "pci8086,704.0",
403 "pci8086,704",
404 "pciclass048000",
405 "pciclass0480";
406
407 reg = <0x19000 0x0 0x0 0x0 0x0>;
408 };
409 };
410
411 isa@1f,0 {
412 #address-cells = <2>;
413 #size-cells = <1>;
414 compatible = "isa";
415 ranges = <1 0 0 0 0 0x100>;
416
417 rtc@70 {
418 compatible = "intel,ce4100-rtc", "motorola,mc146818";
419 interrupts = <8 3>;
420 interrupt-parent = <&ioapic1>;
421 ctrl-reg = <2>;
422 freq-reg = <0x26>;
423 reg = <1 0x70 2>;
424 };
425 };
426 };
427 };
428};
diff --git a/arch/x86/platform/mrst/mrst.c b/arch/x86/platform/mrst/mrst.c
index ea6529e93c6f..5c0207bf959b 100644
--- a/arch/x86/platform/mrst/mrst.c
+++ b/arch/x86/platform/mrst/mrst.c
@@ -31,6 +31,7 @@
31#include <asm/apic.h> 31#include <asm/apic.h>
32#include <asm/io_apic.h> 32#include <asm/io_apic.h>
33#include <asm/mrst.h> 33#include <asm/mrst.h>
34#include <asm/mrst-vrtc.h>
34#include <asm/io.h> 35#include <asm/io.h>
35#include <asm/i8259.h> 36#include <asm/i8259.h>
36#include <asm/intel_scu_ipc.h> 37#include <asm/intel_scu_ipc.h>
@@ -268,6 +269,7 @@ void __init x86_mrst_early_setup(void)
268 269
269 x86_platform.calibrate_tsc = mrst_calibrate_tsc; 270 x86_platform.calibrate_tsc = mrst_calibrate_tsc;
270 x86_platform.i8042_detect = mrst_i8042_detect; 271 x86_platform.i8042_detect = mrst_i8042_detect;
272 x86_init.timers.wallclock_init = mrst_rtc_init;
271 x86_init.pci.init = pci_mrst_init; 273 x86_init.pci.init = pci_mrst_init;
272 x86_init.pci.fixup_irqs = x86_init_noop; 274 x86_init.pci.fixup_irqs = x86_init_noop;
273 275
diff --git a/arch/x86/platform/mrst/vrtc.c b/arch/x86/platform/mrst/vrtc.c
index 32cd7edd71a0..04cf645feb92 100644
--- a/arch/x86/platform/mrst/vrtc.c
+++ b/arch/x86/platform/mrst/vrtc.c
@@ -100,22 +100,14 @@ int vrtc_set_mmss(unsigned long nowtime)
100 100
101void __init mrst_rtc_init(void) 101void __init mrst_rtc_init(void)
102{ 102{
103 unsigned long rtc_paddr; 103 unsigned long vrtc_paddr = sfi_mrtc_array[0].phys_addr;
104 void __iomem *virt_base;
105 104
106 sfi_table_parse(SFI_SIG_MRTC, NULL, NULL, sfi_parse_mrtc); 105 sfi_table_parse(SFI_SIG_MRTC, NULL, NULL, sfi_parse_mrtc);
107 if (!sfi_mrtc_num) 106 if (!sfi_mrtc_num || !vrtc_paddr)
108 return; 107 return;
109 108
110 rtc_paddr = sfi_mrtc_array[0].phys_addr; 109 vrtc_virt_base = (void __iomem *)set_fixmap_offset_nocache(FIX_LNW_VRTC,
111 110 vrtc_paddr);
112 /* vRTC's register address may not be page aligned */
113 set_fixmap_nocache(FIX_LNW_VRTC, rtc_paddr);
114
115 virt_base = (void __iomem *)__fix_to_virt(FIX_LNW_VRTC);
116 virt_base += rtc_paddr & ~PAGE_MASK;
117 vrtc_virt_base = virt_base;
118
119 x86_platform.get_wallclock = vrtc_get_time; 111 x86_platform.get_wallclock = vrtc_get_time;
120 x86_platform.set_wallclock = vrtc_set_mmss; 112 x86_platform.set_wallclock = vrtc_set_mmss;
121} 113}
diff --git a/arch/x86/platform/olpc/Makefile b/arch/x86/platform/olpc/Makefile
index e797428b163b..c2a8cab65e5d 100644
--- a/arch/x86/platform/olpc/Makefile
+++ b/arch/x86/platform/olpc/Makefile
@@ -1,4 +1,4 @@
1obj-$(CONFIG_OLPC) += olpc.o 1obj-$(CONFIG_OLPC) += olpc.o
2obj-$(CONFIG_OLPC_XO1) += olpc-xo1.o 2obj-$(CONFIG_OLPC_XO1) += olpc-xo1.o
3obj-$(CONFIG_OLPC_OPENFIRMWARE) += olpc_ofw.o 3obj-$(CONFIG_OLPC) += olpc_ofw.o
4obj-$(CONFIG_OLPC_OPENFIRMWARE_DT) += olpc_dt.o 4obj-$(CONFIG_OF_PROMTREE) += olpc_dt.o
diff --git a/arch/x86/platform/uv/uv_irq.c b/arch/x86/platform/uv/uv_irq.c
index 7b24460917d5..374a05d8ad22 100644
--- a/arch/x86/platform/uv/uv_irq.c
+++ b/arch/x86/platform/uv/uv_irq.c
@@ -131,7 +131,7 @@ arch_enable_uv_irq(char *irq_name, unsigned int irq, int cpu, int mmr_blade,
131 unsigned long mmr_offset, int limit) 131 unsigned long mmr_offset, int limit)
132{ 132{
133 const struct cpumask *eligible_cpu = cpumask_of(cpu); 133 const struct cpumask *eligible_cpu = cpumask_of(cpu);
134 struct irq_cfg *cfg = get_irq_chip_data(irq); 134 struct irq_cfg *cfg = irq_get_chip_data(irq);
135 unsigned long mmr_value; 135 unsigned long mmr_value;
136 struct uv_IO_APIC_route_entry *entry; 136 struct uv_IO_APIC_route_entry *entry;
137 int mmr_pnode, err; 137 int mmr_pnode, err;
@@ -148,7 +148,7 @@ arch_enable_uv_irq(char *irq_name, unsigned int irq, int cpu, int mmr_blade,
148 else 148 else
149 irq_set_status_flags(irq, IRQ_MOVE_PCNTXT); 149 irq_set_status_flags(irq, IRQ_MOVE_PCNTXT);
150 150
151 set_irq_chip_and_handler_name(irq, &uv_irq_chip, handle_percpu_irq, 151 irq_set_chip_and_handler_name(irq, &uv_irq_chip, handle_percpu_irq,
152 irq_name); 152 irq_name);
153 153
154 mmr_value = 0; 154 mmr_value = 0;
diff --git a/arch/x86/platform/visws/visws_quirks.c b/arch/x86/platform/visws/visws_quirks.c
index 632037671746..fe4cf8294878 100644
--- a/arch/x86/platform/visws/visws_quirks.c
+++ b/arch/x86/platform/visws/visws_quirks.c
@@ -569,11 +569,13 @@ out_unlock:
569static struct irqaction master_action = { 569static struct irqaction master_action = {
570 .handler = piix4_master_intr, 570 .handler = piix4_master_intr,
571 .name = "PIIX4-8259", 571 .name = "PIIX4-8259",
572 .flags = IRQF_NO_THREAD,
572}; 573};
573 574
574static struct irqaction cascade_action = { 575static struct irqaction cascade_action = {
575 .handler = no_action, 576 .handler = no_action,
576 .name = "cascade", 577 .name = "cascade",
578 .flags = IRQF_NO_THREAD,
577}; 579};
578 580
579static inline void set_piix4_virtual_irq_type(void) 581static inline void set_piix4_virtual_irq_type(void)
@@ -606,7 +608,7 @@ static void __init visws_pre_intr_init(void)
606 chip = &cobalt_irq_type; 608 chip = &cobalt_irq_type;
607 609
608 if (chip) 610 if (chip)
609 set_irq_chip(i, chip); 611 irq_set_chip(i, chip);
610 } 612 }
611 613
612 setup_irq(CO_IRQ_8259, &master_action); 614 setup_irq(CO_IRQ_8259, &master_action);
diff --git a/arch/x86/xen/Kconfig b/arch/x86/xen/Kconfig
index 5b54892e4bc3..e4343fe488ed 100644
--- a/arch/x86/xen/Kconfig
+++ b/arch/x86/xen/Kconfig
@@ -48,3 +48,11 @@ config XEN_DEBUG_FS
48 help 48 help
49 Enable statistics output and various tuning options in debugfs. 49 Enable statistics output and various tuning options in debugfs.
