fault.c 25.9 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, ...		*/
Ingo Molnar's avatar
Ingo Molnar committed
13

14
15
#include <asm/traps.h>			/* dotraplinkage, ...		*/
#include <asm/pgalloc.h>		/* pgd_*(), ...			*/
Linus Torvalds's avatar
Linus Torvalds committed
16

17
/*
Ingo Molnar's avatar
Ingo Molnar committed
18
19
20
21
22
23
24
 * 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
25
 */
Ingo Molnar's avatar
Ingo Molnar committed
26
27
28
29
30
31
32
33
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,
};
34

35
/*
36
37
 * Returns 0 if mmiotrace is disabled, or if the fault is not
 * handled by mmiotrace:
38
 */
39
static inline int kmmio_fault(struct pt_regs *regs, unsigned long addr)
40
{
41
42
43
44
	if (unlikely(is_kmmio_active()))
		if (kmmio_handler(regs, addr) == 1)
			return -1;
	return 0;
45
46
}

47
static inline int notify_page_fault(struct pt_regs *regs)
48
{
49
50
51
	int ret = 0;

	/* kprobe_running() needs smp_processor_id() */
52
	if (kprobes_built_in() && !user_mode_vm(regs)) {
53
54
55
56
57
		preempt_disable();
		if (kprobe_running() && kprobe_fault_handler(regs, 14))
			ret = 1;
		preempt_enable();
	}
58

59
	return ret;
60
}
61

62
/*
Ingo Molnar's avatar
Ingo Molnar committed
63
64
65
66
67
68
 * Prefetch quirks:
 *
 * 32-bit mode:
 *
 *   Sometimes AMD Athlon/Opteron CPUs report invalid exceptions on prefetch.
 *   Check that here and ignore it.
69
 *
Ingo Molnar's avatar
Ingo Molnar committed
70
 * 64-bit mode:
71
 *
Ingo Molnar's avatar
Ingo Molnar committed
72
73
74
75
 *   Sometimes the CPU reports invalid exceptions on prefetch.
 *   Check that here and ignore it.
 *
 * Opcode checker based on code by Richard Brunner.
76
 */
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
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
124
125
static int
is_prefetch(struct pt_regs *regs, unsigned long error_code, unsigned long addr)
126
{
Ingo Molnar's avatar
Ingo Molnar committed
127
	unsigned char *max_instr;
128
	unsigned char *instr;
129
	int prefetch = 0;
Linus Torvalds's avatar
Linus Torvalds committed
130

Ingo Molnar's avatar
Ingo Molnar committed
131
132
133
134
	/*
	 * If it was a exec (instruction fetch) fault on NX page, then
	 * do not ignore the fault:
	 */
135
	if (error_code & PF_INSTR)
Linus Torvalds's avatar
Linus Torvalds committed
136
		return 0;
137

138
	instr = (void *)convert_ip_to_linear(current, regs);
139
	max_instr = instr + 15;
Linus Torvalds's avatar
Linus Torvalds committed
140

141
	if (user_mode(regs) && instr >= (unsigned char *)TASK_SIZE)
Linus Torvalds's avatar
Linus Torvalds committed
142
143
		return 0;

144
	while (instr < max_instr) {
Ingo Molnar's avatar
Ingo Molnar committed
145
		unsigned char opcode;
Linus Torvalds's avatar
Linus Torvalds committed
146

147
		if (probe_kernel_address(instr, opcode))
148
			break;
Linus Torvalds's avatar
Linus Torvalds committed
149
150
151

		instr++;

152
		if (!check_prefetch_opcode(regs, instr, opcode, &prefetch))
Linus Torvalds's avatar
Linus Torvalds committed
153
154
155
156
157
			break;
	}
	return prefetch;
}

Ingo Molnar's avatar
Ingo Molnar committed
158
159
160
static void
force_sig_info_fault(int si_signo, int si_code, unsigned long address,
		     struct task_struct *tsk)
161
162
163
{
	siginfo_t info;

Ingo Molnar's avatar
Ingo Molnar committed
164
165
166
167
168
	info.si_signo	= si_signo;
	info.si_errno	= 0;
	info.si_code	= si_code;
	info.si_addr	= (void __user *)address;

169
170
171
	force_sig_info(si_signo, &info, tsk);
}

172
173
174
175
176
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)
177
{
178
179
180
181
	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
182

