aboutsummaryrefslogtreecommitdiffstats
path: root/arch/x86/kernel
diff options
context:
space:
mode:
authorThomas Gleixner <tglx@linutronix.de>2011-05-14 06:06:36 -0400
committerThomas Gleixner <tglx@linutronix.de>2011-05-14 06:06:36 -0400
commita18f22a968de17b29f2310cdb7ba69163e65ec15 (patch)
treea7d56d88fad5e444d7661484109758a2f436129e /arch/x86/kernel
parenta1c57e0fec53defe745e64417eacdbd3618c3e66 (diff)
parent798778b8653f64b7b2162ac70eca10367cff6ce8 (diff)
Merge branch 'consolidate-clksrc-i8253' of master.kernel.org:~rmk/linux-2.6-arm into timers/clocksource
Conflicts: arch/ia64/kernel/cyclone.c arch/mips/kernel/i8253.c arch/x86/kernel/i8253.c Reason: Resolve conflicts so further cleanups do not conflict further Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Diffstat (limited to 'arch/x86/kernel')
-rw-r--r--arch/x86/kernel/Makefile12
-rw-r--r--arch/x86/kernel/acpi/boot.c22
-rw-r--r--arch/x86/kernel/acpi/realmode/wakeup.S21
-rw-r--r--arch/x86/kernel/acpi/realmode/wakeup.h5
-rw-r--r--arch/x86/kernel/acpi/realmode/wakeup.lds.S28
-rw-r--r--arch/x86/kernel/acpi/sleep.c77
-rw-r--r--arch/x86/kernel/acpi/sleep.h5
-rw-r--r--arch/x86/kernel/acpi/wakeup_rm.S12
-rw-r--r--arch/x86/kernel/alternative.c9
-rw-r--r--arch/x86/kernel/amd_iommu_init.c26
-rw-r--r--arch/x86/kernel/amd_nb.c102
-rw-r--r--arch/x86/kernel/apb_timer.c64
-rw-r--r--arch/x86/kernel/aperture_64.c37
-rw-r--r--arch/x86/kernel/apic/apic.c183
-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.c2
-rw-r--r--arch/x86/kernel/apic/io_apic.c491
-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.c12
-rw-r--r--arch/x86/kernel/apm_32.c24
-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/check.c8
-rw-r--r--arch/x86/kernel/cpu/amd.c86
-rw-r--r--arch/x86/kernel/cpu/common.c6
-rw-r--r--arch/x86/kernel/cpu/cpufreq/longhaul.c4
-rw-r--r--arch/x86/kernel/cpu/cpufreq/p4-clockmod.c6
-rw-r--r--arch/x86/kernel/cpu/cpufreq/pcc-cpufreq.c4
-rw-r--r--arch/x86/kernel/cpu/cpufreq/powernow-k8.c18
-rw-r--r--arch/x86/kernel/cpu/cpufreq/speedstep-smi.c4
-rw-r--r--arch/x86/kernel/cpu/intel.c5
-rw-r--r--arch/x86/kernel/cpu/intel_cacheinfo.c80
-rw-r--r--arch/x86/kernel/cpu/mcheck/mce-apei.c42
-rw-r--r--arch/x86/kernel/cpu/mcheck/mce-inject.c2
-rw-r--r--arch/x86/kernel/cpu/mcheck/mce.c25
-rw-r--r--arch/x86/kernel/cpu/mcheck/mce_amd.c7
-rw-r--r--arch/x86/kernel/cpu/mtrr/generic.c2
-rw-r--r--arch/x86/kernel/cpu/mtrr/main.c30
-rw-r--r--arch/x86/kernel/cpu/perf_event.c199
-rw-r--r--arch/x86/kernel/cpu/perf_event_amd.c191
-rw-r--r--arch/x86/kernel/cpu/perf_event_intel.c472
-rw-r--r--arch/x86/kernel/cpu/perf_event_intel_ds.c83
-rw-r--r--arch/x86/kernel/cpu/perf_event_p4.c47
-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/cpu/vmware.c2
-rw-r--r--arch/x86/kernel/crash_dump_32.c3
-rw-r--r--arch/x86/kernel/crash_dump_64.c3
-rw-r--r--arch/x86/kernel/devicetree.c439
-rw-r--r--arch/x86/kernel/dumpstack.c51
-rw-r--r--arch/x86/kernel/dumpstack_32.c15
-rw-r--r--arch/x86/kernel/dumpstack_64.c14
-rw-r--r--arch/x86/kernel/e820.c19
-rw-r--r--arch/x86/kernel/early-quirks.c21
-rw-r--r--arch/x86/kernel/entry_32.S13
-rw-r--r--arch/x86/kernel/entry_64.S17
-rw-r--r--arch/x86/kernel/ftrace.c15
-rw-r--r--arch/x86/kernel/head32.c9
-rw-r--r--arch/x86/kernel/head64.c3
-rw-r--r--arch/x86/kernel/head_32.S10
-rw-r--r--arch/x86/kernel/head_64.S3
-rw-r--r--arch/x86/kernel/hpet.c2
-rw-r--r--arch/x86/kernel/i387.c2
-rw-r--r--arch/x86/kernel/i8237.c30
-rw-r--r--arch/x86/kernel/i8253.c76
-rw-r--r--arch/x86/kernel/i8259.c35
-rw-r--r--arch/x86/kernel/ioport.c20
-rw-r--r--arch/x86/kernel/irq.c92
-rw-r--r--arch/x86/kernel/irq_32.c2
-rw-r--r--arch/x86/kernel/irqinit.c92
-rw-r--r--arch/x86/kernel/kgdb.c15
-rw-r--r--arch/x86/kernel/kprobes.c8
-rw-r--r--arch/x86/kernel/kvm.c2
-rw-r--r--arch/x86/kernel/mca_32.c2
-rw-r--r--arch/x86/kernel/microcode_amd.c188
-rw-r--r--arch/x86/kernel/microcode_core.c41
-rw-r--r--arch/x86/kernel/mpparse.c12
-rw-r--r--arch/x86/kernel/pci-calgary_64.c4
-rw-r--r--arch/x86/kernel/pci-gart_64.c41
-rw-r--r--arch/x86/kernel/process.c11
-rw-r--r--arch/x86/kernel/process_64.c8
-rw-r--r--arch/x86/kernel/ptrace.c36
-rw-r--r--arch/x86/kernel/reboot.c129
-rw-r--r--arch/x86/kernel/reboot_32.S135
-rw-r--r--arch/x86/kernel/rtc.c3
-rw-r--r--arch/x86/kernel/setup.c93
-rw-r--r--arch/x86/kernel/setup_percpu.c11
-rw-r--r--arch/x86/kernel/smpboot.c134
-rw-r--r--arch/x86/kernel/stacktrace.c6
-rw-r--r--arch/x86/kernel/step.c2
-rw-r--r--arch/x86/kernel/syscall_table_32.S4
-rw-r--r--arch/x86/kernel/topology.c2
-rw-r--r--arch/x86/kernel/trampoline.c42
-rw-r--r--arch/x86/kernel/trampoline_32.S15
-rw-r--r--arch/x86/kernel/trampoline_64.S28
-rw-r--r--arch/x86/kernel/tsc.c4
-rw-r--r--arch/x86/kernel/verify_cpu.S2
-rw-r--r--arch/x86/kernel/vmlinux.lds.S18
-rw-r--r--arch/x86/kernel/x8664_ksyms_64.c1
-rw-r--r--arch/x86/kernel/x86_init.c1
-rw-r--r--arch/x86/kernel/xsave.c2
110 files changed, 2958 insertions, 1908 deletions
diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile
index 34244b2cd880..7338ef2218bc 100644
--- a/arch/x86/kernel/Makefile
+++ b/arch/x86/kernel/Makefile
@@ -41,13 +41,13 @@ obj-$(CONFIG_X86_32) += sys_i386_32.o i386_ksyms_32.o
41obj-$(CONFIG_X86_64) += sys_x86_64.o x8664_ksyms_64.o 41obj-$(CONFIG_X86_64) += sys_x86_64.o x8664_ksyms_64.o
42obj-$(CONFIG_X86_64) += syscall_64.o vsyscall_64.o 42obj-$(CONFIG_X86_64) += syscall_64.o vsyscall_64.o
43obj-y += bootflag.o e820.o 43obj-y += bootflag.o e820.o
44obj-y += pci-dma.o quirks.o i8237.o topology.o kdebugfs.o 44obj-y += pci-dma.o quirks.o topology.o kdebugfs.o
45obj-y += alternative.o i8253.o pci-nommu.o hw_breakpoint.o 45obj-y += alternative.o i8253.o pci-nommu.o hw_breakpoint.o
46obj-y += tsc.o io_delay.o rtc.o 46obj-y += tsc.o io_delay.o rtc.o
47obj-y += pci-iommu_table.o 47obj-y += pci-iommu_table.o
48obj-y += resource.o 48obj-y += resource.o
49 49
50obj-$(CONFIG_X86_TRAMPOLINE) += trampoline.o 50obj-y += trampoline.o trampoline_$(BITS).o
51obj-y += process.o 51obj-y += process.o
52obj-y += i387.o xsave.o 52obj-y += i387.o xsave.o
53obj-y += ptrace.o 53obj-y += ptrace.o
@@ -55,10 +55,12 @@ obj-$(CONFIG_X86_32) += tls.o
55obj-$(CONFIG_IA32_EMULATION) += tls.o 55obj-$(CONFIG_IA32_EMULATION) += tls.o
56obj-y += step.o 56obj-y += step.o
57obj-$(CONFIG_INTEL_TXT) += tboot.o 57obj-$(CONFIG_INTEL_TXT) += tboot.o
58obj-$(CONFIG_ISA_DMA_API) += i8237.o
58obj-$(CONFIG_STACKTRACE) += stacktrace.o 59obj-$(CONFIG_STACKTRACE) += stacktrace.o
59obj-y += cpu/ 60obj-y += cpu/
60obj-y += acpi/ 61obj-y += acpi/
61obj-y += reboot.o 62obj-y += reboot.o
63obj-$(CONFIG_X86_32) += reboot_32.o
62obj-$(CONFIG_MCA) += mca_32.o 64obj-$(CONFIG_MCA) += mca_32.o
63obj-$(CONFIG_X86_MSR) += msr.o 65obj-$(CONFIG_X86_MSR) += msr.o
64obj-$(CONFIG_X86_CPUID) += cpuid.o 66obj-$(CONFIG_X86_CPUID) += cpuid.o
@@ -66,10 +68,9 @@ obj-$(CONFIG_PCI) += early-quirks.o
66apm-y := apm_32.o 68apm-y := apm_32.o
67obj-$(CONFIG_APM) += apm.o 69obj-$(CONFIG_APM) += apm.o
68obj-$(CONFIG_SMP) += smp.o 70obj-$(CONFIG_SMP) += smp.o
69obj-$(CONFIG_SMP) += smpboot.o tsc_sync.o 71obj-$(CONFIG_SMP) += smpboot.o
72obj-$(CONFIG_SMP) += tsc_sync.o
70obj-$(CONFIG_SMP) += setup_percpu.o 73obj-$(CONFIG_SMP) += setup_percpu.o
71obj-$(CONFIG_X86_64_SMP) += tsc_sync.o
72obj-$(CONFIG_X86_TRAMPOLINE) += trampoline_$(BITS).o
73obj-$(CONFIG_X86_MPPARSE) += mpparse.o 74obj-$(CONFIG_X86_MPPARSE) += mpparse.o
74obj-y += apic/ 75obj-y += apic/
75obj-$(CONFIG_X86_REBOOTFIXUPS) += reboot_fixups_32.o 76obj-$(CONFIG_X86_REBOOTFIXUPS) += reboot_fixups_32.o
@@ -109,6 +110,7 @@ obj-$(CONFIG_MICROCODE) += microcode.o
109obj-$(CONFIG_X86_CHECK_BIOS_CORRUPTION) += check.o 110obj-$(CONFIG_X86_CHECK_BIOS_CORRUPTION) += check.o
110 111
111obj-$(CONFIG_SWIOTLB) += pci-swiotlb.o 112obj-$(CONFIG_SWIOTLB) += pci-swiotlb.o
113obj-$(CONFIG_OF) += devicetree.o
112 114
113### 115###
114# 64 bit specific files 116# 64 bit specific files
diff --git a/arch/x86/kernel/acpi/boot.c b/arch/x86/kernel/acpi/boot.c
index b3a71137983a..9a966c579af5 100644
--- a/arch/x86/kernel/acpi/boot.c
+++ b/arch/x86/kernel/acpi/boot.c
@@ -72,6 +72,7 @@ u8 acpi_sci_flags __initdata;
72int acpi_sci_override_gsi __initdata; 72int acpi_sci_override_gsi __initdata;
73int acpi_skip_timer_override __initdata; 73int acpi_skip_timer_override __initdata;
74int acpi_use_timer_override __initdata; 74int acpi_use_timer_override __initdata;
75int acpi_fix_pin2_polarity __initdata;
75 76
76#ifdef CONFIG_X86_LOCAL_APIC 77#ifdef CONFIG_X86_LOCAL_APIC
77static u64 acpi_lapic_addr __initdata = APIC_DEFAULT_PHYS_BASE; 78static u64 acpi_lapic_addr __initdata = APIC_DEFAULT_PHYS_BASE;
@@ -415,10 +416,15 @@ acpi_parse_int_src_ovr(struct acpi_subtable_header * header,
415 return 0; 416 return 0;
416 } 417 }
417 418
418 if (acpi_skip_timer_override && 419 if (intsrc->source_irq == 0 && intsrc->global_irq == 2) {
419 intsrc->source_irq == 0 && intsrc->global_irq == 2) { 420 if (acpi_skip_timer_override) {
420 printk(PREFIX "BIOS IRQ0 pin2 override ignored.\n"); 421 printk(PREFIX "BIOS IRQ0 pin2 override ignored.\n");
421 return 0; 422 return 0;
423 }
424 if (acpi_fix_pin2_polarity && (intsrc->inti_flags & ACPI_MADT_POLARITY_MASK)) {
425 intsrc->inti_flags &= ~ACPI_MADT_POLARITY_MASK;
426 printk(PREFIX "BIOS IRQ0 pin2 override: forcing polarity to high active.\n");
427 }
422 } 428 }
423 429
424 mp_override_legacy_irq(intsrc->source_irq, 430 mp_override_legacy_irq(intsrc->source_irq,
@@ -589,14 +595,8 @@ static void acpi_map_cpu2node(acpi_handle handle, int cpu, int physid)
589 nid = acpi_get_node(handle); 595 nid = acpi_get_node(handle);
590 if (nid == -1 || !node_online(nid)) 596 if (nid == -1 || !node_online(nid))
591 return; 597 return;
592#ifdef CONFIG_X86_64 598 set_apicid_to_node(physid, nid);
593 apicid_to_node[physid] = nid;
594 numa_set_node(cpu, nid); 599 numa_set_node(cpu, nid);
595#else /* CONFIG_X86_32 */
596 apicid_2_node[physid] = nid;
597 cpu_to_node_map[cpu] = nid;
598#endif
599
600#endif 600#endif
601} 601}
602 602
diff --git a/arch/x86/kernel/acpi/realmode/wakeup.S b/arch/x86/kernel/acpi/realmode/wakeup.S
index 28595d6df47c..ead21b663117 100644
--- a/arch/x86/kernel/acpi/realmode/wakeup.S
+++ b/arch/x86/kernel/acpi/realmode/wakeup.S
@@ -6,11 +6,17 @@
6#include <asm/page_types.h> 6#include <asm/page_types.h>
7#include <asm/pgtable_types.h> 7#include <asm/pgtable_types.h>
8#include <asm/processor-flags.h> 8#include <asm/processor-flags.h>
9#include "wakeup.h"
9 10
10 .code16 11 .code16
11 .section ".header", "a" 12 .section ".jump", "ax"
13 .globl _start
14_start:
15 cli
16 jmp wakeup_code
12 17
13/* This should match the structure in wakeup.h */ 18/* This should match the structure in wakeup.h */
19 .section ".header", "a"
14 .globl wakeup_header 20 .globl wakeup_header
15wakeup_header: 21wakeup_header:
16video_mode: .short 0 /* Video mode number */ 22video_mode: .short 0 /* Video mode number */
@@ -30,14 +36,11 @@ wakeup_jmp: .byte 0xea /* ljmpw */
30wakeup_jmp_off: .word 3f 36wakeup_jmp_off: .word 3f
31wakeup_jmp_seg: .word 0 37wakeup_jmp_seg: .word 0
32wakeup_gdt: .quad 0, 0, 0 38wakeup_gdt: .quad 0, 0, 0
33signature: .long 0x51ee1111 39signature: .long WAKEUP_HEADER_SIGNATURE
34 40
35 .text 41 .text
36 .globl _start
37 .code16 42 .code16
38wakeup_code: 43wakeup_code:
39_start:
40 cli
41 cld 44 cld
42 45
43 /* Apparently some dimwit BIOS programmers don't know how to 46 /* Apparently some dimwit BIOS programmers don't know how to
@@ -77,12 +80,12 @@ _start:
77 80
78 /* Check header signature... */ 81 /* Check header signature... */
79 movl signature, %eax 82 movl signature, %eax
80 cmpl $0x51ee1111, %eax 83 cmpl $WAKEUP_HEADER_SIGNATURE, %eax
81 jne bogus_real_magic 84 jne bogus_real_magic
82 85
83 /* Check we really have everything... */ 86 /* Check we really have everything... */
84 movl end_signature, %eax 87 movl end_signature, %eax
85 cmpl $0x65a22c82, %eax 88 cmpl $WAKEUP_END_SIGNATURE, %eax
86 jne bogus_real_magic 89 jne bogus_real_magic
87 90
88 /* Call the C code */ 91 /* Call the C code */
@@ -147,3 +150,7 @@ wakeup_heap:
147wakeup_stack: 150wakeup_stack:
148 .space 2048 151 .space 2048
149wakeup_stack_end: 152wakeup_stack_end:
153
154 .section ".signature","a"
155end_signature:
156 .long WAKEUP_END_SIGNATURE
diff --git a/arch/x86/kernel/acpi/realmode/wakeup.h b/arch/x86/kernel/acpi/realmode/wakeup.h
index 69d38d0b2b64..e1828c07e79c 100644
--- a/arch/x86/kernel/acpi/realmode/wakeup.h
+++ b/arch/x86/kernel/acpi/realmode/wakeup.h
@@ -35,7 +35,8 @@ struct wakeup_header {
35extern struct wakeup_header wakeup_header; 35extern struct wakeup_header wakeup_header;
36#endif 36#endif
37 37
38#define HEADER_OFFSET 0x3f00 38#define WAKEUP_HEADER_OFFSET 8
39#define WAKEUP_SIZE 0x4000 39#define WAKEUP_HEADER_SIGNATURE 0x51ee1111
40#define WAKEUP_END_SIGNATURE 0x65a22c82
40 41
41#endif /* ARCH_X86_KERNEL_ACPI_RM_WAKEUP_H */ 42#endif /* ARCH_X86_KERNEL_ACPI_RM_WAKEUP_H */
diff --git a/arch/x86/kernel/acpi/realmode/wakeup.lds.S b/arch/x86/kernel/acpi/realmode/wakeup.lds.S
index 060fff8f5c5b..d4f8010a5b1b 100644
--- a/arch/x86/kernel/acpi/realmode/wakeup.lds.S
+++ b/arch/x86/kernel/acpi/realmode/wakeup.lds.S
@@ -13,9 +13,19 @@ ENTRY(_start)
13SECTIONS 13SECTIONS
14{ 14{
15 . = 0; 15 . = 0;
16 .jump : {
17 *(.jump)
18 } = 0x90909090
19
20 . = WAKEUP_HEADER_OFFSET;
21 .header : {
22 *(.header)
23 }
24
25 . = ALIGN(16);
16 .text : { 26 .text : {
17 *(.text*) 27 *(.text*)
18 } 28 } = 0x90909090
19 29
20 . = ALIGN(16); 30 . = ALIGN(16);
21 .rodata : { 31 .rodata : {
@@ -33,11 +43,6 @@ SECTIONS
33 *(.data*) 43 *(.data*)
34 } 44 }
35 45
36 .signature : {
37 end_signature = .;
38 LONG(0x65a22c82)
39 }
40
41 . = ALIGN(16); 46 . = ALIGN(16);
42 .bss : { 47 .bss : {
43 __bss_start = .; 48 __bss_start = .;
@@ -45,20 +50,13 @@ SECTIONS
45 __bss_end = .; 50 __bss_end = .;
46 } 51 }
47 52
48 . = HEADER_OFFSET; 53 .signature : {
49 .header : { 54 *(.signature)
50 *(.header)
51 } 55 }
52 56
53 . = ALIGN(16);
54 _end = .; 57 _end = .;
55 58
56 /DISCARD/ : { 59 /DISCARD/ : {
57 *(.note*) 60 *(.note*)
58 } 61 }
59
60 /*
61 * The ASSERT() sink to . is intentional, for binutils 2.14 compatibility:
62 */
63 . = ASSERT(_end <= WAKEUP_SIZE, "Wakeup too big!");
64} 62}
diff --git a/arch/x86/kernel/acpi/sleep.c b/arch/x86/kernel/acpi/sleep.c
index 68d1537b8c81..ff93bc1b09c3 100644
--- a/arch/x86/kernel/acpi/sleep.c
+++ b/arch/x86/kernel/acpi/sleep.c
@@ -18,37 +18,28 @@
18#include "realmode/wakeup.h" 18#include "realmode/wakeup.h"
19#include "sleep.h" 19#include "sleep.h"
20 20
21unsigned long acpi_wakeup_address;
22unsigned long acpi_realmode_flags; 21unsigned long acpi_realmode_flags;
23 22
24/* address in low memory of the wakeup routine. */
25static unsigned long acpi_realmode;
26
27#if defined(CONFIG_SMP) && defined(CONFIG_64BIT) 23#if defined(CONFIG_SMP) && defined(CONFIG_64BIT)
28static char temp_stack[4096]; 24static char temp_stack[4096];
29#endif 25#endif
30 26
31/** 27/**
32 * acpi_save_state_mem - save kernel state 28 * acpi_suspend_lowlevel - save kernel state
33 * 29 *
34 * Create an identity mapped page table and copy the wakeup routine to 30 * Create an identity mapped page table and copy the wakeup routine to
35 * low memory. 31 * low memory.
36 *
37 * Note that this is too late to change acpi_wakeup_address.
38 */ 32 */
39int acpi_save_state_mem(void) 33int acpi_suspend_lowlevel(void)
40{ 34{
41 struct wakeup_header *header; 35 struct wakeup_header *header;
36 /* address in low memory of the wakeup routine. */
37 char *acpi_realmode;
42 38
43 if (!acpi_realmode) { 39 acpi_realmode = TRAMPOLINE_SYM(acpi_wakeup_code);
44 printk(KERN_ERR "Could not allocate memory during boot, "
45 "S3 disabled\n");
46 return -ENOMEM;
47 }
48 memcpy((void *)acpi_realmode, &wakeup_code_start, WAKEUP_SIZE);
49 40
50 header = (struct wakeup_header *)(acpi_realmode + HEADER_OFFSET); 41 header = (struct wakeup_header *)(acpi_realmode + WAKEUP_HEADER_OFFSET);
51 if (header->signature != 0x51ee1111) { 42 if (header->signature != WAKEUP_HEADER_SIGNATURE) {
52 printk(KERN_ERR "wakeup header does not match\n"); 43 printk(KERN_ERR "wakeup header does not match\n");
53 return -EINVAL; 44 return -EINVAL;
54 } 45 }
@@ -68,9 +59,7 @@ int acpi_save_state_mem(void)
68 /* GDT[0]: GDT self-pointer */ 59 /* GDT[0]: GDT self-pointer */
69 header->wakeup_gdt[0] = 60 header->wakeup_gdt[0] =
70 (u64)(sizeof(header->wakeup_gdt) - 1) + 61 (u64)(sizeof(header->wakeup_gdt) - 1) +
71 ((u64)(acpi_wakeup_address + 62 ((u64)__pa(&header->wakeup_gdt) << 16);
72 ((char *)&header->wakeup_gdt - (char *)acpi_realmode))
73 << 16);
74 /* GDT[1]: big real mode-like code segment */ 63 /* GDT[1]: big real mode-like code segment */
75 header->wakeup_gdt[1] = 64 header->wakeup_gdt[1] =
76 GDT_ENTRY(0x809b, acpi_wakeup_address, 0xfffff); 65 GDT_ENTRY(0x809b, acpi_wakeup_address, 0xfffff);
@@ -96,7 +85,7 @@ int acpi_save_state_mem(void)
96 header->pmode_cr3 = (u32)__pa(&initial_page_table); 85 header->pmode_cr3 = (u32)__pa(&initial_page_table);
97 saved_magic = 0x12345678; 86 saved_magic = 0x12345678;
98#else /* CONFIG_64BIT */ 87#else /* CONFIG_64BIT */
99 header->trampoline_segment = setup_trampoline() >> 4; 88 header->trampoline_segment = trampoline_address() >> 4;
100#ifdef CONFIG_SMP 89#ifdef CONFIG_SMP
101 stack_start = (unsigned long)temp_stack + sizeof(temp_stack); 90 stack_start = (unsigned long)temp_stack + sizeof(temp_stack);
102 early_gdt_descr.address = 91 early_gdt_descr.address =
@@ -107,56 +96,10 @@ int acpi_save_state_mem(void)
107 saved_magic = 0x123456789abcdef0L; 96 saved_magic = 0x123456789abcdef0L;
108#endif /* CONFIG_64BIT */ 97#endif /* CONFIG_64BIT */
109 98
99 do_suspend_lowlevel();
110 return 0; 100 return 0;
111} 101}
112 102
113/*
114 * acpi_restore_state - undo effects of acpi_save_state_mem
115 */
116void acpi_restore_state_mem(void)
117{
118}
119
120
121/**
122 * acpi_reserve_wakeup_memory - do _very_ early ACPI initialisation
123 *
124 * We allocate a page from the first 1MB of memory for the wakeup
125 * routine for when we come back from a sleep state. The
126 * runtime allocator allows specification of <16MB pages, but not
127 * <1MB pages.
128 */
129void __init acpi_reserve_wakeup_memory(void)
130{
131 phys_addr_t mem;
132
133 if ((&wakeup_code_end - &wakeup_code_start) > WAKEUP_SIZE) {
134 printk(KERN_ERR
135 "ACPI: Wakeup code way too big, S3 disabled.\n");
136 return;
137 }
138
139 mem = memblock_find_in_range(0, 1<<20, WAKEUP_SIZE, PAGE_SIZE);
140
141 if (mem == MEMBLOCK_ERROR) {
142 printk(KERN_ERR "ACPI: Cannot allocate lowmem, S3 disabled.\n");
143 return;
144 }
145 acpi_realmode = (unsigned long) phys_to_virt(mem);
146 acpi_wakeup_address = mem;
147 memblock_x86_reserve_range(mem, mem + WAKEUP_SIZE, "ACPI WAKEUP");
148}
149
150int __init acpi_configure_wakeup_memory(void)
151{
152 if (acpi_realmode)
153 set_memory_x(acpi_realmode, WAKEUP_SIZE >> PAGE_SHIFT);
154
155 return 0;
156}
157arch_initcall(acpi_configure_wakeup_memory);
158
159
160static int __init acpi_sleep_setup(char *str) 103static int __init acpi_sleep_setup(char *str)
161{ 104{
162 while ((str != NULL) && (*str != '\0')) { 105 while ((str != NULL) && (*str != '\0')) {
diff --git a/arch/x86/kernel/acpi/sleep.h b/arch/x86/kernel/acpi/sleep.h
index adbcbaa6f1df..416d4be13fef 100644
--- a/arch/x86/kernel/acpi/sleep.h
+++ b/arch/x86/kernel/acpi/sleep.h
@@ -4,13 +4,12 @@
4 4
5#include <asm/trampoline.h> 5#include <asm/trampoline.h>
6 6
7extern char wakeup_code_start, wakeup_code_end;
8
9extern unsigned long saved_video_mode; 7extern unsigned long saved_video_mode;
10extern long saved_magic; 8extern long saved_magic;
11 9
12extern int wakeup_pmode_return; 10extern int wakeup_pmode_return;
13extern char swsusp_pg_dir[PAGE_SIZE];
14 11
15extern unsigned long acpi_copy_wakeup_routine(unsigned long); 12extern unsigned long acpi_copy_wakeup_routine(unsigned long);
16extern void wakeup_long64(void); 13extern void wakeup_long64(void);
14
15extern void do_suspend_lowlevel(void);
diff --git a/arch/x86/kernel/acpi/wakeup_rm.S b/arch/x86/kernel/acpi/wakeup_rm.S
index 6ff3b5730575..63b8ab524f2c 100644
--- a/arch/x86/kernel/acpi/wakeup_rm.S
+++ b/arch/x86/kernel/acpi/wakeup_rm.S
@@ -2,9 +2,11 @@
2 * Wrapper script for the realmode binary as a transport object 2 * Wrapper script for the realmode binary as a transport object
3 * before copying to low memory. 3 * before copying to low memory.
4 */ 4 */
5 .section ".rodata","a" 5#include <asm/page_types.h>
6 .globl wakeup_code_start, wakeup_code_end 6
7wakeup_code_start: 7 .section ".x86_trampoline","a"
8 .balign PAGE_SIZE
9 .globl acpi_wakeup_code
10acpi_wakeup_code:
8 .incbin "arch/x86/kernel/acpi/realmode/wakeup.bin" 11 .incbin "arch/x86/kernel/acpi/realmode/wakeup.bin"
9wakeup_code_end: 12 .size acpi_wakeup_code, .-acpi_wakeup_code
10 .size wakeup_code_start, .-wakeup_code_start
diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
index 7038b95d363f..4a234677e213 100644
--- a/arch/x86/kernel/alternative.c
+++ b/arch/x86/kernel/alternative.c
@@ -199,7 +199,7 @@ void *text_poke_early(void *addr, const void *opcode, size_t len);
199 199
200/* Replace instructions with better alternatives for this CPU type. 200/* Replace instructions with better alternatives for this CPU type.
201 This runs before SMP is initialized to avoid SMP problems with 201 This runs before SMP is initialized to avoid SMP problems with
202 self modifying code. This implies that assymetric systems where 202 self modifying code. This implies that asymmetric systems where
203 APs have less capabilities than the boot processor are not handled. 203 APs have less capabilities than the boot processor are not handled.
204 Tough. Make sure you disable such features by hand. */ 204 Tough. Make sure you disable such features by hand. */
205 205
@@ -620,7 +620,12 @@ static int __kprobes stop_machine_text_poke(void *data)
620 flush_icache_range((unsigned long)p->addr, 620 flush_icache_range((unsigned long)p->addr,
621 (unsigned long)p->addr + p->len); 621 (unsigned long)p->addr + p->len);
622 } 622 }
623 623 /*
624 * Intel Archiecture Software Developer's Manual section 7.1.3 specifies
625 * that a core serializing instruction such as "cpuid" should be
626 * executed on _each_ core before the new instruction is made visible.
627 */
628 sync_core();
624 return 0; 629 return 0;
625} 630}
626 631
diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c
index 6e11c8134158..246d727b65b7 100644
--- a/arch/x86/kernel/amd_iommu_init.c
+++ b/arch/x86/kernel/amd_iommu_init.c
@@ -21,7 +21,7 @@
21#include <linux/acpi.h> 21#include <linux/acpi.h>
22#include <linux/list.h> 22#include <linux/list.h>
23#include <linux/slab.h> 23#include <linux/slab.h>
24#include <linux/sysdev.h> 24#include <linux/syscore_ops.h>
25#include <linux/interrupt.h> 25#include <linux/interrupt.h>
26#include <linux/msi.h> 26#include <linux/msi.h>
27#include <asm/pci-direct.h> 27#include <asm/pci-direct.h>
@@ -1260,7 +1260,7 @@ static void disable_iommus(void)
1260 * disable suspend until real resume implemented 1260 * disable suspend until real resume implemented
1261 */ 1261 */
1262 1262
1263static int amd_iommu_resume(struct sys_device *dev) 1263static void amd_iommu_resume(void)
1264{ 1264{
1265 struct amd_iommu *iommu; 1265 struct amd_iommu *iommu;
1266 1266
@@ -1276,11 +1276,9 @@ static int amd_iommu_resume(struct sys_device *dev)
1276 */ 1276 */
1277 amd_iommu_flush_all_devices(); 1277 amd_iommu_flush_all_devices();
1278 amd_iommu_flush_all_domains(); 1278 amd_iommu_flush_all_domains();
1279
1280 return 0;
1281} 1279}
1282 1280
1283static int amd_iommu_suspend(struct sys_device *dev, pm_message_t state) 1281static int amd_iommu_suspend(void)
1284{ 1282{
1285 /* disable IOMMUs to go out of the way for BIOS */ 1283 /* disable IOMMUs to go out of the way for BIOS */
1286 disable_iommus(); 1284 disable_iommus();
@@ -1288,17 +1286,11 @@ static int amd_iommu_suspend(struct sys_device *dev, pm_message_t state)
1288 return 0; 1286 return 0;
1289} 1287}
1290 1288
1291static struct sysdev_class amd_iommu_sysdev_class = { 1289static struct syscore_ops amd_iommu_syscore_ops = {
1292 .name = "amd_iommu",
1293 .suspend = amd_iommu_suspend, 1290 .suspend = amd_iommu_suspend,
1294 .resume = amd_iommu_resume, 1291 .resume = amd_iommu_resume,
1295}; 1292};
1296 1293
1297static struct sys_device device_amd_iommu = {
1298 .id = 0,
1299 .cls = &amd_iommu_sysdev_class,
1300};
1301
1302/* 1294/*
1303 * This is the core init function for AMD IOMMU hardware in the system. 1295 * This is the core init function for AMD IOMMU hardware in the system.
1304 * This function is called from the generic x86 DMA layer initialization 1296 * This function is called from the generic x86 DMA layer initialization
@@ -1415,14 +1407,6 @@ static int __init amd_iommu_init(void)
1415 goto free; 1407 goto free;
1416 } 1408 }
1417 1409
1418 ret = sysdev_class_register(&amd_iommu_sysdev_class);
1419 if (ret)
1420 goto free;
1421
1422 ret = sysdev_register(&device_amd_iommu);
1423 if (ret)
1424 goto free;
1425
1426 ret = amd_iommu_init_devices(); 1410 ret = amd_iommu_init_devices();
1427 if (ret) 1411 if (ret)
1428 goto free; 1412 goto free;
@@ -1441,6 +1425,8 @@ static int __init amd_iommu_init(void)
1441 1425
1442 amd_iommu_init_notifier(); 1426 amd_iommu_init_notifier();
1443 1427
1428 register_syscore_ops(&amd_iommu_syscore_ops);
1429
1444 if (iommu_pass_through) 1430 if (iommu_pass_through)
1445 goto out; 1431 goto out;
1446 1432
diff --git a/arch/x86/kernel/amd_nb.c b/arch/x86/kernel/amd_nb.c
index 0a99f7198bc3..4c39baa8facc 100644
--- a/arch/x86/kernel/amd_nb.c
+++ b/arch/x86/kernel/amd_nb.c
@@ -12,14 +12,19 @@
12 12
13static u32 *flush_words; 13static u32 *flush_words;
14 14
15struct pci_device_id amd_nb_misc_ids[] = { 15const struct pci_device_id amd_nb_misc_ids[] = {
16 { PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_K8_NB_MISC) }, 16 { PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_K8_NB_MISC) },
17 { PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_10H_NB_MISC) }, 17 { PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_10H_NB_MISC) },
18 { PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_15H_NB_MISC) }, 18 { PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_15H_NB_F3) },
19 {} 19 {}
20}; 20};
21EXPORT_SYMBOL(amd_nb_misc_ids); 21EXPORT_SYMBOL(amd_nb_misc_ids);
22 22
23static struct pci_device_id amd_nb_link_ids[] = {
24 { PCI_DEVICE(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_15H_NB_F4) },
25 {}
26};
27
23const struct amd_nb_bus_dev_range amd_nb_bus_dev_ranges[] __initconst = { 28const struct amd_nb_bus_dev_range amd_nb_bus_dev_ranges[] __initconst = {
24 { 0x00, 0x18, 0x20 }, 29 { 0x00, 0x18, 0x20 },
25 { 0xff, 0x00, 0x20 }, 30 { 0xff, 0x00, 0x20 },
@@ -31,7 +36,7 @@ struct amd_northbridge_info amd_northbridges;
31EXPORT_SYMBOL(amd_northbridges); 36EXPORT_SYMBOL(amd_northbridges);
32 37
33static struct pci_dev *next_northbridge(struct pci_dev *dev, 38static struct pci_dev *next_northbridge(struct pci_dev *dev,
34 struct pci_device_id *ids) 39 const struct pci_device_id *ids)
35{ 40{
36 do { 41 do {
37 dev = pci_get_device(PCI_ANY_ID, PCI_ANY_ID, dev); 42 dev = pci_get_device(PCI_ANY_ID, PCI_ANY_ID, dev);
@@ -43,9 +48,9 @@ static struct pci_dev *next_northbridge(struct pci_dev *dev,
43 48
44int amd_cache_northbridges(void) 49int amd_cache_northbridges(void)
45{ 50{
46 int i = 0; 51 u16 i = 0;
47 struct amd_northbridge *nb; 52 struct amd_northbridge *nb;
48 struct pci_dev *misc; 53 struct pci_dev *misc, *link;
49 54
50 if (amd_nb_num()) 55 if (amd_nb_num())
51 return 0; 56 return 0;
@@ -64,10 +69,12 @@ int amd_cache_northbridges(void)
64 amd_northbridges.nb = nb; 69 amd_northbridges.nb = nb;
65 amd_northbridges.num = i; 70 amd_northbridges.num = i;
66 71
67 misc = NULL; 72 link = misc = NULL;
68 for (i = 0; i != amd_nb_num(); i++) { 73 for (i = 0; i != amd_nb_num(); i++) {
69 node_to_amd_nb(i)->misc = misc = 74 node_to_amd_nb(i)->misc = misc =
70 next_northbridge(misc, amd_nb_misc_ids); 75 next_northbridge(misc, amd_nb_misc_ids);
76 node_to_amd_nb(i)->link = link =
77 next_northbridge(link, amd_nb_link_ids);
71 } 78 }
72 79
73 /* some CPU families (e.g. family 0x11) do not support GART */ 80 /* some CPU families (e.g. family 0x11) do not support GART */
@@ -85,26 +92,95 @@ int amd_cache_northbridges(void)
85 boot_cpu_data.x86_mask >= 0x1)) 92 boot_cpu_data.x86_mask >= 0x1))
86 amd_northbridges.flags |= AMD_NB_L3_INDEX_DISABLE; 93 amd_northbridges.flags |= AMD_NB_L3_INDEX_DISABLE;
87 94
95 if (boot_cpu_data.x86 == 0x15)
96 amd_northbridges.flags |= AMD_NB_L3_INDEX_DISABLE;
97
98 /* L3 cache partitioning is supported on family 0x15 */
99 if (boot_cpu_data.x86 == 0x15)
100 amd_northbridges.flags |= AMD_NB_L3_PARTITIONING;
101
88 return 0; 102 return 0;
89} 103}
90EXPORT_SYMBOL_GPL(amd_cache_northbridges); 104EXPORT_SYMBOL_GPL(amd_cache_northbridges);
91 105
92/* Ignores subdevice/subvendor but as far as I can figure out 106/*
93 they're useless anyways */ 107 * Ignores subdevice/subvendor but as far as I can figure out
94int __init early_is_amd_nb(u32 device) 108 * they're useless anyways
109 */
110bool __init early_is_amd_nb(u32 device)
95{ 111{
96 struct pci_device_id *id; 112 const struct pci_device_id *id;
97 u32 vendor = device & 0xffff; 113 u32 vendor = device & 0xffff;
114
98 device >>= 16; 115 device >>= 16;
99 for (id = amd_nb_misc_ids; id->vendor; id++) 116 for (id = amd_nb_misc_ids; id->vendor; id++)
100 if (vendor == id->vendor && device == id->device) 117 if (vendor == id->vendor && device == id->device)
101 return 1; 118 return true;
119 return false;
120}
121
122int amd_get_subcaches(int cpu)
123{
124 struct pci_dev *link = node_to_amd_nb(amd_get_nb_id(cpu))->link;
125 unsigned int mask;
126 int cuid = 0;
127
128 if (!amd_nb_has_feature(AMD_NB_L3_PARTITIONING))
129 return 0;
130
131 pci_read_config_dword(link, 0x1d4, &mask);
132
133#ifdef CONFIG_SMP
134 cuid = cpu_data(cpu).compute_unit_id;
135#endif
136 return (mask >> (4 * cuid)) & 0xf;
137}
138
139int amd_set_subcaches(int cpu, int mask)
140{
141 static unsigned int reset, ban;
142 struct amd_northbridge *nb = node_to_amd_nb(amd_get_nb_id(cpu));
143 unsigned int reg;
144 int cuid = 0;
145
146 if (!amd_nb_has_feature(AMD_NB_L3_PARTITIONING) || mask > 0xf)
147 return -EINVAL;
148
149 /* if necessary, collect reset state of L3 partitioning and BAN mode */
150 if (reset == 0) {
151 pci_read_config_dword(nb->link, 0x1d4, &reset);
152 pci_read_config_dword(nb->misc, 0x1b8, &ban);
153 ban &= 0x180000;
154 }
155
156 /* deactivate BAN mode if any subcaches are to be disabled */
157 if (mask != 0xf) {
158 pci_read_config_dword(nb->misc, 0x1b8, &reg);
159 pci_write_config_dword(nb->misc, 0x1b8, reg & ~0x180000);
160 }
161
162#ifdef CONFIG_SMP
163 cuid = cpu_data(cpu).compute_unit_id;
164#endif
165 mask <<= 4 * cuid;
166 mask |= (0xf ^ (1 << cuid)) << 26;
167
168 pci_write_config_dword(nb->link, 0x1d4, mask);
169
170 /* reset BAN mode if L3 partitioning returned to reset state */
171 pci_read_config_dword(nb->link, 0x1d4, &reg);
172 if (reg == reset) {
173 pci_read_config_dword(nb->misc, 0x1b8, &reg);
174 reg &= ~0x180000;
175 pci_write_config_dword(nb->misc, 0x1b8, reg | ban);
176 }
177
102 return 0; 178 return 0;
103} 179}
104 180
105int amd_cache_gart(void) 181static int amd_cache_gart(void)
106{ 182{
107 int i; 183 u16 i;
108 184
109 if (!amd_nb_has_feature(AMD_NB_GART)) 185 if (!amd_nb_has_feature(AMD_NB_GART))
110 return 0; 186 return 0;
diff --git a/arch/x86/kernel/apb_timer.c b/arch/x86/kernel/apb_timer.c
index 29ebf5a3b192..289e92862fd9 100644
--- a/arch/x86/kernel/apb_timer.c
+++ b/arch/x86/kernel/apb_timer.c
@@ -283,7 +283,7 @@ static int __init apbt_clockevent_register(void)
283 memcpy(&adev->evt, &apbt_clockevent, sizeof(struct clock_event_device)); 283 memcpy(&adev->evt, &apbt_clockevent, sizeof(struct clock_event_device));
284 284
285 if (mrst_timer_options == MRST_TIMER_LAPIC_APBT) { 285 if (mrst_timer_options == MRST_TIMER_LAPIC_APBT) {
286 apbt_clockevent.rating = APBT_CLOCKEVENT_RATING - 100; 286 adev->evt.rating = APBT_CLOCKEVENT_RATING - 100;
287 global_clock_event = &adev->evt; 287 global_clock_event = &adev->evt;
288 printk(KERN_DEBUG "%s clockevent registered as global\n", 288 printk(KERN_DEBUG "%s clockevent registered as global\n",
289 global_clock_event->name); 289 global_clock_event->name);
@@ -315,7 +315,7 @@ static void apbt_setup_irq(struct apbt_dev *adev)
315 irq_modify_status(adev->irq, 0, IRQ_MOVE_PCNTXT); 315 irq_modify_status(adev->irq, 0, IRQ_MOVE_PCNTXT);
316 irq_set_affinity(adev->irq, cpumask_of(adev->cpu)); 316 irq_set_affinity(adev->irq, cpumask_of(adev->cpu));
317 /* APB timer irqs are set up as mp_irqs, timer is edge type */ 317 /* APB timer irqs are set up as mp_irqs, timer is edge type */
318 __set_irq_handler(adev->irq, handle_edge_irq, 0, "edge"); 318 __irq_set_handler(adev->irq, handle_edge_irq, 0, "edge");
319 319
320 if (system_state == SYSTEM_BOOTING) { 320 if (system_state == SYSTEM_BOOTING) {
321 if (request_irq(adev->irq, apbt_interrupt_handler, 321 if (request_irq(adev->irq, apbt_interrupt_handler,
@@ -507,64 +507,12 @@ static int apbt_next_event(unsigned long delta,
507 return 0; 507 return 0;
508} 508}
509 509
510/*
511 * APB timer clock is not in sync with pclk on Langwell, which translates to
512 * unreliable read value caused by sampling error. the error does not add up
513 * overtime and only happens when sampling a 0 as a 1 by mistake. so the time
514 * would go backwards. the following code is trying to prevent time traveling
515 * backwards. little bit paranoid.