50 Enabling this option may incur a significant performance overhead. 50 Enabling this option may incur a significant performance overhead.
51
52config XEN_DEBUG
53 bool "Enable Xen debug checks"
54 depends on XEN
55 default n
56 help
57 Enable various WARN_ON checks in the Xen MMU code.
58 Enabling this option WILL incur a significant performance overhead.
diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
index 50542efe45fb..49dbd78ec3cb 100644
--- a/arch/x86/xen/enlighten.c
+++ b/arch/x86/xen/enlighten.c
@@ -1284,15 +1284,14 @@ static int init_hvm_pv_info(int *major, int *minor)
1284 1284
1285 xen_setup_features(); 1285 xen_setup_features();
1286 1286
1287 pv_info = xen_info; 1287 pv_info.name = "Xen HVM";
1288 pv_info.kernel_rpl = 0;
1289 1288
1290 xen_domain_type = XEN_HVM_DOMAIN; 1289 xen_domain_type = XEN_HVM_DOMAIN;
1291 1290
1292 return 0; 1291 return 0;
1293} 1292}
1294 1293
1295void xen_hvm_init_shared_info(void) 1294void __ref xen_hvm_init_shared_info(void)
1296{ 1295{
1297 int cpu; 1296 int cpu;
1298 struct xen_add_to_physmap xatp; 1297 struct xen_add_to_physmap xatp;
@@ -1331,6 +1330,8 @@ static int __cpuinit xen_hvm_cpu_notify(struct notifier_block *self,
1331 switch (action) { 1330 switch (action) {
1332 case CPU_UP_PREPARE: 1331 case CPU_UP_PREPARE:
1333 per_cpu(xen_vcpu, cpu) = &HYPERVISOR_shared_info->vcpu_info[cpu]; 1332 per_cpu(xen_vcpu, cpu) = &HYPERVISOR_shared_info->vcpu_info[cpu];
1333 if (xen_have_vector_callback)
1334 xen_init_lock_cpu(cpu);
1334 break; 1335 break;
1335 default: 1336 default:
1336 break; 1337 break;
@@ -1355,6 +1356,7 @@ static void __init xen_hvm_guest_init(void)
1355 1356
1356 if (xen_feature(XENFEAT_hvm_callback_vector)) 1357 if (xen_feature(XENFEAT_hvm_callback_vector))
1357 xen_have_vector_callback = 1; 1358 xen_have_vector_callback = 1;
1359 xen_hvm_smp_init();
1358 register_cpu_notifier(&xen_hvm_cpu_notifier); 1360 register_cpu_notifier(&xen_hvm_cpu_notifier);
1359 xen_unplug_emulated_devices(); 1361 xen_unplug_emulated_devices();
1360 have_vcpu_info_placement = 0; 1362 have_vcpu_info_placement = 0;
diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c
index f6089421147a..3f6f3347aa17 100644
--- a/arch/x86/xen/mmu.c
+++ b/arch/x86/xen/mmu.c
@@ -46,6 +46,7 @@
46#include <linux/module.h> 46#include <linux/module.h>
47#include <linux/gfp.h> 47#include <linux/gfp.h>
48#include <linux/memblock.h> 48#include <linux/memblock.h>
49#include <linux/seq_file.h>
49 50
50#include <asm/pgtable.h> 51#include <asm/pgtable.h>
51#include <asm/tlbflush.h> 52#include <asm/tlbflush.h>
@@ -416,8 +417,12 @@ static pteval_t pte_pfn_to_mfn(pteval_t val)
416 if (val & _PAGE_PRESENT) { 417 if (val & _PAGE_PRESENT) {
417 unsigned long pfn = (val & PTE_PFN_MASK) >> PAGE_SHIFT; 418 unsigned long pfn = (val & PTE_PFN_MASK) >> PAGE_SHIFT;
418 pteval_t flags = val & PTE_FLAGS_MASK; 419 pteval_t flags = val & PTE_FLAGS_MASK;
419 unsigned long mfn = pfn_to_mfn(pfn); 420 unsigned long mfn;
420 421
422 if (!xen_feature(XENFEAT_auto_translated_physmap))
423 mfn = get_phys_to_machine(pfn);
424 else
425 mfn = pfn;
421 /* 426 /*
422 * If there's no mfn for the pfn, then just create an 427 * If there's no mfn for the pfn, then just create an
423 * empty non-present pte. Unfortunately this loses 428 * empty non-present pte. Unfortunately this loses
@@ -427,8 +432,18 @@ static pteval_t pte_pfn_to_mfn(pteval_t val)
427 if (unlikely(mfn == INVALID_P2M_ENTRY)) { 432 if (unlikely(mfn == INVALID_P2M_ENTRY)) {
428 mfn = 0; 433 mfn = 0;
429 flags = 0; 434 flags = 0;
435 } else {
436 /*
437 * Paramount to do this test _after_ the
438 * INVALID_P2M_ENTRY as INVALID_P2M_ENTRY &
439 * IDENTITY_FRAME_BIT resolves to true.
440 */
441 mfn &= ~FOREIGN_FRAME_BIT;
442 if (mfn & IDENTITY_FRAME_BIT) {
443 mfn &= ~IDENTITY_FRAME_BIT;
444 flags |= _PAGE_IOMAP;
445 }
430 } 446 }
431
432 val = ((pteval_t)mfn << PAGE_SHIFT) | flags; 447 val = ((pteval_t)mfn << PAGE_SHIFT) | flags;
433 } 448 }
434 449
@@ -532,6 +547,41 @@ pte_t xen_make_pte(pteval_t pte)
532} 547}
533PV_CALLEE_SAVE_REGS_THUNK(xen_make_pte); 548PV_CALLEE_SAVE_REGS_THUNK(xen_make_pte);
534 549
550#ifdef CONFIG_XEN_DEBUG
551pte_t xen_make_pte_debug(pteval_t pte)
552{
553 phys_addr_t addr = (pte & PTE_PFN_MASK);
554 phys_addr_t other_addr;
555 bool io_page = false;
556 pte_t _pte;
557
558 if (pte & _PAGE_IOMAP)
559 io_page = true;
560
561 _pte = xen_make_pte(pte);
562
563 if (!addr)
564 return _pte;
565
566 if (io_page &&
567 (xen_initial_domain() || addr >= ISA_END_ADDRESS)) {
568 other_addr = pfn_to_mfn(addr >> PAGE_SHIFT) << PAGE_SHIFT;
569 WARN(addr != other_addr,
570 "0x%lx is using VM_IO, but it is 0x%lx!\n",
571 (unsigned long)addr, (unsigned long)other_addr);
572 } else {
573 pteval_t iomap_set = (_pte.pte & PTE_FLAGS_MASK) & _PAGE_IOMAP;
574 other_addr = (_pte.pte & PTE_PFN_MASK);
575 WARN((addr == other_addr) && (!io_page) && (!iomap_set),
576 "0x%lx is missing VM_IO (and wasn't fixed)!\n",
577 (unsigned long)addr);
578 }
579
580 return _pte;
581}
582PV_CALLEE_SAVE_REGS_THUNK(xen_make_pte_debug);
583#endif
584
535pgd_t xen_make_pgd(pgdval_t pgd) 585pgd_t xen_make_pgd(pgdval_t pgd)
536{ 586{
537 pgd = pte_pfn_to_mfn(pgd); 587 pgd = pte_pfn_to_mfn(pgd);
@@ -1441,7 +1491,7 @@ static __init pte_t mask_rw_pte(pte_t *ptep, pte_t pte)
1441 * early_ioremap fixmap slot, make sure it is RO. 1491 * early_ioremap fixmap slot, make sure it is RO.