183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
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
285
	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;

	if (!pmd_present(*pmd)) {
		set_pmd(pmd, *pmd_k);
		arch_flush_lazy_mmu_mode();
	} else {
		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;
286
}
Linus Torvalds's avatar
Linus Torvalds committed
287

288
static void dump_pagetable(unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
289
{
290
291
292
293
	__typeof__(pte_val(__pte(0))) page;

	page = read_cr3();
	page = ((__typeof__(page) *) __va(page))[address >> PGDIR_SHIFT];
Ingo Molnar's avatar
Ingo Molnar committed
294

295
296
297
298
299
300
#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
301
							& (PTRS_PER_PMD - 1)];
302
303
304
305
306
307
308
309
310
311
312
		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
313
	 * it's allocated already:
314
315
316
317
	 */
	if ((page >> PAGE_SHIFT) < max_low_pfn
	    && (page & _PAGE_PRESENT)
	    && !(page & _PAGE_PSE)) {
Ingo Molnar's avatar
Ingo Molnar committed
318

319
320
		page &= PAGE_MASK;
		page = ((__typeof__(page) *) __va(page))[(address >> PAGE_SHIFT)
Ingo Molnar's avatar
Ingo Molnar committed
321
							& (PTRS_PER_PTE - 1)];
322
323
324
325
		printk("*pte = %0*Lx ", sizeof(page)*2, (u64)page);
	}

	printk("\n");
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
451
}

#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
452
453
454
455
456
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;

457
	pgd = (pgd_t *)read_cr3();
Linus Torvalds's avatar
Linus Torvalds committed
458

459
	pgd = __va((unsigned long)pgd & PHYSICAL_PAGE_MASK);
Ingo Molnar's avatar
Ingo Molnar committed
460

Linus Torvalds's avatar
Linus Torvalds committed
461
	pgd += pgd_index(address);
Ingo Molnar's avatar
Ingo Molnar committed
462
463
464
	if (bad_address(pgd))
		goto bad;

465
	printk("PGD %lx ", pgd_val(*pgd));
Ingo Molnar's avatar
Ingo Molnar committed
466
467
468

	if (!pgd_present(*pgd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
469

470
	pud = pud_offset(pgd, address);
Ingo Molnar's avatar
Ingo Molnar committed
471
472
473
	if (bad_address(pud))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
474
	printk("PUD %lx ", pud_val(*pud));
475
	if (!pud_present(*pud) || pud_large(*pud))
Ingo Molnar's avatar
Ingo Molnar committed
476
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
477
478

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

Linus Torvalds's avatar
Linus Torvalds committed
482
	printk("PMD %lx ", pmd_val(*pmd));
Ingo Molnar's avatar
Ingo Molnar committed
483
484
	if (!pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
485
486

	pte = pte_offset_kernel(pmd, address);
Ingo Molnar's avatar
Ingo Molnar committed
487
488
489
	if (bad_address(pte))
		goto bad;

490
	printk("PTE %lx", pte_val(*pte));
Ingo Molnar's avatar
Ingo Molnar committed
491
out:
Linus Torvalds's avatar
Linus Torvalds committed
492
493
494
495
	printk("\n");
	return;
bad:
	printk("BAD\n");
496
497
}

498
#endif /* CONFIG_X86_64 */
Linus Torvalds's avatar
Linus Torvalds committed
499

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

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

522
	if ((address >> 32) != 0)
Linus Torvalds's avatar
Linus Torvalds committed
523
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
524

Linus Torvalds's avatar
Linus Torvalds committed
525
	address |= 0xffffffffUL << 32;
526
527
	if ((address >= (u64)_stext && address <= (u64)_etext) ||
	    (address >= MODULES_VADDR && address <= MODULES_END)) {
Ingo Molnar's avatar
Ingo Molnar committed
528
		if (!once) {
529
			printk(errata93_warning);
Ingo Molnar's avatar
Ingo Molnar committed
530
			once = 1;
Linus Torvalds's avatar
Linus Torvalds committed
531
		}
532
		regs->ip = address;
Linus Torvalds's avatar
Linus Torvalds committed
533
534
		return 1;
	}
535
#endif
Linus Torvalds's avatar
Linus Torvalds committed
536
	return 0;
537
}
Linus Torvalds's avatar
Linus Torvalds committed
538

539
/*
Ingo Molnar's avatar
Ingo Molnar committed
540
541
542
543
544
 * 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
545
546
547
548
549
 * 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
550
	if ((regs->cs == __USER32_CS || (regs->cs & (1<<2))) && (address >> 32))
551
552
553
554
555
		return 1;
#endif
	return 0;
}

