fault.c 26.1 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_counter.h>		/* perf_swcounter_event		*/
Ingo Molnar's avatar
Ingo Molnar committed
14

15
16
#include <asm/traps.h>			/* dotraplinkage, ...		*/
#include <asm/pgalloc.h>		/* pgd_*(), ...			*/
Linus Torvalds's avatar
Linus Torvalds committed
17

18
/*
Ingo Molnar's avatar
Ingo Molnar committed
19
20
21
22
23
24
25
 * 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
26
 */
Ingo Molnar's avatar
Ingo Molnar committed
27
28
29
30
31
32
33
34
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,
};
35

36
/*
37
38
 * Returns 0 if mmiotrace is disabled, or if the fault is not
 * handled by mmiotrace:
39
 */
40
static inline int kmmio_fault(struct pt_regs *regs, unsigned long addr)
41
{
42
43
44
45
	if (unlikely(is_kmmio_active()))
		if (kmmio_handler(regs, addr) == 1)
			return -1;
	return 0;
46
47
}

48
static inline int notify_page_fault(struct pt_regs *regs)
49
{
50
51
52
	int ret = 0;

	/* kprobe_running() needs smp_processor_id() */
53
	if (kprobes_built_in() && !user_mode_vm(regs)) {
54
55
56
57
58
		preempt_disable();
		if (kprobe_running() && kprobe_fault_handler(regs, 14))
			ret = 1;
		preempt_enable();
	}
59

60
	return ret;
61
}
62

63
/*
Ingo Molnar's avatar
Ingo Molnar committed
64
65
66
67
68
69
 * Prefetch quirks:
 *
 * 32-bit mode:
 *
 *   Sometimes AMD Athlon/Opteron CPUs report invalid exceptions on prefetch.
 *   Check that here and ignore it.
70
 *
Ingo Molnar's avatar
Ingo Molnar committed
71
 * 64-bit mode:
72
 *
Ingo Molnar's avatar
Ingo Molnar committed
73
74
75
76
 *   Sometimes the CPU reports invalid exceptions on prefetch.
 *   Check that here and ignore it.
 *
 * Opcode checker based on code by Richard Brunner.
77
 */
78
79
80
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
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
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.
		 */
		return (!user_mode(regs)) || (regs->cs == __USER_CS);
#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
125
126
static int
is_prefetch(struct pt_regs *regs, unsigned long error_code, unsigned long addr)
127
{
Ingo Molnar's avatar
Ingo Molnar committed
128
	unsigned char *max_instr;
129
	unsigned char *instr;
130
	int prefetch = 0;
Linus Torvalds's avatar
Linus Torvalds committed
131

Ingo Molnar's avatar
Ingo Molnar committed
132
133
134
135
	/*
	 * If it was a exec (instruction fetch) fault on NX page, then
	 * do not ignore the fault:
	 */
136
	if (error_code & PF_INSTR)
Linus Torvalds's avatar
Linus Torvalds committed
137
		return 0;
138

139
	instr = (void *)convert_ip_to_linear(current, regs);
140
	max_instr = instr + 15;
Linus Torvalds's avatar
Linus Torvalds committed
141

142
	if (user_mode(regs) && instr >= (unsigned char *)TASK_SIZE)
Linus Torvalds's avatar
Linus Torvalds committed
143
144
		return 0;

145
	while (instr < max_instr) {
Ingo Molnar's avatar
Ingo Molnar committed
146
		unsigned char opcode;
Linus Torvalds's avatar
Linus Torvalds committed
147

148
		if (probe_kernel_address(instr, opcode))
149
			break;
Linus Torvalds's avatar
Linus Torvalds committed
150
151
152

		instr++;

153
		if (!check_prefetch_opcode(regs, instr, opcode, &prefetch))
Linus Torvalds's avatar
Linus Torvalds committed
154
155
156
157
158
			break;
	}
	return prefetch;
}

Ingo Molnar's avatar
Ingo Molnar committed
159
160
161
static void
force_sig_info_fault(int si_signo, int si_code, unsigned long address,
		     struct task_struct *tsk)
162
163
164
{
	siginfo_t info;

Ingo Molnar's avatar
Ingo Molnar committed
165
166
167
168
169
	info.si_signo	= si_signo;
	info.si_errno	= 0;
	info.si_code	= si_code;
	info.si_addr	= (void __user *)address;