1442 */ 1492 */
1443 if (!is_early_ioremap_ptep(ptep) && 1493 if (!is_early_ioremap_ptep(ptep) &&
1444 pfn >= e820_table_start && pfn < e820_table_end) 1494 pfn >= pgt_buf_start && pfn < pgt_buf_end)
1445 pte = pte_wrprotect(pte); 1495 pte = pte_wrprotect(pte);
1446 1496
1447 return pte; 1497 return pte;
@@ -1940,6 +1990,9 @@ __init void xen_ident_map_ISA(void)
1940 1990
1941static __init void xen_post_allocator_init(void) 1991static __init void xen_post_allocator_init(void)
1942{ 1992{
1993#ifdef CONFIG_XEN_DEBUG
1994 pv_mmu_ops.make_pte = PV_CALLEE_SAVE(xen_make_pte_debug);
1995#endif
1943 pv_mmu_ops.set_pte = xen_set_pte; 1996 pv_mmu_ops.set_pte = xen_set_pte;
1944 pv_mmu_ops.set_pmd = xen_set_pmd; 1997 pv_mmu_ops.set_pmd = xen_set_pmd;
1945 pv_mmu_ops.set_pud = xen_set_pud; 1998 pv_mmu_ops.set_pud = xen_set_pud;
@@ -2072,7 +2125,7 @@ static void xen_zap_pfn_range(unsigned long vaddr, unsigned int order,
2072 in_frames[i] = virt_to_mfn(vaddr); 2125 in_frames[i] = virt_to_mfn(vaddr);
2073 2126
2074 MULTI_update_va_mapping(mcs.mc, vaddr, VOID_PTE, 0); 2127 MULTI_update_va_mapping(mcs.mc, vaddr, VOID_PTE, 0);
2075 set_phys_to_machine(virt_to_pfn(vaddr), INVALID_P2M_ENTRY); 2128 __set_phys_to_machine(virt_to_pfn(vaddr), INVALID_P2M_ENTRY);
2076 2129
2077 if (out_frames) 2130 if (out_frames)
2078 out_frames[i] = virt_to_pfn(vaddr); 2131 out_frames[i] = virt_to_pfn(vaddr);
@@ -2351,6 +2404,18 @@ EXPORT_SYMBOL_GPL(xen_remap_domain_mfn_range);
2351 2404
2352#ifdef CONFIG_XEN_DEBUG_FS 2405#ifdef CONFIG_XEN_DEBUG_FS
2353 2406
2407static int p2m_dump_open(struct inode *inode, struct file *filp)
2408{
2409 return single_open(filp, p2m_dump_show, NULL);
2410}
2411
2412static const struct file_operations p2m_dump_fops = {
2413 .open = p2m_dump_open,
2414 .read = seq_read,
2415 .llseek = seq_lseek,
2416 .release = single_release,
2417};
2418
2354static struct dentry *d_mmu_debug; 2419static struct dentry *d_mmu_debug;
2355 2420
2356static int __init xen_mmu_debugfs(void) 2421static int __init xen_mmu_debugfs(void)
@@ -2406,6 +2471,7 @@ static int __init xen_mmu_debugfs(void)
2406 debugfs_create_u32("prot_commit_batched", 0444, d_mmu_debug, 2471 debugfs_create_u32("prot_commit_batched", 0444, d_mmu_debug,
2407 &mmu_stats.prot_commit_batched); 2472 &mmu_stats.prot_commit_batched);
2408 2473
2474 debugfs_create_file("p2m", 0600, d_mmu_debug, NULL, &p2m_dump_fops);
2409 return 0; 2475 return 0;
2410} 2476}
2411fs_initcall(xen_mmu_debugfs); 2477fs_initcall(xen_mmu_debugfs);
diff --git a/arch/x86/xen/p2m.c b/arch/x86/xen/p2m.c
index fd12d7ce7ff9..215a3ce61068 100644
--- a/arch/x86/xen/p2m.c
+++ b/arch/x86/xen/p2m.c
@@ -23,6 +23,129 @@
23 * P2M_PER_PAGE depends on the architecture, as a mfn is always 23 * P2M_PER_PAGE depends on the architecture, as a mfn is always
24 * unsigned long (8 bytes on 64-bit, 4 bytes on 32), leading to 24 * unsigned long (8 bytes on 64-bit, 4 bytes on 32), leading to
25 * 512 and 1024 entries respectively. 25 * 512 and 1024 entries respectively.
26 *
27 * In short, these structures contain the Machine Frame Number (MFN) of the PFN.
28 *
29 * However not all entries are filled with MFNs. Specifically for all other
30 * leaf entries, or for the top root, or middle one, for which there is a void
31 * entry, we assume it is "missing". So (for example)
32 * pfn_to_mfn(0x90909090)=INVALID_P2M_ENTRY.
33 *
34 * We also have the possibility of setting 1-1 mappings on certain regions, so
35 * that:
36 * pfn_to_mfn(0xc0000)=0xc0000
37 *
38 * The benefit of this is, that we can assume for non-RAM regions (think
39 * PCI BARs, or ACPI spaces), we can create mappings easily b/c we
40 * get the PFN value to match the MFN.
41 *
42 * For this to work efficiently we have one new page p2m_identity and
43 * allocate (via reserved_brk) any other pages we need to cover the sides
44 * (1GB or 4MB boundary violations). All entries in p2m_identity are set to
45 * INVALID_P2M_ENTRY type (Xen toolstack only recognizes that and MFNs,
46 * no other fancy value).
47 *
48 * On lookup we spot that the entry points to p2m_identity and return the
49 * identity value instead of dereferencing and returning INVALID_P2M_ENTRY.
50 * If the entry points to an allocated page, we just proceed as before and
51 * return the PFN. If the PFN has IDENTITY_FRAME_BIT set we unmask that in
52 * appropriate functions (pfn_to_mfn).
53 *
54 * The reason for having the IDENTITY_FRAME_BIT instead of just returning the
55 * PFN is that we could find ourselves where pfn_to_mfn(pfn)==pfn for a
56 * non-identity pfn. To protect ourselves against we elect to set (and get) the
57 * IDENTITY_FRAME_BIT on all identity mapped PFNs.
58 *
59 * This simplistic diagram is used to explain the more subtle piece of code.
60 * There is also a digram of the P2M at the end that can help.
61 * Imagine your E820 looking as so:
62 *
63 * 1GB 2GB
64 * /-------------------+---------\/----\ /----------\ /---+-----\
65 * | System RAM | Sys RAM ||ACPI| | reserved | | Sys RAM |
66 * \-------------------+---------/\----/ \----------/ \---+-----/
67 * ^- 1029MB ^- 2001MB
68 *
69 * [1029MB = 263424 (0x40500), 2001MB = 512256 (0x7D100),
70 * 2048MB = 524288 (0x80000)]
71 *
72 * And dom0_mem=max:3GB,1GB is passed in to the guest, meaning memory past 1GB
73 * is actually not present (would have to kick the balloon driver to put it in).
74 *
75 * When we are told to set the PFNs for identity mapping (see patch: "xen/setup:
76 * Set identity mapping for non-RAM E820 and E820 gaps.") we pass in the start
77 * of the PFN and the end PFN (263424 and 512256 respectively). The first step
78 * is to reserve_brk a top leaf page if the p2m[1] is missing. The top leaf page
79 * covers 512^2 of page estate (1GB) and in case the start or end PFN is not
80 * aligned on 512^2*PAGE_SIZE (1GB) we loop on aligned 1GB PFNs from start pfn
81 * to end pfn. We reserve_brk top leaf pages if they are missing (means they
82 * point to p2m_mid_missing).
