aboutsummaryrefslogtreecommitdiffstats
path: root/arch/x86
diff options
context:
space:
mode:
Diffstat (limited to 'arch/x86')
-rw-r--r--arch/x86/Kconfig13
-rw-r--r--arch/x86/boot/video-bios.c3
-rw-r--r--arch/x86/boot/video-vesa.c4
-rw-r--r--arch/x86/include/asm/atomic_32.h185
-rw-r--r--arch/x86/include/asm/atomic_64.h42
-rw-r--r--arch/x86/include/asm/boot.h6
-rw-r--r--arch/x86/include/asm/fixmap.h10
-rw-r--r--arch/x86/include/asm/msr-index.h4
-rw-r--r--arch/x86/include/asm/nmi.h1
-rw-r--r--arch/x86/include/asm/pci.h2
-rw-r--r--arch/x86/include/asm/percpu.h10
-rw-r--r--arch/x86/include/asm/perf_counter.h3
-rw-r--r--arch/x86/include/asm/proto.h11
-rw-r--r--arch/x86/include/asm/spinlock.h4
-rw-r--r--arch/x86/include/asm/stacktrace.h2
-rw-r--r--arch/x86/include/asm/thread_info.h2
-rw-r--r--arch/x86/kernel/Makefile2
-rw-r--r--arch/x86/kernel/amd_iommu.c4
-rw-r--r--arch/x86/kernel/amd_iommu_init.c13
-rw-r--r--arch/x86/kernel/apic/apic.c2
-rw-r--r--arch/x86/kernel/apic/io_apic.c30
-rw-r--r--arch/x86/kernel/cpu/amd.c4
-rw-r--r--arch/x86/kernel/cpu/common.c3
-rw-r--r--arch/x86/kernel/cpu/cpufreq/powernow-k8.c32
-rw-r--r--arch/x86/kernel/cpu/cpufreq/powernow-k8.h3
-rw-r--r--arch/x86/kernel/cpu/mcheck/mce.c16
-rw-r--r--arch/x86/kernel/cpu/perf_counter.c30
-rw-r--r--arch/x86/kernel/cpu/perfctr-watchdog.c5
-rw-r--r--arch/x86/kernel/dumpstack.c1
-rw-r--r--arch/x86/kernel/dumpstack_32.c6
-rw-r--r--arch/x86/kernel/dumpstack_64.c22
-rw-r--r--arch/x86/kernel/e820.c23
-rw-r--r--arch/x86/kernel/kvm.c2
-rw-r--r--arch/x86/kernel/pci-dma.c2
-rw-r--r--arch/x86/kernel/pci-gart_64.c2
-rw-r--r--arch/x86/kernel/setup.c16
-rw-r--r--arch/x86/kernel/setup_percpu.c219
-rw-r--r--arch/x86/kernel/tlb_uv.c9
-rw-r--r--arch/x86/kernel/traps.c3
-rw-r--r--arch/x86/kvm/mmu.c6
-rw-r--r--arch/x86/kvm/paging_tmpl.h2
-rw-r--r--arch/x86/kvm/vmx.c15
-rw-r--r--arch/x86/kvm/x86.c1
-rw-r--r--arch/x86/kvm/x86_emulate.c2
-rw-r--r--arch/x86/lib/Makefile1
-rw-r--r--arch/x86/lib/atomic64_32.c230
-rw-r--r--arch/x86/lib/clear_page_64.S5
-rw-r--r--arch/x86/lib/copy_user_64.S1
-rw-r--r--arch/x86/lib/delay.c3
-rw-r--r--arch/x86/lib/usercopy_32.c2
-rw-r--r--arch/x86/mm/fault.c9
-rw-r--r--arch/x86/mm/init.c18
-rw-r--r--arch/x86/mm/init_64.c9
-rw-r--r--arch/x86/mm/pageattr.c65
-rw-r--r--arch/x86/oprofile/nmi_int.c2
-rw-r--r--arch/x86/pci/acpi.c59
-rw-r--r--arch/x86/pci/amd_bus.c8
-rw-r--r--arch/x86/power/Makefile2
-rw-r--r--arch/x86/power/cpu.c2
59 files changed, 774 insertions, 419 deletions
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index d1430ef6b4f9..738bdc6b0f8b 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -1913,25 +1913,26 @@ config DMAR_DEFAULT_ON
1913 recommended you say N here while the DMAR code remains 1913 recommended you say N here while the DMAR code remains
1914 experimental. 1914 experimental.
1915 1915
1916config DMAR_GFX_WA 1916config DMAR_BROKEN_GFX_WA
1917 def_bool y 1917 def_bool n
1918 prompt "Support for Graphics workaround" 1918 prompt "Workaround broken graphics drivers (going away soon)"
1919 depends on DMAR 1919 depends on DMAR
1920 ---help--- 1920 ---help---
1921 Current Graphics drivers tend to use physical address 1921 Current Graphics drivers tend to use physical address
1922 for DMA and avoid using DMA APIs. Setting this config 1922 for DMA and avoid using DMA APIs. Setting this config
1923 option permits the IOMMU driver to set a unity map for 1923 option permits the IOMMU driver to set a unity map for
1924 all the OS-visible memory. Hence the driver can continue 1924 all the OS-visible memory. Hence the driver can continue
1925 to use physical addresses for DMA. 1925 to use physical addresses for DMA, at least until this
1926 option is removed in the 2.6.32 kernel.
1926 1927
1927config DMAR_FLOPPY_WA 1928config DMAR_FLOPPY_WA
1928 def_bool y 1929 def_bool y
1929 depends on DMAR 1930 depends on DMAR
1930 ---help--- 1931 ---help---
1931 Floppy disk drivers are know to bypass DMA API calls 1932 Floppy disk drivers are known to bypass DMA API calls
1932 thereby failing to work when IOMMU is enabled. This 1933 thereby failing to work when IOMMU is enabled. This
1933 workaround will setup a 1:1 mapping for the first 1934 workaround will setup a 1:1 mapping for the first
1934 16M to make floppy (an ISA device) work. 1935 16MiB to make floppy (an ISA device) work.
1935 1936
1936config INTR_REMAP 1937config INTR_REMAP
1937 bool "Support for Interrupt Remapping (EXPERIMENTAL)" 1938 bool "Support for Interrupt Remapping (EXPERIMENTAL)"
diff --git a/arch/x86/boot/video-bios.c b/arch/x86/boot/video-bios.c
index d660be492363..49e0c18833e0 100644
--- a/arch/x86/boot/video-bios.c
+++ b/arch/x86/boot/video-bios.c
@@ -37,14 +37,13 @@ static int set_bios_mode(u8 mode)
37 ireg.al = mode; /* AH=0x00 Set Video Mode */ 37 ireg.al = mode; /* AH=0x00 Set Video Mode */
38 intcall(0x10, &ireg, NULL); 38 intcall(0x10, &ireg, NULL);
39 39
40
41 ireg.ah = 0x0f; /* Get Current Video Mode */ 40 ireg.ah = 0x0f; /* Get Current Video Mode */
42 intcall(0x10, &ireg, &oreg); 41 intcall(0x10, &ireg, &oreg);
43 42
44 do_restore = 1; /* Assume video contents were lost */ 43 do_restore = 1; /* Assume video contents were lost */
45 44
46 /* Not all BIOSes are clean with the top bit */ 45 /* Not all BIOSes are clean with the top bit */
47 new_mode = ireg.al & 0x7f; 46 new_mode = oreg.al & 0x7f;
48 47
49 if (new_mode == mode) 48 if (new_mode == mode)
50 return 0; /* Mode change OK */ 49 return 0; /* Mode change OK */
diff --git a/arch/x86/boot/video-vesa.c b/arch/x86/boot/video-vesa.c
index c700147d6ffb..275dd177f198 100644
--- a/arch/x86/boot/video-vesa.c
+++ b/arch/x86/boot/video-vesa.c
@@ -45,7 +45,7 @@ static int vesa_probe(void)
45 ireg.di = (size_t)&vginfo; 45 ireg.di = (size_t)&vginfo;
46 intcall(0x10, &ireg, &oreg); 46 intcall(0x10, &ireg, &oreg);
47 47
48 if (ireg.ax != 0x004f || 48 if (oreg.ax != 0x004f ||
49 vginfo.signature != VESA_MAGIC || 49 vginfo.signature != VESA_MAGIC ||
50 vginfo.version < 0x0102) 50 vginfo.version < 0x0102)
51 return 0; /* Not present */ 51 return 0; /* Not present */
@@ -70,7 +70,7 @@ static int vesa_probe(void)
70 ireg.di = (size_t)&vminfo; 70 ireg.di = (size_t)&vminfo;
71 intcall(0x10, &ireg, &oreg); 71 intcall(0x10, &ireg, &oreg);
72 72
73 if (ireg.ax != 0x004f) 73 if (oreg.ax != 0x004f)
74 continue; 74 continue;
75 75
76 if ((vminfo.mode_attr & 0x15) == 0x05) { 76 if ((vminfo.mode_attr & 0x15) == 0x05) {
diff --git a/arch/x86/include/asm/atomic_32.h b/arch/x86/include/asm/atomic_32.h
index 2503d4e64c2a..dc5a667ff791 100644
--- a/arch/x86/include/asm/atomic_32.h
+++ b/arch/x86/include/asm/atomic_32.h
@@ -19,7 +19,10 @@
19 * 19 *
20 * Atomically reads the value of @v. 20 * Atomically reads the value of @v.
21 */ 21 */
22#define atomic_read(v) ((v)->counter) 22static inline int atomic_read(const atomic_t *v)
23{
24 return v->counter;
25}
23 26
24/** 27/**
25 * atomic_set - set atomic variable 28 * atomic_set - set atomic variable
@@ -28,7 +31,10 @@
28 * 31 *
29 * Atomically sets the value of @v to @i. 32 * Atomically sets the value of @v to @i.
30 */ 33 */
31#define atomic_set(v, i) (((v)->counter) = (i)) 34static inline void atomic_set(atomic_t *v, int i)
35{
36 v->counter = i;
37}
32 38
33/** 39/**
34 * atomic_add - add integer to atomic variable 40 * atomic_add - add integer to atomic variable
@@ -200,8 +206,15 @@ static inline int atomic_sub_return(int i, atomic_t *v)
200 return atomic_add_return(-i, v); 206 return atomic_add_return(-i, v);
201} 207}
202 208
203#define atomic_cmpxchg(v, old, new) (cmpxchg(&((v)->counter), (old), (new))) 209static inline int atomic_cmpxchg(atomic_t *v, int old, int new)
204#define atomic_xchg(v, new) (xchg(&((v)->counter), (new))) 210{
211 return cmpxchg(&v->counter, old, new);
212}
213
214static inline int atomic_xchg(atomic_t *v, int new)
215{
216 return xchg(&v->counter, new);
217}
205 218
206/** 219/**
207 * atomic_add_unless - add unless the number is already a given value 220 * atomic_add_unless - add unless the number is already a given value
@@ -250,45 +263,12 @@ static inline int atomic_add_unless(atomic_t *v, int a, int u)
250/* An 64bit atomic type */ 263/* An 64bit atomic type */
251 264
252typedef struct { 265typedef struct {
253 unsigned long long counter; 266 u64 __aligned(8) counter;
254} atomic64_t; 267} atomic64_t;
255 268
256#define ATOMIC64_INIT(val) { (val) } 269#define ATOMIC64_INIT(val) { (val) }
257 270
258/** 271extern u64 atomic64_cmpxchg(atomic64_t *ptr, u64 old_val, u64 new_val);
259 * atomic64_read - read atomic64 variable
260 * @ptr: pointer of type atomic64_t
261 *
262 * Atomically reads the value of @v.
263 * Doesn't imply a read memory barrier.
264 */
265#define __atomic64_read(ptr) ((ptr)->counter)
266
267static inline unsigned long long
268cmpxchg8b(unsigned long long *ptr, unsigned long long old, unsigned long long new)
269{
270 asm volatile(
271
272 LOCK_PREFIX "cmpxchg8b (%[ptr])\n"
273
274 : "=A" (old)
275
276 : [ptr] "D" (ptr),
277 "A" (old),
278 "b" (ll_low(new)),
279 "c" (ll_high(new))
280
281 : "memory");
282
283 return old;
284}
285
286static inline unsigned long long
287atomic64_cmpxchg(atomic64_t *ptr, unsigned long long old_val,
288 unsigned long long new_val)
289{
290 return cmpxchg8b(&ptr->counter, old_val, new_val);
291}
292 272
293/** 273/**
294 * atomic64_xchg - xchg atomic64 variable 274 * atomic64_xchg - xchg atomic64 variable
@@ -298,18 +278,7 @@ atomic64_cmpxchg(atomic64_t *ptr, unsigned long long old_val,
298 * Atomically xchgs the value of @ptr to @new_val and returns 278 * Atomically xchgs the value of @ptr to @new_val and returns
299 * the old value. 279 * the old value.
300 */ 280 */
301 281extern u64 atomic64_xchg(atomic64_t *ptr, u64 new_val);
302static inline unsigned long long
303atomic64_xchg(atomic64_t *ptr, unsigned long long new_val)
304{
305 unsigned long long old_val;
306
307 do {
308 old_val = atomic_read(ptr);
309 } while (atomic64_cmpxchg(ptr, old_val, new_val) != old_val);
310
311 return old_val;
312}
313 282
314/** 283/**
315 * atomic64_set - set atomic64 variable 284 * atomic64_set - set atomic64 variable
@@ -318,10 +287,7 @@ atomic64_xchg(atomic64_t *ptr, unsigned long long new_val)
318 * 287 *
319 * Atomically sets the value of @ptr to @new_val. 288 * Atomically sets the value of @ptr to @new_val.
320 */ 289 */
321static inline void atomic64_set(atomic64_t *ptr, unsigned long long new_val) 290extern void atomic64_set(atomic64_t *ptr, u64 new_val);
322{
323 atomic64_xchg(ptr, new_val);
324}
325 291
326/** 292/**
327 * atomic64_read - read atomic64 variable 293 * atomic64_read - read atomic64 variable
@@ -329,17 +295,30 @@ static inline void atomic64_set(atomic64_t *ptr, unsigned long long new_val)
329 * 295 *
330 * Atomically reads the value of @ptr and returns it. 296 * Atomically reads the value of @ptr and returns it.
331 */ 297 */
332static inline unsigned long long atomic64_read(atomic64_t *ptr) 298static inline u64 atomic64_read(atomic64_t *ptr)
333{ 299{
334 unsigned long long curr_val; 300 u64 res;
335 301
336 do { 302 /*
337 curr_val = __atomic64_read(ptr); 303 * Note, we inline this atomic64_t primitive because
338 } while (atomic64_cmpxchg(ptr, curr_val, curr_val) != curr_val); 304 * it only clobbers EAX/EDX and leaves the others
339 305 * untouched. We also (somewhat subtly) rely on the
340 return curr_val; 306 * fact that cmpxchg8b returns the current 64-bit value
307 * of the memory location we are touching:
308 */
309 asm volatile(
310 "mov %%ebx, %%eax\n\t"
311 "mov %%ecx, %%edx\n\t"
312 LOCK_PREFIX "cmpxchg8b %1\n"
313 : "=&A" (res)
314 : "m" (*ptr)
315 );
316
317 return res;
341} 318}
342 319
320extern u64 atomic64_read(atomic64_t *ptr);
321
343/** 322/**
344 * atomic64_add_return - add and return 323 * atomic64_add_return - add and return
345 * @delta: integer value to add 324 * @delta: integer value to add
@@ -347,34 +326,14 @@ static inline unsigned long long atomic64_read(atomic64_t *ptr)
347 * 326 *
348 * Atomically adds @delta to @ptr and returns @delta + *@ptr 327 * Atomically adds @delta to @ptr and returns @delta + *@ptr
349 */ 328 */
350static inline unsigned long long 329extern u64 atomic64_add_return(u64 delta, atomic64_t *ptr);
351atomic64_add_return(unsigned long long delta, atomic64_t *ptr)
352{
353 unsigned long long old_val, new_val;
354
355 do {
356 old_val = atomic_read(ptr);
357 new_val = old_val + delta;
358
359 } while (atomic64_cmpxchg(ptr, old_val, new_val) != old_val);
360
361 return new_val;
362}
363
364static inline long atomic64_sub_return(unsigned long long delta, atomic64_t *ptr)
365{
366 return atomic64_add_return(-delta, ptr);
367}
368 330
369static inline long atomic64_inc_return(atomic64_t *ptr) 331/*
370{ 332 * Other variants with different arithmetic operators:
371 return atomic64_add_return(1, ptr); 333 */
372} 334extern u64 atomic64_sub_return(u64 delta, atomic64_t *ptr);
373 335extern u64 atomic64_inc_return(atomic64_t *ptr);
374static inline long atomic64_dec_return(atomic64_t *ptr) 336extern u64 atomic64_dec_return(atomic64_t *ptr);
375{
376 return atomic64_sub_return(1, ptr);
377}
378 337
379/** 338/**
380 * atomic64_add - add integer to atomic64 variable 339 * atomic64_add - add integer to atomic64 variable
@@ -383,10 +342,7 @@ static inline long atomic64_dec_return(atomic64_t *ptr)
383 * 342 *
384 * Atomically adds @delta to @ptr. 343 * Atomically adds @delta to @ptr.
