fault.c 28.2 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1
2
/*
 *  Copyright (C) 1995  Linus Torvalds
Ingo Molnar's avatar
Ingo Molnar committed
3
 *  Copyright (C) 2001, 2002 Andi Kleen, SuSE Labs.
4
 *  Copyright (C) 2008-2009, Red Hat Inc., Ingo Molnar
Linus Torvalds's avatar
Linus Torvalds committed
5
 */
6
7
8
9
10
11
12
#include <linux/magic.h>		/* STACK_END_MAGIC		*/
#include <linux/sched.h>		/* test_thread_flag(), ...	*/
#include <linux/kdebug.h>		/* oops_begin/end, ...		*/
#include <linux/module.h>		/* search_exception_table	*/
#include <linux/bootmem.h>		/* max_low_pfn			*/
#include <linux/kprobes.h>		/* __kprobes, ...		*/
#include <linux/mmiotrace.h>		/* kmmio_handler, ...		*/
13
#include <linux/perf_event.h>		/* perf_sw_event		*/
14
#include <linux/hugetlb.h>		/* hstate_index_to_shift	*/
15
#include <linux/prefetch.h>		/* prefetchw			*/
Ingo Molnar's avatar
Ingo Molnar committed
16

17
18
#include <asm/traps.h>			/* dotraplinkage, ...		*/
#include <asm/pgalloc.h>		/* pgd_*(), ...			*/
Vegard Nossum's avatar
Vegard Nossum committed
19
#include <asm/kmemcheck.h>		/* kmemcheck_*(), ...		*/
Randy Dunlap's avatar
Randy Dunlap committed
20
#include <asm/vsyscall.h>
Linus Torvalds's avatar
Linus Torvalds committed
21

22
/*
Ingo Molnar's avatar
Ingo Molnar committed
23
24
25
26
27
28
29
 * Page fault error code bits:
 *
 *   bit 0 ==	 0: no page found	1: protection fault
 *   bit 1 ==	 0: read access		1: write access
 *   bit 2 ==	 0: kernel-mode access	1: user-mode access
 *   bit 3 ==				1: use of reserved bit detected
 *   bit 4 ==				1: fault was an instruction fetch
30
 */
Ingo Molnar's avatar
Ingo Molnar committed
31
32
33
34
35
36
37
38
enum x86_pf_error_code {

	PF_PROT		=		1 << 0,
	PF_WRITE	=		1 << 1,
	PF_USER		=		1 << 2,
	PF_RSVD		=		1 << 3,
	PF_INSTR	=		1 << 4,
};
39

40
/*
41
42
 * Returns 0 if mmiotrace is disabled, or if the fault is not
 * handled by mmiotrace:
43
 */
44
45
static inline int __kprobes
kmmio_fault(struct pt_regs *regs, unsigned long addr)
46
{
47
48
49
50
	if (unlikely(is_kmmio_active()))
		if (kmmio_handler(regs, addr) == 1)
			return -1;
	return 0;
51
52
}

53
static inline int __kprobes notify_page_fault(struct pt_regs *regs)
54
{
55
56
57
	int ret = 0;

	/* kprobe_running() needs smp_processor_id() */
58
	if (kprobes_built_in() && !user_mode_vm(regs)) {
59
60
61
62
63
		preempt_disable();
		if (kprobe_running() && kprobe_fault_handler(regs, 14))
			ret = 1;
		preempt_enable();
	}
64

65
	return ret;
66
}
67

68
/*
Ingo Molnar's avatar
Ingo Molnar committed
69
70
71
72
73
74
 * Prefetch quirks:
 *
 * 32-bit mode:
 *
 *   Sometimes AMD Athlon/Opteron CPUs report invalid exceptions on prefetch.
 *   Check that here and ignore it.
75
 *
Ingo Molnar's avatar
Ingo Molnar committed
76
 * 64-bit mode:
77
 *
Ingo Molnar's avatar
Ingo Molnar committed
78
79
80
81
 *   Sometimes the CPU reports invalid exceptions on prefetch.
 *   Check that here and ignore it.
 *
 * Opcode checker based on code by Richard Brunner.
82
 */
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
static inline int
check_prefetch_opcode(struct pt_regs *regs, unsigned char *instr,
		      unsigned char opcode, int *prefetch)
{
	unsigned char instr_hi = opcode & 0xf0;
	unsigned char instr_lo = opcode & 0x0f;

	switch (instr_hi) {
	case 0x20:
	case 0x30:
		/*
		 * Values 0x26,0x2E,0x36,0x3E are valid x86 prefixes.
		 * In X86_64 long mode, the CPU will signal invalid
		 * opcode if some of these prefixes are present so
		 * X86_64 will never get here anyway
		 */
		return ((instr_lo & 7) == 0x6);
#ifdef CONFIG_X86_64
	case 0x40:
		/*
		 * In AMD64 long mode 0x40..0x4F are valid REX prefixes
		 * Need to figure out under what instruction mode the
		 * instruction was issued. Could check the LDT for lm,
		 * but for now it's good enough to assume that long
		 * mode only uses well known segments or kernel.
		 */
109
		return (!user_mode(regs) || user_64bit_mode(regs));
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
#endif
	case 0x60:
		/* 0x64 thru 0x67 are valid prefixes in all modes. */
		return (instr_lo & 0xC) == 0x4;
	case 0xF0:
		/* 0xF0, 0xF2, 0xF3 are valid prefixes in all modes. */
		return !instr_lo || (instr_lo>>1) == 1;
	case 0x00:
		/* Prefetch instruction is 0x0F0D or 0x0F18 */
		if (probe_kernel_address(instr, opcode))
			return 0;

		*prefetch = (instr_lo == 0xF) &&
			(opcode == 0x0D || opcode == 0x18);
		return 0;
	default:
		return 0;
	}
}

Ingo Molnar's avatar
Ingo Molnar committed
130
131
static int
is_prefetch(struct pt_regs *regs, unsigned long error_code, unsigned long addr)
132
{
Ingo Molnar's avatar
Ingo Molnar committed
133
	unsigned char *max_instr;
134
	unsigned char *instr;
135
	int prefetch = 0;
Linus Torvalds's avatar
Linus Torvalds committed
136