83 *
84 * With the E820 example above, 263424 is not 1GB aligned so we allocate a
85 * reserve_brk page which will cover the PFNs estate from 0x40000 to 0x80000.
86 * Each entry in the allocate page is "missing" (points to p2m_missing).
87 *
88 * Next stage is to determine if we need to do a more granular boundary check
89 * on the 4MB (or 2MB depending on architecture) off the start and end pfn's.
90 * We check if the start pfn and end pfn violate that boundary check, and if
91 * so reserve_brk a middle (p2m[x][y]) leaf page. This way we have a much finer
92 * granularity of setting which PFNs are missing and which ones are identity.
93 * In our example 263424 and 512256 both fail the check so we reserve_brk two
94 * pages. Populate them with INVALID_P2M_ENTRY (so they both have "missing"
95 * values) and assign them to p2m[1][2] and p2m[1][488] respectively.
96 *
97 * At this point we would at minimum reserve_brk one page, but could be up to
98 * three. Each call to set_phys_range_identity has at maximum a three page
99 * cost. If we were to query the P2M at this stage, all those entries from
100 * start PFN through end PFN (so 1029MB -> 2001MB) would return
101 * INVALID_P2M_ENTRY ("missing").
102 *
103 * The next step is to walk from the start pfn to the end pfn setting
104 * the IDENTITY_FRAME_BIT on each PFN. This is done in set_phys_range_identity.
105 * If we find that the middle leaf is pointing to p2m_missing we can swap it
106 * over to p2m_identity - this way covering 4MB (or 2MB) PFN space. At this
107 * point we do not need to worry about boundary aligment (so no need to
108 * reserve_brk a middle page, figure out which PFNs are "missing" and which
109 * ones are identity), as that has been done earlier. If we find that the
110 * middle leaf is not occupied by p2m_identity or p2m_missing, we dereference
111 * that page (which covers 512 PFNs) and set the appropriate PFN with
112 * IDENTITY_FRAME_BIT. In our example 263424 and 512256 end up there, and we
113 * set from p2m[1][2][256->511] and p2m[1][488][0->256] with
114 * IDENTITY_FRAME_BIT set.
115 *
116 * All other regions that are void (or not filled) either point to p2m_missing
117 * (considered missing) or have the default value of INVALID_P2M_ENTRY (also
118 * considered missing). In our case, p2m[1][2][0->255] and p2m[1][488][257->511]
119 * contain the INVALID_P2M_ENTRY value and are considered "missing."
120 *
121 * This is what the p2m ends up looking (for the E820 above) with this
122 * fabulous drawing:
123 *
124 * p2m /--------------\
125 * /-----\ | &mfn_list[0],| /-----------------\
126 * | 0 |------>| &mfn_list[1],| /---------------\ | ~0, ~0, .. |
127 * |-----| | ..., ~0, ~0 | | ~0, ~0, [x]---+----->| IDENTITY [@256] |
128 * | 1 |---\ \--------------/ | [p2m_identity]+\ | IDENTITY [@257] |
129 * |-----| \ | [p2m_identity]+\\ | .... |
130 * | 2 |--\ \-------------------->| ... | \\ \----------------/
131 * |-----| \ \---------------/ \\
132 * | 3 |\ \ \\ p2m_identity
133 * |-----| \ \-------------------->/---------------\ /-----------------\
134 * | .. +->+ | [p2m_identity]+-->| ~0, ~0, ~0, ... |
135 * \-----/ / | [p2m_identity]+-->| ..., ~0 |
136 * / /---------------\ | .... | \-----------------/
137 * / | IDENTITY[@0] | /-+-[x], ~0, ~0.. |
138 * / | IDENTITY[@256]|<----/ \---------------/
139 * / | ~0, ~0, .... |
140 * | \---------------/
141 * |
142 * p2m_missing p2m_missing
143 * /------------------\ /------------\
144 * | [p2m_mid_missing]+---->| ~0, ~0, ~0 |
145 * | [p2m_mid_missing]+---->| ..., ~0 |
146 * \------------------/ \------------/
147 *
148 * where ~0 is INVALID_P2M_ENTRY. IDENTITY is (PFN | IDENTITY_BIT)
26 */ 149 */
27 150
28#include <linux/init.h> 151#include <linux/init.h>
@@ -30,6 +153,7 @@
30#include <linux/list.h> 153#include <linux/list.h>
31#include <linux/hash.h> 154#include <linux/hash.h>
32#include <linux/sched.h> 155#include <linux/sched.h>
156#include <linux/seq_file.h>
33 157
34#include <asm/cache.h> 158#include <asm/cache.h>
35#include <asm/setup.h> 159#include <asm/setup.h>
@@ -59,9 +183,15 @@ static RESERVE_BRK_ARRAY(unsigned long **, p2m_top, P2M_TOP_PER_PAGE);
59static RESERVE_BRK_ARRAY(unsigned long, p2m_top_mfn, P2M_TOP_PER_PAGE); 183static RESERVE_BRK_ARRAY(unsigned long, p2m_top_mfn, P2M_TOP_PER_PAGE);
60static RESERVE_BRK_ARRAY(unsigned long *, p2m_top_mfn_p, P2M_TOP_PER_PAGE); 184static RESERVE_BRK_ARRAY(unsigned long *, p2m_top_mfn_p, P2M_TOP_PER_PAGE);
61 185
186static RESERVE_BRK_ARRAY(unsigned long, p2m_identity, P2M_PER_PAGE);
187
62RESERVE_BRK(p2m_mid, PAGE_SIZE * (MAX_DOMAIN_PAGES / (P2M_PER_PAGE * P2M_MID_PER_PAGE))); 188RESERVE_BRK(p2m_mid, PAGE_SIZE * (MAX_DOMAIN_PAGES / (P2M_PER_PAGE * P2M_MID_PER_PAGE)));
63RESERVE_BRK(p2m_mid_mfn, PAGE_SIZE * (MAX_DOMAIN_PAGES / (P2M_PER_PAGE * P2M_MID_PER_PAGE))); 189RESERVE_BRK(p2m_mid_mfn, PAGE_SIZE * (MAX_DOMAIN_PAGES / (P2M_PER_PAGE * P2M_MID_PER_PAGE)));
64 190
191/* We might hit two boundary violations at the start and end, at max each
192 * boundary violation will require three middle nodes. */
193RESERVE_BRK(p2m_mid_identity, PAGE_SIZE * 2 * 3);
194
65static inline unsigned p2m_top_index(unsigned long pfn) 195static inline unsigned p2m_top_index(unsigned long pfn)
66{ 196{
67 BUG_ON(pfn >= MAX_P2M_PFN); 197 BUG_ON(pfn >= MAX_P2M_PFN);
@@ -136,7 +266,7 @@ static void p2m_init(unsigned long *p2m)
136 * - After resume we're called from within stop_machine, but the mfn 266 * - After resume we're called from within stop_machine, but the mfn
137 * tree should alreay be completely allocated. 267 * tree should alreay be completely allocated.
