aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2009-06-20 13:49:48 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2009-06-20 13:49:48 -0400
commitc4c5ab3089c8a794eb0bdaa9794d0f055dd82412 (patch)
treee088b2aef48fb3db4d19abbdc5021aa42a8fb0a4
parent7fd5b632db00ebf8a26b5e86d6f01e501466e5ef (diff)
parent1d99100120ead486cd7a2502f19eaf1c1699d806 (diff)
Merge branch 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip
* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (45 commits) x86, mce: fix error path in mce_create_device() x86: use zalloc_cpumask_var for mce_dev_initialized x86: fix duplicated sysfs attribute x86: de-assembler-ize asm/desc.h i386: fix/simplify espfix stack switching, move it into assembly i386: fix return to 16-bit stack from NMI handler x86, ioapic: Don't call disconnect_bsp_APIC if no APIC present x86: Remove duplicated #include's x86: msr.h linux/types.h is only required for __KERNEL__ x86: nmi: Add Intel processor 0x6f4 to NMI perfctr1 workaround x86, mce: mce_intel.c needs <asm/apic.h> x86: apic/io_apic.c: dmar_msi_type should be static x86, io_apic.c: Work around compiler warning x86: mce: Don't touch THERMAL_APIC_VECTOR if no active APIC present x86: mce: Handle banks == 0 case in K7 quirk x86, boot: use .code16gcc instead of .code16 x86: correct the conversion of EFI memory types x86: cap iomem_resource to addressable physical memory x86, mce: rename _64.c files which are no longer 64-bit-specific x86, mce: mce.h cleanup ... Manually fix up trivial conflict in arch/x86/mm/fault.c
Diffstat
-rw-r--r--arch/x86/boot/bioscall.S2
-rw-r--r--arch/x86/include/asm/amd_iommu.h2
-rw-r--r--arch/x86/include/asm/atomic_32.h3
-rw-r--r--arch/x86/include/asm/desc.h26
-rw-r--r--arch/x86/include/asm/mce.h63
-rw-r--r--arch/x86/include/asm/msr.h7
-rw-r--r--arch/x86/include/asm/therm_throt.h9
-rw-r--r--arch/x86/kernel/amd_iommu.c16
-rw-r--r--arch/x86/kernel/amd_iommu_init.c26
-rw-r--r--arch/x86/kernel/apic/io_apic.c9
-rw-r--r--arch/x86/kernel/apic/probe_32.c11
-rw-r--r--arch/x86/kernel/apic/summit_32.c1
-rw-r--r--arch/x86/kernel/cpu/common.c5
-rw-r--r--arch/x86/kernel/cpu/mcheck/Makefile9
-rw-r--r--arch/x86/kernel/cpu/mcheck/k7.c3
-rw-r--r--arch/x86/kernel/cpu/mcheck/mce.c233
-rw-r--r--arch/x86/kernel/cpu/mcheck/mce.h38
-rw-r--r--arch/x86/kernel/cpu/mcheck/mce_amd.c (renamed from arch/x86/kernel/cpu/mcheck/mce_amd_64.c)0
-rw-r--r--arch/x86/kernel/cpu/mcheck/mce_intel.c250
-rw-r--r--arch/x86/kernel/cpu/mcheck/mce_intel_64.c248
-rw-r--r--arch/x86/kernel/cpu/mcheck/non-fatal.c3
-rw-r--r--arch/x86/kernel/cpu/mcheck/p4.c48
-rw-r--r--arch/x86/kernel/cpu/mcheck/p5.c15
-rw-r--r--arch/x86/kernel/cpu/mcheck/p6.c3
-rw-r--r--arch/x86/kernel/cpu/mcheck/therm_throt.c106
-rw-r--r--arch/x86/kernel/cpu/mcheck/winchip.c3
-rw-r--r--arch/x86/kernel/cpu/perfctr-watchdog.c12
-rw-r--r--arch/x86/kernel/crash.c6
-rw-r--r--arch/x86/kernel/efi.c31
-rw-r--r--arch/x86/kernel/entry_32.S64
-rw-r--r--arch/x86/kernel/head_32.S1
-rw-r--r--arch/x86/kernel/head_64.S1
-rw-r--r--arch/x86/kernel/hpet.c3
-rw-r--r--arch/x86/kernel/pci-dma.c2
-rw-r--r--arch/x86/kernel/traps.c3
-rw-r--r--arch/x86/mm/fault.c3
-rw-r--r--arch/x86/mm/init_64.c2
37 files changed, 670 insertions, 597 deletions
diff --git a/arch/x86/boot/bioscall.S b/arch/x86/boot/bioscall.S
index 507793739ea5..1dfbf64e52a2 100644
--- a/arch/x86/boot/bioscall.S
+++ b/arch/x86/boot/bioscall.S
@@ -13,7 +13,7 @@
13 * touching registers they shouldn't be. 13 * touching registers they shouldn't be.
14 */ 14 */
15 15
16 .code16 16 .code16gcc
17 .text 17 .text
18 .globl intcall 18 .globl intcall
19 .type intcall, @function 19 .type intcall, @function
diff --git a/arch/x86/include/asm/amd_iommu.h b/arch/x86/include/asm/amd_iommu.h
index 262e02820049..bdf96f119f06 100644
--- a/arch/x86/include/asm/amd_iommu.h
+++ b/arch/x86/include/asm/amd_iommu.h
@@ -29,9 +29,11 @@ extern void amd_iommu_detect(void);
29extern irqreturn_t amd_iommu_int_handler(int irq, void *data); 29extern irqreturn_t amd_iommu_int_handler(int irq, void *data);
30extern void amd_iommu_flush_all_domains(void); 30extern void amd_iommu_flush_all_domains(void);
31extern void amd_iommu_flush_all_devices(void); 31extern void amd_iommu_flush_all_devices(void);
32extern void amd_iommu_shutdown(void);
32#else 33#else
33static inline int amd_iommu_init(void) { return -ENODEV; } 34static inline int amd_iommu_init(void) { return -ENODEV; }
34static inline void amd_iommu_detect(void) { } 35static inline void amd_iommu_detect(void) { }
36static inline void amd_iommu_shutdown(void) { }
35#endif 37#endif
36 38
37#endif /* _ASM_X86_AMD_IOMMU_H */ 39#endif /* _ASM_X86_AMD_IOMMU_H */
diff --git a/arch/x86/include/asm/atomic_32.h b/arch/x86/include/asm/atomic_32.h
index 8cb9c814e120..2503d4e64c2a 100644
--- a/arch/x86/include/asm/atomic_32.h
+++ b/arch/x86/include/asm/atomic_32.h
@@ -257,7 +257,7 @@ typedef struct {
257 257
258/** 258/**
259 * atomic64_read - read atomic64 variable 259 * atomic64_read - read atomic64 variable
260 * @v: pointer of type atomic64_t 260 * @ptr: pointer of type atomic64_t
261 * 261 *
262 * Atomically reads the value of @v. 262 * Atomically reads the value of @v.
263 * Doesn't imply a read memory barrier. 263 * Doesn't imply a read memory barrier.
@@ -294,7 +294,6 @@ atomic64_cmpxchg(atomic64_t *ptr, unsigned long long old_val,
294 * atomic64_xchg - xchg atomic64 variable 294 * atomic64_xchg - xchg atomic64 variable
295 * @ptr: pointer to type atomic64_t 295 * @ptr: pointer to type atomic64_t
296 * @new_val: value to assign 296 * @new_val: value to assign
297 * @old_val: old value that was there
298 * 297 *
299 * Atomically xchgs the value of @ptr to @new_val and returns 298 * Atomically xchgs the value of @ptr to @new_val and returns
300 * the old value. 299 * the old value.
diff --git a/arch/x86/include/asm/desc.h b/arch/x86/include/asm/desc.h
index c45f415ce315..c993e9e0fed4 100644
--- a/arch/x86/include/asm/desc.h
+++ b/arch/x86/include/asm/desc.h
@@ -1,7 +1,6 @@
1#ifndef _ASM_X86_DESC_H 1#ifndef _ASM_X86_DESC_H
2#define _ASM_X86_DESC_H 2#define _ASM_X86_DESC_H
3 3
4#ifndef __ASSEMBLY__
5#include <asm/desc_defs.h> 4#include <asm/desc_defs.h>
6#include <asm/ldt.h> 5#include <asm/ldt.h>
7#include <asm/mmu.h> 6#include <asm/mmu.h>
@@ -380,29 +379,4 @@ static inline void set_system_intr_gate_ist(int n, void *addr, unsigned ist)
380 _set_gate(n, GATE_INTERRUPT, addr, 0x3, ist, __KERNEL_CS); 379 _set_gate(n, GATE_INTERRUPT, addr, 0x3, ist, __KERNEL_CS);
381} 380}
382 381
383#else
384/*
385 * GET_DESC_BASE reads the descriptor base of the specified segment.
386 *
387 * Args:
388 * idx - descriptor index
389 * gdt - GDT pointer
390 * base - 32bit register to which the base will be written
391 * lo_w - lo word of the "base" register
392 * lo_b - lo byte of the "base" register
393 * hi_b - hi byte of the low word of the "base" register
394 *
395 * Example:
396 * GET_DESC_BASE(GDT_ENTRY_ESPFIX_SS, %ebx, %eax, %ax, %al, %ah)
397 * Will read the base address of GDT_ENTRY_ESPFIX_SS and put it into %eax.
398 */
399#define GET_DESC_BASE(idx, gdt, base, lo_w, lo_b, hi_b) \
400 movb idx * 8 + 4(gdt), lo_b; \
401 movb idx * 8 + 7(gdt), hi_b; \
402 shll $16, base; \
403 movw idx * 8 + 2(gdt), lo_w;
404
405
406#endif /* __ASSEMBLY__ */
407
408#endif /* _ASM_X86_DESC_H */ 382#endif /* _ASM_X86_DESC_H */
diff --git a/arch/x86/include/asm/mce.h b/arch/x86/include/asm/mce.h
index 540a466e50f5..5cdd8d100ec9 100644
--- a/arch/x86/include/asm/mce.h
+++ b/arch/x86/include/asm/mce.h
@@ -102,15 +102,39 @@ struct mce_log {
102 102
103#ifdef __KERNEL__ 103#ifdef __KERNEL__
104 104
105#include <linux/percpu.h>
106#include <linux/init.h>
107#include <asm/atomic.h>
108
105extern int mce_disabled; 109extern int mce_disabled;
110extern int mce_p5_enabled;
106 111
107#include <asm/atomic.h> 112#ifdef CONFIG_X86_MCE
108#include <linux/percpu.h> 113void mcheck_init(struct cpuinfo_x86 *c);
114#else
115static inline void mcheck_init(struct cpuinfo_x86 *c) {}
116#endif
117
118#ifdef CONFIG_X86_OLD_MCE
119extern int nr_mce_banks;
120void amd_mcheck_init(struct cpuinfo_x86 *c);
121void intel_p4_mcheck_init(struct cpuinfo_x86 *c);
122void intel_p6_mcheck_init(struct cpuinfo_x86 *c);
123#endif
124
125#ifdef CONFIG_X86_ANCIENT_MCE
126void intel_p5_mcheck_init(struct cpuinfo_x86 *c);
127void winchip_mcheck_init(struct cpuinfo_x86 *c);
128static inline void enable_p5_mce(void) { mce_p5_enabled = 1; }
129#else
130static inline void intel_p5_mcheck_init(struct cpuinfo_x86 *c) {}
131static inline void winchip_mcheck_init(struct cpuinfo_x86 *c) {}
132static inline void enable_p5_mce(void) {}
133#endif
109 134
110void mce_setup(struct mce *m); 135void mce_setup(struct mce *m);
111void mce_log(struct mce *m); 136void mce_log(struct mce *m);
112DECLARE_PER_CPU(struct sys_device, mce_dev); 137DECLARE_PER_CPU(struct sys_device, mce_dev);
113extern void (*threshold_cpu_callback)(unsigned long action, unsigned int cpu);
114 138
115/* 139/*
116 * To support more than 128 would need to escape the predefined 140 * To support more than 128 would need to escape the predefined