Ingo Molnar's avatar
Ingo Molnar committed
137
138
139
140
	/*
	 * If it was a exec (instruction fetch) fault on NX page, then
	 * do not ignore the fault:
	 */
141
	if (error_code & PF_INSTR)
Linus Torvalds's avatar
Linus Torvalds committed
142
		return 0;
143

144
	instr = (void *)convert_ip_to_linear(current, regs);
145
	max_instr = instr + 15;
Linus Torvalds's avatar
Linus Torvalds committed
146

147
	if (user_mode(regs) && instr >= (unsigned char *)TASK_SIZE)
Linus Torvalds's avatar
Linus Torvalds committed
148
149
		return 0;

150
	while (instr < max_instr) {
Ingo Molnar's avatar
Ingo Molnar committed
151
		unsigned char opcode;
Linus Torvalds's avatar
Linus Torvalds committed
152

153
		if (probe_kernel_address(instr, opcode))
154
			break;
Linus Torvalds's avatar
Linus Torvalds committed
155
156
157

		instr++;

158
		if (!check_prefetch_opcode(regs, instr, opcode, &prefetch))
Linus Torvalds's avatar
Linus Torvalds committed
159
160
161
162
163
			break;
	}
	return prefetch;
}

Ingo Molnar's avatar
Ingo Molnar committed
164
165
static void
force_sig_info_fault(int si_signo, int si_code, unsigned long address,
166
		     struct task_struct *tsk, int fault)
167
{
168
	unsigned lsb = 0;
169
170
	siginfo_t info;

Ingo Molnar's avatar
Ingo Molnar committed
171
172
173
174
	info.si_signo	= si_signo;
	info.si_errno	= 0;
	info.si_code	= si_code;
	info.si_addr	= (void __user *)address;
175
176
177
178
179
	if (fault & VM_FAULT_HWPOISON_LARGE)
		lsb = hstate_index_to_shift(VM_FAULT_GET_HINDEX(fault)); 
	if (fault & VM_FAULT_HWPOISON)
		lsb = PAGE_SHIFT;
	info.si_addr_lsb = lsb;
Ingo Molnar's avatar
Ingo Molnar committed
180

181
182
183
	force_sig_info(si_signo, &info, tsk);
}

184
185
186
187
188
DEFINE_SPINLOCK(pgd_lock);
LIST_HEAD(pgd_list);

#ifdef CONFIG_X86_32
static inline pmd_t *vmalloc_sync_one(pgd_t *pgd, unsigned long address)
189
{
190
191
192
193
	unsigned index = pgd_index(address);
	pgd_t *pgd_k;
	pud_t *pud, *pud_k;
	pmd_t *pmd, *pmd_k;
Ingo Molnar's avatar
Ingo Molnar committed
194

195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
	pgd += index;
	pgd_k = init_mm.pgd + index;

	if (!pgd_present(*pgd_k))
		return NULL;

	/*
	 * set_pgd(pgd, *pgd_k); here would be useless on PAE
	 * and redundant with the set_pmd() on non-PAE. As would
	 * set_pud.
	 */
	pud = pud_offset(pgd, address);
	pud_k = pud_offset(pgd_k, address);
	if (!pud_present(*pud_k))
		return NULL;

	pmd = pmd_offset(pud, address);
	pmd_k = pmd_offset(pud_k, address);
	if (!pmd_present(*pmd_k))
		return NULL;

216
	if (!pmd_present(*pmd))
217
		set_pmd(pmd, *pmd_k);
218
	else
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
		BUG_ON(pmd_page(*pmd) != pmd_page(*pmd_k));

	return pmd_k;
}

void vmalloc_sync_all(void)
{
	unsigned long address;

	if (SHARED_KERNEL_PMD)
		return;

	for (address = VMALLOC_START & PMD_MASK;
	     address >= TASK_SIZE && address < FIXADDR_TOP;
	     address += PMD_SIZE) {
		struct page *page;

236
		spin_lock(&pgd_lock);
237
		list_for_each_entry(page, &pgd_list, lru) {
238
			spinlock_t *pgt_lock;
239
			pmd_t *ret;
240

241
			/* the pgt_lock only for Xen */
242
243
244
245
246
247
248
			pgt_lock = &pgd_page_get_mm(page)->page_table_lock;

			spin_lock(pgt_lock);
			ret = vmalloc_sync_one(page_address(page), address);
			spin_unlock(pgt_lock);

			if (!ret)
249
250
				break;
		}
251
		spin_unlock(&pgd_lock);
252
253
254
255
256
257
258
259
	}
}

/*
 * 32-bit:
 *
 *   Handle a fault on the vmalloc or module mapping area
 */
260
static noinline __kprobes int vmalloc_fault(unsigned long address)
261
262
263
264
265
266
267
268
269
{
	unsigned long pgd_paddr;
	pmd_t *pmd_k;
	pte_t *pte_k;

	/* Make sure we are in vmalloc area: */
	if (!(address >= VMALLOC_START && address < VMALLOC_END))
		return -1;

270
271
	WARN_ON_ONCE(in_nmi());

272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
	/*
	 * Synchronize this task's top level page-table
	 * with the 'reference' page table.
	 *
	 * Do _not_ use "current" here. We might be inside
	 * an interrupt in the middle of a task switch..
	 */
	pgd_paddr = read_cr3();
	pmd_k = vmalloc_sync_one(__va(pgd_paddr), address);
	if (!pmd_k)
		return -1;

	pte_k = pte_offset_kernel(pmd_k, address);
	if (!pte_present(*pte_k))
		return -1;

	return 0;
}

/*
 * Did it hit the DOS screen memory VA from vm86 mode?
 */
static inline void
check_v8086_mode(struct pt_regs *regs, unsigned long address,
		 struct task_struct *tsk)
{
	unsigned long bit;

	if (!v8086_mode(regs))
		return;

	bit = (address - 0xA0000) >> PAGE_SHIFT;
	if (bit < 32)
		tsk->thread.screen_bitmap |= 1 << bit;
306
}
Linus Torvalds's avatar
Linus Torvalds committed
307