385 */ 344 */
386static inline void atomic64_add(unsigned long long delta, atomic64_t *ptr) 345extern void atomic64_add(u64 delta, atomic64_t *ptr);
387{
388 atomic64_add_return(delta, ptr);
389}
390 346
391/** 347/**
392 * atomic64_sub - subtract the atomic64 variable 348 * atomic64_sub - subtract the atomic64 variable
@@ -395,10 +351,7 @@ static inline void atomic64_add(unsigned long long delta, atomic64_t *ptr)
395 * 351 *
396 * Atomically subtracts @delta from @ptr. 352 * Atomically subtracts @delta from @ptr.
397 */ 353 */
398static inline void atomic64_sub(unsigned long long delta, atomic64_t *ptr) 354extern void atomic64_sub(u64 delta, atomic64_t *ptr);
399{
400 atomic64_add(-delta, ptr);
401}
402 355
403/** 356/**
404 * atomic64_sub_and_test - subtract value from variable and test result 357 * atomic64_sub_and_test - subtract value from variable and test result
@@ -409,13 +362,7 @@ static inline void atomic64_sub(unsigned long long delta, atomic64_t *ptr)
409 * true if the result is zero, or false for all 362 * true if the result is zero, or false for all
410 * other cases. 363 * other cases.
411 */ 364 */
412static inline int 365extern int atomic64_sub_and_test(u64 delta, atomic64_t *ptr);
413atomic64_sub_and_test(unsigned long long delta, atomic64_t *ptr)
414{
415 unsigned long long old_val = atomic64_sub_return(delta, ptr);
416
417 return old_val == 0;
418}
419 366
420/** 367/**
421 * atomic64_inc - increment atomic64 variable 368 * atomic64_inc - increment atomic64 variable
@@ -423,10 +370,7 @@ atomic64_sub_and_test(unsigned long long delta, atomic64_t *ptr)
423 * 370 *
424 * Atomically increments @ptr by 1. 371 * Atomically increments @ptr by 1.
425 */ 372 */
426static inline void atomic64_inc(atomic64_t *ptr) 373extern void atomic64_inc(atomic64_t *ptr);
427{
428 atomic64_add(1, ptr);
429}
430 374
431/** 375/**
432 * atomic64_dec - decrement atomic64 variable 376 * atomic64_dec - decrement atomic64 variable
@@ -434,10 +378,7 @@ static inline void atomic64_inc(atomic64_t *ptr)
434 * 378 *
435 * Atomically decrements @ptr by 1. 379 * Atomically decrements @ptr by 1.
436 */ 380 */
437static inline void atomic64_dec(atomic64_t *ptr) 381extern void atomic64_dec(atomic64_t *ptr);
438{
439 atomic64_sub(1, ptr);
440}
441 382
442/** 383/**
443 * atomic64_dec_and_test - decrement and test 384 * atomic64_dec_and_test - decrement and test
@@ -447,10 +388,7 @@ static inline void atomic64_dec(atomic64_t *ptr)
447 * returns true if the result is 0, or false for all other 388 * returns true if the result is 0, or false for all other
448 * cases. 389 * cases.
449 */ 390 */
450static inline int atomic64_dec_and_test(atomic64_t *ptr) 391extern int atomic64_dec_and_test(atomic64_t *ptr);
451{
452 return atomic64_sub_and_test(1, ptr);
453}
454 392
455/** 393/**
456 * atomic64_inc_and_test - increment and test 394 * atomic64_inc_and_test - increment and test
@@ -460,10 +398,7 @@ static inline int atomic64_dec_and_test(atomic64_t *ptr)
460 * and returns true if the result is zero, or false for all 398 * and returns true if the result is zero, or false for all
461 * other cases. 399 * other cases.
462 */ 400 */
463static inline int atomic64_inc_and_test(atomic64_t *ptr) 401extern int atomic64_inc_and_test(atomic64_t *ptr);
464{
465 return atomic64_sub_and_test(-1, ptr);
466}
467 402
468/** 403/**
469 * atomic64_add_negative - add and test if negative 404 * atomic64_add_negative - add and test if negative
@@ -474,13 +409,7 @@ static inline int atomic64_inc_and_test(atomic64_t *ptr)
474 * if the result is negative, or false when 409 * if the result is negative, or false when
475 * result is greater than or equal to zero. 410 * result is greater than or equal to zero.
476 */ 411 */
477static inline int 412extern int atomic64_add_negative(u64 delta, atomic64_t *ptr);
478atomic64_add_negative(unsigned long long delta, atomic64_t *ptr)
479{
480 long long old_val = atomic64_add_return(delta, ptr);
481
482 return old_val < 0;
483}
484 413
485#include <asm-generic/atomic-long.h> 414#include <asm-generic/atomic-long.h>
486#endif /* _ASM_X86_ATOMIC_32_H */ 415#endif /* _ASM_X86_ATOMIC_32_H */
diff --git a/arch/x86/include/asm/atomic_64.h b/arch/x86/include/asm/atomic_64.h
index 0d6360220007..d605dc268e79 100644
--- a/arch/x86/include/asm/atomic_64.h
+++ b/arch/x86/include/asm/atomic_64.h
@@ -18,7 +18,10 @@
18 * 18 *
19 * Atomically reads the value of @v. 19 * Atomically reads the value of @v.
20 */ 20 */
21#define atomic_read(v) ((v)->counter) 21static inline int atomic_read(const atomic_t *v)
22{
23 return v->counter;
24}
22 25
23/** 26/**
24 * atomic_set - set atomic variable 27 * atomic_set - set atomic variable
@@ -27,7 +30,10 @@
27 * 30 *
28 * Atomically sets the value of @v to @i. 31 * Atomically sets the value of @v to @i.
29 */ 32 */
30#define atomic_set(v, i) (((v)->counter) = (i)) 33static inline void atomic_set(atomic_t *v, int i)
34{
35 v->counter = i;
36}
31 37
32/** 38/**
33 * atomic_add - add integer to atomic variable 39 * atomic_add - add integer to atomic variable
@@ -192,7 +198,10 @@ static inline int atomic_sub_return(int i, atomic_t *v)
192 * Atomically reads the value of @v. 198 * Atomically reads the value of @v.
193 * Doesn't imply a read memory barrier. 199 * Doesn't imply a read memory barrier.
194 */ 200 */
195#define atomic64_read(v) ((v)->counter) 201static inline long atomic64_read(const atomic64_t *v)
202{
203 return v->counter;
204}
196 205
197/** 206/**
198 * atomic64_set - set atomic64 variable 207 * atomic64_set - set atomic64 variable
@@ -201,7 +210,10 @@ static inline int atomic_sub_return(int i, atomic_t *v)
201 * 210 *
202 * Atomically sets the value of @v to @i. 211 * Atomically sets the value of @v to @i.
203 */ 212 */
204#define atomic64_set(v, i) (((v)->counter) = (i)) 213static inline void atomic64_set(atomic64_t *v, long i)
214{
215 v->counter = i;
216}
205 217
206/** 218/**
207 * atomic64_add - add integer to atomic64 variable 219 * atomic64_add - add integer to atomic64 variable
@@ -355,11 +367,25 @@ static inline long atomic64_sub_return(long i, atomic64_t *v)
355#define atomic64_inc_return(v) (atomic64_add_return(1, (v))) 367#define atomic64_inc_return(v) (atomic64_add_return(1, (v)))
356#define atomic64_dec_return(v) (atomic64_sub_return(1, (v))) 368#define atomic64_dec_return(v) (atomic64_sub_return(1, (v)))
357 369
358#define atomic64_cmpxchg(v, old, new) (cmpxchg(&((v)->counter), (old), (new))) 370static inline long atomic64_cmpxchg(atomic64_t *v, long old, long new)
359#define atomic64_xchg(v, new) (xchg(&((v)->counter), new)) 371{
372 return cmpxchg(&v->counter, old, new);
373}
374
375static inline long atomic64_xchg(atomic64_t *v, long new)
376{
377 return xchg(&v->counter, new);
378}
360 379
361#define atomic_cmpxchg(v, old, new) (cmpxchg(&((v)->counter), (old), (new))) 380static inline long atomic_cmpxchg(atomic_t *v, int old, int new)
362#define atomic_xchg(v, new) (xchg(&((v)->counter), (new))) 381{
382 return cmpxchg(&v->counter, old, new);
383}
384
385static inline long atomic_xchg(atomic_t *v, int new)
386{
387 return xchg(&v->counter, new);
388}
363 389
364/** 390/**
365 * atomic_add_unless - add unless the number is a given value 391 * atomic_add_unless - add unless the number is a given value
diff --git a/arch/x86/include/asm/boot.h b/arch/x86/include/asm/boot.h
index 418e632d4a80..7a1065958ba9 100644
--- a/arch/x86/include/asm/boot.h
+++ b/arch/x86/include/asm/boot.h
@@ -8,7 +8,7 @@
8 8
9#ifdef __KERNEL__ 9#ifdef __KERNEL__
10 10
11#include <asm/page_types.h> 11#include <asm/pgtable_types.h>
12 12
13/* Physical address where kernel should be loaded. */ 13/* Physical address where kernel should be loaded. */
14#define LOAD_PHYSICAL_ADDR ((CONFIG_PHYSICAL_START \ 14#define LOAD_PHYSICAL_ADDR ((CONFIG_PHYSICAL_START \
@@ -16,10 +16,10 @@
16 & ~(CONFIG_PHYSICAL_ALIGN - 1)) 16 & ~(CONFIG_PHYSICAL_ALIGN - 1))
17 17
18/* Minimum kernel alignment, as a power of two */ 18/* Minimum kernel alignment, as a power of two */
19#ifdef CONFIG_x86_64 19#ifdef CONFIG_X86_64
20#define MIN_KERNEL_ALIGN_LG2 PMD_SHIFT 20#define MIN_KERNEL_ALIGN_LG2 PMD_SHIFT
21#else 21#else
22#define MIN_KERNEL_ALIGN_LG2 (PAGE_SHIFT+1) 22#define MIN_KERNEL_ALIGN_LG2 (PAGE_SHIFT + THREAD_ORDER)
23#endif 23#endif
24#define MIN_KERNEL_ALIGN (_AC(1, UL) << MIN_KERNEL_ALIGN_LG2) 24#define MIN_KERNEL_ALIGN (_AC(1, UL) << MIN_KERNEL_ALIGN_LG2)
25 25
diff --git a/arch/x86/include/asm/fixmap.h b/arch/x86/include/asm/fixmap.h
index 2d81af3974a0..7b2d71df39a6 100644
--- a/arch/x86/include/asm/fixmap.h
+++ b/arch/x86/include/asm/fixmap.h
@@ -111,12 +111,9 @@ enum fixed_addresses {
111#ifdef CONFIG_PARAVIRT 111#ifdef CONFIG_PARAVIRT
112 FIX_PARAVIRT_BOOTMAP, 112 FIX_PARAVIRT_BOOTMAP,
113#endif 113#endif
114 FIX_TEXT_POKE0, /* reserve 2 pages for text_poke() */ 114 FIX_TEXT_POKE1, /* reserve 2 pages for text_poke() */
115 FIX_TEXT_POKE1, 115 FIX_TEXT_POKE0, /* first page is last, because allocation is backward */
116 __end_of_permanent_fixed_addresses, 116 __end_of_permanent_fixed_addresses,
117#ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
118 FIX_OHCI1394_BASE,
119#endif
120 /* 117 /*
121 * 256 temporary boot-time mappings, used by early_ioremap(), 118 * 256 temporary boot-time mappings, used by early_ioremap(),
122 * before ioremap() is functional. 119 * before ioremap() is functional.
@@ -129,6 +126,9 @@ enum fixed_addresses {
129 FIX_BTMAP_END = __end_of_permanent_fixed_addresses + 256 - 126 FIX_BTMAP_END = __end_of_permanent_fixed_addresses + 256 -
130 (__end_of_permanent_fixed_addresses & 255), 127 (__end_of_permanent_fixed_addresses & 255),
131 FIX_BTMAP_BEGIN = FIX_BTMAP_END + NR_FIX_BTMAPS*FIX_BTMAPS_SLOTS - 1, 128 FIX_BTMAP_BEGIN = FIX_BTMAP_END + NR_FIX_BTMAPS*FIX_BTMAPS_SLOTS - 1,
129#ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
130 FIX_OHCI1394_BASE,
131#endif
132#ifdef CONFIG_X86_32 132#ifdef CONFIG_X86_32
133 FIX_WP_TEST, 133 FIX_WP_TEST,
134#endif 134#endif
diff --git a/arch/x86/include/asm/msr-index.h b/arch/x86/include/asm/msr-index.h
index 1692fb5050e3..6be7fc254b59 100644
--- a/arch/x86/include/asm/msr-index.h
+++ b/arch/x86/include/asm/msr-index.h
@@ -246,10 +246,6 @@
246#define MSR_IA32_MISC_ENABLE_TURBO_DISABLE (1ULL << 38) 246#define MSR_IA32_MISC_ENABLE_TURBO_DISABLE (1ULL << 38)
247#define MSR_IA32_MISC_ENABLE_IP_PREF_DISABLE (1ULL << 39) 247#define MSR_IA32_MISC_ENABLE_IP_PREF_DISABLE (1ULL << 39)
248 248
249/* Intel Model 6 */
250#define MSR_P6_EVNTSEL0 0x00000186
251#define MSR_P6_EVNTSEL1 0x00000187
252
253/* P4/Xeon+ specific */ 249/* P4/Xeon+ specific */
254#define MSR_IA32_MCG_EAX 0x00000180 250#define MSR_IA32_MCG_EAX 0x00000180
255#define MSR_IA32_MCG_EBX 0x00000181 251#define MSR_IA32_MCG_EBX 0x00000181
diff --git a/arch/x86/include/asm/nmi.h b/arch/x86/include/asm/nmi.h
index c97264409934..c86e5ed4af51 100644
--- a/arch/x86/include/asm/nmi.h
+++ b/arch/x86/include/asm/nmi.h
@@ -72,7 +72,6 @@ void lapic_watchdog_stop(void);
72int lapic_watchdog_init(unsigned nmi_hz); 72int lapic_watchdog_init(unsigned nmi_hz);
73int lapic_wd_event(unsigned nmi_hz); 73int lapic_wd_event(unsigned nmi_hz);
74unsigned lapic_adjust_nmi_hz(unsigned hz); 74unsigned lapic_adjust_nmi_hz(unsigned hz);
75int lapic_watchdog_ok(void);
76void disable_lapic_nmi_watchdog(void); 75void disable_lapic_nmi_watchdog(void);
77void enable_lapic_nmi_watchdog(void); 76void enable_lapic_nmi_watchdog(void);
78void stop_nmi(void); 77void stop_nmi(void);
diff --git a/arch/x86/include/asm/pci.h b/arch/x86/include/asm/pci.h
index 927958d13c19..1ff685ca221c 100644
--- a/arch/x86/include/asm/pci.h
+++ b/arch/x86/include/asm/pci.h
@@ -91,7 +91,7 @@ extern void pci_iommu_alloc(void);
91 91
92#define PCI_DMA_BUS_IS_PHYS (dma_ops->is_phys) 92#define PCI_DMA_BUS_IS_PHYS (dma_ops->is_phys)
93 93
94#if defined(CONFIG_X86_64) || defined(CONFIG_DMA_API_DEBUG) 94#if defined(CONFIG_X86_64) || defined(CONFIG_DMAR) || defined(CONFIG_DMA_API_DEBUG)
95 95
96#define DECLARE_PCI_UNMAP_ADDR(ADDR_NAME) \ 96#define DECLARE_PCI_UNMAP_ADDR(ADDR_NAME) \
97 dma_addr_t ADDR_NAME; 97 dma_addr_t ADDR_NAME;
diff --git a/arch/x86/include/asm/percpu.h b/arch/x86/include/asm/percpu.h
index 02ecb30982a3..103f1ddb0d85 100644
--- a/arch/x86/include/asm/percpu.h
+++ b/arch/x86/include/asm/percpu.h
@@ -42,6 +42,7 @@
42 42
43#else /* ...!ASSEMBLY */ 43#else /* ...!ASSEMBLY */
44 44
45#include <linux/kernel.h>
45#include <linux/stringify.h> 46#include <linux/stringify.h>
46 47
47#ifdef CONFIG_SMP 48#ifdef CONFIG_SMP
@@ -155,6 +156,15 @@ do { \
155/* We can use this directly for local CPU (faster). */ 156/* We can use this directly for local CPU (faster). */
156DECLARE_PER_CPU(unsigned long, this_cpu_off); 157DECLARE_PER_CPU(unsigned long, this_cpu_off);
157 158
159#ifdef CONFIG_NEED_MULTIPLE_NODES
160void *pcpu_lpage_remapped(void *kaddr);
161#else
162static inline void *pcpu_lpage_remapped(void *kaddr)
163{
164 return NULL;
165}
166#endif
167
158#endif /* !__ASSEMBLY__ */ 168#endif /* !__ASSEMBLY__ */
159 169
160#ifdef CONFIG_SMP 170#ifdef CONFIG_SMP
diff --git a/arch/x86/include/asm/perf_counter.h b/arch/x86/include/asm/perf_counter.h
index 5fb33e160ea0..fa64e401589d 100644
--- a/arch/x86/include/asm/perf_counter.h
+++ b/arch/x86/include/asm/perf_counter.h
@@ -87,6 +87,9 @@ union cpuid10_edx {
87#ifdef CONFIG_PERF_COUNTERS 87#ifdef CONFIG_PERF_COUNTERS
88extern void init_hw_perf_counters(void); 88extern void init_hw_perf_counters(void);
89extern void perf_counters_lapic_init(void); 89extern void perf_counters_lapic_init(void);
90
91#define PERF_COUNTER_INDEX_OFFSET 0
92
90#else 93#else
91static inline void init_hw_perf_counters(void) { } 94static inline void init_hw_perf_counters(void) { }
92static inline void perf_counters_lapic_init(void) { } 95static inline void perf_counters_lapic_init(void) { }
diff --git a/arch/x86/include/asm/proto.h b/arch/x86/include/asm/proto.h
index 49fb3ecf3bb3..621f56d73121 100644
--- a/arch/x86/include/asm/proto.h
+++ b/arch/x86/include/asm/proto.h
@@ -22,7 +22,14 @@ extern int reboot_force;
22 22
23long do_arch_prctl(struct task_struct *task, int code, unsigned long addr); 23long do_arch_prctl(struct task_struct *task, int code, unsigned long addr);
24 24
25#define round_up(x, y) (((x) + (y) - 1) & ~((y) - 1)) 25/*
26#define round_down(x, y) ((x) & ~((y) - 1)) 26 * This looks more complex than it should be. But we need to
27 * get the type for the ~ right in round_down (it needs to be
28 * as wide as the result!), and we want to evaluate the macro
29 * arguments just once each.
