fault.c 26.5 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.
Linus Torvalds's avatar
Linus Torvalds committed
4
5
 */
#include <linux/interrupt.h>
Ingo Molnar's avatar
Ingo Molnar committed
6
7
#include <linux/mmiotrace.h>
#include <linux/bootmem.h>
Linus Torvalds's avatar
Linus Torvalds committed
8
#include <linux/compiler.h>
Harvey Harrison's avatar
Harvey Harrison committed
9
#include <linux/highmem.h>
10
#include <linux/kprobes.h>
11
#include <linux/uaccess.h>
Ingo Molnar's avatar
Ingo Molnar committed
12
13
14
15
16
17
18
#include <linux/vmalloc.h>
#include <linux/vt_kern.h>
#include <linux/signal.h>
#include <linux/kernel.h>
#include <linux/ptrace.h>
#include <linux/string.h>
#include <linux/module.h>
19
#include <linux/kdebug.h>
Ingo Molnar's avatar
Ingo Molnar committed
20
#include <linux/errno.h>
21
#include <linux/magic.h>
Ingo Molnar's avatar
Ingo Molnar committed
22
23
24
25
26
27
28
29
30
#include <linux/sched.h>
#include <linux/types.h>
#include <linux/init.h>
#include <linux/mman.h>
#include <linux/tty.h>
#include <linux/smp.h>
#include <linux/mm.h>

#include <asm-generic/sections.h>
Linus Torvalds's avatar
Linus Torvalds committed
31
32

#include <asm/tlbflush.h>
Ingo Molnar's avatar
Ingo Molnar committed
33
34
35
#include <asm/pgalloc.h>
#include <asm/segment.h>
#include <asm/system.h>
Linus Torvalds's avatar
Linus Torvalds committed
36
#include <asm/proto.h>
37
#include <asm/traps.h>
Ingo Molnar's avatar
Ingo Molnar committed
38
#include <asm/desc.h>
Linus Torvalds's avatar
Linus Torvalds committed
39

40
/*
Ingo Molnar's avatar
Ingo Molnar committed
41
42
43
44
45
46
47
 * 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
48
 */
Ingo Molnar's avatar
Ingo Molnar committed
49
50
51
52
53
54
55
56
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,
};
57

58
59
60
/*
 * (returns 0 if mmiotrace is disabled)
 */
61
static inline int kmmio_fault(struct pt_regs *regs, unsigned long addr)
62
{
63
64
65
66
	if (unlikely(is_kmmio_active()))
		if (kmmio_handler(regs, addr) == 1)
			return -1;
	return 0;
67
68
}

69
static inline int notify_page_fault(struct pt_regs *regs)
70
{
71
72
73
	int ret = 0;

	/* kprobe_running() needs smp_processor_id() */
74
	if (kprobes_built_in() && !user_mode_vm(regs)) {
75
76
77
78
79
		preempt_disable();
		if (kprobe_running() && kprobe_fault_handler(regs, 14))
			ret = 1;
		preempt_enable();
	}
80

81
	return ret;
82
}
83

84
/*
Ingo Molnar's avatar
Ingo Molnar committed
85
86
87
88
89
90
 * Prefetch quirks:
 *
 * 32-bit mode:
 *
 *   Sometimes AMD Athlon/Opteron CPUs report invalid exceptions on prefetch.
 *   Check that here and ignore it.
91
 *
Ingo Molnar's avatar
Ingo Molnar committed
92
 * 64-bit mode:
93
 *
Ingo Molnar's avatar
Ingo Molnar committed
94
95
96
97
 *   Sometimes the CPU reports invalid exceptions on prefetch.
 *   Check that here and ignore it.
 *
 * Opcode checker based on code by Richard Brunner.
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
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
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
146
147
static int
is_prefetch(struct pt_regs *regs, unsigned long error_code, unsigned long addr)
148
{
Ingo Molnar's avatar
Ingo Molnar committed
149
	unsigned char *max_instr;
150
	unsigned char *instr;
151
	int prefetch = 0;
Linus Torvalds's avatar
Linus Torvalds committed
152

Ingo Molnar's avatar
Ingo Molnar committed
153
154
155
156
	/*
	 * If it was a exec (instruction fetch) fault on NX page, then
	 * do not ignore the fault:
	 */
157
	if (error_code & PF_INSTR)
Linus Torvalds's avatar
Linus Torvalds committed
158
		return 0;
159

160
	instr = (void *)convert_ip_to_linear(current, regs);
161
	max_instr = instr + 15;
Linus Torvalds's avatar
Linus Torvalds committed
162

163
	if (user_mode(regs) && instr >= (unsigned char *)TASK_SIZE)
Linus Torvalds's avatar
Linus Torvalds committed
164
165
		return 0;

166
	while (instr < max_instr) {
Ingo Molnar's avatar
Ingo Molnar committed
167
		unsigned char opcode;
Linus Torvalds's avatar
Linus Torvalds committed
168

169
		if (probe_kernel_address(instr, opcode))
170
			break;
Linus Torvalds's avatar
Linus Torvalds committed
171
172
173

		instr++;

174
		if (!check_prefetch_opcode(regs, instr, opcode, &prefetch))
Linus Torvalds's avatar
Linus Torvalds committed
175
176
177
178
179
			break;
	}
	return prefetch;
}