Akinobu Mita's avatar
Akinobu Mita committed
308
static bool low_pfn(unsigned long pfn)
Linus Torvalds's avatar
Linus Torvalds committed
309
{
Akinobu Mita's avatar
Akinobu Mita committed
310
311
	return pfn < max_low_pfn;
}
312

Akinobu Mita's avatar
Akinobu Mita committed
313
314
315
316
317
318
static void dump_pagetable(unsigned long address)
{
	pgd_t *base = __va(read_cr3());
	pgd_t *pgd = &base[pgd_index(address)];
	pmd_t *pmd;
	pte_t *pte;
Ingo Molnar's avatar
Ingo Molnar committed
319

320
#ifdef CONFIG_X86_PAE
Akinobu Mita's avatar
Akinobu Mita committed
321
322
323
	printk("*pdpt = %016Lx ", pgd_val(*pgd));
	if (!low_pfn(pgd_val(*pgd) >> PAGE_SHIFT) || !pgd_present(*pgd))
		goto out;
324
#endif
Akinobu Mita's avatar
Akinobu Mita committed
325
326
	pmd = pmd_offset(pud_offset(pgd, address), address);
	printk(KERN_CONT "*pde = %0*Lx ", sizeof(*pmd) * 2, (u64)pmd_val(*pmd));
327
328
329
330
331

	/*
	 * We must not directly access the pte in the highpte
	 * case if the page table is located in highmem.
	 * And let's rather not kmap-atomic the pte, just in case
Ingo Molnar's avatar
Ingo Molnar committed
332
	 * it's allocated already:
333
	 */
Akinobu Mita's avatar
Akinobu Mita committed
334
335
	if (!low_pfn(pmd_pfn(*pmd)) || !pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
336

Akinobu Mita's avatar
Akinobu Mita committed
337
338
339
	pte = pte_offset_kernel(pmd, address);
	printk("*pte = %0*Lx ", sizeof(*pte) * 2, (u64)pte_val(*pte));
out:
340
	printk("\n");
341
342
343
344
345
346
}

#else /* CONFIG_X86_64: */

void vmalloc_sync_all(void)
{
347
	sync_global_pgds(VMALLOC_START & PGDIR_MASK, VMALLOC_END);
348
349
350
351
352
353
354
355
356
}

/*
 * 64-bit:
 *
 *   Handle a fault on the vmalloc area
 *
 * This assumes no large pages in there.
 */
357
static noinline __kprobes int vmalloc_fault(unsigned long address)
358
359
360
361
362
363
364
365
366
367
{
	pgd_t *pgd, *pgd_ref;
	pud_t *pud, *pud_ref;
	pmd_t *pmd, *pmd_ref;
	pte_t *pte, *pte_ref;

	/* Make sure we are in vmalloc area: */
	if (!(address >= VMALLOC_START && address < VMALLOC_END))
		return -1;

368
369
	WARN_ON_ONCE(in_nmi());

370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
	/*
	 * Copy kernel mappings over when needed. This can also
	 * happen within a race in page table update. In the later
	 * case just flush:
	 */
	pgd = pgd_offset(current->active_mm, address);
	pgd_ref = pgd_offset_k(address);
	if (pgd_none(*pgd_ref))
		return -1;

	if (pgd_none(*pgd))
		set_pgd(pgd, *pgd_ref);
	else
		BUG_ON(pgd_page_vaddr(*pgd) != pgd_page_vaddr(*pgd_ref));

	/*
	 * Below here mismatches are bugs because these lower tables
	 * are shared:
	 */

	pud = pud_offset(pgd, address);
	pud_ref = pud_offset(pgd_ref, address);
	if (pud_none(*pud_ref))
		return -1;

	if (pud_none(*pud) || pud_page_vaddr(*pud) != pud_page_vaddr(*pud_ref))
		BUG();

	pmd = pmd_offset(pud, address);
	pmd_ref = pmd_offset(pud_ref, address);
	if (pmd_none(*pmd_ref))
		return -1;

	if (pmd_none(*pmd) || pmd_page(*pmd) != pmd_page(*pmd_ref))
		BUG();

	pte_ref = pte_offset_kernel(pmd_ref, address);
	if (!pte_present(*pte_ref))
		return -1;

	pte = pte_offset_kernel(pmd, address);

	/*
	 * Don't use pte_page here, because the mappings can point
	 * outside mem_map, and the NUMA hash lookup cannot handle
	 * that:
	 */
	if (!pte_present(*pte) || pte_pfn(*pte) != pte_pfn(*pte_ref))
		BUG();

	return 0;
}

static const char errata93_warning[] =
424
425
426
427
428
KERN_ERR 
"******* Your BIOS seems to not contain a fix for K8 errata #93\n"
"******* Working around it, but it may cause SEGVs or burn power.\n"
"******* Please consider a BIOS update.\n"
"******* Disabling USB legacy in the BIOS may also help.\n";
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447

/*
 * No vm86 mode in 64-bit mode:
 */
static inline void
check_v8086_mode(struct pt_regs *regs, unsigned long address,
		 struct task_struct *tsk)
{
}

static int bad_address(void *p)
{
	unsigned long dummy;

	return probe_kernel_address((unsigned long *)p, dummy);
}

static void dump_pagetable(unsigned long address)
{
Akinobu Mita's avatar
Akinobu Mita committed
448
449
	pgd_t *base = __va(read_cr3() & PHYSICAL_PAGE_MASK);
	pgd_t *pgd = base + pgd_index(address);
Linus Torvalds's avatar
Linus Torvalds committed
450
451
452
453
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;

Ingo Molnar's avatar
Ingo Molnar committed
454
455
456
	if (bad_address(pgd))
		goto bad;