30 */
31#define __round_mask(x,y) ((__typeof__(x))((y)-1))
32#define round_up(x,y) ((((x)-1) | __round_mask(x,y))+1)
33#define round_down(x,y) ((x) & ~__round_mask(x,y))
27 34
28#endif /* _ASM_X86_PROTO_H */ 35#endif /* _ASM_X86_PROTO_H */
diff --git a/arch/x86/include/asm/spinlock.h b/arch/x86/include/asm/spinlock.h
index b7e5db876399..4e77853321db 100644
--- a/arch/x86/include/asm/spinlock.h
+++ b/arch/x86/include/asm/spinlock.h
@@ -302,4 +302,8 @@ static inline void __raw_write_unlock(raw_rwlock_t *rw)
302#define _raw_read_relax(lock) cpu_relax() 302#define _raw_read_relax(lock) cpu_relax()
303#define _raw_write_relax(lock) cpu_relax() 303#define _raw_write_relax(lock) cpu_relax()
304 304
305/* The {read|write|spin}_lock() on x86 are full memory barriers. */
306static inline void smp_mb__after_lock(void) { }
307#define ARCH_HAS_SMP_MB_AFTER_LOCK
308
305#endif /* _ASM_X86_SPINLOCK_H */ 309#endif /* _ASM_X86_SPINLOCK_H */
diff --git a/arch/x86/include/asm/stacktrace.h b/arch/x86/include/asm/stacktrace.h
index f517944b2b17..cf86a5e73815 100644
--- a/arch/x86/include/asm/stacktrace.h
+++ b/arch/x86/include/asm/stacktrace.h
@@ -3,6 +3,8 @@
3 3
4extern int kstack_depth_to_print; 4extern int kstack_depth_to_print;
5 5
6int x86_is_stack_id(int id, char *name);
7
6/* Generic stack tracer with callbacks */ 8/* Generic stack tracer with callbacks */
7 9
8struct stacktrace_ops { 10struct stacktrace_ops {
diff --git a/arch/x86/include/asm/thread_info.h b/arch/x86/include/asm/thread_info.h
index b0783520988b..fad7d40b75f8 100644
--- a/arch/x86/include/asm/thread_info.h
+++ b/arch/x86/include/asm/thread_info.h
@@ -49,7 +49,7 @@ struct thread_info {
49 .exec_domain = &default_exec_domain, \ 49 .exec_domain = &default_exec_domain, \
50 .flags = 0, \ 50 .flags = 0, \
51 .cpu = 0, \ 51 .cpu = 0, \
52 .preempt_count = 1, \ 52 .preempt_count = INIT_PREEMPT_COUNT, \
53 .addr_limit = KERNEL_DS, \ 53 .addr_limit = KERNEL_DS, \
54 .restart_block = { \ 54 .restart_block = { \
55 .fn = do_no_restart_syscall, \ 55 .fn = do_no_restart_syscall, \
diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile
index 6c327b852e23..430d5b24af7b 100644
--- a/arch/x86/kernel/Makefile
+++ b/arch/x86/kernel/Makefile
@@ -26,6 +26,8 @@ CFLAGS_tsc.o := $(nostackp)
26CFLAGS_paravirt.o := $(nostackp) 26CFLAGS_paravirt.o := $(nostackp)
27GCOV_PROFILE_vsyscall_64.o := n 27GCOV_PROFILE_vsyscall_64.o := n
28GCOV_PROFILE_hpet.o := n 28GCOV_PROFILE_hpet.o := n
29GCOV_PROFILE_tsc.o := n
30GCOV_PROFILE_paravirt.o := n
29 31
30obj-y := process_$(BITS).o signal.o entry_$(BITS).o 32obj-y := process_$(BITS).o signal.o entry_$(BITS).o
31obj-y += traps.o irq.o irq_$(BITS).o dumpstack_$(BITS).o 33obj-y += traps.o irq.o irq_$(BITS).o dumpstack_$(BITS).o
diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 9372f0406ad4..6c99f5037801 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -1192,7 +1192,7 @@ out:
1192 return 0; 1192 return 0;
1193} 1193}
1194 1194
1195struct notifier_block device_nb = { 1195static struct notifier_block device_nb = {
1196 .notifier_call = device_change_notifier, 1196 .notifier_call = device_change_notifier,
1197}; 1197};
1198 1198
@@ -1763,7 +1763,7 @@ static void *alloc_coherent(struct device *dev, size_t size,
1763 flag |= __GFP_ZERO; 1763 flag |= __GFP_ZERO;
1764 virt_addr = (void *)__get_free_pages(flag, get_order(size)); 1764 virt_addr = (void *)__get_free_pages(flag, get_order(size));
1765 if (!virt_addr) 1765 if (!virt_addr)
1766 return 0; 1766 return NULL;
1767 1767
1768 paddr = virt_to_phys(virt_addr); 1768 paddr = virt_to_phys(virt_addr);
1769 1769
diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c
index 10b2accd12ea..c1b17e97252e 100644
--- a/arch/x86/kernel/amd_iommu_init.c
+++ b/arch/x86/kernel/amd_iommu_init.c
@@ -472,6 +472,8 @@ static u8 * __init alloc_event_buffer(struct amd_iommu *iommu)
472 if (iommu->evt_buf == NULL) 472 if (iommu->evt_buf == NULL)
473 return NULL; 473 return NULL;
474 474
475 iommu->evt_buf_size = EVT_BUFFER_SIZE;
476
475 return iommu->evt_buf; 477 return iommu->evt_buf;
476} 478}
477 479
@@ -691,6 +693,7 @@ static void __init init_iommu_from_acpi(struct amd_iommu *iommu,
691 693
692 devid = e->devid; 694 devid = e->devid;
693 devid_to = e->ext >> 8; 695 devid_to = e->ext >> 8;
696 set_dev_entry_from_acpi(iommu, devid , e->flags, 0);
694 set_dev_entry_from_acpi(iommu, devid_to, e->flags, 0); 697 set_dev_entry_from_acpi(iommu, devid_to, e->flags, 0);
695 amd_iommu_alias_table[devid] = devid_to; 698 amd_iommu_alias_table[devid] = devid_to;
696 break; 699 break;
@@ -749,11 +752,13 @@ static void __init init_iommu_from_acpi(struct amd_iommu *iommu,
749 752
750 devid = e->devid; 753 devid = e->devid;
751 for (dev_i = devid_start; dev_i <= devid; ++dev_i) { 754 for (dev_i = devid_start; dev_i <= devid; ++dev_i) {
752 if (alias) 755 if (alias) {
753 amd_iommu_alias_table[dev_i] = devid_to; 756 amd_iommu_alias_table[dev_i] = devid_to;
754 set_dev_entry_from_acpi(iommu, 757 set_dev_entry_from_acpi(iommu,
755 amd_iommu_alias_table[dev_i], 758 devid_to, flags, ext_flags);
756 flags, ext_flags); 759 }
760 set_dev_entry_from_acpi(iommu, dev_i,
761 flags, ext_flags);
757 } 762 }
758 break; 763 break;
759 default: 764 default:
diff --git a/arch/x86/kernel/apic/apic.c b/arch/x86/kernel/apic/apic.c
index 8c7c042ecad1..0a1c2830ec66 100644
--- a/arch/x86/kernel/apic/apic.c
+++ b/arch/x86/kernel/apic/apic.c
@@ -140,7 +140,6 @@ int x2apic_mode;
140#ifdef CONFIG_X86_X2APIC 140#ifdef CONFIG_X86_X2APIC
141/* x2apic enabled before OS handover */ 141/* x2apic enabled before OS handover */
142static int x2apic_preenabled; 142static int x2apic_preenabled;
143static int disable_x2apic;
144static __init int setup_nox2apic(char *str) 143static __init int setup_nox2apic(char *str)
145{ 144{
146 if (x2apic_enabled()) { 145 if (x2apic_enabled()) {
@@ -149,7 +148,6 @@ static __init int setup_nox2apic(char *str)
149 return 0; 148 return 0;
150 } 149 }
151 150
152 disable_x2apic = 1;
153 setup_clear_cpu_cap(X86_FEATURE_X2APIC); 151 setup_clear_cpu_cap(X86_FEATURE_X2APIC);
154 return 0; 152 return 0;
155} 153}
diff --git a/arch/x86/kernel/apic/io_apic.c b/arch/x86/kernel/apic/io_apic.c
index 4d0216fcb36c..90b5e6efa938 100644
--- a/arch/x86/kernel/apic/io_apic.c
+++ b/arch/x86/kernel/apic/io_apic.c
@@ -1716,25 +1716,19 @@ __apicdebuginit(void) print_IO_APIC(void)
1716 return; 1716 return;
1717} 1717}
1718 1718
1719__apicdebuginit(void) print_APIC_bitfield(int base) 1719__apicdebuginit(void) print_APIC_field(int base)
1720{ 1720{
1721 unsigned int v; 1721 int i;
1722 int i, j;
1723 1722
1724 if (apic_verbosity == APIC_QUIET) 1723 if (apic_verbosity == APIC_QUIET)
1725 return; 1724 return;
1726 1725
1727 printk(KERN_DEBUG "0123456789abcdef0123456789abcdef\n" KERN_DEBUG); 1726 printk(KERN_DEBUG);
1728 for (i = 0; i < 8; i++) { 1727
1729 v = apic_read(base + i*0x10); 1728 for (i = 0; i < 8; i++)
1730 for (j = 0; j < 32; j++) { 1729 printk(KERN_CONT "%08x", apic_read(base + i*0x10));
1731 if (v & (1<<j)) 1730
1732 printk("1"); 1731 printk(KERN_CONT "\n");
1733 else
1734 printk("0");
1735 }
1736 printk("\n");
1737 }
1738} 1732}
1739 1733
1740__apicdebuginit(void) print_local_APIC(void *dummy) 1734__apicdebuginit(void) print_local_APIC(void *dummy)
@@ -1745,7 +1739,7 @@ __apicdebuginit(void) print_local_APIC(void *dummy)
1745 if (apic_verbosity == APIC_QUIET) 1739 if (apic_verbosity == APIC_QUIET)
1746 return; 1740 return;
1747 1741
1748 printk("\n" KERN_DEBUG "printing local APIC contents on CPU#%d/%d:\n", 1742 printk(KERN_DEBUG "printing local APIC contents on CPU#%d/%d:\n",
1749 smp_processor_id(), hard_smp_processor_id()); 1743 smp_processor_id(), hard_smp_processor_id());
1750 v = apic_read(APIC_ID); 1744 v = apic_read(APIC_ID);
1751 printk(KERN_INFO "... APIC ID: %08x (%01x)\n", v, read_apic_id()); 1745 printk(KERN_INFO "... APIC ID: %08x (%01x)\n", v, read_apic_id());
@@ -1786,11 +1780,11 @@ __apicdebuginit(void) print_local_APIC(void *dummy)
1786 printk(KERN_DEBUG "... APIC SPIV: %08x\n", v); 1780 printk(KERN_DEBUG "... APIC SPIV: %08x\n", v);
1787 1781
1788 printk(KERN_DEBUG "... APIC ISR field:\n"); 1782 printk(KERN_DEBUG "... APIC ISR field:\n");
1789 print_APIC_bitfield(APIC_ISR); 1783 print_APIC_field(APIC_ISR);
1790 printk(KERN_DEBUG "... APIC TMR field:\n"); 1784 printk(KERN_DEBUG "... APIC TMR field:\n");
1791 print_APIC_bitfield(APIC_TMR); 1785 print_APIC_field(APIC_TMR);
1792 printk(KERN_DEBUG "... APIC IRR field:\n"); 1786 printk(KERN_DEBUG "... APIC IRR field:\n");
1793 print_APIC_bitfield(APIC_IRR); 1787 print_APIC_field(APIC_IRR);
1794 1788
1795 if (APIC_INTEGRATED(ver)) { /* !82489DX */ 1789 if (APIC_INTEGRATED(ver)) { /* !82489DX */
1796 if (maxlvt > 3) /* Due to the Pentium erratum 3AP. */ 1790 if (maxlvt > 3) /* Due to the Pentium erratum 3AP. */
diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c
index e5b27d8f1b47..28e5f5956042 100644
--- a/arch/x86/kernel/cpu/amd.c
+++ b/arch/x86/kernel/cpu/amd.c
@@ -258,13 +258,15 @@ static void __cpuinit amd_detect_cmp(struct cpuinfo_x86 *c)
258{ 258{
259#ifdef CONFIG_X86_HT 259#ifdef CONFIG_X86_HT
260 unsigned bits; 260 unsigned bits;
261 int cpu = smp_processor_id();
261 262
262 bits = c->x86_coreid_bits; 263 bits = c->x86_coreid_bits;
263
264 /* Low order bits define the core id (index of core in socket) */ 264 /* Low order bits define the core id (index of core in socket) */
265 c->cpu_core_id = c->initial_apicid & ((1 << bits)-1); 265 c->cpu_core_id = c->initial_apicid & ((1 << bits)-1);
266 /* Convert the initial APIC ID into the socket ID */ 266 /* Convert the initial APIC ID into the socket ID */
267 c->phys_proc_id = c->initial_apicid >> bits; 267 c->phys_proc_id = c->initial_apicid >> bits;
268 /* use socket ID also for last level cache */
269 per_cpu(cpu_llc_id, cpu) = c->phys_proc_id;
268#endif 270#endif
269} 271}
270 272
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index 6b26d4deada0..f1961c07af9a 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -848,9 +848,6 @@ 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;
854} 851}
855 852
856#ifdef CONFIG_X86_64 853#ifdef CONFIG_X86_64
diff --git a/arch/x86/kernel/cpu/cpufreq/powernow-k8.c b/arch/x86/kernel/cpu/cpufreq/powernow-k8.c
index 81cbe64ed6b4..2a50ef891000 100644
--- a/arch/x86/kernel/cpu/cpufreq/powernow-k8.c
+++ b/arch/x86/kernel/cpu/cpufreq/powernow-k8.c
@@ -299,7 +299,7 @@ static int transition_pstate(struct powernow_k8_data *data, u32 pstate)
299static int transition_fid_vid(struct powernow_k8_data *data, 299static int transition_fid_vid(struct powernow_k8_data *data,
300 u32 reqfid, u32 reqvid) 300 u32 reqfid, u32 reqvid)
301{ 301{
302 if (core_voltage_pre_transition(data, reqvid)) 302 if (core_voltage_pre_transition(data, reqvid, reqfid))
303 return 1; 303 return 1;
304 304
305 if (core_frequency_transition(data, reqfid)) 305 if (core_frequency_transition(data, reqfid))
@@ -327,17 +327,20 @@ static int transition_fid_vid(struct powernow_k8_data *data,
327 327
328/* Phase 1 - core voltage transition ... setup voltage */ 328/* Phase 1 - core voltage transition ... setup voltage */
329static int core_voltage_pre_transition(struct powernow_k8_data *data, 329static int core_voltage_pre_transition(struct powernow_k8_data *data,
330 u32 reqvid) 330 u32 reqvid, u32 reqfid)
331{ 331{
332 u32 rvosteps = data->rvo; 332 u32 rvosteps = data->rvo;
333 u32 savefid = data->currfid; 333 u32 savefid = data->currfid;
334 u32 maxvid, lo; 334 u32 maxvid, lo, rvomult = 1;
335 335
336 dprintk("ph1 (cpu%d): start, currfid 0x%x, currvid 0x%x, " 336 dprintk("ph1 (cpu%d): start, currfid 0x%x, currvid 0x%x, "
337 "reqvid 0x%x, rvo 0x%x\n", 337 "reqvid 0x%x, rvo 0x%x\n",
338 smp_processor_id(), 338 smp_processor_id(),
339 data->currfid, data->currvid, reqvid, data->rvo); 339 data->currfid, data->currvid, reqvid, data->rvo);
340 340
341 if ((savefid < LO_FID_TABLE_TOP) && (reqfid < LO_FID_TABLE_TOP))
342 rvomult = 2;
343 rvosteps *= rvomult;
341 rdmsr(MSR_FIDVID_STATUS, lo, maxvid); 344 rdmsr(MSR_FIDVID_STATUS, lo, maxvid);
342 maxvid = 0x1f & (maxvid >> 16); 345 maxvid = 0x1f & (maxvid >> 16);
343 dprintk("ph1 maxvid=0x%x\n", maxvid); 346 dprintk("ph1 maxvid=0x%x\n", maxvid);
@@ -351,7 +354,8 @@ static int core_voltage_pre_transition(struct powernow_k8_data *data,
351 return 1; 354 return 1;
352 } 355 }
353 356
354 while ((rvosteps > 0) && ((data->rvo + data->currvid) > reqvid)) { 357 while ((rvosteps > 0) &&
358 ((rvomult * data->rvo + data->currvid) > reqvid)) {
355 if (data->currvid == maxvid) { 359 if (data->currvid == maxvid) {
356 rvosteps = 0; 360 rvosteps = 0;
357 } else { 361 } else {
@@ -384,13 +388,6 @@ static int core_frequency_transition(struct powernow_k8_data *data, u32 reqfid)
384 u32 vcoreqfid, vcocurrfid, vcofiddiff; 388 u32 vcoreqfid, vcocurrfid, vcofiddiff;
385 u32 fid_interval, savevid = data->currvid; 389 u32 fid_interval, savevid = data->currvid;
386 390
387 if ((reqfid < HI_FID_TABLE_BOTTOM) &&
388 (data->currfid < HI_FID_TABLE_BOTTOM)) {
389 printk(KERN_ERR PFX "ph2: illegal lo-lo transition "
390 "0x%x 0x%x\n", reqfid, data->currfid);
391 return 1;
392 }
393
394 if (data->currfid == reqfid) { 391 if (data->currfid == reqfid) {
395 printk(KERN_ERR PFX "ph2 null fid transition 0x%x\n", 392 printk(KERN_ERR PFX "ph2 null fid transition 0x%x\n",
396 data->currfid); 393 data->currfid);
@@ -407,6 +404,9 @@ static int core_frequency_transition(struct powernow_k8_data *data, u32 reqfid)
407 vcofiddiff = vcocurrfid > vcoreqfid ? vcocurrfid - vcoreqfid 404 vcofiddiff = vcocurrfid > vcoreqfid ? vcocurrfid - vcoreqfid
408 : vcoreqfid - vcocurrfid; 405 : vcoreqfid - vcocurrfid;
409 406
407 if ((reqfid <= LO_FID_TABLE_TOP) && (data->currfid <= LO_FID_TABLE_TOP))