556
557
558
559
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
560

561
	/*
Ingo Molnar's avatar
Ingo Molnar committed
562
	 * Pentium F0 0F C7 C8 bug workaround:
563
564
565
566
567
568
569
570
571
572
573
574
575
	 */
	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;
}

576
577
578
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
579
580
581
static void
show_fault_oops(struct pt_regs *regs, unsigned long error_code,
		unsigned long address)
582
{
583
584
585
586
	if (!oops_may_print())
		return;

	if (error_code & PF_INSTR) {
587
		unsigned int level;
Ingo Molnar's avatar
Ingo Molnar committed
588

589
590
		pte_t *pte = lookup_address(address, &level);

591
592
		if (pte && pte_present(*pte) && !pte_exec(*pte))
			printk(nx_warning, current_uid());
593
594
	}

595
	printk(KERN_ALERT "BUG: unable to handle kernel ");
596
	if (address < PAGE_SIZE)
597
		printk(KERN_CONT "NULL pointer dereference");
598
	else
599
		printk(KERN_CONT "paging request");
Ingo Molnar's avatar
Ingo Molnar committed
600

601
	printk(KERN_CONT " at %p\n", (void *) address);
602
	printk(KERN_ALERT "IP:");
603
	printk_address(regs->ip, 1);
Ingo Molnar's avatar
Ingo Molnar committed
604

605
606
607
	dump_pagetable(address);
}

Ingo Molnar's avatar
Ingo Molnar committed
608
609
610
static noinline void
pgtable_bad(struct pt_regs *regs, unsigned long error_code,
	    unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
611
{
Ingo Molnar's avatar
Ingo Molnar committed
612
613
614
615
616
617
618
	struct task_struct *tsk;
	unsigned long flags;
	int sig;

	flags = oops_begin();
	tsk = current;
	sig = SIGKILL;
619

Linus Torvalds's avatar
Linus Torvalds committed
620
	printk(KERN_ALERT "%s: Corrupted page table at address %lx\n",
621
	       tsk->comm, address);
Linus Torvalds's avatar
Linus Torvalds committed
622
	dump_pagetable(address);
Ingo Molnar's avatar
Ingo Molnar committed
623
624
625
626
627

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

628
	if (__die("Bad pagetable", regs, error_code))
629
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
630

631
	oops_end(flags, regs, sig);
Linus Torvalds's avatar
Linus Torvalds committed
632
633
}

Ingo Molnar's avatar
Ingo Molnar committed
634
635
636
static noinline void
no_context(struct pt_regs *regs, unsigned long error_code,
	   unsigned long address)
637
638
{
	struct task_struct *tsk = current;
639
	unsigned long *stackend;
640
641
642
	unsigned long flags;
	int sig;

Ingo Molnar's avatar
Ingo Molnar committed
643
	/* Are we prepared to handle this kernel fault? */
644
645
646
647
	if (fixup_exception(regs))
		return;

	/*
Ingo Molnar's avatar
Ingo Molnar committed
648
649
650
651
652
653
654
	 * 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:
655
	 *
Ingo Molnar's avatar
Ingo Molnar committed
656
	 *   Hall of shame of CPU/BIOS bugs.
657
658
659
660
661
662
663
664
665
	 */
	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
666
	 * terminate things with extreme prejudice:
667
668
669
670
671
	 */
	flags = oops_begin();

	show_fault_oops(regs, error_code, address);

Ingo Molnar's avatar
Ingo Molnar committed
672
	stackend = end_of_stack(tsk);
673
674
675
	if (*stackend != STACK_END_MAGIC)
		printk(KERN_ALERT "Thread overran stack, or stack corrupted\n");

676
677
678
	tsk->thread.cr2		= address;
	tsk->thread.trap_no	= 14;
	tsk->thread.error_code	= error_code;
679
680
681
682

	sig = SIGKILL;
	if (__die("Oops", regs, error_code))
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
683

684
685
	/* 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
686

687
688
689
	oops_end(flags, regs, sig);
}

Ingo Molnar's avatar
Ingo Molnar committed
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
/*
 * 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)
717
718
719
720
721
722
{
	struct task_struct *tsk = current;

	/* User mode accesses just cause a SIGSEGV */
	if (error_code & PF_USER) {
		/*
Ingo Molnar's avatar
Ingo Molnar committed
723
		 * It's possible to have interrupts off here:
724
725
726
727
728
		 */
		local_irq_enable();