457
	printk("PGD %lx ", pgd_val(*pgd));
Ingo Molnar's avatar
Ingo Molnar committed
458
459
460

	if (!pgd_present(*pgd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
461

462
	pud = pud_offset(pgd, address);
Ingo Molnar's avatar
Ingo Molnar committed
463
464
465
	if (bad_address(pud))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
466
	printk("PUD %lx ", pud_val(*pud));
467
	if (!pud_present(*pud) || pud_large(*pud))
Ingo Molnar's avatar
Ingo Molnar committed
468
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
469
470

	pmd = pmd_offset(pud, address);
Ingo Molnar's avatar
Ingo Molnar committed
471
472
473
	if (bad_address(pmd))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
474
	printk("PMD %lx ", pmd_val(*pmd));
Ingo Molnar's avatar
Ingo Molnar committed
475
476
	if (!pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
477
478

	pte = pte_offset_kernel(pmd, address);
Ingo Molnar's avatar
Ingo Molnar committed
479
480
481
	if (bad_address(pte))
		goto bad;

482
	printk("PTE %lx", pte_val(*pte));
Ingo Molnar's avatar
Ingo Molnar committed
483
out:
Linus Torvalds's avatar
Linus Torvalds committed
484
485
486
487
	printk("\n");
	return;
bad:
	printk("BAD\n");
488
489
}

490
#endif /* CONFIG_X86_64 */
Linus Torvalds's avatar
Linus Torvalds committed
491

Ingo Molnar's avatar
Ingo Molnar committed
492
493
494
495
496
497
498
499
500
501
502
503
504
/*
 * Workaround for K8 erratum #93 & buggy BIOS.
 *
 * BIOS SMM functions are required to use a specific workaround
 * to avoid corruption of the 64bit RIP register on C stepping K8.
 *
 * A lot of BIOS that didn't get tested properly miss this.
 *
 * The OS sees this as a page fault with the upper 32bits of RIP cleared.
 * Try to work around it here.
 *
 * Note we only handle faults in kernel here.
 * Does nothing on 32-bit.
505
 */
506
static int is_errata93(struct pt_regs *regs, unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
507
{
508
#ifdef CONFIG_X86_64
509
	if (address != regs->ip)
Linus Torvalds's avatar
Linus Torvalds committed
510
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
511

512
	if ((address >> 32) != 0)
Linus Torvalds's avatar
Linus Torvalds committed
513
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
514

Linus Torvalds's avatar
Linus Torvalds committed
515
	address |= 0xffffffffUL << 32;
516
517
	if ((address >= (u64)_stext && address <= (u64)_etext) ||
	    (address >= MODULES_VADDR && address <= MODULES_END)) {
518
		printk_once(errata93_warning);
519
		regs->ip = address;
Linus Torvalds's avatar
Linus Torvalds committed
520
521
		return 1;
	}
522
#endif
Linus Torvalds's avatar
Linus Torvalds committed
523
	return 0;
524
}
Linus Torvalds's avatar
Linus Torvalds committed
525

526
/*
Ingo Molnar's avatar
Ingo Molnar committed
527
528
529
530
531
 * Work around K8 erratum #100 K8 in compat mode occasionally jumps
 * to illegal addresses >4GB.
 *
 * We catch this in the page fault handler because these addresses
 * are not reachable. Just detect this case and return.  Any code
532
533
534
535
536
 * segment in LDT is compatibility mode.
 */
static int is_errata100(struct pt_regs *regs, unsigned long address)
{
#ifdef CONFIG_X86_64
Ingo Molnar's avatar
Ingo Molnar committed
537
	if ((regs->cs == __USER32_CS || (regs->cs & (1<<2))) && (address >> 32))
538
539
540
541
542
		return 1;
#endif
	return 0;
}

543
544
545
546
static int is_f00f_bug(struct pt_regs *regs, unsigned long address)
{
#ifdef CONFIG_X86_F00F_BUG
	unsigned long nr;
Ingo Molnar's avatar
Ingo Molnar committed
547

548
	/*
Ingo Molnar's avatar
Ingo Molnar committed
549
	 * Pentium F0 0F C7 C8 bug workaround:
550
551
552
553
554
555
556
557
558
559
560
561
562
	 */
	if (boot_cpu_data.f00f_bug) {
		nr = (address - idt_descr.address) >> 3;

		if (nr == 6) {
			do_invalid_op(regs, 0);
			return 1;
		}
	}
#endif
	return 0;
}

563
564
565
static const char nx_warning[] = KERN_CRIT
"kernel tried to execute NX-protected page - exploit attempt? (uid: %d)\n";

Ingo Molnar's avatar
Ingo Molnar committed
566
567
568
static void
show_fault_oops(struct pt_regs *regs, unsigned long error_code,
		unsigned long address)
569
{
570
571
572
573
	if (!oops_may_print())
		return;

	if (error_code & PF_INSTR) {
574
		unsigned int level;
Ingo Molnar's avatar
Ingo Molnar committed
575

576
577
		pte_t *pte = lookup_address(address, &level);

578
579
		if (pte && pte_present(*pte) && !pte_exec(*pte))
			printk(nx_warning, current_uid());
580
581
	}

582
	printk(KERN_ALERT "BUG: unable to handle kernel ");
583
	if (address < PAGE_SIZE)
584
		printk(KERN_CONT "NULL pointer dereference");
585
	else
586
		printk(KERN_CONT "paging request");
Ingo Molnar's avatar
Ingo Molnar committed
587

588
	printk(KERN_CONT " at %p\n", (void *) address);
589
	printk(KERN_ALERT "IP:");
590
	printk_address(regs->ip, 1);
Ingo Molnar's avatar
Ingo Molnar committed
591

592
593
594
	dump_pagetable(address);
}

Ingo Molnar's avatar
Ingo Molnar committed
595
596
597
static noinline void
pgtable_bad(struct pt_regs *regs, unsigned long error_code,
	    unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
598
{
Ingo Molnar's avatar
Ingo Molnar committed
599
600
601
602
603
604
605
	struct task_struct *tsk;
	unsigned long flags;
	int sig;

	flags = oops_begin();
	tsk = current;
	sig = SIGKILL;
606

Linus Torvalds's avatar
Linus Torvalds committed
607
	printk(KERN_ALERT "%s: Corrupted page table at address %lx\n",
608
	       tsk->comm, address);
Linus Torvalds's avatar
Linus Torvalds committed
609
	dump_pagetable(address);
Ingo Molnar's avatar
Ingo Molnar committed
610
611
612
613
614

	tsk->thread.cr2		= address;
	tsk->thread.trap_no	= 14;
	tsk->thread.error_code	= error_code;