408 vcofiddiff = 0;
409
410 while (vcofiddiff > 2) { 410 while (vcofiddiff > 2) {
411 (data->currfid & 1) ? (fid_interval = 1) : (fid_interval = 2); 411 (data->currfid & 1) ? (fid_interval = 1) : (fid_interval = 2);
412 412
@@ -1081,14 +1081,6 @@ static int transition_frequency_fidvid(struct powernow_k8_data *data,
1081 return 0; 1081 return 0;
1082 } 1082 }
1083 1083
1084 if ((fid < HI_FID_TABLE_BOTTOM) &&
1085 (data->currfid < HI_FID_TABLE_BOTTOM)) {
1086 printk(KERN_ERR PFX
1087 "ignoring illegal change in lo freq table-%x to 0x%x\n",
1088 data->currfid, fid);
1089 return 1;
1090 }
1091
1092 dprintk("cpu %d, changing to fid 0x%x, vid 0x%x\n", 1084 dprintk("cpu %d, changing to fid 0x%x, vid 0x%x\n",
1093 smp_processor_id(), fid, vid); 1085 smp_processor_id(), fid, vid);
1094 freqs.old = find_khz_freq_from_fid(data->currfid); 1086 freqs.old = find_khz_freq_from_fid(data->currfid);
@@ -1267,7 +1259,7 @@ static int __cpuinit powernowk8_cpu_init(struct cpufreq_policy *pol)
1267{ 1259{
1268 static const char ACPI_PSS_BIOS_BUG_MSG[] = 1260 static const char ACPI_PSS_BIOS_BUG_MSG[] =
1269 KERN_ERR FW_BUG PFX "No compatible ACPI _PSS objects found.\n" 1261 KERN_ERR FW_BUG PFX "No compatible ACPI _PSS objects found.\n"
1270 KERN_ERR FW_BUG PFX "Try again with latest BIOS.\n"; 1262 FW_BUG PFX "Try again with latest BIOS.\n";
1271 struct powernow_k8_data *data; 1263 struct powernow_k8_data *data;
1272 struct init_on_cpu init_on_cpu; 1264 struct init_on_cpu init_on_cpu;
1273 int rc; 1265 int rc;
diff --git a/arch/x86/kernel/cpu/cpufreq/powernow-k8.h b/arch/x86/kernel/cpu/cpufreq/powernow-k8.h
index c9c1190b5e1f..02ce824073cb 100644
--- a/arch/x86/kernel/cpu/cpufreq/powernow-k8.h
+++ b/arch/x86/kernel/cpu/cpufreq/powernow-k8.h
@@ -215,7 +215,8 @@ struct pst_s {
215 215
216#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, "powernow-k8", msg) 216#define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, "powernow-k8", msg)
217 217
218static int core_voltage_pre_transition(struct powernow_k8_data *data, u32 reqvid); 218static int core_voltage_pre_transition(struct powernow_k8_data *data,
219 u32 reqvid, u32 regfid);
219static int core_voltage_post_transition(struct powernow_k8_data *data, u32 reqvid); 220static int core_voltage_post_transition(struct powernow_k8_data *data, u32 reqvid);
220static int core_frequency_transition(struct powernow_k8_data *data, u32 reqfid); 221static int core_frequency_transition(struct powernow_k8_data *data, u32 reqfid);
221 222
diff --git a/arch/x86/kernel/cpu/mcheck/mce.c b/arch/x86/kernel/cpu/mcheck/mce.c
index 284d1de968bc..484c1e5f658e 100644
--- a/arch/x86/kernel/cpu/mcheck/mce.c
+++ b/arch/x86/kernel/cpu/mcheck/mce.c
@@ -194,14 +194,14 @@ static void print_mce(struct mce *m)
194 m->cs, m->ip); 194 m->cs, m->ip);
195 if (m->cs == __KERNEL_CS) 195 if (m->cs == __KERNEL_CS)
196 print_symbol("{%s}", m->ip); 196 print_symbol("{%s}", m->ip);
197 printk("\n"); 197 printk(KERN_CONT "\n");
198 } 198 }
199 printk(KERN_EMERG "TSC %llx ", m->tsc); 199 printk(KERN_EMERG "TSC %llx ", m->tsc);
200 if (m->addr) 200 if (m->addr)
201 printk("ADDR %llx ", m->addr); 201 printk(KERN_CONT "ADDR %llx ", m->addr);
202 if (m->misc) 202 if (m->misc)
203 printk("MISC %llx ", m->misc); 203 printk(KERN_CONT "MISC %llx ", m->misc);
204 printk("\n"); 204 printk(KERN_CONT "\n");
205 printk(KERN_EMERG "PROCESSOR %u:%x TIME %llu SOCKET %u APIC %x\n", 205 printk(KERN_EMERG "PROCESSOR %u:%x TIME %llu SOCKET %u APIC %x\n",
206 m->cpuvendor, m->cpuid, m->time, m->socketid, 206 m->cpuvendor, m->cpuid, m->time, m->socketid,
207 m->apicid); 207 m->apicid);
@@ -209,13 +209,13 @@ static void print_mce(struct mce *m)
209 209
210static void print_mce_head(void) 210static void print_mce_head(void)
211{ 211{
212 printk(KERN_EMERG "\n" KERN_EMERG "HARDWARE ERROR\n"); 212 printk(KERN_EMERG "\nHARDWARE ERROR\n");
213} 213}
214 214
215static void print_mce_tail(void) 215static void print_mce_tail(void)
216{ 216{
217 printk(KERN_EMERG "This is not a software problem!\n" 217 printk(KERN_EMERG "This is not a software problem!\n"
218 KERN_EMERG "Run through mcelog --ascii to decode and contact your hardware vendor\n"); 218 "Run through mcelog --ascii to decode and contact your hardware vendor\n");
219} 219}
220 220
221#define PANIC_TIMEOUT 5 /* 5 seconds */ 221#define PANIC_TIMEOUT 5 /* 5 seconds */
@@ -1117,7 +1117,7 @@ static void mcheck_timer(unsigned long data)
1117 *n = min(*n*2, (int)round_jiffies_relative(check_interval*HZ)); 1117 *n = min(*n*2, (int)round_jiffies_relative(check_interval*HZ));
1118 1118
1119 t->expires = jiffies + *n; 1119 t->expires = jiffies + *n;
1120 add_timer(t); 1120 add_timer_on(t, smp_processor_id());
1121} 1121}
1122 1122
1123static void mce_do_trigger(struct work_struct *work) 1123static void mce_do_trigger(struct work_struct *work)
@@ -1321,7 +1321,7 @@ static void mce_init_timer(void)
1321 return; 1321 return;
1322 setup_timer(t, mcheck_timer, smp_processor_id()); 1322 setup_timer(t, mcheck_timer, smp_processor_id());
1323 t->expires = round_jiffies(jiffies + *n); 1323 t->expires = round_jiffies(jiffies + *n);
1324 add_timer(t); 1324 add_timer_on(t, smp_processor_id());
1325} 1325}
1326 1326
1327/* 1327/*
diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c
index 76dfef23f789..36c3dc7b8991 100644
--- a/arch/x86/kernel/cpu/perf_counter.c
+++ b/arch/x86/kernel/cpu/perf_counter.c
@@ -401,7 +401,7 @@ static const u64 amd_hw_cache_event_ids
401 [ C(RESULT_MISS) ] = 0x0041, /* Data Cache Misses */ 401 [ C(RESULT_MISS) ] = 0x0041, /* Data Cache Misses */
402 }, 402 },
403 [ C(OP_WRITE) ] = { 403 [ C(OP_WRITE) ] = {
404 [ C(RESULT_ACCESS) ] = 0x0042, /* Data Cache Refills from L2 */ 404 [ C(RESULT_ACCESS) ] = 0x0142, /* Data Cache Refills :system */
405 [ C(RESULT_MISS) ] = 0, 405 [ C(RESULT_MISS) ] = 0,
406 }, 406 },
407 [ C(OP_PREFETCH) ] = { 407 [ C(OP_PREFETCH) ] = {
@@ -912,6 +912,8 @@ x86_perf_counter_set_period(struct perf_counter *counter,
912 err = checking_wrmsrl(hwc->counter_base + idx, 912 err = checking_wrmsrl(hwc->counter_base + idx,
913 (u64)(-left) & x86_pmu.counter_mask); 913 (u64)(-left) & x86_pmu.counter_mask);
914 914
915 perf_counter_update_userpage(counter);
916
915 return ret; 917 return ret;
916} 918}
917 919
@@ -969,13 +971,6 @@ fixed_mode_idx(struct perf_counter *counter, struct hw_perf_counter *hwc)
969 if (!x86_pmu.num_counters_fixed) 971 if (!x86_pmu.num_counters_fixed)
970 return -1; 972 return -1;
971 973
972 /*
973 * Quirk, IA32_FIXED_CTRs do not work on current Atom processors:
974 */
975 if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL &&
976 boot_cpu_data.x86_model == 28)
977 return -1;
978
979 event = hwc->config & ARCH_PERFMON_EVENT_MASK; 974 event = hwc->config & ARCH_PERFMON_EVENT_MASK;
980 975
981 if (unlikely(event == x86_pmu.event_map(PERF_COUNT_HW_INSTRUCTIONS))) 976 if (unlikely(event == x86_pmu.event_map(PERF_COUNT_HW_INSTRUCTIONS)))
@@ -1041,6 +1036,8 @@ try_generic:
1041 x86_perf_counter_set_period(counter, hwc, idx); 1036 x86_perf_counter_set_period(counter, hwc, idx);
1042 x86_pmu.enable(hwc, idx); 1037 x86_pmu.enable(hwc, idx);
1043 1038
1039 perf_counter_update_userpage(counter);
1040
1044 return 0; 1041 return 0;
1045} 1042}
1046 1043
@@ -1133,6 +1130,8 @@ static void x86_pmu_disable(struct perf_counter *counter)
1133 x86_perf_counter_update(counter, hwc, idx); 1130 x86_perf_counter_update(counter, hwc, idx);
1134 cpuc->counters[idx] = NULL; 1131 cpuc->counters[idx] = NULL;
1135 clear_bit(idx, cpuc->used_mask); 1132 clear_bit(idx, cpuc->used_mask);
1133
1134 perf_counter_update_userpage(counter);
1136} 1135}
1137 1136
1138/* 1137/*
@@ -1428,8 +1427,6 @@ static int intel_pmu_init(void)
1428 */ 1427 */
1429 x86_pmu.num_counters_fixed = max((int)edx.split.num_counters_fixed, 3); 1428 x86_pmu.num_counters_fixed = max((int)edx.split.num_counters_fixed, 3);
1430 1429
1431 rdmsrl(MSR_CORE_PERF_GLOBAL_CTRL, x86_pmu.intel_ctrl);
1432
1433 /* 1430 /*
1434 * Install the hw-cache-events table: 1431 * Install the hw-cache-events table:
1435 */ 1432 */
@@ -1499,21 +1496,22 @@ void __init init_hw_perf_counters(void)
1499 pr_cont("%s PMU driver.\n", x86_pmu.name); 1496 pr_cont("%s PMU driver.\n", x86_pmu.name);
1500 1497
1501 if (x86_pmu.num_counters > X86_PMC_MAX_GENERIC) { 1498 if (x86_pmu.num_counters > X86_PMC_MAX_GENERIC) {
1502 x86_pmu.num_counters = X86_PMC_MAX_GENERIC;
1503 WARN(1, KERN_ERR "hw perf counters %d > max(%d), clipping!", 1499 WARN(1, KERN_ERR "hw perf counters %d > max(%d), clipping!",
1504 x86_pmu.num_counters, X86_PMC_MAX_GENERIC); 1500 x86_pmu.num_counters, X86_PMC_MAX_GENERIC);
1501 x86_pmu.num_counters = X86_PMC_MAX_GENERIC;
1505 } 1502 }
1506 perf_counter_mask = (1 << x86_pmu.num_counters) - 1; 1503 perf_counter_mask = (1 << x86_pmu.num_counters) - 1;
1507 perf_max_counters = x86_pmu.num_counters; 1504 perf_max_counters = x86_pmu.num_counters;
1508 1505
1509 if (x86_pmu.num_counters_fixed > X86_PMC_MAX_FIXED) { 1506 if (x86_pmu.num_counters_fixed > X86_PMC_MAX_FIXED) {
1510 x86_pmu.num_counters_fixed = X86_PMC_MAX_FIXED;
1511 WARN(1, KERN_ERR "hw perf counters fixed %d > max(%d), clipping!", 1507 WARN(1, KERN_ERR "hw perf counters fixed %d > max(%d), clipping!",
1512 x86_pmu.num_counters_fixed, X86_PMC_MAX_FIXED); 1508 x86_pmu.num_counters_fixed, X86_PMC_MAX_FIXED);
1509 x86_pmu.num_counters_fixed = X86_PMC_MAX_FIXED;
1513 } 1510 }
1514 1511
1515 perf_counter_mask |= 1512 perf_counter_mask |=
1516 ((1LL << x86_pmu.num_counters_fixed)-1) << X86_PMC_IDX_FIXED; 1513 ((1LL << x86_pmu.num_counters_fixed)-1) << X86_PMC_IDX_FIXED;
1514 x86_pmu.intel_ctrl = perf_counter_mask;
1517 1515
1518 perf_counters_lapic_init(); 1516 perf_counters_lapic_init();
1519 register_die_notifier(&perf_counter_nmi_notifier); 1517 register_die_notifier(&perf_counter_nmi_notifier);
@@ -1563,6 +1561,7 @@ void callchain_store(struct perf_callchain_entry *entry, u64 ip)
1563 1561
1564static DEFINE_PER_CPU(struct perf_callchain_entry, irq_entry); 1562static DEFINE_PER_CPU(struct perf_callchain_entry, irq_entry);
1565static DEFINE_PER_CPU(struct perf_callchain_entry, nmi_entry); 1563static DEFINE_PER_CPU(struct perf_callchain_entry, nmi_entry);
1564static DEFINE_PER_CPU(int, in_nmi_frame);
1566 1565
1567 1566
1568static void 1567static void
@@ -1578,7 +1577,9 @@ static void backtrace_warning(void *data, char *msg)
1578 1577
1579static int backtrace_stack(void *data, char *name) 1578static int backtrace_stack(void *data, char *name)
1580{ 1579{
1581 /* Process all stacks: */ 1580 per_cpu(in_nmi_frame, smp_processor_id()) =
1581 x86_is_stack_id(NMI_STACK, name);
1582
1582 return 0; 1583 return 0;
1583} 1584}
1584 1585
@@ -1586,6 +1587,9 @@ static void backtrace_address(void *data, unsigned long addr, int reliable)
1586{ 1587{
1587 struct perf_callchain_entry *entry = data; 1588 struct perf_callchain_entry *entry = data;
1588 1589
1590 if (per_cpu(in_nmi_frame, smp_processor_id()))
1591 return;
1592
1589 if (reliable) 1593 if (reliable)
1590 callchain_store(entry, addr); 1594 callchain_store(entry, addr);
1591} 1595}
diff --git a/arch/x86/kernel/cpu/perfctr-watchdog.c b/arch/x86/kernel/cpu/perfctr-watchdog.c
index 5c481f6205bf..e60ed740d2b3 100644
--- a/arch/x86/kernel/cpu/perfctr-watchdog.c
+++ b/arch/x86/kernel/cpu/perfctr-watchdog.c
@@ -803,8 +803,3 @@ int __kprobes lapic_wd_event(unsigned nmi_hz)
803 wd_ops->rearm(wd, nmi_hz); 803 wd_ops->rearm(wd, nmi_hz);
804 return 1; 804 return 1;
805} 805}
806
807int lapic_watchdog_ok(void)
808{
809 return wd_ops != NULL;
810}
diff --git a/arch/x86/kernel/dumpstack.c b/arch/x86/kernel/dumpstack.c
index 95ea5fa7d444..c8405718a4c3 100644
--- a/arch/x86/kernel/dumpstack.c
+++ b/arch/x86/kernel/dumpstack.c
@@ -22,6 +22,7 @@
22#include "dumpstack.h" 22#include "dumpstack.h"
23 23
24int panic_on_unrecovered_nmi; 24int panic_on_unrecovered_nmi;
25int panic_on_io_nmi;
25unsigned int code_bytes = 64; 26unsigned int code_bytes = 64;
26int kstack_depth_to_print = 3 * STACKSLOTS_PER_LINE; 27int kstack_depth_to_print = 3 * STACKSLOTS_PER_LINE;
27static int die_counter; 28static int die_counter;
diff --git a/arch/x86/kernel/dumpstack_32.c b/arch/x86/kernel/dumpstack_32.c
index d593cd1f58dc..bca5fba91c9e 100644
--- a/arch/x86/kernel/dumpstack_32.c
+++ b/arch/x86/kernel/dumpstack_32.c
@@ -19,6 +19,12 @@
19 19
20#include "dumpstack.h" 20#include "dumpstack.h"
21 21
22/* Just a stub for now */
23int x86_is_stack_id(int id, char *name)
24{
25 return 0;
26}
27
22void dump_trace(struct task_struct *task, struct pt_regs *regs, 28void dump_trace(struct task_struct *task, struct pt_regs *regs,
23 unsigned long *stack, unsigned long bp, 29 unsigned long *stack, unsigned long bp,
24 const struct stacktrace_ops *ops, void *data) 30 const struct stacktrace_ops *ops, void *data)
diff --git a/arch/x86/kernel/dumpstack_64.c b/arch/x86/kernel/dumpstack_64.c
index d35db5993fd6..54b0a3276766 100644
--- a/arch/x86/kernel/dumpstack_64.c
+++ b/arch/x86/kernel/dumpstack_64.c
@@ -19,10 +19,8 @@
19 19
20#include "dumpstack.h" 20#include "dumpstack.h"
21 21
22static unsigned long *in_exception_stack(unsigned cpu, unsigned long stack, 22
23 unsigned *usedp, char **idp) 23static char x86_stack_ids[][8] = {
24{
25 static char ids[][8] = {
26 [DEBUG_STACK - 1] = "#DB", 24 [DEBUG_STACK - 1] = "#DB",
27 [NMI_STACK - 1] = "NMI", 25 [NMI_STACK - 1] = "NMI",
28 [DOUBLEFAULT_STACK - 1] = "#DF", 26 [DOUBLEFAULT_STACK - 1] = "#DF",
@@ -33,6 +31,15 @@ static unsigned long *in_exception_stack(unsigned cpu, unsigned long stack,
33 N_EXCEPTION_STACKS + DEBUG_STKSZ / EXCEPTION_STKSZ - 2] = "#DB[?]" 31 N_EXCEPTION_STACKS + DEBUG_STKSZ / EXCEPTION_STKSZ - 2] = "#DB[?]"