516 */
517static cycle_t apbt_read_clocksource(struct clocksource *cs) 510static cycle_t apbt_read_clocksource(struct clocksource *cs)
518{ 511{
519 unsigned long t0, t1, t2; 512 unsigned long current_count;
520 static unsigned long last_read; 513
521 514 current_count = apbt_readl(phy_cs_timer_id, APBTMR_N_CURRENT_VALUE);
522bad_count: 515 return (cycle_t)~current_count;
523 t1 = apbt_readl(phy_cs_timer_id,
524 APBTMR_N_CURRENT_VALUE);
525 t2 = apbt_readl(phy_cs_timer_id,
526 APBTMR_N_CURRENT_VALUE);
527 if (unlikely(t1 < t2)) {
528 pr_debug("APBT: read current count error %lx:%lx:%lx\n",
529 t1, t2, t2 - t1);
530 goto bad_count;
531 }
532 /*
533 * check against cached last read, makes sure time does not go back.
534 * it could be a normal rollover but we will do tripple check anyway
535 */
536 if (unlikely(t2 > last_read)) {
537 /* check if we have a normal rollover */
538 unsigned long raw_intr_status =
539 apbt_readl_reg(APBTMRS_RAW_INT_STATUS);
540 /*
541 * cs timer interrupt is masked but raw intr bit is set if
542 * rollover occurs. then we read EOI reg to clear it.
543 */
544 if (raw_intr_status & (1 << phy_cs_timer_id)) {
545 apbt_readl(phy_cs_timer_id, APBTMR_N_EOI);
546 goto out;
547 }
548 pr_debug("APB CS going back %lx:%lx:%lx ",
549 t2, last_read, t2 - last_read);
550bad_count_x3:
551 pr_debug("triple check enforced\n");
552 t0 = apbt_readl(phy_cs_timer_id,
553 APBTMR_N_CURRENT_VALUE);
554 udelay(1);
555 t1 = apbt_readl(phy_cs_timer_id,
556 APBTMR_N_CURRENT_VALUE);
557 udelay(1);
558 t2 = apbt_readl(phy_cs_timer_id,
559 APBTMR_N_CURRENT_VALUE);
560 if ((t2 > t1) || (t1 > t0)) {
561 printk(KERN_ERR "Error: APB CS tripple check failed\n");
562 goto bad_count_x3;
563 }
564 }
565out:
566 last_read = t2;
567 return (cycle_t)~t2;
568} 516}
569 517
570static int apbt_clocksource_register(void) 518static int apbt_clocksource_register(void)
diff --git a/arch/x86/kernel/aperture_64.c b/arch/x86/kernel/aperture_64.c
index 5955a7800a96..73fb469908c6 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)
@@ -73,7 +73,7 @@ static u32 __init allocate_aperture(void)
73 /* 73 /*
74 * using 512M as goal, in case kexec will load kernel_big 74 * using 512M as goal, in case kexec will load kernel_big
75 * that will do the on position decompress, and could overlap with 75 * that will do the on position decompress, and could overlap with
76 * that positon with gart that is used. 76 * that position with gart that is used.
77 * sequende: 77 * sequende:
78 * kernel_small 78 * kernel_small
79 * ==> kexec (with kdump trigger path or previous doesn't shutdown gart) 79 * ==> kexec (with kdump trigger path or previous doesn't shutdown gart)
@@ -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
@@ -500,7 +499,7 @@ out:
500 * Don't enable translation yet but enable GART IO and CPU 499 * Don't enable translation yet but enable GART IO and CPU
501 * accesses and set DISTLBWALKPRB since GART table memory is UC. 500 * accesses and set DISTLBWALKPRB since GART table memory is UC.
502 */ 501 */
503 u32 ctl = DISTLBWALKPRB | aper_order << 1; 502 u32 ctl = aper_order << 1;
504 503
505 bus = amd_nb_bus_dev_ranges[i].bus; 504 bus = amd_nb_bus_dev_ranges[i].bus;
506 dev_base = amd_nb_bus_dev_ranges[i].dev_base; 505 dev_base = amd_nb_bus_dev_ranges[i].dev_base;
diff --git a/arch/x86/kernel/apic/apic.c b/arch/x86/kernel/apic/apic.c
index 76b96d74978a..fabf01eff771 100644
--- a/arch/x86/kernel/apic/apic.c
+++ b/arch/x86/kernel/apic/apic.c
@@ -24,7 +24,7 @@
24#include <linux/ftrace.h> 24#include <linux/ftrace.h>
25#include <linux/ioport.h> 25#include <linux/ioport.h>
26#include <linux/module.h> 26#include <linux/module.h>
27#include <linux/sysdev.h> 27#include <linux/syscore_ops.h>
28#include <linux/delay.h> 28#include <linux/delay.h>
29#include <linux/timex.h> 29#include <linux/timex.h>
30#include <linux/dmar.h> 30#include <linux/dmar.h>
@@ -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;
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);
1972 } 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
@@ -2036,7 +2046,7 @@ static struct {
2036 unsigned int apic_thmr; 2046 unsigned int apic_thmr;
2037} apic_pm_state; 2047} apic_pm_state;
2038 2048
2039static int lapic_suspend(struct sys_device *dev, pm_message_t state) 2049static int lapic_suspend(void)
2040{ 2050{
2041 unsigned long flags; 2051 unsigned long flags;
2042 int maxlvt; 2052 int maxlvt;
@@ -2074,23 +2084,21 @@ static int lapic_suspend(struct sys_device *dev, pm_message_t state)
2074 return 0; 2084 return 0;
2075} 2085}
2076 2086
2077static int lapic_resume(struct sys_device *dev) 2087static void lapic_resume(void)
2078{ 2088{
2079 unsigned int l, h; 2089 unsigned int l, h;
2080 unsigned long flags; 2090 unsigned long flags;
2081 int maxlvt; 2091 int maxlvt, ret;
2082 int ret = 0;
2083 struct IO_APIC_route_entry **ioapic_entries = NULL; 2092 struct IO_APIC_route_entry **ioapic_entries = NULL;
2084 2093
2085 if (!apic_pm_state.active) 2094 if (!apic_pm_state.active)
2086 return 0; 2095 return;
2087 2096
2088 local_irq_save(flags); 2097 local_irq_save(flags);
2089 if (intr_remapping_enabled) { 2098 if (intr_remapping_enabled) {
2090 ioapic_entries = alloc_ioapic_entries(); 2099 ioapic_entries = alloc_ioapic_entries();
2091 if (!ioapic_entries) { 2100 if (!ioapic_entries) {
2092 WARN(1, "Alloc ioapic_entries in lapic resume failed."); 2101 WARN(1, "Alloc ioapic_entries in lapic resume failed.");
2093 ret = -ENOMEM;
2094 goto restore; 2102 goto restore;
2095 } 2103 }
2096 2104
@@ -2152,8 +2160,6 @@ static int lapic_resume(struct sys_device *dev)
2152 } 2160 }
2153restore: 2161restore:
2154 local_irq_restore(flags); 2162 local_irq_restore(flags);
2155
2156 return ret;
2157} 2163}
2158 2164
2159/* 2165/*
@@ -2161,17 +2167,11 @@ restore:
2161 * are needed on every CPU up until machine_halt/restart/poweroff. 2167 * are needed on every CPU up until machine_halt/restart/poweroff.
2162 */ 2168 */
2163 2169
2164static struct sysdev_class lapic_sysclass = { 2170static struct syscore_ops lapic_syscore_ops = {
2165 .name = "lapic",
2166 .resume = lapic_resume, 2171 .resume = lapic_resume,
2167 .suspend = lapic_suspend, 2172 .suspend = lapic_suspend,
2168}; 2173};
2169 2174
2170static struct sys_device device_lapic = {
2171 .id = 0,
2172 .cls = &lapic_sysclass,
2173};
2174
2175static void __cpuinit apic_pm_activate(void) 2175static void __cpuinit apic_pm_activate(void)
2176{ 2176{
2177 apic_pm_state.active = 1; 2177 apic_pm_state.active = 1;
@@ -2179,16 +2179,11 @@ static void __cpuinit apic_pm_activate(void)
2179 2179
2180static int __init init_lapic_sysfs(void) 2180static int __init init_lapic_sysfs(void)
2181{ 2181{
2182 int error;
2183
2184 if (!cpu_has_apic)
2185 return 0;
2186 /* XXX: remove suspend/resume procs if !apic_pm_state.active? */ 2182 /* XXX: remove suspend/resume procs if !apic_pm_state.active? */
2183 if (cpu_has_apic)
2184 register_syscore_ops(&lapic_syscore_ops);
2187 2185
2188 error = sysdev_class_register(&lapic_sysclass); 2186 return 0;
2189 if (!error)
2190 error = sysdev_register(&device_lapic);
2191 return error;
2192} 2187}
2193 2188
2194/* local apic needs to resume before other devices access its registers. */ 2189/* local apic needs to resume before other devices access its registers. */
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..5260fe91bcb6 100644
--- a/arch/x86/kernel/apic/hw_nmi.c
+++ b/arch/x86/kernel/apic/hw_nmi.c
@@ -16,6 +16,7 @@
16#include <linux/kprobes.h> 16#include <linux/kprobes.h>
17#include <linux/nmi.h> 17#include <linux/nmi.h>
18#include <linux/module.h> 18#include <linux/module.h>
19#include <linux/delay.h>
19 20
20#ifdef CONFIG_HARDLOCKUP_DETECTOR 21#ifdef CONFIG_HARDLOCKUP_DETECTOR
21u64 hw_nmi_get_sample_period(void) 22u64 hw_nmi_get_sample_period(void)
@@ -83,7 +84,6 @@ arch_trigger_all_cpu_backtrace_handler(struct notifier_block *self,
83 arch_spin_lock(&lock); 84 arch_spin_lock(&lock);
84 printk(KERN_WARNING "NMI backtrace for cpu %d\n", cpu); 85 printk(KERN_WARNING "NMI backtrace for cpu %d\n", cpu);
85 show_regs(regs); 86 show_regs(regs);
86 dump_stack();
87 arch_spin_unlock(&lock); 87 arch_spin_unlock(&lock);
88 cpumask_clear_cpu(cpu, to_cpumask(backtrace_mask)); 88 cpumask_clear_cpu(cpu, to_cpumask(backtrace_mask));
89 return NOTIFY_STOP; 89 return NOTIFY_STOP;
diff --git a/arch/x86/kernel/apic/io_apic.c b/arch/x86/kernel/apic/io_apic.c
index ca9e2a3545a9..45fd33d1fd3a 100644
--- a/arch/x86/kernel/apic/io_apic.c
+++ b/arch/x86/kernel/apic/io_apic.c
@@ -30,7 +30,7 @@
30#include <linux/compiler.h> 30#include <linux/compiler.h>
31#include <linux/acpi.h> 31#include <linux/acpi.h>
32#include <linux/module.h> 32#include <linux/module.h>
33#include <linux/sysdev.h> 33#include <linux/syscore_ops.h>
34#include <linux/msi.h> 34#include <linux/msi.h>
35#include <linux/htirq.h> 35#include <linux/htirq.h>
36#include <linux/freezer.h> 36#include <linux/freezer.h>
@@ -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(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;
@@ -1916,7 +1886,7 @@ void disable_IO_APIC(void)
1916 * 1886 *
1917 * With interrupt-remapping, for now we will use virtual wire A mode, 1887 * With interrupt-remapping, for now we will use virtual wire A mode,
1918 * as virtual wire B is little complex (need to configure both 1888 * as virtual wire B is little complex (need to configure both
1919 * IOAPIC RTE aswell as interrupt-remapping table entry). 1889 * IOAPIC RTE as well as interrupt-remapping table entry).
1920 * As this gets called during crash dump, keep this simple for now. 1890 * As this gets called during crash dump, keep this simple for now.
1921 */ 1891 */
1922 if (ioapic_i8259.pin != -1 && !intr_remapping_enabled) { 1892 if (ioapic_i8259.pin != -1 && !intr_remapping_enabled) {
@@ -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;
@@ -2935,7 +2905,7 @@ void __init setup_IO_APIC(void)
2935} 2905}
2936 2906
2937/* 2907/*
2938 * Called after all the initialization is done. If we didnt find any 2908 * Called after all the initialization is done. If we didn't find any
2939 * APIC bugs then we can allow the modify fast path 2909 * APIC bugs then we can allow the modify fast path
2940 */ 2910 */
2941 2911
@@ -2948,89 +2918,84 @@ static int __init io_apic_bug_finalize(void)
2948 2918
2949late_initcall(io_apic_bug_finalize); 2919late_initcall(io_apic_bug_finalize);
2950 2920
2951struct sysfs_ioapic_data { 2921static struct IO_APIC_route_entry *ioapic_saved_data[MAX_IO_APICS];
2952 struct sys_device dev;
2953 struct IO_APIC_route_entry entry[0];
2954};
2955static struct sysfs_ioapic_data * mp_ioapic_data[MAX_IO_APICS];
2956 2922
2957static int ioapic_suspend(struct sys_device *dev, pm_message_t state) 2923static void suspend_ioapic(int ioapic_id)
2958{ 2924{
2959 struct IO_APIC_route_entry *entry; 2925 struct IO_APIC_route_entry *saved_data = ioapic_saved_data[ioapic_id];
2960 struct sysfs_ioapic_data *data;
2961 int i; 2926 int i;
2962 2927
2963 data = container_of(dev, struct sysfs_ioapic_data, dev); 2928 if (!saved_data)
2964 entry = data->entry; 2929 return;
2965 for (i = 0; i < nr_ioapic_registers[dev->id]; i ++, entry ++ ) 2930
2966 *entry = ioapic_read_entry(dev->id, i); 2931 for (i = 0; i < nr_ioapic_registers[ioapic_id]; i++)
2932 saved_data[i] = ioapic_read_entry(ioapic_id, i);
2933}
2934
2935static int ioapic_suspend(void)
2936{
2937 int ioapic_id;
2938
2939 for (ioapic_id = 0; ioapic_id < nr_ioapics; ioapic_id++)
2940 suspend_ioapic(ioapic_id);
2967 2941
2968 return 0; 2942 return 0;
2969} 2943}
2970 2944
2971static int ioapic_resume(struct sys_device *dev) 2945static void resume_ioapic(int ioapic_id)
2972{ 2946{
2973 struct IO_APIC_route_entry *entry; 2947 struct IO_APIC_route_entry *saved_data = ioapic_saved_data[ioapic_id];
2974 struct sysfs_ioapic_data *data;
2975 unsigned long flags; 2948 unsigned long flags;
2976 union IO_APIC_reg_00 reg_00; 2949 union IO_APIC_reg_00 reg_00;
2977 int i; 2950 int i;
2978 2951
2979 data = container_of(dev, struct sysfs_ioapic_data, dev); 2952 if (!saved_data)
2980 entry = data->entry; 2953 return;
2981 2954
2982 raw_spin_lock_irqsave(&ioapic_lock, flags); 2955 raw_spin_lock_irqsave(&ioapic_lock, flags);
2983 reg_00.raw = io_apic_read(dev->id, 0); 2956 reg_00.raw = io_apic_read(ioapic_id, 0);
2984 if (reg_00.bits.ID != mp_ioapics[dev->id].apicid) { 2957 if (reg_00.bits.ID != mp_ioapics[ioapic_id].apicid) {
2985 reg_00.bits.ID = mp_ioapics[dev->id].apicid; 2958 reg_00.bits.ID = mp_ioapics[ioapic_id].apicid;
2986 io_apic_write(dev->id, 0, reg_00.raw); 2959 io_apic_write(ioapic_id, 0, reg_00.raw);
2987 } 2960 }
2988 raw_spin_unlock_irqrestore(&ioapic_lock, flags); 2961 raw_spin_unlock_irqrestore(&ioapic_lock, flags);
2989 for (i = 0; i < nr_ioapic_registers[dev->id]; i++) 2962 for (i = 0; i < nr_ioapic_registers[ioapic_id]; i++)
2990 ioapic_write_entry(dev->id, i, entry[i]); 2963 ioapic_write_entry(ioapic_id, i, saved_data[i]);
2964}
2991 2965
2992 return 0; 2966static void ioapic_resume(void)
2967{
2968 int ioapic_id;
2969
2970 for (ioapic_id = nr_ioapics - 1; ioapic_id >= 0; ioapic_id--)
2971 resume_ioapic(ioapic_id);
2993} 2972}
2994 2973
2995static struct sysdev_class ioapic_sysdev_class = { 2974static struct syscore_ops ioapic_syscore_ops = {
2996 .name = "ioapic",
2997 .suspend = ioapic_suspend, 2975 .suspend = ioapic_suspend,
2998 .resume = ioapic_resume, 2976 .resume = ioapic_resume,
2999}; 2977};
3000 2978
3001static int __init ioapic_init_sysfs(void) 2979static int __init ioapic_init_ops(void)
3002{ 2980{
3003 struct sys_device * dev; 2981 int i;
3004 int i, size, error;
3005 2982
3006 error = sysdev_class_register(&ioapic_sysdev_class); 2983 for (i = 0; i < nr_ioapics; i++) {
3007 if (error) 2984 unsigned int size;
3008 return error;
3009 2985
3010 for (i = 0; i < nr_ioapics; i++ ) { 2986 size = nr_ioapic_registers[i]
3011 size = sizeof(struct sys_device) + nr_ioapic_registers[i]
3012 * sizeof(struct IO_APIC_route_entry); 2987 * sizeof(struct IO_APIC_route_entry);
3013 mp_ioapic_data[i] = kzalloc(size, GFP_KERNEL); 2988 ioapic_saved_data[i] = kzalloc(size, GFP_KERNEL);
3014 if (!mp_ioapic_data[i]) { 2989 if (!ioapic_saved_data[i])
3015 printk(KERN_ERR "Can't suspend/resume IOAPIC %d\n", i); 2990 pr_err("IOAPIC %d: suspend/resume impossible!\n", i);
3016 continue;
3017 }
3018 dev = &mp_ioapic_data[i]->dev;
3019 dev->id = i;
3020 dev->cls = &ioapic_sysdev_class;
3021 error = sysdev_register(dev);
3022 if (error) {
3023 kfree(mp_ioapic_data[i]);
3024 mp_ioapic_data[i] = NULL;
3025 printk(KERN_ERR "Can't suspend/resume IOAPIC %d\n", i);
3026 continue;
3027 }
3028 } 2991 }
3029 2992
2993 register_syscore_ops(&ioapic_syscore_ops);
2994
3030 return 0; 2995 return 0;
3031} 2996}
3032 2997
3033device_initcall(ioapic_init_sysfs); 2998device_initcall(ioapic_init_ops);
3034 2999
3035/* 3000/*
3036 * Dynamic irq allocate and deallocation 3001 * Dynamic irq allocate and deallocation
@@ -3060,7 +3025,7 @@ unsigned int create_irq_nr(unsigned int from, int node)
3060 raw_spin_unlock_irqrestore(&vector_lock, flags); 3025 raw_spin_unlock_irqrestore(&vector_lock, flags);
3061 3026
3062 if (ret) { 3027 if (ret) {
3063 set_irq_chip_data(irq, cfg); 3028 irq_set_chip_data(irq, cfg);
3064 irq_clear_status_flags(irq, IRQ_NOREQUEST); 3029 irq_clear_status_flags(irq, IRQ_NOREQUEST);
3065 } else { 3030 } else {
3066 free_irq_at(irq, cfg); 3031 free_irq_at(irq, cfg);
@@ -3085,7 +3050,7 @@ int create_irq(void)
3085 3050
3086void destroy_irq(unsigned int irq) 3051void destroy_irq(unsigned int irq)
3087{ 3052{
3088 struct irq_cfg *cfg = get_irq_chip_data(irq); 3053 struct irq_cfg *cfg = irq_get_chip_data(irq);
3089 unsigned long flags; 3054 unsigned long flags;
3090 3055
3091 irq_set_status_flags(irq, IRQ_NOREQUEST|IRQ_NOPROBE); 3056 irq_set_status_flags(irq, IRQ_NOREQUEST|IRQ_NOPROBE);
@@ -3119,7 +3084,7 @@ static int msi_compose_msg(struct pci_dev *pdev, unsigned int irq,
3119 3084
3120 dest = apic->cpu_mask_to_apicid_and(cfg->domain, apic->target_cpus()); 3085 dest = apic->cpu_mask_to_apicid_and(cfg->domain, apic->target_cpus());
3121 3086
3122 if (irq_remapped(get_irq_chip_data(irq))) { 3087 if (irq_remapped(cfg)) {
3123 struct irte irte; 3088 struct irte irte;
3124 int ir_index; 3089 int ir_index;
3125 u16 sub_handle; 3090 u16 sub_handle;
@@ -3291,6 +3256,7 @@ static int msi_alloc_irte(struct pci_dev *dev, int irq, int nvec)
3291 3256
3292static int setup_msi_irq(struct pci_dev *dev, struct msi_desc *msidesc, int irq) 3257static int setup_msi_irq(struct pci_dev *dev, struct msi_desc *msidesc, int irq)
3293{ 3258{
3259 struct irq_chip *chip = &msi_chip;
3294 struct msi_msg msg; 3260 struct msi_msg msg;
3295 int ret; 3261 int ret;
3296 3262
@@ -3298,14 +3264,15 @@ static int setup_msi_irq(struct pci_dev *dev, struct msi_desc *msidesc, int irq)
3298 if (ret < 0) 3264 if (ret < 0)
3299 return ret; 3265 return ret;
3300 3266
3301 set_irq_msi(irq, msidesc); 3267 irq_set_msi_desc(irq, msidesc);
3302 write_msi_msg(irq, &msg); 3268 write_msi_msg(irq, &msg);
3303 3269
3304 if (irq_remapped(get_irq_chip_data(irq))) { 3270 if (irq_remapped(irq_get_chip_data(irq))) {
3305 irq_set_status_flags(irq, IRQ_MOVE_PCNTXT); 3271 irq_set_status_flags(irq, IRQ_MOVE_PCNTXT);
3306 set_irq_chip_and_handler_name(irq, &msi_ir_chip, handle_edge_irq, "edge"); 3272 chip = &msi_ir_chip;
3307 } else 3273 }
3308 set_irq_chip_and_handler_name(irq, &msi_chip, handle_edge_irq, "edge"); 3274
3275 irq_set_chip_and_handler_name(irq, chip, handle_edge_irq, "edge");
3309 3276
3310 dev_printk(KERN_DEBUG, &dev->dev, "irq %d for MSI/MSI-X\n", irq); 3277 dev_printk(KERN_DEBUG, &dev->dev, "irq %d for MSI/MSI-X\n", irq);
3311 3278
@@ -3423,8 +3390,8 @@ int arch_setup_dmar_msi(unsigned int irq)
3423 if (ret < 0) 3390 if (ret < 0)
3424 return ret; 3391 return ret;
3425 dmar_msi_write(irq, &msg); 3392 dmar_msi_write(irq, &msg);
3426 set_irq_chip_and_handler_name(irq, &dmar_msi_type, handle_edge_irq, 3393 irq_set_chip_and_handler_name(irq, &dmar_msi_type, handle_edge_irq,
3427 "edge"); 3394 "edge");
3428 return 0; 3395 return 0;
3429} 3396}
3430#endif 3397#endif
@@ -3482,6 +3449,7 @@ static struct irq_chip hpet_msi_type = {
3482 3449
3483int arch_setup_hpet_msi(unsigned int irq, unsigned int id) 3450int arch_setup_hpet_msi(unsigned int irq, unsigned int id)
3484{ 3451{
3452 struct irq_chip *chip = &hpet_msi_type;
3485 struct msi_msg msg; 3453 struct msi_msg msg;
3486 int ret; 3454 int ret;
3487 3455
@@ -3501,15 +3469,12 @@ int arch_setup_hpet_msi(unsigned int irq, unsigned int id)
3501 if (ret < 0) 3469 if (ret < 0)
3502 return ret; 3470 return ret;
3503 3471
3504 hpet_msi_write(get_irq_data(irq), &msg); 3472 hpet_msi_write(irq_get_handler_data(irq), &msg);
3505 irq_set_status_flags(irq, IRQ_MOVE_PCNTXT); 3473 irq_set_status_flags(irq, IRQ_MOVE_PCNTXT);
3506 if (irq_remapped(get_irq_chip_data(irq))) 3474 if (irq_remapped(irq_get_chip_data(irq)))
3507 set_irq_chip_and_handler_name(irq, &ir_hpet_msi_type, 3475 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 3476
3477 irq_set_chip_and_handler_name(irq, chip, handle_edge_irq, "edge");
3513 return 0; 3478 return 0;
3514} 3479}
3515#endif 3480#endif
@@ -3596,7 +3561,7 @@ int arch_setup_ht_irq(unsigned int irq, struct pci_dev *dev)
3596 3561
3597 write_ht_irq_msg(irq, &msg); 3562 write_ht_irq_msg(irq, &msg);
3598 3563
3599 set_irq_chip_and_handler_name(irq, &ht_irq_chip, 3564 irq_set_chip_and_handler_name(irq, &ht_irq_chip,
3600 handle_edge_irq, "edge"); 3565 handle_edge_irq, "edge");
3601 3566
3602 dev_printk(KERN_DEBUG, &dev->dev, "irq %d for HT\n", irq); 3567 dev_printk(KERN_DEBUG, &dev->dev, "irq %d for HT\n", irq);
@@ -3605,7 +3570,40 @@ int arch_setup_ht_irq(unsigned int irq, struct pci_dev *dev)
3605} 3570}
3606#endif /* CONFIG_HT_IRQ */ 3571#endif /* CONFIG_HT_IRQ */
3607 3572
3608int __init io_apic_get_redir_entries (int ioapic) 3573static int
3574io_apic_setup_irq_pin(unsigned int irq, int node, struct io_apic_irq_attr *attr)
3575{
3576 struct irq_cfg *cfg = alloc_irq_and_cfg_at(irq, node);
3577 int ret;
3578
3579 if (!cfg)
3580 return -EINVAL;
3581 ret = __add_pin_to_irq_node(cfg, node, attr->ioapic, attr->ioapic_pin);
3582 if (!ret)
3583 setup_ioapic_irq(attr->ioapic, attr->ioapic_pin, irq, cfg,
3584 attr->trigger, attr->polarity);
3585 return ret;
3586}
3587
3588int io_apic_setup_irq_pin_once(unsigned int irq, int node,
3589 struct io_apic_irq_attr *attr)
3590{
3591 unsigned int id = attr->ioapic, pin = attr->ioapic_pin;
3592 int ret;
3593
3594 /* Avoid redundant programming */
3595 if (test_bit(pin, mp_ioapic_routing[id].pin_programmed)) {
3596 pr_debug("Pin %d-%d already programmed\n",
3597 mp_ioapics[id].apicid, pin);
3598 return 0;
3599 }
3600 ret = io_apic_setup_irq_pin(irq, node, attr);
3601 if (!ret)
3602 set_bit(pin, mp_ioapic_routing[id].pin_programmed);
3603 return ret;
3604}
3605
3606static int __init io_apic_get_redir_entries(int ioapic)
3609{ 3607{
3610 union IO_APIC_reg_01 reg_01; 3608 union IO_APIC_reg_01 reg_01;
3611 unsigned long flags; 3609 unsigned long flags;
@@ -3659,96 +3657,24 @@ int __init arch_probe_nr_irqs(void)
3659} 3657}
3660#endif 3658#endif
3661 3659
3662static int __io_apic_set_pci_routing(struct device *dev, int irq, 3660int io_apic_set_pci_routing(struct device *dev, int irq,
3663 struct io_apic_irq_attr *irq_attr) 3661 struct io_apic_irq_attr *irq_attr)
3664{ 3662{
3665 struct irq_cfg *cfg;
3666 int node; 3663 int node;
3667 int ioapic, pin;
3668 int trigger, polarity;
3669 3664
3670 ioapic = irq_attr->ioapic;
3671 if (!IO_APIC_IRQ(irq)) { 3665 if (!IO_APIC_IRQ(irq)) {
3672 apic_printk(APIC_QUIET,KERN_ERR "IOAPIC[%d]: Invalid reference to IRQ 0\n", 3666 apic_printk(APIC_QUIET,KERN_ERR "IOAPIC[%d]: Invalid reference to IRQ 0\n",
3673 ioapic); 3667 irq_attr->ioapic);
3674 return -EINVAL; 3668 return -EINVAL;
3675 } 3669 }
3676 3670
3677 if (dev) 3671 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
3690 /*
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 3672
3703 return 0; 3673 return io_apic_setup_irq_pin_once(irq, node, irq_attr);
3704} 3674}
3705 3675
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 3676#ifdef CONFIG_X86_32
3730 if ((boot_cpu_data.x86_vendor == X86_VENDOR_INTEL) && 3677static 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{ 3678{
3753 union IO_APIC_reg_00 reg_00; 3679 union IO_APIC_reg_00 reg_00;
3754 static physid_mask_t apic_id_map = PHYSID_MASK_NONE; 3680 static physid_mask_t apic_id_map = PHYSID_MASK_NONE;
@@ -3821,9 +3747,33 @@ int __init io_apic_get_unique_id(int ioapic, int apic_id)
3821 3747
3822 return apic_id; 3748 return apic_id;
3823} 3749}
3750
3751static u8 __init io_apic_unique_id(u8 id)
3752{
3753 if ((boot_cpu_data.x86_vendor == X86_VENDOR_INTEL) &&
3754 !APIC_XAPIC(apic_version[boot_cpu_physical_apicid]))
3755 return io_apic_get_unique_id(nr_ioapics, id);
3756 else
3757 return id;
3758}
3759#else
3760static u8 __init io_apic_unique_id(u8 id)
3761{
3762 int i;
3763 DECLARE_BITMAP(used, 256);
3764
3765 bitmap_zero(used, 256);
3766 for (i = 0; i < nr_ioapics; i++) {
3767 struct mpc_ioapic *ia = &mp_ioapics[i];
3768 __set_bit(ia->apicid, used);
3769 }
3770 if (!test_bit(id, used))
3771 return id;
3772 return find_first_zero_bit(used, 256);
3773}
3824#endif 3774#endif
3825 3775
3826int __init io_apic_get_version(int ioapic) 3776static int __init io_apic_get_version(int ioapic)
3827{ 3777{
3828 union IO_APIC_reg_01 reg_01; 3778 union IO_APIC_reg_01 reg_01;
3829 unsigned long flags; 3779 unsigned long flags;
@@ -3868,8 +3818,8 @@ int acpi_get_override_irq(u32 gsi, int *trigger, int *polarity)
3868void __init setup_ioapic_dest(void) 3818void __init setup_ioapic_dest(void)
3869{ 3819{
3870 int pin, ioapic, irq, irq_entry; 3820 int pin, ioapic, irq, irq_entry;
3871 struct irq_desc *desc;
3872 const struct cpumask *mask; 3821 const struct cpumask *mask;
3822 struct irq_data *idata;
3873 3823
3874 if (skip_ioapic_setup == 1) 3824 if (skip_ioapic_setup == 1)
3875 return; 3825 return;
@@ -3884,21 +3834,20 @@ void __init setup_ioapic_dest(void)
3884 if ((ioapic > 0) && (irq > 16)) 3834 if ((ioapic > 0) && (irq > 16))
3885 continue; 3835 continue;
3886 3836
3887 desc = irq_to_desc(irq); 3837 idata = irq_get_irq_data(irq);
3888 3838
3889 /* 3839 /*
3890 * Honour affinities which have been set in early boot 3840 * Honour affinities which have been set in early boot
3891 */ 3841 */
3892 if (desc->status & 3842 if (!irqd_can_balance(idata) || irqd_affinity_was_set(idata))
3893 (IRQ_NO_BALANCING | IRQ_AFFINITY_SET)) 3843 mask = idata->affinity;
3894 mask = desc->irq_data.affinity;
3895 else 3844 else
3896 mask = apic->target_cpus(); 3845 mask = apic->target_cpus();
3897 3846
3898 if (intr_remapping_enabled) 3847 if (intr_remapping_enabled)
3899 ir_ioapic_set_affinity(&desc->irq_data, mask, false); 3848 ir_ioapic_set_affinity(idata, mask, false);
3900 else 3849 else
3901 ioapic_set_affinity(&desc->irq_data, mask, false); 3850 ioapic_set_affinity(idata, mask, false);
3902 } 3851 }
3903 3852
3904} 3853}
@@ -4026,10 +3975,10 @@ int mp_find_ioapic_pin(int ioapic, u32 gsi)
4026 return gsi - mp_gsi_routing[ioapic].gsi_base; 3975 return gsi - mp_gsi_routing[ioapic].gsi_base;
4027} 3976}
4028 3977
4029static int bad_ioapic(unsigned long address) 3978static __init int bad_ioapic(unsigned long address)
4030{ 3979{
4031 if (nr_ioapics >= MAX_IO_APICS) { 3980 if (nr_ioapics >= MAX_IO_APICS) {
4032 printk(KERN_WARNING "WARING: Max # of I/O APICs (%d) exceeded " 3981 printk(KERN_WARNING "WARNING: Max # of I/O APICs (%d) exceeded "
4033 "(found %d), skipping\n", MAX_IO_APICS, nr_ioapics); 3982 "(found %d), skipping\n", MAX_IO_APICS, nr_ioapics);
4034 return 1; 3983 return 1;
4035 } 3984 }
@@ -4086,20 +4035,16 @@ void __init mp_register_ioapic(int id, u32 address, u32 gsi_base)
4086/* Enable IOAPIC early just for system timer */ 4035/* Enable IOAPIC early just for system timer */
4087void __init pre_init_apic_IRQ0(void) 4036void __init pre_init_apic_IRQ0(void)
4088{ 4037{
4089 struct irq_cfg *cfg; 4038 struct io_apic_irq_attr attr = { 0, 0, 0, 0 };
4090 4039
4091 printk(KERN_INFO "Early APIC setup for system timer0\n"); 4040 printk(KERN_INFO "Early APIC setup for system timer0\n");
4092#ifndef CONFIG_SMP 4041#ifndef CONFIG_SMP
4093 physid_set_mask_of_physid(boot_cpu_physical_apicid, 4042 physid_set_mask_of_physid(boot_cpu_physical_apicid,
4094 &phys_cpu_present_map); 4043 &phys_cpu_present_map);
4095#endif 4044#endif
4096 /* Make sure the irq descriptor is set up */
4097 cfg = alloc_irq_and_cfg_at(0, 0);
4098
4099 setup_local_APIC(); 4045 setup_local_APIC();
4100 4046
4101 add_pin_to_irq_node(cfg, 0, 0, 0); 4047 io_apic_setup_irq_pin(0, 0, &attr);
4102 set_irq_chip_and_handler_name(0, &ioapic_chip, handle_edge_irq, "edge"); 4048 irq_set_chip_and_handler_name(0, &ioapic_chip, handle_edge_irq,
4103 4049 "edge");
4104 setup_ioapic_irq(0, 0, 0, cfg, 0, 0);
4105} 4050}
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..33b10a0fc095 100644
--- a/arch/x86/kernel/apic/x2apic_uv_x.c
+++ b/arch/x86/kernel/apic/x2apic_uv_x.c
@@ -23,6 +23,8 @@
23#include <linux/io.h> 23#include <linux/io.h>
24#include <linux/pci.h> 24#include <linux/pci.h>
25#include <linux/kdebug.h> 25#include <linux/kdebug.h>
26#include <linux/delay.h>
27#include <linux/crash_dump.h>
26 28
27#include <asm/uv/uv_mmrs.h> 29#include <asm/uv/uv_mmrs.h>
28#include <asm/uv/uv_hub.h> 30#include <asm/uv/uv_hub.h>
@@ -34,6 +36,7 @@
34#include <asm/ipi.h> 36#include <asm/ipi.h>
35#include <asm/smp.h> 37#include <asm/smp.h>
36#include <asm/x86_init.h> 38#include <asm/x86_init.h>
39#include <asm/emergency-restart.h>
37 40
38DEFINE_PER_CPU(int, x2apic_extra_bits); 41DEFINE_PER_CPU(int, x2apic_extra_bits);
39 42
@@ -338,8 +341,6 @@ struct apic __refdata apic_x2apic_uv_x = {
338 .ioapic_phys_id_map = NULL, 341 .ioapic_phys_id_map = NULL,
339 .setup_apic_routing = NULL, 342 .setup_apic_routing = NULL,
340 .multi_timer_check = NULL, 343 .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, 344 .cpu_present_to_apicid = default_cpu_present_to_apicid,
344 .apicid_to_cpu_present = NULL, 345 .apicid_to_cpu_present = NULL,
345 .setup_portio_remap = NULL, 346 .setup_portio_remap = NULL,
@@ -812,4 +813,11 @@ void __init uv_system_init(void)
812 813
813 /* register Legacy VGA I/O redirection handler */ 814 /* register Legacy VGA I/O redirection handler */
814 pci_register_set_vga_state(uv_set_vga_state); 815 pci_register_set_vga_state(uv_set_vga_state);
816
817 /*
818 * For a kdump kernel the reset must be BOOT_ACPI, not BOOT_EFI, as
819 * EFI is not enabled in the kdump kernel.
820 */
821 if (is_kdump_kernel())
822 reboot_type = BOOT_ACPI;
815} 823}
diff --git a/arch/x86/kernel/apm_32.c b/arch/x86/kernel/apm_32.c
index 0e4f24c2a746..adee12e0da1f 100644
--- a/arch/x86/kernel/apm_32.c
+++ b/arch/x86/kernel/apm_32.c
@@ -66,7 +66,7 @@
66 * 1.5: Fix segment register reloading (in case of bad segments saved 66 * 1.5: Fix segment register reloading (in case of bad segments saved
67 * across BIOS call). 67 * across BIOS call).