@@ -145,12 +169,8 @@ int mce_available(struct cpuinfo_x86 *c);
145DECLARE_PER_CPU(unsigned, mce_exception_count); 169DECLARE_PER_CPU(unsigned, mce_exception_count);
146DECLARE_PER_CPU(unsigned, mce_poll_count); 170DECLARE_PER_CPU(unsigned, mce_poll_count);
147 171
148void mce_log_therm_throt_event(__u64 status);
149
150extern atomic_t mce_entry; 172extern atomic_t mce_entry;
151 173
152void do_machine_check(struct pt_regs *, long);
153
154typedef DECLARE_BITMAP(mce_banks_t, MAX_NR_BANKS); 174typedef DECLARE_BITMAP(mce_banks_t, MAX_NR_BANKS);
155DECLARE_PER_CPU(mce_banks_t, mce_poll_banks); 175DECLARE_PER_CPU(mce_banks_t, mce_poll_banks);
156 176
@@ -167,13 +187,32 @@ void mce_notify_process(void);
167DECLARE_PER_CPU(struct mce, injectm); 187DECLARE_PER_CPU(struct mce, injectm);
168extern struct file_operations mce_chrdev_ops; 188extern struct file_operations mce_chrdev_ops;
169 189
170#ifdef CONFIG_X86_MCE 190/*
171void mcheck_init(struct cpuinfo_x86 *c); 191 * Exception handler
172#else 192 */
173#define mcheck_init(c) do { } while (0) 193
174#endif 194/* Call the installed machine check handler for this CPU setup. */
195extern void (*machine_check_vector)(struct pt_regs *, long error_code);
196void do_machine_check(struct pt_regs *, long);
197
198/*
199 * Threshold handler
200 */
175 201
176extern void (*mce_threshold_vector)(void); 202extern void (*mce_threshold_vector)(void);
203extern void (*threshold_cpu_callback)(unsigned long action, unsigned int cpu);
204
205/*
206 * Thermal handler
207 */
208
209void intel_init_thermal(struct cpuinfo_x86 *c);
210
211#ifdef CONFIG_X86_NEW_MCE
212void mce_log_therm_throt_event(__u64 status);
213#else
214static inline void mce_log_therm_throt_event(__u64 status) {}
215#endif
177 216
178#endif /* __KERNEL__ */ 217#endif /* __KERNEL__ */
179#endif /* _ASM_X86_MCE_H */ 218#endif /* _ASM_X86_MCE_H */
diff --git a/arch/x86/include/asm/msr.h b/arch/x86/include/asm/msr.h
index 22603764e7db..48ad9d29484a 100644
--- a/arch/x86/include/asm/msr.h
+++ b/arch/x86/include/asm/msr.h
@@ -3,13 +3,10 @@
3 3
4#include <asm/msr-index.h> 4#include <asm/msr-index.h>
5 5
6#ifndef __ASSEMBLY__
7# include <linux/types.h>
8#endif
9
10#ifdef __KERNEL__ 6#ifdef __KERNEL__
11#ifndef __ASSEMBLY__ 7#ifndef __ASSEMBLY__
12 8
9#include <linux/types.h>
13#include <asm/asm.h> 10#include <asm/asm.h>
14#include <asm/errno.h> 11#include <asm/errno.h>
15#include <asm/cpumask.h> 12#include <asm/cpumask.h>
@@ -264,6 +261,4 @@ static inline int wrmsr_safe_on_cpu(unsigned int cpu, u32 msr_no, u32 l, u32 h)
264#endif /* CONFIG_SMP */ 261#endif /* CONFIG_SMP */
265#endif /* __ASSEMBLY__ */ 262#endif /* __ASSEMBLY__ */
266#endif /* __KERNEL__ */ 263#endif /* __KERNEL__ */
267
268
269#endif /* _ASM_X86_MSR_H */ 264#endif /* _ASM_X86_MSR_H */
diff --git a/arch/x86/include/asm/therm_throt.h b/arch/x86/include/asm/therm_throt.h
deleted file mode 100644
index c62349ee7860..000000000000
--- a/arch/x86/include/asm/therm_throt.h
+++ /dev/null
@@ -1,9 +0,0 @@
1#ifndef _ASM_X86_THERM_THROT_H
2#define _ASM_X86_THERM_THROT_H
3
4#include <asm/atomic.h>
5
6extern atomic_t therm_throt_en;
7int therm_throt_process(int curr);
8
9#endif /* _ASM_X86_THERM_THROT_H */
diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 1c60554537c3..9372f0406ad4 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -434,6 +434,16 @@ static void iommu_flush_tlb(struct amd_iommu *iommu, u16 domid)
434 iommu_queue_inv_iommu_pages(iommu, address, domid, 0, 1); 434 iommu_queue_inv_iommu_pages(iommu, address, domid, 0, 1);
435} 435}
436 436
437/* Flush the whole IO/TLB for a given protection domain - including PDE */
438static void iommu_flush_tlb_pde(struct amd_iommu *iommu, u16 domid)
439{
440 u64 address = CMD_INV_IOMMU_ALL_PAGES_ADDRESS;
441
442 INC_STATS_COUNTER(domain_flush_single);
443
444 iommu_queue_inv_iommu_pages(iommu, address, domid, 1, 1);
445}
446
437/* 447/*
438 * This function is used to flush the IO/TLB for a given protection domain 448 * This function is used to flush the IO/TLB for a given protection domain
439 * on every IOMMU in the system 449 * on every IOMMU in the system
@@ -1078,7 +1088,13 @@ static void attach_device(struct amd_iommu *iommu,
1078 amd_iommu_pd_table[devid] = domain; 1088 amd_iommu_pd_table[devid] = domain;
1079 write_unlock_irqrestore(&amd_iommu_devtable_lock, flags); 1089 write_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
1080 1090
1091 /*
1092 * We might boot into a crash-kernel here. The crashed kernel
1093 * left the caches in the IOMMU dirty. So we have to flush
1094 * here to evict all dirty stuff.
1095 */
1081 iommu_queue_inv_dev_entry(iommu, devid); 1096 iommu_queue_inv_dev_entry(iommu, devid);
1097 iommu_flush_tlb_pde(iommu, domain->id);
1082} 1098}
1083 1099
1084/* 1100/*
diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c
index 238989ec077d..10b2accd12ea 100644
--- a/arch/x86/kernel/amd_iommu_init.c
+++ b/arch/x86/kernel/amd_iommu_init.c
@@ -260,6 +260,14 @@ static void iommu_enable(struct amd_iommu *iommu)
260 260
261static void iommu_disable(struct amd_iommu *iommu) 261static void iommu_disable(struct amd_iommu *iommu)
262{ 262{
263 /* Disable command buffer */
264 iommu_feature_disable(iommu, CONTROL_CMDBUF_EN);
265
266 /* Disable event logging and event interrupts */
267 iommu_feature_disable(iommu, CONTROL_EVT_INT_EN);
268 iommu_feature_disable(iommu, CONTROL_EVT_LOG_EN);
269
270 /* Disable IOMMU hardware itself */
263 iommu_feature_disable(iommu, CONTROL_IOMMU_EN); 271 iommu_feature_disable(iommu, CONTROL_IOMMU_EN);
264} 272}
265 273
@@ -478,6 +486,10 @@ static void iommu_enable_event_buffer(struct amd_iommu *iommu)
478 memcpy_toio(iommu->mmio_base + MMIO_EVT_BUF_OFFSET, 486 memcpy_toio(iommu->mmio_base + MMIO_EVT_BUF_OFFSET,
479 &entry, sizeof(entry)); 487 &entry, sizeof(entry));
480 488
489 /* set head and tail to zero manually */
490 writel(0x00, iommu->mmio_base + MMIO_EVT_HEAD_OFFSET);
491 writel(0x00, iommu->mmio_base + MMIO_EVT_TAIL_OFFSET);
492
481 iommu_feature_enable(iommu, CONTROL_EVT_LOG_EN); 493 iommu_feature_enable(iommu, CONTROL_EVT_LOG_EN);
482} 494}
483 495
@@ -1042,6 +1054,7 @@ static void enable_iommus(void)
1042 struct amd_iommu *iommu; 1054 struct amd_iommu *iommu;
1043 1055
1044 for_each_iommu(iommu) { 1056 for_each_iommu(iommu) {
1057 iommu_disable(iommu);
1045 iommu_set_device_table(iommu); 1058 iommu_set_device_table(iommu);
1046 iommu_enable_command_buffer(iommu); 1059 iommu_enable_command_buffer(iommu);
1047 iommu_enable_event_buffer(iommu); 1060 iommu_enable_event_buffer(iommu);
@@ -1066,12 +1079,6 @@ static void disable_iommus(void)
1066 1079
1067static int amd_iommu_resume(struct sys_device *dev) 1080static int amd_iommu_resume(struct sys_device *dev)
1068{ 1081{
1069 /*
1070 * Disable IOMMUs before reprogramming the hardware registers.
1071 * IOMMU is still enabled from the resume kernel.
1072 */
1073 disable_iommus();
1074
1075 /* re-load the hardware */ 1082 /* re-load the hardware */
1076 enable_iommus(); 1083 enable_iommus();
1077 1084
@@ -1079,8 +1086,8 @@ static int amd_iommu_resume(struct sys_device *dev)
1079 * we have to flush after the IOMMUs are enabled because a 1086 * we have to flush after the IOMMUs are enabled because a
1080 * disabled IOMMU will never execute the commands we send 1087 * disabled IOMMU will never execute the commands we send
1081 */ 1088 */
1082 amd_iommu_flush_all_domains();
1083 amd_iommu_flush_all_devices(); 1089 amd_iommu_flush_all_devices();
1090 amd_iommu_flush_all_domains();
1084 1091
1085 return 0; 1092 return 0;
1086} 1093}
@@ -1273,6 +1280,11 @@ free:
1273 goto out; 1280 goto out;
1274} 1281}
1275 1282
1283void amd_iommu_shutdown(void)
1284{
1285 disable_iommus();
1286}
1287
1276/**************************************************************************** 1288/****************************************************************************
1277 * 1289 *
1278 * Early detect code. This code runs at IOMMU detection time in the DMA 1290 * Early detect code. This code runs at IOMMU detection time in the DMA
diff --git a/arch/x86/kernel/apic/io_apic.c b/arch/x86/kernel/apic/io_apic.c
index ef8d9290c7ea..b7a79207295e 100644
--- a/arch/x86/kernel/apic/io_apic.c
+++ b/arch/x86/kernel/apic/io_apic.c
@@ -462,7 +462,8 @@ static struct IO_APIC_route_entry ioapic_read_entry(int apic, int pin)
462static void 462static void
463__ioapic_write_entry(int apic, int pin, struct IO_APIC_route_entry e) 463__ioapic_write_entry(int apic, int pin, struct IO_APIC_route_entry e)
464{ 464{
465 union entry_union eu; 465 union entry_union eu = {{0, 0}};
466
466 eu.entry = e; 467 eu.entry = e;
467 io_apic_write(apic, 0x11 + 2*pin, eu.w2); 468 io_apic_write(apic, 0x11 + 2*pin, eu.w2);
468 io_apic_write(apic, 0x10 + 2*pin, eu.w1); 469 io_apic_write(apic, 0x10 + 2*pin, eu.w1);
@@ -2003,7 +2004,9 @@ void disable_IO_APIC(void)
2003 /* 2004 /*
2004 * Use virtual wire A mode when interrupt remapping is enabled. 2005 * Use virtual wire A mode when interrupt remapping is enabled.