615
	if (__die("Bad pagetable", regs, error_code))
616
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
617

618
	oops_end(flags, regs, sig);
Linus Torvalds's avatar
Linus Torvalds committed
619
620
}

Ingo Molnar's avatar
Ingo Molnar committed
621
622
623
static noinline void
no_context(struct pt_regs *regs, unsigned long error_code,
	   unsigned long address)
624
625
{
	struct task_struct *tsk = current;
626
	unsigned long *stackend;
627
628
629
	unsigned long flags;
	int sig;

Ingo Molnar's avatar
Ingo Molnar committed
630
	/* Are we prepared to handle this kernel fault? */
631
632
633
634
	if (fixup_exception(regs))
		return;

	/*
Ingo Molnar's avatar
Ingo Molnar committed
635
636
637
638
639
640
641
	 * 32-bit:
	 *
	 *   Valid to do another page fault here, because if this fault
	 *   had been triggered by is_prefetch fixup_exception would have
	 *   handled it.
	 *
	 * 64-bit:
642
	 *
Ingo Molnar's avatar
Ingo Molnar committed
643
	 *   Hall of shame of CPU/BIOS bugs.
644
645
646
647
648
649
650
651
652
	 */
	if (is_prefetch(regs, error_code, address))
		return;

	if (is_errata93(regs, address))
		return;

	/*
	 * Oops. The kernel tried to access some bad page. We'll have to
Ingo Molnar's avatar
Ingo Molnar committed
653
	 * terminate things with extreme prejudice:
654
655
656
657
658
	 */
	flags = oops_begin();

	show_fault_oops(regs, error_code, address);

Ingo Molnar's avatar
Ingo Molnar committed
659
	stackend = end_of_stack(tsk);
660
	if (tsk != &init_task && *stackend != STACK_END_MAGIC)
661
662
		printk(KERN_ALERT "Thread overran stack, or stack corrupted\n");

663
664
665
	tsk->thread.cr2		= address;
	tsk->thread.trap_no	= 14;
	tsk->thread.error_code	= error_code;
666
667
668
669

	sig = SIGKILL;
	if (__die("Oops", regs, error_code))
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
670

671
672
	/* Executive summary in case the body of the oops scrolled away */
	printk(KERN_EMERG "CR2: %016lx\n", address);
Ingo Molnar's avatar
Ingo Molnar committed
673

674
675
676
	oops_end(flags, regs, sig);
}

Ingo Molnar's avatar
Ingo Molnar committed
677
678
679
680
681
682
683
684
685
686
687
688
689
690
/*
 * Print out info about fatal segfaults, if the show_unhandled_signals
 * sysctl is set:
 */
static inline void
show_signal_msg(struct pt_regs *regs, unsigned long error_code,
		unsigned long address, struct task_struct *tsk)
{
	if (!unhandled_signal(tsk, SIGSEGV))
		return;

	if (!printk_ratelimit())
		return;

691
	printk("%s%s[%d]: segfault at %lx ip %p sp %p error %lx",
Ingo Molnar's avatar
Ingo Molnar committed
692
693
694
695
696
697
698
699
700
701
702
703
		task_pid_nr(tsk) > 1 ? KERN_INFO : KERN_EMERG,
		tsk->comm, task_pid_nr(tsk), address,
		(void *)regs->ip, (void *)regs->sp, error_code);

	print_vma_addr(KERN_CONT " in ", regs->ip);

	printk(KERN_CONT "\n");
}

static void
__bad_area_nosemaphore(struct pt_regs *regs, unsigned long error_code,
		       unsigned long address, int si_code)
704
705
706
707
708
709
{
	struct task_struct *tsk = current;

	/* User mode accesses just cause a SIGSEGV */
	if (error_code & PF_USER) {
		/*
Ingo Molnar's avatar
Ingo Molnar committed
710
		 * It's possible to have interrupts off here:
711
712
713
714
715
		 */
		local_irq_enable();

		/*
		 * Valid to do another page fault here because this one came
Ingo Molnar's avatar
Ingo Molnar committed
716
		 * from user space:
717
718
719
720
721
722
723
		 */
		if (is_prefetch(regs, error_code, address))
			return;

		if (is_errata100(regs, address))
			return;

724
725
726
727
728
729
730
731
732
733
734
735
#ifdef CONFIG_X86_64
		/*
		 * Instruction fetch faults in the vsyscall page might need
		 * emulation.
		 */
		if (unlikely((error_code & PF_INSTR) &&
			     ((address & ~0xfff) == VSYSCALL_START))) {
			if (emulate_vsyscall(regs, address))
				return;
		}
#endif

Ingo Molnar's avatar
Ingo Molnar committed
736
737
738
739
740
741
742
		if (unlikely(show_unhandled_signals))
			show_signal_msg(regs, error_code, address, tsk);

		/* Kernel addresses are always protection faults: */
		tsk->thread.cr2		= address;
		tsk->thread.error_code	= error_code | (address >= TASK_SIZE);
		tsk->thread.trap_no	= 14;
743

744
		force_sig_info_fault(SIGSEGV, si_code, address, tsk, 0);
Ingo Molnar's avatar
Ingo Molnar committed
745

746
747
748
749
750
751
752
753
754
		return;
	}

	if (is_f00f_bug(regs, address))
		return;

	no_context(regs, error_code, address);
}

Ingo Molnar's avatar
Ingo Molnar committed
755
756
757
static noinline void
bad_area_nosemaphore(struct pt_regs *regs, unsigned long error_code,
		     unsigned long address)
758
759
760
761
{
	__bad_area_nosemaphore(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
762
763
764
static void
__bad_area(struct pt_regs *regs, unsigned long error_code,
	   unsigned long address, int si_code)
765
766
767
768
769
770
771
772
773
774
775
776
{
	struct mm_struct *mm = current->mm;

	/*
	 * Something tried to access memory that isn't in our memory map..
	 * Fix it, but check if it's kernel or user first..
	 */
	up_read(&mm->mmap_sem);