68 * Stephen Rothwell 68 * Stephen Rothwell
69 * 1.6: Cope with complier/assembler differences. 69 * 1.6: Cope with compiler/assembler differences.
70 * Only try to turn off the first display device. 70 * Only try to turn off the first display device.
71 * Fix OOPS at power off with no APM BIOS by Jan Echternach 71 * Fix OOPS at power off with no APM BIOS by Jan Echternach
72 * <echter@informatik.uni-rostock.de> 72 * <echter@informatik.uni-rostock.de>
@@ -227,6 +227,8 @@
227#include <linux/suspend.h> 227#include <linux/suspend.h>
228#include <linux/kthread.h> 228#include <linux/kthread.h>
229#include <linux/jiffies.h> 229#include <linux/jiffies.h>
230#include <linux/acpi.h>
231#include <linux/syscore_ops.h>
230 232
231#include <asm/system.h> 233#include <asm/system.h>
232#include <asm/uaccess.h> 234#include <asm/uaccess.h>
@@ -975,20 +977,10 @@ recalc:
975 977
976static void apm_power_off(void) 978static void apm_power_off(void)
977{ 979{
978 unsigned char po_bios_call[] = {
979 0xb8, 0x00, 0x10, /* movw $0x1000,ax */
980 0x8e, 0xd0, /* movw ax,ss */
981 0xbc, 0x00, 0xf0, /* movw $0xf000,sp */
982 0xb8, 0x07, 0x53, /* movw $0x5307,ax */
983 0xbb, 0x01, 0x00, /* movw $0x0001,bx */
984 0xb9, 0x03, 0x00, /* movw $0x0003,cx */
985 0xcd, 0x15 /* int $0x15 */
986 };
987
988 /* Some bioses don't like being called from CPU != 0 */ 980 /* Some bioses don't like being called from CPU != 0 */
989 if (apm_info.realmode_power_off) { 981 if (apm_info.realmode_power_off) {
990 set_cpus_allowed_ptr(current, cpumask_of(0)); 982 set_cpus_allowed_ptr(current, cpumask_of(0));
991 machine_real_restart(po_bios_call, sizeof(po_bios_call)); 983 machine_real_restart(MRR_APM);
992 } else { 984 } else {
993 (void)set_system_power_state(APM_STATE_OFF); 985 (void)set_system_power_state(APM_STATE_OFF);
994 } 986 }
@@ -1247,6 +1239,7 @@ static int suspend(int vetoable)
1247 1239
1248 local_irq_disable(); 1240 local_irq_disable();
1249 sysdev_suspend(PMSG_SUSPEND); 1241 sysdev_suspend(PMSG_SUSPEND);
1242 syscore_suspend();
1250 1243
1251 local_irq_enable(); 1244 local_irq_enable();
1252 1245
@@ -1264,6 +1257,7 @@ static int suspend(int vetoable)
1264 apm_error("suspend", err); 1257 apm_error("suspend", err);
1265 err = (err == APM_SUCCESS) ? 0 : -EIO; 1258 err = (err == APM_SUCCESS) ? 0 : -EIO;
1266 1259
1260 syscore_resume();
1267 sysdev_resume(); 1261 sysdev_resume();
1268 local_irq_enable(); 1262 local_irq_enable();
1269 1263
@@ -1289,6 +1283,7 @@ static void standby(void)
1289 1283
1290 local_irq_disable(); 1284 local_irq_disable();
1291 sysdev_suspend(PMSG_SUSPEND); 1285 sysdev_suspend(PMSG_SUSPEND);
1286 syscore_suspend();
1292 local_irq_enable(); 1287 local_irq_enable();
1293 1288
1294 err = set_system_power_state(APM_STATE_STANDBY); 1289 err = set_system_power_state(APM_STATE_STANDBY);
@@ -1296,6 +1291,7 @@ static void standby(void)
1296 apm_error("standby", err); 1291 apm_error("standby", err);
1297 1292
1298 local_irq_disable(); 1293 local_irq_disable();
1294 syscore_resume();
1299 sysdev_resume(); 1295 sysdev_resume();
1300 local_irq_enable(); 1296 local_irq_enable();
1301 1297
@@ -2331,12 +2327,11 @@ static int __init apm_init(void)
2331 apm_info.disabled = 1; 2327 apm_info.disabled = 1;
2332 return -ENODEV; 2328 return -ENODEV;
2333 } 2329 }
2334 if (pm_flags & PM_ACPI) { 2330 if (!acpi_disabled) {
2335 printk(KERN_NOTICE "apm: overridden by ACPI.\n"); 2331 printk(KERN_NOTICE "apm: overridden by ACPI.\n");
2336 apm_info.disabled = 1; 2332 apm_info.disabled = 1;
2337 return -ENODEV; 2333 return -ENODEV;
2338 } 2334 }
2339 pm_flags |= PM_APM;
2340 2335
2341 /* 2336 /*
2342 * Set up the long jump entry point to the APM BIOS, which is called 2337 * Set up the long jump entry point to the APM BIOS, which is called
@@ -2428,7 +2423,6 @@ static void __exit apm_exit(void)
2428 kthread_stop(kapmd_task); 2423 kthread_stop(kapmd_task);
2429 kapmd_task = NULL; 2424 kapmd_task = NULL;
2430 } 2425 }
2431 pm_flags &= ~PM_APM;
2432} 2426}
2433 2427
2434module_init(apm_init); 2428module_init(apm_init);
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/check.c b/arch/x86/kernel/check.c
index 13a389179514..452932d34730 100644
--- a/arch/x86/kernel/check.c
+++ b/arch/x86/kernel/check.c
@@ -106,8 +106,8 @@ void __init setup_bios_corruption_check(void)
106 addr += size; 106 addr += size;
107 } 107 }
108 108
109 printk(KERN_INFO "Scanning %d areas for low memory corruption\n", 109 if (num_scan_areas)
110 num_scan_areas); 110 printk(KERN_INFO "Scanning %d areas for low memory corruption\n", num_scan_areas);
111} 111}
112 112
113 113
@@ -143,12 +143,12 @@ static void check_corruption(struct work_struct *dummy)
143{ 143{
144 check_for_bios_corruption(); 144 check_for_bios_corruption();
145 schedule_delayed_work(&bios_check_work, 145 schedule_delayed_work(&bios_check_work,
146 round_jiffies_relative(corruption_check_period*HZ)); 146 round_jiffies_relative(corruption_check_period*HZ));
147} 147}
148 148
149static int start_periodic_check_for_corruption(void) 149static int start_periodic_check_for_corruption(void)
150{ 150{
151 if (!memory_corruption_check || corruption_check_period == 0) 151 if (!num_scan_areas || !memory_corruption_check || corruption_check_period == 0)
152 return 0; 152 return 0;
153 153
154 printk(KERN_INFO "Scanning for low memory corruption every %d seconds\n", 154 printk(KERN_INFO "Scanning for low memory corruption every %d seconds\n",
diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c
index 7c7bedb83c5a..bb9eb29a52dd 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 }
@@ -261,7 +265,7 @@ static int __cpuinit nearby_node(int apicid)
261#ifdef CONFIG_X86_HT 265#ifdef CONFIG_X86_HT
262static void __cpuinit amd_get_topology(struct cpuinfo_x86 *c) 266static void __cpuinit amd_get_topology(struct cpuinfo_x86 *c)
263{ 267{
264 u32 nodes; 268 u32 nodes, cores_per_cu = 1;
265 u8 node_id; 269 u8 node_id;
266 int cpu = smp_processor_id(); 270 int cpu = smp_processor_id();
267 271
@@ -276,6 +280,7 @@ static void __cpuinit amd_get_topology(struct cpuinfo_x86 *c)
276 /* get compute unit information */ 280 /* get compute unit information */
277 smp_num_siblings = ((ebx >> 8) & 3) + 1; 281 smp_num_siblings = ((ebx >> 8) & 3) + 1;
278 c->compute_unit_id = ebx & 0xff; 282 c->compute_unit_id = ebx & 0xff;
283 cores_per_cu += ((ebx >> 8) & 3);
279 } else if (cpu_has(c, X86_FEATURE_NODEID_MSR)) { 284 } else if (cpu_has(c, X86_FEATURE_NODEID_MSR)) {
280 u64 value; 285 u64 value;
281 286
@@ -288,15 +293,18 @@ static void __cpuinit amd_get_topology(struct cpuinfo_x86 *c)
288 /* fixup multi-node processor information */ 293 /* fixup multi-node processor information */
289 if (nodes > 1) { 294 if (nodes > 1) {
290 u32 cores_per_node; 295 u32 cores_per_node;
296 u32 cus_per_node;
291 297
292 set_cpu_cap(c, X86_FEATURE_AMD_DCM); 298 set_cpu_cap(c, X86_FEATURE_AMD_DCM);
293 cores_per_node = c->x86_max_cores / nodes; 299 cores_per_node = c->x86_max_cores / nodes;
300 cus_per_node = cores_per_node / cores_per_cu;
294 301
295 /* store NodeID, use llc_shared_map to store sibling info */ 302 /* store NodeID, use llc_shared_map to store sibling info */
296 per_cpu(cpu_llc_id, cpu) = node_id; 303 per_cpu(cpu_llc_id, cpu) = node_id;
297 304
298 /* core id to be in range from 0 to (cores_per_node - 1) */ 305 /* core id has to be in the [0 .. cores_per_node - 1] range */
299 c->cpu_core_id = c->cpu_core_id % cores_per_node; 306 c->cpu_core_id %= cores_per_node;
307 c->compute_unit_id %= cus_per_node;
300 } 308 }
301} 309}
302#endif 310#endif
@@ -334,31 +342,40 @@ EXPORT_SYMBOL_GPL(amd_get_nb_id);
334 342
335static void __cpuinit srat_detect_node(struct cpuinfo_x86 *c) 343static void __cpuinit srat_detect_node(struct cpuinfo_x86 *c)
336{ 344{
337#if defined(CONFIG_NUMA) && defined(CONFIG_X86_64) 345#ifdef CONFIG_NUMA
338 int cpu = smp_processor_id(); 346 int cpu = smp_processor_id();
339 int node; 347 int node;
340 unsigned apicid = c->apicid; 348 unsigned apicid = c->apicid;
341 349
342 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);
343 353
344 if (apicid_to_node[apicid] != NUMA_NO_NODE)
345 node = apicid_to_node[apicid];
346 if (!node_online(node)) { 354 if (!node_online(node)) {
347 /* Two possibilities here: 355 /*
348 - The CPU is missing memory and no node was created. 356 * Two possibilities here:
349 In that case try picking one from a nearby CPU 357 *
350 - The APIC IDs differ from the HyperTransport node IDs 358 * - The CPU is missing memory and no node was created. In
351 which the K8 northbridge parsing fills in. 359 * that case try picking one from a nearby CPU.
352 Assume they are all increased by a constant offset, 360 *
353 but in the same order as the HT nodeids. 361 * - The APIC IDs differ from the HyperTransport node IDs
354 If that doesn't result in a usable node fall back to the 362 * which the K8 northbridge parsing fills in. Assume
355 path for the previous case. */ 363 * they are all increased by a constant offset, but in
356 364 * 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 */
357 int ht_nodeid = c->initial_apicid; 374 int ht_nodeid = c->initial_apicid;
358 375
359 if (ht_nodeid >= 0 && 376 if (ht_nodeid >= 0 &&
360 apicid_to_node[ht_nodeid] != NUMA_NO_NODE) 377 __apicid_to_node[ht_nodeid] != NUMA_NO_NODE)
361 node = apicid_to_node[ht_nodeid]; 378 node = __apicid_to_node[ht_nodeid];
362 /* Pick a nearby node */ 379 /* Pick a nearby node */
363 if (!node_online(node)) 380 if (!node_online(node))
364 node = nearby_node(apicid); 381 node = nearby_node(apicid);
@@ -594,6 +611,29 @@ static void __cpuinit init_amd(struct cpuinfo_x86 *c)
594 } 611 }
595 } 612 }
596#endif 613#endif
614
615 /* As a rule processors have APIC timer running in deep C states */
616 if (c->x86 >= 0xf && !cpu_has_amd_erratum(amd_erratum_400))
617 set_cpu_cap(c, X86_FEATURE_ARAT);
618
619 /*
620 * Disable GART TLB Walk Errors on Fam10h. We do this here
621 * because this is always needed when GART is enabled, even in a
622 * kernel which has no MCE support built in.
623 */
624 if (c->x86 == 0x10) {
625 /*
626 * BIOS should disable GartTlbWlk Errors themself. If
627 * it doesn't do it here as suggested by the BKDG.
628 *
629 * Fixes: https://bugzilla.kernel.org/show_bug.cgi?id=33012
630 */
631 u64 mask;
632
633 rdmsrl(MSR_AMD64_MCx_MASK(4), mask);
634 mask |= (1 << 10);
635 wrmsrl(MSR_AMD64_MCx_MASK(4), mask);
636 }
597} 637}
598 638
599#ifdef CONFIG_X86_32 639#ifdef CONFIG_X86_32
@@ -658,7 +698,7 @@ cpu_dev_register(amd_cpu_dev);
658 */ 698 */
659 699
660const int amd_erratum_400[] = 700const int amd_erratum_400[] =
661 AMD_OSVW_ERRATUM(1, AMD_MODEL_RANGE(0xf, 0x41, 0x2, 0xff, 0xf), 701 AMD_OSVW_ERRATUM(1, AMD_MODEL_RANGE(0x0f, 0x4, 0x2, 0xff, 0xf),
662 AMD_MODEL_RANGE(0x10, 0x2, 0x1, 0xff, 0xf)); 702 AMD_MODEL_RANGE(0x10, 0x2, 0x1, 0xff, 0xf));
663EXPORT_SYMBOL_GPL(amd_erratum_400); 703EXPORT_SYMBOL_GPL(amd_erratum_400);
664 704
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/cpufreq/longhaul.c b/arch/x86/kernel/cpu/cpufreq/longhaul.c
index 03162dac6271..cf48cdd6907d 100644
--- a/arch/x86/kernel/cpu/cpufreq/longhaul.c
+++ b/arch/x86/kernel/cpu/cpufreq/longhaul.c
@@ -444,7 +444,7 @@ static int __cpuinit longhaul_get_ranges(void)
444 return -EINVAL; 444 return -EINVAL;
445 } 445 }
446 /* Get max multiplier - as we always did. 446 /* Get max multiplier - as we always did.
447 * Longhaul MSR is usefull only when voltage scaling is enabled. 447 * Longhaul MSR is useful only when voltage scaling is enabled.
448 * C3 is booting at max anyway. */ 448 * C3 is booting at max anyway. */
449 maxmult = mult; 449 maxmult = mult;
450 /* Get min multiplier */ 450 /* Get min multiplier */
@@ -1011,7 +1011,7 @@ static void __exit longhaul_exit(void)
1011 * trigger frequency transition in some cases. */ 1011 * trigger frequency transition in some cases. */
1012module_param(disable_acpi_c3, int, 0644); 1012module_param(disable_acpi_c3, int, 0644);
1013MODULE_PARM_DESC(disable_acpi_c3, "Don't use ACPI C3 support"); 1013MODULE_PARM_DESC(disable_acpi_c3, "Don't use ACPI C3 support");
1014/* Change CPU voltage with frequency. Very usefull to save 1014/* Change CPU voltage with frequency. Very useful to save
1015 * power, but most VIA C3 processors aren't supporting it. */ 1015 * power, but most VIA C3 processors aren't supporting it. */
1016module_param(scale_voltage, int, 0644); 1016module_param(scale_voltage, int, 0644);
1017MODULE_PARM_DESC(scale_voltage, "Scale voltage of processor"); 1017MODULE_PARM_DESC(scale_voltage, "Scale voltage of processor");
diff --git a/arch/x86/kernel/cpu/cpufreq/p4-clockmod.c b/arch/x86/kernel/cpu/cpufreq/p4-clockmod.c
index bd1cac747f67..52c93648e492 100644
--- a/arch/x86/kernel/cpu/cpufreq/p4-clockmod.c
+++ b/arch/x86/kernel/cpu/cpufreq/p4-clockmod.c
@@ -158,9 +158,9 @@ static unsigned int cpufreq_p4_get_frequency(struct cpuinfo_x86 *c)
158{ 158{
159 if (c->x86 == 0x06) { 159 if (c->x86 == 0x06) {
160 if (cpu_has(c, X86_FEATURE_EST)) 160 if (cpu_has(c, X86_FEATURE_EST))
161 printk(KERN_WARNING PFX "Warning: EST-capable CPU " 161 printk_once(KERN_WARNING PFX "Warning: EST-capable "
162 "detected. The acpi-cpufreq module offers " 162 "CPU detected. The acpi-cpufreq module offers "
163 "voltage scaling in addition of frequency " 163 "voltage scaling in addition to frequency "
164 "scaling. You should use that instead of " 164 "scaling. You should use that instead of "
165 "p4-clockmod, if possible.\n"); 165 "p4-clockmod, if possible.\n");
166 switch (c->x86_model) { 166 switch (c->x86_model) {
diff --git a/arch/x86/kernel/cpu/cpufreq/pcc-cpufreq.c b/arch/x86/kernel/cpu/cpufreq/pcc-cpufreq.c
index 4f6f679f2799..755a31e0f5b0 100644
--- a/arch/x86/kernel/cpu/cpufreq/pcc-cpufreq.c
+++ b/arch/x86/kernel/cpu/cpufreq/pcc-cpufreq.c
@@ -195,7 +195,7 @@ static unsigned int pcc_get_freq(unsigned int cpu)
195cmd_incomplete: 195cmd_incomplete:
196 iowrite16(0, &pcch_hdr->status); 196 iowrite16(0, &pcch_hdr->status);
197 spin_unlock(&pcc_lock); 197 spin_unlock(&pcc_lock);
198 return -EINVAL; 198 return 0;
199} 199}
200 200
201static int pcc_cpufreq_target(struct cpufreq_policy *policy, 201static int pcc_cpufreq_target(struct cpufreq_policy *policy,
@@ -315,8 +315,6 @@ static int __init pcc_cpufreq_do_osc(acpi_handle *handle)
315 315
316 input.count = 4; 316 input.count = 4;
317 input.pointer = in_params; 317 input.pointer = in_params;
318 input.count = 4;
319 input.pointer = in_params;
320 in_params[0].type = ACPI_TYPE_BUFFER; 318 in_params[0].type = ACPI_TYPE_BUFFER;
321 in_params[0].buffer.length = 16; 319 in_params[0].buffer.length = 16;
322 in_params[0].buffer.pointer = OSC_UUID; 320 in_params[0].buffer.pointer = OSC_UUID;
diff --git a/arch/x86/kernel/cpu/cpufreq/powernow-k8.c b/arch/x86/kernel/cpu/cpufreq/powernow-k8.c
index 35c7e65e59be..2368e38327b3 100644
--- a/arch/x86/kernel/cpu/cpufreq/powernow-k8.c
+++ b/arch/x86/kernel/cpu/cpufreq/powernow-k8.c
@@ -630,8 +630,7 @@ static void print_basics(struct powernow_k8_data *data)
630 data->powernow_table[j].frequency/1000); 630 data->powernow_table[j].frequency/1000);
631 } else { 631 } else {
632 printk(KERN_INFO PFX 632 printk(KERN_INFO PFX
633 " %d : fid 0x%x (%d MHz), vid 0x%x\n", 633 "fid 0x%x (%d MHz), vid 0x%x\n",
634 j,
635 data->powernow_table[j].index & 0xff, 634 data->powernow_table[j].index & 0xff,
636 data->powernow_table[j].frequency/1000, 635 data->powernow_table[j].frequency/1000,
637 data->powernow_table[j].index >> 8); 636 data->powernow_table[j].index >> 8);
@@ -1276,7 +1275,7 @@ static int __cpuinit powernowk8_cpu_init(struct cpufreq_policy *pol)
1276 1275
1277 if (powernow_k8_cpu_init_acpi(data)) { 1276 if (powernow_k8_cpu_init_acpi(data)) {
1278 /* 1277 /*
1279 * Use the PSB BIOS structure. This is only availabe on 1278 * Use the PSB BIOS structure. This is only available on
1280 * an UP version, and is deprecated by AMD. 1279 * an UP version, and is deprecated by AMD.
1281 */ 1280 */
1282 if (num_online_cpus() != 1) { 1281 if (num_online_cpus() != 1) {
@@ -1537,6 +1536,7 @@ static struct notifier_block cpb_nb = {
1537static int __cpuinit powernowk8_init(void) 1536static int __cpuinit powernowk8_init(void)
1538{ 1537{
1539 unsigned int i, supported_cpus = 0, cpu; 1538 unsigned int i, supported_cpus = 0, cpu;
1539 int rv;
1540 1540
1541 for_each_online_cpu(i) { 1541 for_each_online_cpu(i) {
1542 int rc; 1542 int rc;
@@ -1555,14 +1555,14 @@ static int __cpuinit powernowk8_init(void)
1555 1555
1556 cpb_capable = true; 1556 cpb_capable = true;
1557 1557
1558 register_cpu_notifier(&cpb_nb);
1559
1560 msrs = msrs_alloc(); 1558 msrs = msrs_alloc();
1561 if (!msrs) { 1559 if (!msrs) {
1562 printk(KERN_ERR "%s: Error allocating msrs!\n", __func__); 1560 printk(KERN_ERR "%s: Error allocating msrs!\n", __func__);
1563 return -ENOMEM; 1561 return -ENOMEM;
1564 } 1562 }
1565 1563
1564 register_cpu_notifier(&cpb_nb);
1565
1566 rdmsr_on_cpus(cpu_online_mask, MSR_K7_HWCR, msrs); 1566 rdmsr_on_cpus(cpu_online_mask, MSR_K7_HWCR, msrs);
1567 1567
1568 for_each_cpu(cpu, cpu_online_mask) { 1568 for_each_cpu(cpu, cpu_online_mask) {
@@ -1574,7 +1574,13 @@ static int __cpuinit powernowk8_init(void)
1574 (cpb_enabled ? "on" : "off")); 1574 (cpb_enabled ? "on" : "off"));
1575 } 1575 }
1576 1576
1577 return cpufreq_register_driver(&cpufreq_amd64_driver); 1577 rv = cpufreq_register_driver(&cpufreq_amd64_driver);
1578 if (rv < 0 && boot_cpu_has(X86_FEATURE_CPB)) {
1579 unregister_cpu_notifier(&cpb_nb);
1580 msrs_free(msrs);
1581 msrs = NULL;
1582 }
1583 return rv;
1578} 1584}
1579 1585
1580/* driver entry point for term */ 1586/* driver entry point for term */
diff --git a/arch/x86/kernel/cpu/cpufreq/speedstep-smi.c b/arch/x86/kernel/cpu/cpufreq/speedstep-smi.c
index 8abd869baabf..91bc25b67bc1 100644
--- a/arch/x86/kernel/cpu/cpufreq/speedstep-smi.c
+++ b/arch/x86/kernel/cpu/cpufreq/speedstep-smi.c
@@ -292,7 +292,7 @@ static int speedstep_cpu_init(struct cpufreq_policy *policy)
292 292
293 result = speedstep_smi_ownership(); 293 result = speedstep_smi_ownership();
294 if (result) { 294 if (result) {
295 dprintk("fails in aquiring ownership of a SMI interface.\n"); 295 dprintk("fails in acquiring ownership of a SMI interface.\n");
296 return -EINVAL; 296 return -EINVAL;
297 } 297 }
298 298
@@ -360,7 +360,7 @@ static int speedstep_resume(struct cpufreq_policy *policy)
360 int result = speedstep_smi_ownership(); 360 int result = speedstep_smi_ownership();
361 361
362 if (result) 362 if (result)
363 dprintk("fails in re-aquiring ownership of a SMI interface.\n"); 363 dprintk("fails in re-acquiring ownership of a SMI interface.\n");
364 364
365 return result; 365 return result;
366} 366}
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 ec2c19a7b8ef..1ce1af2899df 100644
--- a/arch/x86/kernel/cpu/intel_cacheinfo.c
+++ b/arch/x86/kernel/cpu/intel_cacheinfo.c
@@ -304,8 +304,9 @@ amd_cpuid4(int leaf, union _cpuid4_leaf_eax *eax,
304 304
305struct _cache_attr { 305struct _cache_attr {
306 struct attribute attr; 306 struct attribute attr;
307 ssize_t (*show)(struct _cpuid4_info *, char *); 307 ssize_t (*show)(struct _cpuid4_info *, char *, unsigned int);
308 ssize_t (*store)(struct _cpuid4_info *, const char *, size_t count); 308 ssize_t (*store)(struct _cpuid4_info *, const char *, size_t count,
309 unsigned int);
309}; 310};
310 311
311#ifdef CONFIG_AMD_NB 312#ifdef CONFIG_AMD_NB
@@ -400,7 +401,8 @@ static ssize_t show_cache_disable(struct _cpuid4_info *this_leaf, char *buf,
400 401
401#define SHOW_CACHE_DISABLE(slot) \ 402#define SHOW_CACHE_DISABLE(slot) \
402static ssize_t \ 403static ssize_t \
403show_cache_disable_##slot(struct _cpuid4_info *this_leaf, char *buf) \ 404show_cache_disable_##slot(struct _cpuid4_info *this_leaf, char *buf, \
405 unsigned int cpu) \
404{ \ 406{ \
405 return show_cache_disable(this_leaf, buf, slot); \ 407 return show_cache_disable(this_leaf, buf, slot); \
406} 408}
@@ -512,7 +514,8 @@ static ssize_t store_cache_disable(struct _cpuid4_info *this_leaf,
512#define STORE_CACHE_DISABLE(slot) \ 514#define STORE_CACHE_DISABLE(slot) \
513static ssize_t \ 515static ssize_t \
514store_cache_disable_##slot(struct _cpuid4_info *this_leaf, \ 516store_cache_disable_##slot(struct _cpuid4_info *this_leaf, \
515 const char *buf, size_t count) \ 517 const char *buf, size_t count, \
518 unsigned int cpu) \
516{ \ 519{ \
517 return store_cache_disable(this_leaf, buf, count, slot); \ 520 return store_cache_disable(this_leaf, buf, count, slot); \
518} 521}
@@ -524,6 +527,39 @@ static struct _cache_attr cache_disable_0 = __ATTR(cache_disable_0, 0644,
524static struct _cache_attr cache_disable_1 = __ATTR(cache_disable_1, 0644, 527static struct _cache_attr cache_disable_1 = __ATTR(cache_disable_1, 0644,
525 show_cache_disable_1, store_cache_disable_1); 528 show_cache_disable_1, store_cache_disable_1);
526 529
530static ssize_t
531show_subcaches(struct _cpuid4_info *this_leaf, char *buf, unsigned int cpu)
532{
533 if (!this_leaf->l3 || !amd_nb_has_feature(AMD_NB_L3_PARTITIONING))
534 return -EINVAL;
535
536 return sprintf(buf, "%x\n", amd_get_subcaches(cpu));
537}
538
539static ssize_t
540store_subcaches(struct _cpuid4_info *this_leaf, const char *buf, size_t count,
541 unsigned int cpu)
542{
543 unsigned long val;
544
545 if (!capable(CAP_SYS_ADMIN))
546 return -EPERM;
547
548 if (!this_leaf->l3 || !amd_nb_has_feature(AMD_NB_L3_PARTITIONING))
549 return -EINVAL;
550
551 if (strict_strtoul(buf, 16, &val) < 0)
552 return -EINVAL;
553
554 if (amd_set_subcaches(cpu, val))
555 return -EINVAL;
556
557 return count;
558}
559
560static struct _cache_attr subcaches =
561 __ATTR(subcaches, 0644, show_subcaches, store_subcaches);
562
527#else /* CONFIG_AMD_NB */ 563#else /* CONFIG_AMD_NB */
528#define amd_init_l3_cache(x, y) 564#define amd_init_l3_cache(x, y)
529#endif /* CONFIG_AMD_NB */ 565#endif /* CONFIG_AMD_NB */
@@ -532,9 +568,9 @@ static int
532__cpuinit cpuid4_cache_lookup_regs(int index, 568__cpuinit cpuid4_cache_lookup_regs(int index,
533 struct _cpuid4_info_regs *this_leaf) 569 struct _cpuid4_info_regs *this_leaf)
534{ 570{
535 union _cpuid4_leaf_eax eax; 571 union _cpuid4_leaf_eax eax;
536 union _cpuid4_leaf_ebx ebx; 572 union _cpuid4_leaf_ebx ebx;
537 union _cpuid4_leaf_ecx ecx; 573 union _cpuid4_leaf_ecx ecx;
538 unsigned edx; 574 unsigned edx;
539 575
540 if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD) { 576 if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD) {
@@ -732,11 +768,11 @@ static void __cpuinit cache_shared_cpu_map_setup(unsigned int cpu, int index)
732 struct cpuinfo_x86 *c = &cpu_data(cpu); 768 struct cpuinfo_x86 *c = &cpu_data(cpu);
733 769
734 if ((index == 3) && (c->x86_vendor == X86_VENDOR_AMD)) { 770 if ((index == 3) && (c->x86_vendor == X86_VENDOR_AMD)) {
735 for_each_cpu(i, c->llc_shared_map) { 771 for_each_cpu(i, cpu_llc_shared_mask(cpu)) {
736 if (!per_cpu(ici_cpuid4_info, i)) 772 if (!per_cpu(ici_cpuid4_info, i))
737 continue; 773 continue;
738 this_leaf = CPUID4_INFO_IDX(i, index); 774 this_leaf = CPUID4_INFO_IDX(i, index);
739 for_each_cpu(sibling, c->llc_shared_map) { 775 for_each_cpu(sibling, cpu_llc_shared_mask(cpu)) {
740 if (!cpu_online(sibling)) 776 if (!cpu_online(sibling))
741 continue; 777 continue;
742 set_bit(sibling, this_leaf->shared_cpu_map); 778 set_bit(sibling, this_leaf->shared_cpu_map);
@@ -870,8 +906,8 @@ static DEFINE_PER_CPU(struct _index_kobject *, ici_index_kobject);
870#define INDEX_KOBJECT_PTR(x, y) (&((per_cpu(ici_index_kobject, x))[y])) 906#define INDEX_KOBJECT_PTR(x, y) (&((per_cpu(ici_index_kobject, x))[y]))
871 907
872#define show_one_plus(file_name, object, val) \ 908#define show_one_plus(file_name, object, val) \
873static ssize_t show_##file_name \ 909static ssize_t show_##file_name(struct _cpuid4_info *this_leaf, char *buf, \
874 (struct _cpuid4_info *this_leaf, char *buf) \ 910 unsigned int cpu) \
875{ \ 911{ \
876 return sprintf(buf, "%lu\n", (unsigned long)this_leaf->object + val); \ 912 return sprintf(buf, "%lu\n", (unsigned long)this_leaf->object + val); \
877} 913}
@@ -882,7 +918,8 @@ show_one_plus(physical_line_partition, ebx.split.physical_line_partition, 1);
882show_one_plus(ways_of_associativity, ebx.split.ways_of_associativity, 1); 918show_one_plus(ways_of_associativity, ebx.split.ways_of_associativity, 1);
883show_one_plus(number_of_sets, ecx.split.number_of_sets, 1); 919show_one_plus(number_of_sets, ecx.split.number_of_sets, 1);
884 920
885static ssize_t show_size(struct _cpuid4_info *this_leaf, char *buf) 921static ssize_t show_size(struct _cpuid4_info *this_leaf, char *buf,
922 unsigned int cpu)
886{ 923{
887 return sprintf(buf, "%luK\n", this_leaf->size / 1024); 924 return sprintf(buf, "%luK\n", this_leaf->size / 1024);
888} 925}
@@ -906,17 +943,20 @@ static ssize_t show_shared_cpu_map_func(struct _cpuid4_info *this_leaf,
906 return n; 943 return n;
907} 944}
908 945
909static inline ssize_t show_shared_cpu_map(struct _cpuid4_info *leaf, char *buf) 946static inline ssize_t show_shared_cpu_map(struct _cpuid4_info *leaf, char *buf,
947 unsigned int cpu)
910{ 948{
911 return show_shared_cpu_map_func(leaf, 0, buf); 949 return show_shared_cpu_map_func(leaf, 0, buf);
912} 950}
913 951
914static inline ssize_t show_shared_cpu_list(struct _cpuid4_info *leaf, char *buf) 952static inline ssize_t show_shared_cpu_list(struct _cpuid4_info *leaf, char *buf,
953 unsigned int cpu)
915{ 954{
916 return show_shared_cpu_map_func(leaf, 1, buf); 955 return show_shared_cpu_map_func(leaf, 1, buf);
917} 956}
918 957
919static ssize_t show_type(struct _cpuid4_info *this_leaf, char *buf) 958static ssize_t show_type(struct _cpuid4_info *this_leaf, char *buf,
959 unsigned int cpu)
920{ 960{
921 switch (this_leaf->eax.split.type) { 961 switch (this_leaf->eax.split.type) {
922 case CACHE_TYPE_DATA: 962 case CACHE_TYPE_DATA:
@@ -974,6 +1014,9 @@ static struct attribute ** __cpuinit amd_l3_attrs(void)
974 if (amd_nb_has_feature(AMD_NB_L3_INDEX_DISABLE)) 1014 if (amd_nb_has_feature(AMD_NB_L3_INDEX_DISABLE))
975 n += 2; 1015 n += 2;
976 1016
1017 if (amd_nb_has_feature(AMD_NB_L3_PARTITIONING))
1018 n += 1;
1019
977 attrs = kzalloc(n * sizeof (struct attribute *), GFP_KERNEL); 1020 attrs = kzalloc(n * sizeof (struct attribute *), GFP_KERNEL);
978 if (attrs == NULL) 1021 if (attrs == NULL)
979 return attrs = default_attrs; 1022 return attrs = default_attrs;
@@ -986,6 +1029,9 @@ static struct attribute ** __cpuinit amd_l3_attrs(void)
986 attrs[n++] = &cache_disable_1.attr; 1029 attrs[n++] = &cache_disable_1.attr;
987 } 1030 }
988 1031
1032 if (amd_nb_has_feature(AMD_NB_L3_PARTITIONING))
1033 attrs[n++] = &subcaches.attr;
1034
989 return attrs; 1035 return attrs;
990} 1036}
991#endif 1037#endif
@@ -998,7 +1044,7 @@ static ssize_t show(struct kobject *kobj, struct attribute *attr, char *buf)
998 1044
999 ret = fattr->show ? 1045 ret = fattr->show ?
1000 fattr->show(CPUID4_INFO_IDX(this_leaf->cpu, this_leaf->index), 1046 fattr->show(CPUID4_INFO_IDX(this_leaf->cpu, this_leaf->index),
1001 buf) : 1047 buf, this_leaf->cpu) :
1002 0; 1048 0;
1003 return ret; 1049 return ret;
1004} 1050}
@@ -1012,7 +1058,7 @@ static ssize_t store(struct kobject *kobj, struct attribute *attr,
1012 1058
1013 ret = fattr->store ? 1059 ret = fattr->store ?
1014 fattr->store(CPUID4_INFO_IDX(this_leaf->cpu, this_leaf->index), 1060 fattr->store(CPUID4_INFO_IDX(this_leaf->cpu, this_leaf->index),
1015 buf, count) : 1061 buf, count, this_leaf->cpu) :
1016 0; 1062 0;
1017 return ret; 1063 return ret;
1018} 1064}
diff --git a/arch/x86/kernel/cpu/mcheck/mce-apei.c b/arch/x86/kernel/cpu/mcheck/mce-apei.c
index 8209472b27a5..83930deec3c6 100644
--- a/arch/x86/kernel/cpu/mcheck/mce-apei.c
+++ b/arch/x86/kernel/cpu/mcheck/mce-apei.c
@@ -106,24 +106,34 @@ int apei_write_mce(struct mce *m)
106ssize_t apei_read_mce(struct mce *m, u64 *record_id) 106ssize_t apei_read_mce(struct mce *m, u64 *record_id)
107{ 107{
108 struct cper_mce_record rcd; 108 struct cper_mce_record rcd;
109 ssize_t len; 109 int rc, pos;
110 110
111 len = erst_read_next(&rcd.hdr, sizeof(rcd)); 111 rc = erst_get_record_id_begin(&pos);
112 if (len <= 0) 112 if (rc)
113 return len; 113 return rc;
114 /* Can not skip other records in storage via ERST unless clear them */ 114retry:
115 else if (len != sizeof(rcd) || 115 rc = erst_get_record_id_next(&pos, record_id);
116 uuid_le_cmp(rcd.hdr.creator_id, CPER_CREATOR_MCE)) { 116 if (rc)
117 if (printk_ratelimit()) 117 goto out;
118 pr_warning( 118 /* no more record */
119 "MCE-APEI: Can not skip the unknown record in ERST"); 119 if (*record_id == APEI_ERST_INVALID_RECORD_ID)
120 return -EIO; 120 goto out;
121 } 121 rc = erst_read(*record_id, &rcd.hdr, sizeof(rcd));
122 122 /* someone else has cleared the record, try next one */
123 if (rc == -ENOENT)
124 goto retry;
125 else if (rc < 0)
126 goto out;
127 /* try to skip other type records in storage */
128 else if (rc != sizeof(rcd) ||
129 uuid_le_cmp(rcd.hdr.creator_id, CPER_CREATOR_MCE))
130 goto retry;
123 memcpy(m, &rcd.mce, sizeof(*m)); 131 memcpy(m, &rcd.mce, sizeof(*m));
124 *record_id = rcd.hdr.record_id; 132 rc = sizeof(*m);
133out:
134 erst_get_record_id_end();
125 135
126 return sizeof(*m); 136 return rc;
127} 137}
128 138
129/* Check whether there is record in ERST */ 139/* Check whether there is record in ERST */
diff --git a/arch/x86/kernel/cpu/mcheck/mce-inject.c b/arch/x86/kernel/cpu/mcheck/mce-inject.c
index a77971979564..0ed633c5048b 100644
--- a/arch/x86/kernel/cpu/mcheck/mce-inject.c
+++ b/arch/x86/kernel/cpu/mcheck/mce-inject.c
@@ -32,7 +32,7 @@ static void inject_mce(struct mce *m)
32{ 32{
33 struct mce *i = &per_cpu(injectm, m->extcpu); 33 struct mce *i = &per_cpu(injectm, m->extcpu);
34 34
35 /* Make sure noone reads partially written injectm */ 35 /* Make sure no one reads partially written injectm */
36 i->finished = 0; 36 i->finished = 0;
37 mb(); 37 mb();
38 m->finished = 0; 38 m->finished = 0;
diff --git a/arch/x86/kernel/cpu/mcheck/mce.c b/arch/x86/kernel/cpu/mcheck/mce.c
index d916183b7f9c..3385ea26f684 100644
--- a/arch/x86/kernel/cpu/mcheck/mce.c
+++ b/arch/x86/kernel/cpu/mcheck/mce.c
@@ -21,6 +21,7 @@
21#include <linux/percpu.h> 21#include <linux/percpu.h>
22#include <linux/string.h> 22#include <linux/string.h>
23#include <linux/sysdev.h> 23#include <linux/sysdev.h>
24#include <linux/syscore_ops.h>
24#include <linux/delay.h> 25#include <linux/delay.h>
25#include <linux/ctype.h> 26#include <linux/ctype.h>
26#include <linux/sched.h> 27#include <linux/sched.h>
@@ -881,7 +882,7 @@ reset:
881 * Check if the address reported by the CPU is in a format we can parse. 882 * Check if the address reported by the CPU is in a format we can parse.
882 * It would be possible to add code for most other cases, but all would 883 * It would be possible to add code for most other cases, but all would
883 * be somewhat complicated (e.g. segment offset would require an instruction 884 * be somewhat complicated (e.g. segment offset would require an instruction
884 * parser). So only support physical addresses upto page granuality for now. 885 * parser). So only support physical addresses up to page granuality for now.