2005 */ 2006 */
2006 disconnect_bsp_APIC(!intr_remapping_enabled && ioapic_i8259.pin != -1); 2007 if (cpu_has_apic)
2008 disconnect_bsp_APIC(!intr_remapping_enabled &&
2009 ioapic_i8259.pin != -1);
2007} 2010}
2008 2011
2009#ifdef CONFIG_X86_32 2012#ifdef CONFIG_X86_32
@@ -3567,7 +3570,7 @@ static int dmar_msi_set_affinity(unsigned int irq, const struct cpumask *mask)
3567 3570
3568#endif /* CONFIG_SMP */ 3571#endif /* CONFIG_SMP */
3569 3572
3570struct irq_chip dmar_msi_type = { 3573static struct irq_chip dmar_msi_type = {
3571 .name = "DMAR_MSI", 3574 .name = "DMAR_MSI",
3572 .unmask = dmar_msi_unmask, 3575 .unmask = dmar_msi_unmask,
3573 .mask = dmar_msi_mask, 3576 .mask = dmar_msi_mask,
diff --git a/arch/x86/kernel/apic/probe_32.c b/arch/x86/kernel/apic/probe_32.c
index 440a8bccd91a..0c0182cc947d 100644
--- a/arch/x86/kernel/apic/probe_32.c
+++ b/arch/x86/kernel/apic/probe_32.c
@@ -20,23 +20,12 @@
20#include <asm/apic.h> 20#include <asm/apic.h>
21#include <asm/setup.h> 21#include <asm/setup.h>
22 22
23#include <linux/threads.h>
24#include <linux/cpumask.h>
25#include <asm/mpspec.h>
26#include <asm/fixmap.h>
27#include <asm/apicdef.h>
28#include <linux/kernel.h>
29#include <linux/string.h>
30#include <linux/smp.h> 23#include <linux/smp.h>
31#include <linux/init.h>
32#include <asm/ipi.h> 24#include <asm/ipi.h>
33 25
34#include <linux/smp.h>
35#include <linux/init.h>
36#include <linux/interrupt.h> 26#include <linux/interrupt.h>
37#include <asm/acpi.h> 27#include <asm/acpi.h>
38#include <asm/e820.h> 28#include <asm/e820.h>
39#include <asm/setup.h>
40 29
41#ifdef CONFIG_HOTPLUG_CPU 30#ifdef CONFIG_HOTPLUG_CPU
42#define DEFAULT_SEND_IPI (1) 31#define DEFAULT_SEND_IPI (1)
diff --git a/arch/x86/kernel/apic/summit_32.c b/arch/x86/kernel/apic/summit_32.c
index 344eee4ac0a4..eafdfbd1ea95 100644
--- a/arch/x86/kernel/apic/summit_32.c
+++ b/arch/x86/kernel/apic/summit_32.c
@@ -44,7 +44,6 @@
44#include <asm/ipi.h> 44#include <asm/ipi.h>
45#include <linux/kernel.h> 45#include <linux/kernel.h>
46#include <linux/string.h> 46#include <linux/string.h>
47#include <linux/init.h>
48#include <linux/gfp.h> 47#include <linux/gfp.h>
49#include <linux/smp.h> 48#include <linux/smp.h>
50 49
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index 9fa33886c0d7..6b26d4deada0 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -108,7 +108,7 @@ DEFINE_PER_CPU_PAGE_ALIGNED(struct gdt_page, gdt_page) = { .gdt = {
108 /* data */ 108 /* data */
109 [GDT_ENTRY_APMBIOS_BASE+2] = { { { 0x0000ffff, 0x00409200 } } }, 109 [GDT_ENTRY_APMBIOS_BASE+2] = { { { 0x0000ffff, 0x00409200 } } },
110 110
111 [GDT_ENTRY_ESPFIX_SS] = { { { 0x00000000, 0x00c09200 } } }, 111 [GDT_ENTRY_ESPFIX_SS] = { { { 0x0000ffff, 0x00cf9200 } } },
112 [GDT_ENTRY_PERCPU] = { { { 0x0000ffff, 0x00cf9200 } } }, 112 [GDT_ENTRY_PERCPU] = { { { 0x0000ffff, 0x00cf9200 } } },
113 GDT_STACK_CANARY_INIT 113 GDT_STACK_CANARY_INIT
114#endif 114#endif
@@ -848,6 +848,9 @@ static void __cpuinit identify_cpu(struct cpuinfo_x86 *c)
848#if defined(CONFIG_NUMA) && defined(CONFIG_X86_64) 848#if defined(CONFIG_NUMA) && defined(CONFIG_X86_64)
849 numa_add_cpu(smp_processor_id()); 849 numa_add_cpu(smp_processor_id());
850#endif 850#endif
851
852 /* Cap the iomem address space to what is addressable on all CPUs */
853 iomem_resource.end &= (1ULL << c->x86_phys_bits) - 1;
851} 854}
852 855
853#ifdef CONFIG_X86_64 856#ifdef CONFIG_X86_64
diff --git a/arch/x86/kernel/cpu/mcheck/Makefile b/arch/x86/kernel/cpu/mcheck/Makefile
index 45004faf67ea..188a1ca5ad2b 100644
--- a/arch/x86/kernel/cpu/mcheck/Makefile
+++ b/arch/x86/kernel/cpu/mcheck/Makefile
@@ -1,11 +1,12 @@
1obj-y = mce.o therm_throt.o 1obj-y = mce.o
2 2
3obj-$(CONFIG_X86_NEW_MCE) += mce-severity.o 3obj-$(CONFIG_X86_NEW_MCE) += mce-severity.o
4obj-$(CONFIG_X86_OLD_MCE) += k7.o p4.o p6.o 4obj-$(CONFIG_X86_OLD_MCE) += k7.o p4.o p6.o
5obj-$(CONFIG_X86_ANCIENT_MCE) += winchip.o p5.o 5obj-$(CONFIG_X86_ANCIENT_MCE) += winchip.o p5.o
6obj-$(CONFIG_X86_MCE_P4THERMAL) += mce_intel.o 6obj-$(CONFIG_X86_MCE_INTEL) += mce_intel.o
7obj-$(CONFIG_X86_MCE_INTEL) += mce_intel_64.o mce_intel.o 7obj-$(CONFIG_X86_MCE_AMD) += mce_amd.o
8obj-$(CONFIG_X86_MCE_AMD) += mce_amd_64.o
9obj-$(CONFIG_X86_MCE_NONFATAL) += non-fatal.o 8obj-$(CONFIG_X86_MCE_NONFATAL) += non-fatal.o
10obj-$(CONFIG_X86_MCE_THRESHOLD) += threshold.o 9obj-$(CONFIG_X86_MCE_THRESHOLD) += threshold.o
11obj-$(CONFIG_X86_MCE_INJECT) += mce-inject.o 10obj-$(CONFIG_X86_MCE_INJECT) += mce-inject.o
11
12obj-$(CONFIG_X86_THERMAL_VECTOR) += therm_throt.o
diff --git a/arch/x86/kernel/cpu/mcheck/k7.c b/arch/x86/kernel/cpu/mcheck/k7.c
index 89e510424152..b945d5dbc609 100644
--- a/arch/x86/kernel/cpu/mcheck/k7.c
+++ b/arch/x86/kernel/cpu/mcheck/k7.c
@@ -10,10 +10,9 @@
10 10
11#include <asm/processor.h> 11#include <asm/processor.h>
12#include <asm/system.h> 12#include <asm/system.h>
13#include <asm/mce.h>
13#include <asm/msr.h> 14#include <asm/msr.h>
14 15
15#include "mce.h"
16
17/* Machine Check Handler For AMD Athlon/Duron: */ 16/* Machine Check Handler For AMD Athlon/Duron: */
18static void k7_machine_check(struct pt_regs *regs, long error_code) 17static void k7_machine_check(struct pt_regs *regs, long error_code)
19{ 18{
diff --git a/arch/x86/kernel/cpu/mcheck/mce.c b/arch/x86/kernel/cpu/mcheck/mce.c
index fabba15e4558..284d1de968bc 100644
--- a/arch/x86/kernel/cpu/mcheck/mce.c
+++ b/arch/x86/kernel/cpu/mcheck/mce.c
@@ -44,7 +44,6 @@
44#include <asm/msr.h> 44#include <asm/msr.h>
45 45
46#include "mce-internal.h" 46#include "mce-internal.h"
47#include "mce.h"
48 47
49/* Handle unconfigured int18 (should never happen) */ 48/* Handle unconfigured int18 (should never happen) */
50static void unexpected_machine_check(struct pt_regs *regs, long error_code) 49static void unexpected_machine_check(struct pt_regs *regs, long error_code)
@@ -57,7 +56,7 @@ static void unexpected_machine_check(struct pt_regs *regs, long error_code)
57void (*machine_check_vector)(struct pt_regs *, long error_code) = 56void (*machine_check_vector)(struct pt_regs *, long error_code) =
58 unexpected_machine_check; 57 unexpected_machine_check;
59 58
60int mce_disabled; 59int mce_disabled __read_mostly;
61 60
62#ifdef CONFIG_X86_NEW_MCE 61#ifdef CONFIG_X86_NEW_MCE
63 62
@@ -76,21 +75,22 @@ DEFINE_PER_CPU(unsigned, mce_exception_count);
76 * 2: SIGBUS or log uncorrected errors (if possible), log corrected errors 75 * 2: SIGBUS or log uncorrected errors (if possible), log corrected errors
77 * 3: never panic or SIGBUS, log all errors (for testing only) 76 * 3: never panic or SIGBUS, log all errors (for testing only)
78 */ 77 */
79static int tolerant = 1; 78static int tolerant __read_mostly = 1;
80static int banks; 79static int banks __read_mostly;
81static u64 *bank; 80static u64 *bank __read_mostly;
82static unsigned long notify_user; 81static int rip_msr __read_mostly;
83static int rip_msr; 82static int mce_bootlog __read_mostly = -1;
84static int mce_bootlog = -1; 83static int monarch_timeout __read_mostly = -1;
85static int monarch_timeout = -1; 84static int mce_panic_timeout __read_mostly;
86static int mce_panic_timeout; 85static int mce_dont_log_ce __read_mostly;
87static int mce_dont_log_ce; 86int mce_cmci_disabled __read_mostly;
88int mce_cmci_disabled; 87int mce_ignore_ce __read_mostly;
89int mce_ignore_ce; 88int mce_ser __read_mostly;
90int mce_ser; 89
91 90/* User mode helper program triggered by machine check event */
92static char trigger[128]; 91static unsigned long mce_need_notify;
93static char *trigger_argv[2] = { trigger, NULL }; 92static char mce_helper[128];
93static char *mce_helper_argv[2] = { mce_helper, NULL };
94 94
95static unsigned long dont_init_banks; 95static unsigned long dont_init_banks;
96 96
@@ -180,7 +180,7 @@ void mce_log(struct mce *mce)
180 wmb(); 180 wmb();
181 181
182 mce->finished = 1; 182 mce->finished = 1;
183 set_bit(0, &notify_user); 183 set_bit(0, &mce_need_notify);
184} 184}
185 185
186static void print_mce(struct mce *m) 186static void print_mce(struct mce *m)
@@ -691,18 +691,21 @@ static atomic_t global_nwo;
691 * in the entry order. 691 * in the entry order.
692 * TBD double check parallel CPU hotunplug 692 * TBD double check parallel CPU hotunplug
693 */ 693 */
694static int mce_start(int no_way_out, int *order) 694static int mce_start(int *no_way_out)
695{ 695{
696 int nwo; 696 int order;
697 int cpus = num_online_cpus(); 697 int cpus = num_online_cpus();
698 u64 timeout = (u64)monarch_timeout * NSEC_PER_USEC; 698 u64 timeout = (u64)monarch_timeout * NSEC_PER_USEC;
699 699
700 if (!timeout) { 700 if (!timeout)
701 *order = -1; 701 return -1;
702 return no_way_out;
703 }
704 702
705 atomic_add(no_way_out, &global_nwo); 703 atomic_add(*no_way_out, &global_nwo);
704 /*
705 * global_nwo should be updated before mce_callin
706 */
707 smp_wmb();
708 order = atomic_add_return(1, &mce_callin);
706 709
707 /* 710 /*
708 * Wait for everyone. 711 * Wait for everyone.
@@ -710,40 +713,43 @@ static int mce_start(int no_way_out, int *order)
710 while (atomic_read(&mce_callin) != cpus) { 713 while (atomic_read(&mce_callin) != cpus) {
711 if (mce_timed_out(&timeout)) { 714 if (mce_timed_out(&timeout)) {
712 atomic_set(&global_nwo, 0); 715 atomic_set(&global_nwo, 0);
713 *order = -1; 716 return -1;
714 return no_way_out;
715 } 717 }
716 ndelay(SPINUNIT); 718 ndelay(SPINUNIT);
717 } 719 }
718 720
719 /* 721 /*
720 * Cache the global no_way_out state. 722 * mce_callin should be read before global_nwo
721 */ 723 */
722 nwo = atomic_read(&global_nwo); 724 smp_rmb();
723 725
724 /* 726 if (order == 1) {
725 * Monarch starts executing now, the others wait. 727 /*
726 */ 728 * Monarch: Starts executing now, the others wait.
727 if (*order == 1) { 729 */
728 atomic_set(&mce_executing, 1); 730 atomic_set(&mce_executing, 1);
729 return nwo; 731 } else {
732 /*
733 * Subject: Now start the scanning loop one by one in
734 * the original callin order.
735 * This way when there are any shared banks it will be
736 * only seen by one CPU before cleared, avoiding duplicates.
737 */
738 while (atomic_read(&mce_executing) < order) {
739 if (mce_timed_out(&timeout)) {
740 atomic_set(&global_nwo, 0);
741 return -1;
742 }
743 ndelay(SPINUNIT);
744 }
730 } 745 }
731 746
732 /* 747 /*
733 * Now start the scanning loop one by one 748 * Cache the global no_way_out state.
734 * in the original callin order.
735 * This way when there are any shared banks it will
736 * be only seen by one CPU before cleared, avoiding duplicates.
737 */ 749 */
738 while (atomic_read(&mce_executing) < *order) { 750 *no_way_out = atomic_read(&global_nwo);
739 if (mce_timed_out(&timeout)) { 751
740 atomic_set(&global_nwo, 0); 752 return order;
741 *order = -1;
742 return no_way_out;
743 }
744 ndelay(SPINUNIT);
745 }
746 return nwo;
747} 753}
748 754
749/* 755/*
@@ -863,7 +869,6 @@ void do_machine_check(struct pt_regs *regs, long error_code)
863 * check handler. 869 * check handler.
864 */ 870 */
865 int order; 871 int order;
866
867 /* 872 /*
868 * If no_way_out gets set, there is no safe way to recover from this 873 * If no_way_out gets set, there is no safe way to recover from this
869 * MCE. If tolerant is cranked up, we'll try anyway. 874 * MCE. If tolerant is cranked up, we'll try anyway.
@@ -887,7 +892,6 @@ void do_machine_check(struct pt_regs *regs, long error_code)
887 if (!banks) 892 if (!banks)
888 goto out; 893 goto out;
889 894
890 order = atomic_add_return(1, &mce_callin);
891 mce_setup(&m); 895 mce_setup(&m);
892 896
893 m.mcgstatus = mce_rdmsrl(MSR_IA32_MCG_STATUS); 897 m.mcgstatus = mce_rdmsrl(MSR_IA32_MCG_STATUS);
@@ -909,7 +913,7 @@ void do_machine_check(struct pt_regs *regs, long error_code)
909 * This way we don't report duplicated events on shared banks 913 * This way we don't report duplicated events on shared banks
910 * because the first one to see it will clear it. 914 * because the first one to see it will clear it.
911 */ 915 */
912 no_way_out = mce_start(no_way_out, &order); 916 order = mce_start(&no_way_out);
913 for (i = 0; i < banks; i++) { 917 for (i = 0; i < banks; i++) {
914 __clear_bit(i, toclear); 918 __clear_bit(i, toclear);
915 if (!bank[i]) 919 if (!bank[i])
@@ -1118,7 +1122,7 @@ static void mcheck_timer(unsigned long data)
1118 1122
1119static void mce_do_trigger(struct work_struct *work) 1123static void mce_do_trigger(struct work_struct *work)
1120{ 1124{
1121 call_usermodehelper(trigger, trigger_argv, NULL, UMH_NO_WAIT); 1125 call_usermodehelper(mce_helper, mce_helper_argv, NULL, UMH_NO_WAIT);
1122} 1126}
1123 1127
1124static DECLARE_WORK(mce_trigger_work, mce_do_trigger); 1128static DECLARE_WORK(mce_trigger_work, mce_do_trigger);
@@ -1135,7 +1139,7 @@ int mce_notify_irq(void)
1135 1139
1136 clear_thread_flag(TIF_MCE_NOTIFY); 1140 clear_thread_flag(TIF_MCE_NOTIFY);
1137 1141
1138 if (test_and_clear_bit(0, &notify_user)) { 1142 if (test_and_clear_bit(0, &mce_need_notify)) {
1139 wake_up_interruptible(&mce_wait); 1143 wake_up_interruptible(&mce_wait);
1140 1144
1141 /* 1145 /*
@@ -1143,7 +1147,7 @@ int mce_notify_irq(void)
1143 * work_pending is always cleared before the function is 1147 * work_pending is always cleared before the function is
1144 * executed. 1148 * executed.
