fault.c 27.9 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_*(), ...		*/
Linus Torvalds's avatar
Linus Torvalds committed
20

21
/*
Ingo Molnar's avatar
Ingo Molnar committed
22
23
24
25
26
27
28
 * 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
29
 */
Ingo Molnar's avatar
Ingo Molnar committed
30
31
32
33
34
35
36
37
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,
};
38

39
/*
40
41
 * Returns 0 if mmiotrace is disabled, or if the fault is not
 * handled by mmiotrace:
42
 */
43
44
static inline int __kprobes
kmmio_fault(struct pt_regs *regs, unsigned long addr)
45
{
46
47
48
49
	if (unlikely(is_kmmio_active()))
		if (kmmio_handler(regs, addr) == 1)
			return -1;
	return 0;
50
51
}

52
static inline int __kprobes notify_page_fault(struct pt_regs *regs)
53
{
54
55
56
	int ret = 0;

	/* kprobe_running() needs smp_processor_id() */
57
	if (kprobes_built_in() && !user_mode_vm(regs)) {
58
59
60
61
62
		preempt_disable();
		if (kprobe_running() && kprobe_fault_handler(regs, 14))
			ret = 1;
		preempt_enable();
	}
63

64
	return ret;
65
}
66

67
/*
Ingo Molnar's avatar
Ingo Molnar committed
68
69
70
71
72
73
 * Prefetch quirks:
 *
 * 32-bit mode:
 *
 *   Sometimes AMD Athlon/Opteron CPUs report invalid exceptions on prefetch.
 *   Check that here and ignore it.
74
 *
Ingo Molnar's avatar
Ingo Molnar committed
75
 * 64-bit mode:
76
 *
Ingo Molnar's avatar
Ingo Molnar committed
77
78
79
80
 *   Sometimes the CPU reports invalid exceptions on prefetch.
 *   Check that here and ignore it.
 *
 * Opcode checker based on code by Richard Brunner.
81
 */
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
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.
		 */
108
		return (!user_mode(regs) || user_64bit_mode(regs));
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
#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
129
130
static int
is_prefetch(struct pt_regs *regs, unsigned long error_code, unsigned long addr)
131
{
Ingo Molnar's avatar
Ingo Molnar committed
132
	unsigned char *max_instr;
133
	unsigned char *instr;
134
	int prefetch = 0;
Linus Torvalds's avatar
Linus Torvalds committed
135

Ingo Molnar's avatar
Ingo Molnar committed
136
137
138
139
	/*
	 * If it was a exec (instruction fetch) fault on NX page, then
	 * do not ignore the fault:
	 */
140
	if (error_code & PF_INSTR)
Linus Torvalds's avatar
Linus Torvalds committed
141
		return 0;
142

143
	instr = (void *)convert_ip_to_linear(current, regs);
144
	max_instr = instr + 15;
Linus Torvalds's avatar
Linus Torvalds committed
145

146
	if (user_mode(regs) && instr >= (unsigned char *)TASK_SIZE)
Linus Torvalds's avatar
Linus Torvalds committed
147
148
		return 0;

149
	while (instr < max_instr) {
Ingo Molnar's avatar
Ingo Molnar committed
150
		unsigned char opcode;
Linus Torvalds's avatar
Linus Torvalds committed
151

152
		if (probe_kernel_address(instr, opcode))
153
			break;
Linus Torvalds's avatar
Linus Torvalds committed
154
155
156

		instr++;

157
		if (!check_prefetch_opcode(regs, instr, opcode, &prefetch))
Linus Torvalds's avatar
Linus Torvalds committed
158
159
160
161
162
			break;
	}
	return prefetch;
}

Ingo Molnar's avatar
Ingo Molnar committed
163
164
static void
force_sig_info_fault(int si_signo, int si_code, unsigned long address,
165
		     struct task_struct *tsk, int fault)
166
{
167
	unsigned lsb = 0;
168
169
	siginfo_t info;

Ingo Molnar's avatar
Ingo Molnar committed
170
171
172
173
	info.si_signo	= si_signo;
	info.si_errno	= 0;
	info.si_code	= si_code;
	info.si_addr	= (void __user *)address;
174
175
176
177
178
	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
179

180
181
182
	force_sig_info(si_signo, &info, tsk);
}

183
184
185
186
187
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)
188
{
189
190
191
192
	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
193