885 */ 886 */
886static int mce_usable_address(struct mce *m) 887static int mce_usable_address(struct mce *m)
887{ 888{
@@ -1625,7 +1626,7 @@ out:
1625static unsigned int mce_poll(struct file *file, poll_table *wait) 1626static unsigned int mce_poll(struct file *file, poll_table *wait)
1626{ 1627{
1627 poll_wait(file, &mce_wait, wait); 1628 poll_wait(file, &mce_wait, wait);
1628 if (rcu_dereference_check_mce(mcelog.next)) 1629 if (rcu_access_index(mcelog.next))
1629 return POLLIN | POLLRDNORM; 1630 return POLLIN | POLLRDNORM;
1630 if (!mce_apei_read_done && apei_check_mce()) 1631 if (!mce_apei_read_done && apei_check_mce())
1631 return POLLIN | POLLRDNORM; 1632 return POLLIN | POLLRDNORM;
@@ -1749,14 +1750,14 @@ static int mce_disable_error_reporting(void)
1749 return 0; 1750 return 0;
1750} 1751}
1751 1752
1752static int mce_suspend(struct sys_device *dev, pm_message_t state) 1753static int mce_suspend(void)
1753{ 1754{
1754 return mce_disable_error_reporting(); 1755 return mce_disable_error_reporting();
1755} 1756}
1756 1757
1757static int mce_shutdown(struct sys_device *dev) 1758static void mce_shutdown(void)
1758{ 1759{
1759 return mce_disable_error_reporting(); 1760 mce_disable_error_reporting();
1760} 1761}
1761 1762
1762/* 1763/*
@@ -1764,14 +1765,18 @@ static int mce_shutdown(struct sys_device *dev)
1764 * Only one CPU is active at this time, the others get re-added later using 1765 * Only one CPU is active at this time, the others get re-added later using
1765 * CPU hotplug: 1766 * CPU hotplug:
1766 */ 1767 */
1767static int mce_resume(struct sys_device *dev) 1768static void mce_resume(void)
1768{ 1769{
1769 __mcheck_cpu_init_generic(); 1770 __mcheck_cpu_init_generic();
1770 __mcheck_cpu_init_vendor(__this_cpu_ptr(&cpu_info)); 1771 __mcheck_cpu_init_vendor(__this_cpu_ptr(&cpu_info));
1771
1772 return 0;
1773} 1772}
1774 1773
1774static struct syscore_ops mce_syscore_ops = {
1775 .suspend = mce_suspend,
1776 .shutdown = mce_shutdown,
1777 .resume = mce_resume,
1778};
1779
1775static void mce_cpu_restart(void *data) 1780static void mce_cpu_restart(void *data)
1776{ 1781{
1777 del_timer_sync(&__get_cpu_var(mce_timer)); 1782 del_timer_sync(&__get_cpu_var(mce_timer));
@@ -1808,9 +1813,6 @@ static void mce_enable_ce(void *all)
1808} 1813}
1809 1814
1810static struct sysdev_class mce_sysclass = { 1815static struct sysdev_class mce_sysclass = {
1811 .suspend = mce_suspend,
1812 .shutdown = mce_shutdown,
1813 .resume = mce_resume,
1814 .name = "machinecheck", 1816 .name = "machinecheck",
1815}; 1817};
1816 1818
@@ -2139,6 +2141,7 @@ static __init int mcheck_init_device(void)
2139 return err; 2141 return err;
2140 } 2142 }
2141 2143
2144 register_syscore_ops(&mce_syscore_ops);
2142 register_hotcpu_notifier(&mce_cpu_notifier); 2145 register_hotcpu_notifier(&mce_cpu_notifier);
2143 misc_register(&mce_log_device); 2146 misc_register(&mce_log_device);
2144 2147
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/mtrr/generic.c b/arch/x86/kernel/cpu/mtrr/generic.c
index 9f27228ceffd..a71efcdbb092 100644
--- a/arch/x86/kernel/cpu/mtrr/generic.c
+++ b/arch/x86/kernel/cpu/mtrr/generic.c
@@ -1,6 +1,6 @@
1/* 1/*
2 * This only handles 32bit MTRR on 32bit hosts. This is strictly wrong 2 * This only handles 32bit MTRR on 32bit hosts. This is strictly wrong
3 * because MTRRs can span upto 40 bits (36bits on most modern x86) 3 * because MTRRs can span up to 40 bits (36bits on most modern x86)
4 */ 4 */
5#define DEBUG 5#define DEBUG
6 6
diff --git a/arch/x86/kernel/cpu/mtrr/main.c b/arch/x86/kernel/cpu/mtrr/main.c
index bebabec5b448..929739a653d1 100644
--- a/arch/x86/kernel/cpu/mtrr/main.c
+++ b/arch/x86/kernel/cpu/mtrr/main.c
@@ -45,6 +45,7 @@
45#include <linux/cpu.h> 45#include <linux/cpu.h>
46#include <linux/pci.h> 46#include <linux/pci.h>
47#include <linux/smp.h> 47#include <linux/smp.h>
48#include <linux/syscore_ops.h>
48 49
49#include <asm/processor.h> 50#include <asm/processor.h>
50#include <asm/e820.h> 51#include <asm/e820.h>
@@ -292,14 +293,24 @@ set_mtrr(unsigned int reg, unsigned long base, unsigned long size, mtrr_type typ
292 293
293 /* 294 /*
294 * HACK! 295 * HACK!
295 * We use this same function to initialize the mtrrs on boot. 296 *
296 * The state of the boot cpu's mtrrs has been saved, and we want 297 * We use this same function to initialize the mtrrs during boot,
297 * to replicate across all the APs. 298 * resume, runtime cpu online and on an explicit request to set a
298 * If we're doing that @reg is set to something special... 299 * specific MTRR.
300 *
301 * During boot or suspend, the state of the boot cpu's mtrrs has been
302 * saved, and we want to replicate that across all the cpus that come
303 * online (either at the end of boot or resume or during a runtime cpu
304 * online). If we're doing that, @reg is set to something special and on
305 * this cpu we still do mtrr_if->set_all(). During boot/resume, this
306 * is unnecessary if at this point we are still on the cpu that started
307 * the boot/resume sequence. But there is no guarantee that we are still
308 * on the same cpu. So we do mtrr_if->set_all() on this cpu aswell to be
309 * sure that we are in sync with everyone else.
299 */ 310 */
300 if (reg != ~0U) 311 if (reg != ~0U)
301 mtrr_if->set(reg, base, size, type); 312 mtrr_if->set(reg, base, size, type);
302 else if (!mtrr_aps_delayed_init) 313 else
303 mtrr_if->set_all(); 314 mtrr_if->set_all();
304 315
305 /* Wait for the others */ 316 /* Wait for the others */
@@ -630,7 +641,7 @@ struct mtrr_value {
630 641
631static struct mtrr_value mtrr_value[MTRR_MAX_VAR_RANGES]; 642static struct mtrr_value mtrr_value[MTRR_MAX_VAR_RANGES];
632 643
633static int mtrr_save(struct sys_device *sysdev, pm_message_t state) 644static int mtrr_save(void)
634{ 645{
635 int i; 646 int i;
636 647
@@ -642,7 +653,7 @@ static int mtrr_save(struct sys_device *sysdev, pm_message_t state)
642 return 0; 653 return 0;
643} 654}
644 655
645static int mtrr_restore(struct sys_device *sysdev) 656static void mtrr_restore(void)
646{ 657{
647 int i; 658 int i;
648 659
@@ -653,12 +664,11 @@ static int mtrr_restore(struct sys_device *sysdev)
653 mtrr_value[i].ltype); 664 mtrr_value[i].ltype);
654 } 665 }
655 } 666 }
656 return 0;
657} 667}
658 668
659 669
660 670
661static struct sysdev_driver mtrr_sysdev_driver = { 671static struct syscore_ops mtrr_syscore_ops = {
662 .suspend = mtrr_save, 672 .suspend = mtrr_save,
663 .resume = mtrr_restore, 673 .resume = mtrr_restore,
664}; 674};
@@ -839,7 +849,7 @@ static int __init mtrr_init_finialize(void)
839 * TBD: is there any system with such CPU which supports 849 * TBD: is there any system with such CPU which supports
840 * suspend/resume? If no, we should remove the code. 850 * suspend/resume? If no, we should remove the code.
841 */ 851 */
842 sysdev_driver_register(&cpu_sysdev_class, &mtrr_sysdev_driver); 852 register_syscore_ops(&mtrr_syscore_ops);
843 853
844 return 0; 854 return 0;
845} 855}
diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index 9d977a2ea693..e638689279d3 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;
@@ -166,7 +176,7 @@ 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)
171 181
172#define EVENT_CONSTRAINT_END \ 182#define EVENT_CONSTRAINT_END \
@@ -175,6 +185,28 @@ struct cpu_hw_events {
175#define for_each_event_constraint(e, c) \ 185#define for_each_event_constraint(e, c) \
176 for ((e) = (c); (e)->weight; (e)++) 186 for ((e) = (c); (e)->weight; (e)++)
177 187
188/*
189 * Extra registers for specific events.
190 * Some events need large masks and require external MSRs.
191 * Define a mapping to these extra registers.
192 */
193struct extra_reg {
194 unsigned int event;
195 unsigned int msr;
196 u64 config_mask;
197 u64 valid_mask;
198};
199
200#define EVENT_EXTRA_REG(e, ms, m, vm) { \
201 .event = (e), \
202 .msr = (ms), \
203 .config_mask = (m), \
204 .valid_mask = (vm), \
205 }
206#define INTEL_EVENT_EXTRA_REG(event, msr, vm) \
207 EVENT_EXTRA_REG(event, msr, ARCH_PERFMON_EVENTSEL_EVENT, vm)
208#define EVENT_EXTRA_END EVENT_EXTRA_REG(0, 0, 0, 0)
209
178union perf_capabilities { 210union perf_capabilities {
179 struct { 211 struct {
180 u64 lbr_format : 6; 212 u64 lbr_format : 6;
@@ -219,6 +251,7 @@ struct x86_pmu {
219 void (*put_event_constraints)(struct cpu_hw_events *cpuc, 251 void (*put_event_constraints)(struct cpu_hw_events *cpuc,
220 struct perf_event *event); 252 struct perf_event *event);
221 struct event_constraint *event_constraints; 253 struct event_constraint *event_constraints;
254 struct event_constraint *percore_constraints;
222 void (*quirks)(void); 255 void (*quirks)(void);
223 int perfctr_second_write; 256 int perfctr_second_write;
224 257
@@ -247,6 +280,11 @@ struct x86_pmu {
247 */ 280 */
248 unsigned long lbr_tos, lbr_from, lbr_to; /* MSR base regs */ 281 unsigned long lbr_tos, lbr_from, lbr_to; /* MSR base regs */
249 int lbr_nr; /* hardware stack size */ 282 int lbr_nr; /* hardware stack size */
283
284 /*
285 * Extra registers for events
286 */
287 struct extra_reg *extra_regs;
250}; 288};
251 289
252static struct x86_pmu x86_pmu __read_mostly; 290static struct x86_pmu x86_pmu __read_mostly;
@@ -271,6 +309,10 @@ static u64 __read_mostly hw_cache_event_ids
271 [PERF_COUNT_HW_CACHE_MAX] 309 [PERF_COUNT_HW_CACHE_MAX]
272 [PERF_COUNT_HW_CACHE_OP_MAX] 310 [PERF_COUNT_HW_CACHE_OP_MAX]
273 [PERF_COUNT_HW_CACHE_RESULT_MAX]; 311 [PERF_COUNT_HW_CACHE_RESULT_MAX];
312static u64 __read_mostly hw_cache_extra_regs
313 [PERF_COUNT_HW_CACHE_MAX]
314 [PERF_COUNT_HW_CACHE_OP_MAX]
315 [PERF_COUNT_HW_CACHE_RESULT_MAX];
274 316
275/* 317/*
276 * Propagate event elapsed time into the generic event. 318 * Propagate event elapsed time into the generic event.
@@ -298,7 +340,7 @@ x86_perf_event_update(struct perf_event *event)
298 */ 340 */
299again: 341again:
300 prev_raw_count = local64_read(&hwc->prev_count); 342 prev_raw_count = local64_read(&hwc->prev_count);
301 rdmsrl(hwc->event_base + idx, new_raw_count); 343 rdmsrl(hwc->event_base, new_raw_count);
302 344
303 if (local64_cmpxchg(&hwc->prev_count, prev_raw_count, 345 if (local64_cmpxchg(&hwc->prev_count, prev_raw_count,
304 new_raw_count) != prev_raw_count) 346 new_raw_count) != prev_raw_count)
@@ -321,6 +363,49 @@ again:
321 return new_raw_count; 363 return new_raw_count;
322} 364}
323 365
366/* using X86_FEATURE_PERFCTR_CORE to later implement ALTERNATIVE() here */
367static inline int x86_pmu_addr_offset(int index)
368{
369 if (boot_cpu_has(X86_FEATURE_PERFCTR_CORE))
370 return index << 1;
371 return index;
372}
373
374static inline unsigned int x86_pmu_config_addr(int index)
375{
376 return x86_pmu.eventsel + x86_pmu_addr_offset(index);
377}
378
379static inline unsigned int x86_pmu_event_addr(int index)
380{
381 return x86_pmu.perfctr + x86_pmu_addr_offset(index);
382}
383
384/*
385 * Find and validate any extra registers to set up.
386 */
387static int x86_pmu_extra_regs(u64 config, struct perf_event *event)
388{
389 struct extra_reg *er;
390
391 event->hw.extra_reg = 0;
392 event->hw.extra_config = 0;
393
394 if (!x86_pmu.extra_regs)
395 return 0;
396
397 for (er = x86_pmu.extra_regs; er->msr; er++) {
398 if (er->event != (config & er->config_mask))
399 continue;
400 if (event->attr.config1 & ~er->valid_mask)
401 return -EINVAL;
402 event->hw.extra_reg = er->msr;
403 event->hw.extra_config = event->attr.config1;
404 break;
405 }
406 return 0;
407}
408
324static atomic_t active_events; 409static atomic_t active_events;
325static DEFINE_MUTEX(pmc_reserve_mutex); 410static DEFINE_MUTEX(pmc_reserve_mutex);
326 411
@@ -331,12 +416,12 @@ static bool reserve_pmc_hardware(void)
331 int i; 416 int i;
332 417
333 for (i = 0; i < x86_pmu.num_counters; i++) { 418 for (i = 0; i < x86_pmu.num_counters; i++) {
334 if (!reserve_perfctr_nmi(x86_pmu.perfctr + i)) 419 if (!reserve_perfctr_nmi(x86_pmu_event_addr(i)))
335 goto perfctr_fail; 420 goto perfctr_fail;
336 } 421 }
337 422
338 for (i = 0; i < x86_pmu.num_counters; i++) { 423 for (i = 0; i < x86_pmu.num_counters; i++) {
339 if (!reserve_evntsel_nmi(x86_pmu.eventsel + i)) 424 if (!reserve_evntsel_nmi(x86_pmu_config_addr(i)))
340 goto eventsel_fail; 425 goto eventsel_fail;
341 } 426 }
342 427
@@ -344,13 +429,13 @@ static bool reserve_pmc_hardware(void)
344 429
345eventsel_fail: 430eventsel_fail:
346 for (i--; i >= 0; i--) 431 for (i--; i >= 0; i--)
347 release_evntsel_nmi(x86_pmu.eventsel + i); 432 release_evntsel_nmi(x86_pmu_config_addr(i));
348 433
349 i = x86_pmu.num_counters; 434 i = x86_pmu.num_counters;
350 435
351perfctr_fail: 436perfctr_fail:
352 for (i--; i >= 0; i--) 437 for (i--; i >= 0; i--)
353 release_perfctr_nmi(x86_pmu.perfctr + i); 438 release_perfctr_nmi(x86_pmu_event_addr(i));
354 439
355 return false; 440 return false;
356} 441}
@@ -360,8 +445,8 @@ static void release_pmc_hardware(void)
360 int i; 445 int i;
361 446
362 for (i = 0; i < x86_pmu.num_counters; i++) { 447 for (i = 0; i < x86_pmu.num_counters; i++) {
363 release_perfctr_nmi(x86_pmu.perfctr + i); 448 release_perfctr_nmi(x86_pmu_event_addr(i));
364 release_evntsel_nmi(x86_pmu.eventsel + i); 449 release_evntsel_nmi(x86_pmu_config_addr(i));
365 } 450 }
366} 451}
367 452
@@ -382,7 +467,7 @@ static bool check_hw_exists(void)
382 * complain and bail. 467 * complain and bail.
383 */ 468 */
384 for (i = 0; i < x86_pmu.num_counters; i++) { 469 for (i = 0; i < x86_pmu.num_counters; i++) {
385 reg = x86_pmu.eventsel + i; 470 reg = x86_pmu_config_addr(i);
386 ret = rdmsrl_safe(reg, &val); 471 ret = rdmsrl_safe(reg, &val);
387 if (ret) 472 if (ret)
388 goto msr_fail; 473 goto msr_fail;
@@ -407,20 +492,25 @@ static bool check_hw_exists(void)
407 * that don't trap on the MSR access and always return 0s. 492 * that don't trap on the MSR access and always return 0s.
408 */ 493 */
409 val = 0xabcdUL; 494 val = 0xabcdUL;
410 ret = checking_wrmsrl(x86_pmu.perfctr, val); 495 ret = checking_wrmsrl(x86_pmu_event_addr(0), val);
411 ret |= rdmsrl_safe(x86_pmu.perfctr, &val_new); 496 ret |= rdmsrl_safe(x86_pmu_event_addr(0), &val_new);
412 if (ret || val != val_new) 497 if (ret || val != val_new)
413 goto msr_fail; 498 goto msr_fail;
414 499
415 return true; 500 return true;
416 501
417bios_fail: 502bios_fail:
418 printk(KERN_CONT "Broken BIOS detected, using software events only.\n"); 503 /*
504 * We still allow the PMU driver to operate:
505 */
506 printk(KERN_CONT "Broken BIOS detected, complain to your hardware vendor.\n");
419 printk(KERN_ERR FW_BUG "the BIOS has corrupted hw-PMU resources (MSR %x is %Lx)\n", reg, val); 507 printk(KERN_ERR FW_BUG "the BIOS has corrupted hw-PMU resources (MSR %x is %Lx)\n", reg, val);
420 return false; 508
509 return true;
421 510
422msr_fail: 511msr_fail:
423 printk(KERN_CONT "Broken PMU hardware detected, using software events only.\n"); 512 printk(KERN_CONT "Broken PMU hardware detected, using software events only.\n");
513
424 return false; 514 return false;
425} 515}
426 516
@@ -442,8 +532,9 @@ static inline int x86_pmu_initialized(void)
442} 532}
443 533
444static inline int 534static inline int
445set_ext_hw_attr(struct hw_perf_event *hwc, struct perf_event_attr *attr) 535set_ext_hw_attr(struct hw_perf_event *hwc, struct perf_event *event)
446{ 536{
537 struct perf_event_attr *attr = &event->attr;
447 unsigned int cache_type, cache_op, cache_result; 538 unsigned int cache_type, cache_op, cache_result;
448 u64 config, val; 539 u64 config, val;
449 540
@@ -470,8 +561,8 @@ set_ext_hw_attr(struct hw_perf_event *hwc, struct perf_event_attr *attr)
470 return -EINVAL; 561 return -EINVAL;
471 562
472 hwc->config |= val; 563 hwc->config |= val;
473 564 attr->config1 = hw_cache_extra_regs[cache_type][cache_op][cache_result];
474 return 0; 565 return x86_pmu_extra_regs(val, event);
475} 566}
476 567
477static int x86_setup_perfctr(struct perf_event *event) 568static int x86_setup_perfctr(struct perf_event *event)
@@ -495,11 +586,15 @@ static int x86_setup_perfctr(struct perf_event *event)
495 return -EOPNOTSUPP; 586 return -EOPNOTSUPP;
496 } 587 }
497 588
589 /*
590 * Do not allow config1 (extended registers) to propagate,
591 * there's no sane user-space generalization yet:
592 */
498 if (attr->type == PERF_TYPE_RAW) 593 if (attr->type == PERF_TYPE_RAW)
499 return 0; 594 return 0;
500 595
501 if (attr->type == PERF_TYPE_HW_CACHE) 596 if (attr->type == PERF_TYPE_HW_CACHE)
502 return set_ext_hw_attr(hwc, attr); 597 return set_ext_hw_attr(hwc, event);
503 598
504 if (attr->config >= x86_pmu.max_events) 599 if (attr->config >= x86_pmu.max_events)
505 return -EINVAL; 600 return -EINVAL;
@@ -518,8 +613,8 @@ static int x86_setup_perfctr(struct perf_event *event)
518 /* 613 /*
519 * Branch tracing: 614 * Branch tracing:
520 */ 615 */
521 if ((attr->config == PERF_COUNT_HW_BRANCH_INSTRUCTIONS) && 616 if (attr->config == PERF_COUNT_HW_BRANCH_INSTRUCTIONS &&
522 (hwc->sample_period == 1)) { 617 !attr->freq && hwc->sample_period == 1) {
523 /* BTS is not supported by this architecture. */ 618 /* BTS is not supported by this architecture. */
524 if (!x86_pmu.bts_active) 619 if (!x86_pmu.bts_active)
525 return -EOPNOTSUPP; 620 return -EOPNOTSUPP;
@@ -617,11 +712,11 @@ static void x86_pmu_disable_all(void)
617 712
618 if (!test_bit(idx, cpuc->active_mask)) 713 if (!test_bit(idx, cpuc->active_mask))
619 continue; 714 continue;
620 rdmsrl(x86_pmu.eventsel + idx, val); 715 rdmsrl(x86_pmu_config_addr(idx), val);
621 if (!(val & ARCH_PERFMON_EVENTSEL_ENABLE)) 716 if (!(val & ARCH_PERFMON_EVENTSEL_ENABLE))
622 continue; 717 continue;
623 val &= ~ARCH_PERFMON_EVENTSEL_ENABLE; 718 val &= ~ARCH_PERFMON_EVENTSEL_ENABLE;
624 wrmsrl(x86_pmu.eventsel + idx, val); 719 wrmsrl(x86_pmu_config_addr(idx), val);
625 } 720 }
626} 721}
627 722
@@ -642,21 +737,26 @@ static void x86_pmu_disable(struct pmu *pmu)
642 x86_pmu.disable_all(); 737 x86_pmu.disable_all();
643} 738}
644 739
740static inline void __x86_pmu_enable_event(struct hw_perf_event *hwc,
741 u64 enable_mask)
742{
743 if (hwc->extra_reg)
744 wrmsrl(hwc->extra_reg, hwc->extra_config);
745 wrmsrl(hwc->config_base, hwc->config | enable_mask);
746}
747
645static void x86_pmu_enable_all(int added) 748static void x86_pmu_enable_all(int added)
646{ 749{
647 struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events); 750 struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
648 int idx; 751 int idx;
649 752
650 for (idx = 0; idx < x86_pmu.num_counters; idx++) { 753 for (idx = 0; idx < x86_pmu.num_counters; idx++) {
651 struct perf_event *event = cpuc->events[idx]; 754 struct hw_perf_event *hwc = &cpuc->events[idx]->hw;
652 u64 val;
653 755
654 if (!test_bit(idx, cpuc->active_mask)) 756 if (!test_bit(idx, cpuc->active_mask))
655 continue; 757 continue;
656 758
657 val = event->hw.config; 759 __x86_pmu_enable_event(hwc, ARCH_PERFMON_EVENTSEL_ENABLE);
658 val |= ARCH_PERFMON_EVENTSEL_ENABLE;
659 wrmsrl(x86_pmu.eventsel + idx, val);
660 } 760 }
661} 761}
662 762
@@ -821,15 +921,10 @@ static inline void x86_assign_hw_event(struct perf_event *event,
821 hwc->event_base = 0; 921 hwc->event_base = 0;
822 } else if (hwc->idx >= X86_PMC_IDX_FIXED) { 922 } else if (hwc->idx >= X86_PMC_IDX_FIXED) {
823 hwc->config_base = MSR_ARCH_PERFMON_FIXED_CTR_CTRL; 923 hwc->config_base = MSR_ARCH_PERFMON_FIXED_CTR_CTRL;
824 /* 924 hwc->event_base = MSR_ARCH_PERFMON_FIXED_CTR0 + (hwc->idx - X86_PMC_IDX_FIXED);
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 { 925 } else {
831 hwc->config_base = x86_pmu.eventsel; 926 hwc->config_base = x86_pmu_config_addr(hwc->idx);
832 hwc->event_base = x86_pmu.perfctr; 927 hwc->event_base = x86_pmu_event_addr(hwc->idx);
833 } 928 }
834} 929}
835 930
@@ -915,17 +1010,11 @@ static void x86_pmu_enable(struct pmu *pmu)
915 x86_pmu.enable_all(added); 1010 x86_pmu.enable_all(added);
916} 1011}
917 1012
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) 1013static inline void x86_pmu_disable_event(struct perf_event *event)
925{ 1014{
926 struct hw_perf_event *hwc = &event->hw; 1015 struct hw_perf_event *hwc = &event->hw;
927 1016
928 wrmsrl(hwc->config_base + hwc->idx, hwc->config); 1017 wrmsrl(hwc->config_base, hwc->config);
929} 1018}
930 1019
931static DEFINE_PER_CPU(u64 [X86_PMC_IDX_MAX], pmc_prev_left); 1020static DEFINE_PER_CPU(u64 [X86_PMC_IDX_MAX], pmc_prev_left);
@@ -978,7 +1067,7 @@ x86_perf_event_set_period(struct perf_event *event)
978 */ 1067 */
979 local64_set(&hwc->prev_count, (u64)-left); 1068 local64_set(&hwc->prev_count, (u64)-left);
980 1069
981 wrmsrl(hwc->event_base + idx, (u64)(-left) & x86_pmu.cntval_mask); 1070 wrmsrl(hwc->event_base, (u64)(-left) & x86_pmu.cntval_mask);
982 1071
983 /* 1072 /*
984 * Due to erratum on certan cpu we need 1073 * Due to erratum on certan cpu we need
@@ -986,7 +1075,7 @@ x86_perf_event_set_period(struct perf_event *event)
986 * is updated properly 1075 * is updated properly
987 */ 1076 */
988 if (x86_pmu.perfctr_second_write) { 1077 if (x86_pmu.perfctr_second_write) {
989 wrmsrl(hwc->event_base + idx, 1078 wrmsrl(hwc->event_base,
990 (u64)(-left) & x86_pmu.cntval_mask); 1079 (u64)(-left) & x86_pmu.cntval_mask);
991 } 1080 }
992 1081
@@ -1029,7 +1118,7 @@ static int x86_pmu_add(struct perf_event *event, int flags)
1029 1118
1030 /* 1119 /*
1031 * If group events scheduling transaction was started, 1120 * If group events scheduling transaction was started,
1032 * skip the schedulability test here, it will be peformed 1121 * skip the schedulability test here, it will be performed
1033 * at commit time (->commit_txn) as a whole 1122 * at commit time (->commit_txn) as a whole
1034 */ 1123 */
1035 if (cpuc->group_flag & PERF_EVENT_TXN) 1124 if (cpuc->group_flag & PERF_EVENT_TXN)
@@ -1113,8 +1202,8 @@ void perf_event_print_debug(void)
1113 pr_info("CPU#%d: active: %016llx\n", cpu, *(u64 *)cpuc->active_mask); 1202 pr_info("CPU#%d: active: %016llx\n", cpu, *(u64 *)cpuc->active_mask);
1114 1203
1115 for (idx = 0; idx < x86_pmu.num_counters; idx++) { 1204 for (idx = 0; idx < x86_pmu.num_counters; idx++) {
1116 rdmsrl(x86_pmu.eventsel + idx, pmc_ctrl); 1205 rdmsrl(x86_pmu_config_addr(idx), pmc_ctrl);
1117 rdmsrl(x86_pmu.perfctr + idx, pmc_count); 1206 rdmsrl(x86_pmu_event_addr(idx), pmc_count);
1118 1207
1119 prev_left = per_cpu(pmc_prev_left[idx], cpu); 1208 prev_left = per_cpu(pmc_prev_left[idx], cpu);
1120 1209
@@ -1199,6 +1288,16 @@ static int x86_pmu_handle_irq(struct pt_regs *regs)
1199 1288
1200 cpuc = &__get_cpu_var(cpu_hw_events); 1289 cpuc = &__get_cpu_var(cpu_hw_events);
1201 1290
1291 /*
1292 * Some chipsets need to unmask the LVTPC in a particular spot
1293 * inside the nmi handler. As a result, the unmasking was pushed
1294 * into all the nmi handlers.
1295 *
1296 * This generic handler doesn't seem to have any issues where the
1297 * unmasking occurs so it was left at the top.
1298 */
1299 apic_write(APIC_LVTPC, APIC_DM_NMI);
1300
1202 for (idx = 0; idx < x86_pmu.num_counters; idx++) { 1301 for (idx = 0; idx < x86_pmu.num_counters; idx++) {
1203 if (!test_bit(idx, cpuc->active_mask)) { 1302 if (!test_bit(idx, cpuc->active_mask)) {
1204 /* 1303 /*
@@ -1285,8 +1384,6 @@ perf_event_nmi_handler(struct notifier_block *self,
1285 return NOTIFY_DONE; 1384 return NOTIFY_DONE;
1286 } 1385 }
1287 1386
1288 apic_write(APIC_LVTPC, APIC_DM_NMI);
1289
1290 handled = x86_pmu.handle_irq(args->regs); 1387 handled = x86_pmu.handle_irq(args->regs);
1291 if (!handled) 1388 if (!handled)
1292 return NOTIFY_DONE; 1389 return NOTIFY_DONE;
@@ -1389,7 +1486,7 @@ static void __init pmu_check_apic(void)
1389 pr_info("no hardware sampling interrupt available.\n"); 1486 pr_info("no hardware sampling interrupt available.\n");
1390} 1487}
1391 1488
1392int __init init_hw_perf_events(void) 1489static int __init init_hw_perf_events(void)
1393{ 1490{
1394 struct event_constraint *c; 1491 struct event_constraint *c;
1395 int err; 1492 int err;
@@ -1608,7 +1705,7 @@ out:
1608 return ret; 1705 return ret;
1609} 1706}
1610 1707
1611int x86_pmu_event_init(struct perf_event *event) 1708static int x86_pmu_event_init(struct perf_event *event)
1612{ 1709{
1613 struct pmu *tmp; 1710 struct pmu *tmp;
1614 int err; 1711 int err;
@@ -1710,7 +1807,7 @@ perf_callchain_kernel(struct perf_callchain_entry *entry, struct pt_regs *regs)
1710 1807
1711 perf_callchain_store(entry, regs->ip); 1808 perf_callchain_store(entry, regs->ip);
1712 1809
1713 dump_trace(NULL, regs, NULL, &backtrace_ops, entry); 1810 dump_trace(NULL, regs, NULL, 0, &backtrace_ops, entry);
1714} 1811}
1715 1812
1716#ifdef CONFIG_COMPAT 1813#ifdef CONFIG_COMPAT
diff --git a/arch/x86/kernel/cpu/perf_event_amd.c b/arch/x86/kernel/cpu/perf_event_amd.c
index 67e2202a6039..cf4e369cea67 100644
--- a/arch/x86/kernel/cpu/perf_event_amd.c
+++ b/arch/x86/kernel/cpu/perf_event_amd.c
@@ -8,7 +8,7 @@ static __initconst const u64 amd_hw_cache_event_ids
8 [ C(L1D) ] = { 8 [ C(L1D) ] = {
9 [ C(OP_READ) ] = { 9 [ C(OP_READ) ] = {
10 [ C(RESULT_ACCESS) ] = 0x0040, /* Data Cache Accesses */ 10 [ C(RESULT_ACCESS) ] = 0x0040, /* Data Cache Accesses */
11 [ C(RESULT_MISS) ] = 0x0041, /* Data Cache Misses */ 11 [ C(RESULT_MISS) ] = 0x0141, /* Data Cache Misses */
12 }, 12 },
13 [ C(OP_WRITE) ] = { 13 [ C(OP_WRITE) ] = {
14 [ C(RESULT_ACCESS) ] = 0x0142, /* Data Cache Refills :system */ 14 [ C(RESULT_ACCESS) ] = 0x0142, /* Data Cache Refills :system */
@@ -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 +390,195 @@ 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 * 0x000 FP PERF_CTL[3], PERF_CTL[5:3] (*)
431 * 0x003 FP PERF_CTL[3]
432 * 0x004 FP PERF_CTL[3], PERF_CTL[5:3] (*)
433 * 0x00B FP PERF_CTL[3]
434 * 0x00D FP PERF_CTL[3]
435 * 0x023 DE PERF_CTL[2:0]
436 * 0x02D LS PERF_CTL[3]
437 * 0x02E LS PERF_CTL[3,0]
438 * 0x043 CU PERF_CTL[2:0]
439 * 0x045 CU PERF_CTL[2:0]
440 * 0x046 CU PERF_CTL[2:0]
441 * 0x054 CU PERF_CTL[2:0]
442 * 0x055 CU PERF_CTL[2:0]
443 * 0x08F IC PERF_CTL[0]
444 * 0x187 DE PERF_CTL[0]
445 * 0x188 DE PERF_CTL[0]
446 * 0x0DB EX PERF_CTL[5:0]
447 * 0x0DC LS PERF_CTL[5:0]
448 * 0x0DD LS PERF_CTL[5:0]
449 * 0x0DE LS PERF_CTL[5:0]
450 * 0x0DF LS PERF_CTL[5:0]
451 * 0x1D6 EX PERF_CTL[5:0]
452 * 0x1D8 EX PERF_CTL[5:0]
453 *
454 * (*) depending on the umask all FPU counters may be used
455 */
456
457static struct event_constraint amd_f15_PMC0 = EVENT_CONSTRAINT(0, 0x01, 0);
458static struct event_constraint amd_f15_PMC20 = EVENT_CONSTRAINT(0, 0x07, 0);
459static struct event_constraint amd_f15_PMC3 = EVENT_CONSTRAINT(0, 0x08, 0);
460static struct event_constraint amd_f15_PMC30 = EVENT_CONSTRAINT(0, 0x09, 0);
461static struct event_constraint amd_f15_PMC50 = EVENT_CONSTRAINT(0, 0x3F, 0);
462static struct event_constraint amd_f15_PMC53 = EVENT_CONSTRAINT(0, 0x38, 0);
463
464static struct event_constraint *
465amd_get_event_constraints_f15h(struct cpu_hw_events *cpuc, struct perf_event *event)
466{
467 struct hw_perf_event *hwc = &event->hw;
468 unsigned int event_code = amd_get_event_code(hwc);
469
470 switch (event_code & AMD_EVENT_TYPE_MASK) {
471 case AMD_EVENT_FP:
472 switch (event_code) {
473 case 0x000:
474 if (!(hwc->config & 0x0000F000ULL))
475 break;
476 if (!(hwc->config & 0x00000F00ULL))
477 break;
478 return &amd_f15_PMC3;
479 case 0x004:
480 if (hweight_long(hwc->config & ARCH_PERFMON_EVENTSEL_UMASK) <= 1)
481 break;
482 return &amd_f15_PMC3;
483 case 0x003:
484 case 0x00B:
485 case 0x00D:
486 return &amd_f15_PMC3;
487 }
488 return &amd_f15_PMC53;
489 case AMD_EVENT_LS:
490 case AMD_EVENT_DC:
491 case AMD_EVENT_EX_LS:
492 switch (event_code) {
493 case 0x023:
494 case 0x043:
495 case 0x045:
496 case 0x046:
497 case 0x054:
498 case 0x055:
499 return &amd_f15_PMC20;
500 case 0x02D:
501 return &amd_f15_PMC3;
502 case 0x02E:
503 return &amd_f15_PMC30;
504 default:
505 return &amd_f15_PMC50;
506 }
507 case AMD_EVENT_CU:
508 case AMD_EVENT_IC_DE:
509 case AMD_EVENT_DE:
510 switch (event_code) {
511 case 0x08F:
512 case 0x187:
513 case 0x188:
514 return &amd_f15_PMC0;
515 case 0x0DB ... 0x0DF:
516 case 0x1D6:
517 case 0x1D8:
518 return &amd_f15_PMC50;
519 default:
520 return &amd_f15_PMC20;
521 }
522 case AMD_EVENT_NB:
523 /* not yet implemented */
524 return &emptyconstraint;
525 default:
526 return &emptyconstraint;
527 }
528}
529
530static __initconst const struct x86_pmu amd_pmu_f15h = {
531 .name = "AMD Family 15h",
532 .handle_irq = x86_pmu_handle_irq,
533 .disable_all = x86_pmu_disable_all,
534 .enable_all = x86_pmu_enable_all,
535 .enable = x86_pmu_enable_event,
536 .disable = x86_pmu_disable_event,
537 .hw_config = amd_pmu_hw_config,
538 .schedule_events = x86_schedule_events,
539 .eventsel = MSR_F15H_PERF_CTL,
540 .perfctr = MSR_F15H_PERF_CTR,
541 .event_map = amd_pmu_event_map,
542 .max_events = ARRAY_SIZE(amd_perfmon_event_map),
543 .num_counters = 6,
544 .cntval_bits = 48,
545 .cntval_mask = (1ULL << 48) - 1,
546 .apic = 1,
547 /* use highest bit to detect overflow */
548 .max_period = (1ULL << 47) - 1,
549 .get_event_constraints = amd_get_event_constraints_f15h,
550 /* nortbridge counters not yet implemented: */
551#if 0
552 .put_event_constraints = amd_put_event_constraints,
553
554 .cpu_prepare = amd_pmu_cpu_prepare,
555 .cpu_starting = amd_pmu_cpu_starting,
556 .cpu_dead = amd_pmu_cpu_dead,
557#endif
558};
559
388static __init int amd_pmu_init(void) 560static __init int amd_pmu_init(void)
389{ 561{
390 /* Performance-monitoring supported from K7 and later: */ 562 /* Performance-monitoring supported from K7 and later: */
391 if (boot_cpu_data.x86 < 6) 563 if (boot_cpu_data.x86 < 6)
392 return -ENODEV; 564 return -ENODEV;
393 565
394 x86_pmu = amd_pmu; 566 /*
567 * If core performance counter extensions exists, it must be
568 * family 15h, otherwise fail. See x86_pmu_addr_offset().
569 */
570 switch (boot_cpu_data.x86) {
571 case 0x15:
572 if (!cpu_has_perfctr_core)
573 return -ENODEV;
574 x86_pmu = amd_pmu_f15h;
575 break;
576 default:
577 if (cpu_has_perfctr_core)
578 return -ENODEV;
579 x86_pmu = amd_pmu;
580 break;
581 }
395 582
396 /* Events are common for all AMDs */ 583 /* Events are common for all AMDs */
397 memcpy(hw_cache_event_ids, amd_hw_cache_event_ids, 584 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..447a28de6f09 100644
--- a/arch/x86/kernel/cpu/perf_event_intel.c
+++ b/arch/x86/kernel/cpu/perf_event_intel.c
@@ -1,9 +1,31 @@
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 */
6static const u64 intel_perfmon_event_map[] = 28static u64 intel_perfmon_event_map[PERF_COUNT_HW_MAX] __read_mostly =
7{ 29{
8 [PERF_COUNT_HW_CPU_CYCLES] = 0x003c, 30 [PERF_COUNT_HW_CPU_CYCLES] = 0x003c,
9 [PERF_COUNT_HW_INSTRUCTIONS] = 0x00c0, 31 [PERF_COUNT_HW_INSTRUCTIONS] = 0x00c0,
@@ -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 +150,103 @@ 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 [ C(OP_READ) ] = {
188 /* OFFCORE_RESPONSE.ANY_DATA.LOCAL_CACHE */
189 [ C(RESULT_ACCESS) ] = 0x01b7,
190 /* OFFCORE_RESPONSE.ANY_DATA.ANY_LLC_MISS */
191 [ C(RESULT_MISS) ] = 0x01b7,
192 },
193 [ C(OP_WRITE) ] = {
194 /* OFFCORE_RESPONSE.ANY_RFO.LOCAL_CACHE */
195 [ C(RESULT_ACCESS) ] = 0x01b7,
196 /* OFFCORE_RESPONSE.ANY_RFO.ANY_LLC_MISS */
197 [ C(RESULT_MISS) ] = 0x01b7,
198 },
199 [ C(OP_PREFETCH) ] = {
200 /* OFFCORE_RESPONSE.PREFETCH.LOCAL_CACHE */
201 [ C(RESULT_ACCESS) ] = 0x01b7,
202 /* OFFCORE_RESPONSE.PREFETCH.ANY_LLC_MISS */
203 [ C(RESULT_MISS) ] = 0x01b7,
204 },
205 },
206 [ C(DTLB) ] = {
207 [ C(OP_READ) ] = {
208 [ C(RESULT_ACCESS) ] = 0x81d0, /* MEM_UOP_RETIRED.ALL_LOADS */
209 [ C(RESULT_MISS) ] = 0x0108, /* DTLB_LOAD_MISSES.CAUSES_A_WALK */
210 },
211 [ C(OP_WRITE) ] = {
212 [ C(RESULT_ACCESS) ] = 0x82d0, /* MEM_UOP_RETIRED.ALL_STORES */
213 [ C(RESULT_MISS) ] = 0x0149, /* DTLB_STORE_MISSES.MISS_CAUSES_A_WALK */
214 },
215 [ C(OP_PREFETCH) ] = {
216 [ C(RESULT_ACCESS) ] = 0x0,
217 [ C(RESULT_MISS) ] = 0x0,
218 },
219 },
220 [ C(ITLB) ] = {
221 [ C(OP_READ) ] = {
222 [ C(RESULT_ACCESS) ] = 0x1085, /* ITLB_MISSES.STLB_HIT */
223 [ C(RESULT_MISS) ] = 0x0185, /* ITLB_MISSES.CAUSES_A_WALK */
224 },
225 [ C(OP_WRITE) ] = {
226 [ C(RESULT_ACCESS) ] = -1,
227 [ C(RESULT_MISS) ] = -1,
228 },
229 [ C(OP_PREFETCH) ] = {
230 [ C(RESULT_ACCESS) ] = -1,
231 [ C(RESULT_MISS) ] = -1,
232 },
233 },
234 [ C(BPU ) ] = {
235 [ C(OP_READ) ] = {
236 [ C(RESULT_ACCESS) ] = 0x00c4, /* BR_INST_RETIRED.ALL_BRANCHES */
237 [ C(RESULT_MISS) ] = 0x00c5, /* BR_MISP_RETIRED.ALL_BRANCHES */
238 },
239 [ C(OP_WRITE) ] = {
240 [ C(RESULT_ACCESS) ] = -1,
241 [ C(RESULT_MISS) ] = -1,
242 },
243 [ C(OP_PREFETCH) ] = {
244 [ C(RESULT_ACCESS) ] = -1,
245 [ C(RESULT_MISS) ] = -1,
246 },
247 },
248};
249
92static __initconst const u64 westmere_hw_cache_event_ids 250static __initconst const u64 westmere_hw_cache_event_ids
93 [PERF_COUNT_HW_CACHE_MAX] 251 [PERF_COUNT_HW_CACHE_MAX]
94 [PERF_COUNT_HW_CACHE_OP_MAX] 252 [PERF_COUNT_HW_CACHE_OP_MAX]
@@ -124,16 +282,26 @@ static __initconst const u64 westmere_hw_cache_event_ids
124 }, 282 },
125 [ C(LL ) ] = { 283 [ C(LL ) ] = {
126 [ C(OP_READ) ] = { 284 [ C(OP_READ) ] = {
127 [ C(RESULT_ACCESS) ] = 0x0324, /* L2_RQSTS.LOADS */ 285 /* OFFCORE_RESPONSE.ANY_DATA.LOCAL_CACHE */
128 [ C(RESULT_MISS) ] = 0x0224, /* L2_RQSTS.LD_MISS */ 286 [ C(RESULT_ACCESS) ] = 0x01b7,
287 /* OFFCORE_RESPONSE.ANY_DATA.ANY_LLC_MISS */
288 [ C(RESULT_MISS) ] = 0x01b7,
129 }, 289 },
290 /*
291 * Use RFO, not WRITEBACK, because a write miss would typically occur
292 * on RFO.