34#endif 32#endif
35 }; 33 };
34
35int x86_is_stack_id(int id, char *name)
36{
37 return x86_stack_ids[id - 1] == name;
38}
39
40static unsigned long *in_exception_stack(unsigned cpu, unsigned long stack,
41 unsigned *usedp, char **idp)
42{
36 unsigned k; 43 unsigned k;
37 44
38 /* 45 /*
@@ -61,7 +68,7 @@ static unsigned long *in_exception_stack(unsigned cpu, unsigned long stack,
61 if (*usedp & (1U << k)) 68 if (*usedp & (1U << k))
62 break; 69 break;
63 *usedp |= 1U << k; 70 *usedp |= 1U << k;
64 *idp = ids[k]; 71 *idp = x86_stack_ids[k];
65 return (unsigned long *)end; 72 return (unsigned long *)end;
66 } 73 }
67 /* 74 /*
@@ -81,12 +88,13 @@ static unsigned long *in_exception_stack(unsigned cpu, unsigned long stack,
81 do { 88 do {
82 ++j; 89 ++j;
83 end -= EXCEPTION_STKSZ; 90 end -= EXCEPTION_STKSZ;
84 ids[j][4] = '1' + (j - N_EXCEPTION_STACKS); 91 x86_stack_ids[j][4] = '1' +
92 (j - N_EXCEPTION_STACKS);
85 } while (stack < end - EXCEPTION_STKSZ); 93 } while (stack < end - EXCEPTION_STKSZ);
86 if (*usedp & (1U << j)) 94 if (*usedp & (1U << j))
87 break; 95 break;
88 *usedp |= 1U << j; 96 *usedp |= 1U << j;
89 *idp = ids[j]; 97 *idp = x86_stack_ids[j];
90 return (unsigned long *)end; 98 return (unsigned long *)end;
91 } 99 }
92#endif 100#endif
diff --git a/arch/x86/kernel/e820.c b/arch/x86/kernel/e820.c
index 7271fa33d791..5cb5725b2bae 100644
--- a/arch/x86/kernel/e820.c
+++ b/arch/x86/kernel/e820.c
@@ -627,10 +627,9 @@ __init void e820_setup_gap(void)
627#ifdef CONFIG_X86_64 627#ifdef CONFIG_X86_64
628 if (!found) { 628 if (!found) {
629 gapstart = (max_pfn << PAGE_SHIFT) + 1024*1024; 629 gapstart = (max_pfn << PAGE_SHIFT) + 1024*1024;
630 printk(KERN_ERR "PCI: Warning: Cannot find a gap in the 32bit " 630 printk(KERN_ERR
631 "address range\n" 631 "PCI: Warning: Cannot find a gap in the 32bit address range\n"
632 KERN_ERR "PCI: Unassigned devices with 32bit resource " 632 "PCI: Unassigned devices with 32bit resource registers may break!\n");
633 "registers may break!\n");
634 } 633 }
635#endif 634#endif
636 635
@@ -1383,6 +1382,8 @@ static unsigned long ram_alignment(resource_size_t pos)
1383 return 32*1024*1024; 1382 return 32*1024*1024;
1384} 1383}
1385 1384
1385#define MAX_RESOURCE_SIZE ((resource_size_t)-1)
1386
1386void __init e820_reserve_resources_late(void) 1387void __init e820_reserve_resources_late(void)
1387{ 1388{
1388 int i; 1389 int i;
@@ -1400,17 +1401,19 @@ void __init e820_reserve_resources_late(void)
1400 * avoid stolen RAM: 1401 * avoid stolen RAM:
1401 */ 1402 */
1402 for (i = 0; i < e820.nr_map; i++) { 1403 for (i = 0; i < e820.nr_map; i++) {
1403 struct e820entry *entry = &e820_saved.map[i]; 1404 struct e820entry *entry = &e820.map[i];
1404 resource_size_t start, end; 1405 u64 start, end;
1405 1406
1406 if (entry->type != E820_RAM) 1407 if (entry->type != E820_RAM)
1407 continue; 1408 continue;
1408 start = entry->addr + entry->size; 1409 start = entry->addr + entry->size;
1409 end = round_up(start, ram_alignment(start)); 1410 end = round_up(start, ram_alignment(start)) - 1;
1410 if (start == end) 1411 if (end > MAX_RESOURCE_SIZE)
1412 end = MAX_RESOURCE_SIZE;
1413 if (start >= end)
1411 continue; 1414 continue;
1412 reserve_region_with_split(&iomem_resource, start, 1415 reserve_region_with_split(&iomem_resource, start, end,
1413 end - 1, "RAM buffer"); 1416 "RAM buffer");
1414 } 1417 }
1415} 1418}
1416 1419
diff --git a/arch/x86/kernel/kvm.c b/arch/x86/kernel/kvm.c
index a78ecad0c900..c664d515f613 100644
--- a/arch/x86/kernel/kvm.c
+++ b/arch/x86/kernel/kvm.c
@@ -200,7 +200,7 @@ static void kvm_leave_lazy_mmu(void)
200 state->mode = paravirt_get_lazy_mode(); 200 state->mode = paravirt_get_lazy_mode();
201} 201}
202 202
203static void paravirt_ops_setup(void) 203static void __init paravirt_ops_setup(void)
204{ 204{
205 pv_info.name = "KVM"; 205 pv_info.name = "KVM";
206 pv_info.paravirt_enabled = 1; 206 pv_info.paravirt_enabled = 1;
diff --git a/arch/x86/kernel/pci-dma.c b/arch/x86/kernel/pci-dma.c
index 47630479b067..1a041bcf506b 100644
--- a/arch/x86/kernel/pci-dma.c
+++ b/arch/x86/kernel/pci-dma.c
@@ -211,11 +211,11 @@ static __init int iommu_setup(char *p)
211#ifdef CONFIG_SWIOTLB 211#ifdef CONFIG_SWIOTLB
212 if (!strncmp(p, "soft", 4)) 212 if (!strncmp(p, "soft", 4))
213 swiotlb = 1; 213 swiotlb = 1;
214#endif
214 if (!strncmp(p, "pt", 2)) { 215 if (!strncmp(p, "pt", 2)) {
215 iommu_pass_through = 1; 216 iommu_pass_through = 1;
216 return 1; 217 return 1;
217 } 218 }
218#endif
219 219
220 gart_parse_options(p); 220 gart_parse_options(p);
221 221
diff --git a/arch/x86/kernel/pci-gart_64.c b/arch/x86/kernel/pci-gart_64.c
index cfd9f9063896..d2e56b8f48e7 100644
--- a/arch/x86/kernel/pci-gart_64.c
+++ b/arch/x86/kernel/pci-gart_64.c
@@ -675,7 +675,7 @@ static __init int init_k8_gatt(struct agp_kern_info *info)
675 nommu: 675 nommu:
676 /* Should not happen anymore */ 676 /* Should not happen anymore */
677 printk(KERN_WARNING "PCI-DMA: More than 4GB of RAM and no IOMMU\n" 677 printk(KERN_WARNING "PCI-DMA: More than 4GB of RAM and no IOMMU\n"
678 KERN_WARNING "falling back to iommu=soft.\n"); 678 "falling back to iommu=soft.\n");
679 return -1; 679 return -1;
680} 680}
681 681
diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c
index be5ae80f897f..de2cab132844 100644
--- a/arch/x86/kernel/setup.c
+++ b/arch/x86/kernel/setup.c
@@ -289,6 +289,20 @@ void * __init extend_brk(size_t size, size_t align)
289 return ret; 289 return ret;
290} 290}
291 291
292#ifdef CONFIG_X86_64
293static void __init init_gbpages(void)
294{
295 if (direct_gbpages && cpu_has_gbpages)
296 printk(KERN_INFO "Using GB pages for direct mapping\n");
297 else
298 direct_gbpages = 0;
299}
300#else
301static inline void init_gbpages(void)
302{
303}
304#endif
305
292static void __init reserve_brk(void) 306static void __init reserve_brk(void)
293{ 307{
294 if (_brk_end > _brk_start) 308 if (_brk_end > _brk_start)
@@ -871,6 +885,8 @@ void __init setup_arch(char **cmdline_p)
871 885
872 reserve_brk(); 886 reserve_brk();
873 887
888 init_gbpages();
889
874 /* max_pfn_mapped is updated here */ 890 /* max_pfn_mapped is updated here */
875 max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT); 891 max_low_pfn_mapped = init_memory_mapping(0, max_low_pfn<<PAGE_SHIFT);
876 max_pfn_mapped = max_low_pfn_mapped; 892 max_pfn_mapped = max_low_pfn_mapped;
diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c
index 9c3f0823e6aa..29a3eef7cf4a 100644
--- a/arch/x86/kernel/setup_percpu.c
+++ b/arch/x86/kernel/setup_percpu.c
@@ -124,7 +124,7 @@ static void * __init pcpu_alloc_bootmem(unsigned int cpu, unsigned long size,
124} 124}
125 125
126/* 126/*
127 * Remap allocator 127 * Large page remap allocator
128 * 128 *
129 * This allocator uses PMD page as unit. A PMD page is allocated for 129 * This allocator uses PMD page as unit. A PMD page is allocated for
130 * each cpu and each is remapped into vmalloc area using PMD mapping. 130 * each cpu and each is remapped into vmalloc area using PMD mapping.
@@ -137,105 +137,185 @@ static void * __init pcpu_alloc_bootmem(unsigned int cpu, unsigned long size,
137 * better than only using 4k mappings while still being NUMA friendly. 137 * better than only using 4k mappings while still being NUMA friendly.
138 */ 138 */
139#ifdef CONFIG_NEED_MULTIPLE_NODES 139#ifdef CONFIG_NEED_MULTIPLE_NODES
140static size_t pcpur_size __initdata; 140struct pcpul_ent {
141static void **pcpur_ptrs __initdata; 141 unsigned int cpu;
142 void *ptr;
143};
144
145static size_t pcpul_size;
146static struct pcpul_ent *pcpul_map;
147static struct vm_struct pcpul_vm;
142 148
143static struct page * __init pcpur_get_page(unsigned int cpu, int pageno) 149static struct page * __init pcpul_get_page(unsigned int cpu, int pageno)
144{ 150{
145 size_t off = (size_t)pageno << PAGE_SHIFT; 151 size_t off = (size_t)pageno << PAGE_SHIFT;
146 152
147 if (off >= pcpur_size) 153 if (off >= pcpul_size)
148 return NULL; 154 return NULL;
149 155
150 return virt_to_page(pcpur_ptrs[cpu] + off); 156 return virt_to_page(pcpul_map[cpu].ptr + off);
151} 157}
152 158
153static ssize_t __init setup_pcpu_remap(size_t static_size) 159static ssize_t __init setup_pcpu_lpage(size_t static_size, bool chosen)
154{ 160{
155 static struct vm_struct vm; 161 size_t map_size, dyn_size;
156 size_t ptrs_size, dyn_size;
157 unsigned int cpu; 162 unsigned int cpu;
163 int i, j;
158 ssize_t ret; 164 ssize_t ret;
159 165
160 /* 166 if (!chosen) {
161 * If large page isn't supported, there's no benefit in doing 167 size_t vm_size = VMALLOC_END - VMALLOC_START;
162 * this. Also, on non-NUMA, embedding is better. 168 size_t tot_size = num_possible_cpus() * PMD_SIZE;
163 * 169
164 * NOTE: disabled for now. 170 /* on non-NUMA, embedding is better */
165 */ 171 if (!pcpu_need_numa())
166 if (true || !cpu_has_pse || !pcpu_need_numa()) 172 return -EINVAL;
173
174 /* don't consume more than 20% of vmalloc area */
175 if (tot_size > vm_size / 5) {
176 pr_info("PERCPU: too large chunk size %zuMB for "
177 "large page remap\n", tot_size >> 20);
178 return -EINVAL;
179 }
180 }
181
182 /* need PSE */
183 if (!cpu_has_pse) {
184 pr_warning("PERCPU: lpage allocator requires PSE\n");
167 return -EINVAL; 185 return -EINVAL;
186 }
168 187
169 /* 188 /*
170 * Currently supports only single page. Supporting multiple 189 * Currently supports only single page. Supporting multiple
171 * pages won't be too difficult if it ever becomes necessary. 190 * pages won't be too difficult if it ever becomes necessary.
172 */ 191 */
173 pcpur_size = PFN_ALIGN(static_size + PERCPU_MODULE_RESERVE + 192 pcpul_size = PFN_ALIGN(static_size + PERCPU_MODULE_RESERVE +
174 PERCPU_DYNAMIC_RESERVE); 193 PERCPU_DYNAMIC_RESERVE);
175 if (pcpur_size > PMD_SIZE) { 194 if (pcpul_size > PMD_SIZE) {
176 pr_warning("PERCPU: static data is larger than large page, " 195 pr_warning("PERCPU: static data is larger than large page, "
177 "can't use large page\n"); 196 "can't use large page\n");
178 return -EINVAL; 197 return -EINVAL;
179 } 198 }
180 dyn_size = pcpur_size - static_size - PERCPU_FIRST_CHUNK_RESERVE; 199 dyn_size = pcpul_size - static_size - PERCPU_FIRST_CHUNK_RESERVE;
181 200
182 /* allocate pointer array and alloc large pages */ 201 /* allocate pointer array and alloc large pages */
183 ptrs_size = PFN_ALIGN(num_possible_cpus() * sizeof(pcpur_ptrs[0])); 202 map_size = PFN_ALIGN(num_possible_cpus() * sizeof(pcpul_map[0]));
184 pcpur_ptrs = alloc_bootmem(ptrs_size); 203 pcpul_map = alloc_bootmem(map_size);
185 204
186 for_each_possible_cpu(cpu) { 205 for_each_possible_cpu(cpu) {
187 pcpur_ptrs[cpu] = pcpu_alloc_bootmem(cpu, PMD_SIZE, PMD_SIZE); 206 pcpul_map[cpu].cpu = cpu;
188 if (!pcpur_ptrs[cpu]) 207 pcpul_map[cpu].ptr = pcpu_alloc_bootmem(cpu, PMD_SIZE,
208 PMD_SIZE);
209 if (!pcpul_map[cpu].ptr) {
210 pr_warning("PERCPU: failed to allocate large page "
211 "for cpu%u\n", cpu);
189 goto enomem; 212 goto enomem;
213 }
190 214
191 /* 215 /*
192 * Only use pcpur_size bytes and give back the rest. 216 * Only use pcpul_size bytes and give back the rest.