194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
	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;

215
	if (!pmd_present(*pmd))
216
		set_pmd(pmd, *pmd_k);
217
	else
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
		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;

235
		spin_lock(&pgd_lock);
236
		list_for_each_entry(page, &pgd_list, lru) {
237
			spinlock_t *pgt_lock;
238
			pmd_t *ret;
239

240
			/* the pgt_lock only for Xen */
241
242
243
244
245
246
247
			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)
248
249
				break;
		}
250
		spin_unlock(&pgd_lock);
251
252
253
254
255
256
257
258
	}
}

/*
 * 32-bit:
 *
 *   Handle a fault on the vmalloc or module mapping area
 */
259
static noinline __kprobes int vmalloc_fault(unsigned long address)
260
261
262
263
264
265
266
267
268
{
	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;

269
270
	WARN_ON_ONCE(in_nmi());

271
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
	/*
	 * 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;
305
}
Linus Torvalds's avatar
Linus Torvalds committed
306

Akinobu Mita's avatar
Akinobu Mita committed
307
static bool low_pfn(unsigned long pfn)
Linus Torvalds's avatar
Linus Torvalds committed
308
{
Akinobu Mita's avatar
Akinobu Mita committed
309
310
	return pfn < max_low_pfn;
}
311

Akinobu Mita's avatar
Akinobu Mita committed
312
313
314
315
316
317
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
318

319
#ifdef CONFIG_X86_PAE
Akinobu Mita's avatar
Akinobu Mita committed
320
321
322
	printk("*pdpt = %016Lx ", pgd_val(*pgd));
	if (!low_pfn(pgd_val(*pgd) >> PAGE_SHIFT) || !pgd_present(*pgd))
		goto out;
323
#endif
Akinobu Mita's avatar
Akinobu Mita committed
324
325
	pmd = pmd_offset(pud_offset(pgd, address), address);
	printk(KERN_CONT "*pde = %0*Lx ", sizeof(*pmd) * 2, (u64)pmd_val(*pmd));
326
327
328
329
330

	/*
	 * 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
331
	 * it's allocated already:
332
	 */
Akinobu Mita's avatar
Akinobu Mita committed
333
334
	if (!low_pfn(pmd_pfn(*pmd)) || !pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
335

Akinobu Mita's avatar
Akinobu Mita committed
336
337
338
	pte = pte_offset_kernel(pmd, address);
	printk("*pte = %0*Lx ", sizeof(*pte) * 2, (u64)pte_val(*pte));
out:
339
	printk("\n");
340
341
342
343
344
345
}

#else /* CONFIG_X86_64: */

void vmalloc_sync_all(void)
{
346
	sync_global_pgds(VMALLOC_START & PGDIR_MASK, VMALLOC_END);
347
348
349
350
351
352
353
354
355
}

/*
 * 64-bit:
 *
 *   Handle a fault on the vmalloc area
 *
 * This assumes no large pages in there.
 */
356
static noinline __kprobes int vmalloc_fault(unsigned long address)
357
358
359
360
361
362
363
364
365
366
{
	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;

367
368
	WARN_ON_ONCE(in_nmi());

369
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
	/*
	 * 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[] =
423
424
425
426
427
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";
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446

/*
 * 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
447
448
	pgd_t *base = __va(read_cr3() & PHYSICAL_PAGE_MASK);
	pgd_t *pgd = base + pgd_index(address);
Linus Torvalds's avatar
Linus Torvalds committed
449
450
451
452
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;

Ingo Molnar's avatar
Ingo Molnar committed
453
454
455
	if (bad_address(pgd))
		goto bad;

456
	printk("PGD %lx ", pgd_val(*pgd));
Ingo Molnar's avatar
Ingo Molnar committed
457
458
459

	if (!pgd_present(*pgd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
460

461
	pud = pud_offset(pgd, address);
Ingo Molnar's avatar
Ingo Molnar committed
462
463
464
	if (bad_address(pud))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
465
	printk("PUD %lx ", pud_val(*pud));
466
	if (!pud_present(*pud) || pud_large(*pud))
Ingo Molnar's avatar
Ingo Molnar committed
467
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
468
469

	pmd = pmd_offset(pud, address);
Ingo Molnar's avatar
Ingo Molnar committed
470
471
472
	if (bad_address(pmd))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
473
	printk("PMD %lx ", pmd_val(*pmd));
Ingo Molnar's avatar
Ingo Molnar committed
474
475
	if (!pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
476
477

	pte = pte_offset_kernel(pmd, address);
Ingo Molnar's avatar
Ingo Molnar committed
478
479
480
	if (bad_address(pte))
		goto bad;