293 */
130 [ C(OP_WRITE) ] = { 294 [ C(OP_WRITE) ] = {
131 [ C(RESULT_ACCESS) ] = 0x0c24, /* L2_RQSTS.RFOS */ 295 /* OFFCORE_RESPONSE.ANY_RFO.LOCAL_CACHE */
132 [ C(RESULT_MISS) ] = 0x0824, /* L2_RQSTS.RFO_MISS */ 296 [ C(RESULT_ACCESS) ] = 0x01b7,
297 /* OFFCORE_RESPONSE.ANY_RFO.ANY_LLC_MISS */
298 [ C(RESULT_MISS) ] = 0x01b7,
133 }, 299 },
134 [ C(OP_PREFETCH) ] = { 300 [ C(OP_PREFETCH) ] = {
135 [ C(RESULT_ACCESS) ] = 0x4f2e, /* LLC Reference */ 301 /* OFFCORE_RESPONSE.PREFETCH.LOCAL_CACHE */
136 [ C(RESULT_MISS) ] = 0x412e, /* LLC Misses */ 302 [ C(RESULT_ACCESS) ] = 0x01b7,
303 /* OFFCORE_RESPONSE.PREFETCH.ANY_LLC_MISS */
304 [ C(RESULT_MISS) ] = 0x01b7,
137 }, 305 },
138 }, 306 },
139 [ C(DTLB) ] = { 307 [ C(DTLB) ] = {
@@ -180,6 +348,59 @@ static __initconst const u64 westmere_hw_cache_event_ids
180 }, 348 },
181}; 349};
182 350
351/*
352 * Nehalem/Westmere MSR_OFFCORE_RESPONSE bits;
353 * See IA32 SDM Vol 3B 30.6.1.3
354 */
355
356#define NHM_DMND_DATA_RD (1 << 0)
357#define NHM_DMND_RFO (1 << 1)
358#define NHM_DMND_IFETCH (1 << 2)
359#define NHM_DMND_WB (1 << 3)
360#define NHM_PF_DATA_RD (1 << 4)
361#define NHM_PF_DATA_RFO (1 << 5)
362#define NHM_PF_IFETCH (1 << 6)
363#define NHM_OFFCORE_OTHER (1 << 7)
364#define NHM_UNCORE_HIT (1 << 8)
365#define NHM_OTHER_CORE_HIT_SNP (1 << 9)
366#define NHM_OTHER_CORE_HITM (1 << 10)
367 /* reserved */
368#define NHM_REMOTE_CACHE_FWD (1 << 12)
369#define NHM_REMOTE_DRAM (1 << 13)
370#define NHM_LOCAL_DRAM (1 << 14)
371#define NHM_NON_DRAM (1 << 15)
372
373#define NHM_ALL_DRAM (NHM_REMOTE_DRAM|NHM_LOCAL_DRAM)
374
375#define NHM_DMND_READ (NHM_DMND_DATA_RD)
376#define NHM_DMND_WRITE (NHM_DMND_RFO|NHM_DMND_WB)
377#define NHM_DMND_PREFETCH (NHM_PF_DATA_RD|NHM_PF_DATA_RFO)
378
379#define NHM_L3_HIT (NHM_UNCORE_HIT|NHM_OTHER_CORE_HIT_SNP|NHM_OTHER_CORE_HITM)
380#define NHM_L3_MISS (NHM_NON_DRAM|NHM_ALL_DRAM|NHM_REMOTE_CACHE_FWD)
381#define NHM_L3_ACCESS (NHM_L3_HIT|NHM_L3_MISS)
382
383static __initconst const u64 nehalem_hw_cache_extra_regs
384 [PERF_COUNT_HW_CACHE_MAX]
385 [PERF_COUNT_HW_CACHE_OP_MAX]
386 [PERF_COUNT_HW_CACHE_RESULT_MAX] =
387{
388 [ C(LL ) ] = {
389 [ C(OP_READ) ] = {
390 [ C(RESULT_ACCESS) ] = NHM_DMND_READ|NHM_L3_ACCESS,
391 [ C(RESULT_MISS) ] = NHM_DMND_READ|NHM_L3_MISS,
392 },
393 [ C(OP_WRITE) ] = {
394 [ C(RESULT_ACCESS) ] = NHM_DMND_WRITE|NHM_L3_ACCESS,
395 [ C(RESULT_MISS) ] = NHM_DMND_WRITE|NHM_L3_MISS,
396 },
397 [ C(OP_PREFETCH) ] = {
398 [ C(RESULT_ACCESS) ] = NHM_DMND_PREFETCH|NHM_L3_ACCESS,
399 [ C(RESULT_MISS) ] = NHM_DMND_PREFETCH|NHM_L3_MISS,
400 },
401 }
402};
403
183static __initconst const u64 nehalem_hw_cache_event_ids 404static __initconst const u64 nehalem_hw_cache_event_ids
184 [PERF_COUNT_HW_CACHE_MAX] 405 [PERF_COUNT_HW_CACHE_MAX]
185 [PERF_COUNT_HW_CACHE_OP_MAX] 406 [PERF_COUNT_HW_CACHE_OP_MAX]
@@ -187,12 +408,12 @@ static __initconst const u64 nehalem_hw_cache_event_ids
187{ 408{
188 [ C(L1D) ] = { 409 [ C(L1D) ] = {
189 [ C(OP_READ) ] = { 410 [ C(OP_READ) ] = {
190 [ C(RESULT_ACCESS) ] = 0x0f40, /* L1D_CACHE_LD.MESI */ 411 [ C(RESULT_ACCESS) ] = 0x010b, /* MEM_INST_RETIRED.LOADS */
191 [ C(RESULT_MISS) ] = 0x0140, /* L1D_CACHE_LD.I_STATE */ 412 [ C(RESULT_MISS) ] = 0x0151, /* L1D.REPL */
192 }, 413 },
193 [ C(OP_WRITE) ] = { 414 [ C(OP_WRITE) ] = {
194 [ C(RESULT_ACCESS) ] = 0x0f41, /* L1D_CACHE_ST.MESI */ 415 [ C(RESULT_ACCESS) ] = 0x020b, /* MEM_INST_RETURED.STORES */
195 [ C(RESULT_MISS) ] = 0x0141, /* L1D_CACHE_ST.I_STATE */ 416 [ C(RESULT_MISS) ] = 0x0251, /* L1D.M_REPL */
196 }, 417 },
197 [ C(OP_PREFETCH) ] = { 418 [ C(OP_PREFETCH) ] = {
198 [ C(RESULT_ACCESS) ] = 0x014e, /* L1D_PREFETCH.REQUESTS */ 419 [ C(RESULT_ACCESS) ] = 0x014e, /* L1D_PREFETCH.REQUESTS */
@@ -215,16 +436,26 @@ static __initconst const u64 nehalem_hw_cache_event_ids
215 }, 436 },
216 [ C(LL ) ] = { 437 [ C(LL ) ] = {
217 [ C(OP_READ) ] = { 438 [ C(OP_READ) ] = {
218 [ C(RESULT_ACCESS) ] = 0x0324, /* L2_RQSTS.LOADS */ 439 /* OFFCORE_RESPONSE.ANY_DATA.LOCAL_CACHE */
219 [ C(RESULT_MISS) ] = 0x0224, /* L2_RQSTS.LD_MISS */ 440 [ C(RESULT_ACCESS) ] = 0x01b7,
441 /* OFFCORE_RESPONSE.ANY_DATA.ANY_LLC_MISS */
442 [ C(RESULT_MISS) ] = 0x01b7,
220 }, 443 },
444 /*
445 * Use RFO, not WRITEBACK, because a write miss would typically occur
446 * on RFO.
447 */
221 [ C(OP_WRITE) ] = { 448 [ C(OP_WRITE) ] = {
222 [ C(RESULT_ACCESS) ] = 0x0c24, /* L2_RQSTS.RFOS */ 449 /* OFFCORE_RESPONSE.ANY_RFO.LOCAL_CACHE */
223 [ C(RESULT_MISS) ] = 0x0824, /* L2_RQSTS.RFO_MISS */ 450 [ C(RESULT_ACCESS) ] = 0x01b7,
451 /* OFFCORE_RESPONSE.ANY_RFO.ANY_LLC_MISS */
452 [ C(RESULT_MISS) ] = 0x01b7,
224 }, 453 },
225 [ C(OP_PREFETCH) ] = { 454 [ C(OP_PREFETCH) ] = {
226 [ C(RESULT_ACCESS) ] = 0x4f2e, /* LLC Reference */ 455 /* OFFCORE_RESPONSE.PREFETCH.LOCAL_CACHE */
227 [ C(RESULT_MISS) ] = 0x412e, /* LLC Misses */ 456 [ C(RESULT_ACCESS) ] = 0x01b7,
457 /* OFFCORE_RESPONSE.PREFETCH.ANY_LLC_MISS */
458 [ C(RESULT_MISS) ] = 0x01b7,
228 }, 459 },
229 }, 460 },
230 [ C(DTLB) ] = { 461 [ C(DTLB) ] = {
@@ -691,8 +922,8 @@ static void intel_pmu_reset(void)
691 printk("clearing PMU state on CPU#%d\n", smp_processor_id()); 922 printk("clearing PMU state on CPU#%d\n", smp_processor_id());
692 923
693 for (idx = 0; idx < x86_pmu.num_counters; idx++) { 924 for (idx = 0; idx < x86_pmu.num_counters; idx++) {
694 checking_wrmsrl(x86_pmu.eventsel + idx, 0ull); 925 checking_wrmsrl(x86_pmu_config_addr(idx), 0ull);
695 checking_wrmsrl(x86_pmu.perfctr + idx, 0ull); 926 checking_wrmsrl(x86_pmu_event_addr(idx), 0ull);
696 } 927 }
697 for (idx = 0; idx < x86_pmu.num_counters_fixed; idx++) 928 for (idx = 0; idx < x86_pmu.num_counters_fixed; idx++)
698 checking_wrmsrl(MSR_ARCH_PERFMON_FIXED_CTR0 + idx, 0ull); 929 checking_wrmsrl(MSR_ARCH_PERFMON_FIXED_CTR0 + idx, 0ull);
@@ -719,6 +950,16 @@ static int intel_pmu_handle_irq(struct pt_regs *regs)
719 950
720 cpuc = &__get_cpu_var(cpu_hw_events); 951 cpuc = &__get_cpu_var(cpu_hw_events);
721 952
953 /*
954 * Some chipsets need to unmask the LVTPC in a particular spot
955 * inside the nmi handler. As a result, the unmasking was pushed
956 * into all the nmi handlers.
957 *
958 * This handler doesn't seem to have any issues with the unmasking
959 * so it was left at the top.
960 */
961 apic_write(APIC_LVTPC, APIC_DM_NMI);
962
722 intel_pmu_disable_all(); 963 intel_pmu_disable_all();
723 handled = intel_pmu_drain_bts_buffer(); 964 handled = intel_pmu_drain_bts_buffer();
724 status = intel_pmu_get_status(); 965 status = intel_pmu_get_status();
@@ -784,6 +1025,9 @@ intel_bts_constraints(struct perf_event *event)
784 struct hw_perf_event *hwc = &event->hw; 1025 struct hw_perf_event *hwc = &event->hw;
785 unsigned int hw_event, bts_event; 1026 unsigned int hw_event, bts_event;
786 1027
1028 if (event->attr.freq)
1029 return NULL;
1030
787 hw_event = hwc->config & INTEL_ARCH_EVENT_MASK; 1031 hw_event = hwc->config & INTEL_ARCH_EVENT_MASK;
788 bts_event = x86_pmu.event_map(PERF_COUNT_HW_BRANCH_INSTRUCTIONS); 1032 bts_event = x86_pmu.event_map(PERF_COUNT_HW_BRANCH_INSTRUCTIONS);
789 1033
@@ -794,6 +1038,67 @@ intel_bts_constraints(struct perf_event *event)
794} 1038}
795 1039
796static struct event_constraint * 1040static struct event_constraint *
1041intel_percore_constraints(struct cpu_hw_events *cpuc, struct perf_event *event)
1042{
1043 struct hw_perf_event *hwc = &event->hw;
1044 unsigned int e = hwc->config & ARCH_PERFMON_EVENTSEL_EVENT;
1045 struct event_constraint *c;
1046 struct intel_percore *pc;
1047 struct er_account *era;
1048 int i;
1049 int free_slot;
1050 int found;
1051
1052 if (!x86_pmu.percore_constraints || hwc->extra_alloc)
1053 return NULL;
1054
1055 for (c = x86_pmu.percore_constraints; c->cmask; c++) {
1056 if (e != c->code)
1057 continue;
1058
1059 /*
1060 * Allocate resource per core.
1061 */
1062 pc = cpuc->per_core;
1063 if (!pc)
1064 break;
1065 c = &emptyconstraint;
1066 raw_spin_lock(&pc->lock);
1067 free_slot = -1;
1068 found = 0;
1069 for (i = 0; i < MAX_EXTRA_REGS; i++) {
1070 era = &pc->regs[i];
1071 if (era->ref > 0 && hwc->extra_reg == era->extra_reg) {
1072 /* Allow sharing same config */
1073 if (hwc->extra_config == era->extra_config) {
1074 era->ref++;
1075 cpuc->percore_used = 1;
1076 hwc->extra_alloc = 1;
1077 c = NULL;
1078 }
1079 /* else conflict */
1080 found = 1;
1081 break;
1082 } else if (era->ref == 0 && free_slot == -1)
1083 free_slot = i;
1084 }
1085 if (!found && free_slot != -1) {
1086 era = &pc->regs[free_slot];
1087 era->ref = 1;
1088 era->extra_reg = hwc->extra_reg;
1089 era->extra_config = hwc->extra_config;
1090 cpuc->percore_used = 1;
1091 hwc->extra_alloc = 1;
1092 c = NULL;
1093 }
1094 raw_spin_unlock(&pc->lock);
1095 return c;
1096 }
1097
1098 return NULL;
1099}
1100
1101static struct event_constraint *
797intel_get_event_constraints(struct cpu_hw_events *cpuc, struct perf_event *event) 1102intel_get_event_constraints(struct cpu_hw_events *cpuc, struct perf_event *event)
798{ 1103{
799 struct event_constraint *c; 1104 struct event_constraint *c;
@@ -806,9 +1111,51 @@ intel_get_event_constraints(struct cpu_hw_events *cpuc, struct perf_event *event
806 if (c) 1111 if (c)
807 return c; 1112 return c;
808 1113
1114 c = intel_percore_constraints(cpuc, event);
1115 if (c)
1116 return c;
1117
809 return x86_get_event_constraints(cpuc, event); 1118 return x86_get_event_constraints(cpuc, event);
810} 1119}
811 1120
1121static void intel_put_event_constraints(struct cpu_hw_events *cpuc,
1122 struct perf_event *event)
1123{
1124 struct extra_reg *er;
1125 struct intel_percore *pc;
1126 struct er_account *era;
1127 struct hw_perf_event *hwc = &event->hw;
1128 int i, allref;
1129
1130 if (!cpuc->percore_used)
1131 return;
1132
1133 for (er = x86_pmu.extra_regs; er->msr; er++) {
1134 if (er->event != (hwc->config & er->config_mask))
1135 continue;
1136
1137 pc = cpuc->per_core;
1138 raw_spin_lock(&pc->lock);
1139 for (i = 0; i < MAX_EXTRA_REGS; i++) {
1140 era = &pc->regs[i];
1141 if (era->ref > 0 &&
1142 era->extra_config == hwc->extra_config &&
1143 era->extra_reg == er->msr) {
1144 era->ref--;
1145 hwc->extra_alloc = 0;
1146 break;
1147 }
1148 }
1149 allref = 0;
1150 for (i = 0; i < MAX_EXTRA_REGS; i++)
1151 allref += pc->regs[i].ref;
1152 if (allref == 0)
1153 cpuc->percore_used = 0;
1154 raw_spin_unlock(&pc->lock);
1155 break;
1156 }
1157}
1158
812static int intel_pmu_hw_config(struct perf_event *event) 1159static int intel_pmu_hw_config(struct perf_event *event)
813{ 1160{
814 int ret = x86_pmu_hw_config(event); 1161 int ret = x86_pmu_hw_config(event);
@@ -880,20 +1227,67 @@ static __initconst const struct x86_pmu core_pmu = {
880 */ 1227 */
881 .max_period = (1ULL << 31) - 1, 1228 .max_period = (1ULL << 31) - 1,
882 .get_event_constraints = intel_get_event_constraints, 1229 .get_event_constraints = intel_get_event_constraints,
1230 .put_event_constraints = intel_put_event_constraints,
883 .event_constraints = intel_core_event_constraints, 1231 .event_constraints = intel_core_event_constraints,
884}; 1232};
885 1233
1234static int intel_pmu_cpu_prepare(int cpu)
1235{
1236 struct cpu_hw_events *cpuc = &per_cpu(cpu_hw_events, cpu);
1237
1238 if (!cpu_has_ht_siblings())
1239 return NOTIFY_OK;
1240
1241 cpuc->per_core = kzalloc_node(sizeof(struct intel_percore),
1242 GFP_KERNEL, cpu_to_node(cpu));
1243 if (!cpuc->per_core)
1244 return NOTIFY_BAD;
1245
1246 raw_spin_lock_init(&cpuc->per_core->lock);
1247 cpuc->per_core->core_id = -1;
1248 return NOTIFY_OK;
1249}
1250
886static void intel_pmu_cpu_starting(int cpu) 1251static void intel_pmu_cpu_starting(int cpu)
887{ 1252{
1253 struct cpu_hw_events *cpuc = &per_cpu(cpu_hw_events, cpu);
1254 int core_id = topology_core_id(cpu);
1255 int i;
1256
888 init_debug_store_on_cpu(cpu); 1257 init_debug_store_on_cpu(cpu);
889 /* 1258 /*
890 * Deal with CPUs that don't clear their LBRs on power-up. 1259 * Deal with CPUs that don't clear their LBRs on power-up.
891 */ 1260 */
892 intel_pmu_lbr_reset(); 1261 intel_pmu_lbr_reset();
1262
1263 if (!cpu_has_ht_siblings())
1264 return;
1265
1266 for_each_cpu(i, topology_thread_cpumask(cpu)) {
1267 struct intel_percore *pc = per_cpu(cpu_hw_events, i).per_core;
1268
1269 if (pc && pc->core_id == core_id) {
1270 kfree(cpuc->per_core);
1271 cpuc->per_core = pc;
1272 break;
1273 }
1274 }
1275
1276 cpuc->per_core->core_id = core_id;
1277 cpuc->per_core->refcnt++;
893} 1278}
894 1279
895static void intel_pmu_cpu_dying(int cpu) 1280static void intel_pmu_cpu_dying(int cpu)
896{ 1281{
1282 struct cpu_hw_events *cpuc = &per_cpu(cpu_hw_events, cpu);
1283 struct intel_percore *pc = cpuc->per_core;
1284
1285 if (pc) {
1286 if (pc->core_id == -1 || --pc->refcnt == 0)
1287 kfree(pc);
1288 cpuc->per_core = NULL;
1289 }
1290
897 fini_debug_store_on_cpu(cpu); 1291 fini_debug_store_on_cpu(cpu);
898} 1292}
899 1293
@@ -918,7 +1312,9 @@ static __initconst const struct x86_pmu intel_pmu = {
918 */ 1312 */
919 .max_period = (1ULL << 31) - 1, 1313 .max_period = (1ULL << 31) - 1,
920 .get_event_constraints = intel_get_event_constraints, 1314 .get_event_constraints = intel_get_event_constraints,
1315 .put_event_constraints = intel_put_event_constraints,
921 1316
1317 .cpu_prepare = intel_pmu_cpu_prepare,
922 .cpu_starting = intel_pmu_cpu_starting, 1318 .cpu_starting = intel_pmu_cpu_starting,
923 .cpu_dying = intel_pmu_cpu_dying, 1319 .cpu_dying = intel_pmu_cpu_dying,
924}; 1320};
@@ -939,7 +1335,7 @@ static void intel_clovertown_quirks(void)
939 * AJ106 could possibly be worked around by not allowing LBR 1335 * AJ106 could possibly be worked around by not allowing LBR
940 * usage from PEBS, including the fixup. 1336 * usage from PEBS, including the fixup.
941 * AJ68 could possibly be worked around by always programming 1337 * AJ68 could possibly be worked around by always programming
942 * a pebs_event_reset[0] value and coping with the lost events. 1338 * a pebs_event_reset[0] value and coping with the lost events.
943 * 1339 *
944 * But taken together it might just make sense to not enable PEBS on 1340 * But taken together it might just make sense to not enable PEBS on
945 * these chips. 1341 * these chips.
@@ -1024,6 +1420,7 @@ static __init int intel_pmu_init(void)
1024 intel_pmu_lbr_init_core(); 1420 intel_pmu_lbr_init_core();
1025 1421
1026 x86_pmu.event_constraints = intel_core2_event_constraints; 1422 x86_pmu.event_constraints = intel_core2_event_constraints;
1423 x86_pmu.pebs_constraints = intel_core2_pebs_event_constraints;
1027 pr_cont("Core2 events, "); 1424 pr_cont("Core2 events, ");
1028 break; 1425 break;
1029 1426
@@ -1032,11 +1429,28 @@ static __init int intel_pmu_init(void)
1032 case 46: /* 45 nm nehalem-ex, "Beckton" */ 1429 case 46: /* 45 nm nehalem-ex, "Beckton" */
1033 memcpy(hw_cache_event_ids, nehalem_hw_cache_event_ids, 1430 memcpy(hw_cache_event_ids, nehalem_hw_cache_event_ids,
1034 sizeof(hw_cache_event_ids)); 1431 sizeof(hw_cache_event_ids));
1432 memcpy(hw_cache_extra_regs, nehalem_hw_cache_extra_regs,
1433 sizeof(hw_cache_extra_regs));
1035 1434
1036 intel_pmu_lbr_init_nhm(); 1435 intel_pmu_lbr_init_nhm();
1037 1436
1038 x86_pmu.event_constraints = intel_nehalem_event_constraints; 1437 x86_pmu.event_constraints = intel_nehalem_event_constraints;
1438 x86_pmu.pebs_constraints = intel_nehalem_pebs_event_constraints;
1439 x86_pmu.percore_constraints = intel_nehalem_percore_constraints;
1039 x86_pmu.enable_all = intel_pmu_nhm_enable_all; 1440 x86_pmu.enable_all = intel_pmu_nhm_enable_all;
1441 x86_pmu.extra_regs = intel_nehalem_extra_regs;
1442
1443 if (ebx & 0x40) {
1444 /*
1445 * Erratum AAJ80 detected, we work it around by using
1446 * the BR_MISP_EXEC.ANY event. This will over-count
1447 * branch-misses, but it's still much better than the
1448 * architectural event which is often completely bogus:
1449 */
1450 intel_perfmon_event_map[PERF_COUNT_HW_BRANCH_MISSES] = 0x7f89;
1451
1452 pr_cont("erratum AAJ80 worked around, ");
1453 }
1040 pr_cont("Nehalem events, "); 1454 pr_cont("Nehalem events, ");
1041 break; 1455 break;
1042 1456
@@ -1047,21 +1461,39 @@ static __init int intel_pmu_init(void)
1047 intel_pmu_lbr_init_atom(); 1461 intel_pmu_lbr_init_atom();
1048 1462
1049 x86_pmu.event_constraints = intel_gen_event_constraints; 1463 x86_pmu.event_constraints = intel_gen_event_constraints;
1464 x86_pmu.pebs_constraints = intel_atom_pebs_event_constraints;
1050 pr_cont("Atom events, "); 1465 pr_cont("Atom events, ");
1051 break; 1466 break;
1052 1467
1053 case 37: /* 32 nm nehalem, "Clarkdale" */ 1468 case 37: /* 32 nm nehalem, "Clarkdale" */
1054 case 44: /* 32 nm nehalem, "Gulftown" */ 1469 case 44: /* 32 nm nehalem, "Gulftown" */
1470 case 47: /* 32 nm Xeon E7 */
1055 memcpy(hw_cache_event_ids, westmere_hw_cache_event_ids, 1471 memcpy(hw_cache_event_ids, westmere_hw_cache_event_ids,
1056 sizeof(hw_cache_event_ids)); 1472 sizeof(hw_cache_event_ids));
1473 memcpy(hw_cache_extra_regs, nehalem_hw_cache_extra_regs,
1474 sizeof(hw_cache_extra_regs));
1057 1475
1058 intel_pmu_lbr_init_nhm(); 1476 intel_pmu_lbr_init_nhm();
1059 1477
1060 x86_pmu.event_constraints = intel_westmere_event_constraints; 1478 x86_pmu.event_constraints = intel_westmere_event_constraints;
1479 x86_pmu.percore_constraints = intel_westmere_percore_constraints;
1061 x86_pmu.enable_all = intel_pmu_nhm_enable_all; 1480 x86_pmu.enable_all = intel_pmu_nhm_enable_all;
1481 x86_pmu.pebs_constraints = intel_westmere_pebs_event_constraints;
1482 x86_pmu.extra_regs = intel_westmere_extra_regs;
1062 pr_cont("Westmere events, "); 1483 pr_cont("Westmere events, ");
1063 break; 1484 break;
1064 1485
1486 case 42: /* SandyBridge */
1487 memcpy(hw_cache_event_ids, snb_hw_cache_event_ids,
1488 sizeof(hw_cache_event_ids));
1489
1490 intel_pmu_lbr_init_nhm();
1491
1492 x86_pmu.event_constraints = intel_snb_event_constraints;
1493 x86_pmu.pebs_constraints = intel_snb_pebs_events;
1494 pr_cont("SandyBridge events, ");
1495 break;
1496
1065 default: 1497 default:
1066 /* 1498 /*
1067 * default constraints for v2 and up 1499 * 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..bab491b8ee25 100644
--- a/arch/x86/kernel/cpu/perf_event_intel_ds.c
+++ b/arch/x86/kernel/cpu/perf_event_intel_ds.c
@@ -361,30 +361,70 @@ static int intel_pmu_drain_bts_buffer(void)
361/* 361/*
362 * PEBS 362 * PEBS
363 */ 363 */
364static struct event_constraint intel_core2_pebs_event_constraints[] = {
365 INTEL_UEVENT_CONSTRAINT(0x00c0, 0x1), /* INST_RETIRED.ANY */
366 INTEL_UEVENT_CONSTRAINT(0xfec1, 0x1), /* X87_OPS_RETIRED.ANY */
367 INTEL_UEVENT_CONSTRAINT(0x00c5, 0x1), /* BR_INST_RETIRED.MISPRED */
368 INTEL_UEVENT_CONSTRAINT(0x1fc7, 0x1), /* SIMD_INST_RETURED.ANY */
369 INTEL_EVENT_CONSTRAINT(0xcb, 0x1), /* MEM_LOAD_RETIRED.* */
370 EVENT_CONSTRAINT_END
371};
372
373static struct event_constraint intel_atom_pebs_event_constraints[] = {
374 INTEL_UEVENT_CONSTRAINT(0x00c0, 0x1), /* INST_RETIRED.ANY */
375 INTEL_UEVENT_CONSTRAINT(0x00c5, 0x1), /* MISPREDICTED_BRANCH_RETIRED */
376 INTEL_EVENT_CONSTRAINT(0xcb, 0x1), /* MEM_LOAD_RETIRED.* */
377 EVENT_CONSTRAINT_END
378};
364 379
365static struct event_constraint intel_core_pebs_events[] = { 380static struct event_constraint intel_nehalem_pebs_event_constraints[] = {
366 PEBS_EVENT_CONSTRAINT(0x00c0, 0x1), /* INSTR_RETIRED.ANY */ 381 INTEL_EVENT_CONSTRAINT(0x0b, 0xf), /* MEM_INST_RETIRED.* */
367 PEBS_EVENT_CONSTRAINT(0xfec1, 0x1), /* X87_OPS_RETIRED.ANY */ 382 INTEL_EVENT_CONSTRAINT(0x0f, 0xf), /* MEM_UNCORE_RETIRED.* */
368 PEBS_EVENT_CONSTRAINT(0x00c5, 0x1), /* BR_INST_RETIRED.MISPRED */ 383 INTEL_UEVENT_CONSTRAINT(0x010c, 0xf), /* MEM_STORE_RETIRED.DTLB_MISS */
369 PEBS_EVENT_CONSTRAINT(0x1fc7, 0x1), /* SIMD_INST_RETURED.ANY */ 384 INTEL_EVENT_CONSTRAINT(0xc0, 0xf), /* INST_RETIRED.ANY */
370 PEBS_EVENT_CONSTRAINT(0x01cb, 0x1), /* MEM_LOAD_RETIRED.L1D_MISS */ 385 INTEL_EVENT_CONSTRAINT(0xc2, 0xf), /* UOPS_RETIRED.* */
371 PEBS_EVENT_CONSTRAINT(0x02cb, 0x1), /* MEM_LOAD_RETIRED.L1D_LINE_MISS */ 386 INTEL_EVENT_CONSTRAINT(0xc4, 0xf), /* BR_INST_RETIRED.* */
372 PEBS_EVENT_CONSTRAINT(0x04cb, 0x1), /* MEM_LOAD_RETIRED.L2_MISS */ 387 INTEL_UEVENT_CONSTRAINT(0x02c5, 0xf), /* BR_MISP_RETIRED.NEAR_CALL */
373 PEBS_EVENT_CONSTRAINT(0x08cb, 0x1), /* MEM_LOAD_RETIRED.L2_LINE_MISS */ 388 INTEL_EVENT_CONSTRAINT(0xc7, 0xf), /* SSEX_UOPS_RETIRED.* */
374 PEBS_EVENT_CONSTRAINT(0x10cb, 0x1), /* MEM_LOAD_RETIRED.DTLB_MISS */ 389 INTEL_UEVENT_CONSTRAINT(0x20c8, 0xf), /* ITLB_MISS_RETIRED */
390 INTEL_EVENT_CONSTRAINT(0xcb, 0xf), /* MEM_LOAD_RETIRED.* */
391 INTEL_EVENT_CONSTRAINT(0xf7, 0xf), /* FP_ASSIST.* */
375 EVENT_CONSTRAINT_END 392 EVENT_CONSTRAINT_END
376}; 393};
377 394
378static struct event_constraint intel_nehalem_pebs_events[] = { 395static struct event_constraint intel_westmere_pebs_event_constraints[] = {
379 PEBS_EVENT_CONSTRAINT(0x00c0, 0xf), /* INSTR_RETIRED.ANY */ 396 INTEL_EVENT_CONSTRAINT(0x0b, 0xf), /* MEM_INST_RETIRED.* */
380 PEBS_EVENT_CONSTRAINT(0xfec1, 0xf), /* X87_OPS_RETIRED.ANY */ 397 INTEL_EVENT_CONSTRAINT(0x0f, 0xf), /* MEM_UNCORE_RETIRED.* */
381 PEBS_EVENT_CONSTRAINT(0x00c5, 0xf), /* BR_INST_RETIRED.MISPRED */ 398 INTEL_UEVENT_CONSTRAINT(0x010c, 0xf), /* MEM_STORE_RETIRED.DTLB_MISS */
382 PEBS_EVENT_CONSTRAINT(0x1fc7, 0xf), /* SIMD_INST_RETURED.ANY */ 399 INTEL_EVENT_CONSTRAINT(0xc0, 0xf), /* INSTR_RETIRED.* */
383 PEBS_EVENT_CONSTRAINT(0x01cb, 0xf), /* MEM_LOAD_RETIRED.L1D_MISS */ 400 INTEL_EVENT_CONSTRAINT(0xc2, 0xf), /* UOPS_RETIRED.* */
384 PEBS_EVENT_CONSTRAINT(0x02cb, 0xf), /* MEM_LOAD_RETIRED.L1D_LINE_MISS */ 401 INTEL_EVENT_CONSTRAINT(0xc4, 0xf), /* BR_INST_RETIRED.* */
385 PEBS_EVENT_CONSTRAINT(0x04cb, 0xf), /* MEM_LOAD_RETIRED.L2_MISS */ 402 INTEL_EVENT_CONSTRAINT(0xc5, 0xf), /* BR_MISP_RETIRED.* */
386 PEBS_EVENT_CONSTRAINT(0x08cb, 0xf), /* MEM_LOAD_RETIRED.L2_LINE_MISS */ 403 INTEL_EVENT_CONSTRAINT(0xc7, 0xf), /* SSEX_UOPS_RETIRED.* */
387 PEBS_EVENT_CONSTRAINT(0x10cb, 0xf), /* MEM_LOAD_RETIRED.DTLB_MISS */ 404 INTEL_UEVENT_CONSTRAINT(0x20c8, 0xf), /* ITLB_MISS_RETIRED */
405 INTEL_EVENT_CONSTRAINT(0xcb, 0xf), /* MEM_LOAD_RETIRED.* */
406 INTEL_EVENT_CONSTRAINT(0xf7, 0xf), /* FP_ASSIST.* */
407 EVENT_CONSTRAINT_END
408};
409
410static struct event_constraint intel_snb_pebs_events[] = {
411 INTEL_UEVENT_CONSTRAINT(0x01c0, 0x2), /* INST_RETIRED.PRECDIST */
412 INTEL_UEVENT_CONSTRAINT(0x01c2, 0xf), /* UOPS_RETIRED.ALL */
413 INTEL_UEVENT_CONSTRAINT(0x02c2, 0xf), /* UOPS_RETIRED.RETIRE_SLOTS */
414 INTEL_EVENT_CONSTRAINT(0xc4, 0xf), /* BR_INST_RETIRED.* */
415 INTEL_EVENT_CONSTRAINT(0xc5, 0xf), /* BR_MISP_RETIRED.* */
416 INTEL_EVENT_CONSTRAINT(0xcd, 0x8), /* MEM_TRANS_RETIRED.* */
417 INTEL_UEVENT_CONSTRAINT(0x11d0, 0xf), /* MEM_UOP_RETIRED.STLB_MISS_LOADS */
418 INTEL_UEVENT_CONSTRAINT(0x12d0, 0xf), /* MEM_UOP_RETIRED.STLB_MISS_STORES */
419 INTEL_UEVENT_CONSTRAINT(0x21d0, 0xf), /* MEM_UOP_RETIRED.LOCK_LOADS */
420 INTEL_UEVENT_CONSTRAINT(0x22d0, 0xf), /* MEM_UOP_RETIRED.LOCK_STORES */
421 INTEL_UEVENT_CONSTRAINT(0x41d0, 0xf), /* MEM_UOP_RETIRED.SPLIT_LOADS */
422 INTEL_UEVENT_CONSTRAINT(0x42d0, 0xf), /* MEM_UOP_RETIRED.SPLIT_STORES */
423 INTEL_UEVENT_CONSTRAINT(0x81d0, 0xf), /* MEM_UOP_RETIRED.ANY_LOADS */
424 INTEL_UEVENT_CONSTRAINT(0x82d0, 0xf), /* MEM_UOP_RETIRED.ANY_STORES */
425 INTEL_EVENT_CONSTRAINT(0xd1, 0xf), /* MEM_LOAD_UOPS_RETIRED.* */
426 INTEL_EVENT_CONSTRAINT(0xd2, 0xf), /* MEM_LOAD_UOPS_LLC_HIT_RETIRED.* */
427 INTEL_UEVENT_CONSTRAINT(0x02d4, 0xf), /* MEM_LOAD_UOPS_MISC_RETIRED.LLC_MISS */
388 EVENT_CONSTRAINT_END 428 EVENT_CONSTRAINT_END
389}; 429};
390 430
@@ -695,20 +735,17 @@ static void intel_ds_init(void)
695 printk(KERN_CONT "PEBS fmt0%c, ", pebs_type); 735 printk(KERN_CONT "PEBS fmt0%c, ", pebs_type);
696 x86_pmu.pebs_record_size = sizeof(struct pebs_record_core); 736 x86_pmu.pebs_record_size = sizeof(struct pebs_record_core);
697 x86_pmu.drain_pebs = intel_pmu_drain_pebs_core; 737 x86_pmu.drain_pebs = intel_pmu_drain_pebs_core;
698 x86_pmu.pebs_constraints = intel_core_pebs_events;
699 break; 738 break;
700 739
701 case 1: 740 case 1:
702 printk(KERN_CONT "PEBS fmt1%c, ", pebs_type); 741 printk(KERN_CONT "PEBS fmt1%c, ", pebs_type);
703 x86_pmu.pebs_record_size = sizeof(struct pebs_record_nhm); 742 x86_pmu.pebs_record_size = sizeof(struct pebs_record_nhm);
704 x86_pmu.drain_pebs = intel_pmu_drain_pebs_nhm; 743 x86_pmu.drain_pebs = intel_pmu_drain_pebs_nhm;
705 x86_pmu.pebs_constraints = intel_nehalem_pebs_events;
706 break; 744 break;
707 745
708 default: 746 default:
709 printk(KERN_CONT "no PEBS fmt%d%c, ", format, pebs_type); 747 printk(KERN_CONT "no PEBS fmt%d%c, ", format, pebs_type);
710 x86_pmu.pebs = 0; 748 x86_pmu.pebs = 0;
711 break;
712 } 749 }
713 } 750 }
714} 751}
diff --git a/arch/x86/kernel/cpu/perf_event_p4.c b/arch/x86/kernel/cpu/perf_event_p4.c
index f7a0993c1e7c..e93fcd55fae1 100644
--- a/arch/x86/kernel/cpu/perf_event_p4.c
+++ b/arch/x86/kernel/cpu/perf_event_p4.c
@@ -1,5 +1,5 @@
1/* 1/*
2 * Netburst Perfomance Events (P4, old Xeon) 2 * Netburst Performance Events (P4, old Xeon)
3 * 3 *
4 * Copyright (C) 2010 Parallels, Inc., Cyrill Gorcunov <gorcunov@openvz.org> 4 * Copyright (C) 2010 Parallels, Inc., Cyrill Gorcunov <gorcunov@openvz.org>
5 * Copyright (C) 2010 Intel Corporation, Lin Ming <ming.m.lin@intel.com> 5 * Copyright (C) 2010 Intel Corporation, Lin Ming <ming.m.lin@intel.com>
@@ -679,7 +679,7 @@ static int p4_validate_raw_event(struct perf_event *event)
679 */ 679 */
680 680
681 /* 681 /*
682 * if an event is shared accross the logical threads 682 * if an event is shared across the logical threads
683 * the user needs special permissions to be able to use it 683 * the user needs special permissions to be able to use it
684 */ 684 */
685 if (p4_ht_active() && p4_event_bind_map[v].shared) { 685 if (p4_ht_active() && p4_event_bind_map[v].shared) {
@@ -764,15 +764,21 @@ 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
773 /* it might be unflagged overflow */ 773 /*
774 rdmsrl(hwc->event_base + hwc->idx, v); 774 * In some circumstances the overflow might issue an NMI but did
775 if (!(v & ARCH_P4_CNTRVAL_MASK)) 775 * not set P4_CCCR_OVF bit. Because a counter holds a negative value
776 * we simply check for high bit being set, if it's cleared it means
777 * the counter has reached zero value and continued counting before
778 * real NMI signal was received:
779 */
780 rdmsrl(hwc->event_base, v);
781 if (!(v & ARCH_P4_UNFLAGGED_BIT))
776 return 1; 782 return 1;
777 783
778 return 0; 784 return 0;
@@ -785,13 +791,13 @@ static void p4_pmu_disable_pebs(void)
785 * 791 *
786 * It's still allowed that two threads setup same cache 792 * It's still allowed that two threads setup same cache
787 * events so we can't simply clear metrics until we knew 793 * events so we can't simply clear metrics until we knew
788 * noone is depending on us, so we need kind of counter 794 * no one is depending on us, so we need kind of counter
789 * for "ReplayEvent" users. 795 * for "ReplayEvent" users.