	__bad_area_nosemaphore(regs, error_code, address, si_code);
}

Ingo Molnar's avatar
Ingo Molnar committed
777
778
static noinline void
bad_area(struct pt_regs *regs, unsigned long error_code, unsigned long address)
779
780
781
782
{
	__bad_area(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
783
784
785
static noinline void
bad_area_access_error(struct pt_regs *regs, unsigned long error_code,
		      unsigned long address)
786
787
788
789
790
{
	__bad_area(regs, error_code, address, SEGV_ACCERR);
}

/* TODO: fixup for "mm-invoke-oom-killer-from-page-fault.patch" */
Ingo Molnar's avatar
Ingo Molnar committed
791
792
793
static void
out_of_memory(struct pt_regs *regs, unsigned long error_code,
	      unsigned long address)
794
795
796
{
	/*
	 * We ran out of memory, call the OOM killer, and return the userspace
Ingo Molnar's avatar
Ingo Molnar committed
797
	 * (which will retry the fault, or kill us if we got oom-killed):
798
799
	 */
	up_read(&current->mm->mmap_sem);
Ingo Molnar's avatar
Ingo Molnar committed
800

801
802
803
	pagefault_out_of_memory();
}

Ingo Molnar's avatar
Ingo Molnar committed
804
static void
805
806
do_sigbus(struct pt_regs *regs, unsigned long error_code, unsigned long address,
	  unsigned int fault)
807
808
809
{
	struct task_struct *tsk = current;
	struct mm_struct *mm = tsk->mm;
810
	int code = BUS_ADRERR;
811
812
813

	up_read(&mm->mmap_sem);

Ingo Molnar's avatar
Ingo Molnar committed
814
	/* Kernel mode? Handle exceptions or die: */
815
	if (!(error_code & PF_USER)) {
816
		no_context(regs, error_code, address);
817
818
		return;
	}
Ingo Molnar's avatar
Ingo Molnar committed
819

820
	/* User-space => ok to do another page fault: */
821
822
	if (is_prefetch(regs, error_code, address))
		return;
Ingo Molnar's avatar
Ingo Molnar committed
823
824
825
826
827

	tsk->thread.cr2		= address;
	tsk->thread.error_code	= error_code;
	tsk->thread.trap_no	= 14;

828
#ifdef CONFIG_MEMORY_FAILURE
829
	if (fault & (VM_FAULT_HWPOISON|VM_FAULT_HWPOISON_LARGE)) {
830
831
832
833
834
835
		printk(KERN_ERR
	"MCE: Killing %s:%d due to hardware memory corruption fault at %lx\n",
			tsk->comm, tsk->pid, address);
		code = BUS_MCEERR_AR;
	}
#endif
836
	force_sig_info_fault(SIGBUS, code, address, tsk, fault);
837
838
}

839
static noinline int
Ingo Molnar's avatar
Ingo Molnar committed
840
841
mm_fault_error(struct pt_regs *regs, unsigned long error_code,
	       unsigned long address, unsigned int fault)
842
{
843
844
845
846
847
848
849
850
851
852
853
854
855
856
	/*
	 * Pagefault was interrupted by SIGKILL. We have no reason to
	 * continue pagefault.
	 */
	if (fatal_signal_pending(current)) {
		if (!(fault & VM_FAULT_RETRY))
			up_read(&current->mm->mmap_sem);
		if (!(error_code & PF_USER))
			no_context(regs, error_code, address);
		return 1;
	}
	if (!(fault & VM_FAULT_ERROR))
		return 0;

Ingo Molnar's avatar
Ingo Molnar committed
857
	if (fault & VM_FAULT_OOM) {
858
859
860
861
		/* Kernel mode? Handle exceptions or die: */
		if (!(error_code & PF_USER)) {
			up_read(&current->mm->mmap_sem);
			no_context(regs, error_code, address);
862
			return 1;
863
864
		}

865
		out_of_memory(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
866
	} else {
867
868
		if (fault & (VM_FAULT_SIGBUS|VM_FAULT_HWPOISON|
			     VM_FAULT_HWPOISON_LARGE))
869
			do_sigbus(regs, error_code, address, fault);
Ingo Molnar's avatar
Ingo Molnar committed
870
871
872
		else
			BUG();
	}
873
	return 1;
874
875
}

876
877
878
879
static int spurious_fault_check(unsigned long error_code, pte_t *pte)
{
	if ((error_code & PF_WRITE) && !pte_write(*pte))
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
880

881
882
883
884
885
886
	if ((error_code & PF_INSTR) && !pte_exec(*pte))
		return 0;

	return 1;
}

887
/*
Ingo Molnar's avatar
Ingo Molnar committed
888
889
890
891
892
893
894
895
 * Handle a spurious fault caused by a stale TLB entry.
 *
 * This allows us to lazily refresh the TLB when increasing the
 * permissions of a kernel page (RO -> RW or NX -> X).  Doing it
 * eagerly is very expensive since that implies doing a full
 * cross-processor TLB flush, even if no stale TLB entries exist
 * on other processors.
 *
896
897
898
 * There are no security implications to leaving a stale TLB when
 * increasing the permissions on a page.
 */
899
static noinline __kprobes int
Ingo Molnar's avatar
Ingo Molnar committed
900
spurious_fault(unsigned long error_code, unsigned long address)
901
902
903
904
905
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;
906
	int ret;
907
908
909
910
911
912
913
914
915
916
917
918
919

	/* Reserved-bit violation or user access to kernel space? */
	if (error_code & (PF_USER | PF_RSVD))
		return 0;

	pgd = init_mm.pgd + pgd_index(address);
	if (!pgd_present(*pgd))
		return 0;

	pud = pud_offset(pgd, address);
	if (!pud_present(*pud))
		return 0;

920
921
922
	if (pud_large(*pud))
		return spurious_fault_check(error_code, (pte_t *) pud);

923
924
925
926
	pmd = pmd_offset(pud, address);
	if (!pmd_present(*pmd))
		return 0;

927
928
929
	if (pmd_large(*pmd))
		return spurious_fault_check(error_code, (pte_t *) pmd);