481
	printk("PTE %lx", pte_val(*pte));
Ingo Molnar's avatar
Ingo Molnar committed
482
out:
Linus Torvalds's avatar
Linus Torvalds committed
483
484
485
486
	printk("\n");
	return;
bad:
	printk("BAD\n");
487
488
}

489
#endif /* CONFIG_X86_64 */
Linus Torvalds's avatar
Linus Torvalds committed
490

Ingo Molnar's avatar
Ingo Molnar committed
491
492
493
494
495
496
497
498
499
500
501
502
503
/*
 * 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.
504
 */
505
static int is_errata93(struct pt_regs *regs, unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
506
{
507
#ifdef CONFIG_X86_64
508
	if (address != regs->ip)
Linus Torvalds's avatar
Linus Torvalds committed
509
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
510

511
	if ((address >> 32) != 0)
Linus Torvalds's avatar
Linus Torvalds committed
512
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
513

Linus Torvalds's avatar
Linus Torvalds committed
514
	address |= 0xffffffffUL << 32;
515
516
	if ((address >= (u64)_stext && address <= (u64)_etext) ||
	    (address >= MODULES_VADDR && address <= MODULES_END)) {
517
		printk_once(errata93_warning);
518
		regs->ip = address;
Linus Torvalds's avatar
Linus Torvalds committed
519
520
		return 1;
	}
521
#endif
Linus Torvalds's avatar
Linus Torvalds committed
522
	return 0;
523
}
Linus Torvalds's avatar
Linus Torvalds committed
524

525
/*
Ingo Molnar's avatar
Ingo Molnar committed
526
527
528
529
530
 * 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
531
532
533
534
535
 * 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
536
	if ((regs->cs == __USER32_CS || (regs->cs & (1<<2))) && (address >> 32))
537
538
539
540
541
		return 1;
#endif
	return 0;
}

542
543
544
545
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
546

547
	/*
Ingo Molnar's avatar
Ingo Molnar committed
548
	 * Pentium F0 0F C7 C8 bug workaround:
549
550
551
552
553
554
555
556
557
558
559
560
561
	 */
	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;
}

562
563
564
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
565
566
567
static void
show_fault_oops(struct pt_regs *regs, unsigned long error_code,
		unsigned long address)
568
{
569
570
571
572
	if (!oops_may_print())
		return;

	if (error_code & PF_INSTR) {
573
		unsigned int level;
Ingo Molnar's avatar
Ingo Molnar committed
574

575
576
		pte_t *pte = lookup_address(address, &level);

577
578
		if (pte && pte_present(*pte) && !pte_exec(*pte))
			printk(nx_warning, current_uid());
579
580
	}

581
	printk(KERN_ALERT "BUG: unable to handle kernel ");
582
	if (address < PAGE_SIZE)
583
		printk(KERN_CONT "NULL pointer dereference");
584
	else
585
		printk(KERN_CONT "paging request");
Ingo Molnar's avatar
Ingo Molnar committed
586

587
	printk(KERN_CONT " at %p\n", (void *) address);
588
	printk(KERN_ALERT "IP:");
589
	printk_address(regs->ip, 1);
Ingo Molnar's avatar
Ingo Molnar committed
590

591
592
593
	dump_pagetable(address);
}

Ingo Molnar's avatar
Ingo Molnar committed
594
595
596
static noinline void
pgtable_bad(struct pt_regs *regs, unsigned long error_code,
	    unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
597
{
Ingo Molnar's avatar
Ingo Molnar committed
598
599
600
601
602
603
604
	struct task_struct *tsk;
	unsigned long flags;
	int sig;

	flags = oops_begin();
	tsk = current;
	sig = SIGKILL;
605

Linus Torvalds's avatar
Linus Torvalds committed
606
	printk(KERN_ALERT "%s: Corrupted page table at address %lx\n",
607
	       tsk->comm, address);
Linus Torvalds's avatar
Linus Torvalds committed
608
	dump_pagetable(address);
Ingo Molnar's avatar
Ingo Molnar committed
609
610
611
612
613