790 * 796 *
791 * What is more complex -- RAW events, if user (for some 797 * What is more complex -- RAW events, if user (for some
792 * reason) will pass some cache event metric with improper 798 * reason) will pass some cache event metric with improper
793 * event opcode -- it's fine from hardware point of view 799 * event opcode -- it's fine from hardware point of view
794 * but completely nonsence from "meaning" of such action. 800 * but completely nonsense from "meaning" of such action.
795 * 801 *
796 * So at moment let leave metrics turned on forever -- it's 802 * So at moment let leave metrics turned on forever -- it's
797 * ok for now but need to be revisited! 803 * ok for now but need to be revisited!
@@ -810,7 +816,7 @@ static inline void p4_pmu_disable_event(struct perf_event *event)
810 * state we need to clear P4_CCCR_OVF, otherwise interrupt get 816 * state we need to clear P4_CCCR_OVF, otherwise interrupt get
811 * asserted again and again 817 * asserted again and again
812 */ 818 */
813 (void)checking_wrmsrl(hwc->config_base + hwc->idx, 819 (void)checking_wrmsrl(hwc->config_base,
814 (u64)(p4_config_unpack_cccr(hwc->config)) & 820 (u64)(p4_config_unpack_cccr(hwc->config)) &
815 ~P4_CCCR_ENABLE & ~P4_CCCR_OVF & ~P4_CCCR_RESERVED); 821 ~P4_CCCR_ENABLE & ~P4_CCCR_OVF & ~P4_CCCR_RESERVED);
816} 822}
@@ -880,7 +886,7 @@ static void p4_pmu_enable_event(struct perf_event *event)
880 p4_pmu_enable_pebs(hwc->config); 886 p4_pmu_enable_pebs(hwc->config);
881 887
882 (void)checking_wrmsrl(escr_addr, escr_conf); 888 (void)checking_wrmsrl(escr_addr, escr_conf);
883 (void)checking_wrmsrl(hwc->config_base + hwc->idx, 889 (void)checking_wrmsrl(hwc->config_base,
884 (cccr & ~P4_CCCR_RESERVED) | P4_CCCR_ENABLE); 890 (cccr & ~P4_CCCR_RESERVED) | P4_CCCR_ENABLE);
885} 891}
886 892
@@ -941,14 +947,23 @@ static int p4_pmu_handle_irq(struct pt_regs *regs)
941 if (!x86_perf_event_set_period(event)) 947 if (!x86_perf_event_set_period(event))
942 continue; 948 continue;
943 if (perf_event_overflow(event, 1, &data, regs)) 949 if (perf_event_overflow(event, 1, &data, regs))
944 p4_pmu_disable_event(event); 950 x86_pmu_stop(event, 0);
945 } 951 }
946 952
947 if (handled) { 953 if (handled)
948 /* p4 quirk: unmask it again */
949 apic_write(APIC_LVTPC, apic_read(APIC_LVTPC) & ~APIC_LVT_MASKED);
950 inc_irq_stat(apic_perf_irqs); 954 inc_irq_stat(apic_perf_irqs);
951 } 955
956 /*
957 * When dealing with the unmasking of the LVTPC on P4 perf hw, it has
958 * been observed that the OVF bit flag has to be cleared first _before_
959 * the LVTPC can be unmasked.
960 *
961 * The reason is the NMI line will continue to be asserted while the OVF
962 * bit is set. This causes a second NMI to generate if the LVTPC is
963 * unmasked before the OVF bit is cleared, leading to unknown NMI
964 * messages.
965 */
966 apic_write(APIC_LVTPC, APIC_DM_NMI);
952 967
953 return handled; 968 return handled;
954} 969}
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/cpu/vmware.c b/arch/x86/kernel/cpu/vmware.c
index 227b0448960d..d22d0c4edcfd 100644
--- a/arch/x86/kernel/cpu/vmware.c
+++ b/arch/x86/kernel/cpu/vmware.c
@@ -86,7 +86,7 @@ static void __init vmware_platform_setup(void)
86} 86}
87 87
88/* 88/*
89 * While checking the dmi string infomation, just checking the product 89 * While checking the dmi string information, just checking the product
90 * serial key should be enough, as this will always have a VMware 90 * serial key should be enough, as this will always have a VMware
91 * specific string when running under VMware hypervisor. 91 * specific string when running under VMware hypervisor.
92 */ 92 */
diff --git a/arch/x86/kernel/crash_dump_32.c b/arch/x86/kernel/crash_dump_32.c
index d5cd13945d5a..642f75a68cd5 100644
--- a/arch/x86/kernel/crash_dump_32.c
+++ b/arch/x86/kernel/crash_dump_32.c
@@ -14,9 +14,6 @@
14 14
15static void *kdump_buf_page; 15static void *kdump_buf_page;
16 16
17/* Stores the physical address of elf header of crash image. */
18unsigned long long elfcorehdr_addr = ELFCORE_ADDR_MAX;
19
20static inline bool is_crashed_pfn_valid(unsigned long pfn) 17static inline bool is_crashed_pfn_valid(unsigned long pfn)
21{ 18{
22#ifndef CONFIG_X86_PAE 19#ifndef CONFIG_X86_PAE
diff --git a/arch/x86/kernel/crash_dump_64.c b/arch/x86/kernel/crash_dump_64.c
index 994828899e09..afa64adb75ee 100644
--- a/arch/x86/kernel/crash_dump_64.c
+++ b/arch/x86/kernel/crash_dump_64.c
@@ -10,9 +10,6 @@
10#include <linux/uaccess.h> 10#include <linux/uaccess.h>
11#include <linux/io.h> 11#include <linux/io.h>
12 12
13/* Stores the physical address of elf header of crash image. */
14unsigned long long elfcorehdr_addr = ELFCORE_ADDR_MAX;
15
16/** 13/**
17 * copy_oldmem_page - copy one page from "oldmem" 14 * copy_oldmem_page - copy one page from "oldmem"
18 * @pfn: page frame number to be copied 15 * @pfn: page frame number to be copied
diff --git a/arch/x86/kernel/devicetree.c b/arch/x86/kernel/devicetree.c
new file mode 100644
index 000000000000..e90f08458e6b
--- /dev/null
+++ b/arch/x86/kernel/devicetree.c
@@ -0,0 +1,439 @@
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 0;
69 if (type == IRQ_TYPE_NONE)
70 return virq;
71 irq_set_irq_type(virq, type);
72 return virq;
73}
74EXPORT_SYMBOL_GPL(irq_create_of_mapping);
75
76unsigned long pci_address_to_pio(phys_addr_t address)
77{
78 /*
79 * The ioport address can be directly used by inX / outX
80 */
81 BUG_ON(address >= (1 << 16));
82 return (unsigned long)address;
83}
84EXPORT_SYMBOL_GPL(pci_address_to_pio);
85
86void __init early_init_dt_scan_chosen_arch(unsigned long node)
87{
88 BUG();
89}
90
91void __init early_init_dt_add_memory_arch(u64 base, u64 size)
92{
93 BUG();
94}
95
96void * __init early_init_dt_alloc_memory_arch(u64 size, u64 align)
97{
98 return __alloc_bootmem(size, align, __pa(MAX_DMA_ADDRESS));
99}
100
101void __init add_dtb(u64 data)
102{
103 initial_dtb = data + offsetof(struct setup_data, data);
104}
105
106/*
107 * CE4100 ids. Will be moved to machine_device_initcall() once we have it.
108 */
109static struct of_device_id __initdata ce4100_ids[] = {
110 { .compatible = "intel,ce4100-cp", },
111 { .compatible = "isa", },
112 { .compatible = "pci", },
113 {},
114};
115
116static int __init add_bus_probe(void)
117{
118 if (!of_have_populated_dt())
119 return 0;
120
121 return of_platform_bus_probe(NULL, ce4100_ids, NULL);
122}
123module_init(add_bus_probe);
124
125#ifdef CONFIG_PCI
126static int x86_of_pci_irq_enable(struct pci_dev *dev)
127{
128 struct of_irq oirq;
129 u32 virq;
130 int ret;
131 u8 pin;
132
133 ret = pci_read_config_byte(dev, PCI_INTERRUPT_PIN, &pin);
134 if (ret)
135 return ret;
136 if (!pin)
137 return 0;
138
139 ret = of_irq_map_pci(dev, &oirq);
140 if (ret)
141 return ret;
142
143 virq = irq_create_of_mapping(oirq.controller, oirq.specifier,
144 oirq.size);
145 if (virq == 0)
146 return -EINVAL;
147 dev->irq = virq;
148 return 0;
149}
150
151static void x86_of_pci_irq_disable(struct pci_dev *dev)
152{
153}
154
155void __cpuinit x86_of_pci_init(void)
156{
157 struct device_node *np;
158
159 pcibios_enable_irq = x86_of_pci_irq_enable;
160 pcibios_disable_irq = x86_of_pci_irq_disable;
161
162 for_each_node_by_type(np, "pci") {
163 const void *prop;
164 struct pci_bus *bus;
165 unsigned int bus_min;
166 struct device_node *child;
167
168 prop = of_get_property(np, "bus-range", NULL);
169 if (!prop)
170 continue;
171 bus_min = be32_to_cpup(prop);
172
173 bus = pci_find_bus(0, bus_min);
174 if (!bus) {
175 printk(KERN_ERR "Can't find a node for bus %s.\n",
176 np->full_name);
177 continue;
178 }
179
180 if (bus->self)
181 bus->self->dev.of_node = np;
182 else
183 bus->dev.of_node = np;
184
185 for_each_child_of_node(np, child) {
186 struct pci_dev *dev;
187 u32 devfn;
188
189 prop = of_get_property(child, "reg", NULL);
190 if (!prop)
191 continue;
192
193 devfn = (be32_to_cpup(prop) >> 8) & 0xff;
194 dev = pci_get_slot(bus, devfn);
195 if (!dev)
196 continue;
197 dev->dev.of_node = child;
198 pci_dev_put(dev);
199 }
200 }
201}
202#endif
203
204static void __init dtb_setup_hpet(void)
205{
206#ifdef CONFIG_HPET_TIMER
207 struct device_node *dn;
208 struct resource r;
209 int ret;
210
211 dn = of_find_compatible_node(NULL, NULL, "intel,ce4100-hpet");
212 if (!dn)
213 return;
214 ret = of_address_to_resource(dn, 0, &r);
215 if (ret) {
216 WARN_ON(1);
217 return;
218 }
219 hpet_address = r.start;
220#endif
221}
222
223static void __init dtb_lapic_setup(void)
224{
225#ifdef CONFIG_X86_LOCAL_APIC
226 struct device_node *dn;
227 struct resource r;
228 int ret;
229
230 dn = of_find_compatible_node(NULL, NULL, "intel,ce4100-lapic");
231 if (!dn)
232 return;
233
234 ret = of_address_to_resource(dn, 0, &r);
235 if (WARN_ON(ret))
236 return;
237
238 /* Did the boot loader setup the local APIC ? */
239 if (!cpu_has_apic) {
240 if (apic_force_enable(r.start))
241 return;
242 }
243 smp_found_config = 1;
244 pic_mode = 1;
245 register_lapic_address(r.start);
246 generic_processor_info(boot_cpu_physical_apicid,
247 GET_APIC_VERSION(apic_read(APIC_LVR)));
248#endif
249}
250
251#ifdef CONFIG_X86_IO_APIC
252static unsigned int ioapic_id;
253
254static void __init dtb_add_ioapic(struct device_node *dn)
255{
256 struct resource r;
257 int ret;
258
259 ret = of_address_to_resource(dn, 0, &r);
260 if (ret) {
261 printk(KERN_ERR "Can't obtain address from node %s.\n",
262 dn->full_name);
263 return;
264 }
265 mp_register_ioapic(++ioapic_id, r.start, gsi_top);
266}
267
268static void __init dtb_ioapic_setup(void)
269{
270 struct device_node *dn;
271
272 for_each_compatible_node(dn, NULL, "intel,ce4100-ioapic")
273 dtb_add_ioapic(dn);
274
275 if (nr_ioapics) {
276 of_ioapic = 1;
277 return;
278 }
279 printk(KERN_ERR "Error: No information about IO-APIC in OF.\n");
280}
281#else
282static void __init dtb_ioapic_setup(void) {}
283#endif
284
285static void __init dtb_apic_setup(void)
286{
287 dtb_lapic_setup();
288 dtb_ioapic_setup();
289}
290
291#ifdef CONFIG_OF_FLATTREE
292static void __init x86_flattree_get_config(void)
293{
294 u32 size, map_len;
295 void *new_dtb;
296
297 if (!initial_dtb)
298 return;
299
300 map_len = max(PAGE_SIZE - (initial_dtb & ~PAGE_MASK),
301 (u64)sizeof(struct boot_param_header));
302
303 initial_boot_params = early_memremap(initial_dtb, map_len);
304 size = be32_to_cpu(initial_boot_params->totalsize);
305 if (map_len < size) {
306 early_iounmap(initial_boot_params, map_len);
307 initial_boot_params = early_memremap(initial_dtb, size);
308 map_len = size;
309 }
310
311 new_dtb = alloc_bootmem(size);
312 memcpy(new_dtb, initial_boot_params, size);
313 early_iounmap(initial_boot_params, map_len);
314
315 initial_boot_params = new_dtb;
316
317 /* root level address cells */
318 of_scan_flat_dt(early_init_dt_scan_root, NULL);
319
320 unflatten_device_tree();
321}
322#else
323static inline void x86_flattree_get_config(void) { }
324#endif
325
326void __init x86_dtb_init(void)
327{
328 x86_flattree_get_config();
329
330 if (!of_have_populated_dt())
331 return;
332
333 dtb_setup_hpet();
334 dtb_apic_setup();
335}
336
337#ifdef CONFIG_X86_IO_APIC
338
339struct of_ioapic_type {
340 u32 out_type;
341 u32 trigger;
342 u32 polarity;
343};
344
345static struct of_ioapic_type of_ioapic_type[] =
346{
347 {
348 .out_type = IRQ_TYPE_EDGE_RISING,
349 .trigger = IOAPIC_EDGE,
350 .polarity = 1,
351 },
352 {
353 .out_type = IRQ_TYPE_LEVEL_LOW,
354 .trigger = IOAPIC_LEVEL,
355 .polarity = 0,
356 },
357 {
358 .out_type = IRQ_TYPE_LEVEL_HIGH,
359 .trigger = IOAPIC_LEVEL,
360 .polarity = 1,
361 },
362 {
363 .out_type = IRQ_TYPE_EDGE_FALLING,
364 .trigger = IOAPIC_EDGE,
365 .polarity = 0,
366 },
367};
368
369static int ioapic_xlate(struct irq_domain *id, const u32 *intspec, u32 intsize,
370 u32 *out_hwirq, u32 *out_type)
371{
372 struct io_apic_irq_attr attr;
373 struct of_ioapic_type *it;
374 u32 line, idx, type;
375
376 if (intsize < 2)
377 return -EINVAL;
378
379 line = *intspec;
380 idx = (u32) id->priv;
381 *out_hwirq = line + mp_gsi_routing[idx].gsi_base;
382
383 intspec++;
384 type = *intspec;
385
386 if (type >= ARRAY_SIZE(of_ioapic_type))
387 return -EINVAL;
388
389 it = of_ioapic_type + type;
390 *out_type = it->out_type;
391
392 set_io_apic_irq_attr(&attr, idx, line, it->trigger, it->polarity);
393
394 return io_apic_setup_irq_pin_once(*out_hwirq, cpu_to_node(0), &attr);
395}
396
397static void __init ioapic_add_ofnode(struct device_node *np)
398{
399 struct resource r;
400 int i, ret;
401
402 ret = of_address_to_resource(np, 0, &r);
403 if (ret) {
404 printk(KERN_ERR "Failed to obtain address for %s\n",
405 np->full_name);
406 return;
407 }
408
409 for (i = 0; i < nr_ioapics; i++) {
410 if (r.start == mp_ioapics[i].apicaddr) {
411 struct irq_domain *id;
412
413 id = kzalloc(sizeof(*id), GFP_KERNEL);
414 BUG_ON(!id);
415 id->controller = np;
416 id->xlate = ioapic_xlate;
417 id->priv = (void *)i;
418 add_interrupt_host(id);
419 return;
420 }
421 }
422 printk(KERN_ERR "IOxAPIC at %s is not registered.\n", np->full_name);
423}
424
425void __init x86_add_irq_domains(void)
426{
427 struct device_node *dp;
428
429 if (!of_have_populated_dt())
430 return;
431
432 for_each_node_with_property(dp, "interrupt-controller") {
433 if (of_device_is_compatible(dp, "intel,ce4100-ioapic"))
434 ioapic_add_ofnode(dp);
435 }
436}
437#else
438void __init x86_add_irq_domains(void) { }
439#endif
diff --git a/arch/x86/kernel/dumpstack.c b/arch/x86/kernel/dumpstack.c
index df20723a6a1b..e2a3f0606da4 100644
--- a/arch/x86/kernel/dumpstack.c
+++ b/arch/x86/kernel/dumpstack.c
@@ -27,7 +27,7 @@ static int die_counter;
27 27
28void printk_address(unsigned long address, int reliable) 28void printk_address(unsigned long address, int reliable)
29{ 29{
30 printk(" [<%p>] %s%pS\n", (void *) address, 30 printk(" [<%p>] %s%pB\n", (void *) address,
31 reliable ? "" : "? ", (void *) address); 31 reliable ? "" : "? ", (void *) address);
32} 32}
33 33
@@ -175,21 +175,21 @@ static const struct stacktrace_ops print_trace_ops = {
175 175
176void 176void
177show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs, 177show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs,
178 unsigned long *stack, char *log_lvl) 178 unsigned long *stack, unsigned long bp, char *log_lvl)
179{ 179{
180 printk("%sCall Trace:\n", log_lvl); 180 printk("%sCall Trace:\n", log_lvl);
181 dump_trace(task, regs, stack, &print_trace_ops, log_lvl); 181 dump_trace(task, regs, stack, bp, &print_trace_ops, log_lvl);
182} 182}
183 183
184void show_trace(struct task_struct *task, struct pt_regs *regs, 184void show_trace(struct task_struct *task, struct pt_regs *regs,
185 unsigned long *stack) 185 unsigned long *stack, unsigned long bp)
186{ 186{
187 show_trace_log_lvl(task, regs, stack, ""); 187 show_trace_log_lvl(task, regs, stack, bp, "");
188} 188}
189 189
190void show_stack(struct task_struct *task, unsigned long *sp) 190void show_stack(struct task_struct *task, unsigned long *sp)
191{ 191{
192 show_stack_log_lvl(task, NULL, sp, ""); 192 show_stack_log_lvl(task, NULL, sp, 0, "");
193} 193}
194 194
195/* 195/*
@@ -197,14 +197,16 @@ void show_stack(struct task_struct *task, unsigned long *sp)
197 */ 197 */
198void dump_stack(void) 198void dump_stack(void)
199{ 199{
200 unsigned long bp;
200 unsigned long stack; 201 unsigned long stack;
201 202
203 bp = stack_frame(current, NULL);
202 printk("Pid: %d, comm: %.20s %s %s %.*s\n", 204 printk("Pid: %d, comm: %.20s %s %s %.*s\n",
203 current->pid, current->comm, print_tainted(), 205 current->pid, current->comm, print_tainted(),
204 init_utsname()->release, 206 init_utsname()->release,
205 (int)strcspn(init_utsname()->version, " "), 207 (int)strcspn(init_utsname()->version, " "),
206 init_utsname()->version); 208 init_utsname()->version);
207 show_trace(NULL, NULL, &stack); 209 show_trace(NULL, NULL, &stack, bp);
208} 210}
209EXPORT_SYMBOL(dump_stack); 211EXPORT_SYMBOL(dump_stack);
210 212
@@ -320,41 +322,6 @@ void die(const char *str, struct pt_regs *regs, long err)
320 oops_end(flags, regs, sig); 322 oops_end(flags, regs, sig);
321} 323}
322 324
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)
349{
350 if (!s)
351 return -EINVAL;
352 if (!strcmp(s, "panic"))
353 panic_on_oops = 1;
354 return 0;
355}
356early_param("oops", oops_setup);
357
358static int __init kstack_setup(char *s) 325static int __init kstack_setup(char *s)
359{ 326{
360 if (!s) 327 if (!s)
diff --git a/arch/x86/kernel/dumpstack_32.c b/arch/x86/kernel/dumpstack_32.c
index 74cc1eda384b..3b97a80ce329 100644
--- a/arch/x86/kernel/dumpstack_32.c
+++ b/arch/x86/kernel/dumpstack_32.c
@@ -17,12 +17,11 @@
17#include <asm/stacktrace.h> 17#include <asm/stacktrace.h>
18 18
19 19
20void dump_trace(struct task_struct *task, 20void dump_trace(struct task_struct *task, struct pt_regs *regs,
21 struct pt_regs *regs, unsigned long *stack, 21 unsigned long *stack, unsigned long bp,
22 const struct stacktrace_ops *ops, void *data) 22 const struct stacktrace_ops *ops, void *data)
23{ 23{
24 int graph = 0; 24 int graph = 0;
25 unsigned long bp;
26 25
27 if (!task) 26 if (!task)
28 task = current; 27 task = current;
@@ -35,7 +34,9 @@ void dump_trace(struct task_struct *task,
35 stack = (unsigned long *)task->thread.sp; 34 stack = (unsigned long *)task->thread.sp;
36 } 35 }
37 36
38 bp = stack_frame(task, regs); 37 if (!bp)
38 bp = stack_frame(task, regs);
39
39 for (;;) { 40 for (;;) {
40 struct thread_info *context; 41 struct thread_info *context;
41 42
@@ -55,7 +56,7 @@ EXPORT_SYMBOL(dump_trace);
55 56
56void 57void
57show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs, 58show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs,
58 unsigned long *sp, char *log_lvl) 59 unsigned long *sp, unsigned long bp, char *log_lvl)
59{ 60{
60 unsigned long *stack; 61 unsigned long *stack;
61 int i; 62 int i;
@@ -77,7 +78,7 @@ show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs,
77 touch_nmi_watchdog(); 78 touch_nmi_watchdog();
78 } 79 }
79 printk(KERN_CONT "\n"); 80 printk(KERN_CONT "\n");
80 show_trace_log_lvl(task, regs, sp, log_lvl); 81 show_trace_log_lvl(task, regs, sp, bp, log_lvl);
81} 82}
82 83
83 84
@@ -102,7 +103,7 @@ void show_registers(struct pt_regs *regs)
102 u8 *ip; 103 u8 *ip;
103 104
104 printk(KERN_EMERG "Stack:\n"); 105 printk(KERN_EMERG "Stack:\n");
105 show_stack_log_lvl(NULL, regs, &regs->sp, KERN_EMERG); 106 show_stack_log_lvl(NULL, regs, &regs->sp, 0, KERN_EMERG);
106 107
107 printk(KERN_EMERG "Code: "); 108 printk(KERN_EMERG "Code: ");
108 109
diff --git a/arch/x86/kernel/dumpstack_64.c b/arch/x86/kernel/dumpstack_64.c
index a6b6fcf7f0ae..e71c98d3c0d2 100644
--- a/arch/x86/kernel/dumpstack_64.c
+++ b/arch/x86/kernel/dumpstack_64.c
@@ -139,8 +139,8 @@ fixup_bp_irq_link(unsigned long bp, unsigned long *stack,
139 * severe exception (double fault, nmi, stack fault, debug, mce) hardware stack 139 * severe exception (double fault, nmi, stack fault, debug, mce) hardware stack
140 */ 140 */
141 141
142void dump_trace(struct task_struct *task, 142void dump_trace(struct task_struct *task, struct pt_regs *regs,
143 struct pt_regs *regs, unsigned long *stack, 143 unsigned long *stack, unsigned long bp,
144 const struct stacktrace_ops *ops, void *data) 144 const struct stacktrace_ops *ops, void *data)
145{ 145{
146 const unsigned cpu = get_cpu(); 146 const unsigned cpu = get_cpu();
@@ -150,7 +150,6 @@ void dump_trace(struct task_struct *task,
150 struct thread_info *tinfo; 150 struct thread_info *tinfo;
151 int graph = 0; 151 int graph = 0;
152 unsigned long dummy; 152 unsigned long dummy;
153 unsigned long bp;
154 153
155 if (!task) 154 if (!task)
156 task = current; 155 task = current;
@@ -161,7 +160,8 @@ void dump_trace(struct task_struct *task,
161 stack = (unsigned long *)task->thread.sp; 160 stack = (unsigned long *)task->thread.sp;
162 } 161 }
163 162
164 bp = stack_frame(task, regs); 163 if (!bp)
164 bp = stack_frame(task, regs);
165 /* 165 /*
166 * Print function call entries in all stacks, starting at the 166 * Print function call entries in all stacks, starting at the
167 * current stack address. If the stacks consist of nested 167 * current stack address. If the stacks consist of nested
@@ -225,7 +225,7 @@ EXPORT_SYMBOL(dump_trace);
225 225
226void 226void
227show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs, 227show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs,
228 unsigned long *sp, char *log_lvl) 228 unsigned long *sp, unsigned long bp, char *log_lvl)
229{ 229{
230 unsigned long *irq_stack_end; 230 unsigned long *irq_stack_end;
231 unsigned long *irq_stack; 231 unsigned long *irq_stack;
@@ -269,7 +269,7 @@ show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs,
269 preempt_enable(); 269 preempt_enable();
270 270
271 printk(KERN_CONT "\n"); 271 printk(KERN_CONT "\n");
272 show_trace_log_lvl(task, regs, sp, log_lvl); 272 show_trace_log_lvl(task, regs, sp, bp, log_lvl);
273} 273}
274 274
275void show_registers(struct pt_regs *regs) 275void show_registers(struct pt_regs *regs)
@@ -298,7 +298,7 @@ void show_registers(struct pt_regs *regs)
298 298
299 printk(KERN_EMERG "Stack:\n"); 299 printk(KERN_EMERG "Stack:\n");
300 show_stack_log_lvl(NULL, regs, (unsigned long *)sp, 300 show_stack_log_lvl(NULL, regs, (unsigned long *)sp,
301 KERN_EMERG); 301 0, KERN_EMERG);
302 302
303 printk(KERN_EMERG "Code: "); 303 printk(KERN_EMERG "Code: ");
304 304
diff --git a/arch/x86/kernel/e820.c b/arch/x86/kernel/e820.c
index 294f26da0c0c..3e2ef8425316 100644
--- a/arch/x86/kernel/e820.c
+++ b/arch/x86/kernel/e820.c
@@ -11,6 +11,7 @@
11#include <linux/kernel.h> 11#include <linux/kernel.h>
12#include <linux/types.h> 12#include <linux/types.h>
13#include <linux/init.h> 13#include <linux/init.h>
14#include <linux/crash_dump.h>
14#include <linux/bootmem.h> 15#include <linux/bootmem.h>
15#include <linux/pfn.h> 16#include <linux/pfn.h>
16#include <linux/suspend.h> 17#include <linux/suspend.h>
@@ -667,21 +668,15 @@ __init void e820_setup_gap(void)
667 * boot_params.e820_map, others are passed via SETUP_E820_EXT node of 668 * boot_params.e820_map, others are passed via SETUP_E820_EXT node of
668 * linked list of struct setup_data, which is parsed here. 669 * linked list of struct setup_data, which is parsed here.
669 */ 670 */
670void __init parse_e820_ext(struct setup_data *sdata, unsigned long pa_data) 671void __init parse_e820_ext(struct setup_data *sdata)
671{ 672{
672 u32 map_len;
673 int entries; 673 int entries;
674 struct e820entry *extmap; 674 struct e820entry *extmap;
675 675
676 entries = sdata->len / sizeof(struct e820entry); 676 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); 677 extmap = (struct e820entry *)(sdata->data);
681 __append_e820_map(extmap, entries); 678 __append_e820_map(extmap, entries);
682 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map); 679 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"); 680 printk(KERN_INFO "extended physical RAM map:\n");
686 e820_print_map("extended"); 681 e820_print_map("extended");
687} 682}
@@ -847,15 +842,21 @@ static int __init parse_memopt(char *p)
847 if (!p) 842 if (!p)
848 return -EINVAL; 843 return -EINVAL;
849 844
850#ifdef CONFIG_X86_32
851 if (!strcmp(p, "nopentium")) { 845 if (!strcmp(p, "nopentium")) {
846#ifdef CONFIG_X86_32
852 setup_clear_cpu_cap(X86_FEATURE_PSE); 847 setup_clear_cpu_cap(X86_FEATURE_PSE);
853 return 0; 848 return 0;
854 } 849#else
850 printk(KERN_WARNING "mem=nopentium ignored! (only supported on x86_32)\n");
851 return -EINVAL;
855#endif 852#endif
853 }
856 854
857 userdef = 1; 855 userdef = 1;
858 mem_size = memparse(p, &p); 856 mem_size = memparse(p, &p);
857 /* don't remove all of memory when handling "mem={invalid}" param */
858 if (mem_size == 0)
859 return -EINVAL;
859 e820_remove_range(mem_size, ULLONG_MAX - mem_size, E820_RAM, 1); 860 e820_remove_range(mem_size, ULLONG_MAX - mem_size, E820_RAM, 1);
860 861
861 return 0; 862 return 0;
diff --git a/arch/x86/kernel/early-quirks.c b/arch/x86/kernel/early-quirks.c
index 76b8cd953dee..3755ef494390 100644
--- a/arch/x86/kernel/early-quirks.c
+++ b/arch/x86/kernel/early-quirks.c
@@ -143,15 +143,10 @@ static void __init ati_bugs(int num, int slot, int func)
143 143
144static u32 __init ati_sbx00_rev(int num, int slot, int func) 144static u32 __init ati_sbx00_rev(int num, int slot, int func)
145{ 145{
146 u32 old, d; 146 u32 d;
147 147
148 d = read_pci_config(num, slot, func, 0x70);
149 old = d;
150 d &= ~(1<<8);
151 write_pci_config(num, slot, func, 0x70, d);
152 d = read_pci_config(num, slot, func, 0x8); 148 d = read_pci_config(num, slot, func, 0x8);
153 d &= 0xff; 149 d &= 0xff;
154 write_pci_config(num, slot, func, 0x70, old);
155 150
156 return d; 151 return d;
157} 152}
@@ -160,11 +155,19 @@ static void __init ati_bugs_contd(int num, int slot, int func)
160{ 155{
161 u32 d, rev; 156 u32 d, rev;
162 157
163 if (acpi_use_timer_override) 158 rev = ati_sbx00_rev(num, slot, func);
159 if (rev >= 0x40)
160 acpi_fix_pin2_polarity = 1;
161
162 /*
163 * SB600: revisions 0x11, 0x12, 0x13, 0x14, ...
164 * SB700: revisions 0x39, 0x3a, ...
165 * SB800: revisions 0x40, 0x41, ...
166 */
167 if (rev >= 0x39)
164 return; 168 return;
165 169
166 rev = ati_sbx00_rev(num, slot, func); 170 if (acpi_use_timer_override)
167 if (rev > 0x13)
168 return; 171 return;
169 172
170 /* check for IRQ0 interrupt swap */ 173 /* check for IRQ0 interrupt swap */
diff --git a/arch/x86/kernel/entry_32.S b/arch/x86/kernel/entry_32.S
index c8b4efad7ebb..5c1a91974918 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
@@ -1409,11 +1411,10 @@ 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(async_page_fault)
1417#endif 1418#endif
1418 1419
1419/* 1420/*
diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S
index aed1ffbeb0c9..8a445a0c989e 100644
--- a/arch/x86/kernel/entry_64.S
+++ b/arch/x86/kernel/entry_64.S
@@ -18,7 +18,7 @@
18 * A note on terminology: 18 * A note on terminology:
19 * - top of stack: Architecture defined interrupt frame from SS to RIP 19 * - top of stack: Architecture defined interrupt frame from SS to RIP
20 * at the top of the kernel process stack. 20 * at the top of the kernel process stack.
21 * - partial stack frame: partially saved registers upto R11. 21 * - partial stack frame: partially saved registers up to R11.
22 * - full stack frame: Like partial stack frame, but all register saved. 22 * - full stack frame: Like partial stack frame, but all register saved.
23 * 23 *
24 * Some macro usage: 24 * Some macro usage:
@@ -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)
@@ -420,7 +422,7 @@ ENTRY(ret_from_fork)
420END(ret_from_fork) 422END(ret_from_fork)
421 423
422/* 424/*
423 * System call entry. Upto 6 arguments in registers are supported. 425 * System call entry. Up to 6 arguments in registers are supported.
424 * 426 *
425 * SYSCALL does not save anything on the stack and does not change the 427 * SYSCALL does not save anything on the stack and does not change the
426 * stack pointer. 428 * stack pointer.
@@ -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/head32.c b/arch/x86/kernel/head32.c
index 7f138b3c3c52..d6d6bb361931 100644
--- a/arch/x86/kernel/head32.c
+++ b/arch/x86/kernel/head32.c
@@ -34,15 +34,6 @@ void __init i386_start_kernel(void)
34{ 34{
35 memblock_init(); 35 memblock_init();
36 36
37#ifdef CONFIG_X86_TRAMPOLINE
38 /*
39 * But first pinch a few for the stack/trampoline stuff
40 * FIXME: Don't need the extra page at 4K, but need to fix
41 * trampoline before removing it. (see the GDT stuff)
42 */
43 memblock_x86_reserve_range(PAGE_SIZE, PAGE_SIZE + PAGE_SIZE, "EX TRAMPOLINE");
44#endif
45
46 memblock_x86_reserve_range(__pa_symbol(&_text), __pa_symbol(&__bss_stop), "TEXT DATA BSS"); 37 memblock_x86_reserve_range(__pa_symbol(&_text), __pa_symbol(&__bss_stop), "TEXT DATA BSS");
47 38
48#ifdef CONFIG_BLK_DEV_INITRD 39#ifdef CONFIG_BLK_DEV_INITRD
diff --git a/arch/x86/kernel/head64.c b/arch/x86/kernel/head64.c
index 2d2673c28aff..5655c2272adb 100644
--- a/arch/x86/kernel/head64.c
+++ b/arch/x86/kernel/head64.c
@@ -77,9 +77,6 @@ void __init x86_64_start_kernel(char * real_mode_data)
77 /* Make NULL pointers segfault */ 77 /* Make NULL pointers segfault */
78 zap_identity_mappings(); 78 zap_identity_mappings();
79 79
80 /* Cleanup the over mapped high alias */
81 cleanup_highmap();
82
83 max_pfn_mapped = KERNEL_IMAGE_SIZE >> PAGE_SHIFT; 80 max_pfn_mapped = KERNEL_IMAGE_SIZE >> PAGE_SHIFT;
84 81
85 for (i = 0; i < NUM_EXCEPTION_VECTORS; i++) { 82 for (i = 0; i < NUM_EXCEPTION_VECTORS; i++) {
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/head_64.S b/arch/x86/kernel/head_64.S
index 239046bd447f..e11e39478a49 100644
--- a/arch/x86/kernel/head_64.S
+++ b/arch/x86/kernel/head_64.S
@@ -136,10 +136,9 @@ ident_complete:
136 /* Fixup phys_base */ 136 /* Fixup phys_base */
137 addq %rbp, phys_base(%rip) 137 addq %rbp, phys_base(%rip)
138 138
139#ifdef CONFIG_X86_TRAMPOLINE 139 /* Fixup trampoline */
140 addq %rbp, trampoline_level4_pgt + 0(%rip) 140 addq %rbp, trampoline_level4_pgt + 0(%rip)
141 addq %rbp, trampoline_level4_pgt + (511*8)(%rip) 141 addq %rbp, trampoline_level4_pgt + (511*8)(%rip)
142#endif
143 142
144 /* Due to ENTRY(), sometimes the empty space gets filled with 143 /* Due to ENTRY(), sometimes the empty space gets filled with
145 * zeros. Better take a jmp than relying on empty space being 144 * zeros. Better take a jmp than relying on empty space being
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/i387.c b/arch/x86/kernel/i387.c
index e60c38cc0eed..12aff2537682 100644
--- a/arch/x86/kernel/i387.c
+++ b/arch/x86/kernel/i387.c
@@ -145,7 +145,7 @@ EXPORT_SYMBOL_GPL(fpu_finit);
145 * The _current_ task is using the FPU for the first time 145 * The _current_ task is using the FPU for the first time
146 * so initialize it and set the mxcsr to its default 146 * so initialize it and set the mxcsr to its default
147 * value at reset if we support XMM instructions and then 147 * value at reset if we support XMM instructions and then
148 * remeber the current task has used the FPU. 148 * remember the current task has used the FPU.
149 */ 149 */
150int init_fpu(struct task_struct *tsk) 150int init_fpu(struct task_struct *tsk)
151{ 151{
diff --git a/arch/x86/kernel/i8237.c b/arch/x86/kernel/i8237.c
index b42ca694dc68..8eeaa81de066 100644
--- a/arch/x86/kernel/i8237.c
+++ b/arch/x86/kernel/i8237.c
@@ -10,7 +10,7 @@
10 */ 10 */
11 11
12#include <linux/init.h> 12#include <linux/init.h>
13#include <linux/sysdev.h> 13#include <linux/syscore_ops.h>
14 14
15#include <asm/dma.h> 15#include <asm/dma.h>
16 16
@@ -21,7 +21,7 @@
21 * in asm/dma.h. 21 * in asm/dma.h.
22 */ 22 */
23 23
24static int i8237A_resume(struct sys_device *dev) 24static void i8237A_resume(void)
25{ 25{
26 unsigned long flags; 26 unsigned long flags;
27 int i; 27 int i;
@@ -41,31 +41,15 @@ static int i8237A_resume(struct sys_device *dev)
41 enable_dma(4); 41 enable_dma(4);
42 42
43 release_dma_lock(flags); 43 release_dma_lock(flags);
44
45 return 0;
46} 44}
47 45
48static int i8237A_suspend(struct sys_device *dev, pm_message_t state) 46static struct syscore_ops i8237_syscore_ops = {
49{
50 return 0;
51}
52
53static struct sysdev_class i8237_sysdev_class = {
54 .name = "i8237",
55 .suspend = i8237A_suspend,
56 .resume = i8237A_resume, 47 .resume = i8237A_resume,
57}; 48};
58 49
59static struct sys_device device_i8237A = { 50static int __init i8237A_init_ops(void)
60 .id = 0,
61 .cls = &i8237_sysdev_class,
62};
63
64static int __init i8237A_init_sysfs(void)
65{ 51{
66 int error = sysdev_class_register(&i8237_sysdev_class); 52 register_syscore_ops(&i8237_syscore_ops);
67 if (!error) 53 return 0;
68 error = sysdev_register(&device_i8237A);
69 return error;
70} 54}
71device_initcall(i8237A_init_sysfs); 55device_initcall(i8237A_init_ops);
diff --git a/arch/x86/kernel/i8253.c b/arch/x86/kernel/i8253.c
index 212fe6590aab..577e90cadaeb 100644
--- a/arch/x86/kernel/i8253.c
+++ b/arch/x86/kernel/i8253.c
@@ -117,79 +117,6 @@ void __init setup_pit_timer(void)
117} 117}
118 118
119#ifndef CONFIG_X86_64 119#ifndef CONFIG_X86_64
120/*
121 * Since the PIT overflows every tick, its not very useful
122 * to just read by itself. So use jiffies to emulate a free
123 * running counter:
124 */
125static cycle_t pit_read(struct clocksource *cs)
126{
127 static int old_count;
128 static u32 old_jifs;
129 unsigned long flags;
130 int count;
131 u32 jifs;
132
133 raw_spin_lock_irqsave(&i8253_lock, flags);
134 /*
135 * Although our caller may have the read side of xtime_lock,
136 * this is now a seqlock, and we are cheating in this routine
137 * by having side effects on state that we cannot undo if
138 * there is a collision on the seqlock and our caller has to
139 * retry. (Namely, old_jifs and old_count.) So we must treat
140 * jiffies as volatile despite the lock. We read jiffies
141 * before latching the timer count to guarantee that although
142 * the jiffies value might be older than the count (that is,
143 * the counter may underflow between the last point where
144 * jiffies was incremented and the point where we latch the
145 * count), it cannot be newer.
146 */
147 jifs = jiffies;
148 outb_pit(0x00, PIT_MODE); /* latch the count ASAP */
149 count = inb_pit(PIT_CH0); /* read the latched count */
150 count |= inb_pit(PIT_CH0) << 8;
151
152 /* VIA686a test code... reset the latch if count > max + 1 */
153 if (count > LATCH) {
154 outb_pit(0x34, PIT_MODE);
155 outb_pit(LATCH & 0xff, PIT_CH0);
156 outb_pit(LATCH >> 8, PIT_CH0);
157 count = LATCH - 1;
158 }
159
160 /*
161 * It's possible for count to appear to go the wrong way for a
162 * couple of reasons:
163 *
164 * 1. The timer counter underflows, but we haven't handled the
165 * resulting interrupt and incremented jiffies yet.
166 * 2. Hardware problem with the timer, not giving us continuous time,
167 * the counter does small "jumps" upwards on some Pentium systems,
168 * (see c't 95/10 page 335 for Neptun bug.)
169 *
170 * Previous attempts to handle these cases intelligently were
171 * buggy, so we just do the simple thing now.