138 */ 268 */
139void xen_build_mfn_list_list(void) 269void __ref xen_build_mfn_list_list(void)
140{ 270{
141 unsigned long pfn; 271 unsigned long pfn;
142 272
@@ -221,6 +351,9 @@ void __init xen_build_dynamic_phys_to_machine(void)
221 p2m_top = extend_brk(PAGE_SIZE, PAGE_SIZE); 351 p2m_top = extend_brk(PAGE_SIZE, PAGE_SIZE);
222 p2m_top_init(p2m_top); 352 p2m_top_init(p2m_top);
223 353
354 p2m_identity = extend_brk(PAGE_SIZE, PAGE_SIZE);
355 p2m_init(p2m_identity);
356
224 /* 357 /*
225 * The domain builder gives us a pre-constructed p2m array in 358 * The domain builder gives us a pre-constructed p2m array in
226 * mfn_list for all the pages initially given to us, so we just 359 * mfn_list for all the pages initially given to us, so we just
@@ -266,6 +399,14 @@ unsigned long get_phys_to_machine(unsigned long pfn)
266 mididx = p2m_mid_index(pfn); 399 mididx = p2m_mid_index(pfn);
267 idx = p2m_index(pfn); 400 idx = p2m_index(pfn);
268 401
402 /*
403 * The INVALID_P2M_ENTRY is filled in both p2m_*identity
404 * and in p2m_*missing, so returning the INVALID_P2M_ENTRY
405 * would be wrong.
406 */
407 if (p2m_top[topidx][mididx] == p2m_identity)
408 return IDENTITY_FRAME(pfn);
409
269 return p2m_top[topidx][mididx][idx]; 410 return p2m_top[topidx][mididx][idx];
270} 411}
271EXPORT_SYMBOL_GPL(get_phys_to_machine); 412EXPORT_SYMBOL_GPL(get_phys_to_machine);
@@ -335,9 +476,11 @@ static bool alloc_p2m(unsigned long pfn)
335 p2m_top_mfn_p[topidx] = mid_mfn; 476 p2m_top_mfn_p[topidx] = mid_mfn;
336 } 477 }
337 478
338 if (p2m_top[topidx][mididx] == p2m_missing) { 479 if (p2m_top[topidx][mididx] == p2m_identity ||
480 p2m_top[topidx][mididx] == p2m_missing) {
339 /* p2m leaf page is missing */ 481 /* p2m leaf page is missing */
340 unsigned long *p2m; 482 unsigned long *p2m;
483 unsigned long *p2m_orig = p2m_top[topidx][mididx];
341 484
342 p2m = alloc_p2m_page(); 485 p2m = alloc_p2m_page();
343 if (!p2m) 486 if (!p2m)
@@ -345,7 +488,7 @@ static bool alloc_p2m(unsigned long pfn)
345 488
346 p2m_init(p2m); 489 p2m_init(p2m);
347 490
348 if (cmpxchg(&mid[mididx], p2m_missing, p2m) != p2m_missing) 491 if (cmpxchg(&mid[mididx], p2m_orig, p2m) != p2m_orig)
349 free_p2m_page(p2m); 492 free_p2m_page(p2m);
350 else 493 else
351 mid_mfn[mididx] = virt_to_mfn(p2m); 494 mid_mfn[mididx] = virt_to_mfn(p2m);
@@ -354,11 +497,91 @@ static bool alloc_p2m(unsigned long pfn)
354 return true; 497 return true;
355} 498}
356 499
500bool __early_alloc_p2m(unsigned long pfn)
501{
502 unsigned topidx, mididx, idx;
503
504 topidx = p2m_top_index(pfn);
505 mididx = p2m_mid_index(pfn);
506 idx = p2m_index(pfn);
507
508 /* Pfff.. No boundary cross-over, lets get out. */
509 if (!idx)
510 return false;
511
512 WARN(p2m_top[topidx][mididx] == p2m_identity,
513 "P2M[%d][%d] == IDENTITY, should be MISSING (or alloced)!\n",
514 topidx, mididx);
515
516 /*
517 * Could be done by xen_build_dynamic_phys_to_machine..
518 */
519 if (p2m_top[topidx][mididx] != p2m_missing)
520 return false;
521
522 /* Boundary cross-over for the edges: */
523 if (idx) {
524 unsigned long *p2m = extend_brk(PAGE_SIZE, PAGE_SIZE);
525
526 p2m_init(p2m);
527
528 p2m_top[topidx][mididx] = p2m;
529
530 }
531 return idx != 0;
532}
533unsigned long set_phys_range_identity(unsigned long pfn_s,
534 unsigned long pfn_e)
535{
536 unsigned long pfn;
537
538 if (unlikely(pfn_s >= MAX_P2M_PFN || pfn_e >= MAX_P2M_PFN))
539 return 0;
540
541 if (unlikely(xen_feature(XENFEAT_auto_translated_physmap)))
542 return pfn_e - pfn_s;
543
544 if (pfn_s > pfn_e)
545 return 0;
546
547 for (pfn = (pfn_s & ~(P2M_MID_PER_PAGE * P2M_PER_PAGE - 1));
548 pfn < ALIGN(pfn_e, (P2M_MID_PER_PAGE * P2M_PER_PAGE));
549 pfn += P2M_MID_PER_PAGE * P2M_PER_PAGE)
550 {
551 unsigned topidx = p2m_top_index(pfn);
552 if (p2m_top[topidx] == p2m_mid_missing) {
553 unsigned long **mid = extend_brk(PAGE_SIZE, PAGE_SIZE);
554
555 p2m_mid_init(mid);
556
557 p2m_top[topidx] = mid;
558 }
559 }
560
561 __early_alloc_p2m(pfn_s);
562 __early_alloc_p2m(pfn_e);
563
564 for (pfn = pfn_s; pfn < pfn_e; pfn++)
565 if (!__set_phys_to_machine(pfn, IDENTITY_FRAME(pfn)))
566 break;
567
568 if (!WARN((pfn - pfn_s) != (pfn_e - pfn_s),
569 "Identity mapping failed. We are %ld short of 1-1 mappings!\n",
570 (pfn_e - pfn_s) - (pfn - pfn_s)))
571 printk(KERN_DEBUG "1-1 mapping on %lx->%lx\n", pfn_s, pfn);
572
573 return pfn - pfn_s;
574}
575
357/* Try to install p2m mapping; fail if intermediate bits missing */ 576/* Try to install p2m mapping; fail if intermediate bits missing */
358bool __set_phys_to_machine(unsigned long pfn, unsigned long mfn) 577bool __set_phys_to_machine(unsigned long pfn, unsigned long mfn)
359{ 578{
360 unsigned topidx, mididx, idx; 579 unsigned topidx, mididx, idx;
361 580
581 if (unlikely(xen_feature(XENFEAT_auto_translated_physmap))) {
582 BUG_ON(pfn != mfn && mfn != INVALID_P2M_ENTRY);
583 return true;
584 }
362 if (unlikely(pfn >= MAX_P2M_PFN)) { 585 if (unlikely(pfn >= MAX_P2M_PFN)) {
363 BUG_ON(mfn != INVALID_P2M_ENTRY); 586 BUG_ON(mfn != INVALID_P2M_ENTRY);
364 return true; 587 return true;
@@ -368,6 +591,21 @@ bool __set_phys_to_machine(unsigned long pfn, unsigned long mfn)
368 mididx = p2m_mid_index(pfn); 591 mididx = p2m_mid_index(pfn);
369 idx = p2m_index(pfn); 592 idx = p2m_index(pfn);
370 593
594 /* For sparse holes were the p2m leaf has real PFN along with
595 * PCI holes, stick in the PFN as the MFN value.