170
171
172
	force_sig_info(si_signo, &info, tsk);
}

173
174
175
176
177
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)
178
{
179
180
181
182
	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
183

184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
	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;

205
	if (!pmd_present(*pmd))
206
		set_pmd(pmd, *pmd_k);
207
	else
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
		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) {

		unsigned long flags;
		struct page *page;

		spin_lock_irqsave(&pgd_lock, flags);
		list_for_each_entry(page, &pgd_list, lru) {
			if (!vmalloc_sync_one(page_address(page), address))
				break;
		}
		spin_unlock_irqrestore(&pgd_lock, flags);
	}
}

/*
 * 32-bit:
 *
 *   Handle a fault on the vmalloc or module mapping area
 */
static noinline int vmalloc_fault(unsigned long address)
{
	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;

	/*
	 * 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;
285
}
Linus Torvalds's avatar
Linus Torvalds committed
286

287
static void dump_pagetable(unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
288
{
289
290
291
292
	__typeof__(pte_val(__pte(0))) page;

	page = read_cr3();
	page = ((__typeof__(page) *) __va(page))[address >> PGDIR_SHIFT];
Ingo Molnar's avatar
Ingo Molnar committed
293

294
295
296
297
298
299
#ifdef CONFIG_X86_PAE
	printk("*pdpt = %016Lx ", page);
	if ((page >> PAGE_SHIFT) < max_low_pfn
	    && page & _PAGE_PRESENT) {
		page &= PAGE_MASK;
		page = ((__typeof__(page) *) __va(page))[(address >> PMD_SHIFT)
Ingo Molnar's avatar
Ingo Molnar committed
300
							& (PTRS_PER_PMD - 1)];
301
302
303
304
305
306
307
308
309
310
311
		printk(KERN_CONT "*pde = %016Lx ", page);
		page &= ~_PAGE_NX;
	}
#else
	printk("*pde = %08lx ", page);
#endif

	/*
	 * 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
312
	 * it's allocated already:
313
314
315
316
	 */
	if ((page >> PAGE_SHIFT) < max_low_pfn
	    && (page & _PAGE_PRESENT)
	    && !(page & _PAGE_PSE)) {
Ingo Molnar's avatar
Ingo Molnar committed
317

318
319
		page &= PAGE_MASK;
		page = ((__typeof__(page) *) __va(page))[(address >> PAGE_SHIFT)
Ingo Molnar's avatar
Ingo Molnar committed
320
							& (PTRS_PER_PTE - 1)];
321
322
323
324
		printk("*pte = %0*Lx ", sizeof(page)*2, (u64)page);
	}

	printk("\n");
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
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
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
}

#else /* CONFIG_X86_64: */

void vmalloc_sync_all(void)
{
	unsigned long address;

	for (address = VMALLOC_START & PGDIR_MASK; address <= VMALLOC_END;
	     address += PGDIR_SIZE) {

		const pgd_t *pgd_ref = pgd_offset_k(address);
		unsigned long flags;
		struct page *page;

		if (pgd_none(*pgd_ref))
			continue;

		spin_lock_irqsave(&pgd_lock, flags);
		list_for_each_entry(page, &pgd_list, lru) {
			pgd_t *pgd;
			pgd = (pgd_t *)page_address(page) + pgd_index(address);
			if (pgd_none(*pgd))
				set_pgd(pgd, *pgd_ref);
			else
				BUG_ON(pgd_page_vaddr(*pgd) != pgd_page_vaddr(*pgd_ref));
		}
		spin_unlock_irqrestore(&pgd_lock, flags);
	}
}

/*
 * 64-bit:
 *
 *   Handle a fault on the vmalloc area
 *
 * This assumes no large pages in there.
 */
static noinline int vmalloc_fault(unsigned long address)
{
	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;

	/*
	 * 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[] =
KERN_ERR "******* Your BIOS seems to not contain a fix for K8 errata #93\n"
KERN_ERR "******* Working around it, but it may cause SEGVs or burn power.\n"
KERN_ERR "******* Please consider a BIOS update.\n"
KERN_ERR "******* Disabling USB legacy in the BIOS may also help.\n";

/*
 * 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)
{
Linus Torvalds's avatar
Linus Torvalds committed
451
452
453
454
455
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;