193 * 217 *
194 * Ingo: The 2MB up-rounding bootmem is needed to make 218 * Ingo: The 2MB up-rounding bootmem is needed to make
195 * sure the partial 2MB page is still fully RAM - it's 219 * sure the partial 2MB page is still fully RAM - it's
196 * not well-specified to have a PAT-incompatible area 220 * not well-specified to have a PAT-incompatible area
197 * (unmapped RAM, device memory, etc.) in that hole. 221 * (unmapped RAM, device memory, etc.) in that hole.
198 */ 222 */
199 free_bootmem(__pa(pcpur_ptrs[cpu] + pcpur_size), 223 free_bootmem(__pa(pcpul_map[cpu].ptr + pcpul_size),
200 PMD_SIZE - pcpur_size); 224 PMD_SIZE - pcpul_size);
201 225
202 memcpy(pcpur_ptrs[cpu], __per_cpu_load, static_size); 226 memcpy(pcpul_map[cpu].ptr, __per_cpu_load, static_size);
203 } 227 }
204 228
205 /* allocate address and map */ 229 /* allocate address and map */
206 vm.flags = VM_ALLOC; 230 pcpul_vm.flags = VM_ALLOC;
207 vm.size = num_possible_cpus() * PMD_SIZE; 231 pcpul_vm.size = num_possible_cpus() * PMD_SIZE;
208 vm_area_register_early(&vm, PMD_SIZE); 232 vm_area_register_early(&pcpul_vm, PMD_SIZE);
209 233
210 for_each_possible_cpu(cpu) { 234 for_each_possible_cpu(cpu) {
211 pmd_t *pmd; 235 pmd_t *pmd, pmd_v;
212 236
213 pmd = populate_extra_pmd((unsigned long)vm.addr 237 pmd = populate_extra_pmd((unsigned long)pcpul_vm.addr +
214 + cpu * PMD_SIZE); 238 cpu * PMD_SIZE);
215 set_pmd(pmd, pfn_pmd(page_to_pfn(virt_to_page(pcpur_ptrs[cpu])), 239 pmd_v = pfn_pmd(page_to_pfn(virt_to_page(pcpul_map[cpu].ptr)),
216 PAGE_KERNEL_LARGE)); 240 PAGE_KERNEL_LARGE);
241 set_pmd(pmd, pmd_v);
217 } 242 }
218 243
219 /* we're ready, commit */ 244 /* we're ready, commit */
220 pr_info("PERCPU: Remapped at %p with large pages, static data " 245 pr_info("PERCPU: Remapped at %p with large pages, static data "
221 "%zu bytes\n", vm.addr, static_size); 246 "%zu bytes\n", pcpul_vm.addr, static_size);
222 247
223 ret = pcpu_setup_first_chunk(pcpur_get_page, static_size, 248 ret = pcpu_setup_first_chunk(pcpul_get_page, static_size,
224 PERCPU_FIRST_CHUNK_RESERVE, dyn_size, 249 PERCPU_FIRST_CHUNK_RESERVE, dyn_size,
225 PMD_SIZE, vm.addr, NULL); 250 PMD_SIZE, pcpul_vm.addr, NULL);
226 goto out_free_ar; 251
252 /* sort pcpul_map array for pcpu_lpage_remapped() */
253 for (i = 0; i < num_possible_cpus() - 1; i++)
254 for (j = i + 1; j < num_possible_cpus(); j++)
255 if (pcpul_map[i].ptr > pcpul_map[j].ptr) {
256 struct pcpul_ent tmp = pcpul_map[i];
257 pcpul_map[i] = pcpul_map[j];
258 pcpul_map[j] = tmp;
259 }
260
261 return ret;
227 262
228enomem: 263enomem:
229 for_each_possible_cpu(cpu) 264 for_each_possible_cpu(cpu)
230 if (pcpur_ptrs[cpu]) 265 if (pcpul_map[cpu].ptr)
231 free_bootmem(__pa(pcpur_ptrs[cpu]), PMD_SIZE); 266 free_bootmem(__pa(pcpul_map[cpu].ptr), pcpul_size);
232 ret = -ENOMEM; 267 free_bootmem(__pa(pcpul_map), map_size);
233out_free_ar: 268 return -ENOMEM;
234 free_bootmem(__pa(pcpur_ptrs), ptrs_size); 269}
235 return ret; 270
271/**
272 * pcpu_lpage_remapped - determine whether a kaddr is in pcpul recycled area
273 * @kaddr: the kernel address in question
274 *
275 * Determine whether @kaddr falls in the pcpul recycled area. This is
276 * used by pageattr to detect VM aliases and break up the pcpu PMD
277 * mapping such that the same physical page is not mapped under
278 * different attributes.
279 *
280 * The recycled area is always at the tail of a partially used PMD
281 * page.
282 *
283 * RETURNS:
284 * Address of corresponding remapped pcpu address if match is found;
285 * otherwise, NULL.
286 */
287void *pcpu_lpage_remapped(void *kaddr)
288{
289 void *pmd_addr = (void *)((unsigned long)kaddr & PMD_MASK);
290 unsigned long offset = (unsigned long)kaddr & ~PMD_MASK;
291 int left = 0, right = num_possible_cpus() - 1;
292 int pos;
293
294 /* pcpul in use at all? */
295 if (!pcpul_map)
296 return NULL;
297
298 /* okay, perform binary search */
299 while (left <= right) {
300 pos = (left + right) / 2;
301
302 if (pcpul_map[pos].ptr < pmd_addr)
303 left = pos + 1;
304 else if (pcpul_map[pos].ptr > pmd_addr)
305 right = pos - 1;
306 else {
307 /* it shouldn't be in the area for the first chunk */
308 WARN_ON(offset < pcpul_size);
309
310 return pcpul_vm.addr +
311 pcpul_map[pos].cpu * PMD_SIZE + offset;
312 }
313 }
314
315 return NULL;
236} 316}
237#else 317#else
238static ssize_t __init setup_pcpu_remap(size_t static_size) 318static ssize_t __init setup_pcpu_lpage(size_t static_size, bool chosen)
239{ 319{
240 return -EINVAL; 320 return -EINVAL;
241} 321}
@@ -249,7 +329,7 @@ static ssize_t __init setup_pcpu_remap(size_t static_size)
249 * mapping so that it can use PMD mapping without additional TLB 329 * mapping so that it can use PMD mapping without additional TLB
250 * pressure. 330 * pressure.
251 */ 331 */
252static ssize_t __init setup_pcpu_embed(size_t static_size) 332static ssize_t __init setup_pcpu_embed(size_t static_size, bool chosen)
253{ 333{
254 size_t reserve = PERCPU_MODULE_RESERVE + PERCPU_DYNAMIC_RESERVE; 334 size_t reserve = PERCPU_MODULE_RESERVE + PERCPU_DYNAMIC_RESERVE;
255 335
@@ -258,7 +338,7 @@ static ssize_t __init setup_pcpu_embed(size_t static_size)
258 * this. Also, embedding allocation doesn't play well with 338 * this. Also, embedding allocation doesn't play well with
259 * NUMA. 339 * NUMA.
260 */ 340 */
261 if (!cpu_has_pse || pcpu_need_numa()) 341 if (!chosen && (!cpu_has_pse || pcpu_need_numa()))
262 return -EINVAL; 342 return -EINVAL;
263 343
264 return pcpu_embed_first_chunk(static_size, PERCPU_FIRST_CHUNK_RESERVE, 344 return pcpu_embed_first_chunk(static_size, PERCPU_FIRST_CHUNK_RESERVE,
@@ -308,8 +388,11 @@ static ssize_t __init setup_pcpu_4k(size_t static_size)
308 void *ptr; 388 void *ptr;
309 389
310 ptr = pcpu_alloc_bootmem(cpu, PAGE_SIZE, PAGE_SIZE); 390 ptr = pcpu_alloc_bootmem(cpu, PAGE_SIZE, PAGE_SIZE);
311 if (!ptr) 391 if (!ptr) {
392 pr_warning("PERCPU: failed to allocate "
393 "4k page for cpu%u\n", cpu);
312 goto enomem; 394 goto enomem;
395 }
313 396
314 memcpy(ptr, __per_cpu_load + i * PAGE_SIZE, PAGE_SIZE); 397 memcpy(ptr, __per_cpu_load + i * PAGE_SIZE, PAGE_SIZE);
315 pcpu4k_pages[j++] = virt_to_page(ptr); 398 pcpu4k_pages[j++] = virt_to_page(ptr);
@@ -333,6 +416,16 @@ out_free_ar:
333 return ret; 416 return ret;
334} 417}
335 418
419/* for explicit first chunk allocator selection */
420static char pcpu_chosen_alloc[16] __initdata;
421
422static int __init percpu_alloc_setup(char *str)
423{
424 strncpy(pcpu_chosen_alloc, str, sizeof(pcpu_chosen_alloc) - 1);
425 return 0;
426}
427early_param("percpu_alloc", percpu_alloc_setup);
428
336static inline void setup_percpu_segment(int cpu) 429static inline void setup_percpu_segment(int cpu)
337{ 430{
338#ifdef CONFIG_X86_32 431#ifdef CONFIG_X86_32
@@ -346,11 +439,6 @@ static inline void setup_percpu_segment(int cpu)
346#endif 439#endif
347} 440}
348 441
349/*
350 * Great future plan:
351 * Declare PDA itself and support (irqstack,tss,pgd) as per cpu data.
352 * Always point %gs to its beginning
353 */
354void __init setup_per_cpu_areas(void) 442void __init setup_per_cpu_areas(void)
355{ 443{
356 size_t static_size = __per_cpu_end - __per_cpu_start; 444 size_t static_size = __per_cpu_end - __per_cpu_start;
@@ -367,9 +455,26 @@ void __init setup_per_cpu_areas(void)
367 * of large page mappings. Please read comments on top of 455 * of large page mappings. Please read comments on top of
368 * each allocator for details. 456 * each allocator for details.
369 */ 457 */
370 ret = setup_pcpu_remap(static_size); 458 ret = -EINVAL;
371 if (ret < 0) 459 if (strlen(pcpu_chosen_alloc)) {
372 ret = setup_pcpu_embed(static_size); 460 if (strcmp(pcpu_chosen_alloc, "4k")) {
461 if (!strcmp(pcpu_chosen_alloc, "lpage"))
462 ret = setup_pcpu_lpage(static_size, true);
463 else if (!strcmp(pcpu_chosen_alloc, "embed"))
464 ret = setup_pcpu_embed(static_size, true);
465 else
466 pr_warning("PERCPU: unknown allocator %s "
467 "specified\n", pcpu_chosen_alloc);
468 if (ret < 0)
469 pr_warning("PERCPU: %s allocator failed (%zd), "
470 "falling back to 4k\n",
471 pcpu_chosen_alloc, ret);
472 }
473 } else {
474 ret = setup_pcpu_lpage(static_size, false);
475 if (ret < 0)
476 ret = setup_pcpu_embed(static_size, false);
477 }
373 if (ret < 0) 478 if (ret < 0)
374 ret = setup_pcpu_4k(static_size); 479 ret = setup_pcpu_4k(static_size);
375 if (ret < 0) 480 if (ret < 0)
diff --git a/arch/x86/kernel/tlb_uv.c b/arch/x86/kernel/tlb_uv.c
index 124d40c575df..8ccabb8a2f6a 100644
--- a/arch/x86/kernel/tlb_uv.c
+++ b/arch/x86/kernel/tlb_uv.c
@@ -711,7 +711,6 @@ uv_activation_descriptor_init(int node, int pnode)
711 unsigned long pa; 711 unsigned long pa;
712 unsigned long m; 712 unsigned long m;
713 unsigned long n; 713 unsigned long n;
714 unsigned long mmr_image;
715 struct bau_desc *adp; 714 struct bau_desc *adp;
716 struct bau_desc *ad2; 715 struct bau_desc *ad2;
717 716
@@ -727,12 +726,8 @@ uv_activation_descriptor_init(int node, int pnode)
727 n = pa >> uv_nshift; 726 n = pa >> uv_nshift;
728 m = pa & uv_mmask; 727 m = pa & uv_mmask;
729 728
730 mmr_image = uv_read_global_mmr64(pnode, UVH_LB_BAU_SB_DESCRIPTOR_BASE); 729 uv_write_global_mmr64(pnode, UVH_LB_BAU_SB_DESCRIPTOR_BASE,
731 if (mmr_image) { 730 (n << UV_DESC_BASE_PNODE_SHIFT | m));
732 uv_write_global_mmr64(pnode, (unsigned long)
733 UVH_LB_BAU_SB_DESCRIPTOR_BASE,
734 (n << UV_DESC_BASE_PNODE_SHIFT | m));
735 }
736 731
737 /* 732 /*
738 * initializing all 8 (UV_ITEMS_PER_DESCRIPTOR) descriptors for each 733 * initializing all 8 (UV_ITEMS_PER_DESCRIPTOR) descriptors for each
diff --git a/arch/x86/kernel/traps.c b/arch/x86/kernel/traps.c
index a0f48f5671c0..5204332f475d 100644
--- a/arch/x86/kernel/traps.c
+++ b/arch/x86/kernel/traps.c
@@ -346,6 +346,9 @@ io_check_error(unsigned char reason, struct pt_regs *regs)
346 printk(KERN_EMERG "NMI: IOCK error (debug interrupt?)\n"); 346 printk(KERN_EMERG "NMI: IOCK error (debug interrupt?)\n");
347 show_registers(regs); 347 show_registers(regs);
348 348
349 if (panic_on_io_nmi)
350 panic("NMI IOCK error: Not continuing");
351
349 /* Re-enable the IOCK line, wait for a few seconds */ 352 /* Re-enable the IOCK line, wait for a few seconds */
350 reason = (reason & 0xf) | 8; 353 reason = (reason & 0xf) | 8;
351 outb(reason, 0x61); 354 outb(reason, 0x61);
diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
index 5c3d6e81a7dc..7030b5f911bf 100644
--- a/arch/x86/kvm/mmu.c
+++ b/arch/x86/kvm/mmu.c
@@ -2157,7 +2157,7 @@ static void reset_rsvds_bits_mask(struct kvm_vcpu *vcpu, int level)
2157 else 2157 else
2158 /* 32 bits PSE 4MB page */ 2158 /* 32 bits PSE 4MB page */
2159 context->rsvd_bits_mask[1][1] = rsvd_bits(13, 21); 2159 context->rsvd_bits_mask[1][1] = rsvd_bits(13, 21);
2160 context->rsvd_bits_mask[1][0] = ~0ull; 2160 context->rsvd_bits_mask[1][0] = context->rsvd_bits_mask[1][0];
2161 break; 2161 break;
2162 case PT32E_ROOT_LEVEL: 2162 case PT32E_ROOT_LEVEL:
2163 context->rsvd_bits_mask[0][2] = 2163 context->rsvd_bits_mask[0][2] =
@@ -2170,7 +2170,7 @@ static void reset_rsvds_bits_mask(struct kvm_vcpu *vcpu, int level)
2170 context->rsvd_bits_mask[1][1] = exb_bit_rsvd | 2170 context->rsvd_bits_mask[1][1] = exb_bit_rsvd |
2171 rsvd_bits(maxphyaddr, 62) | 2171 rsvd_bits(maxphyaddr, 62) |
2172 rsvd_bits(13, 20); /* large page */ 2172 rsvd_bits(13, 20); /* large page */
2173 context->rsvd_bits_mask[1][0] = ~0ull; 2173 context->rsvd_bits_mask[1][0] = context->rsvd_bits_mask[1][0];
2174 break; 2174 break;
2175 case PT64_ROOT_LEVEL: 2175 case PT64_ROOT_LEVEL:
2176 context->rsvd_bits_mask[0][3] = exb_bit_rsvd | 2176 context->rsvd_bits_mask[0][3] = exb_bit_rsvd |
@@ -2186,7 +2186,7 @@ static void reset_rsvds_bits_mask(struct kvm_vcpu *vcpu, int level)
2186 context->rsvd_bits_mask[1][1] = exb_bit_rsvd | 2186 context->rsvd_bits_mask[1][1] = exb_bit_rsvd |
2187 rsvd_bits(maxphyaddr, 51) | 2187 rsvd_bits(maxphyaddr, 51) |
2188 rsvd_bits(13, 20); /* large page */ 2188 rsvd_bits(13, 20); /* large page */
2189 context->rsvd_bits_mask[1][0] = ~0ull; 2189 context->rsvd_bits_mask[1][0] = context->rsvd_bits_mask[1][0];
2190 break; 2190 break;
2191 } 2191 }
2192} 2192}
diff --git a/arch/x86/kvm/paging_tmpl.h b/arch/x86/kvm/paging_tmpl.h
index 258e4591e1ca..67785f635399 100644
--- a/arch/x86/kvm/paging_tmpl.h
+++ b/arch/x86/kvm/paging_tmpl.h
@@ -281,7 +281,7 @@ static u64 *FNAME(fetch)(struct kvm_vcpu *vcpu, gva_t addr,
281{ 281{
282 unsigned access = gw->pt_access; 282 unsigned access = gw->pt_access;
283 struct kvm_mmu_page *shadow_page; 283 struct kvm_mmu_page *shadow_page;
284 u64 spte, *sptep; 284 u64 spte, *sptep = NULL;
285 int direct; 285 int direct;
286 gfn_t table_gfn; 286 gfn_t table_gfn;
287 int r; 287 int r;
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index e770bf349ec4..356a0ce85c68 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -3012,6 +3012,12 @@ static int handle_vmcall(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
3012 return 1; 3012 return 1;
3013} 3013}
3014 3014
3015static int handle_vmx_insn(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
3016{
3017 kvm_queue_exception(vcpu, UD_VECTOR);
3018 return 1;
3019}
3020
3015static int handle_invlpg(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run) 3021static int handle_invlpg(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
3016{ 3022{
3017 unsigned long exit_qualification = vmcs_readl(EXIT_QUALIFICATION); 3023 unsigned long exit_qualification = vmcs_readl(EXIT_QUALIFICATION);