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

614
	if (__die("Bad pagetable", regs, error_code))
615
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
616

617
	oops_end(flags, regs, sig);
Linus Torvalds's avatar
Linus Torvalds committed
618
619
}

Ingo Molnar's avatar
Ingo Molnar committed
620
621
622
static noinline void
no_context(struct pt_regs *regs, unsigned long error_code,
	   unsigned long address)
623
624
{
	struct task_struct *tsk = current;
625
	unsigned long *stackend;
626
627
628
	unsigned long flags;
	int sig;

Ingo Molnar's avatar
Ingo Molnar committed
629
	/* Are we prepared to handle this kernel fault? */
630
631
632
633
	if (fixup_exception(regs))
		return;

	/*
Ingo Molnar's avatar
Ingo Molnar committed
634
635
636
637
638
639
640
	 * 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:
641
	 *
Ingo Molnar's avatar
Ingo Molnar committed
642
	 *   Hall of shame of CPU/BIOS bugs.
643
644
645
646
647
648
649
650
651
	 */
	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
652
	 * terminate things with extreme prejudice:
653
654
655
656
657
	 */
	flags = oops_begin();

	show_fault_oops(regs, error_code, address);

Ingo Molnar's avatar
Ingo Molnar committed
658
	stackend = end_of_stack(tsk);
659
	if (tsk != &init_task && *stackend != STACK_END_MAGIC)
660
661
		printk(KERN_ALERT "Thread overran stack, or stack corrupted\n");

662
663
664
	tsk->thread.cr2		= address;
	tsk->thread.trap_no	= 14;
	tsk->thread.error_code	= error_code;
665
666
667
668

	sig = SIGKILL;
	if (__die("Oops", regs, error_code))
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
669

670
671
	/* 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
672

673
674
675
	oops_end(flags, regs, sig);
}

Ingo Molnar's avatar
Ingo Molnar committed
676
677
678
679
680
681
682
683
684
685
686
687
688
689
/*
 * 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;

690
	printk("%s%s[%d]: segfault at %lx ip %p sp %p error %lx",
Ingo Molnar's avatar
Ingo Molnar committed
691
692
693
694
695
696
697
698
699
700
701
702
		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)
703
704
705
706
707
708
{
	struct task_struct *tsk = current;

	/* User mode accesses just cause a SIGSEGV */
	if (error_code & PF_USER) {
		/*
Ingo Molnar's avatar
Ingo Molnar committed
709
		 * It's possible to have interrupts off here:
710
711
712
713
714
		 */
		local_irq_enable();

		/*
		 * Valid to do another page fault here because this one came
Ingo Molnar's avatar
Ingo Molnar committed
715
		 * from user space:
716
717
718
719
720
721
722
		 */
		if (is_prefetch(regs, error_code, address))
			return;

		if (is_errata100(regs, address))
			return;

Ingo Molnar's avatar
Ingo Molnar committed
723
724
725
726
727
728
729
		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;
730

731
		force_sig_info_fault(SIGSEGV, si_code, address, tsk, 0);
Ingo Molnar's avatar
Ingo Molnar committed
732

733
734
735
736
737
738
739
740
741
		return;
	}

	if (is_f00f_bug(regs, address))
		return;

	no_context(regs, error_code, address);
}

Ingo Molnar's avatar
Ingo Molnar committed
742
743
744
static noinline void
bad_area_nosemaphore(struct pt_regs *regs, unsigned long error_code,
		     unsigned long address)
745
746
747
748
{
	__bad_area_nosemaphore(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
749
750
751
static void
__bad_area(struct pt_regs *regs, unsigned long error_code,
	   unsigned long address, int si_code)
752
753
754
755
756
757
758
759
760
761
762
763
{
	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
764
765
static noinline void
bad_area(struct pt_regs *regs, unsigned long error_code, unsigned long address)
766
767
768
769
{
	__bad_area(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
770
771
772
static noinline void
bad_area_access_error(struct pt_regs *regs, unsigned long error_code,
		      unsigned long address)
773
774
775
776
777
{
	__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
778
779
780
static void
out_of_memory(struct pt_regs *regs, unsigned long error_code,
	      unsigned long address)
781
782
783
{
	/*
	 * We ran out of memory, call the OOM killer, and return the userspace
Ingo Molnar's avatar
Ingo Molnar committed
784
	 * (which will retry the fault, or kill us if we got oom-killed):
785
786
	 */
	up_read(&current->mm->mmap_sem);
Ingo Molnar's avatar
Ingo Molnar committed
787