1145 */ 1149 */
1146 if (trigger[0] && !work_pending(&mce_trigger_work)) 1150 if (mce_helper[0] && !work_pending(&mce_trigger_work))
1147 schedule_work(&mce_trigger_work); 1151 schedule_work(&mce_trigger_work);
1148 1152
1149 if (__ratelimit(&ratelimit)) 1153 if (__ratelimit(&ratelimit))
@@ -1245,7 +1249,7 @@ static void mce_cpu_quirks(struct cpuinfo_x86 *c)
1245 * Various K7s with broken bank 0 around. Always disable 1249 * Various K7s with broken bank 0 around. Always disable
1246 * by default. 1250 * by default.
1247 */ 1251 */
1248 if (c->x86 == 6) 1252 if (c->x86 == 6 && banks > 0)
1249 bank[0] = 0; 1253 bank[0] = 0;
1250 } 1254 }
1251 1255
@@ -1282,8 +1286,7 @@ static void __cpuinit mce_ancient_init(struct cpuinfo_x86 *c)
1282 return; 1286 return;
1283 switch (c->x86_vendor) { 1287 switch (c->x86_vendor) {
1284 case X86_VENDOR_INTEL: 1288 case X86_VENDOR_INTEL:
1285 if (mce_p5_enabled()) 1289 intel_p5_mcheck_init(c);
1286 intel_p5_mcheck_init(c);
1287 break; 1290 break;
1288 case X86_VENDOR_CENTAUR: 1291 case X86_VENDOR_CENTAUR:
1289 winchip_mcheck_init(c); 1292 winchip_mcheck_init(c);
@@ -1609,8 +1612,9 @@ static int mce_resume(struct sys_device *dev)
1609static void mce_cpu_restart(void *data) 1612static void mce_cpu_restart(void *data)
1610{ 1613{
1611 del_timer_sync(&__get_cpu_var(mce_timer)); 1614 del_timer_sync(&__get_cpu_var(mce_timer));
1612 if (mce_available(&current_cpu_data)) 1615 if (!mce_available(&current_cpu_data))
1613 mce_init(); 1616 return;
1617 mce_init();
1614 mce_init_timer(); 1618 mce_init_timer();
1615} 1619}
1616 1620
@@ -1620,6 +1624,26 @@ static void mce_restart(void)
1620 on_each_cpu(mce_cpu_restart, NULL, 1); 1624 on_each_cpu(mce_cpu_restart, NULL, 1);
1621} 1625}
1622 1626
1627/* Toggle features for corrected errors */
1628static void mce_disable_ce(void *all)
1629{
1630 if (!mce_available(&current_cpu_data))
1631 return;
1632 if (all)
1633 del_timer_sync(&__get_cpu_var(mce_timer));
1634 cmci_clear();
1635}
1636
1637static void mce_enable_ce(void *all)
1638{
1639 if (!mce_available(&current_cpu_data))
1640 return;
1641 cmci_reenable();
1642 cmci_recheck();
1643 if (all)
1644 mce_init_timer();
1645}
1646
1623static struct sysdev_class mce_sysclass = { 1647static struct sysdev_class mce_sysclass = {
1624 .suspend = mce_suspend, 1648 .suspend = mce_suspend,
1625 .shutdown = mce_shutdown, 1649 .shutdown = mce_shutdown,
@@ -1659,9 +1683,9 @@ static ssize_t set_bank(struct sys_device *s, struct sysdev_attribute *attr,
1659static ssize_t 1683static ssize_t
1660show_trigger(struct sys_device *s, struct sysdev_attribute *attr, char *buf) 1684show_trigger(struct sys_device *s, struct sysdev_attribute *attr, char *buf)
1661{ 1685{
1662 strcpy(buf, trigger); 1686 strcpy(buf, mce_helper);
1663 strcat(buf, "\n"); 1687 strcat(buf, "\n");
1664 return strlen(trigger) + 1; 1688 return strlen(mce_helper) + 1;
1665} 1689}
1666 1690
1667static ssize_t set_trigger(struct sys_device *s, struct sysdev_attribute *attr, 1691static ssize_t set_trigger(struct sys_device *s, struct sysdev_attribute *attr,
@@ -1670,10 +1694,10 @@ static ssize_t set_trigger(struct sys_device *s, struct sysdev_attribute *attr,
1670 char *p; 1694 char *p;
1671 int len; 1695 int len;
1672 1696
1673 strncpy(trigger, buf, sizeof(trigger)); 1697 strncpy(mce_helper, buf, sizeof(mce_helper));
1674 trigger[sizeof(trigger)-1] = 0; 1698 mce_helper[sizeof(mce_helper)-1] = 0;
1675 len = strlen(trigger); 1699 len = strlen(mce_helper);
1676 p = strchr(trigger, '\n'); 1700 p = strchr(mce_helper, '\n');
1677 1701
1678 if (*p) 1702 if (*p)
1679 *p = 0; 1703 *p = 0;
@@ -1681,6 +1705,52 @@ static ssize_t set_trigger(struct sys_device *s, struct sysdev_attribute *attr,
1681 return len; 1705 return len;
1682} 1706}
1683 1707
1708static ssize_t set_ignore_ce(struct sys_device *s,
1709 struct sysdev_attribute *attr,
1710 const char *buf, size_t size)
1711{
1712 u64 new;
1713
1714 if (strict_strtoull(buf, 0, &new) < 0)
1715 return -EINVAL;
1716
1717 if (mce_ignore_ce ^ !!new) {
1718 if (new) {
1719 /* disable ce features */
1720 on_each_cpu(mce_disable_ce, (void *)1, 1);
1721 mce_ignore_ce = 1;
1722 } else {
1723 /* enable ce features */
1724 mce_ignore_ce = 0;
1725 on_each_cpu(mce_enable_ce, (void *)1, 1);
1726 }
1727 }
1728 return size;
1729}
1730
1731static ssize_t set_cmci_disabled(struct sys_device *s,
1732 struct sysdev_attribute *attr,
1733 const char *buf, size_t size)
1734{
1735 u64 new;
1736
1737 if (strict_strtoull(buf, 0, &new) < 0)
1738 return -EINVAL;
1739
1740 if (mce_cmci_disabled ^ !!new) {
1741 if (new) {
1742 /* disable cmci */
1743 on_each_cpu(mce_disable_ce, NULL, 1);
1744 mce_cmci_disabled = 1;
1745 } else {
1746 /* enable cmci */
1747 mce_cmci_disabled = 0;
1748 on_each_cpu(mce_enable_ce, NULL, 1);
1749 }
1750 }
1751 return size;
1752}
1753
1684static ssize_t store_int_with_restart(struct sys_device *s, 1754static ssize_t store_int_with_restart(struct sys_device *s,
1685 struct sysdev_attribute *attr, 1755 struct sysdev_attribute *attr,
1686 const char *buf, size_t size) 1756 const char *buf, size_t size)
@@ -1693,6 +1763,7 @@ static ssize_t store_int_with_restart(struct sys_device *s,
1693static SYSDEV_ATTR(trigger, 0644, show_trigger, set_trigger); 1763static SYSDEV_ATTR(trigger, 0644, show_trigger, set_trigger);
1694static SYSDEV_INT_ATTR(tolerant, 0644, tolerant); 1764static SYSDEV_INT_ATTR(tolerant, 0644, tolerant);
1695static SYSDEV_INT_ATTR(monarch_timeout, 0644, monarch_timeout); 1765static SYSDEV_INT_ATTR(monarch_timeout, 0644, monarch_timeout);
1766static SYSDEV_INT_ATTR(dont_log_ce, 0644, mce_dont_log_ce);
1696 1767
1697static struct sysdev_ext_attribute attr_check_interval = { 1768static struct sysdev_ext_attribute attr_check_interval = {
1698 _SYSDEV_ATTR(check_interval, 0644, sysdev_show_int, 1769 _SYSDEV_ATTR(check_interval, 0644, sysdev_show_int,
@@ -1700,9 +1771,24 @@ static struct sysdev_ext_attribute attr_check_interval = {
1700 &check_interval 1771 &check_interval
1701}; 1772};
1702 1773
1774static struct sysdev_ext_attribute attr_ignore_ce = {
1775 _SYSDEV_ATTR(ignore_ce, 0644, sysdev_show_int, set_ignore_ce),
1776 &mce_ignore_ce
1777};
1778
1779static struct sysdev_ext_attribute attr_cmci_disabled = {
1780 _SYSDEV_ATTR(cmci_disabled, 0644, sysdev_show_int, set_cmci_disabled),
1781 &mce_cmci_disabled
1782};
1783
1703static struct sysdev_attribute *mce_attrs[] = { 1784static struct sysdev_attribute *mce_attrs[] = {
1704 &attr_tolerant.attr, &attr_check_interval.attr, &attr_trigger, 1785 &attr_tolerant.attr,
1786 &attr_check_interval.attr,
1787 &attr_trigger,
1705 &attr_monarch_timeout.attr, 1788 &attr_monarch_timeout.attr,
1789 &attr_dont_log_ce.attr,
1790 &attr_ignore_ce.attr,
1791 &attr_cmci_disabled.attr,
1706 NULL 1792 NULL
1707}; 1793};
1708 1794
@@ -1712,7 +1798,7 @@ static cpumask_var_t mce_dev_initialized;
1712static __cpuinit int mce_create_device(unsigned int cpu) 1798static __cpuinit int mce_create_device(unsigned int cpu)
1713{ 1799{
1714 int err; 1800 int err;
1715 int i; 1801 int i, j;
1716 1802
1717 if (!mce_available(&boot_cpu_data)) 1803 if (!mce_available(&boot_cpu_data))
1718 return -EIO; 1804 return -EIO;
@@ -1730,9 +1816,9 @@ static __cpuinit int mce_create_device(unsigned int cpu)
1730 if (err) 1816 if (err)
1731 goto error; 1817 goto error;
1732 } 1818 }
1733 for (i = 0; i < banks; i++) { 1819 for (j = 0; j < banks; j++) {
1734 err = sysdev_create_file(&per_cpu(mce_dev, cpu), 1820 err = sysdev_create_file(&per_cpu(mce_dev, cpu),
1735 &bank_attrs[i]); 1821 &bank_attrs[j]);
1736 if (err) 1822 if (err)
1737 goto error2; 1823 goto error2;
1738 } 1824 }
@@ -1740,8 +1826,8 @@ static __cpuinit int mce_create_device(unsigned int cpu)
1740 1826
1741 return 0; 1827 return 0;
1742error2: 1828error2:
1743 while (--i >= 0) 1829 while (--j >= 0)
1744 sysdev_remove_file(&per_cpu(mce_dev, cpu), &bank_attrs[i]); 1830 sysdev_remove_file(&per_cpu(mce_dev, cpu), &bank_attrs[j]);
1745error: 1831error:
1746 while (--i >= 0) 1832 while (--i >= 0)
1747 sysdev_remove_file(&per_cpu(mce_dev, cpu), mce_attrs[i]); 1833 sysdev_remove_file(&per_cpu(mce_dev, cpu), mce_attrs[i]);
@@ -1883,7 +1969,7 @@ static __init int mce_init_device(void)
1883 if (!mce_available(&boot_cpu_data)) 1969 if (!mce_available(&boot_cpu_data))
1884 return -EIO; 1970 return -EIO;
1885 1971
1886 alloc_cpumask_var(&mce_dev_initialized, GFP_KERNEL); 1972 zalloc_cpumask_var(&mce_dev_initialized, GFP_KERNEL);
1887 1973
1888 err = mce_init_banks(); 1974 err = mce_init_banks();
1889 if (err) 1975 if (err)
@@ -1915,7 +2001,7 @@ EXPORT_SYMBOL_GPL(nr_mce_banks); /* non-fatal.o */
1915/* This has to be run for each processor */ 2001/* This has to be run for each processor */
1916void mcheck_init(struct cpuinfo_x86 *c) 2002void mcheck_init(struct cpuinfo_x86 *c)
1917{ 2003{
1918 if (mce_disabled == 1) 2004 if (mce_disabled)
1919 return; 2005 return;
1920 2006
1921 switch (c->x86_vendor) { 2007 switch (c->x86_vendor) {
@@ -1945,10 +2031,9 @@ void mcheck_init(struct cpuinfo_x86 *c)
1945 2031
1946static int __init mcheck_enable(char *str) 2032static int __init mcheck_enable(char *str)
1947{ 2033{
1948 mce_disabled = -1; 2034 mce_p5_enabled = 1;
1949 return 1; 2035 return 1;
1950} 2036}
1951
1952__setup("mce", mcheck_enable); 2037__setup("mce", mcheck_enable);
1953 2038
1954#endif /* CONFIG_X86_OLD_MCE */ 2039#endif /* CONFIG_X86_OLD_MCE */
diff --git a/arch/x86/kernel/cpu/mcheck/mce.h b/arch/x86/kernel/cpu/mcheck/mce.h
deleted file mode 100644
index 84a552b458c8..000000000000
--- a/arch/x86/kernel/cpu/mcheck/mce.h
+++ /dev/null
@@ -1,38 +0,0 @@
1#include <linux/init.h>
2#include <asm/mce.h>
3
4#ifdef CONFIG_X86_OLD_MCE
5void amd_mcheck_init(struct cpuinfo_x86 *c);
6void intel_p4_mcheck_init(struct cpuinfo_x86 *c);
7void intel_p6_mcheck_init(struct cpuinfo_x86 *c);
8#endif
9
10#ifdef CONFIG_X86_ANCIENT_MCE
11void intel_p5_mcheck_init(struct cpuinfo_x86 *c);
12void winchip_mcheck_init(struct cpuinfo_x86 *c);
13extern int mce_p5_enable;
14static inline int mce_p5_enabled(void) { return mce_p5_enable; }
15static inline void enable_p5_mce(void) { mce_p5_enable = 1; }
16#else
17static inline void intel_p5_mcheck_init(struct cpuinfo_x86 *c) {}
18static inline void winchip_mcheck_init(struct cpuinfo_x86 *c) {}
19static inline int mce_p5_enabled(void) { return 0; }
20static inline void enable_p5_mce(void) { }
21#endif
22
23/* Call the installed machine check handler for this CPU setup. */
24extern void (*machine_check_vector)(struct pt_regs *, long error_code);
25
26#ifdef CONFIG_X86_OLD_MCE
27
28extern int nr_mce_banks;
29
30void intel_set_thermal_handler(void);
31
32#else
33
34static inline void intel_set_thermal_handler(void) { }
35
36#endif
37
38void intel_init_thermal(struct cpuinfo_x86 *c);
diff --git a/arch/x86/kernel/cpu/mcheck/mce_amd_64.c b/arch/x86/kernel/cpu/mcheck/mce_amd.c
index ddae21620bda..ddae21620bda 100644
--- a/arch/x86/kernel/cpu/mcheck/mce_amd_64.c
+++ b/arch/x86/kernel/cpu/mcheck/mce_amd.c
diff --git a/arch/x86/kernel/cpu/mcheck/mce_intel.c b/arch/x86/kernel/cpu/mcheck/mce_intel.c
index 2b011d2d8579..e1acec0f7a32 100644
--- a/arch/x86/kernel/cpu/mcheck/mce_intel.c
+++ b/arch/x86/kernel/cpu/mcheck/mce_intel.c
@@ -1,74 +1,226 @@
1/* 1/*
2 * Common code for Intel machine checks 2 * Intel specific MCE features.