930
931
932
933
934
935
	/*
	 * Note: don't use pte_present() here, since it returns true
	 * if the _PAGE_PROTNONE bit is set.  However, this aliases the
	 * _PAGE_GLOBAL bit, which for kernel pages give false positives
	 * when CONFIG_DEBUG_PAGEALLOC is used.
	 */
936
	pte = pte_offset_kernel(pmd, address);
937
	if (!(pte_flags(*pte) & _PAGE_PRESENT))
938
939
		return 0;

940
941
942
943
944
	ret = spurious_fault_check(error_code, pte);
	if (!ret)
		return 0;

	/*
Ingo Molnar's avatar
Ingo Molnar committed
945
946
	 * Make sure we have permissions in PMD.
	 * If not, then there's a bug in the page tables:
947
948
949
	 */
	ret = spurious_fault_check(error_code, (pte_t *) pmd);
	WARN_ONCE(!ret, "PMD has incorrect permission bits\n");
Ingo Molnar's avatar
Ingo Molnar committed
950

951
	return ret;
952
953
}

954
int show_unhandled_signals = 1;
Linus Torvalds's avatar
Linus Torvalds committed
955

Ingo Molnar's avatar
Ingo Molnar committed
956
static inline int
957
access_error(unsigned long error_code, struct vm_area_struct *vma)
958
{
959
	if (error_code & PF_WRITE) {
Ingo Molnar's avatar
Ingo Molnar committed
960
		/* write, present and write, not present: */
961
962
		if (unlikely(!(vma->vm_flags & VM_WRITE)))
			return 1;
Ingo Molnar's avatar
Ingo Molnar committed
963
		return 0;
964
965
	}

Ingo Molnar's avatar
Ingo Molnar committed
966
967
968
969
970
971
972
973
	/* read, present: */
	if (unlikely(error_code & PF_PROT))
		return 1;

	/* read, not present: */
	if (unlikely(!(vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE))))
		return 1;

974
975
976
	return 0;
}

977
978
static int fault_in_kernel_space(unsigned long address)
{
979
	return address >= TASK_SIZE_MAX;
980
981
}

Linus Torvalds's avatar
Linus Torvalds committed
982
983
984
985
986
/*
 * This routine handles page faults.  It determines the address,
 * and the problem, and then passes it off to one of the appropriate
 * routines.
 */
987
988
dotraplinkage void __kprobes
do_page_fault(struct pt_regs *regs, unsigned long error_code)
Linus Torvalds's avatar
Linus Torvalds committed
989
{
Ingo Molnar's avatar
Ingo Molnar committed
990
	struct vm_area_struct *vma;
Linus Torvalds's avatar
Linus Torvalds committed
991
	struct task_struct *tsk;
Ingo Molnar's avatar
Ingo Molnar committed
992
	unsigned long address;
Linus Torvalds's avatar
Linus Torvalds committed
993
	struct mm_struct *mm;
994
	int fault;
995
	int write = error_code & PF_WRITE;
996
	unsigned int flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE |
997
					(write ? FAULT_FLAG_WRITE : 0);
Linus Torvalds's avatar
Linus Torvalds committed
998

999
1000
	tsk = current;
	mm = tsk->mm;
Ingo Molnar's avatar
Ingo Molnar committed
1001
1002

	/* Get the faulting address: */
1003
	address = read_cr2();
Linus Torvalds's avatar
Linus Torvalds committed
1004

Vegard Nossum's avatar
Vegard Nossum committed
1005
1006
1007
1008
1009
1010
	/*
	 * Detect and handle instructions that would cause a page fault for
	 * both a tracked kernel page and a userspace page.
	 */
	if (kmemcheck_active(regs))
		kmemcheck_hide(regs);
1011
	prefetchw(&mm->mmap_sem);
Vegard Nossum's avatar
Vegard Nossum committed
1012

1013
	if (unlikely(kmmio_fault(regs, address)))
1014
		return;
Linus Torvalds's avatar
Linus Torvalds committed
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026

	/*
	 * We fault-in kernel-space virtual memory on-demand. The
	 * 'reference' page table is init_mm.pgd.
	 *
	 * NOTE! We MUST NOT take any locks for this case. We may
	 * be in an interrupt or a critical region, and should
	 * only copy the information from the master page table,
	 * nothing more.
	 *
	 * This verifies that the fault happens in kernel space
	 * (error_code & 4) == 0, and that the fault was not a
1027
	 * protection error (error_code & 9) == 0.
Linus Torvalds's avatar
Linus Torvalds committed
1028
	 */
1029
	if (unlikely(fault_in_kernel_space(address))) {
Vegard Nossum's avatar
Vegard Nossum committed
1030
1031
1032
1033
1034
1035
1036
		if (!(error_code & (PF_RSVD | PF_USER | PF_PROT))) {
			if (vmalloc_fault(address) >= 0)
				return;

			if (kmemcheck_fault(regs, address, error_code))
				return;
		}
1037

Ingo Molnar's avatar
Ingo Molnar committed
1038
		/* Can handle a stale RO->RW TLB: */
1039
		if (spurious_fault(error_code, address))
1040
1041
			return;

Ingo Molnar's avatar
Ingo Molnar committed
1042
		/* kprobes don't want to hook the spurious faults: */
1043
1044
		if (notify_page_fault(regs))
			return;
1045
1046
		/*
		 * Don't take the mm semaphore here. If we fixup a prefetch
Ingo Molnar's avatar
Ingo Molnar committed
1047
		 * fault we could otherwise deadlock:
1048
		 */
1049
		bad_area_nosemaphore(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
1050

1051
		return;
1052
1053
	}

Ingo Molnar's avatar
Ingo Molnar committed
1054
	/* kprobes don't want to hook the spurious faults: */
1055
	if (unlikely(notify_page_fault(regs)))
1056
		return;
1057
	/*
1058
1059
1060
1061
	 * It's safe to allow irq's after cr2 has been saved and the
	 * vmalloc fault has been handled.
	 *
	 * User-mode registers count as a user access even for any
Ingo Molnar's avatar
Ingo Molnar committed
1062
	 * potential system fault or CPU buglet:
1063
	 */
1064
1065
1066
	if (user_mode_vm(regs)) {
		local_irq_enable();
		error_code |= PF_USER;
Ingo Molnar's avatar
Ingo Molnar committed
1067
1068
1069
1070
	} else {
		if (regs->flags & X86_EFLAGS_IF)
			local_irq_enable();
	}
1071