788
789
790
	pagefault_out_of_memory();
}

Ingo Molnar's avatar
Ingo Molnar committed
791
static void
792
793
do_sigbus(struct pt_regs *regs, unsigned long error_code, unsigned long address,
	  unsigned int fault)
794
795
796
{
	struct task_struct *tsk = current;
	struct mm_struct *mm = tsk->mm;
797
	int code = BUS_ADRERR;
798
799
800

	up_read(&mm->mmap_sem);

Ingo Molnar's avatar
Ingo Molnar committed
801
	/* Kernel mode? Handle exceptions or die: */
802
	if (!(error_code & PF_USER)) {
803
		no_context(regs, error_code, address);
804
805
		return;
	}
Ingo Molnar's avatar
Ingo Molnar committed
806

807
	/* User-space => ok to do another page fault: */
808
809
	if (is_prefetch(regs, error_code, address))
		return;
Ingo Molnar's avatar
Ingo Molnar committed
810
811
812
813
814

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

815
#ifdef CONFIG_MEMORY_FAILURE
816
	if (fault & (VM_FAULT_HWPOISON|VM_FAULT_HWPOISON_LARGE)) {
817
818
819
820
821
822
		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
823
	force_sig_info_fault(SIGBUS, code, address, tsk, fault);
824
825
}

826
static noinline int
Ingo Molnar's avatar
Ingo Molnar committed
827
828
mm_fault_error(struct pt_regs *regs, unsigned long error_code,
	       unsigned long address, unsigned int fault)
829
{
830
831
832
833
834
835
836
837
838
839
840
841
842
843
	/*
	 * 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
844
	if (fault & VM_FAULT_OOM) {
845
846
847
848
		/* Kernel mode? Handle exceptions or die: */
		if (!(error_code & PF_USER)) {
			up_read(&current->mm->mmap_sem);
			no_context(regs, error_code, address);
849
			return 1;
850
851
		}

852
		out_of_memory(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
853
	} else {
854
855
		if (fault & (VM_FAULT_SIGBUS|VM_FAULT_HWPOISON|
			     VM_FAULT_HWPOISON_LARGE))
856
			do_sigbus(regs, error_code, address, fault);
Ingo Molnar's avatar
Ingo Molnar committed
857
858
859
		else
			BUG();
	}
860
	return 1;
861
862
}

863
864
865
866
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
867

868
869
870
871
872
873
	if ((error_code & PF_INSTR) && !pte_exec(*pte))
		return 0;

	return 1;
}

874
/*
Ingo Molnar's avatar
Ingo Molnar committed
875
876
877
878
879
880
881
882
 * 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.
 *
883
884
885
 * There are no security implications to leaving a stale TLB when
 * increasing the permissions on a page.
 */
886
static noinline __kprobes int
Ingo Molnar's avatar
Ingo Molnar committed
887
spurious_fault(unsigned long error_code, unsigned long address)
888
889
890
891
892
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;
893
	int ret;
894
895
896
897
898
899
900
901
902
903
904
905
906

	/* 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;

907
908
909
	if (pud_large(*pud))
		return spurious_fault_check(error_code, (pte_t *) pud);

910
911
912
913
	pmd = pmd_offset(pud, address);
	if (!pmd_present(*pmd))
		return 0;

914
915
916
	if (pmd_large(*pmd))
		return spurious_fault_check(error_code, (pte_t *) pmd);

917
918
919
920
921
922
	/*
	 * 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.
	 */
923
	pte = pte_offset_kernel(pmd, address);
924
	if (!(pte_flags(*pte) & _PAGE_PRESENT))
925
926
		return 0;

927
928
929
930
931
	ret = spurious_fault_check(error_code, pte);
	if (!ret)
		return 0;

	/*
Ingo Molnar's avatar
Ingo Molnar committed
932
933
	 * Make sure we have permissions in PMD.
	 * If not, then there's a bug in the page tables:
934
935
936
	 */
	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
937

938
	return ret;
939
940
}

941
int show_unhandled_signals = 1;
Linus Torvalds's avatar
Linus Torvalds committed
942