		/*
		 * Valid to do another page fault here because this one came
Ingo Molnar's avatar
Ingo Molnar committed
729
		 * from user space:
730
731
732
733
734
735
736
		 */
		if (is_prefetch(regs, error_code, address))
			return;

		if (is_errata100(regs, address))
			return;

Ingo Molnar's avatar
Ingo Molnar committed
737
738
739
740
741
742
743
		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;
744
745

		force_sig_info_fault(SIGSEGV, si_code, address, tsk);
Ingo Molnar's avatar
Ingo Molnar committed
746

747
748
749
750
751
752
753
754
755
		return;
	}

	if (is_f00f_bug(regs, address))
		return;

	no_context(regs, error_code, address);
}

Ingo Molnar's avatar
Ingo Molnar committed
756
757
758
static noinline void
bad_area_nosemaphore(struct pt_regs *regs, unsigned long error_code,
		     unsigned long address)
759
760
761
762
{
	__bad_area_nosemaphore(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
763
764
765
static void
__bad_area(struct pt_regs *regs, unsigned long error_code,
	   unsigned long address, int si_code)
766
767
768
769
770
771
772
773
774
775
776
777
{
	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
778
779
static noinline void
bad_area(struct pt_regs *regs, unsigned long error_code, unsigned long address)
780
781
782
783
{
	__bad_area(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
784
785
786
static noinline void
bad_area_access_error(struct pt_regs *regs, unsigned long error_code,
		      unsigned long address)
787
788
789
790
791
{
	__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
792
793
794
static void
out_of_memory(struct pt_regs *regs, unsigned long error_code,
	      unsigned long address)
795
796
797
{
	/*
	 * We ran out of memory, call the OOM killer, and return the userspace
Ingo Molnar's avatar
Ingo Molnar committed
798
	 * (which will retry the fault, or kill us if we got oom-killed):
799
800
	 */
	up_read(&current->mm->mmap_sem);
Ingo Molnar's avatar
Ingo Molnar committed
801

802
803
804
	pagefault_out_of_memory();
}

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

	up_read(&mm->mmap_sem);

Ingo Molnar's avatar
Ingo Molnar committed
813
	/* Kernel mode? Handle exceptions or die: */
814
815
	if (!(error_code & PF_USER))
		no_context(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
816

817
	/* User-space => ok to do another page fault: */
818
819
	if (is_prefetch(regs, error_code, address))
		return;
Ingo Molnar's avatar
Ingo Molnar committed
820
821
822
823
824

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

825
826
827
	force_sig_info_fault(SIGBUS, BUS_ADRERR, address, tsk);
}

Ingo Molnar's avatar
Ingo Molnar committed
828
829
830
static noinline void
mm_fault_error(struct pt_regs *regs, unsigned long error_code,
	       unsigned long address, unsigned int fault)
831
{
Ingo Molnar's avatar
Ingo Molnar committed
832
	if (fault & VM_FAULT_OOM) {
833
		out_of_memory(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
834
835
836
837
838
839
	} else {
		if (fault & VM_FAULT_SIGBUS)
			do_sigbus(regs, error_code, address);
		else
			BUG();
	}
840
841
}

842
843
844
845
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
846

847
848
849
850
851
852
	if ((error_code & PF_INSTR) && !pte_exec(*pte))
		return 0;

	return 1;
}

853
/*
Ingo Molnar's avatar
Ingo Molnar committed
854
855
856
857
858
859
860
861
 * 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.
 *
862
863
864
 * There are no security implications to leaving a stale TLB when
 * increasing the permissions on a page.
 */
Ingo Molnar's avatar
Ingo Molnar committed
865
866
static noinline int
spurious_fault(unsigned long error_code, unsigned long address)
867
868
869
870
871
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;
872
	int ret;
873
874
875
876
877
878
879
880
881
882
883
884
885

	/* 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;

886
887
888
	if (pud_large(*pud))
		return spurious_fault_check(error_code, (pte_t *) pud);

889
890
891
892
	pmd = pmd_offset(pud, address);
	if (!pmd_present(*pmd))
		return 0;

893
894
895
	if (pmd_large(*pmd))
		return spurious_fault_check(error_code, (pte_t *) pmd);

896
897
898
899
	pte = pte_offset_kernel(pmd, address);
	if (!pte_present(*pte))
		return 0;