3 * Copyright 2004 Zwane Mwaikambo <zwane@linuxpower.ca>
4 * Copyright (C) 2008, 2009 Intel Corporation
5 * Author: Andi Kleen
3 */ 6 */
4#include <linux/interrupt.h>
5#include <linux/kernel.h>
6#include <linux/types.h>
7#include <linux/init.h>
8#include <linux/smp.h>
9 7
10#include <asm/therm_throt.h> 8#include <linux/init.h>
11#include <asm/processor.h> 9#include <linux/interrupt.h>
12#include <asm/system.h> 10#include <linux/percpu.h>
13#include <asm/apic.h> 11#include <asm/apic.h>
12#include <asm/processor.h>
14#include <asm/msr.h> 13#include <asm/msr.h>
14#include <asm/mce.h>
15
16/*
17 * Support for Intel Correct Machine Check Interrupts. This allows
18 * the CPU to raise an interrupt when a corrected machine check happened.
19 * Normally we pick those up using a regular polling timer.
20 * Also supports reliable discovery of shared banks.
21 */
15 22
16#include "mce.h" 23static DEFINE_PER_CPU(mce_banks_t, mce_banks_owned);
17 24
18void intel_init_thermal(struct cpuinfo_x86 *c) 25/*
26 * cmci_discover_lock protects against parallel discovery attempts
27 * which could race against each other.
28 */
29static DEFINE_SPINLOCK(cmci_discover_lock);
30
31#define CMCI_THRESHOLD 1
32
33static int cmci_supported(int *banks)
19{ 34{
20 unsigned int cpu = smp_processor_id(); 35 u64 cap;
21 int tm2 = 0;
22 u32 l, h;
23 36
24 /* Thermal monitoring depends on ACPI and clock modulation*/ 37 if (mce_cmci_disabled || mce_ignore_ce)
25 if (!cpu_has(c, X86_FEATURE_ACPI) || !cpu_has(c, X86_FEATURE_ACC)) 38 return 0;
26 return;
27 39
28 /* 40 /*
29 * First check if its enabled already, in which case there might 41 * Vendor check is not strictly needed, but the initial
30 * be some SMM goo which handles it, so we can't even put a handler 42 * initialization is vendor keyed and this
31 * since it might be delivered via SMI already: 43 * makes sure none of the backdoors are entered otherwise.
32 */ 44 */
33 rdmsr(MSR_IA32_MISC_ENABLE, l, h); 45 if (boot_cpu_data.x86_vendor != X86_VENDOR_INTEL)
34 h = apic_read(APIC_LVTTHMR); 46 return 0;
35 if ((l & MSR_IA32_MISC_ENABLE_TM1) && (h & APIC_DM_SMI)) { 47 if (!cpu_has_apic || lapic_get_maxlvt() < 6)
36 printk(KERN_DEBUG 48 return 0;
37 "CPU%d: Thermal monitoring handled by SMI\n", cpu); 49 rdmsrl(MSR_IA32_MCG_CAP, cap);
38 return; 50 *banks = min_t(unsigned, MAX_NR_BANKS, cap & 0xff);
51 return !!(cap & MCG_CMCI_P);
52}
53
54/*
55 * The interrupt handler. This is called on every event.
56 * Just call the poller directly to log any events.
57 * This could in theory increase the threshold under high load,
58 * but doesn't for now.
59 */
60static void intel_threshold_interrupt(void)
61{
62 machine_check_poll(MCP_TIMESTAMP, &__get_cpu_var(mce_banks_owned));
63 mce_notify_irq();
64}
65
66static void print_update(char *type, int *hdr, int num)
67{
68 if (*hdr == 0)
69 printk(KERN_INFO "CPU %d MCA banks", smp_processor_id());
70 *hdr = 1;
71 printk(KERN_CONT " %s:%d", type, num);
72}
73
74/*
75 * Enable CMCI (Corrected Machine Check Interrupt) for available MCE banks
76 * on this CPU. Use the algorithm recommended in the SDM to discover shared
77 * banks.
78 */
79static void cmci_discover(int banks, int boot)
80{
81 unsigned long *owned = (void *)&__get_cpu_var(mce_banks_owned);
82 unsigned long flags;
83 int hdr = 0;
84 int i;
85
86 spin_lock_irqsave(&cmci_discover_lock, flags);
87 for (i = 0; i < banks; i++) {
88 u64 val;
89
90 if (test_bit(i, owned))
91 continue;
92
93 rdmsrl(MSR_IA32_MC0_CTL2 + i, val);
94
95 /* Already owned by someone else? */
96 if (val & CMCI_EN) {
97 if (test_and_clear_bit(i, owned) || boot)
98 print_update("SHD", &hdr, i);
99 __clear_bit(i, __get_cpu_var(mce_poll_banks));
100 continue;
101 }
102
103 val |= CMCI_EN | CMCI_THRESHOLD;
104 wrmsrl(MSR_IA32_MC0_CTL2 + i, val);
105 rdmsrl(MSR_IA32_MC0_CTL2 + i, val);
106
107 /* Did the enable bit stick? -- the bank supports CMCI */
108 if (val & CMCI_EN) {
109 if (!test_and_set_bit(i, owned) || boot)
110 print_update("CMCI", &hdr, i);
111 __clear_bit(i, __get_cpu_var(mce_poll_banks));
112 } else {
113 WARN_ON(!test_bit(i, __get_cpu_var(mce_poll_banks)));
114 }
39 } 115 }
116 spin_unlock_irqrestore(&cmci_discover_lock, flags);
117 if (hdr)
118 printk(KERN_CONT "\n");
119}
120
121/*
122 * Just in case we missed an event during initialization check
123 * all the CMCI owned banks.
124 */
125void cmci_recheck(void)
126{
127 unsigned long flags;
128 int banks;
129
130 if (!mce_available(&current_cpu_data) || !cmci_supported(&banks))
131 return;
132 local_irq_save(flags);
133 machine_check_poll(MCP_TIMESTAMP, &__get_cpu_var(mce_banks_owned));
134 local_irq_restore(flags);
135}
40 136
41 if (cpu_has(c, X86_FEATURE_TM2) && (l & MSR_IA32_MISC_ENABLE_TM2)) 137/*
42 tm2 = 1; 138 * Disable CMCI on this CPU for all banks it owns when it goes down.
139 * This allows other CPUs to claim the banks on rediscovery.
140 */
141void cmci_clear(void)
142{
143 unsigned long flags;
144 int i;
145 int banks;
146 u64 val;
43 147
44 /* Check whether a vector already exists */ 148 if (!cmci_supported(&banks))
45 if (h & APIC_VECTOR_MASK) {
46 printk(KERN_DEBUG
47 "CPU%d: Thermal LVT vector (%#x) already installed\n",
48 cpu, (h & APIC_VECTOR_MASK));
49 return; 149 return;
150 spin_lock_irqsave(&cmci_discover_lock, flags);
151 for (i = 0; i < banks; i++) {
152 if (!test_bit(i, __get_cpu_var(mce_banks_owned)))
153 continue;
154 /* Disable CMCI */
155 rdmsrl(MSR_IA32_MC0_CTL2 + i, val);
156 val &= ~(CMCI_EN|CMCI_THRESHOLD_MASK);
157 wrmsrl(MSR_IA32_MC0_CTL2 + i, val);
158 __clear_bit(i, __get_cpu_var(mce_banks_owned));
50 } 159 }
160 spin_unlock_irqrestore(&cmci_discover_lock, flags);
161}
162
163/*
164 * After a CPU went down cycle through all the others and rediscover
165 * Must run in process context.
166 */
167void cmci_rediscover(int dying)
168{
169 int banks;
170 int cpu;
171 cpumask_var_t old;
172
173 if (!cmci_supported(&banks))
174 return;
175 if (!alloc_cpumask_var(&old, GFP_KERNEL))
176 return;
177 cpumask_copy(old, &current->cpus_allowed);
51 178
52 /* We'll mask the thermal vector in the lapic till we're ready: */ 179 for_each_online_cpu(cpu) {
53 h = THERMAL_APIC_VECTOR | APIC_DM_FIXED | APIC_LVT_MASKED; 180 if (cpu == dying)
54 apic_write(APIC_LVTTHMR, h); 181 continue;
182 if (set_cpus_allowed_ptr(current, cpumask_of(cpu)))
183 continue;
184 /* Recheck banks in case CPUs don't all have the same */
185 if (cmci_supported(&banks))
186 cmci_discover(banks, 0);
187 }
55 188
56 rdmsr(MSR_IA32_THERM_INTERRUPT, l, h); 189 set_cpus_allowed_ptr(current, old);
57 wrmsr(MSR_IA32_THERM_INTERRUPT, 190 free_cpumask_var(old);
58 l | (THERM_INT_LOW_ENABLE | THERM_INT_HIGH_ENABLE), h); 191}
59 192
60 intel_set_thermal_handler(); 193/*
194 * Reenable CMCI on this CPU in case a CPU down failed.
195 */
196void cmci_reenable(void)
197{
198 int banks;
199 if (cmci_supported(&banks))
200 cmci_discover(banks, 0);
201}
61 202
62 rdmsr(MSR_IA32_MISC_ENABLE, l, h); 203static void intel_init_cmci(void)
63 wrmsr(MSR_IA32_MISC_ENABLE, l | MSR_IA32_MISC_ENABLE_TM1, h); 204{
205 int banks;
64 206
65 /* Unmask the thermal vector: */ 207 if (!cmci_supported(&banks))
66 l = apic_read(APIC_LVTTHMR); 208 return;
67 apic_write(APIC_LVTTHMR, l & ~APIC_LVT_MASKED);
68 209
69 printk(KERN_INFO "CPU%d: Thermal monitoring enabled (%s)\n", 210 mce_threshold_vector = intel_threshold_interrupt;
70 cpu, tm2 ? "TM2" : "TM1"); 211 cmci_discover(banks, 1);
212 /*
213 * For CPU #0 this runs with still disabled APIC, but that's
214 * ok because only the vector is set up. We still do another
215 * check for the banks later for CPU #0 just to make sure
216 * to not miss any events.
217 */
218 apic_write(APIC_LVTCMCI, THRESHOLD_APIC_VECTOR|APIC_DM_FIXED);
219 cmci_recheck();
220}
71 221
72 /* enable thermal throttle processing */ 222void mce_intel_feature_init(struct cpuinfo_x86 *c)
73 atomic_set(&therm_throt_en, 1); 223{
224 intel_init_thermal(c);
225 intel_init_cmci();
74} 226}
diff --git a/arch/x86/kernel/cpu/mcheck/mce_intel_64.c b/arch/x86/kernel/cpu/mcheck/mce_intel_64.c
deleted file mode 100644
index f2ef6952c400..000000000000
--- a/arch/x86/kernel/cpu/mcheck/mce_intel_64.c
+++ /dev/null
@@ -1,248 +0,0 @@
1/*
2 * Intel specific MCE features.
3 * Copyright 2004 Zwane Mwaikambo <zwane@linuxpower.ca>
4 * Copyright (C) 2008, 2009 Intel Corporation
5 * Author: Andi Kleen
6 */
7
8#include <linux/init.h>
9#include <linux/interrupt.h>
10#include <linux/percpu.h>
11#include <asm/processor.h>
12#include <asm/apic.h>
13#include <asm/msr.h>
14#include <asm/mce.h>
15#include <asm/hw_irq.h>
16#include <asm/idle.h>
17#include <asm/therm_throt.h>
18
19#include "mce.h"
20
21asmlinkage void smp_thermal_interrupt(void)
22{
23 __u64 msr_val;
24
25 ack_APIC_irq();
26
27 exit_idle();
28 irq_enter();
29
30 rdmsrl(MSR_IA32_THERM_STATUS, msr_val);
31 if (therm_throt_process(msr_val & THERM_STATUS_PROCHOT))
32 mce_log_therm_throt_event(msr_val);
33
34 inc_irq_stat(irq_thermal_count);
35 irq_exit();
36}
37
38/*
39 * Support for Intel Correct Machine Check Interrupts. This allows
40 * the CPU to raise an interrupt when a corrected machine check happened.
41 * Normally we pick those up using a regular polling timer.
42 * Also supports reliable discovery of shared banks.