Ingo Molnar's avatar
Ingo Molnar committed
943
static inline int
944
access_error(unsigned long error_code, struct vm_area_struct *vma)
945
{
946
	if (error_code & PF_WRITE) {
Ingo Molnar's avatar
Ingo Molnar committed
947
		/* write, present and write, not present: */
948
949
		if (unlikely(!(vma->vm_flags & VM_WRITE)))
			return 1;
Ingo Molnar's avatar
Ingo Molnar committed
950
		return 0;
951
952
	}

Ingo Molnar's avatar
Ingo Molnar committed
953
954
955
956
957
958
959
960
	/* 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;

961
962
963
	return 0;
}

964
965
static int fault_in_kernel_space(unsigned long address)
{
966
	return address >= TASK_SIZE_MAX;
967
968
}

Linus Torvalds's avatar
Linus Torvalds committed
969
970
971
972
973
/*
 * This routine handles page faults.  It determines the address,
 * and the problem, and then passes it off to one of the appropriate
 * routines.
 */
974
975
dotraplinkage void __kprobes
do_page_fault(struct pt_regs *regs, unsigned long error_code)
Linus Torvalds's avatar
Linus Torvalds committed
976
{
Ingo Molnar's avatar
Ingo Molnar committed
977
	struct vm_area_struct *vma;
Linus Torvalds's avatar
Linus Torvalds committed
978
	struct task_struct *tsk;
Ingo Molnar's avatar
Ingo Molnar committed
979
	unsigned long address;
Linus Torvalds's avatar
Linus Torvalds committed
980
	struct mm_struct *mm;
981
	int fault;
982
	int write = error_code & PF_WRITE;
983
	unsigned int flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE |
984
					(write ? FAULT_FLAG_WRITE : 0);
Linus Torvalds's avatar
Linus Torvalds committed
985

986
987
	tsk = current;
	mm = tsk->mm;
Ingo Molnar's avatar
Ingo Molnar committed
988
989

	/* Get the faulting address: */
990
	address = read_cr2();
Linus Torvalds's avatar
Linus Torvalds committed
991

Vegard Nossum's avatar
Vegard Nossum committed
992
993
994
995
996
997
	/*
	 * 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);
998
	prefetchw(&mm->mmap_sem);
Vegard Nossum's avatar
Vegard Nossum committed
999

1000
	if (unlikely(kmmio_fault(regs, address)))
1001
		return;
Linus Torvalds's avatar
Linus Torvalds committed
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013

	/*
	 * 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
1014
	 * protection error (error_code & 9) == 0.
Linus Torvalds's avatar
Linus Torvalds committed
1015
	 */
1016
	if (unlikely(fault_in_kernel_space(address))) {
Vegard Nossum's avatar
Vegard Nossum committed
1017
1018
1019
1020
1021
1022
1023
		if (!(error_code & (PF_RSVD | PF_USER | PF_PROT))) {
			if (vmalloc_fault(address) >= 0)
				return;

			if (kmemcheck_fault(regs, address, error_code))
				return;
		}
1024

Ingo Molnar's avatar
Ingo Molnar committed
1025
		/* Can handle a stale RO->RW TLB: */
1026
		if (spurious_fault(error_code, address))
1027
1028
			return;

Ingo Molnar's avatar
Ingo Molnar committed
1029
		/* kprobes don't want to hook the spurious faults: */
1030
1031
		if (notify_page_fault(regs))
			return;
1032
1033
		/*
		 * Don't take the mm semaphore here. If we fixup a prefetch
Ingo Molnar's avatar
Ingo Molnar committed
1034
		 * fault we could otherwise deadlock:
1035
		 */
1036
		bad_area_nosemaphore(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
1037

1038
		return;
1039
1040
	}

Ingo Molnar's avatar
Ingo Molnar committed
1041
	/* kprobes don't want to hook the spurious faults: */
1042
	if (unlikely(notify_page_fault(regs)))
1043
		return;
1044
	/*
1045
1046
1047
1048
	 * 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
1049
	 * potential system fault or CPU buglet:
1050
	 */
1051
1052
1053
	if (user_mode_vm(regs)) {
		local_irq_enable();
		error_code |= PF_USER;
Ingo Molnar's avatar
Ingo Molnar committed
1054
1055
1056
1057
	} else {
		if (regs->flags & X86_EFLAGS_IF)
			local_irq_enable();
	}
1058