@@ -3198,6 +3204,15 @@ static int (*kvm_vmx_exit_handlers[])(struct kvm_vcpu *vcpu,
3198 [EXIT_REASON_HLT] = handle_halt, 3204 [EXIT_REASON_HLT] = handle_halt,
3199 [EXIT_REASON_INVLPG] = handle_invlpg, 3205 [EXIT_REASON_INVLPG] = handle_invlpg,
3200 [EXIT_REASON_VMCALL] = handle_vmcall, 3206 [EXIT_REASON_VMCALL] = handle_vmcall,
3207 [EXIT_REASON_VMCLEAR] = handle_vmx_insn,
3208 [EXIT_REASON_VMLAUNCH] = handle_vmx_insn,
3209 [EXIT_REASON_VMPTRLD] = handle_vmx_insn,
3210 [EXIT_REASON_VMPTRST] = handle_vmx_insn,
3211 [EXIT_REASON_VMREAD] = handle_vmx_insn,
3212 [EXIT_REASON_VMRESUME] = handle_vmx_insn,
3213 [EXIT_REASON_VMWRITE] = handle_vmx_insn,
3214 [EXIT_REASON_VMOFF] = handle_vmx_insn,
3215 [EXIT_REASON_VMON] = handle_vmx_insn,
3201 [EXIT_REASON_TPR_BELOW_THRESHOLD] = handle_tpr_below_threshold, 3216 [EXIT_REASON_TPR_BELOW_THRESHOLD] = handle_tpr_below_threshold,
3202 [EXIT_REASON_APIC_ACCESS] = handle_apic_access, 3217 [EXIT_REASON_APIC_ACCESS] = handle_apic_access,
3203 [EXIT_REASON_WBINVD] = handle_wbinvd, 3218 [EXIT_REASON_WBINVD] = handle_wbinvd,
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 249540f98513..fe5474aec41a 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -898,6 +898,7 @@ int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata)
898 case MSR_VM_HSAVE_PA: 898 case MSR_VM_HSAVE_PA:
899 case MSR_P6_EVNTSEL0: 899 case MSR_P6_EVNTSEL0:
900 case MSR_P6_EVNTSEL1: 900 case MSR_P6_EVNTSEL1:
901 case MSR_K7_EVNTSEL0:
901 data = 0; 902 data = 0;
902 break; 903 break;
903 case MSR_MTRRcap: 904 case MSR_MTRRcap:
diff --git a/arch/x86/kvm/x86_emulate.c b/arch/x86/kvm/x86_emulate.c
index c1b6c232e02b..616de4628d60 100644
--- a/arch/x86/kvm/x86_emulate.c
+++ b/arch/x86/kvm/x86_emulate.c
@@ -1361,7 +1361,7 @@ static inline int writeback(struct x86_emulate_ctxt *ctxt,
1361 return 0; 1361 return 0;
1362} 1362}
1363 1363
1364void toggle_interruptibility(struct x86_emulate_ctxt *ctxt, u32 mask) 1364static void toggle_interruptibility(struct x86_emulate_ctxt *ctxt, u32 mask)
1365{ 1365{
1366 u32 int_shadow = kvm_x86_ops->get_interrupt_shadow(ctxt->vcpu, mask); 1366 u32 int_shadow = kvm_x86_ops->get_interrupt_shadow(ctxt->vcpu, mask);
1367 /* 1367 /*
diff --git a/arch/x86/lib/Makefile b/arch/x86/lib/Makefile
index f9d35632666b..07c31899c9c2 100644
--- a/arch/x86/lib/Makefile
+++ b/arch/x86/lib/Makefile
@@ -10,6 +10,7 @@ lib-y += usercopy_$(BITS).o getuser.o putuser.o
10lib-y += memcpy_$(BITS).o 10lib-y += memcpy_$(BITS).o
11 11
12ifeq ($(CONFIG_X86_32),y) 12ifeq ($(CONFIG_X86_32),y)
13 obj-y += atomic64_32.o
13 lib-y += checksum_32.o 14 lib-y += checksum_32.o
14 lib-y += strstr_32.o 15 lib-y += strstr_32.o
15 lib-y += semaphore_32.o string_32.o 16 lib-y += semaphore_32.o string_32.o
diff --git a/arch/x86/lib/atomic64_32.c b/arch/x86/lib/atomic64_32.c
new file mode 100644
index 000000000000..824fa0be55a3
--- /dev/null
+++ b/arch/x86/lib/atomic64_32.c
@@ -0,0 +1,230 @@
1#include <linux/compiler.h>
2#include <linux/module.h>
3#include <linux/types.h>
4
5#include <asm/processor.h>
6#include <asm/cmpxchg.h>
7#include <asm/atomic.h>
8
9static noinline u64 cmpxchg8b(u64 *ptr, u64 old, u64 new)
10{
11 u32 low = new;
12 u32 high = new >> 32;
13
14 asm volatile(
15 LOCK_PREFIX "cmpxchg8b %1\n"
16 : "+A" (old), "+m" (*ptr)
17 : "b" (low), "c" (high)
18 );
19 return old;
20}
21
22u64 atomic64_cmpxchg(atomic64_t *ptr, u64 old_val, u64 new_val)
23{
24 return cmpxchg8b(&ptr->counter, old_val, new_val);
25}
26EXPORT_SYMBOL(atomic64_cmpxchg);
27
28/**
29 * atomic64_xchg - xchg atomic64 variable
30 * @ptr: pointer to type atomic64_t
31 * @new_val: value to assign
32 *
33 * Atomically xchgs the value of @ptr to @new_val and returns
34 * the old value.
35 */
36u64 atomic64_xchg(atomic64_t *ptr, u64 new_val)
37{
38 /*
39 * Try first with a (possibly incorrect) assumption about
40 * what we have there. We'll do two loops most likely,
41 * but we'll get an ownership MESI transaction straight away
42 * instead of a read transaction followed by a
43 * flush-for-ownership transaction:
44 */
45 u64 old_val, real_val = 0;
46
47 do {
48 old_val = real_val;
49
50 real_val = atomic64_cmpxchg(ptr, old_val, new_val);
51
52 } while (real_val != old_val);
53
54 return old_val;
55}
56EXPORT_SYMBOL(atomic64_xchg);
57
58/**
59 * atomic64_set - set atomic64 variable
60 * @ptr: pointer to type atomic64_t
61 * @new_val: value to assign
62 *
63 * Atomically sets the value of @ptr to @new_val.
64 */
65void atomic64_set(atomic64_t *ptr, u64 new_val)
66{
67 atomic64_xchg(ptr, new_val);
68}
69EXPORT_SYMBOL(atomic64_set);
70
71/**
72EXPORT_SYMBOL(atomic64_read);
73 * atomic64_add_return - add and return
74 * @delta: integer value to add
75 * @ptr: pointer to type atomic64_t
76 *
77 * Atomically adds @delta to @ptr and returns @delta + *@ptr
78 */
79noinline u64 atomic64_add_return(u64 delta, atomic64_t *ptr)
80{
81 /*
82 * Try first with a (possibly incorrect) assumption about
83 * what we have there. We'll do two loops most likely,
84 * but we'll get an ownership MESI transaction straight away
85 * instead of a read transaction followed by a
86 * flush-for-ownership transaction:
87 */
88 u64 old_val, new_val, real_val = 0;
89
90 do {
91 old_val = real_val;
92 new_val = old_val + delta;
93
94 real_val = atomic64_cmpxchg(ptr, old_val, new_val);
95
96 } while (real_val != old_val);
97
98 return new_val;
99}
100EXPORT_SYMBOL(atomic64_add_return);
101
102u64 atomic64_sub_return(u64 delta, atomic64_t *ptr)
103{
104 return atomic64_add_return(-delta, ptr);
105}
106EXPORT_SYMBOL(atomic64_sub_return);
107
108u64 atomic64_inc_return(atomic64_t *ptr)
109{
110 return atomic64_add_return(1, ptr);
111}
112EXPORT_SYMBOL(atomic64_inc_return);
113
114u64 atomic64_dec_return(atomic64_t *ptr)
115{
116 return atomic64_sub_return(1, ptr);
117}
118EXPORT_SYMBOL(atomic64_dec_return);
119
120/**
121 * atomic64_add - add integer to atomic64 variable
122 * @delta: integer value to add
123 * @ptr: pointer to type atomic64_t
124 *
125 * Atomically adds @delta to @ptr.
126 */
127void atomic64_add(u64 delta, atomic64_t *ptr)
128{
129 atomic64_add_return(delta, ptr);
130}
131EXPORT_SYMBOL(atomic64_add);
132
133/**
134 * atomic64_sub - subtract the atomic64 variable
135 * @delta: integer value to subtract
136 * @ptr: pointer to type atomic64_t
137 *
138 * Atomically subtracts @delta from @ptr.
139 */
140void atomic64_sub(u64 delta, atomic64_t *ptr)
141{
142 atomic64_add(-delta, ptr);
143}
144EXPORT_SYMBOL(atomic64_sub);
145
146/**
147 * atomic64_sub_and_test - subtract value from variable and test result
148 * @delta: integer value to subtract
149 * @ptr: pointer to type atomic64_t
150 *
151 * Atomically subtracts @delta from @ptr and returns
152 * true if the result is zero, or false for all
153 * other cases.
154 */
155int atomic64_sub_and_test(u64 delta, atomic64_t *ptr)
156{
157 u64 new_val = atomic64_sub_return(delta, ptr);
158
159 return new_val == 0;
160}
161EXPORT_SYMBOL(atomic64_sub_and_test);
162
163/**
164 * atomic64_inc - increment atomic64 variable
165 * @ptr: pointer to type atomic64_t
166 *
167 * Atomically increments @ptr by 1.
168 */
169void atomic64_inc(atomic64_t *ptr)
170{
171 atomic64_add(1, ptr);
172}
173EXPORT_SYMBOL(atomic64_inc);
174
175/**
176 * atomic64_dec - decrement atomic64 variable
177 * @ptr: pointer to type atomic64_t
178 *
179 * Atomically decrements @ptr by 1.
180 */
181void atomic64_dec(atomic64_t *ptr)
182{
183 atomic64_sub(1, ptr);
184}
185EXPORT_SYMBOL(atomic64_dec);
186
187/**
188 * atomic64_dec_and_test - decrement and test
189 * @ptr: pointer to type atomic64_t
190 *
191 * Atomically decrements @ptr by 1 and
192 * returns true if the result is 0, or false for all other
193 * cases.
194 */
195int atomic64_dec_and_test(atomic64_t *ptr)
196{
197 return atomic64_sub_and_test(1, ptr);
198}
199EXPORT_SYMBOL(atomic64_dec_and_test);
200
201/**
202 * atomic64_inc_and_test - increment and test
203 * @ptr: pointer to type atomic64_t
204 *
205 * Atomically increments @ptr by 1
206 * and returns true if the result is zero, or false for all
207 * other cases.
208 */
209int atomic64_inc_and_test(atomic64_t *ptr)
210{
211 return atomic64_sub_and_test(-1, ptr);
212}
213EXPORT_SYMBOL(atomic64_inc_and_test);
214
215/**
216 * atomic64_add_negative - add and test if negative
217 * @delta: integer value to add
218 * @ptr: pointer to type atomic64_t
219 *
220 * Atomically adds @delta to @ptr and returns true
221 * if the result is negative, or false when
222 * result is greater than or equal to zero.
223 */
224int atomic64_add_negative(u64 delta, atomic64_t *ptr)
225{
226 s64 new_val = atomic64_add_return(delta, ptr);
227
228 return new_val < 0;
229}
230EXPORT_SYMBOL(atomic64_add_negative);
diff --git a/arch/x86/lib/clear_page_64.S b/arch/x86/lib/clear_page_64.S
index 9a10a78bb4a4..ebeafcce04a9 100644
--- a/arch/x86/lib/clear_page_64.S
+++ b/arch/x86/lib/clear_page_64.S
@@ -5,15 +5,14 @@
5 * Zero a page. 5 * Zero a page.
6 * rdi page 6 * rdi page
7 */ 7 */
8 ALIGN 8ENTRY(clear_page_c)
9clear_page_c:
10 CFI_STARTPROC 9 CFI_STARTPROC
11 movl $4096/8,%ecx 10 movl $4096/8,%ecx
12 xorl %eax,%eax 11 xorl %eax,%eax
13 rep stosq 12 rep stosq
14 ret 13 ret
15 CFI_ENDPROC 14 CFI_ENDPROC
16ENDPROC(clear_page) 15ENDPROC(clear_page_c)
17 16
18ENTRY(clear_page) 17ENTRY(clear_page)
19 CFI_STARTPROC 18 CFI_STARTPROC
diff --git a/arch/x86/lib/copy_user_64.S b/arch/x86/lib/copy_user_64.S
index f118c110af32..6ba0f7bb85ea 100644
--- a/arch/x86/lib/copy_user_64.S
+++ b/arch/x86/lib/copy_user_64.S
@@ -75,6 +75,7 @@ ENTRY(copy_to_user)
75 jae bad_to_user 75 jae bad_to_user
76 ALTERNATIVE_JUMP X86_FEATURE_REP_GOOD,copy_user_generic_unrolled,copy_user_generic_string 76 ALTERNATIVE_JUMP X86_FEATURE_REP_GOOD,copy_user_generic_unrolled,copy_user_generic_string
77 CFI_ENDPROC 77 CFI_ENDPROC
78ENDPROC(copy_to_user)
78 79
79/* Standard copy_from_user with segment limit checking */ 80/* Standard copy_from_user with segment limit checking */
80ENTRY(copy_from_user) 81ENTRY(copy_from_user)
diff --git a/arch/x86/lib/delay.c b/arch/x86/lib/delay.c
index f4568605d7d5..ff485d361182 100644
--- a/arch/x86/lib/delay.c
+++ b/arch/x86/lib/delay.c
@@ -55,8 +55,10 @@ static void delay_tsc(unsigned long loops)
55 55
56 preempt_disable(); 56 preempt_disable();
57 cpu = smp_processor_id(); 57 cpu = smp_processor_id();
58 rdtsc_barrier();
58 rdtscl(bclock); 59 rdtscl(bclock);
59 for (;;) { 60 for (;;) {
61 rdtsc_barrier();
60 rdtscl(now); 62 rdtscl(now);
61 if ((now - bclock) >= loops) 63 if ((now - bclock) >= loops)
62 break; 64 break;
@@ -78,6 +80,7 @@ static void delay_tsc(unsigned long loops)
78 if (unlikely(cpu != smp_processor_id())) { 80 if (unlikely(cpu != smp_processor_id())) {
79 loops -= (now - bclock); 81 loops -= (now - bclock);
80 cpu = smp_processor_id(); 82 cpu = smp_processor_id();
83 rdtsc_barrier();
81 rdtscl(bclock); 84 rdtscl(bclock);
82 } 85 }
83 } 86 }
diff --git a/arch/x86/lib/usercopy_32.c b/arch/x86/lib/usercopy_32.c
index 7c8ca91bb9ec..1f118d462acc 100644
--- a/arch/x86/lib/usercopy_32.c
+++ b/arch/x86/lib/usercopy_32.c
@@ -751,7 +751,7 @@ survive:
751 751
752 if (retval == -ENOMEM && is_global_init(current)) { 752 if (retval == -ENOMEM && is_global_init(current)) {
753 up_read(&current->mm->mmap_sem); 753 up_read(&current->mm->mmap_sem);
754 congestion_wait(WRITE, HZ/50); 754 congestion_wait(BLK_RW_ASYNC, HZ/50);
755 goto survive; 755 goto survive;
756 } 756 }
757 757
diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c
index 78a5fff857be..85307cc6e45f 100644
--- a/arch/x86/mm/fault.c
+++ b/arch/x86/mm/fault.c
@@ -426,10 +426,11 @@ static noinline int vmalloc_fault(unsigned long address)
426} 426}
427 427
428static const char errata93_warning[] = 428static const char errata93_warning[] =
429KERN_ERR "******* Your BIOS seems to not contain a fix for K8 errata #93\n" 429KERN_ERR
430KERN_ERR "******* Working around it, but it may cause SEGVs or burn power.\n" 430"******* Your BIOS seems to not contain a fix for K8 errata #93\n"
431KERN_ERR "******* Please consider a BIOS update.\n" 431"******* Working around it, but it may cause SEGVs or burn power.\n"
432KERN_ERR "******* Disabling USB legacy in the BIOS may also help.\n"; 432"******* Please consider a BIOS update.\n"
433"******* Disabling USB legacy in the BIOS may also help.\n";
433 434
434/* 435/*
435 * No vm86 mode in 64-bit mode: 436 * No vm86 mode in 64-bit mode:
diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c
index f53b57e4086f..0607119cef94 100644
--- a/arch/x86/mm/init.c
+++ b/arch/x86/mm/init.c
@@ -12,6 +12,7 @@
12#include <asm/system.h> 12#include <asm/system.h>
13#include <asm/tlbflush.h> 13#include <asm/tlbflush.h>
14#include <asm/tlb.h> 14#include <asm/tlb.h>
15#include <asm/proto.h>
15 16
16DEFINE_PER_CPU(struct mmu_gather, mmu_gathers); 17DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
17 18
@@ -177,20 +178,6 @@ static int __meminit save_mr(struct map_range *mr, int nr_range,
177 return nr_range; 178 return nr_range;
178} 179}
179 180
180#ifdef CONFIG_X86_64
181static void __init init_gbpages(void)
182{
183 if (direct_gbpages && cpu_has_gbpages)
184 printk(KERN_INFO "Using GB pages for direct mapping\n");
185 else
186 direct_gbpages = 0;
187}
188#else
189static inline void init_gbpages(void)
190{
191}
192#endif
193
194/* 181/*
195 * Setup the direct mapping of the physical memory at PAGE_OFFSET. 182 * Setup the direct mapping of the physical memory at PAGE_OFFSET.