900
901
902
903
904
	ret = spurious_fault_check(error_code, pte);
	if (!ret)
		return 0;

	/*
Ingo Molnar's avatar
Ingo Molnar committed
905
906
	 * Make sure we have permissions in PMD.
	 * If not, then there's a bug in the page tables:
907
908
909
	 */
	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
910

911
	return ret;
912
913
}

914
int show_unhandled_signals = 1;
Linus Torvalds's avatar
Linus Torvalds committed
915

Ingo Molnar's avatar
Ingo Molnar committed
916
917
static inline int
access_error(unsigned long error_code, int write, struct vm_area_struct *vma)
918
919
{
	if (write) {
Ingo Molnar's avatar
Ingo Molnar committed
920
		/* write, present and write, not present: */
921
922
		if (unlikely(!(vma->vm_flags & VM_WRITE)))
			return 1;
Ingo Molnar's avatar
Ingo Molnar committed
923
		return 0;
924
925
	}

Ingo Molnar's avatar
Ingo Molnar committed
926
927
928
929
930
931
932
933
	/* 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;

934
935
936
	return 0;
}

937
938
static int fault_in_kernel_space(unsigned long address)
{
939
	return address >= TASK_SIZE_MAX;
940
941
}

Linus Torvalds's avatar
Linus Torvalds committed
942
943
944
945
946
/*
 * This routine handles page faults.  It determines the address,
 * and the problem, and then passes it off to one of the appropriate
 * routines.
 */
947
948
dotraplinkage void __kprobes
do_page_fault(struct pt_regs *regs, unsigned long error_code)
Linus Torvalds's avatar
Linus Torvalds committed
949
{
Ingo Molnar's avatar
Ingo Molnar committed
950
	struct vm_area_struct *vma;
Linus Torvalds's avatar
Linus Torvalds committed
951
	struct task_struct *tsk;
Ingo Molnar's avatar
Ingo Molnar committed
952
	unsigned long address;
Linus Torvalds's avatar
Linus Torvalds committed
953
	struct mm_struct *mm;
954
	int write;
955
	int fault;
Linus Torvalds's avatar
Linus Torvalds committed
956

957
958
	tsk = current;
	mm = tsk->mm;
Ingo Molnar's avatar
Ingo Molnar committed
959

960
961
	prefetchw(&mm->mmap_sem);

Ingo Molnar's avatar
Ingo Molnar committed
962
	/* Get the faulting address: */
963
	address = read_cr2();
Linus Torvalds's avatar
Linus Torvalds committed
964

965
	if (unlikely(kmmio_fault(regs, address)))
966
		return;
Linus Torvalds's avatar
Linus Torvalds committed
967
968
969
970
971
972
973
974
975
976
977
978

	/*
	 * 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
979
	 * protection error (error_code & 9) == 0.
Linus Torvalds's avatar
Linus Torvalds committed
980
	 */
981
	if (unlikely(fault_in_kernel_space(address))) {
982
983
984
		if (!(error_code & (PF_RSVD|PF_USER|PF_PROT)) &&
		    vmalloc_fault(address) >= 0)
			return;
985

Ingo Molnar's avatar
Ingo Molnar committed
986
		/* Can handle a stale RO->RW TLB: */
987
		if (spurious_fault(error_code, address))
988
989
			return;

Ingo Molnar's avatar
Ingo Molnar committed
990
		/* kprobes don't want to hook the spurious faults: */
991
992
		if (notify_page_fault(regs))
			return;
993
994
		/*
		 * Don't take the mm semaphore here. If we fixup a prefetch
Ingo Molnar's avatar
Ingo Molnar committed
995
		 * fault we could otherwise deadlock:
996
		 */
997
		bad_area_nosemaphore(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
998

999
		return;
1000
1001
	}

Ingo Molnar's avatar
Ingo Molnar committed
1002
	/* kprobes don't want to hook the spurious faults: */
1003
	if (unlikely(notify_page_fault(regs)))
1004
		return;
1005
	/*
1006
1007
1008
1009
	 * 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
1010
	 * potential system fault or CPU buglet:
1011
	 */
1012
1013
1014
	if (user_mode_vm(regs)) {
		local_irq_enable();
		error_code |= PF_USER;
Ingo Molnar's avatar
Ingo Molnar committed
1015
1016
1017
1018
	} else {
		if (regs->flags & X86_EFLAGS_IF)
			local_irq_enable();
	}
1019