1059
	if (unlikely(error_code & PF_RSVD))
1060
		pgtable_bad(regs, error_code, address);
Linus Torvalds's avatar
Linus Torvalds committed
1061

1062
	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, 0, regs, address);
1063

Linus Torvalds's avatar
Linus Torvalds committed
1064
	/*
Ingo Molnar's avatar
Ingo Molnar committed
1065
1066
	 * 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
1067
	 */
1068
1069
1070
1071
	if (unlikely(in_atomic() || !mm)) {
		bad_area_nosemaphore(regs, error_code, address);
		return;
	}
Linus Torvalds's avatar
Linus Torvalds committed
1072

1073
1074
	/*
	 * When running in the kernel we expect faults to occur only to
Ingo Molnar's avatar
Ingo Molnar committed
1075
1076
1077
1078
1079
1080
1081
	 * 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
1082
1083
	 *
	 * As the vast majority of faults will be valid we will only perform
Ingo Molnar's avatar
Ingo Molnar committed
1084
1085
1086
1087
	 * 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
1088
	 */
1089
	if (unlikely(!down_read_trylock(&mm->mmap_sem))) {
1090
		if ((error_code & PF_USER) == 0 &&
1091
1092
1093
1094
		    !search_exception_tables(regs->ip)) {
			bad_area_nosemaphore(regs, error_code, address);
			return;
		}
1095
retry:
Linus Torvalds's avatar
Linus Torvalds committed
1096
		down_read(&mm->mmap_sem);
1097
1098
	} else {
		/*
Ingo Molnar's avatar
Ingo Molnar committed
1099
1100
1101
		 * The above down_read_trylock() might have succeeded in
		 * which case we'll have missed the might_sleep() from
		 * down_read():
1102
1103
		 */
		might_sleep();
Linus Torvalds's avatar
Linus Torvalds committed
1104
1105
1106
	}

	vma = find_vma(mm, address);
1107
1108
1109
1110
1111
	if (unlikely(!vma)) {
		bad_area(regs, error_code, address);
		return;
	}
	if (likely(vma->vm_start <= address))
Linus Torvalds's avatar
Linus Torvalds committed
1112
		goto good_area;
1113
1114
1115
1116
	if (unlikely(!(vma->vm_flags & VM_GROWSDOWN))) {
		bad_area(regs, error_code, address);
		return;
	}
1117
	if (error_code & PF_USER) {
1118
1119
1120
		/*
		 * Accessing the stack below %sp is always a bug.
		 * The large cushion allows instructions like enter
Ingo Molnar's avatar
Ingo Molnar committed
1121
		 * and pusha to work. ("enter $65535, $31" pushes
1122
		 * 32 pointers and then decrements %sp by 65535.)
1123
		 */
1124
1125
1126
1127
		if (unlikely(address + 65536 + 32 * sizeof(unsigned long) < regs->sp)) {
			bad_area(regs, error_code, address);
			return;
		}
Linus Torvalds's avatar
Linus Torvalds committed
1128
	}
1129
1130
1131
1132
1133
1134
1135
1136
1137
	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
1138
good_area:
1139
	if (unlikely(access_error(error_code, vma))) {
1140
1141
		bad_area_access_error(regs, error_code, address);
		return;
Linus Torvalds's avatar
Linus Torvalds committed
1142
1143
1144
1145
1146
	}

	/*
	 * 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
1147
	 * the fault:
Linus Torvalds's avatar
Linus Torvalds committed
1148
	 */
1149
	fault = handle_mm_fault(mm, vma, address, flags);
Ingo Molnar's avatar
Ingo Molnar committed
1150

1151
1152
1153
	if (unlikely(fault & (VM_FAULT_RETRY|VM_FAULT_ERROR))) {
		if (mm_fault_error(regs, error_code, address, fault))
			return;
1154
1155
	}

1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
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++;
			perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, 0,
				      regs, address);
		} else {
			tsk->min_flt++;
			perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, 0,
				      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;
		}
1177
	}
1178

1179
1180
	check_v8086_mode(regs, address, tsk);

Linus Torvalds's avatar
Linus Torvalds committed
1181
1182
	up_read(&mm->mmap_sem);
}