456
	pgd = (pgd_t *)read_cr3();
Linus Torvalds's avatar
Linus Torvalds committed
457

458
	pgd = __va((unsigned long)pgd & PHYSICAL_PAGE_MASK);
Ingo Molnar's avatar
Ingo Molnar committed
459

Linus Torvalds's avatar
Linus Torvalds committed
460
	pgd += pgd_index(address);
Ingo Molnar's avatar
Ingo Molnar committed
461
462
463
	if (bad_address(pgd))
		goto bad;

464
	printk("PGD %lx ", pgd_val(*pgd));
Ingo Molnar's avatar
Ingo Molnar committed
465
466
467

	if (!pgd_present(*pgd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
468

469
	pud = pud_offset(pgd, address);
Ingo Molnar's avatar
Ingo Molnar committed
470
471
472
	if (bad_address(pud))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
473
	printk("PUD %lx ", pud_val(*pud));
474
	if (!pud_present(*pud) || pud_large(*pud))
Ingo Molnar's avatar
Ingo Molnar committed
475
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
476
477

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

Linus Torvalds's avatar
Linus Torvalds committed
481
	printk("PMD %lx ", pmd_val(*pmd));
Ingo Molnar's avatar
Ingo Molnar committed
482
483
	if (!pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
484
485

	pte = pte_offset_kernel(pmd, address);
Ingo Molnar's avatar
Ingo Molnar committed
486
487
488
	if (bad_address(pte))
		goto bad;

489
	printk("PTE %lx", pte_val(*pte));
Ingo Molnar's avatar
Ingo Molnar committed
490
out:
Linus Torvalds's avatar
Linus Torvalds committed
491
492
493
494
	printk("\n");
	return;
bad:
	printk("BAD\n");
495
496
}

497
#endif /* CONFIG_X86_64 */
Linus Torvalds's avatar
Linus Torvalds committed
498

Ingo Molnar's avatar
Ingo Molnar committed
499
500
501
502
503
504
505
506
507
508
509
510
511
/*
 * 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.
512
 */
513
static int is_errata93(struct pt_regs *regs, unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
514
{
515
#ifdef CONFIG_X86_64
516
	if (address != regs->ip)
Linus Torvalds's avatar
Linus Torvalds committed
517
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
518

519
	if ((address >> 32) != 0)
Linus Torvalds's avatar
Linus Torvalds committed
520
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
521

Linus Torvalds's avatar
Linus Torvalds committed
522
	address |= 0xffffffffUL << 32;
523
524
	if ((address >= (u64)_stext && address <= (u64)_etext) ||
	    (address >= MODULES_VADDR && address <= MODULES_END)) {
525
		printk_once(errata93_warning);
526
		regs->ip = address;
Linus Torvalds's avatar
Linus Torvalds committed
527
528
		return 1;
	}
529
#endif
Linus Torvalds's avatar
Linus Torvalds committed
530
	return 0;
531
}
Linus Torvalds's avatar
Linus Torvalds committed
532

533
/*
Ingo Molnar's avatar
Ingo Molnar committed
534
535
536
537
538
 * 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
539
540
541
542
543
 * 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
544
	if ((regs->cs == __USER32_CS || (regs->cs & (1<<2))) && (address >> 32))
545
546
547
548
549
		return 1;
#endif
	return 0;
}

550
551
552
553
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
554

555
	/*
Ingo Molnar's avatar
Ingo Molnar committed
556
	 * Pentium F0 0F C7 C8 bug workaround:
557
558
559
560
561
562
563
564
565
566
567
568
569
	 */
	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;
}

570
571
572
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
573
574
575
static void
show_fault_oops(struct pt_regs *regs, unsigned long error_code,
		unsigned long address)
576
{
577
578
579
580
	if (!oops_may_print())
		return;

	if (error_code & PF_INSTR) {
581
		unsigned int level;
Ingo Molnar's avatar
Ingo Molnar committed
582

583
584
		pte_t *pte = lookup_address(address, &level);

585
586
		if (pte && pte_present(*pte) && !pte_exec(*pte))
			printk(nx_warning, current_uid());
587
588
	}

589
	printk(KERN_ALERT "BUG: unable to handle kernel ");
590
	if (address < PAGE_SIZE)
591
		printk(KERN_CONT "NULL pointer dereference");
592
	else
593
		printk(KERN_CONT "paging request");
Ingo Molnar's avatar
Ingo Molnar committed
594