43 */
44
45static DEFINE_PER_CPU(mce_banks_t, mce_banks_owned);
46
47/*
48 * cmci_discover_lock protects against parallel discovery attempts
49 * which could race against each other.
50 */
51static DEFINE_SPINLOCK(cmci_discover_lock);
52
53#define CMCI_THRESHOLD 1
54
55static int cmci_supported(int *banks)
56{
57 u64 cap;
58
59 if (mce_cmci_disabled || mce_ignore_ce)
60 return 0;
61
62 /*
63 * Vendor check is not strictly needed, but the initial
64 * initialization is vendor keyed and this
65 * makes sure none of the backdoors are entered otherwise.
66 */
67 if (boot_cpu_data.x86_vendor != X86_VENDOR_INTEL)
68 return 0;
69 if (!cpu_has_apic || lapic_get_maxlvt() < 6)
70 return 0;
71 rdmsrl(MSR_IA32_MCG_CAP, cap);
72 *banks = min_t(unsigned, MAX_NR_BANKS, cap & 0xff);
73 return !!(cap & MCG_CMCI_P);
74}
75
76/*
77 * The interrupt handler. This is called on every event.
78 * Just call the poller directly to log any events.
79 * This could in theory increase the threshold under high load,
80 * but doesn't for now.
81 */
82static void intel_threshold_interrupt(void)
83{
84 machine_check_poll(MCP_TIMESTAMP, &__get_cpu_var(mce_banks_owned));
85 mce_notify_irq();
86}
87
88static void print_update(char *type, int *hdr, int num)
89{
90 if (*hdr == 0)
91 printk(KERN_INFO "CPU %d MCA banks", smp_processor_id());
92 *hdr = 1;
93 printk(KERN_CONT " %s:%d", type, num);
94}
95
96/*
97 * Enable CMCI (Corrected Machine Check Interrupt) for available MCE banks
98 * on this CPU. Use the algorithm recommended in the SDM to discover shared
99 * banks.
100 */
101static void cmci_discover(int banks, int boot)
102{
103 unsigned long *owned = (void *)&__get_cpu_var(mce_banks_owned);
104 unsigned long flags;
105 int hdr = 0;
106 int i;
107
108 spin_lock_irqsave(&cmci_discover_lock, flags);
109 for (i = 0; i < banks; i++) {
110 u64 val;
111
112 if (test_bit(i, owned))
113 continue;
114
115 rdmsrl(MSR_IA32_MC0_CTL2 + i, val);
116
117 /* Already owned by someone else? */
118 if (val & CMCI_EN) {
119 if (test_and_clear_bit(i, owned) || boot)
120 print_update("SHD", &hdr, i);
121 __clear_bit(i, __get_cpu_var(mce_poll_banks));
122 continue;
123 }
124
125 val |= CMCI_EN | CMCI_THRESHOLD;
126 wrmsrl(MSR_IA32_MC0_CTL2 + i, val);
127 rdmsrl(MSR_IA32_MC0_CTL2 + i, val);
128
129 /* Did the enable bit stick? -- the bank supports CMCI */
130 if (val & CMCI_EN) {
131 if (!test_and_set_bit(i, owned) || boot)
132 print_update("CMCI", &hdr, i);
133 __clear_bit(i, __get_cpu_var(mce_poll_banks));
134 } else {
135 WARN_ON(!test_bit(i, __get_cpu_var(mce_poll_banks)));
136 }
137 }
138 spin_unlock_irqrestore(&cmci_discover_lock, flags);
139 if (hdr)
140 printk(KERN_CONT "\n");
141}
142
143/*
144 * Just in case we missed an event during initialization check
145 * all the CMCI owned banks.
146 */
147void cmci_recheck(void)
148{
149 unsigned long flags;
150 int banks;
151
152 if (!mce_available(&current_cpu_data) || !cmci_supported(&banks))
153 return;
154 local_irq_save(flags);
155 machine_check_poll(MCP_TIMESTAMP, &__get_cpu_var(mce_banks_owned));
156 local_irq_restore(flags);
157}
158
159/*
160 * Disable CMCI on this CPU for all banks it owns when it goes down.
161 * This allows other CPUs to claim the banks on rediscovery.
162 */
163void cmci_clear(void)
164{
165 unsigned long flags;
166 int i;
167 int banks;
168 u64 val;
169
170 if (!cmci_supported(&banks))
171 return;
172 spin_lock_irqsave(&cmci_discover_lock, flags);
173 for (i = 0; i < banks; i++) {
174 if (!test_bit(i, __get_cpu_var(mce_banks_owned)))
175 continue;
176 /* Disable CMCI */
177 rdmsrl(MSR_IA32_MC0_CTL2 + i, val);
178 val &= ~(CMCI_EN|CMCI_THRESHOLD_MASK);
179 wrmsrl(MSR_IA32_MC0_CTL2 + i, val);
180 __clear_bit(i, __get_cpu_var(mce_banks_owned));
181 }
182 spin_unlock_irqrestore(&cmci_discover_lock, flags);
183}
184
185/*
186 * After a CPU went down cycle through all the others and rediscover
187 * Must run in process context.
188 */
189void cmci_rediscover(int dying)
190{
191 int banks;
192 int cpu;
193 cpumask_var_t old;
194
195 if (!cmci_supported(&banks))
196 return;
197 if (!alloc_cpumask_var(&old, GFP_KERNEL))
198 return;
199 cpumask_copy(old, &current->cpus_allowed);
200
201 for_each_online_cpu(cpu) {
202 if (cpu == dying)
203 continue;
204 if (set_cpus_allowed_ptr(current, cpumask_of(cpu)))
205 continue;
206 /* Recheck banks in case CPUs don't all have the same */
207 if (cmci_supported(&banks))
208 cmci_discover(banks, 0);
209 }
210
211 set_cpus_allowed_ptr(current, old);
212 free_cpumask_var(old);
213}
214
215/*
216 * Reenable CMCI on this CPU in case a CPU down failed.
217 */
218void cmci_reenable(void)
219{
220 int banks;
221 if (cmci_supported(&banks))
222 cmci_discover(banks, 0);
223}
224
225static void intel_init_cmci(void)
226{
227 int banks;
228
229 if (!cmci_supported(&banks))
230 return;
231
232 mce_threshold_vector = intel_threshold_interrupt;
233 cmci_discover(banks, 1);
234 /*
235 * For CPU #0 this runs with still disabled APIC, but that's
236 * ok because only the vector is set up. We still do another
237 * check for the banks later for CPU #0 just to make sure
238 * to not miss any events.
239 */
240 apic_write(APIC_LVTCMCI, THRESHOLD_APIC_VECTOR|APIC_DM_FIXED);
241 cmci_recheck();
242}
243
244void mce_intel_feature_init(struct cpuinfo_x86 *c)
245{
246 intel_init_thermal(c);
247 intel_init_cmci();
248}
diff --git a/arch/x86/kernel/cpu/mcheck/non-fatal.c b/arch/x86/kernel/cpu/mcheck/non-fatal.c
index 70b710420f74..f5f2d6f71fb6 100644
--- a/arch/x86/kernel/cpu/mcheck/non-fatal.c
+++ b/arch/x86/kernel/cpu/mcheck/non-fatal.c
@@ -17,10 +17,9 @@
17 17
18#include <asm/processor.h> 18#include <asm/processor.h>
19#include <asm/system.h> 19#include <asm/system.h>
20#include <asm/mce.h>
20#include <asm/msr.h> 21#include <asm/msr.h>
21 22
22#include "mce.h"
23
24static int firstbank; 23static int firstbank;
25 24
26#define MCE_RATE (15*HZ) /* timer rate is 15s */ 25#define MCE_RATE (15*HZ) /* timer rate is 15s */
diff --git a/arch/x86/kernel/cpu/mcheck/p4.c b/arch/x86/kernel/cpu/mcheck/p4.c
index 82cee108a2d3..4482aea9aa2e 100644
--- a/arch/x86/kernel/cpu/mcheck/p4.c
+++ b/arch/x86/kernel/cpu/mcheck/p4.c
@@ -1,21 +1,15 @@
1/* 1/*
2 * P4 specific Machine Check Exception Reporting 2 * P4 specific Machine Check Exception Reporting
3 */ 3 */
4
5#include <linux/interrupt.h>
6#include <linux/kernel.h> 4#include <linux/kernel.h>
7#include <linux/types.h> 5#include <linux/types.h>
8#include <linux/init.h> 6#include <linux/init.h>
9#include <linux/smp.h> 7#include <linux/smp.h>
10 8
11#include <asm/therm_throt.h>
12#include <asm/processor.h> 9#include <asm/processor.h>
13#include <asm/system.h> 10#include <asm/mce.h>
14#include <asm/apic.h>
15#include <asm/msr.h> 11#include <asm/msr.h>
16 12
17#include "mce.h"
18
19/* as supported by the P4/Xeon family */ 13/* as supported by the P4/Xeon family */
20struct intel_mce_extended_msrs { 14struct intel_mce_extended_msrs {
21 u32 eax; 15 u32 eax;
@@ -33,46 +27,6 @@ struct intel_mce_extended_msrs {
33 27
34static int mce_num_extended_msrs; 28static int mce_num_extended_msrs;
35 29
36
37#ifdef CONFIG_X86_MCE_P4THERMAL
38
39static void unexpected_thermal_interrupt(struct pt_regs *regs)
40{
41 printk(KERN_ERR "CPU%d: Unexpected LVT TMR interrupt!\n",
42 smp_processor_id());
43 add_taint(TAINT_MACHINE_CHECK);
44}
45
46/* P4/Xeon Thermal transition interrupt handler: */
47static void intel_thermal_interrupt(struct pt_regs *regs)
48{
49 __u64 msr_val;
50
51 ack_APIC_irq();
52
53 rdmsrl(MSR_IA32_THERM_STATUS, msr_val);
54 therm_throt_process(msr_val & THERM_STATUS_PROCHOT);
55}
56
57/* Thermal interrupt handler for this CPU setup: */
58static void (*vendor_thermal_interrupt)(struct pt_regs *regs) =
59 unexpected_thermal_interrupt;
60
61void smp_thermal_interrupt(struct pt_regs *regs)
62{
63 irq_enter();
64 vendor_thermal_interrupt(regs);
65 __get_cpu_var(irq_stat).irq_thermal_count++;
66 irq_exit();
67}
68
69void intel_set_thermal_handler(void)
70{
71 vendor_thermal_interrupt = intel_thermal_interrupt;
72}
73
74#endif /* CONFIG_X86_MCE_P4THERMAL */
75
76/* P4/Xeon Extended MCE MSR retrieval, return 0 if unsupported */ 30/* P4/Xeon Extended MCE MSR retrieval, return 0 if unsupported */
77static void intel_get_extended_msrs(struct intel_mce_extended_msrs *r) 31static void intel_get_extended_msrs(struct intel_mce_extended_msrs *r)
78{ 32{
diff --git a/arch/x86/kernel/cpu/mcheck/p5.c b/arch/x86/kernel/cpu/mcheck/p5.c
index 015f481ab1b0..5c0e6533d9bc 100644
--- a/arch/x86/kernel/cpu/mcheck/p5.c
+++ b/arch/x86/kernel/cpu/mcheck/p5.c
@@ -10,12 +10,11 @@
10 10
11#include <asm/processor.h> 11#include <asm/processor.h>
12#include <asm/system.h> 12#include <asm/system.h>
13#include <asm/mce.h>
13#include <asm/msr.h> 14#include <asm/msr.h>
14 15
15#include "mce.h"
16
17/* By default disabled */ 16/* By default disabled */
18int mce_p5_enable; 17int mce_p5_enabled __read_mostly;
19 18
20/* Machine check handler for Pentium class Intel CPUs: */ 19/* Machine check handler for Pentium class Intel CPUs: */
21static void pentium_machine_check(struct pt_regs *regs, long error_code) 20static void pentium_machine_check(struct pt_regs *regs, long error_code)
@@ -43,15 +42,13 @@ void intel_p5_mcheck_init(struct cpuinfo_x86 *c)
43{ 42{
44 u32 l, h; 43 u32 l, h;
45 44
46 /* Check for MCE support: */ 45 /* Default P5 to off as its often misconnected: */
47 if (!cpu_has(c, X86_FEATURE_MCE)) 46 if (!mce_p5_enabled)
48 return; 47 return;
49 48
50#ifdef CONFIG_X86_OLD_MCE 49 /* Check for MCE support: */
51 /* Default P5 to off as its often misconnected: */ 50 if (!cpu_has(c, X86_FEATURE_MCE))
52 if (mce_disabled != -1)
53 return; 51 return;
54#endif
55 52
56 machine_check_vector = pentium_machine_check; 53 machine_check_vector = pentium_machine_check;
57 /* Make sure the vector pointer is visible before we enable MCEs: */ 54 /* Make sure the vector pointer is visible before we enable MCEs: */
diff --git a/arch/x86/kernel/cpu/mcheck/p6.c b/arch/x86/kernel/cpu/mcheck/p6.c
index 43c24e667457..01e4f8178183 100644
--- a/arch/x86/kernel/cpu/mcheck/p6.c
+++ b/arch/x86/kernel/cpu/mcheck/p6.c
@@ -10,10 +10,9 @@
10 10
11#include <asm/processor.h> 11#include <asm/processor.h>
12#include <asm/system.h> 12#include <asm/system.h>
13#include <asm/mce.h>
13#include <asm/msr.h> 14#include <asm/msr.h>
14 15
15#include "mce.h"
16
17/* Machine Check Handler For PII/PIII */ 16/* Machine Check Handler For PII/PIII */
18static void intel_machine_check(struct pt_regs *regs, long error_code) 17static void intel_machine_check(struct pt_regs *regs, long error_code)
19{ 18{
diff --git a/arch/x86/kernel/cpu/mcheck/therm_throt.c b/arch/x86/kernel/cpu/mcheck/therm_throt.c
index 7b1ae2e20ba5..bff8dd191dd5 100644
--- a/arch/x86/kernel/cpu/mcheck/therm_throt.c
+++ b/arch/x86/kernel/cpu/mcheck/therm_throt.c
@@ -13,13 +13,23 @@
13 * Credits: Adapted from Zwane Mwaikambo's original code in mce_intel.c. 13 * Credits: Adapted from Zwane Mwaikambo's original code in mce_intel.c.