1072
	if (unlikely(error_code & PF_RSVD))
1073
		pgtable_bad(regs, error_code, address);
Linus Torvalds's avatar
Linus Torvalds committed
1074

1075
	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address);
1076

Linus Torvalds's avatar
Linus Torvalds committed
1077
	/*
Ingo Molnar's avatar
Ingo Molnar committed
1078
1079
	 * If we're in an interrupt, have no user context or are running
	 * in an atomic region then we must not take the fault:
Linus Torvalds's avatar
Linus Torvalds committed
1080
	 */
1081
1082
1083
1084
	if (unlikely(in_atomic() || !mm)) {
		bad_area_nosemaphore(regs, error_code, address);
		return;
	}
Linus Torvalds's avatar
Linus Torvalds committed
1085

1086
1087
	/*
	 * When running in the kernel we expect faults to occur only to
Ingo Molnar's avatar
Ingo Molnar committed
1088
1089
1090
1091
1092
1093
1094
	 * addresses in user space.  All other faults represent errors in
	 * the kernel and should generate an OOPS.  Unfortunately, in the
	 * case of an erroneous fault occurring in a code path which already
	 * holds mmap_sem we will deadlock attempting to validate the fault
	 * against the address space.  Luckily the kernel only validly
	 * references user space from well defined areas of code, which are
	 * listed in the exceptions table.
Linus Torvalds's avatar
Linus Torvalds committed
1095
1096
	 *
	 * As the vast majority of faults will be valid we will only perform
Ingo Molnar's avatar
Ingo Molnar committed
1097
1098
1099
1100
	 * the source reference check when there is a possibility of a
	 * deadlock. Attempt to lock the address space, if we cannot we then
	 * validate the source. If this is invalid we can skip the address
	 * space check, thus avoiding the deadlock:
Linus Torvalds's avatar
Linus Torvalds committed
1101
	 */
1102
	if (unlikely(!down_read_trylock(&mm->mmap_sem))) {
1103
		if ((error_code & PF_USER) == 0 &&
1104
1105
1106
1107
		    !search_exception_tables(regs->ip)) {
			bad_area_nosemaphore(regs, error_code, address);
			return;
		}
1108
retry:
Linus Torvalds's avatar
Linus Torvalds committed
1109
		down_read(&mm->mmap_sem);
1110
1111
	} else {
		/*
Ingo Molnar's avatar
Ingo Molnar committed
1112
1113
1114
		 * The above down_read_trylock() might have succeeded in
		 * which case we'll have missed the might_sleep() from
		 * down_read():
1115
1116
		 */
		might_sleep();
Linus Torvalds's avatar
Linus Torvalds committed
1117
1118
1119
	}

	vma = find_vma(mm, address);
1120
1121
1122
1123
1124
	if (unlikely(!vma)) {
		bad_area(regs, error_code, address);
		return;
	}
	if (likely(vma->vm_start <= address))
Linus Torvalds's avatar
Linus Torvalds committed
1125
		goto good_area;
1126
1127
1128
1129
	if (unlikely(!(vma->vm_flags & VM_GROWSDOWN))) {
		bad_area(regs, error_code, address);
		return;
	}
1130
	if (error_code & PF_USER) {
1131
1132
1133
		/*
		 * Accessing the stack below %sp is always a bug.
		 * The large cushion allows instructions like enter
Ingo Molnar's avatar
Ingo Molnar committed
1134
		 * and pusha to work. ("enter $65535, $31" pushes
1135
		 * 32 pointers and then decrements %sp by 65535.)
1136
		 */
1137
1138
1139
1140
		if (unlikely(address + 65536 + 32 * sizeof(unsigned long) < regs->sp)) {
			bad_area(regs, error_code, address);
			return;
		}
Linus Torvalds's avatar
Linus Torvalds committed
1141
	}
1142
1143
1144
1145
1146
1147
1148
1149
1150
	if (unlikely(expand_stack(vma, address))) {
		bad_area(regs, error_code, address);
		return;
	}

	/*
	 * Ok, we have a good vm_area for this memory access, so
	 * we can handle it..
	 */
Linus Torvalds's avatar
Linus Torvalds committed
1151
good_area:
1152
	if (unlikely(access_error(error_code, vma))) {
1153
1154
		bad_area_access_error(regs, error_code, address);
		return;
Linus Torvalds's avatar
Linus Torvalds committed
1155
1156
1157
1158
1159
	}

	/*
	 * If for any reason at all we couldn't handle the fault,
	 * make sure we exit gracefully rather than endlessly redo
Ingo Molnar's avatar
Ingo Molnar committed
1160
	 * the fault:
Linus Torvalds's avatar
Linus Torvalds committed
1161
	 */
1162
	fault = handle_mm_fault(mm, vma, address, flags);
Ingo Molnar's avatar
Ingo Molnar committed
1163

1164
1165
1166
	if (unlikely(fault & (VM_FAULT_RETRY|VM_FAULT_ERROR))) {
		if (mm_fault_error(regs, error_code, address, fault))
			return;
1167
1168
	}

1169
1170
1171
1172
1173
1174
1175
1176
	/*
	 * Major/minor page fault accounting is only done on the
	 * initial attempt. If we go through a retry, it is extremely
	 * likely that the page will be found in page cache at that point.
	 */
	if (flags & FAULT_FLAG_ALLOW_RETRY) {
		if (fault & VM_FAULT_MAJOR) {
			tsk->maj_flt++;
1177
			perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1,
1178
1179
1180
				      regs, address);
		} else {
			tsk->min_flt++;
1181
			perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1,
1182
1183
1184
1185
1186
1187
1188
1189
				      regs, address);
		}
		if (fault & VM_FAULT_RETRY) {
			/* Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk
			 * of starvation. */
			flags &= ~FAULT_FLAG_ALLOW_RETRY;
			goto retry;
		}
1190
	}
1191

1192
1193
	check_v8086_mode(regs, address, tsk);

Linus Torvalds's avatar
Linus Torvalds committed
1194
1195
	up_read(&mm->mmap_sem);
}