595
	printk(KERN_CONT " at %p\n", (void *) address);
596
	printk(KERN_ALERT "IP:");
597
	printk_address(regs->ip, 1);
Ingo Molnar's avatar
Ingo Molnar committed
598

599
600
601
	dump_pagetable(address);
}

Ingo Molnar's avatar
Ingo Molnar committed
602
603
604
static noinline void
pgtable_bad(struct pt_regs *regs, unsigned long error_code,
	    unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
605
{
Ingo Molnar's avatar
Ingo Molnar committed
606
607
608
609
610
611
612
	struct task_struct *tsk;
	unsigned long flags;
	int sig;

	flags = oops_begin();
	tsk = current;
	sig = SIGKILL;
613

Linus Torvalds's avatar
Linus Torvalds committed
614
	printk(KERN_ALERT "%s: Corrupted page table at address %lx\n",
615
	       tsk->comm, address);
Linus Torvalds's avatar
Linus Torvalds committed
616
	dump_pagetable(address);
Ingo Molnar's avatar
Ingo Molnar committed
617
618
619
620
621

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

622
	if (__die("Bad pagetable", regs, error_code))
623
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
624

625
	oops_end(flags, regs, sig);
Linus Torvalds's avatar
Linus Torvalds committed
626
627
}

Ingo Molnar's avatar
Ingo Molnar committed
628
629
630
static noinline void
no_context(struct pt_regs *regs, unsigned long error_code,
	   unsigned long address)
631
632
{
	struct task_struct *tsk = current;
633
	unsigned long *stackend;
634
635
636
	unsigned long flags;
	int sig;

Ingo Molnar's avatar
Ingo Molnar committed
637
	/* Are we prepared to handle this kernel fault? */
638
639
640
641
	if (fixup_exception(regs))
		return;

	/*
Ingo Molnar's avatar
Ingo Molnar committed
642
643
644
645
646
647
648
	 * 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:
649
	 *
Ingo Molnar's avatar
Ingo Molnar committed
650
	 *   Hall of shame of CPU/BIOS bugs.
651
652
653
654
655
656
657
658
659
	 */
	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
660
	 * terminate things with extreme prejudice:
661
662
663
664
665
	 */
	flags = oops_begin();

	show_fault_oops(regs, error_code, address);

Ingo Molnar's avatar
Ingo Molnar committed
666
	stackend = end_of_stack(tsk);
667
668
669
	if (*stackend != STACK_END_MAGIC)
		printk(KERN_ALERT "Thread overran stack, or stack corrupted\n");

670
671
672
	tsk->thread.cr2		= address;
	tsk->thread.trap_no	= 14;
	tsk->thread.error_code	= error_code;
673
674
675
676

	sig = SIGKILL;
	if (__die("Oops", regs, error_code))
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
677

678
679
	/* 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
680

681
682
683
	oops_end(flags, regs, sig);
}

Ingo Molnar's avatar
Ingo Molnar committed
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
/*
 * 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;

	printk(KERN_CONT "%s%s[%d]: segfault at %lx ip %p sp %p error %lx",
		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)
711
712
713
714
715
716
{
	struct task_struct *tsk = current;

	/* User mode accesses just cause a SIGSEGV */
	if (error_code & PF_USER) {
		/*
Ingo Molnar's avatar
Ingo Molnar committed
717
		 * It's possible to have interrupts off here:
718
719
720
721
722
		 */
		local_irq_enable();

		/*
		 * Valid to do another page fault here because this one came
Ingo Molnar's avatar
Ingo Molnar committed
723
		 * from user space:
724
725
726
727
728
729
730
		 */
		if (is_prefetch(regs, error_code, address))
			return;

		if (is_errata100(regs, address))
			return;

Ingo Molnar's avatar
Ingo Molnar committed
731
732
733
734
735
736
737
		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;
738
739

		force_sig_info_fault(SIGSEGV, si_code, address, tsk);
Ingo Molnar's avatar
Ingo Molnar committed
740