596 */
597 if (mfn != INVALID_P2M_ENTRY && (mfn & IDENTITY_FRAME_BIT)) {
598 if (p2m_top[topidx][mididx] == p2m_identity)
599 return true;
600
601 /* Swap over from MISSING to IDENTITY if needed. */
602 if (p2m_top[topidx][mididx] == p2m_missing) {
603 WARN_ON(cmpxchg(&p2m_top[topidx][mididx], p2m_missing,
604 p2m_identity) != p2m_missing);
605 return true;
606 }
607 }
608
371 if (p2m_top[topidx][mididx] == p2m_missing) 609 if (p2m_top[topidx][mididx] == p2m_missing)
372 return mfn == INVALID_P2M_ENTRY; 610 return mfn == INVALID_P2M_ENTRY;
373 611
@@ -378,11 +616,6 @@ bool __set_phys_to_machine(unsigned long pfn, unsigned long mfn)
378 616
379bool set_phys_to_machine(unsigned long pfn, unsigned long mfn) 617bool set_phys_to_machine(unsigned long pfn, unsigned long mfn)
380{ 618{
381 if (unlikely(xen_feature(XENFEAT_auto_translated_physmap))) {
382 BUG_ON(pfn != mfn && mfn != INVALID_P2M_ENTRY);
383 return true;
384 }
385
386 if (unlikely(!__set_phys_to_machine(pfn, mfn))) { 619 if (unlikely(!__set_phys_to_machine(pfn, mfn))) {
387 if (!alloc_p2m(pfn)) 620 if (!alloc_p2m(pfn))
388 return false; 621 return false;
@@ -421,7 +654,7 @@ int m2p_add_override(unsigned long mfn, struct page *page)
421{ 654{
422 unsigned long flags; 655 unsigned long flags;
423 unsigned long pfn; 656 unsigned long pfn;
424 unsigned long address; 657 unsigned long uninitialized_var(address);
425 unsigned level; 658 unsigned level;
426 pte_t *ptep = NULL; 659 pte_t *ptep = NULL;
427 660
@@ -455,7 +688,7 @@ int m2p_remove_override(struct page *page)
455 unsigned long flags; 688 unsigned long flags;
456 unsigned long mfn; 689 unsigned long mfn;
457 unsigned long pfn; 690 unsigned long pfn;
458 unsigned long address; 691 unsigned long uninitialized_var(address);
459 unsigned level; 692 unsigned level;
460 pte_t *ptep = NULL; 693 pte_t *ptep = NULL;
461 694
@@ -520,3 +753,80 @@ unsigned long m2p_find_override_pfn(unsigned long mfn, unsigned long pfn)
520 return ret; 753 return ret;
521} 754}
522EXPORT_SYMBOL_GPL(m2p_find_override_pfn); 755EXPORT_SYMBOL_GPL(m2p_find_override_pfn);
756
757#ifdef CONFIG_XEN_DEBUG_FS
758
759int p2m_dump_show(struct seq_file *m, void *v)
760{
761 static const char * const level_name[] = { "top", "middle",
762 "entry", "abnormal" };
763 static const char * const type_name[] = { "identity", "missing",
764 "pfn", "abnormal"};
765#define TYPE_IDENTITY 0
766#define TYPE_MISSING 1
767#define TYPE_PFN 2
768#define TYPE_UNKNOWN 3
769 unsigned long pfn, prev_pfn_type = 0, prev_pfn_level = 0;
770 unsigned int uninitialized_var(prev_level);
771 unsigned int uninitialized_var(prev_type);
772
773 if (!p2m_top)
774 return 0;
775
776 for (pfn = 0; pfn < MAX_DOMAIN_PAGES; pfn++) {
777 unsigned topidx = p2m_top_index(pfn);
778 unsigned mididx = p2m_mid_index(pfn);
779 unsigned idx = p2m_index(pfn);
780 unsigned lvl, type;
781
782 lvl = 4;
783 type = TYPE_UNKNOWN;
784 if (p2m_top[topidx] == p2m_mid_missing) {
785 lvl = 0; type = TYPE_MISSING;
786 } else if (p2m_top[topidx] == NULL) {
787 lvl = 0; type = TYPE_UNKNOWN;
788 } else if (p2m_top[topidx][mididx] == NULL) {
789 lvl = 1; type = TYPE_UNKNOWN;
790 } else if (p2m_top[topidx][mididx] == p2m_identity) {
791 lvl = 1; type = TYPE_IDENTITY;
792 } else if (p2m_top[topidx][mididx] == p2m_missing) {
793 lvl = 1; type = TYPE_MISSING;
794 } else if (p2m_top[topidx][mididx][idx] == 0) {
795 lvl = 2; type = TYPE_UNKNOWN;
796 } else if (p2m_top[topidx][mididx][idx] == IDENTITY_FRAME(pfn)) {
797 lvl = 2; type = TYPE_IDENTITY;
798 } else if (p2m_top[topidx][mididx][idx] == INVALID_P2M_ENTRY) {
799 lvl = 2; type = TYPE_MISSING;
800 } else if (p2m_top[topidx][mididx][idx] == pfn) {
801 lvl = 2; type = TYPE_PFN;
802 } else if (p2m_top[topidx][mididx][idx] != pfn) {
803 lvl = 2; type = TYPE_PFN;
804 }
805 if (pfn == 0) {
806 prev_level = lvl;
807 prev_type = type;
808 }
809 if (pfn == MAX_DOMAIN_PAGES-1) {
810 lvl = 3;
811 type = TYPE_UNKNOWN;
812 }
813 if (prev_type != type) {
814 seq_printf(m, " [0x%lx->0x%lx] %s\n",
815 prev_pfn_type, pfn, type_name[prev_type]);
816 prev_pfn_type = pfn;
817 prev_type = type;
818 }
819 if (prev_level != lvl) {
820 seq_printf(m, " [0x%lx->0x%lx] level %s\n",
821 prev_pfn_level, pfn, level_name[prev_level]);
822 prev_pfn_level = pfn;
823 prev_level = lvl;
824 }
825 }
826 return 0;
827#undef TYPE_IDENTITY
828#undef TYPE_MISSING
829#undef TYPE_PFN
830#undef TYPE_UNKNOWN
831}
832#endif
diff --git a/arch/x86/xen/setup.c b/arch/x86/xen/setup.c
index a8a66a50d446..fa0269a99377 100644
--- a/arch/x86/xen/setup.c
+++ b/arch/x86/xen/setup.c
@@ -52,6 +52,8 @@ phys_addr_t xen_extra_mem_start, xen_extra_mem_size;
52 52
53static __init void xen_add_extra_mem(unsigned long pages) 53static __init void xen_add_extra_mem(unsigned long pages)
54{ 54{
55 unsigned long pfn;
56
55 u64 size = (u64)pages * PAGE_SIZE; 57 u64 size = (u64)pages * PAGE_SIZE;
56 u64 extra_start = xen_extra_mem_start + xen_extra_mem_size; 58 u64 extra_start = xen_extra_mem_start + xen_extra_mem_size;
57 59
@@ -66,6 +68,9 @@ static __init void xen_add_extra_mem(unsigned long pages)
66 xen_extra_mem_size += size; 68 xen_extra_mem_size += size;
67 69
68 xen_max_p2m_pfn = PFN_DOWN(extra_start + size); 70 xen_max_p2m_pfn = PFN_DOWN(extra_start + size);
71
72 for (pfn = PFN_DOWN(extra_start); pfn <= xen_max_p2m_pfn; pfn++)
73 __set_phys_to_machine(pfn, INVALID_P2M_ENTRY);
69} 74}
70 75
71static unsigned long __init xen_release_chunk(phys_addr_t start_addr, 76static unsigned long __init xen_release_chunk(phys_addr_t start_addr,
@@ -104,7 +109,7 @@ static unsigned long __init xen_release_chunk(phys_addr_t start_addr,
104 WARN(ret != 1, "Failed to release memory %lx-%lx err=%d\n", 109 WARN(ret != 1, "Failed to release memory %lx-%lx err=%d\n",
105 start, end, ret); 110 start, end, ret);
106 if (ret == 1) { 111 if (ret == 1) {
107 set_phys_to_machine(pfn, INVALID_P2M_ENTRY); 112 __set_phys_to_machine(pfn, INVALID_P2M_ENTRY);
108 len++; 113 len++;
109 } 114 }
110 } 115 }
@@ -138,12 +143,55 @@ static unsigned long __init xen_return_unused_memory(unsigned long max_pfn,
138 return released; 143 return released;
139} 144}
140 145
146static unsigned long __init xen_set_identity(const struct e820entry *list,
147 ssize_t map_size)
148{
149 phys_addr_t last = xen_initial_domain() ? 0 : ISA_END_ADDRESS;
150 phys_addr_t start_pci = last;
151 const struct e820entry *entry;
152 unsigned long identity = 0;
153 int i;
154
155 for (i = 0, entry = list; i < map_size; i++, entry++) {
156 phys_addr_t start = entry->addr;
157 phys_addr_t end = start + entry->size;
158
159 if (start < last)
160 start = last;
161
162 if (end <= start)
163 continue;
164
165 /* Skip over the 1MB region. */
166 if (last > end)
167 continue;
168
169 if (entry->type == E820_RAM) {
170 if (start > start_pci)
171 identity += set_phys_range_identity(
172 PFN_UP(start_pci), PFN_DOWN(start));
173
174 /* Without saving 'last' we would gooble RAM too
175 * at the end of the loop. */
176 last = end;
177 start_pci = end;
178 continue;
179 }
180 start_pci = min(start, start_pci);
181 last = end;
182 }
183 if (last > start_pci)
184 identity += set_phys_range_identity(
185 PFN_UP(start_pci), PFN_DOWN(last));
186 return identity;
187}
141/** 188/**
142 * machine_specific_memory_setup - Hook for machine specific memory setup. 189 * machine_specific_memory_setup - Hook for machine specific memory setup.