196 * This runs before bootmem is initialized and gets pages directly from 183 * This runs before bootmem is initialized and gets pages directly from
@@ -210,9 +197,6 @@ unsigned long __init_refok init_memory_mapping(unsigned long start,
210 197
211 printk(KERN_INFO "init_memory_mapping: %016lx-%016lx\n", start, end); 198 printk(KERN_INFO "init_memory_mapping: %016lx-%016lx\n", start, end);
212 199
213 if (!after_bootmem)
214 init_gbpages();
215
216#if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_KMEMCHECK) 200#if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_KMEMCHECK)
217 /* 201 /*
218 * For CONFIG_DEBUG_PAGEALLOC, identity mapping will use small pages. 202 * For CONFIG_DEBUG_PAGEALLOC, identity mapping will use small pages.
diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c
index c4378f4fd4a5..6176fe8f29e0 100644
--- a/arch/x86/mm/init_64.c
+++ b/arch/x86/mm/init_64.c
@@ -598,6 +598,15 @@ void __init paging_init(void)
598 598
599 sparse_memory_present_with_active_regions(MAX_NUMNODES); 599 sparse_memory_present_with_active_regions(MAX_NUMNODES);
600 sparse_init(); 600 sparse_init();
601
602 /*
603 * clear the default setting with node 0
604 * note: don't use nodes_clear here, that is really clearing when
605 * numa support is not compiled in, and later node_set_state
606 * will not set it back.
607 */
608 node_clear_state(0, N_NORMAL_MEMORY);
609
601 free_area_init_nodes(max_zone_pfns); 610 free_area_init_nodes(max_zone_pfns);
602} 611}
603 612
diff --git a/arch/x86/mm/pageattr.c b/arch/x86/mm/pageattr.c
index 3cfe9ced8a4c..1b734d7a8966 100644
--- a/arch/x86/mm/pageattr.c
+++ b/arch/x86/mm/pageattr.c
@@ -11,6 +11,7 @@
11#include <linux/interrupt.h> 11#include <linux/interrupt.h>
12#include <linux/seq_file.h> 12#include <linux/seq_file.h>
13#include <linux/debugfs.h> 13#include <linux/debugfs.h>
14#include <linux/pfn.h>
14 15
15#include <asm/e820.h> 16#include <asm/e820.h>
16#include <asm/processor.h> 17#include <asm/processor.h>
@@ -681,8 +682,9 @@ static int __change_page_attr_set_clr(struct cpa_data *cpa, int checkalias);
681static int cpa_process_alias(struct cpa_data *cpa) 682static int cpa_process_alias(struct cpa_data *cpa)
682{ 683{
683 struct cpa_data alias_cpa; 684 struct cpa_data alias_cpa;
684 int ret = 0; 685 unsigned long laddr = (unsigned long)__va(cpa->pfn << PAGE_SHIFT);
685 unsigned long temp_cpa_vaddr, vaddr; 686 unsigned long vaddr, remapped;
687 int ret;
686 688
687 if (cpa->pfn >= max_pfn_mapped) 689 if (cpa->pfn >= max_pfn_mapped)
688 return 0; 690 return 0;
@@ -706,42 +708,55 @@ static int cpa_process_alias(struct cpa_data *cpa)
706 PAGE_OFFSET + (max_pfn_mapped << PAGE_SHIFT)))) { 708 PAGE_OFFSET + (max_pfn_mapped << PAGE_SHIFT)))) {
707 709
708 alias_cpa = *cpa; 710 alias_cpa = *cpa;
709 temp_cpa_vaddr = (unsigned long) __va(cpa->pfn << PAGE_SHIFT); 711 alias_cpa.vaddr = &laddr;
710 alias_cpa.vaddr = &temp_cpa_vaddr;
711 alias_cpa.flags &= ~(CPA_PAGES_ARRAY | CPA_ARRAY); 712 alias_cpa.flags &= ~(CPA_PAGES_ARRAY | CPA_ARRAY);
712 713
713
714 ret = __change_page_attr_set_clr(&alias_cpa, 0); 714 ret = __change_page_attr_set_clr(&alias_cpa, 0);
715 if (ret)
716 return ret;
715 } 717 }
716 718
717#ifdef CONFIG_X86_64 719#ifdef CONFIG_X86_64
718 if (ret)
719 return ret;
720 /* 720 /*
721 * No need to redo, when the primary call touched the high 721 * If the primary call didn't touch the high mapping already
722 * mapping already: 722 * and the physical address is inside the kernel map, we need
723 */
724 if (within(vaddr, (unsigned long) _text, _brk_end))
725 return 0;
726
727 /*
728 * If the physical address is inside the kernel map, we need
729 * to touch the high mapped kernel as well: 723 * to touch the high mapped kernel as well:
730 */ 724 */
731 if (!within(cpa->pfn, highmap_start_pfn(), highmap_end_pfn())) 725 if (!within(vaddr, (unsigned long)_text, _brk_end) &&
732 return 0; 726 within(cpa->pfn, highmap_start_pfn(), highmap_end_pfn())) {
727 unsigned long temp_cpa_vaddr = (cpa->pfn << PAGE_SHIFT) +
728 __START_KERNEL_map - phys_base;
729 alias_cpa = *cpa;
730 alias_cpa.vaddr = &temp_cpa_vaddr;
731 alias_cpa.flags &= ~(CPA_PAGES_ARRAY | CPA_ARRAY);
733 732
734 alias_cpa = *cpa; 733 /*
735 temp_cpa_vaddr = (cpa->pfn << PAGE_SHIFT) + __START_KERNEL_map - phys_base; 734 * The high mapping range is imprecise, so ignore the
736 alias_cpa.vaddr = &temp_cpa_vaddr; 735 * return value.
737 alias_cpa.flags &= ~(CPA_PAGES_ARRAY | CPA_ARRAY); 736 */
737 __change_page_attr_set_clr(&alias_cpa, 0);
738 }
739#endif
738 740
739 /* 741 /*
740 * The high mapping range is imprecise, so ignore the return value. 742 * If the PMD page was partially used for per-cpu remapping,
743 * the recycled area needs to be split and modified. Because
744 * the area is always proper subset of a PMD page
745 * cpa->numpages is guaranteed to be 1 for these areas, so
746 * there's no need to loop over and check for further remaps.
741 */ 747 */
742 __change_page_attr_set_clr(&alias_cpa, 0); 748 remapped = (unsigned long)pcpu_lpage_remapped((void *)laddr);
743#endif 749 if (remapped) {
744 return ret; 750 WARN_ON(cpa->numpages > 1);
751 alias_cpa = *cpa;
752 alias_cpa.vaddr = &remapped;
753 alias_cpa.flags &= ~(CPA_PAGES_ARRAY | CPA_ARRAY);
754 ret = __change_page_attr_set_clr(&alias_cpa, 0);
755 if (ret)
756 return ret;
757 }
758
759 return 0;
745} 760}
746 761
747static int __change_page_attr_set_clr(struct cpa_data *cpa, int checkalias) 762static int __change_page_attr_set_clr(struct cpa_data *cpa, int checkalias)
diff --git a/arch/x86/oprofile/nmi_int.c b/arch/x86/oprofile/nmi_int.c
index b07dd8d0b321..89b9a5cd63da 100644
--- a/arch/x86/oprofile/nmi_int.c
+++ b/arch/x86/oprofile/nmi_int.c
@@ -390,7 +390,7 @@ static int __init p4_init(char **cpu_type)
390static int force_arch_perfmon; 390static int force_arch_perfmon;
391static int force_cpu_type(const char *str, struct kernel_param *kp) 391static int force_cpu_type(const char *str, struct kernel_param *kp)
392{ 392{
393 if (!strcmp(str, "archperfmon")) { 393 if (!strcmp(str, "arch_perfmon")) {
394 force_arch_perfmon = 1; 394 force_arch_perfmon = 1;
395 printk(KERN_INFO "oprofile: forcing architectural perfmon\n"); 395 printk(KERN_INFO "oprofile: forcing architectural perfmon\n");
396 } 396 }
diff --git a/arch/x86/pci/acpi.c b/arch/x86/pci/acpi.c
index b26626dc517c..1014eb4bfc37 100644
--- a/arch/x86/pci/acpi.c
+++ b/arch/x86/pci/acpi.c
@@ -68,6 +68,10 @@ setup_resource(struct acpi_resource *acpi_res, void *data)
68 unsigned long flags; 68 unsigned long flags;
69 struct resource *root; 69 struct resource *root;
70 int max_root_bus_resources = PCI_BUS_NUM_RESOURCES; 70 int max_root_bus_resources = PCI_BUS_NUM_RESOURCES;
71 u64 start, end;
72
73 if (bus_has_transparent_bridge(info->bus))
74 max_root_bus_resources -= 3;
71 75
72 status = resource_to_addr(acpi_res, &addr); 76 status = resource_to_addr(acpi_res, &addr);
73 if (!ACPI_SUCCESS(status)) 77 if (!ACPI_SUCCESS(status))
@@ -84,25 +88,24 @@ setup_resource(struct acpi_resource *acpi_res, void *data)
84 } else 88 } else
85 return AE_OK; 89 return AE_OK;
86 90
87 res = &info->res[info->res_num]; 91 start = addr.minimum + addr.translation_offset;
88 res->name = info->name; 92 end = start + addr.address_length - 1;
89 res->flags = flags;
90 res->start = addr.minimum + addr.translation_offset;
91 res->end = res->start + addr.address_length - 1;
92 res->child = NULL;
93
94 if (bus_has_transparent_bridge(info->bus))
95 max_root_bus_resources -= 3;
96 if (info->res_num >= max_root_bus_resources) { 93 if (info->res_num >= max_root_bus_resources) {
97 printk(KERN_WARNING "PCI: Failed to allocate 0x%lx-0x%lx " 94 printk(KERN_WARNING "PCI: Failed to allocate 0x%lx-0x%lx "
98 "from %s for %s due to _CRS returning more than " 95 "from %s for %s due to _CRS returning more than "
99 "%d resource descriptors\n", (unsigned long) res->start, 96 "%d resource descriptors\n", (unsigned long) start,
100 (unsigned long) res->end, root->name, info->name, 97 (unsigned long) end, root->name, info->name,
101 max_root_bus_resources); 98 max_root_bus_resources);
102 info->res_num++;
103 return AE_OK; 99 return AE_OK;
104 } 100 }
105 101
102 res = &info->res[info->res_num];
103 res->name = info->name;
104 res->flags = flags;
105 res->start = start;
106 res->end = end;
107 res->child = NULL;
108
106 if (insert_resource(root, res)) { 109 if (insert_resource(root, res)) {
107 printk(KERN_ERR "PCI: Failed to allocate 0x%lx-0x%lx " 110 printk(KERN_ERR "PCI: Failed to allocate 0x%lx-0x%lx "
108 "from %s for %s\n", (unsigned long) res->start, 111 "from %s for %s\n", (unsigned long) res->start,
@@ -115,23 +118,6 @@ setup_resource(struct acpi_resource *acpi_res, void *data)
115} 118}
116 119
117static void 120static void
118adjust_transparent_bridge_resources(struct pci_bus *bus)
119{
120 struct pci_dev *dev;
121
122 list_for_each_entry(dev, &bus->devices, bus_list) {
123 int i;
124 u16 class = dev->class >> 8;
125
126 if (class == PCI_CLASS_BRIDGE_PCI && dev->transparent) {
127 for(i = 3; i < PCI_BUS_NUM_RESOURCES; i++)
128 dev->subordinate->resource[i] =
129 dev->bus->resource[i - 3];
130 }
131 }
132}
133
134static void
135get_current_resources(struct acpi_device *device, int busnum, 121get_current_resources(struct acpi_device *device, int busnum,
136 int domain, struct pci_bus *bus) 122 int domain, struct pci_bus *bus)
137{ 123{
@@ -158,8 +144,6 @@ get_current_resources(struct acpi_device *device, int busnum,
158 info.res_num = 0; 144 info.res_num = 0;
159 acpi_walk_resources(device->handle, METHOD_NAME__CRS, setup_resource, 145 acpi_walk_resources(device->handle, METHOD_NAME__CRS, setup_resource,
160 &info); 146 &info);
161 if (info.res_num)
162 adjust_transparent_bridge_resources(bus);
163 147
164 return; 148 return;
165 149
@@ -222,8 +206,15 @@ struct pci_bus * __devinit pci_acpi_scan_root(struct acpi_device *device, int do
222 */ 206 */
223 memcpy(bus->sysdata, sd, sizeof(*sd)); 207 memcpy(bus->sysdata, sd, sizeof(*sd));
224 kfree(sd); 208 kfree(sd);
225 } else 209 } else {
226 bus = pci_scan_bus_parented(NULL, busnum, &pci_root_ops, sd); 210 bus = pci_create_bus(NULL, busnum, &pci_root_ops, sd);
211 if (bus) {
212 if (pci_probe & PCI_USE__CRS)
213 get_current_resources(device, busnum, domain,
214 bus);
215 bus->subordinate = pci_scan_child_bus(bus);
216 }
217 }
227 218
228 if (!bus) 219 if (!bus)
229 kfree(sd); 220 kfree(sd);
@@ -238,8 +229,6 @@ struct pci_bus * __devinit pci_acpi_scan_root(struct acpi_device *device, int do
238#endif 229#endif
239 } 230 }
240 231
241 if (bus && (pci_probe & PCI_USE__CRS))
242 get_current_resources(device, busnum, domain, bus);
243 return bus; 232 return bus;
244} 233}
245 234
diff --git a/arch/x86/pci/amd_bus.c b/arch/x86/pci/amd_bus.c
index f893d6a6e803..3ffa10df20b9 100644
--- a/arch/x86/pci/amd_bus.c
+++ b/arch/x86/pci/amd_bus.c
@@ -100,8 +100,9 @@ void x86_pci_root_bus_res_quirks(struct pci_bus *b)
100 int j; 100 int j;
101 struct pci_root_info *info; 101 struct pci_root_info *info;
102 102
103 /* don't go for it if _CRS is used */ 103 /* don't go for it if _CRS is used already */
104 if (pci_probe & PCI_USE__CRS) 104 if (b->resource[0] != &ioport_resource ||
105 b->resource[1] != &iomem_resource)
105 return; 106 return;
106 107
107 /* if only one root bus, don't need to anything */ 108 /* if only one root bus, don't need to anything */
@@ -116,6 +117,9 @@ void x86_pci_root_bus_res_quirks(struct pci_bus *b)
116 if (i == pci_root_num) 117 if (i == pci_root_num)
117 return; 118 return;
118 119
120 printk(KERN_DEBUG "PCI: peer root bus %02x res updated from pci conf\n",
121 b->number);
122
119 info = &pci_root_info[i]; 123 info = &pci_root_info[i];
120 for (j = 0; j < info->res_num; j++) { 124 for (j = 0; j < info->res_num; j++) {
121 struct resource *res; 125 struct resource *res;
diff --git a/arch/x86/power/Makefile b/arch/x86/power/Makefile
index de2abbd07544..a6a198c33623 100644
--- a/arch/x86/power/Makefile
+++ b/arch/x86/power/Makefile
@@ -1,7 +1,7 @@
1# __restore_processor_state() restores %gs after S3 resume and so should not 1# __restore_processor_state() restores %gs after S3 resume and so should not
2# itself be stack-protected 2# itself be stack-protected
3nostackp := $(call cc-option, -fno-stack-protector) 3nostackp := $(call cc-option, -fno-stack-protector)
4CFLAGS_cpu_$(BITS).o := $(nostackp) 4CFLAGS_cpu.o := $(nostackp)
5 5
6obj-$(CONFIG_PM_SLEEP) += cpu.o 6obj-$(CONFIG_PM_SLEEP) += cpu.o
7obj-$(CONFIG_HIBERNATION) += hibernate_$(BITS).o hibernate_asm_$(BITS).o 7obj-$(CONFIG_HIBERNATION) += hibernate_$(BITS).o hibernate_asm_$(BITS).o
diff --git a/arch/x86/power/cpu.c b/arch/x86/power/cpu.c
index d277ef1eea51..b3d20b9cac63 100644
--- a/arch/x86/power/cpu.c
+++ b/arch/x86/power/cpu.c
@@ -244,7 +244,7 @@ static void __restore_processor_state(struct saved_context *ctxt)
244 do_fpu_end(); 244 do_fpu_end();
245 mtrr_ap_init(); 245 mtrr_ap_init();
246 246
247#ifdef CONFIG_X86_32 247#ifdef CONFIG_X86_OLD_MCE
248 mcheck_init(&boot_cpu_data); 248 mcheck_init(&boot_cpu_data);
249#endif 249#endif
250} 250}
generated by cgit v1.2.2 (git 2.25.0) at 2024-10-25 19:36:50 -0400