741
742
743
744
745
746
747
748
749
		return;
	}

	if (is_f00f_bug(regs, address))
		return;

	no_context(regs, error_code, address);
}

Ingo Molnar's avatar
Ingo Molnar committed
750
751
752
static noinline void
bad_area_nosemaphore(struct pt_regs *regs, unsigned long error_code,
		     unsigned long address)
753
754
755
756
{
	__bad_area_nosemaphore(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
757
758
759
static void
__bad_area(struct pt_regs *regs, unsigned long error_code,
	   unsigned long address, int si_code)
760
761
762
763
764
765
766
767
768
769
770
771
{
	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
772
773
static noinline void
bad_area(struct pt_regs *regs, unsigned long error_code, unsigned long address)
774
775
776
777
{
	__bad_area(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
778
779
780
static noinline void
bad_area_access_error(struct pt_regs *regs, unsigned long error_code,
		      unsigned long address)
781
782
783
784
785
{
	__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
786
787
788
static void
out_of_memory(struct pt_regs *regs, unsigned long error_code,
	      unsigned long address)
789
790
791
{
	/*
	 * We ran out of memory, call the OOM killer, and return the userspace
Ingo Molnar's avatar
Ingo Molnar committed
792
	 * (which will retry the fault, or kill us if we got oom-killed):
793
794
	 */
	up_read(&current->mm->mmap_sem);
Ingo Molnar's avatar
Ingo Molnar committed
795

796
797
798
	pagefault_out_of_memory();
}

Ingo Molnar's avatar
Ingo Molnar committed
799
800
static void
do_sigbus(struct pt_regs *regs, unsigned long error_code, unsigned long address)
801
802
803
804
805
806
{
	struct task_struct *tsk = current;
	struct mm_struct *mm = tsk->mm;

	up_read(&mm->mmap_sem);

Ingo Molnar's avatar
Ingo Molnar committed
807
	/* Kernel mode? Handle exceptions or die: */
808
809
	if (!(error_code & PF_USER))
		no_context(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
810

811
	/* User-space => ok to do another page fault: */
812
813
	if (is_prefetch(regs, error_code, address))
		return;
Ingo Molnar's avatar
Ingo Molnar committed
814
815
816
817
818

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

819
820
821
	force_sig_info_fault(SIGBUS, BUS_ADRERR, address, tsk);
}

Ingo Molnar's avatar
Ingo Molnar committed
822
823
824
static noinline void
mm_fault_error(struct pt_regs *regs, unsigned long error_code,
	       unsigned long address, unsigned int fault)
825
{
Ingo Molnar's avatar
Ingo Molnar committed
826
	if (fault & VM_FAULT_OOM) {
827
		out_of_memory(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
828
829
830
831
832
833
	} else {
		if (fault & VM_FAULT_SIGBUS)
			do_sigbus(regs, error_code, address);
		else
			BUG();
	}
834
835
}

836
837
838
839
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
840

841
842
843
844
845
846
	if ((error_code & PF_INSTR) && !pte_exec(*pte))
		return 0;

	return 1;
}

847
/*
Ingo Molnar's avatar
Ingo Molnar committed
848
849
850
851
852
853
854
855
 * 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.
 *
856
857
858
 * There are no security implications to leaving a stale TLB when
 * increasing the permissions on a page.
 */
Ingo Molnar's avatar
Ingo Molnar committed
859
860
static noinline int
spurious_fault(unsigned long error_code, unsigned long address)
861
862
863
864
865
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;
866
	int ret;
867
868
869
870
871
872
873
874
875
876
877
878
879

	/* 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;

880
881
882
	if (pud_large(*pud))
		return spurious_fault_check(error_code, (pte_t *) pud);

883
884
885
886
	pmd = pmd_offset(pud, address);
	if (!pmd_present(*pmd))
		return 0;

887
888
889
	if (pmd_large(*pmd))
		return spurious_fault_check(error_code, (pte_t *) pmd);

890
891
892
893
	pte = pte_offset_kernel(pmd, address);
	if (!pte_present(*pte))
		return 0;

894
895
896
897
898
	ret = spurious_fault_check(error_code, pte);
	if (!ret)
		return 0;

	/*
Ingo Molnar's avatar
Ingo Molnar committed
899
900
	 * Make sure we have permissions in PMD.
	 * If not, then there's a bug in the page tables:
901
902
903
	 */
	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
904

905
	return ret;
906
907
}

908
int show_unhandled_signals = 1;
Linus Torvalds's avatar
Linus Torvalds committed
909