143 **/ 190 **/
144char * __init xen_memory_setup(void) 191char * __init xen_memory_setup(void)
145{ 192{
146 static struct e820entry map[E820MAX] __initdata; 193 static struct e820entry map[E820MAX] __initdata;
194 static struct e820entry map_raw[E820MAX] __initdata;
147 195
148 unsigned long max_pfn = xen_start_info->nr_pages; 196 unsigned long max_pfn = xen_start_info->nr_pages;
149 unsigned long long mem_end; 197 unsigned long long mem_end;
@@ -151,6 +199,7 @@ char * __init xen_memory_setup(void)
151 struct xen_memory_map memmap; 199 struct xen_memory_map memmap;
152 unsigned long extra_pages = 0; 200 unsigned long extra_pages = 0;
153 unsigned long extra_limit; 201 unsigned long extra_limit;
202 unsigned long identity_pages = 0;
154 int i; 203 int i;
155 int op; 204 int op;
156 205
@@ -176,6 +225,7 @@ char * __init xen_memory_setup(void)
176 } 225 }
177 BUG_ON(rc); 226 BUG_ON(rc);
178 227
228 memcpy(map_raw, map, sizeof(map));
179 e820.nr_map = 0; 229 e820.nr_map = 0;
180 xen_extra_mem_start = mem_end; 230 xen_extra_mem_start = mem_end;
181 for (i = 0; i < memmap.nr_entries; i++) { 231 for (i = 0; i < memmap.nr_entries; i++) {
@@ -194,6 +244,15 @@ char * __init xen_memory_setup(void)
194 end -= delta; 244 end -= delta;
195 245
196 extra_pages += PFN_DOWN(delta); 246 extra_pages += PFN_DOWN(delta);
247 /*
248 * Set RAM below 4GB that is not for us to be unusable.
249 * This prevents "System RAM" address space from being
250 * used as potential resource for I/O address (happens
251 * when 'allocate_resource' is called).
252 */
253 if (delta &&
254 (xen_initial_domain() && end < 0x100000000ULL))
255 e820_add_region(end, delta, E820_UNUSABLE);
197 } 256 }
198 257
199 if (map[i].size > 0 && end > xen_extra_mem_start) 258 if (map[i].size > 0 && end > xen_extra_mem_start)
@@ -251,6 +310,13 @@ char * __init xen_memory_setup(void)
251 310
252 xen_add_extra_mem(extra_pages); 311 xen_add_extra_mem(extra_pages);
253 312
313 /*
314 * Set P2M for all non-RAM pages and E820 gaps to be identity
315 * type PFNs. We supply it with the non-sanitized version
316 * of the E820.
317 */
318 identity_pages = xen_set_identity(map_raw, memmap.nr_entries);
319 printk(KERN_INFO "Set %ld page(s) to 1-1 mapping.\n", identity_pages);
254 return "Xen"; 320 return "Xen";
255} 321}
256 322
diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c
index 72a4c7959045..30612441ed99 100644
--- a/arch/x86/xen/smp.c
+++ b/arch/x86/xen/smp.c
@@ -509,3 +509,41 @@ void __init xen_smp_init(void)
509 xen_fill_possible_map(); 509 xen_fill_possible_map();
510 xen_init_spinlocks(); 510 xen_init_spinlocks();
511} 511}
512
513static void __init xen_hvm_smp_prepare_cpus(unsigned int max_cpus)
514{
515 native_smp_prepare_cpus(max_cpus);
516 WARN_ON(xen_smp_intr_init(0));
517
518 if (!xen_have_vector_callback)
519 return;
520 xen_init_lock_cpu(0);
521 xen_init_spinlocks();
522}
523
524static int __cpuinit xen_hvm_cpu_up(unsigned int cpu)
525{
526 int rc;
527 rc = native_cpu_up(cpu);
528 WARN_ON (xen_smp_intr_init(cpu));
529 return rc;
530}
531
532static void xen_hvm_cpu_die(unsigned int cpu)
533{
534 unbind_from_irqhandler(per_cpu(xen_resched_irq, cpu), NULL);
535 unbind_from_irqhandler(per_cpu(xen_callfunc_irq, cpu), NULL);
536 unbind_from_irqhandler(per_cpu(xen_debug_irq, cpu), NULL);
537 unbind_from_irqhandler(per_cpu(xen_callfuncsingle_irq, cpu), NULL);
538 native_cpu_die(cpu);
539}
540
541void __init xen_hvm_smp_init(void)
542{
543 smp_ops.smp_prepare_cpus = xen_hvm_smp_prepare_cpus;
544 smp_ops.smp_send_reschedule = xen_smp_send_reschedule;
545 smp_ops.cpu_up = xen_hvm_cpu_up;
546 smp_ops.cpu_die = xen_hvm_cpu_die;
547 smp_ops.send_call_func_ipi = xen_smp_send_call_function_ipi;
548 smp_ops.send_call_func_single_ipi = xen_smp_send_call_function_single_ipi;
549}
diff --git a/arch/x86/xen/suspend.c b/arch/x86/xen/suspend.c
index 9bbd63a129b5..45329c8c226e 100644
--- a/arch/x86/xen/suspend.c
+++ b/arch/x86/xen/suspend.c
@@ -12,7 +12,7 @@
12#include "xen-ops.h" 12#include "xen-ops.h"
13#include "mmu.h" 13#include "mmu.h"
14 14
15void xen_pre_suspend(void) 15void xen_arch_pre_suspend(void)
16{ 16{
17 xen_start_info->store_mfn = mfn_to_pfn(xen_start_info->store_mfn); 17 xen_start_info->store_mfn = mfn_to_pfn(xen_start_info->store_mfn);
18 xen_start_info->console.domU.mfn = 18 xen_start_info->console.domU.mfn =
@@ -26,8 +26,9 @@ void xen_pre_suspend(void)
26 BUG(); 26 BUG();
27} 27}
28 28
29void xen_hvm_post_suspend(int suspend_cancelled) 29void xen_arch_hvm_post_suspend(int suspend_cancelled)
30{ 30{
31#ifdef CONFIG_XEN_PVHVM
31 int cpu; 32 int cpu;
32 xen_hvm_init_shared_info(); 33 xen_hvm_init_shared_info();
33 xen_callback_vector(); 34 xen_callback_vector();
@@ -37,9 +38,10 @@ void xen_hvm_post_suspend(int suspend_cancelled)
37 xen_setup_runstate_info(cpu); 38 xen_setup_runstate_info(cpu);
38 } 39 }
39 } 40 }
41#endif
40} 42}
41 43
42void xen_post_suspend(int suspend_cancelled) 44void xen_arch_post_suspend(int suspend_cancelled)
43{ 45{
44 xen_build_mfn_list_list(); 46 xen_build_mfn_list_list();
45 47
diff --git a/arch/x86/xen/time.c b/arch/x86/xen/time.c
index 067759e3d6a5..2e2d370a47b1 100644
--- a/arch/x86/xen/time.c
+++ b/arch/x86/xen/time.c