172 */
173 if (count > old_count && jifs == old_jifs)
174 count = old_count;
175
176 old_count = count;
177 old_jifs = jifs;
178
179 raw_spin_unlock_irqrestore(&i8253_lock, flags);
180
181 count = (LATCH - 1) - count;
182
183 return (cycle_t)(jifs * LATCH) + count;
184}
185
186static struct clocksource pit_cs = {
187 .name = "pit",
188 .rating = 110,
189 .read = pit_read,
190 .mask = CLOCKSOURCE_MASK(32),
191};
192
193static int __init init_pit_clocksource(void) 120static int __init init_pit_clocksource(void)
194{ 121{
195 /* 122 /*
@@ -203,8 +130,7 @@ static int __init init_pit_clocksource(void)
203 pit_ce.mode != CLOCK_EVT_MODE_PERIODIC) 130 pit_ce.mode != CLOCK_EVT_MODE_PERIODIC)
204 return 0; 131 return 0;
205 132
206 return clocksource_register_hz(&pit_cs, CLOCK_TICK_RATE); 133 return clocksource_i8253_init();
207} 134}
208arch_initcall(init_pit_clocksource); 135arch_initcall(init_pit_clocksource);
209
210#endif /* !CONFIG_X86_64 */ 136#endif /* !CONFIG_X86_64 */
diff --git a/arch/x86/kernel/i8259.c b/arch/x86/kernel/i8259.c
index 20757cb2efa3..65b8f5c2eebf 100644
--- a/arch/x86/kernel/i8259.c
+++ b/arch/x86/kernel/i8259.c
@@ -8,7 +8,7 @@
8#include <linux/random.h> 8#include <linux/random.h>
9#include <linux/init.h> 9#include <linux/init.h>
10#include <linux/kernel_stat.h> 10#include <linux/kernel_stat.h>
11#include <linux/sysdev.h> 11#include <linux/syscore_ops.h>
12#include <linux/bitops.h> 12#include <linux/bitops.h>
13#include <linux/acpi.h> 13#include <linux/acpi.h>
14#include <linux/io.h> 14#include <linux/io.h>
@@ -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}
@@ -245,20 +245,19 @@ static void save_ELCR(char *trigger)
245 trigger[1] = inb(0x4d1) & 0xDE; 245 trigger[1] = inb(0x4d1) & 0xDE;
246} 246}
247 247
248static int i8259A_resume(struct sys_device *dev) 248static void i8259A_resume(void)
249{ 249{
250 init_8259A(i8259A_auto_eoi); 250 init_8259A(i8259A_auto_eoi);
251 restore_ELCR(irq_trigger); 251 restore_ELCR(irq_trigger);
252 return 0;
253} 252}
254 253
255static int i8259A_suspend(struct sys_device *dev, pm_message_t state) 254static int i8259A_suspend(void)
256{ 255{
257 save_ELCR(irq_trigger); 256 save_ELCR(irq_trigger);
258 return 0; 257 return 0;
259} 258}
260 259
261static int i8259A_shutdown(struct sys_device *dev) 260static void i8259A_shutdown(void)
262{ 261{
263 /* Put the i8259A into a quiescent state that 262 /* Put the i8259A into a quiescent state that
264 * the kernel initialization code can get it 263 * the kernel initialization code can get it
@@ -266,21 +265,14 @@ static int i8259A_shutdown(struct sys_device *dev)
266 */ 265 */
267 outb(0xff, PIC_MASTER_IMR); /* mask all of 8259A-1 */ 266 outb(0xff, PIC_MASTER_IMR); /* mask all of 8259A-1 */
268 outb(0xff, PIC_SLAVE_IMR); /* mask all of 8259A-1 */ 267 outb(0xff, PIC_SLAVE_IMR); /* mask all of 8259A-1 */
269 return 0;
270} 268}
271 269
272static struct sysdev_class i8259_sysdev_class = { 270static struct syscore_ops i8259_syscore_ops = {
273 .name = "i8259",
274 .suspend = i8259A_suspend, 271 .suspend = i8259A_suspend,
275 .resume = i8259A_resume, 272 .resume = i8259A_resume,
276 .shutdown = i8259A_shutdown, 273 .shutdown = i8259A_shutdown,
277}; 274};
278 275
279static struct sys_device device_i8259A = {
280 .id = 0,
281 .cls = &i8259_sysdev_class,
282};
283
284static void mask_8259A(void) 276static void mask_8259A(void)
285{ 277{
286 unsigned long flags; 278 unsigned long flags;
@@ -399,17 +391,12 @@ struct legacy_pic default_legacy_pic = {
399 391
400struct legacy_pic *legacy_pic = &default_legacy_pic; 392struct legacy_pic *legacy_pic = &default_legacy_pic;
401 393
402static int __init i8259A_init_sysfs(void) 394static int __init i8259A_init_ops(void)
403{ 395{
404 int error; 396 if (legacy_pic == &default_legacy_pic)
405 397 register_syscore_ops(&i8259_syscore_ops);
406 if (legacy_pic != &default_legacy_pic)
407 return 0;
408 398
409 error = sysdev_class_register(&i8259_sysdev_class); 399 return 0;
410 if (!error)
411 error = sysdev_register(&device_i8259A);
412 return error;
413} 400}
414 401
415device_initcall(i8259A_init_sysfs); 402device_initcall(i8259A_init_ops);
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..1cb0b9fc78dc 100644
--- a/arch/x86/kernel/irq.c
+++ b/arch/x86/kernel/irq.c
@@ -8,6 +8,7 @@
8#include <linux/seq_file.h> 8#include <linux/seq_file.h>
9#include <linux/smp.h> 9#include <linux/smp.h>
10#include <linux/ftrace.h> 10#include <linux/ftrace.h>
11#include <linux/delay.h>
11 12
12#include <asm/apic.h> 13#include <asm/apic.h>
13#include <asm/io_apic.h> 14#include <asm/io_apic.h>
@@ -44,9 +45,9 @@ void ack_bad_irq(unsigned int irq)
44 45
45#define irq_stats(x) (&per_cpu(irq_stat, x)) 46#define irq_stats(x) (&per_cpu(irq_stat, x))
46/* 47/*
47 * /proc/interrupts printing: 48 * /proc/interrupts printing for arch specific interrupts
48 */ 49 */
49static int show_other_interrupts(struct seq_file *p, int prec) 50int arch_show_interrupts(struct seq_file *p, int prec)
50{ 51{
51 int j; 52 int j;
52 53
@@ -122,59 +123,6 @@ static int show_other_interrupts(struct seq_file *p, int prec)
122 return 0; 123 return 0;
123} 124}
124 125
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/* 126/*
179 * /proc/stat helpers 127 * /proc/stat helpers
180 */ 128 */
@@ -276,15 +224,6 @@ void smp_x86_platform_ipi(struct pt_regs *regs)
276 224
277EXPORT_SYMBOL_GPL(vector_used_by_percpu_irq); 225EXPORT_SYMBOL_GPL(vector_used_by_percpu_irq);
278 226
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 227#ifdef CONFIG_HOTPLUG_CPU
289/* A cpu has been removed from cpu_online_mask. Reset irq affinities. */ 228/* A cpu has been removed from cpu_online_mask. Reset irq affinities. */
290void fixup_irqs(void) 229void fixup_irqs(void)
@@ -293,6 +232,7 @@ void fixup_irqs(void)
293 static int warned; 232 static int warned;
294 struct irq_desc *desc; 233 struct irq_desc *desc;
295 struct irq_data *data; 234 struct irq_data *data;
235 struct irq_chip *chip;
296 236
297 for_each_irq_desc(irq, desc) { 237 for_each_irq_desc(irq, desc) {
298 int break_affinity = 0; 238 int break_affinity = 0;
@@ -307,10 +247,10 @@ void fixup_irqs(void)
307 /* interrupt's are disabled at this point */ 247 /* interrupt's are disabled at this point */
308 raw_spin_lock(&desc->lock); 248 raw_spin_lock(&desc->lock);
309 249
310 data = &desc->irq_data; 250 data = irq_desc_get_irq_data(desc);
311 affinity = data->affinity; 251 affinity = data->affinity;
312 if (!irq_has_action(irq) || 252 if (!irq_has_action(irq) ||
313 cpumask_equal(affinity, cpu_online_mask)) { 253 cpumask_subset(affinity, cpu_online_mask)) {
314 raw_spin_unlock(&desc->lock); 254 raw_spin_unlock(&desc->lock);
315 continue; 255 continue;
316 } 256 }
@@ -327,16 +267,17 @@ void fixup_irqs(void)
327 affinity = cpu_all_mask; 267 affinity = cpu_all_mask;
328 } 268 }
329 269
330 if (!(desc->status & IRQ_MOVE_PCNTXT) && data->chip->irq_mask) 270 chip = irq_data_get_irq_chip(data);
331 data->chip->irq_mask(data); 271 if (!irqd_can_move_in_process_context(data) && chip->irq_mask)
272 chip->irq_mask(data);
332 273
333 if (data->chip->irq_set_affinity) 274 if (chip->irq_set_affinity)
334 data->chip->irq_set_affinity(data, affinity, true); 275 chip->irq_set_affinity(data, affinity, true);
335 else if (!(warned++)) 276 else if (!(warned++))
336 set_affinity = 0; 277 set_affinity = 0;
337 278
338 if (!(desc->status & IRQ_MOVE_PCNTXT) && data->chip->irq_unmask) 279 if (!irqd_can_move_in_process_context(data) && chip->irq_unmask)
339 data->chip->irq_unmask(data); 280 chip->irq_unmask(data);
340 281
341 raw_spin_unlock(&desc->lock); 282 raw_spin_unlock(&desc->lock);
342 283
@@ -368,10 +309,11 @@ void fixup_irqs(void)
368 irq = __this_cpu_read(vector_irq[vector]); 309 irq = __this_cpu_read(vector_irq[vector]);
369 310
370 desc = irq_to_desc(irq); 311 desc = irq_to_desc(irq);
371 data = &desc->irq_data; 312 data = irq_desc_get_irq_data(desc);
313 chip = irq_data_get_irq_chip(data);
372 raw_spin_lock(&desc->lock); 314 raw_spin_lock(&desc->lock);
373 if (data->chip->irq_retrigger) 315 if (chip->irq_retrigger)
374 data->chip->irq_retrigger(data); 316 chip->irq_retrigger(data);
375 raw_spin_unlock(&desc->lock); 317 raw_spin_unlock(&desc->lock);
376 } 318 }
377 } 319 }
diff --git a/arch/x86/kernel/irq_32.c b/arch/x86/kernel/irq_32.c
index 9974d21048fd..72090705a656 100644
--- a/arch/x86/kernel/irq_32.c
+++ b/arch/x86/kernel/irq_32.c
@@ -172,7 +172,7 @@ asmlinkage void do_softirq(void)
172 172
173 call_on_stack(__do_softirq, isp); 173 call_on_stack(__do_softirq, isp);
174 /* 174 /*
175 * Shouldnt happen, we returned above if in_interrupt(): 175 * Shouldn't happen, we returned above if in_interrupt():
176 */ 176 */
177 WARN_ON_ONCE(softirq_count()); 177 WARN_ON_ONCE(softirq_count());
178 } 178 }
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..5f9ecff328b5 100644
--- a/arch/x86/kernel/kgdb.c
+++ b/arch/x86/kernel/kgdb.c
@@ -121,8 +121,8 @@ char *dbg_get_reg(int regno, void *mem, struct pt_regs *regs)
121 memcpy(mem, (void *)regs + dbg_reg_def[regno].offset, 121 memcpy(mem, (void *)regs + dbg_reg_def[regno].offset,
122 dbg_reg_def[regno].size); 122 dbg_reg_def[regno].size);
123 123
124 switch (regno) {
125#ifdef CONFIG_X86_32 124#ifdef CONFIG_X86_32
125 switch (regno) {
126 case GDB_SS: 126 case GDB_SS:
127 if (!user_mode_vm(regs)) 127 if (!user_mode_vm(regs))
128 *(unsigned long *)mem = __KERNEL_DS; 128 *(unsigned long *)mem = __KERNEL_DS;
@@ -135,8 +135,8 @@ char *dbg_get_reg(int regno, void *mem, struct pt_regs *regs)
135 case GDB_FS: 135 case GDB_FS:
136 *(unsigned long *)mem = 0xFFFF; 136 *(unsigned long *)mem = 0xFFFF;
137 break; 137 break;
138#endif
139 } 138 }
139#endif
140 return dbg_reg_def[regno].name; 140 return dbg_reg_def[regno].name;
141} 141}
142 142
@@ -278,7 +278,7 @@ static int hw_break_release_slot(int breakno)
278 pevent = per_cpu_ptr(breakinfo[breakno].pev, cpu); 278 pevent = per_cpu_ptr(breakinfo[breakno].pev, cpu);
279 if (dbg_release_bp_slot(*pevent)) 279 if (dbg_release_bp_slot(*pevent))
280 /* 280 /*
281 * The debugger is responisble for handing the retry on 281 * The debugger is responsible for handing the retry on
282 * remove failure. 282 * remove failure.
283 */ 283 */
284 return -1; 284 return -1;
@@ -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/kvm.c b/arch/x86/kernel/kvm.c
index 8dc44662394b..33c07b0b122e 100644
--- a/arch/x86/kernel/kvm.c
+++ b/arch/x86/kernel/kvm.c
@@ -493,7 +493,7 @@ static void __init kvm_smp_prepare_boot_cpu(void)
493 native_smp_prepare_boot_cpu(); 493 native_smp_prepare_boot_cpu();
494} 494}
495 495
496static void kvm_guest_cpu_online(void *dummy) 496static void __cpuinit kvm_guest_cpu_online(void *dummy)
497{ 497{
498 kvm_guest_cpu_init(); 498 kvm_guest_cpu_init();
499} 499}
diff --git a/arch/x86/kernel/mca_32.c b/arch/x86/kernel/mca_32.c
index 63eaf6596233..177183cbb6ae 100644
--- a/arch/x86/kernel/mca_32.c
+++ b/arch/x86/kernel/mca_32.c
@@ -259,7 +259,7 @@ static int __init mca_init(void)
259 /* 259 /*
260 * WARNING: Be careful when making changes here. Putting an adapter 260 * WARNING: Be careful when making changes here. Putting an adapter
261 * and the motherboard simultaneously into setup mode may result in 261 * and the motherboard simultaneously into setup mode may result in
262 * damage to chips (according to The Indispensible PC Hardware Book 262 * damage to chips (according to The Indispensable PC Hardware Book
263 * by Hans-Peter Messmer). Also, we disable system interrupts (so 263 * by Hans-Peter Messmer). Also, we disable system interrupts (so
264 * that we are not disturbed in the middle of this). 264 * that we are not disturbed in the middle of this).
265 */ 265 */
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..f9242800bc84 100644
--- a/arch/x86/kernel/microcode_core.c
+++ b/arch/x86/kernel/microcode_core.c
@@ -82,6 +82,7 @@
82#include <linux/cpu.h> 82#include <linux/cpu.h>
83#include <linux/fs.h> 83#include <linux/fs.h>
84#include <linux/mm.h> 84#include <linux/mm.h>
85#include <linux/syscore_ops.h>
85 86
86#include <asm/microcode.h> 87#include <asm/microcode.h>
87#include <asm/processor.h> 88#include <asm/processor.h>
@@ -417,8 +418,10 @@ static int mc_sysdev_add(struct sys_device *sys_dev)
417 if (err) 418 if (err)
418 return err; 419 return err;
419 420
420 if (microcode_init_cpu(cpu) == UCODE_ERROR) 421 if (microcode_init_cpu(cpu) == UCODE_ERROR) {
421 err = -EINVAL; 422 sysfs_remove_group(&sys_dev->kobj, &mc_attr_group);
423 return -EINVAL;
424 }
422 425
423 return err; 426 return err;
424} 427}
@@ -436,33 +439,25 @@ static int mc_sysdev_remove(struct sys_device *sys_dev)
436 return 0; 439 return 0;
437} 440}
438 441
439static int mc_sysdev_resume(struct sys_device *dev) 442static struct sysdev_driver mc_sysdev_driver = {
443 .add = mc_sysdev_add,
444 .remove = mc_sysdev_remove,
445};
446
447/**
448 * mc_bp_resume - Update boot CPU microcode during resume.
449 */
450static void mc_bp_resume(void)
440{ 451{
441 int cpu = dev->id; 452 int cpu = smp_processor_id();
442 struct ucode_cpu_info *uci = ucode_cpu_info + cpu; 453 struct ucode_cpu_info *uci = ucode_cpu_info + cpu;
443 454
444 if (!cpu_online(cpu))
445 return 0;
446
447 /*
448 * All non-bootup cpus are still disabled,
449 * so only CPU 0 will apply ucode here.
450 *
451 * Moreover, there can be no concurrent
452 * updates from any other places at this point.
453 */
454 WARN_ON(cpu != 0);
455
456 if (uci->valid && uci->mc) 455 if (uci->valid && uci->mc)
457 microcode_ops->apply_microcode(cpu); 456 microcode_ops->apply_microcode(cpu);
458
459 return 0;
460} 457}
461 458
462static struct sysdev_driver mc_sysdev_driver = { 459static struct syscore_ops mc_syscore_ops = {
463 .add = mc_sysdev_add, 460 .resume = mc_bp_resume,
464 .remove = mc_sysdev_remove,
465 .resume = mc_sysdev_resume,
466}; 461};
467 462
468static __cpuinit int 463static __cpuinit int
@@ -540,6 +535,7 @@ static int __init microcode_init(void)
540 if (error) 535 if (error)
541 return error; 536 return error;
542 537
538 register_syscore_ops(&mc_syscore_ops);
543 register_hotcpu_notifier(&mc_cpu_notifier); 539 register_hotcpu_notifier(&mc_cpu_notifier);
544 540
545 pr_info("Microcode Update Driver: v" MICROCODE_VERSION 541 pr_info("Microcode Update Driver: v" MICROCODE_VERSION
@@ -554,6 +550,7 @@ static void __exit microcode_exit(void)
554 microcode_dev_exit(); 550 microcode_dev_exit();
555 551
556 unregister_hotcpu_notifier(&mc_cpu_notifier); 552 unregister_hotcpu_notifier(&mc_cpu_notifier);
553 unregister_syscore_ops(&mc_syscore_ops);
557 554
558 get_online_cpus(); 555 get_online_cpus();
559 mutex_lock(&microcode_mutex); 556 mutex_lock(&microcode_mutex);
diff --git a/arch/x86/kernel/mpparse.c b/arch/x86/kernel/mpparse.c
index 01b0f6d06451..5a532ce646bf 100644
--- a/arch/x86/kernel/mpparse.c
+++ b/arch/x86/kernel/mpparse.c
@@ -714,10 +714,6 @@ static void __init check_irq_src(struct mpc_intsrc *m, int *nr_m_spare)
714 *nr_m_spare += 1; 714 *nr_m_spare += 1;
715 } 715 }
716} 716}
717#else /* CONFIG_X86_IO_APIC */
718static
719inline void __init check_irq_src(struct mpc_intsrc *m, int *nr_m_spare) {}
720#endif /* CONFIG_X86_IO_APIC */
721 717
722static int 718static int
723check_slot(unsigned long mpc_new_phys, unsigned long mpc_new_length, int count) 719check_slot(unsigned long mpc_new_phys, unsigned long mpc_new_length, int count)
@@ -731,6 +727,10 @@ check_slot(unsigned long mpc_new_phys, unsigned long mpc_new_length, int count)
731 727
732 return ret; 728 return ret;
733} 729}
730#else /* CONFIG_X86_IO_APIC */
731static
732inline void __init check_irq_src(struct mpc_intsrc *m, int *nr_m_spare) {}
733#endif /* CONFIG_X86_IO_APIC */
734 734
735static int __init replace_intsrc_all(struct mpc_table *mpc, 735static int __init replace_intsrc_all(struct mpc_table *mpc,
736 unsigned long mpc_new_phys, 736 unsigned long mpc_new_phys,
@@ -883,7 +883,7 @@ static int __init update_mp_table(void)
883 883
884 if (!mpc_new_phys) { 884 if (!mpc_new_phys) {
885 unsigned char old, new; 885 unsigned char old, new;
886 /* check if we can change the postion */ 886 /* check if we can change the position */
887 mpc->checksum = 0; 887 mpc->checksum = 0;
888 old = mpf_checksum((unsigned char *)mpc, mpc->length); 888 old = mpf_checksum((unsigned char *)mpc, mpc->length);
889 mpc->checksum = 0xff; 889 mpc->checksum = 0xff;
@@ -892,7 +892,7 @@ static int __init update_mp_table(void)
892 printk(KERN_INFO "mpc is readonly, please try alloc_mptable instead\n"); 892 printk(KERN_INFO "mpc is readonly, please try alloc_mptable instead\n");
893 return 0; 893 return 0;
894 } 894 }
895 printk(KERN_INFO "use in-positon replacing\n"); 895 printk(KERN_INFO "use in-position replacing\n");
896 } else { 896 } else {
897 mpf->physptr = mpc_new_phys; 897 mpf->physptr = mpc_new_phys;
898 mpc_new = phys_to_virt(mpc_new_phys); 898 mpc_new = phys_to_virt(mpc_new_phys);
diff --git a/arch/x86/kernel/pci-calgary_64.c b/arch/x86/kernel/pci-calgary_64.c
index f56a117cef68..e8c33a302006 100644
--- a/arch/x86/kernel/pci-calgary_64.c
+++ b/arch/x86/kernel/pci-calgary_64.c
@@ -1279,7 +1279,7 @@ static int __init calgary_bus_has_devices(int bus, unsigned short pci_dev)
1279 1279
1280 if (pci_dev == PCI_DEVICE_ID_IBM_CALIOC2) { 1280 if (pci_dev == PCI_DEVICE_ID_IBM_CALIOC2) {
1281 /* 1281 /*
1282 * FIXME: properly scan for devices accross the 1282 * FIXME: properly scan for devices across the
1283 * PCI-to-PCI bridge on every CalIOC2 port. 1283 * PCI-to-PCI bridge on every CalIOC2 port.
1284 */ 1284 */
1285 return 1; 1285 return 1;
@@ -1295,7 +1295,7 @@ static int __init calgary_bus_has_devices(int bus, unsigned short pci_dev)
1295 1295
1296/* 1296/*
1297 * calgary_init_bitmap_from_tce_table(): 1297 * calgary_init_bitmap_from_tce_table():
1298 * Funtion for kdump case. In the second/kdump kernel initialize 1298 * Function for kdump case. In the second/kdump kernel initialize
1299 * the bitmap based on the tce table entries obtained from first kernel 1299 * the bitmap based on the tce table entries obtained from first kernel
1300 */ 1300 */
1301static void calgary_init_bitmap_from_tce_table(struct iommu_table *tbl) 1301static void calgary_init_bitmap_from_tce_table(struct iommu_table *tbl)
diff --git a/arch/x86/kernel/pci-gart_64.c b/arch/x86/kernel/pci-gart_64.c
index c01ffa5b9b87..b117efd24f71 100644
--- a/arch/x86/kernel/pci-gart_64.c
+++ b/arch/x86/kernel/pci-gart_64.c
@@ -27,7 +27,7 @@
27#include <linux/kdebug.h> 27#include <linux/kdebug.h>
28#include <linux/scatterlist.h> 28#include <linux/scatterlist.h>
29#include <linux/iommu-helper.h> 29#include <linux/iommu-helper.h>
30#include <linux/sysdev.h> 30#include <linux/syscore_ops.h>
31#include <linux/io.h> 31#include <linux/io.h>
32#include <linux/gfp.h> 32#include <linux/gfp.h>
33#include <asm/atomic.h> 33#include <asm/atomic.h>
@@ -81,6 +81,9 @@ static u32 gart_unmapped_entry;
81#define AGPEXTERN 81#define AGPEXTERN
82#endif 82#endif
83 83
84/* GART can only remap to physical addresses < 1TB */
85#define GART_MAX_PHYS_ADDR (1ULL << 40)
86
84/* backdoor interface to AGP driver */ 87/* backdoor interface to AGP driver */
85AGPEXTERN int agp_memory_reserved; 88AGPEXTERN int agp_memory_reserved;
86AGPEXTERN __u32 *agp_gatt_table; 89AGPEXTERN __u32 *agp_gatt_table;
@@ -212,9 +215,13 @@ static dma_addr_t dma_map_area(struct device *dev, dma_addr_t phys_mem,
212 size_t size, int dir, unsigned long align_mask) 215 size_t size, int dir, unsigned long align_mask)
213{ 216{
214 unsigned long npages = iommu_num_pages(phys_mem, size, PAGE_SIZE); 217 unsigned long npages = iommu_num_pages(phys_mem, size, PAGE_SIZE);
215 unsigned long iommu_page = alloc_iommu(dev, npages, align_mask); 218 unsigned long iommu_page;
216 int i; 219 int i;
217 220
221 if (unlikely(phys_mem + size > GART_MAX_PHYS_ADDR))
222 return bad_dma_addr;
223
224 iommu_page = alloc_iommu(dev, npages, align_mask);
218 if (iommu_page == -1) { 225 if (iommu_page == -1) {
219 if (!nonforced_iommu(dev, phys_mem, size)) 226 if (!nonforced_iommu(dev, phys_mem, size))
220 return phys_mem; 227 return phys_mem;
@@ -589,7 +596,7 @@ void set_up_gart_resume(u32 aper_order, u32 aper_alloc)
589 aperture_alloc = aper_alloc; 596 aperture_alloc = aper_alloc;
590} 597}
591 598
592static void gart_fixup_northbridges(struct sys_device *dev) 599static void gart_fixup_northbridges(void)
593{ 600{
594 int i; 601 int i;
595 602
@@ -613,33 +620,20 @@ static void gart_fixup_northbridges(struct sys_device *dev)
613 } 620 }
614} 621}
615 622
616static int gart_resume(struct sys_device *dev) 623static void gart_resume(void)
617{ 624{
618 pr_info("PCI-DMA: Resuming GART IOMMU\n"); 625 pr_info("PCI-DMA: Resuming GART IOMMU\n");
619 626
620 gart_fixup_northbridges(dev); 627 gart_fixup_northbridges();
621 628
622 enable_gart_translations(); 629 enable_gart_translations();
623
624 return 0;
625} 630}
626 631
627static int gart_suspend(struct sys_device *dev, pm_message_t state) 632static struct syscore_ops gart_syscore_ops = {
628{
629 return 0;
630}
631
632static struct sysdev_class gart_sysdev_class = {
633 .name = "gart",
634 .suspend = gart_suspend,
635 .resume = gart_resume, 633 .resume = gart_resume,
636 634
637}; 635};
638 636
639static struct sys_device device_gart = {
640 .cls = &gart_sysdev_class,
641};
642
643/* 637/*
644 * Private Northbridge GATT initialization in case we cannot use the 638 * Private Northbridge GATT initialization in case we cannot use the
645 * AGP driver for some reason. 639 * AGP driver for some reason.
@@ -650,7 +644,7 @@ static __init int init_amd_gatt(struct agp_kern_info *info)
650 unsigned aper_base, new_aper_base; 644 unsigned aper_base, new_aper_base;
651 struct pci_dev *dev; 645 struct pci_dev *dev;
652 void *gatt; 646 void *gatt;
653 int i, error; 647 int i;
654 648
655 pr_info("PCI-DMA: Disabling AGP.\n"); 649 pr_info("PCI-DMA: Disabling AGP.\n");
656 650
@@ -685,12 +679,7 @@ static __init int init_amd_gatt(struct agp_kern_info *info)
685 679
686 agp_gatt_table = gatt; 680 agp_gatt_table = gatt;
687 681
688 error = sysdev_class_register(&gart_sysdev_class); 682 register_syscore_ops(&gart_syscore_ops);
689 if (!error)
690 error = sysdev_register(&device_gart);
691 if (error)
692 panic("Could not register gart_sysdev -- "
693 "would corrupt data on next suspend");
694 683
695 flush_gart(); 684 flush_gart();
696 685
diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c
index ff4554198981..d46cbe46b7ab 100644
--- a/arch/x86/kernel/process.c
+++ b/arch/x86/kernel/process.c
@@ -87,7 +87,7 @@ void exit_thread(void)
87void show_regs(struct pt_regs *regs) 87void show_regs(struct pt_regs *regs)
88{ 88{
89 show_registers(regs); 89 show_registers(regs);
90 show_trace(NULL, regs, (unsigned long *)kernel_stack_pointer(regs)); 90 show_trace(NULL, regs, (unsigned long *)kernel_stack_pointer(regs), 0);
91} 91}
92 92
93void show_regs_common(void) 93void show_regs_common(void)
@@ -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/process_64.c b/arch/x86/kernel/process_64.c
index bd387e8f73b4..6c9dd922ac0d 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -501,6 +501,10 @@ void set_personality_64bit(void)
501 /* Make sure to be in 64bit mode */ 501 /* Make sure to be in 64bit mode */
502 clear_thread_flag(TIF_IA32); 502 clear_thread_flag(TIF_IA32);
503 503
504 /* Ensure the corresponding mm is not marked. */
505 if (current->mm)
506 current->mm->context.ia32_compat = 0;
507
504 /* TBD: overwrites user setup. Should have two bits. 508 /* TBD: overwrites user setup. Should have two bits.
505 But 64bit processes have always behaved this way, 509 But 64bit processes have always behaved this way,
506 so it's not too bad. The main problem is just that 510 so it's not too bad. The main problem is just that
@@ -516,6 +520,10 @@ void set_personality_ia32(void)
516 set_thread_flag(TIF_IA32); 520 set_thread_flag(TIF_IA32);
517 current->personality |= force_personality32; 521 current->personality |= force_personality32;
518 522
523 /* Mark the associated mm as containing 32-bit tasks. */
524 if (current->mm)
525 current->mm->context.ia32_compat = 1;
526
519 /* Prepare the first "return" to user space */ 527 /* Prepare the first "return" to user space */
520 current_thread_info()->status |= TS_COMPAT; 528 current_thread_info()->status |= TS_COMPAT;
521} 529}
diff --git a/arch/x86/kernel/ptrace.c b/arch/x86/kernel/ptrace.c
index 45892dc4b72a..f65e5b521dbd 100644
--- a/arch/x86/kernel/ptrace.c
+++ b/arch/x86/kernel/ptrace.c
@@ -608,6 +608,9 @@ static int ptrace_write_dr7(struct task_struct *tsk, unsigned long data)
608 unsigned len, type; 608 unsigned len, type;
609 struct perf_event *bp; 609 struct perf_event *bp;
610 610
611 if (ptrace_get_breakpoints(tsk) < 0)
612 return -ESRCH;
613
611 data &= ~DR_CONTROL_RESERVED; 614 data &= ~DR_CONTROL_RESERVED;
612 old_dr7 = ptrace_get_dr7(thread->ptrace_bps); 615 old_dr7 = ptrace_get_dr7(thread->ptrace_bps);
613restore: 616restore:
@@ -655,6 +658,9 @@ restore:
655 } 658 }
656 goto restore; 659 goto restore;
657 } 660 }
661
662 ptrace_put_breakpoints(tsk);
663
658 return ((orig_ret < 0) ? orig_ret : rc); 664 return ((orig_ret < 0) ? orig_ret : rc);
659} 665}
660 666
@@ -668,10 +674,17 @@ static unsigned long ptrace_get_debugreg(struct task_struct *tsk, int n)
668 674
669 if (n < HBP_NUM) { 675 if (n < HBP_NUM) {
670 struct perf_event *bp; 676 struct perf_event *bp;
677
678 if (ptrace_get_breakpoints(tsk) < 0)
679 return -ESRCH;
680
671 bp = thread->ptrace_bps[n]; 681 bp = thread->ptrace_bps[n];
672 if (!bp) 682 if (!bp)
673 return 0; 683 val = 0;
674 val = bp->hw.info.address; 684 else
685 val = bp->hw.info.address;
686
687 ptrace_put_breakpoints(tsk);
675 } else if (n == 6) { 688 } else if (n == 6) {
676 val = thread->debugreg6; 689 val = thread->debugreg6;
677 } else if (n == 7) { 690 } else if (n == 7) {
@@ -686,6 +699,10 @@ static int ptrace_set_breakpoint_addr(struct task_struct *tsk, int nr,
686 struct perf_event *bp; 699 struct perf_event *bp;
687 struct thread_struct *t = &tsk->thread; 700 struct thread_struct *t = &tsk->thread;
688 struct perf_event_attr attr; 701 struct perf_event_attr attr;
702 int err = 0;
703
704 if (ptrace_get_breakpoints(tsk) < 0)
705 return -ESRCH;
689 706
690 if (!t->ptrace_bps[nr]) { 707 if (!t->ptrace_bps[nr]) {
691 ptrace_breakpoint_init(&attr); 708 ptrace_breakpoint_init(&attr);
@@ -709,24 +726,23 @@ static int ptrace_set_breakpoint_addr(struct task_struct *tsk, int nr,
709 * writing for the user. And anyway this is the previous 726 * writing for the user. And anyway this is the previous
710 * behaviour. 727 * behaviour.
711 */ 728 */
712 if (IS_ERR(bp)) 729 if (IS_ERR(bp)) {
713 return PTR_ERR(bp); 730 err = PTR_ERR(bp);
731 goto put;
732 }
714 733
715 t->ptrace_bps[nr] = bp; 734 t->ptrace_bps[nr] = bp;
716 } else { 735 } else {
717 int err;
718
719 bp = t->ptrace_bps[nr]; 736 bp = t->ptrace_bps[nr];
720 737
721 attr = bp->attr; 738 attr = bp->attr;
722 attr.bp_addr = addr; 739 attr.bp_addr = addr;
723 err = modify_user_hw_breakpoint(bp, &attr); 740 err = modify_user_hw_breakpoint(bp, &attr);
724 if (err)
725 return err;
726 } 741 }
727 742
728 743put:
729 return 0; 744 ptrace_put_breakpoints(tsk);
745 return err;
730} 746}
731 747
732/* 748/*
diff --git a/arch/x86/kernel/reboot.c b/arch/x86/kernel/reboot.c
index fc7aae1e2bc7..08c44b08bf5b 100644
--- a/arch/x86/kernel/reboot.c
+++ b/arch/x86/kernel/reboot.c
@@ -6,6 +6,7 @@
6#include <linux/dmi.h> 6#include <linux/dmi.h>
7#include <linux/sched.h> 7#include <linux/sched.h>
8#include <linux/tboot.h> 8#include <linux/tboot.h>
9#include <linux/delay.h>
9#include <acpi/reboot.h> 10#include <acpi/reboot.h>
10#include <asm/io.h> 11#include <asm/io.h>
11#include <asm/apic.h> 12#include <asm/apic.h>
@@ -285,6 +286,14 @@ static struct dmi_system_id __initdata reboot_dmi_table[] = {
285 DMI_MATCH(DMI_BOARD_NAME, "P4S800"), 286 DMI_MATCH(DMI_BOARD_NAME, "P4S800"),
286 }, 287 },
287 }, 288 },
289 { /* Handle problems with rebooting on VersaLogic Menlow boards */
290 .callback = set_bios_reboot,
291 .ident = "VersaLogic Menlow based board",
292 .matches = {
293 DMI_MATCH(DMI_BOARD_VENDOR, "VersaLogic Corporation"),
294 DMI_MATCH(DMI_BOARD_NAME, "VersaLogic Menlow board"),
295 },
296 },
288 { } 297 { }
289}; 298};
290 299
@@ -295,68 +304,16 @@ static int __init reboot_init(void)
295} 304}
296core_initcall(reboot_init); 305core_initcall(reboot_init);
297 306
298/* The following code and data reboots the machine by switching to real 307extern const unsigned char machine_real_restart_asm[];
299 mode and jumping to the BIOS reset entry point, as if the CPU has 308extern const u64 machine_real_restart_gdt[3];
300 really been reset. The previous version asked the keyboard
301 controller to pulse the CPU reset line, which is more thorough, but
302 doesn't work with at least one type of 486 motherboard. It is easy
303 to stop this code working; hence the copious comments. */
304static const unsigned long long
305real_mode_gdt_entries [3] =
306{
307 0x0000000000000000ULL, /* Null descriptor */
308 0x00009b000000ffffULL, /* 16-bit real-mode 64k code at 0x00000000 */
309 0x000093000100ffffULL /* 16-bit real-mode 64k data at 0x00000100 */
310};
311 309
312static const struct desc_ptr 310void machine_real_restart(unsigned int type)
313real_mode_gdt = { sizeof (real_mode_gdt_entries) - 1, (long)real_mode_gdt_entries },
314real_mode_idt = { 0x3ff, 0 };
315
316/* This is 16-bit protected mode code to disable paging and the cache,
317 switch to real mode and jump to the BIOS reset code.
318
319 The instruction that switches to real mode by writing to CR0 must be
320 followed immediately by a far jump instruction, which set CS to a
321 valid value for real mode, and flushes the prefetch queue to avoid
322 running instructions that have already been decoded in protected
323 mode.
324
325 Clears all the flags except ET, especially PG (paging), PE
326 (protected-mode enable) and TS (task switch for coprocessor state
327 save). Flushes the TLB after paging has been disabled. Sets CD and
328 NW, to disable the cache on a 486, and invalidates the cache. This
329 is more like the state of a 486 after reset. I don't know if
330 something else should be done for other chips.
331
332 More could be done here to set up the registers as if a CPU reset had
333 occurred; hopefully real BIOSs don't assume much. */
334static const unsigned char real_mode_switch [] =
335{
336 0x66, 0x0f, 0x20, 0xc0, /* movl %cr0,%eax */
337 0x66, 0x83, 0xe0, 0x11, /* andl $0x00000011,%eax */
338 0x66, 0x0d, 0x00, 0x00, 0x00, 0x60, /* orl $0x60000000,%eax */
339 0x66, 0x0f, 0x22, 0xc0, /* movl %eax,%cr0 */
340 0x66, 0x0f, 0x22, 0xd8, /* movl %eax,%cr3 */
341 0x66, 0x0f, 0x20, 0xc3, /* movl %cr0,%ebx */
342 0x66, 0x81, 0xe3, 0x00, 0x00, 0x00, 0x60, /* andl $0x60000000,%ebx */
343 0x74, 0x02, /* jz f */
344 0x0f, 0x09, /* wbinvd */
345 0x24, 0x10, /* f: andb $0x10,al */
346 0x66, 0x0f, 0x22, 0xc0 /* movl %eax,%cr0 */
347};
348static const unsigned char jump_to_bios [] =
349{ 311{
350 0xea, 0x00, 0x00, 0xff, 0xff /* ljmp $0xffff,$0x0000 */ 312 void *restart_va;
351}; 313 unsigned long restart_pa;
314 void (*restart_lowmem)(unsigned int);
315 u64 *lowmem_gdt;
352 316
353/*
354 * Switch to real mode and then execute the code
355 * specified by the code and length parameters.
356 * We assume that length will aways be less that 100!
357 */
358void machine_real_restart(const unsigned char *code, int length)
359{
360 local_irq_disable(); 317 local_irq_disable();
361 318
362 /* Write zero to CMOS register number 0x0f, which the BIOS POST 319 /* Write zero to CMOS register number 0x0f, which the BIOS POST
@@ -384,41 +341,23 @@ void machine_real_restart(const unsigned char *code, int length)
384 too. */ 341 too. */
385 *((unsigned short *)0x472) = reboot_mode; 342 *((unsigned short *)0x472) = reboot_mode;
386 343
387 /* For the switch to real mode, copy some code to low memory. It has 344 /* Patch the GDT in the low memory trampoline */
388 to be in the first 64k because it is running in 16-bit mode, and it 345 lowmem_gdt = TRAMPOLINE_SYM(machine_real_restart_gdt);
389 has to have the same physical and virtual address, because it turns 346
390 off paging. Copy it near the end of the first page, out of the way 347 restart_va = TRAMPOLINE_SYM(machine_real_restart_asm);
391 of BIOS variables. */ 348 restart_pa = virt_to_phys(restart_va);
392 memcpy((void *)(0x1000 - sizeof(real_mode_switch) - 100), 349 restart_lowmem = (void (*)(unsigned int))restart_pa;
393 real_mode_switch, sizeof (real_mode_switch)); 350
394 memcpy((void *)(0x1000 - 100), code, length); 351 /* GDT[0]: GDT self-pointer */
395 352 lowmem_gdt[0] =
396 /* Set up the IDT for real mode. */ 353 (u64)(sizeof(machine_real_restart_gdt) - 1) +
397 load_idt(&real_mode_idt); 354 ((u64)virt_to_phys(lowmem_gdt) << 16);
398 355 /* GDT[1]: 64K real mode code segment */
399 /* Set up a GDT from which we can load segment descriptors for real 356 lowmem_gdt[1] =
400 mode. The GDT is not used in real mode; it is just needed here to 357 GDT_ENTRY(0x009b, restart_pa, 0xffff);
401 prepare the descriptors. */ 358
402 load_gdt(&real_mode_gdt); 359 /* Jump to the identity-mapped low memory code */
403 360 restart_lowmem(type);
404 /* Load the data segment registers, and thus the descriptors ready for
405 real mode. The base address of each segment is 0x100, 16 times the
406 selector value being loaded here. This is so that the segment
407 registers don't have to be reloaded after switching to real mode:
408 the values are consistent for real mode operation already. */
409 __asm__ __volatile__ ("movl $0x0010,%%eax\n"
410 "\tmovl %%eax,%%ds\n"
411 "\tmovl %%eax,%%es\n"
412 "\tmovl %%eax,%%fs\n"
413 "\tmovl %%eax,%%gs\n"
414 "\tmovl %%eax,%%ss" : : : "eax");
415
416 /* Jump to the 16-bit code that we copied earlier. It disables paging
417 and the cache, switches to real mode, and jumps to the BIOS reset
418 entry point. */
419 __asm__ __volatile__ ("ljmp $0x0008,%0"
420 :
421 : "i" ((void *)(0x1000 - sizeof (real_mode_switch) - 100)));
422} 361}
423#ifdef CONFIG_APM_MODULE 362#ifdef CONFIG_APM_MODULE
424EXPORT_SYMBOL(machine_real_restart); 363EXPORT_SYMBOL(machine_real_restart);
@@ -573,7 +512,7 @@ static void native_machine_emergency_restart(void)
573 512
574#ifdef CONFIG_X86_32 513#ifdef CONFIG_X86_32
575 case BOOT_BIOS: 514 case BOOT_BIOS:
576 machine_real_restart(jump_to_bios, sizeof(jump_to_bios)); 515 machine_real_restart(MRR_BIOS);
577 516
578 reboot_type = BOOT_KBD; 517 reboot_type = BOOT_KBD;
579 break; 518 break;
diff --git a/arch/x86/kernel/reboot_32.S b/arch/x86/kernel/reboot_32.S
new file mode 100644
index 000000000000..1d5c46df0d78
--- /dev/null
+++ b/arch/x86/kernel/reboot_32.S
@@ -0,0 +1,135 @@
1#include <linux/linkage.h>
2#include <linux/init.h>
3#include <asm/segment.h>
4#include <asm/page_types.h>
5
6/*
7 * The following code and data reboots the machine by switching to real
8 * mode and jumping to the BIOS reset entry point, as if the CPU has
9 * really been reset. The previous version asked the keyboard
10 * controller to pulse the CPU reset line, which is more thorough, but
11 * doesn't work with at least one type of 486 motherboard. It is easy
12 * to stop this code working; hence the copious comments.