1020
	if (unlikely(error_code & PF_RSVD))
1021
		pgtable_bad(regs, error_code, address);
Linus Torvalds's avatar
Linus Torvalds committed
1022
1023

	/*
Ingo Molnar's avatar
Ingo Molnar committed
1024
1025
	 * 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
1026
	 */
1027
1028
1029
1030
	if (unlikely(in_atomic() || !mm)) {
		bad_area_nosemaphore(regs, error_code, address);
		return;
	}
Linus Torvalds's avatar
Linus Torvalds committed
1031

1032
1033
	/*
	 * When running in the kernel we expect faults to occur only to
Ingo Molnar's avatar
Ingo Molnar committed
1034
1035
1036
1037
1038
1039
1040
	 * 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
1041
1042
	 *
	 * As the vast majority of faults will be valid we will only perform
Ingo Molnar's avatar
Ingo Molnar committed
1043
1044
1045
1046
	 * 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
1047
	 */
1048
	if (unlikely(!down_read_trylock(&mm->mmap_sem))) {
1049
		if ((error_code & PF_USER) == 0 &&
1050
1051
1052
1053
		    !search_exception_tables(regs->ip)) {
			bad_area_nosemaphore(regs, error_code, address);
			return;
		}
Linus Torvalds's avatar
Linus Torvalds committed
1054
		down_read(&mm->mmap_sem);
1055
1056
	} else {
		/*
Ingo Molnar's avatar
Ingo Molnar committed
1057
1058
1059
		 * The above down_read_trylock() might have succeeded in
		 * which case we'll have missed the might_sleep() from
		 * down_read():
1060
1061
		 */
		might_sleep();
Linus Torvalds's avatar
Linus Torvalds committed
1062
1063
1064
	}

	vma = find_vma(mm, address);
1065
1066
1067
1068
1069
	if (unlikely(!vma)) {
		bad_area(regs, error_code, address);
		return;
	}
	if (likely(vma->vm_start <= address))
Linus Torvalds's avatar
Linus Torvalds committed
1070
		goto good_area;
1071
1072
1073
1074
	if (unlikely(!(vma->vm_flags & VM_GROWSDOWN))) {
		bad_area(regs, error_code, address);
		return;
	}
1075
	if (error_code & PF_USER) {
1076
1077
1078
		/*
		 * Accessing the stack below %sp is always a bug.
		 * The large cushion allows instructions like enter
Ingo Molnar's avatar
Ingo Molnar committed
1079
		 * and pusha to work. ("enter $65535, $31" pushes
1080
		 * 32 pointers and then decrements %sp by 65535.)
1081
		 */
1082
1083
1084
1085
		if (unlikely(address + 65536 + 32 * sizeof(unsigned long) < regs->sp)) {
			bad_area(regs, error_code, address);
			return;
		}
Linus Torvalds's avatar
Linus Torvalds committed
1086
	}
1087
1088
1089
1090
1091
1092
1093
1094
1095
	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
1096
good_area:
1097
	write = error_code & PF_WRITE;
Ingo Molnar's avatar
Ingo Molnar committed
1098

1099
1100
1101
	if (unlikely(access_error(error_code, write, vma))) {
		bad_area_access_error(regs, error_code, address);
		return;
Linus Torvalds's avatar
Linus Torvalds committed
1102
1103
1104
1105
1106
	}

	/*
	 * 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
1107
	 * the fault:
Linus Torvalds's avatar
Linus Torvalds committed
1108
	 */
Nick Piggin's avatar
Nick Piggin committed
1109
	fault = handle_mm_fault(mm, vma, address, write);
Ingo Molnar's avatar
Ingo Molnar committed
1110

Nick Piggin's avatar
Nick Piggin committed
1111
	if (unlikely(fault & VM_FAULT_ERROR)) {
1112
1113
		mm_fault_error(regs, error_code, address, fault);
		return;
Linus Torvalds's avatar
Linus Torvalds committed
1114
	}
Ingo Molnar's avatar
Ingo Molnar committed
1115

Nick Piggin's avatar
Nick Piggin committed
1116
1117
1118
1119
	if (fault & VM_FAULT_MAJOR)
		tsk->maj_flt++;
	else
		tsk->min_flt++;
1120

1121
1122
	check_v8086_mode(regs, address, tsk);

Linus Torvalds's avatar
Linus Torvalds committed
1123
1124
	up_read(&mm->mmap_sem);
}