@@ -397,7 +397,9 @@ void xen_setup_timer(int cpu)
397 name = "<timer kasprintf failed>"; 397 name = "<timer kasprintf failed>";
398 398
399 irq = bind_virq_to_irqhandler(VIRQ_TIMER, cpu, xen_timer_interrupt, 399 irq = bind_virq_to_irqhandler(VIRQ_TIMER, cpu, xen_timer_interrupt,
400 IRQF_DISABLED|IRQF_PERCPU|IRQF_NOBALANCING|IRQF_TIMER, 400 IRQF_DISABLED|IRQF_PERCPU|
401 IRQF_NOBALANCING|IRQF_TIMER|
402 IRQF_FORCE_RESUME,
401 name, NULL); 403 name, NULL);
402 404
403 evt = &per_cpu(xen_clock_events, cpu); 405 evt = &per_cpu(xen_clock_events, cpu);
diff --git a/arch/x86/xen/xen-head.S b/arch/x86/xen/xen-head.S
index 1a5ff24e29c0..aaa7291c9259 100644
--- a/arch/x86/xen/xen-head.S
+++ b/arch/x86/xen/xen-head.S
@@ -28,9 +28,9 @@ ENTRY(startup_xen)
28 __FINIT 28 __FINIT
29 29
30.pushsection .text 30.pushsection .text
31 .align PAGE_SIZE_asm 31 .align PAGE_SIZE
32ENTRY(hypercall_page) 32ENTRY(hypercall_page)
33 .skip PAGE_SIZE_asm 33 .skip PAGE_SIZE
34.popsection 34.popsection
35 35
36 ELFNOTE(Xen, XEN_ELFNOTE_GUEST_OS, .asciz "linux") 36 ELFNOTE(Xen, XEN_ELFNOTE_GUEST_OS, .asciz "linux")
diff --git a/arch/x86/xen/xen-ops.h b/arch/x86/xen/xen-ops.h
index 9d41bf985757..3112f55638c4 100644
--- a/arch/x86/xen/xen-ops.h
+++ b/arch/x86/xen/xen-ops.h
@@ -64,10 +64,12 @@ void xen_setup_vcpu_info_placement(void);
64 64
65#ifdef CONFIG_SMP 65#ifdef CONFIG_SMP
66void xen_smp_init(void); 66void xen_smp_init(void);
67void __init xen_hvm_smp_init(void);
67 68
68extern cpumask_var_t xen_cpu_initialized_map; 69extern cpumask_var_t xen_cpu_initialized_map;
69#else 70#else
70static inline void xen_smp_init(void) {} 71static inline void xen_smp_init(void) {}
72static inline void xen_hvm_smp_init(void) {}
71#endif 73#endif
72 74
73#ifdef CONFIG_PARAVIRT_SPINLOCKS 75#ifdef CONFIG_PARAVIRT_SPINLOCKS
diff --git a/arch/xtensa/include/asm/rwsem.h b/arch/xtensa/include/asm/rwsem.h
index e39edf5c86f2..249619e7e7f2 100644
--- a/arch/xtensa/include/asm/rwsem.h
+++ b/arch/xtensa/include/asm/rwsem.h
@@ -17,44 +17,12 @@
17#error "Please don't include <asm/rwsem.h> directly, use <linux/rwsem.h> instead." 17#error "Please don't include <asm/rwsem.h> directly, use <linux/rwsem.h> instead."
18#endif 18#endif
19 19
20#include <linux/list.h>
21#include <linux/spinlock.h>
22#include <asm/atomic.h>
23#include <asm/system.h>
24
25/*
26 * the semaphore definition
27 */
28struct rw_semaphore {
29 signed long count;
30#define RWSEM_UNLOCKED_VALUE 0x00000000 20#define RWSEM_UNLOCKED_VALUE 0x00000000
31#define RWSEM_ACTIVE_BIAS 0x00000001 21#define RWSEM_ACTIVE_BIAS 0x00000001
32#define RWSEM_ACTIVE_MASK 0x0000ffff 22#define RWSEM_ACTIVE_MASK 0x0000ffff
33#define RWSEM_WAITING_BIAS (-0x00010000) 23#define RWSEM_WAITING_BIAS (-0x00010000)
34#define RWSEM_ACTIVE_READ_BIAS RWSEM_ACTIVE_BIAS 24#define RWSEM_ACTIVE_READ_BIAS RWSEM_ACTIVE_BIAS
35#define RWSEM_ACTIVE_WRITE_BIAS (RWSEM_WAITING_BIAS + RWSEM_ACTIVE_BIAS) 25#define RWSEM_ACTIVE_WRITE_BIAS (RWSEM_WAITING_BIAS + RWSEM_ACTIVE_BIAS)
36 spinlock_t wait_lock;
37 struct list_head wait_list;
38};
39
40#define __RWSEM_INITIALIZER(name) \
41 { RWSEM_UNLOCKED_VALUE, SPIN_LOCK_UNLOCKED, \
42 LIST_HEAD_INIT((name).wait_list) }
43
44#define DECLARE_RWSEM(name) \
45 struct rw_semaphore name = __RWSEM_INITIALIZER(name)
46
47extern struct rw_semaphore *rwsem_down_read_failed(struct rw_semaphore *sem);
48extern struct rw_semaphore *rwsem_down_write_failed(struct rw_semaphore *sem);
49extern struct rw_semaphore *rwsem_wake(struct rw_semaphore *sem);
50extern struct rw_semaphore *rwsem_downgrade_wake(struct rw_semaphore *sem);
51
52static inline void init_rwsem(struct rw_semaphore *sem)
53{
54 sem->count = RWSEM_UNLOCKED_VALUE;
55 spin_lock_init(&sem->wait_lock);
56 INIT_LIST_HEAD(&sem->wait_list);
57}
58 26
59/* 27/*
60 * lock for reading 28 * lock for reading
@@ -160,9 +128,4 @@ static inline int rwsem_atomic_update(int delta, struct rw_semaphore *sem)
160 return atomic_add_return(delta, (atomic_t *)(&sem->count)); 128 return atomic_add_return(delta, (atomic_t *)(&sem->count));
161} 129}
162 130
163static inline int rwsem_is_locked(struct rw_semaphore *sem)
164{
165 return (sem->count != 0);
166}
167
168#endif /* _XTENSA_RWSEM_H */ 131#endif /* _XTENSA_RWSEM_H */
diff --git a/arch/xtensa/kernel/time.c b/arch/xtensa/kernel/time.c
index 19df764f6399..f3e5eb43f71c 100644
--- a/arch/xtensa/kernel/time.c
+++ b/arch/xtensa/kernel/time.c
@@ -96,16 +96,12 @@ again:
96 update_process_times(user_mode(get_irq_regs())); 96 update_process_times(user_mode(get_irq_regs()));
97#endif 97#endif
98 98
99 write_seqlock(&xtime_lock); 99 xtime_update(1); /* Linux handler in kernel/time/timekeeping */
100
101 do_timer(1); /* Linux handler in kernel/timer.c */
102 100
103 /* Note that writing CCOMPARE clears the interrupt. */ 101 /* Note that writing CCOMPARE clears the interrupt. */
104 102
105 next += CCOUNT_PER_JIFFY; 103 next += CCOUNT_PER_JIFFY;
106 set_linux_timer(next); 104 set_linux_timer(next);
107
108 write_sequnlock(&xtime_lock);
109 } 105 }
110 106
111 /* Allow platform to do something useful (Wdog). */ 107 /* Allow platform to do something useful (Wdog). */
generated by cgit v1.2.2 (git 2.25.0) at 2024-12-04 06:02:19 -0500