13 *
14 * This code is called with the restart type (0 = BIOS, 1 = APM) in %eax.
15 */
16 .section ".x86_trampoline","a"
17 .balign 16
18 .code32
19ENTRY(machine_real_restart_asm)
20r_base = .
21 /* Get our own relocated address */
22 call 1f
231: popl %ebx
24 subl $(1b - r_base), %ebx
25
26 /* Compute the equivalent real-mode segment */
27 movl %ebx, %ecx
28 shrl $4, %ecx
29
30 /* Patch post-real-mode segment jump */
31 movw (dispatch_table - r_base)(%ebx,%eax,2),%ax
32 movw %ax, (101f - r_base)(%ebx)
33 movw %cx, (102f - r_base)(%ebx)
34
35 /* Set up the IDT for real mode. */
36 lidtl (machine_real_restart_idt - r_base)(%ebx)
37
38 /*
39 * Set up a GDT from which we can load segment descriptors for real
40 * mode. The GDT is not used in real mode; it is just needed here to
41 * prepare the descriptors.
42 */
43 lgdtl (machine_real_restart_gdt - r_base)(%ebx)
44
45 /*
46 * Load the data segment registers with 16-bit compatible values
47 */
48 movl $16, %ecx
49 movl %ecx, %ds
50 movl %ecx, %es
51 movl %ecx, %fs
52 movl %ecx, %gs
53 movl %ecx, %ss
54 ljmpl $8, $1f - r_base
55
56/*
57 * This is 16-bit protected mode code to disable paging and the cache,
58 * switch to real mode and jump to the BIOS reset code.
59 *
60 * The instruction that switches to real mode by writing to CR0 must be
61 * followed immediately by a far jump instruction, which set CS to a
62 * valid value for real mode, and flushes the prefetch queue to avoid
63 * running instructions that have already been decoded in protected
64 * mode.
65 *
66 * Clears all the flags except ET, especially PG (paging), PE
67 * (protected-mode enable) and TS (task switch for coprocessor state
68 * save). Flushes the TLB after paging has been disabled. Sets CD and
69 * NW, to disable the cache on a 486, and invalidates the cache. This
70 * is more like the state of a 486 after reset. I don't know if
71 * something else should be done for other chips.
72 *
73 * More could be done here to set up the registers as if a CPU reset had
74 * occurred; hopefully real BIOSs don't assume much. This is not the
75 * actual BIOS entry point, anyway (that is at 0xfffffff0).
76 *
77 * Most of this work is probably excessive, but it is what is tested.
78 */
79 .code16
801:
81 xorl %ecx, %ecx
82 movl %cr0, %eax
83 andl $0x00000011, %eax
84 orl $0x60000000, %eax
85 movl %eax, %cr0
86 movl %ecx, %cr3
87 movl %cr0, %edx
88 andl $0x60000000, %edx /* If no cache bits -> no wbinvd */
89 jz 2f
90 wbinvd
912:
92 andb $0x10, %al
93 movl %eax, %cr0
94 .byte 0xea /* ljmpw */
95101: .word 0 /* Offset */
96102: .word 0 /* Segment */
97
98bios:
99 ljmpw $0xf000, $0xfff0
100
101apm:
102 movw $0x1000, %ax
103 movw %ax, %ss
104 movw $0xf000, %sp
105 movw $0x5307, %ax
106 movw $0x0001, %bx
107 movw $0x0003, %cx
108 int $0x15
109
110END(machine_real_restart_asm)
111
112 .balign 16
113 /* These must match <asm/reboot.h */
114dispatch_table:
115 .word bios - r_base
116 .word apm - r_base
117END(dispatch_table)
118
119 .balign 16
120machine_real_restart_idt:
121 .word 0xffff /* Length - real mode default value */
122 .long 0 /* Base - real mode default value */
123END(machine_real_restart_idt)
124
125 .balign 16
126ENTRY(machine_real_restart_gdt)
127 .quad 0 /* Self-pointer, filled in by PM code */
128 .quad 0 /* 16-bit code segment, filled in by PM code */
129 /*
130 * 16-bit data segment with the selector value 16 = 0x10 and
131 * base value 0x100; since this is consistent with real mode
132 * semantics we don't have to reload the segments once CR0.PE = 0.
133 */
134 .quad GDT_ENTRY(0x0093, 0x100, 0xffff)
135END(machine_real_restart_gdt)
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..4be9b398470e 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.
@@ -297,6 +298,9 @@ static void __init init_gbpages(void)
297static inline void init_gbpages(void) 298static inline void init_gbpages(void)
298{ 299{
299} 300}
301static void __init cleanup_highmap(void)
302{
303}
300#endif 304#endif
301 305
302static void __init reserve_brk(void) 306static void __init reserve_brk(void)
@@ -429,16 +433,30 @@ static void __init parse_setup_data(void)
429 return; 433 return;
430 pa_data = boot_params.hdr.setup_data; 434 pa_data = boot_params.hdr.setup_data;
431 while (pa_data) { 435 while (pa_data) {
432 data = early_memremap(pa_data, PAGE_SIZE); 436 u32 data_len, map_len;
437
438 map_len = max(PAGE_SIZE - (pa_data & ~PAGE_MASK),
439 (u64)sizeof(struct setup_data));
440 data = early_memremap(pa_data, map_len);
441 data_len = data->len + sizeof(struct setup_data);
442 if (data_len > map_len) {
443 early_iounmap(data, map_len);
444 data = early_memremap(pa_data, data_len);
445 map_len = data_len;
446 }
447
433 switch (data->type) { 448 switch (data->type) {
434 case SETUP_E820_EXT: 449 case SETUP_E820_EXT:
435 parse_e820_ext(data, pa_data); 450 parse_e820_ext(data);
451 break;
452 case SETUP_DTB:
453 add_dtb(pa_data);
436 break; 454 break;
437 default: 455 default:
438 break; 456 break;
439 } 457 }
440 pa_data = data->next; 458 pa_data = data->next;
441 early_iounmap(data, PAGE_SIZE); 459 early_iounmap(data, map_len);
442 } 460 }
443} 461}
444 462
@@ -601,28 +619,6 @@ void __init reserve_standard_io_resources(void)
601 619
602} 620}
603 621
604/*
605 * Note: elfcorehdr_addr is not just limited to vmcore. It is also used by
606 * is_kdump_kernel() to determine if we are booting after a panic. Hence
607 * ifdef it under CONFIG_CRASH_DUMP and not CONFIG_PROC_VMCORE.
608 */
609
610#ifdef CONFIG_CRASH_DUMP
611/* elfcorehdr= specifies the location of elf core header
612 * stored by the crashed kernel. This option will be passed
613 * by kexec loader to the capture kernel.
614 */
615static int __init setup_elfcorehdr(char *arg)
616{
617 char *end;
618 if (!arg)
619 return -EINVAL;
620 elfcorehdr_addr = memparse(arg, &end);
621 return end > arg ? 0 : -EINVAL;
622}
623early_param("elfcorehdr", setup_elfcorehdr);
624#endif
625
626static __init void reserve_ibft_region(void) 622static __init void reserve_ibft_region(void)
627{ 623{
628 unsigned long addr, size = 0; 624 unsigned long addr, size = 0;
@@ -680,15 +676,6 @@ static int __init parse_reservelow(char *p)
680 676
681early_param("reservelow", parse_reservelow); 677early_param("reservelow", parse_reservelow);
682 678
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/* 679/*
693 * Determine if we were loaded by an EFI loader. If so, then we have also been 680 * 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 681 * passed the efi memmap, systab, etc., so we should use these data structures
@@ -704,8 +691,6 @@ static u64 __init get_max_mapped(void)
704 691
705void __init setup_arch(char **cmdline_p) 692void __init setup_arch(char **cmdline_p)
706{ 693{
707 int acpi = 0;
708 int amd = 0;
709 unsigned long flags; 694 unsigned long flags;
710 695
711#ifdef CONFIG_X86_32 696#ifdef CONFIG_X86_32
@@ -922,6 +907,8 @@ void __init setup_arch(char **cmdline_p)
922 */ 907 */
923 reserve_brk(); 908 reserve_brk();
924 909
910 cleanup_highmap();
911
925 memblock.current_limit = get_max_mapped(); 912 memblock.current_limit = get_max_mapped();
926 memblock_x86_fill(); 913 memblock_x86_fill();
927 914
@@ -935,15 +922,8 @@ void __init setup_arch(char **cmdline_p)
935 printk(KERN_DEBUG "initial memory mapped : 0 - %08lx\n", 922 printk(KERN_DEBUG "initial memory mapped : 0 - %08lx\n",
936 max_pfn_mapped<<PAGE_SHIFT); 923 max_pfn_mapped<<PAGE_SHIFT);
937 924
938 reserve_trampoline_memory(); 925 setup_trampolines();
939 926
940#ifdef CONFIG_ACPI_SLEEP
941 /*
942 * Reserve low memory region for sleep support.
943 * even before init_memory_mapping
944 */
945 acpi_reserve_wakeup_memory();
946#endif
947 init_gbpages(); 927 init_gbpages();
948 928
949 /* max_pfn_mapped is updated here */ 929 /* max_pfn_mapped is updated here */
@@ -984,19 +964,7 @@ void __init setup_arch(char **cmdline_p)
984 964
985 early_acpi_boot_init(); 965 early_acpi_boot_init();
986 966
987#ifdef CONFIG_ACPI_NUMA 967 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(); 968 memblock_find_dma_reserve();
1001 dma32_reserve_bootmem(); 969 dma32_reserve_bootmem();
1002 970
@@ -1008,6 +976,11 @@ void __init setup_arch(char **cmdline_p)
1008 paging_init(); 976 paging_init();
1009 x86_init.paging.pagetable_setup_done(swapper_pg_dir); 977 x86_init.paging.pagetable_setup_done(swapper_pg_dir);
1010 978
979 if (boot_cpu_data.cpuid_level >= 0) {
980 /* A CPU has %cr4 if and only if it has CPUID */
981 mmu_cr4_features = read_cr4();
982 }
983
1011#ifdef CONFIG_X86_32 984#ifdef CONFIG_X86_32
1012 /* sync back kernel address range */ 985 /* sync back kernel address range */
1013 clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY, 986 clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY,
@@ -1029,8 +1002,8 @@ void __init setup_arch(char **cmdline_p)
1029 * Read APIC and some other early information from ACPI tables. 1002 * Read APIC and some other early information from ACPI tables.
1030 */ 1003 */
1031 acpi_boot_init(); 1004 acpi_boot_init();
1032
1033 sfi_init(); 1005 sfi_init();
1006 x86_dtb_init();
1034 1007
1035 /* 1008 /*
1036 * get boot-time SMP configuration: 1009 * get boot-time SMP configuration:
@@ -1040,9 +1013,7 @@ void __init setup_arch(char **cmdline_p)
1040 1013
1041 prefill_possible_map(); 1014 prefill_possible_map();
1042 1015
1043#ifdef CONFIG_X86_64
1044 init_cpu_to_node(); 1016 init_cpu_to_node();
1045#endif
1046 1017
1047 init_apic_mappings(); 1018 init_apic_mappings();
1048 ioapic_and_gsi_init(); 1019 ioapic_and_gsi_init();
@@ -1066,6 +1037,8 @@ void __init setup_arch(char **cmdline_p)
1066#endif 1037#endif
1067 x86_init.oem.banner(); 1038 x86_init.oem.banner();
1068 1039
1040 x86_init.timers.wallclock_init();
1041
1069 mcheck_init(); 1042 mcheck_init();
1070 1043
1071 local_irq_save(flags); 1044 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 08776a953487..c2871d3c71b6 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
@@ -414,6 +336,7 @@ void __cpuinit set_cpu_sibling_map(int cpu)
414 336
415 if (cpu_has(c, X86_FEATURE_TOPOEXT)) { 337 if (cpu_has(c, X86_FEATURE_TOPOEXT)) {
416 if (c->phys_proc_id == o->phys_proc_id && 338 if (c->phys_proc_id == o->phys_proc_id &&
339 per_cpu(cpu_llc_id, cpu) == per_cpu(cpu_llc_id, i) &&
417 c->compute_unit_id == o->compute_unit_id) 340 c->compute_unit_id == o->compute_unit_id)
418 link_thread_siblings(cpu, i); 341 link_thread_siblings(cpu, i);
419 } else if (c->phys_proc_id == o->phys_proc_id && 342 } else if (c->phys_proc_id == o->phys_proc_id &&
@@ -425,7 +348,7 @@ void __cpuinit set_cpu_sibling_map(int cpu)
425 cpumask_set_cpu(cpu, cpu_sibling_mask(cpu)); 348 cpumask_set_cpu(cpu, cpu_sibling_mask(cpu));
426 } 349 }
427 350
428 cpumask_set_cpu(cpu, c->llc_shared_map); 351 cpumask_set_cpu(cpu, cpu_llc_shared_mask(cpu));
429 352
430 if (__this_cpu_read(cpu_info.x86_max_cores) == 1) { 353 if (__this_cpu_read(cpu_info.x86_max_cores) == 1) {
431 cpumask_copy(cpu_core_mask(cpu), cpu_sibling_mask(cpu)); 354 cpumask_copy(cpu_core_mask(cpu), cpu_sibling_mask(cpu));
@@ -436,8 +359,8 @@ void __cpuinit set_cpu_sibling_map(int cpu)
436 for_each_cpu(i, cpu_sibling_setup_mask) { 359 for_each_cpu(i, cpu_sibling_setup_mask) {
437 if (per_cpu(cpu_llc_id, cpu) != BAD_APICID && 360 if (per_cpu(cpu_llc_id, cpu) != BAD_APICID &&
438 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)) {
439 cpumask_set_cpu(i, c->llc_shared_map); 362 cpumask_set_cpu(i, cpu_llc_shared_mask(cpu));
440 cpumask_set_cpu(cpu, cpu_data(i).llc_shared_map); 363 cpumask_set_cpu(cpu, cpu_llc_shared_mask(i));
441 } 364 }
442 if (c->phys_proc_id == cpu_data(i).phys_proc_id) { 365 if (c->phys_proc_id == cpu_data(i).phys_proc_id) {
443 cpumask_set_cpu(i, cpu_core_mask(cpu)); 366 cpumask_set_cpu(i, cpu_core_mask(cpu));
@@ -476,7 +399,7 @@ const struct cpumask *cpu_coregroup_mask(int cpu)
476 !(cpu_has(c, X86_FEATURE_AMD_DCM))) 399 !(cpu_has(c, X86_FEATURE_AMD_DCM)))
477 return cpu_core_mask(cpu); 400 return cpu_core_mask(cpu);
478 else 401 else
479 return c->llc_shared_map; 402 return cpu_llc_shared_mask(cpu);
480} 403}
481 404
482static void impress_friends(void) 405static void impress_friends(void)
@@ -788,7 +711,7 @@ do_rest:
788 stack_start = c_idle.idle->thread.sp; 711 stack_start = c_idle.idle->thread.sp;
789 712
790 /* start_ip had better be page-aligned! */ 713 /* start_ip had better be page-aligned! */
791 start_ip = setup_trampoline(); 714 start_ip = trampoline_address();
792 715
793 /* So we see what's up */ 716 /* So we see what's up */
794 announce_cpu(cpu, apicid); 717 announce_cpu(cpu, apicid);
@@ -798,6 +721,8 @@ do_rest:
798 * the targeted processor. 721 * the targeted processor.
799 */ 722 */
800 723
724 printk(KERN_DEBUG "smpboot cpu %d: start_ip = %lx\n", cpu, start_ip);
725
801 atomic_set(&init_deasserted, 0); 726 atomic_set(&init_deasserted, 0);
802 727
803 if (get_uv_system_type() != UV_NON_UNIQUE_APIC) { 728 if (get_uv_system_type() != UV_NON_UNIQUE_APIC) {
@@ -851,8 +776,8 @@ do_rest:
851 pr_debug("CPU%d: has booted.\n", cpu); 776 pr_debug("CPU%d: has booted.\n", cpu);
852 else { 777 else {
853 boot_error = 1; 778 boot_error = 1;
854 if (*((volatile unsigned char *)trampoline_base) 779 if (*(volatile u32 *)TRAMPOLINE_SYM(trampoline_status)
855 == 0xA5) 780 == 0xA5A5A5A5)
856 /* trampoline started but...? */ 781 /* trampoline started but...? */
857 pr_err("CPU%d: Stuck ??\n", cpu); 782 pr_err("CPU%d: Stuck ??\n", cpu);
858 else 783 else
@@ -878,7 +803,7 @@ do_rest:
878 } 803 }
879 804
880 /* mark "stuck" area as not stuck */ 805 /* mark "stuck" area as not stuck */
881 *((volatile unsigned long *)trampoline_base) = 0; 806 *(volatile u32 *)TRAMPOLINE_SYM(trampoline_status) = 0;
882 807
883 if (get_uv_system_type() != UV_NON_UNIQUE_APIC) { 808 if (get_uv_system_type() != UV_NON_UNIQUE_APIC) {
884 /* 809 /*
@@ -945,6 +870,14 @@ int __cpuinit native_cpu_up(unsigned int cpu)
945 return 0; 870 return 0;
946} 871}
947 872
873/**
874 * arch_disable_smp_support() - disables SMP support for x86 at runtime
875 */
876void arch_disable_smp_support(void)
877{
878 disable_ioapic_support();
879}
880
948/* 881/*
949 * Fall back to non SMP mode after errors. 882 * Fall back to non SMP mode after errors.
950 * 883 *
@@ -960,7 +893,6 @@ static __init void disable_smp(void)
960 physid_set_mask_of_physid(boot_cpu_physical_apicid, &phys_cpu_present_map); 893 physid_set_mask_of_physid(boot_cpu_physical_apicid, &phys_cpu_present_map);
961 else 894 else
962 physid_set_mask_of_physid(0, &phys_cpu_present_map); 895 physid_set_mask_of_physid(0, &phys_cpu_present_map);
963 map_cpu_to_logical_apicid();
964 cpumask_set_cpu(0, cpu_sibling_mask(0)); 896 cpumask_set_cpu(0, cpu_sibling_mask(0));
965 cpumask_set_cpu(0, cpu_core_mask(0)); 897 cpumask_set_cpu(0, cpu_core_mask(0));
966} 898}
@@ -1045,7 +977,7 @@ static int __init smp_sanity_check(unsigned max_cpus)
1045 "(tell your hw vendor)\n"); 977 "(tell your hw vendor)\n");
1046 } 978 }
1047 smpboot_clear_io_apic(); 979 smpboot_clear_io_apic();
1048 arch_disable_smp_support(); 980 disable_ioapic_support();
1049 return -1; 981 return -1;
1050 } 982 }
1051 983
@@ -1089,21 +1021,19 @@ void __init native_smp_prepare_cpus(unsigned int max_cpus)
1089 1021
1090 preempt_disable(); 1022 preempt_disable();
1091 smp_cpu_index_default(); 1023 smp_cpu_index_default();
1092 memcpy(__this_cpu_ptr(&cpu_info), &boot_cpu_data, sizeof(cpu_info)); 1024
1093 cpumask_copy(cpu_callin_mask, cpumask_of(0));
1094 mb();
1095 /* 1025 /*
1096 * Setup boot CPU information 1026 * Setup boot CPU information
1097 */ 1027 */
1098 smp_store_cpu_info(0); /* Final full version of the data */ 1028 smp_store_cpu_info(0); /* Final full version of the data */
1099#ifdef CONFIG_X86_32 1029 cpumask_copy(cpu_callin_mask, cpumask_of(0));
1100 boot_cpu_logical_apicid = logical_smp_processor_id(); 1030 mb();
1101#endif 1031
1102 current_thread_info()->cpu = 0; /* needed? */ 1032 current_thread_info()->cpu = 0; /* needed? */
1103 for_each_possible_cpu(i) { 1033 for_each_possible_cpu(i) {
1104 zalloc_cpumask_var(&per_cpu(cpu_sibling_map, i), GFP_KERNEL); 1034 zalloc_cpumask_var(&per_cpu(cpu_sibling_map, i), GFP_KERNEL);
1105 zalloc_cpumask_var(&per_cpu(cpu_core_map, i), GFP_KERNEL); 1035 zalloc_cpumask_var(&per_cpu(cpu_core_map, i), GFP_KERNEL);
1106 zalloc_cpumask_var(&cpu_data(i).llc_shared_map, GFP_KERNEL); 1036 zalloc_cpumask_var(&per_cpu(cpu_llc_shared_map, i), GFP_KERNEL);
1107 } 1037 }
1108 set_cpu_sibling_map(0); 1038 set_cpu_sibling_map(0);
1109 1039
@@ -1139,8 +1069,6 @@ void __init native_smp_prepare_cpus(unsigned int max_cpus)
1139 1069
1140 bsp_end_local_APIC_setup(); 1070 bsp_end_local_APIC_setup();
1141 1071
1142 map_cpu_to_logical_apicid();
1143
1144 if (apic->setup_portio_remap) 1072 if (apic->setup_portio_remap)
1145 apic->setup_portio_remap(); 1073 apic->setup_portio_remap();
1146 1074
diff --git a/arch/x86/kernel/stacktrace.c b/arch/x86/kernel/stacktrace.c
index 938c8e10a19a..6515733a289d 100644
--- a/arch/x86/kernel/stacktrace.c
+++ b/arch/x86/kernel/stacktrace.c
@@ -73,7 +73,7 @@ static const struct stacktrace_ops save_stack_ops_nosched = {
73 */ 73 */
74void save_stack_trace(struct stack_trace *trace) 74void save_stack_trace(struct stack_trace *trace)
75{ 75{
76 dump_trace(current, NULL, NULL, &save_stack_ops, trace); 76 dump_trace(current, NULL, NULL, 0, &save_stack_ops, trace);
77 if (trace->nr_entries < trace->max_entries) 77 if (trace->nr_entries < trace->max_entries)
78 trace->entries[trace->nr_entries++] = ULONG_MAX; 78 trace->entries[trace->nr_entries++] = ULONG_MAX;
79} 79}
@@ -81,14 +81,14 @@ EXPORT_SYMBOL_GPL(save_stack_trace);
81 81
82void save_stack_trace_regs(struct stack_trace *trace, struct pt_regs *regs) 82void save_stack_trace_regs(struct stack_trace *trace, struct pt_regs *regs)
83{ 83{
84 dump_trace(current, regs, NULL, &save_stack_ops, trace); 84 dump_trace(current, regs, NULL, 0, &save_stack_ops, trace);
85 if (trace->nr_entries < trace->max_entries) 85 if (trace->nr_entries < trace->max_entries)
86 trace->entries[trace->nr_entries++] = ULONG_MAX; 86 trace->entries[trace->nr_entries++] = ULONG_MAX;
87} 87}
88 88
89void save_stack_trace_tsk(struct task_struct *tsk, struct stack_trace *trace) 89void save_stack_trace_tsk(struct task_struct *tsk, struct stack_trace *trace)
90{ 90{
91 dump_trace(tsk, NULL, NULL, &save_stack_ops_nosched, trace); 91 dump_trace(tsk, NULL, NULL, 0, &save_stack_ops_nosched, trace);
92 if (trace->nr_entries < trace->max_entries) 92 if (trace->nr_entries < trace->max_entries)
93 trace->entries[trace->nr_entries++] = ULONG_MAX; 93 trace->entries[trace->nr_entries++] = ULONG_MAX;
94} 94}
diff --git a/arch/x86/kernel/step.c b/arch/x86/kernel/step.c
index 58de45ee08b6..7977f0cfe339 100644
--- a/arch/x86/kernel/step.c
+++ b/arch/x86/kernel/step.c
@@ -166,7 +166,7 @@ static void enable_step(struct task_struct *child, bool block)
166 * Make sure block stepping (BTF) is not enabled unless it should be. 166 * Make sure block stepping (BTF) is not enabled unless it should be.
167 * Note that we don't try to worry about any is_setting_trap_flag() 167 * Note that we don't try to worry about any is_setting_trap_flag()
168 * instructions after the first when using block stepping. 168 * instructions after the first when using block stepping.
169 * So noone should try to use debugger block stepping in a program 169 * So no one should try to use debugger block stepping in a program
170 * that uses user-mode single stepping itself. 170 * that uses user-mode single stepping itself.
171 */ 171 */
172 if (enable_single_step(child) && block) { 172 if (enable_single_step(child) && block) {
diff --git a/arch/x86/kernel/syscall_table_32.S b/arch/x86/kernel/syscall_table_32.S
index b35786dc9b8f..abce34d5c79d 100644
--- a/arch/x86/kernel/syscall_table_32.S
+++ b/arch/x86/kernel/syscall_table_32.S
@@ -340,3 +340,7 @@ 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
346 .long sys_syncfs
diff --git a/arch/x86/kernel/topology.c b/arch/x86/kernel/topology.c
index 7e4515957a1c..8927486a4649 100644
--- a/arch/x86/kernel/topology.c
+++ b/arch/x86/kernel/topology.c
@@ -39,7 +39,7 @@ int __ref arch_register_cpu(int num)
39 /* 39 /*
40 * CPU0 cannot be offlined due to several 40 * CPU0 cannot be offlined due to several
41 * restrictions and assumptions in kernel. This basically 41 * restrictions and assumptions in kernel. This basically
42 * doesnt add a control file, one cannot attempt to offline 42 * doesn't add a control file, one cannot attempt to offline
43 * BSP. 43 * BSP.
44 * 44 *
45 * Also certain PCI quirks require not to enable hotplug control 45 * Also certain PCI quirks require not to enable hotplug control
diff --git a/arch/x86/kernel/trampoline.c b/arch/x86/kernel/trampoline.c
index a375616d77f7..a91ae7709b49 100644
--- a/arch/x86/kernel/trampoline.c
+++ b/arch/x86/kernel/trampoline.c
@@ -2,39 +2,41 @@
2#include <linux/memblock.h> 2#include <linux/memblock.h>
3 3
4#include <asm/trampoline.h> 4#include <asm/trampoline.h>
5#include <asm/cacheflush.h>
5#include <asm/pgtable.h> 6#include <asm/pgtable.h>
6 7
7#if defined(CONFIG_X86_64) && defined(CONFIG_ACPI_SLEEP) 8unsigned char *x86_trampoline_base;
8#define __trampinit
9#define __trampinitdata
10#else
11#define __trampinit __cpuinit
12#define __trampinitdata __cpuinitdata
13#endif
14 9
15/* ready for x86_64 and x86 */ 10void __init setup_trampolines(void)
16unsigned char *__trampinitdata trampoline_base;
17
18void __init reserve_trampoline_memory(void)
19{ 11{
20 phys_addr_t mem; 12 phys_addr_t mem;
13 size_t size = PAGE_ALIGN(x86_trampoline_end - x86_trampoline_start);
21 14
22 /* Has to be in very low memory so we can execute real-mode AP code. */ 15 /* Has to be in very low memory so we can execute real-mode AP code. */
23 mem = memblock_find_in_range(0, 1<<20, TRAMPOLINE_SIZE, PAGE_SIZE); 16 mem = memblock_find_in_range(0, 1<<20, size, PAGE_SIZE);
24 if (mem == MEMBLOCK_ERROR) 17 if (mem == MEMBLOCK_ERROR)
25 panic("Cannot allocate trampoline\n"); 18 panic("Cannot allocate trampoline\n");
26 19
27 trampoline_base = __va(mem); 20 x86_trampoline_base = __va(mem);
28 memblock_x86_reserve_range(mem, mem + TRAMPOLINE_SIZE, "TRAMPOLINE"); 21 memblock_x86_reserve_range(mem, mem + size, "TRAMPOLINE");
22
23 printk(KERN_DEBUG "Base memory trampoline at [%p] %llx size %zu\n",
24 x86_trampoline_base, (unsigned long long)mem, size);
25
26 memcpy(x86_trampoline_base, x86_trampoline_start, size);
29} 27}
30 28
31/* 29/*
32 * Currently trivial. Write the real->protected mode 30 * setup_trampolines() gets called very early, to guarantee the
33 * bootstrap into the page concerned. The caller 31 * availability of low memory. This is before the proper kernel page
34 * has made sure it's suitably aligned. 32 * tables are set up, so we cannot set page permissions in that
33 * function. Thus, we use an arch_initcall instead.
35 */ 34 */
36unsigned long __trampinit setup_trampoline(void) 35static int __init configure_trampolines(void)
37{ 36{
38 memcpy(trampoline_base, trampoline_data, TRAMPOLINE_SIZE); 37 size_t size = PAGE_ALIGN(x86_trampoline_end - x86_trampoline_start);
39 return virt_to_phys(trampoline_base); 38
39 set_memory_x((unsigned long)x86_trampoline_base, size >> PAGE_SHIFT);
40 return 0;
40} 41}
42arch_initcall(configure_trampolines);
diff --git a/arch/x86/kernel/trampoline_32.S b/arch/x86/kernel/trampoline_32.S
index 8508237e8e43..451c0a7ef7fd 100644
--- a/arch/x86/kernel/trampoline_32.S
+++ b/arch/x86/kernel/trampoline_32.S
@@ -32,9 +32,11 @@
32#include <asm/segment.h> 32#include <asm/segment.h>
33#include <asm/page_types.h> 33#include <asm/page_types.h>
34 34
35/* We can free up trampoline after bootup if cpu hotplug is not supported. */ 35#ifdef CONFIG_SMP
36__CPUINITRODATA 36
37.code16 37 .section ".x86_trampoline","a"
38 .balign PAGE_SIZE
39 .code16
38 40
39ENTRY(trampoline_data) 41ENTRY(trampoline_data)
40r_base = . 42r_base = .
@@ -44,7 +46,7 @@ r_base = .
44 46
45 cli # We should be safe anyway 47 cli # We should be safe anyway
46 48
47 movl $0xA5A5A5A5, trampoline_data - r_base 49 movl $0xA5A5A5A5, trampoline_status - r_base
48 # write marker for master knows we're running 50 # write marker for master knows we're running
49 51
50 /* GDT tables in non default location kernel can be beyond 16MB and 52 /* GDT tables in non default location kernel can be beyond 16MB and
@@ -72,5 +74,10 @@ boot_idt_descr:
72 .word 0 # idt limit = 0 74 .word 0 # idt limit = 0
73 .long 0 # idt base = 0L 75 .long 0 # idt base = 0L
74 76
77ENTRY(trampoline_status)
78 .long 0
79
75.globl trampoline_end 80.globl trampoline_end
76trampoline_end: 81trampoline_end:
82
83#endif /* CONFIG_SMP */
diff --git a/arch/x86/kernel/trampoline_64.S b/arch/x86/kernel/trampoline_64.S
index 075d130efcf9..09ff51799e96 100644
--- a/arch/x86/kernel/trampoline_64.S
+++ b/arch/x86/kernel/trampoline_64.S
@@ -32,13 +32,9 @@
32#include <asm/segment.h> 32#include <asm/segment.h>
33#include <asm/processor-flags.h> 33#include <asm/processor-flags.h>
34 34
35#ifdef CONFIG_ACPI_SLEEP 35 .section ".x86_trampoline","a"
36.section .rodata, "a", @progbits 36 .balign PAGE_SIZE
37#else 37 .code16
38/* We can free up the trampoline after bootup if cpu hotplug is not supported. */
39__CPUINITRODATA
40#endif
41.code16
42 38
43ENTRY(trampoline_data) 39ENTRY(trampoline_data)
44r_base = . 40r_base = .
@@ -50,7 +46,7 @@ r_base = .
50 mov %ax, %ss 46 mov %ax, %ss
51 47
52 48
53 movl $0xA5A5A5A5, trampoline_data - r_base 49 movl $0xA5A5A5A5, trampoline_status - r_base
54 # write marker for master knows we're running 50 # write marker for master knows we're running
55 51
56 # Setup stack 52 # Setup stack
@@ -64,10 +60,13 @@ r_base = .
64 movzx %ax, %esi # Find the 32bit trampoline location 60 movzx %ax, %esi # Find the 32bit trampoline location
65 shll $4, %esi 61 shll $4, %esi
66 62
67 # Fixup the vectors 63 # Fixup the absolute vectors
68 addl %esi, startup_32_vector - r_base 64 leal (startup_32 - r_base)(%esi), %eax
69 addl %esi, startup_64_vector - r_base 65 movl %eax, startup_32_vector - r_base
70 addl %esi, tgdt + 2 - r_base # Fixup the gdt pointer 66 leal (startup_64 - r_base)(%esi), %eax
67 movl %eax, startup_64_vector - r_base
68 leal (tgdt - r_base)(%esi), %eax
69 movl %eax, (tgdt + 2 - r_base)
71 70
72 /* 71 /*
73 * GDT tables in non default location kernel can be beyond 16MB and 72 * GDT tables in non default location kernel can be beyond 16MB and
@@ -129,6 +128,7 @@ no_longmode:
129 jmp no_longmode 128 jmp no_longmode
130#include "verify_cpu.S" 129#include "verify_cpu.S"
131 130
131 .balign 4
132 # Careful these need to be in the same 64K segment as the above; 132 # Careful these need to be in the same 64K segment as the above;
133tidt: 133tidt:
134 .word 0 # idt limit = 0 134 .word 0 # idt limit = 0
@@ -156,6 +156,10 @@ startup_64_vector:
156 .long startup_64 - r_base 156 .long startup_64 - r_base
157 .word __KERNEL_CS, 0 157 .word __KERNEL_CS, 0
158 158
159 .balign 4
160ENTRY(trampoline_status)
161 .long 0
162
159trampoline_stack: 163trampoline_stack:
160 .org 0x1000 164 .org 0x1000
161trampoline_stack_end: 165trampoline_stack_end:
diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c
index ffe5755caa8b..9335bf7dd2e7 100644
--- a/arch/x86/kernel/tsc.c
+++ b/arch/x86/kernel/tsc.c
@@ -427,7 +427,7 @@ unsigned long native_calibrate_tsc(void)
427 * the delta to the previous read. We keep track of the min 427 * the delta to the previous read. We keep track of the min
428 * and max values of that delta. The delta is mostly defined 428 * and max values of that delta. The delta is mostly defined
429 * by the IO time of the PIT access, so we can detect when a 429 * by the IO time of the PIT access, so we can detect when a
430 * SMI/SMM disturbance happend between the two reads. If the 430 * SMI/SMM disturbance happened between the two reads. If the
431 * maximum time is significantly larger than the minimum time, 431 * maximum time is significantly larger than the minimum time,
432 * then we discard the result and have another try. 432 * then we discard the result and have another try.
433 * 433 *
@@ -900,7 +900,7 @@ static DECLARE_DELAYED_WORK(tsc_irqwork, tsc_refine_calibration_work);
900 * timer based, instead of loop based, we don't block the boot 900 * timer based, instead of loop based, we don't block the boot
901 * process while this longer calibration is done. 901 * process while this longer calibration is done.
902 * 902 *
903 * If there are any calibration anomolies (too many SMIs, etc), 903 * If there are any calibration anomalies (too many SMIs, etc),
904 * or the refined calibration is off by 1% of the fast early 904 * or the refined calibration is off by 1% of the fast early
905 * calibration, we throw out the new calibration and use the 905 * calibration, we throw out the new calibration and use the
906 * early calibration. 906 * early calibration.
diff --git a/arch/x86/kernel/verify_cpu.S b/arch/x86/kernel/verify_cpu.S
index 0edefc19a113..b9242bacbe59 100644
--- a/arch/x86/kernel/verify_cpu.S
+++ b/arch/x86/kernel/verify_cpu.S
@@ -18,7 +18,7 @@
18 * This file is expected to run in 32bit code. Currently: 18 * This file is expected to run in 32bit code. Currently:
19 * 19 *
20 * arch/x86/boot/compressed/head_64.S: Boot cpu verification 20 * arch/x86/boot/compressed/head_64.S: Boot cpu verification
21 * arch/x86/kernel/trampoline_64.S: secondary processor verfication 21 * arch/x86/kernel/trampoline_64.S: secondary processor verification
22 * arch/x86/kernel/head_32.S: processor startup 22 * arch/x86/kernel/head_32.S: processor startup
23 * 23 *
24 * verify_cpu, returns the status of longmode and SSE in register %eax. 24 * verify_cpu, returns the status of longmode and SSE in register %eax.
diff --git a/arch/x86/kernel/vmlinux.lds.S b/arch/x86/kernel/vmlinux.lds.S
index bf4700755184..624a2016198e 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)
@@ -230,7 +231,7 @@ SECTIONS
230 * output PHDR, so the next output section - .init.text - should 231 * output PHDR, so the next output section - .init.text - should
231 * start another segment - init. 232 * start another segment - init.
232 */ 233 */
233 PERCPU_VADDR(0, :percpu) 234 PERCPU_VADDR(INTERNODE_CACHE_BYTES, 0, :percpu)
234#endif 235#endif
235 236
236 INIT_TEXT_SECTION(PAGE_SIZE) 237 INIT_TEXT_SECTION(PAGE_SIZE)
@@ -240,6 +241,18 @@ SECTIONS
240 241
241 INIT_DATA_SECTION(16) 242 INIT_DATA_SECTION(16)
242 243
244 /*
245 * Code and data for a variety of lowlevel trampolines, to be
246 * copied into base memory (< 1 MiB) during initialization.
247 * Since it is copied early, the main copy can be discarded
248 * afterwards.
249 */
250 .x86_trampoline : AT(ADDR(.x86_trampoline) - LOAD_OFFSET) {
251 x86_trampoline_start = .;
252 *(.x86_trampoline)
253 x86_trampoline_end = .;
254 }
255
243 .x86_cpu_dev.init : AT(ADDR(.x86_cpu_dev.init) - LOAD_OFFSET) { 256 .x86_cpu_dev.init : AT(ADDR(.x86_cpu_dev.init) - LOAD_OFFSET) {
244 __x86_cpu_dev_start = .; 257 __x86_cpu_dev_start = .;
245 *(.x86_cpu_dev.init) 258 *(.x86_cpu_dev.init)
@@ -291,6 +304,7 @@ SECTIONS
291 *(.iommu_table) 304 *(.iommu_table)
292 __iommu_table_end = .; 305 __iommu_table_end = .;
293 } 306 }
307
294 . = ALIGN(8); 308 . = ALIGN(8);
295 /* 309 /*
296 * .exit.text is discard at runtime, not link time, to deal with 310 * .exit.text is discard at runtime, not link time, to deal with
@@ -305,7 +319,7 @@ SECTIONS
305 } 319 }
306 320
307#if !defined(CONFIG_X86_64) || !defined(CONFIG_SMP) 321#if !defined(CONFIG_X86_64) || !defined(CONFIG_SMP)
308 PERCPU(THREAD_SIZE) 322 PERCPU(INTERNODE_CACHE_BYTES, PAGE_SIZE)
309#endif 323#endif
310 324
311 . = ALIGN(PAGE_SIZE); 325 . = 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/kernel/xsave.c b/arch/x86/kernel/xsave.c
index 547128546cc3..a3911343976b 100644
--- a/arch/x86/kernel/xsave.c
+++ b/arch/x86/kernel/xsave.c
@@ -53,7 +53,7 @@ void __sanitize_i387_state(struct task_struct *tsk)
53 53
54 /* 54 /*
55 * None of the feature bits are in init state. So nothing else 55 * None of the feature bits are in init state. So nothing else
56 * to do for us, as the memory layout is upto date. 56 * to do for us, as the memory layout is up to date.
57 */ 57 */
58 if ((xstate_bv & pcntxt_mask) == pcntxt_mask) 58 if ((xstate_bv & pcntxt_mask) == pcntxt_mask)
59 return; 59 return;
generated by cgit v1.2.2 (git 2.25.0) at 2025-08-28 14:14:08 -0400