14 * Inspired by Ross Biro's and Al Borchers' counter code. 14 * Inspired by Ross Biro's and Al Borchers' counter code.
15 */ 15 */
16#include <linux/interrupt.h>
16#include <linux/notifier.h> 17#include <linux/notifier.h>
17#include <linux/jiffies.h> 18#include <linux/jiffies.h>
19#include <linux/kernel.h>
18#include <linux/percpu.h> 20#include <linux/percpu.h>
19#include <linux/sysdev.h> 21#include <linux/sysdev.h>
22#include <linux/types.h>
23#include <linux/init.h>
24#include <linux/smp.h>
20#include <linux/cpu.h> 25#include <linux/cpu.h>
21 26
22#include <asm/therm_throt.h> 27#include <asm/processor.h>
28#include <asm/system.h>
29#include <asm/apic.h>
30#include <asm/idle.h>
31#include <asm/mce.h>
32#include <asm/msr.h>
23 33
24/* How long to wait between reporting thermal events */ 34/* How long to wait between reporting thermal events */
25#define CHECK_INTERVAL (300 * HZ) 35#define CHECK_INTERVAL (300 * HZ)
@@ -27,7 +37,7 @@
27static DEFINE_PER_CPU(__u64, next_check) = INITIAL_JIFFIES; 37static DEFINE_PER_CPU(__u64, next_check) = INITIAL_JIFFIES;
28static DEFINE_PER_CPU(unsigned long, thermal_throttle_count); 38static DEFINE_PER_CPU(unsigned long, thermal_throttle_count);
29 39
30atomic_t therm_throt_en = ATOMIC_INIT(0); 40static atomic_t therm_throt_en = ATOMIC_INIT(0);
31 41
32#ifdef CONFIG_SYSFS 42#ifdef CONFIG_SYSFS
33#define define_therm_throt_sysdev_one_ro(_name) \ 43#define define_therm_throt_sysdev_one_ro(_name) \
@@ -82,7 +92,7 @@ static struct attribute_group thermal_throttle_attr_group = {
82 * 1 : Event should be logged further, and a message has been 92 * 1 : Event should be logged further, and a message has been
83 * printed to the syslog. 93 * printed to the syslog.
84 */ 94 */
85int therm_throt_process(int curr) 95static int therm_throt_process(int curr)
86{ 96{
87 unsigned int cpu = smp_processor_id(); 97 unsigned int cpu = smp_processor_id();
88 __u64 tmp_jiffs = get_jiffies_64(); 98 __u64 tmp_jiffs = get_jiffies_64();
@@ -186,6 +196,94 @@ static __init int thermal_throttle_init_device(void)
186 196
187 return 0; 197 return 0;
188} 198}
189
190device_initcall(thermal_throttle_init_device); 199device_initcall(thermal_throttle_init_device);
200
191#endif /* CONFIG_SYSFS */ 201#endif /* CONFIG_SYSFS */
202
203/* Thermal transition interrupt handler */
204static void intel_thermal_interrupt(void)
205{
206 __u64 msr_val;
207
208 rdmsrl(MSR_IA32_THERM_STATUS, msr_val);
209 if (therm_throt_process(msr_val & THERM_STATUS_PROCHOT))
210 mce_log_therm_throt_event(msr_val);
211}
212
213static void unexpected_thermal_interrupt(void)
214{
215 printk(KERN_ERR "CPU%d: Unexpected LVT TMR interrupt!\n",
216 smp_processor_id());
217 add_taint(TAINT_MACHINE_CHECK);
218}
219
220static void (*smp_thermal_vector)(void) = unexpected_thermal_interrupt;
221
222asmlinkage void smp_thermal_interrupt(struct pt_regs *regs)
223{
224 exit_idle();
225 irq_enter();
226 inc_irq_stat(irq_thermal_count);
227 smp_thermal_vector();
228 irq_exit();
229 /* Ack only at the end to avoid potential reentry */
230 ack_APIC_irq();
231}
232
233void intel_init_thermal(struct cpuinfo_x86 *c)
234{
235 unsigned int cpu = smp_processor_id();
236 int tm2 = 0;
237 u32 l, h;
238
239 /* Thermal monitoring depends on ACPI and clock modulation*/
240 if (!cpu_has(c, X86_FEATURE_ACPI) || !cpu_has(c, X86_FEATURE_ACC))
241 return;
242
243 /*
244 * First check if its enabled already, in which case there might
245 * be some SMM goo which handles it, so we can't even put a handler
246 * since it might be delivered via SMI already:
247 */
248 rdmsr(MSR_IA32_MISC_ENABLE, l, h);
249 h = apic_read(APIC_LVTTHMR);
250 if ((l & MSR_IA32_MISC_ENABLE_TM1) && (h & APIC_DM_SMI)) {
251 printk(KERN_DEBUG
252 "CPU%d: Thermal monitoring handled by SMI\n", cpu);
253 return;
254 }
255
256 if (cpu_has(c, X86_FEATURE_TM2) && (l & MSR_IA32_MISC_ENABLE_TM2))
257 tm2 = 1;
258
259 /* Check whether a vector already exists */
260 if (h & APIC_VECTOR_MASK) {
261 printk(KERN_DEBUG
262 "CPU%d: Thermal LVT vector (%#x) already installed\n",
263 cpu, (h & APIC_VECTOR_MASK));
264 return;
265 }
266
267 /* We'll mask the thermal vector in the lapic till we're ready: */
268 h = THERMAL_APIC_VECTOR | APIC_DM_FIXED | APIC_LVT_MASKED;
269 apic_write(APIC_LVTTHMR, h);
270
271 rdmsr(MSR_IA32_THERM_INTERRUPT, l, h);
272 wrmsr(MSR_IA32_THERM_INTERRUPT,
273 l | (THERM_INT_LOW_ENABLE | THERM_INT_HIGH_ENABLE), h);
274
275 smp_thermal_vector = intel_thermal_interrupt;
276
277 rdmsr(MSR_IA32_MISC_ENABLE, l, h);
278 wrmsr(MSR_IA32_MISC_ENABLE, l | MSR_IA32_MISC_ENABLE_TM1, h);
279
280 /* Unmask the thermal vector: */
281 l = apic_read(APIC_LVTTHMR);
282 apic_write(APIC_LVTTHMR, l & ~APIC_LVT_MASKED);
283
284 printk(KERN_INFO "CPU%d: Thermal monitoring enabled (%s)\n",
285 cpu, tm2 ? "TM2" : "TM1");
286
287 /* enable thermal throttle processing */
288 atomic_set(&therm_throt_en, 1);
289}
diff --git a/arch/x86/kernel/cpu/mcheck/winchip.c b/arch/x86/kernel/cpu/mcheck/winchip.c
index 81b02487090b..54060f565974 100644
--- a/arch/x86/kernel/cpu/mcheck/winchip.c
+++ b/arch/x86/kernel/cpu/mcheck/winchip.c
@@ -9,10 +9,9 @@
9 9
10#include <asm/processor.h> 10#include <asm/processor.h>
11#include <asm/system.h> 11#include <asm/system.h>
12#include <asm/mce.h>
12#include <asm/msr.h> 13#include <asm/msr.h>
13 14
14#include "mce.h"
15
16/* Machine check handler for WinChip C6: */ 15/* Machine check handler for WinChip C6: */
17static void winchip_machine_check(struct pt_regs *regs, long error_code) 16static void winchip_machine_check(struct pt_regs *regs, long error_code)
18{ 17{
diff --git a/arch/x86/kernel/cpu/perfctr-watchdog.c b/arch/x86/kernel/cpu/perfctr-watchdog.c
index d6f5b9fbde32..5c481f6205bf 100644
--- a/arch/x86/kernel/cpu/perfctr-watchdog.c
+++ b/arch/x86/kernel/cpu/perfctr-watchdog.c
@@ -716,11 +716,15 @@ static void probe_nmi_watchdog(void)
716 wd_ops = &k7_wd_ops; 716 wd_ops = &k7_wd_ops;
717 break; 717 break;
718 case X86_VENDOR_INTEL: 718 case X86_VENDOR_INTEL:
719 /* 719 /* Work around where perfctr1 doesn't have a working enable
720 * Work around Core Duo (Yonah) errata AE49 where perfctr1 720 * bit as described in the following errata:
721 * doesn't have a working enable bit. 721 * AE49 Core Duo and Intel Core Solo 65 nm
722 * AN49 Intel Pentium Dual-Core
723 * AF49 Dual-Core Intel Xeon Processor LV
722 */ 724 */
723 if (boot_cpu_data.x86 == 6 && boot_cpu_data.x86_model == 14) { 725 if ((boot_cpu_data.x86 == 6 && boot_cpu_data.x86_model == 14) ||
726 ((boot_cpu_data.x86 == 6 && boot_cpu_data.x86_model == 15 &&
727 boot_cpu_data.x86_mask == 4))) {
724 intel_arch_wd_ops.perfctr = MSR_ARCH_PERFMON_PERFCTR0; 728 intel_arch_wd_ops.perfctr = MSR_ARCH_PERFMON_PERFCTR0;
725 intel_arch_wd_ops.evntsel = MSR_ARCH_PERFMON_EVENTSEL0; 729 intel_arch_wd_ops.evntsel = MSR_ARCH_PERFMON_EVENTSEL0;
726 } 730 }
diff --git a/arch/x86/kernel/crash.c b/arch/x86/kernel/crash.c
index ff958248e61d..5e409dc298a4 100644
--- a/arch/x86/kernel/crash.c
+++ b/arch/x86/kernel/crash.c
@@ -27,6 +27,7 @@
27#include <asm/cpu.h> 27#include <asm/cpu.h>
28#include <asm/reboot.h> 28#include <asm/reboot.h>
29#include <asm/virtext.h> 29#include <asm/virtext.h>
30#include <asm/iommu.h>
30 31
31 32
32#if defined(CONFIG_SMP) && defined(CONFIG_X86_LOCAL_APIC) 33#if defined(CONFIG_SMP) && defined(CONFIG_X86_LOCAL_APIC)
@@ -103,5 +104,10 @@ void native_machine_crash_shutdown(struct pt_regs *regs)
103#ifdef CONFIG_HPET_TIMER 104#ifdef CONFIG_HPET_TIMER
104 hpet_disable(); 105 hpet_disable();
105#endif 106#endif
107
108#ifdef CONFIG_X86_64
109 pci_iommu_shutdown();
110#endif
111
106 crash_save_cpu(regs, safe_smp_processor_id()); 112 crash_save_cpu(regs, safe_smp_processor_id());
107} 113}
diff --git a/arch/x86/kernel/efi.c b/arch/x86/kernel/efi.c
index 1736acc4d7aa..96f7ac0bbf01 100644
--- a/arch/x86/kernel/efi.c
+++ b/arch/x86/kernel/efi.c
@@ -240,10 +240,35 @@ static void __init do_add_efi_memmap(void)
240 unsigned long long size = md->num_pages << EFI_PAGE_SHIFT; 240 unsigned long long size = md->num_pages << EFI_PAGE_SHIFT;
241 int e820_type; 241 int e820_type;
242 242
243 if (md->attribute & EFI_MEMORY_WB) 243 switch (md->type) {
244 e820_type = E820_RAM; 244 case EFI_LOADER_CODE:
245 else 245 case EFI_LOADER_DATA:
246 case EFI_BOOT_SERVICES_CODE:
247 case EFI_BOOT_SERVICES_DATA:
248 case EFI_CONVENTIONAL_MEMORY:
249 if (md->attribute & EFI_MEMORY_WB)
250 e820_type = E820_RAM;
251 else
252 e820_type = E820_RESERVED;
253 break;
254 case EFI_ACPI_RECLAIM_MEMORY:
255 e820_type = E820_ACPI;
256 break;
257 case EFI_ACPI_MEMORY_NVS:
258 e820_type = E820_NVS;
259 break;
260 case EFI_UNUSABLE_MEMORY:
261 e820_type = E820_UNUSABLE;
262 break;
263 default:
264 /*
265 * EFI_RESERVED_TYPE EFI_RUNTIME_SERVICES_CODE
266 * EFI_RUNTIME_SERVICES_DATA EFI_MEMORY_MAPPED_IO
267 * EFI_MEMORY_MAPPED_IO_PORT_SPACE EFI_PAL_CODE
268 */
246 e820_type = E820_RESERVED; 269 e820_type = E820_RESERVED;
270 break;
271 }
247 e820_add_region(start, size, e820_type); 272 e820_add_region(start, size, e820_type);
248 } 273 }
249 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map); 274 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
diff --git a/arch/x86/kernel/entry_32.S b/arch/x86/kernel/entry_32.S
index c929add475c9..9f8ce77dbc64 100644
--- a/arch/x86/kernel/entry_32.S
+++ b/arch/x86/kernel/entry_32.S
@@ -48,7 +48,6 @@
48#include <asm/segment.h> 48#include <asm/segment.h>
49#include <asm/smp.h> 49#include <asm/smp.h>
50#include <asm/page_types.h> 50#include <asm/page_types.h>
51#include <asm/desc.h>
52#include <asm/percpu.h> 51#include <asm/percpu.h>
53#include <asm/dwarf2.h> 52#include <asm/dwarf2.h>
54#include <asm/processor-flags.h> 53#include <asm/processor-flags.h>
@@ -84,7 +83,7 @@
84#define preempt_stop(clobbers) DISABLE_INTERRUPTS(clobbers); TRACE_IRQS_OFF 83#define preempt_stop(clobbers) DISABLE_INTERRUPTS(clobbers); TRACE_IRQS_OFF
85#else 84#else
86#define preempt_stop(clobbers) 85#define preempt_stop(clobbers)
87#define resume_kernel restore_nocheck 86#define resume_kernel restore_all
88#endif 87#endif
89 88
90.macro TRACE_IRQS_IRET 89.macro TRACE_IRQS_IRET
@@ -372,7 +371,7 @@ END(ret_from_exception)
372ENTRY(resume_kernel) 371ENTRY(resume_kernel)
373 DISABLE_INTERRUPTS(CLBR_ANY) 372 DISABLE_INTERRUPTS(CLBR_ANY)
374 cmpl $0,TI_preempt_count(%ebp) # non-zero preempt_count ? 373 cmpl $0,TI_preempt_count(%ebp) # non-zero preempt_count ?