Ingo Molnar's avatar
Ingo Molnar committed
910
911
static inline int
access_error(unsigned long error_code, int write, struct vm_area_struct *vma)
912
913
{
	if (write) {
Ingo Molnar's avatar
Ingo Molnar committed
914
		/* write, present and write, not present: */
915
916
		if (unlikely(!(vma->vm_flags & VM_WRITE)))
			return 1;
Ingo Molnar's avatar
Ingo Molnar committed
917
		return 0;
918
919
	}

Ingo Molnar's avatar
Ingo Molnar committed
920
921
922
923
924
925
926
927
	/* 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;

928
929
930
	return 0;
}

931
932
static int fault_in_kernel_space(unsigned long address)
{
933
	return address >= TASK_SIZE_MAX;
934
935
}

Linus Torvalds's avatar
Linus Torvalds committed
936
937
938
939
940
/*
 * This routine handles page faults.  It determines the address,
 * and the problem, and then passes it off to one of the appropriate
 * routines.
 */
941
942
dotraplinkage void __kprobes
do_page_fault(struct pt_regs *regs, unsigned long error_code)
Linus Torvalds's avatar
Linus Torvalds committed
943
{
Ingo Molnar's avatar
Ingo Molnar committed
944
	struct vm_area_struct *vma;
Linus Torvalds's avatar
Linus Torvalds committed
945
	struct task_struct *tsk;
Ingo Molnar's avatar
Ingo Molnar committed
946
	unsigned long address;
Linus Torvalds's avatar
Linus Torvalds committed
947
	struct mm_struct *mm;
948
	int write;
949
	int fault;
Linus Torvalds's avatar
Linus Torvalds committed
950

951
952
	tsk = current;
	mm = tsk->mm;
Ingo Molnar's avatar
Ingo Molnar committed
953

954
955
	prefetchw(&mm->mmap_sem);

Ingo Molnar's avatar
Ingo Molnar committed
956
	/* Get the faulting address: */
957
	address = read_cr2();
Linus Torvalds's avatar
Linus Torvalds committed
958

959
	if (unlikely(kmmio_fault(regs, address)))
960
		return;
Linus Torvalds's avatar
Linus Torvalds committed
961
962
963
964
965
966
967
968
969
970
971
972

	/*
	 * 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
973
	 * protection error (error_code & 9) == 0.
Linus Torvalds's avatar
Linus Torvalds committed
974
	 */
975
	if (unlikely(fault_in_kernel_space(address))) {
976
977
978
		if (!(error_code & (PF_RSVD|PF_USER|PF_PROT)) &&
		    vmalloc_fault(address) >= 0)
			return;
979

Ingo Molnar's avatar
Ingo Molnar committed
980
		/* Can handle a stale RO->RW TLB: */
981
		if (spurious_fault(error_code, address))
982
983
			return;

Ingo Molnar's avatar
Ingo Molnar committed
984
		/* kprobes don't want to hook the spurious faults: */
985
986
		if (notify_page_fault(regs))
			return;
987
988
		/*
		 * Don't take the mm semaphore here. If we fixup a prefetch
Ingo Molnar's avatar
Ingo Molnar committed
989
		 * fault we could otherwise deadlock:
990
		 */
991
		bad_area_nosemaphore(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
992

993
		return;
994
995
	}

Ingo Molnar's avatar
Ingo Molnar committed
996
	/* kprobes don't want to hook the spurious faults: */
997
	if (unlikely(notify_page_fault(regs)))
998
		return;
999
	/*
1000
1001
1002
1003
	 * 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
1004
	 * potential system fault or CPU buglet:
1005
	 */
1006
1007
1008
	if (user_mode_vm(regs)) {
		local_irq_enable();
		error_code |= PF_USER;
Ingo Molnar's avatar
Ingo Molnar committed
1009
1010
1011
1012
	} else {
		if (regs->flags & X86_EFLAGS_IF)
			local_irq_enable();
	}
1013

1014
	if (unlikely(error_code & PF_RSVD))
1015
		pgtable_bad(regs, error_code, address);
Linus Torvalds's avatar
Linus Torvalds committed
1016

1017
	perf_swcounter_event(PERF_COUNT_SW_PAGE_FAULTS, 1, 0, regs, address);
1018

Linus Torvalds's avatar
Linus Torvalds committed
1019
	/*
Ingo Molnar's avatar
Ingo Molnar committed
1020
1021
	 * 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
1022
	 */
1023
1024
1025
1026
	if (unlikely(in_atomic() || !mm)) {
		bad_area_nosemaphore(regs, error_code, address);
		return;
	}
Linus Torvalds's avatar
Linus Torvalds committed
1027