375 jnz restore_nocheck 374 jnz restore_all
376need_resched: 375need_resched:
377 movl TI_flags(%ebp), %ecx # need_resched set ? 376 movl TI_flags(%ebp), %ecx # need_resched set ?
378 testb $_TIF_NEED_RESCHED, %cl 377 testb $_TIF_NEED_RESCHED, %cl
@@ -540,6 +539,8 @@ syscall_exit:
540 jne syscall_exit_work 539 jne syscall_exit_work
541 540
542restore_all: 541restore_all:
542 TRACE_IRQS_IRET
543restore_all_notrace:
543 movl PT_EFLAGS(%esp), %eax # mix EFLAGS, SS and CS 544 movl PT_EFLAGS(%esp), %eax # mix EFLAGS, SS and CS
544 # Warning: PT_OLDSS(%esp) contains the wrong/random values if we 545 # Warning: PT_OLDSS(%esp) contains the wrong/random values if we
545 # are returning to the kernel. 546 # are returning to the kernel.
@@ -551,8 +552,6 @@ restore_all:
551 CFI_REMEMBER_STATE 552 CFI_REMEMBER_STATE
552 je ldt_ss # returning to user-space with LDT SS 553 je ldt_ss # returning to user-space with LDT SS
553restore_nocheck: 554restore_nocheck:
554 TRACE_IRQS_IRET
555restore_nocheck_notrace:
556 RESTORE_REGS 4 # skip orig_eax/error_code 555 RESTORE_REGS 4 # skip orig_eax/error_code
557 CFI_ADJUST_CFA_OFFSET -4 556 CFI_ADJUST_CFA_OFFSET -4
558irq_return: 557irq_return:
@@ -588,22 +587,34 @@ ldt_ss:
588 jne restore_nocheck 587 jne restore_nocheck
589#endif 588#endif
590 589
591 /* If returning to userspace with 16bit stack, 590/*
592 * try to fix the higher word of ESP, as the CPU 591 * Setup and switch to ESPFIX stack
593 * won't restore it. 592 *
594 * This is an "official" bug of all the x86-compatible 593 * We're returning to userspace with a 16 bit stack. The CPU will not
595 * CPUs, which we can try to work around to make 594 * restore the high word of ESP for us on executing iret... This is an
596 * dosemu and wine happy. */ 595 * "official" bug of all the x86-compatible CPUs, which we can work
597 movl PT_OLDESP(%esp), %eax 596 * around to make dosemu and wine happy. We do this by preloading the
598 movl %esp, %edx 597 * high word of ESP with the high word of the userspace ESP while
599 call patch_espfix_desc 598 * compensating for the offset by changing to the ESPFIX segment with
599 * a base address that matches for the difference.
600 */
601 mov %esp, %edx /* load kernel esp */
602 mov PT_OLDESP(%esp), %eax /* load userspace esp */
603 mov %dx, %ax /* eax: new kernel esp */
604 sub %eax, %edx /* offset (low word is 0) */
605 PER_CPU(gdt_page, %ebx)
606 shr $16, %edx
607 mov %dl, GDT_ENTRY_ESPFIX_SS * 8 + 4(%ebx) /* bits 16..23 */
608 mov %dh, GDT_ENTRY_ESPFIX_SS * 8 + 7(%ebx) /* bits 24..31 */
600 pushl $__ESPFIX_SS 609 pushl $__ESPFIX_SS
601 CFI_ADJUST_CFA_OFFSET 4 610 CFI_ADJUST_CFA_OFFSET 4
602 pushl %eax 611 push %eax /* new kernel esp */
603 CFI_ADJUST_CFA_OFFSET 4 612 CFI_ADJUST_CFA_OFFSET 4
613 /* Disable interrupts, but do not irqtrace this section: we
614 * will soon execute iret and the tracer was already set to
615 * the irqstate after the iret */
604 DISABLE_INTERRUPTS(CLBR_EAX) 616 DISABLE_INTERRUPTS(CLBR_EAX)
605 TRACE_IRQS_OFF 617 lss (%esp), %esp /* switch to espfix segment */
606 lss (%esp), %esp
607 CFI_ADJUST_CFA_OFFSET -8 618 CFI_ADJUST_CFA_OFFSET -8
608 jmp restore_nocheck 619 jmp restore_nocheck
609 CFI_ENDPROC 620 CFI_ENDPROC
@@ -716,15 +727,24 @@ PTREGSCALL(vm86)
716PTREGSCALL(vm86old) 727PTREGSCALL(vm86old)
717 728
718.macro FIXUP_ESPFIX_STACK 729.macro FIXUP_ESPFIX_STACK
719 /* since we are on a wrong stack, we cant make it a C code :( */ 730/*
731 * Switch back for ESPFIX stack to the normal zerobased stack
732 *
733 * We can't call C functions using the ESPFIX stack. This code reads
734 * the high word of the segment base from the GDT and swiches to the
735 * normal stack and adjusts ESP with the matching offset.
736 */
737 /* fixup the stack */
720 PER_CPU(gdt_page, %ebx) 738 PER_CPU(gdt_page, %ebx)
721 GET_DESC_BASE(GDT_ENTRY_ESPFIX_SS, %ebx, %eax, %ax, %al, %ah) 739 mov GDT_ENTRY_ESPFIX_SS * 8 + 4(%ebx), %al /* bits 16..23 */
722 addl %esp, %eax 740 mov GDT_ENTRY_ESPFIX_SS * 8 + 7(%ebx), %ah /* bits 24..31 */
741 shl $16, %eax
742 addl %esp, %eax /* the adjusted stack pointer */
723 pushl $__KERNEL_DS 743 pushl $__KERNEL_DS
724 CFI_ADJUST_CFA_OFFSET 4 744 CFI_ADJUST_CFA_OFFSET 4
725 pushl %eax 745 pushl %eax
726 CFI_ADJUST_CFA_OFFSET 4 746 CFI_ADJUST_CFA_OFFSET 4
727 lss (%esp), %esp 747 lss (%esp), %esp /* switch to the normal stack segment */
728 CFI_ADJUST_CFA_OFFSET -8 748 CFI_ADJUST_CFA_OFFSET -8
729.endm 749.endm
730.macro UNWIND_ESPFIX_STACK 750.macro UNWIND_ESPFIX_STACK
@@ -1329,7 +1349,7 @@ nmi_stack_correct:
1329 xorl %edx,%edx # zero error code 1349 xorl %edx,%edx # zero error code
1330 movl %esp,%eax # pt_regs pointer 1350 movl %esp,%eax # pt_regs pointer
1331 call do_nmi 1351 call do_nmi
1332 jmp restore_nocheck_notrace 1352 jmp restore_all_notrace
1333 CFI_ENDPROC 1353 CFI_ENDPROC
1334 1354
1335nmi_stack_fixup: 1355nmi_stack_fixup:
diff --git a/arch/x86/kernel/head_32.S b/arch/x86/kernel/head_32.S
index dc5ed4bdd88d..8663afb56535 100644
--- a/arch/x86/kernel/head_32.S
+++ b/arch/x86/kernel/head_32.S
@@ -13,7 +13,6 @@
13#include <asm/segment.h> 13#include <asm/segment.h>
14#include <asm/page_types.h> 14#include <asm/page_types.h>
15#include <asm/pgtable_types.h> 15#include <asm/pgtable_types.h>
16#include <asm/desc.h>
17#include <asm/cache.h> 16#include <asm/cache.h>
18#include <asm/thread_info.h> 17#include <asm/thread_info.h>
19#include <asm/asm-offsets.h> 18#include <asm/asm-offsets.h>
diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S
index 54b29bb24e71..fa54f78e2a05 100644
--- a/arch/x86/kernel/head_64.S
+++ b/arch/x86/kernel/head_64.S
@@ -12,7 +12,6 @@
12#include <linux/linkage.h> 12#include <linux/linkage.h>
13#include <linux/threads.h> 13#include <linux/threads.h>
14#include <linux/init.h> 14#include <linux/init.h>
15#include <asm/desc.h>
16#include <asm/segment.h> 15#include <asm/segment.h>
17#include <asm/pgtable.h> 16#include <asm/pgtable.h>
18#include <asm/page.h> 17#include <asm/page.h>
diff --git a/arch/x86/kernel/hpet.c b/arch/x86/kernel/hpet.c
index 81408b93f887..dedc2bddf7a5 100644
--- a/arch/x86/kernel/hpet.c
+++ b/arch/x86/kernel/hpet.c
@@ -510,7 +510,8 @@ static int hpet_setup_irq(struct hpet_dev *dev)
510{ 510{
511 511
512 if (request_irq(dev->irq, hpet_interrupt_handler, 512 if (request_irq(dev->irq, hpet_interrupt_handler,
513 IRQF_DISABLED|IRQF_NOBALANCING, dev->name, dev)) 513 IRQF_TIMER | IRQF_DISABLED | IRQF_NOBALANCING,
514 dev->name, dev))
514 return -1; 515 return -1;
515 516
516 disable_irq(dev->irq); 517 disable_irq(dev->irq);
diff --git a/arch/x86/kernel/pci-dma.c b/arch/x86/kernel/pci-dma.c
index 745579bc8256..328592fb6044 100644
--- a/arch/x86/kernel/pci-dma.c
+++ b/arch/x86/kernel/pci-dma.c
@@ -290,6 +290,8 @@ static int __init pci_iommu_init(void)
290void pci_iommu_shutdown(void) 290void pci_iommu_shutdown(void)
291{ 291{
292 gart_iommu_shutdown(); 292 gart_iommu_shutdown();
293
294 amd_iommu_shutdown();
293} 295}
294/* Must execute after PCI subsystem */ 296/* Must execute after PCI subsystem */
295fs_initcall(pci_iommu_init); 297fs_initcall(pci_iommu_init);
diff --git a/arch/x86/kernel/traps.c b/arch/x86/kernel/traps.c
index 5f935f0d5861..a0f48f5671c0 100644
--- a/arch/x86/kernel/traps.c
+++ b/arch/x86/kernel/traps.c
@@ -54,6 +54,7 @@
54#include <asm/traps.h> 54#include <asm/traps.h>
55#include <asm/desc.h> 55#include <asm/desc.h>
56#include <asm/i387.h> 56#include <asm/i387.h>
57#include <asm/mce.h>
57 58
58#include <asm/mach_traps.h> 59#include <asm/mach_traps.h>
59 60
@@ -65,8 +66,6 @@
65#include <asm/setup.h> 66#include <asm/setup.h>
66#include <asm/traps.h> 67#include <asm/traps.h>
67 68
68#include "cpu/mcheck/mce.h"
69
70asmlinkage int system_call(void); 69asmlinkage int system_call(void);
71 70
72/* Do we ignore FPU interrupts ? */ 71/* Do we ignore FPU interrupts ? */
diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c
index baa0e86adfbc..c403526d5d15 100644
--- a/arch/x86/mm/fault.c
+++ b/arch/x86/mm/fault.c
@@ -952,8 +952,6 @@ do_page_fault(struct pt_regs *regs, unsigned long error_code)
952 tsk = current; 952 tsk = current;
953 mm = tsk->mm; 953 mm = tsk->mm;
954 954
955 prefetchw(&mm->mmap_sem);
956
957 /* Get the faulting address: */ 955 /* Get the faulting address: */
958 address = read_cr2(); 956 address = read_cr2();
959 957
@@ -963,6 +961,7 @@ do_page_fault(struct pt_regs *regs, unsigned long error_code)
963 */ 961 */
964 if (kmemcheck_active(regs)) 962 if (kmemcheck_active(regs))
965 kmemcheck_hide(regs); 963 kmemcheck_hide(regs);
964 prefetchw(&mm->mmap_sem);
966 965
967 if (unlikely(kmmio_fault(regs, address))) 966 if (unlikely(kmmio_fault(regs, address)))
968 return; 967 return;
diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c
index 9c543290a813..c4378f4fd4a5 100644
--- a/arch/x86/mm/init_64.c
+++ b/arch/x86/mm/init_64.c
@@ -527,7 +527,7 @@ phys_pud_update(pgd_t *pgd, unsigned long addr, unsigned long end,
527 return phys_pud_init(pud, addr, end, page_size_mask); 527 return phys_pud_init(pud, addr, end, page_size_mask);
528} 528}
529 529
530unsigned long __init 530unsigned long __meminit
531kernel_physical_mapping_init(unsigned long start, 531kernel_physical_mapping_init(unsigned long start,
532 unsigned long end, 532 unsigned long end,
533 unsigned long page_size_mask) 533 unsigned long page_size_mask)
generated by cgit v1.2.2 (git 2.25.0) at 2024-10-26 04:29:34 -0400