1028
1029
	/*
	 * When running in the kernel we expect faults to occur only to
Ingo Molnar's avatar
Ingo Molnar committed
1030
1031
1032
1033
1034
1035
1036
	 * 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
1037
1038
	 *
	 * As the vast majority of faults will be valid we will only perform
Ingo Molnar's avatar
Ingo Molnar committed
1039
1040
1041
1042
	 * 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
1043
	 */
1044
	if (unlikely(!down_read_trylock(&mm->mmap_sem))) {
1045
		if ((error_code & PF_USER) == 0 &&
1046
1047
1048
1049
		    !search_exception_tables(regs->ip)) {
			bad_area_nosemaphore(regs, error_code, address);
			return;
		}
Linus Torvalds's avatar
Linus Torvalds committed
1050
		down_read(&mm->mmap_sem);
1051
1052
	} else {
		/*
Ingo Molnar's avatar
Ingo Molnar committed
1053
1054
1055
		 * The above down_read_trylock() might have succeeded in
		 * which case we'll have missed the might_sleep() from
		 * down_read():
1056
1057
		 */
		might_sleep();
Linus Torvalds's avatar
Linus Torvalds committed
1058
1059
1060
	}

	vma = find_vma(mm, address);
1061
1062
1063
1064
1065
	if (unlikely(!vma)) {
		bad_area(regs, error_code, address);
		return;
	}
	if (likely(vma->vm_start <= address))
Linus Torvalds's avatar
Linus Torvalds committed
1066
		goto good_area;
1067
1068
1069
1070
	if (unlikely(!(vma->vm_flags & VM_GROWSDOWN))) {
		bad_area(regs, error_code, address);
		return;
	}
1071
	if (error_code & PF_USER) {
1072
1073
1074
		/*
		 * Accessing the stack below %sp is always a bug.
		 * The large cushion allows instructions like enter
Ingo Molnar's avatar
Ingo Molnar committed
1075
		 * and pusha to work. ("enter $65535, $31" pushes
1076
		 * 32 pointers and then decrements %sp by 65535.)
1077
		 */
1078
1079
1080
1081
		if (unlikely(address + 65536 + 32 * sizeof(unsigned long) < regs->sp)) {
			bad_area(regs, error_code, address);
			return;
		}
Linus Torvalds's avatar
Linus Torvalds committed
1082
	}
1083
1084
1085
1086
1087
1088
1089
1090
1091
	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
1092
good_area:
1093
	write = error_code & PF_WRITE;
Ingo Molnar's avatar
Ingo Molnar committed
1094

1095
1096
1097
	if (unlikely(access_error(error_code, write, vma))) {
		bad_area_access_error(regs, error_code, address);
		return;
Linus Torvalds's avatar
Linus Torvalds committed
1098
1099
1100
1101
1102
	}

	/*
	 * 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
1103
	 * the fault:
Linus Torvalds's avatar
Linus Torvalds committed
1104
	 */
Nick Piggin's avatar
Nick Piggin committed
1105
	fault = handle_mm_fault(mm, vma, address, write);
Ingo Molnar's avatar
Ingo Molnar committed
1106

Nick Piggin's avatar
Nick Piggin committed
1107
	if (unlikely(fault & VM_FAULT_ERROR)) {
1108
1109
		mm_fault_error(regs, error_code, address, fault);
		return;
Linus Torvalds's avatar
Linus Torvalds committed
1110
	}
Ingo Molnar's avatar
Ingo Molnar committed
1111

1112
	if (fault & VM_FAULT_MAJOR) {
Nick Piggin's avatar
Nick Piggin committed
1113
		tsk->maj_flt++;
1114
		perf_swcounter_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, 0,
1115
				     regs, address);
1116
	} else {
Nick Piggin's avatar
Nick Piggin committed
1117
		tsk->min_flt++;
1118
		perf_swcounter_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, 0,
1119
				     regs, address);
1120
	}
1121

1122
1123
	check_v8086_mode(regs, address, tsk);

Linus Torvalds's avatar
Linus Torvalds committed
1124
1125
	up_read(&mm->mmap_sem);
}