Ingo Molnar's avatar
Ingo Molnar committed
180
181
182
static void
force_sig_info_fault(int si_signo, int si_code, unsigned long address,
		     struct task_struct *tsk)
183
184
185
{
	siginfo_t info;

Ingo Molnar's avatar
Ingo Molnar committed
186
187
188
189
190
	info.si_signo	= si_signo;
	info.si_errno	= 0;
	info.si_code	= si_code;
	info.si_addr	= (void __user *)address;

191
192
193
	force_sig_info(si_signo, &info, tsk);
}

194
195
196
197
198
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)
199
{
200
201
202
203
	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
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
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
	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;
308
}
Linus Torvalds's avatar
Linus Torvalds committed
309

310
static void dump_pagetable(unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
311
{
312
313
314
315
	__typeof__(pte_val(__pte(0))) page;

	page = read_cr3();
	page = ((__typeof__(page) *) __va(page))[address >> PGDIR_SHIFT];
Ingo Molnar's avatar
Ingo Molnar committed
316

317
318
319
320
321
322
#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
323
							& (PTRS_PER_PMD - 1)];
324
325
326
327
328
329
330
331
332
333
334
		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
335
	 * it's allocated already:
336
337
338
339
	 */
	if ((page >> PAGE_SHIFT) < max_low_pfn
	    && (page & _PAGE_PRESENT)
	    && !(page & _PAGE_PSE)) {
Ingo Molnar's avatar
Ingo Molnar committed
340

341
342
		page &= PAGE_MASK;
		page = ((__typeof__(page) *) __va(page))[(address >> PAGE_SHIFT)
Ingo Molnar's avatar
Ingo Molnar committed
343
							& (PTRS_PER_PTE - 1)];
344
345
346
347
		printk("*pte = %0*Lx ", sizeof(page)*2, (u64)page);
	}

	printk("\n");
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
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
}

#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
474
475
476
477
478
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;

479
	pgd = (pgd_t *)read_cr3();
Linus Torvalds's avatar
Linus Torvalds committed
480

481
	pgd = __va((unsigned long)pgd & PHYSICAL_PAGE_MASK);
Ingo Molnar's avatar
Ingo Molnar committed
482

Linus Torvalds's avatar
Linus Torvalds committed
483
	pgd += pgd_index(address);
Ingo Molnar's avatar
Ingo Molnar committed
484
485
486
	if (bad_address(pgd))
		goto bad;

487
	printk("PGD %lx ", pgd_val(*pgd));
Ingo Molnar's avatar
Ingo Molnar committed
488
489
490

	if (!pgd_present(*pgd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
491

492
	pud = pud_offset(pgd, address);
Ingo Molnar's avatar
Ingo Molnar committed
493
494
495
	if (bad_address(pud))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
496
	printk("PUD %lx ", pud_val(*pud));
497
	if (!pud_present(*pud) || pud_large(*pud))
Ingo Molnar's avatar
Ingo Molnar committed
498
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
499
500

	pmd = pmd_offset(pud, address);
Ingo Molnar's avatar
Ingo Molnar committed
501
502
503
	if (bad_address(pmd))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
504
	printk("PMD %lx ", pmd_val(*pmd));
Ingo Molnar's avatar
Ingo Molnar committed
505
506
	if (!pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
507
508

	pte = pte_offset_kernel(pmd, address);
Ingo Molnar's avatar
Ingo Molnar committed
509
510
511
	if (bad_address(pte))
		goto bad;

512
	printk("PTE %lx", pte_val(*pte));
Ingo Molnar's avatar
Ingo Molnar committed
513
out:
Linus Torvalds's avatar
Linus Torvalds committed
514
515
516
517
	printk("\n");
	return;
bad:
	printk("BAD\n");
518
519
}

520
#endif /* CONFIG_X86_64 */
Linus Torvalds's avatar
Linus Torvalds committed
521

Ingo Molnar's avatar
Ingo Molnar committed
522
523
524
525
526
527
528
529
530
531
532
533
534
/*
 * 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.
535
 */
536
static int is_errata93(struct pt_regs *regs, unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
537
{
538
#ifdef CONFIG_X86_64
Ingo Molnar's avatar
Ingo Molnar committed
539
540
	static int once;

541
	if (address != regs->ip)
Linus Torvalds's avatar
Linus Torvalds committed
542
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
543

544
	if ((address >> 32) != 0)
Linus Torvalds's avatar
Linus Torvalds committed
545
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
546

Linus Torvalds's avatar
Linus Torvalds committed
547
	address |= 0xffffffffUL << 32;
548
549
	if ((address >= (u64)_stext && address <= (u64)_etext) ||
	    (address >= MODULES_VADDR && address <= MODULES_END)) {
Ingo Molnar's avatar
Ingo Molnar committed
550
		if (!once) {
551
			printk(errata93_warning);
Ingo Molnar's avatar
Ingo Molnar committed
552
			once = 1;
Linus Torvalds's avatar
Linus Torvalds committed
553
		}
554
		regs->ip = address;
Linus Torvalds's avatar
Linus Torvalds committed
555
556
		return 1;
	}
557
#endif
Linus Torvalds's avatar
Linus Torvalds committed
558
	return 0;
559
}
Linus Torvalds's avatar
Linus Torvalds committed
560

561
/*
Ingo Molnar's avatar
Ingo Molnar committed
562
563
564
565
566
 * 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
567
568
569
570
571
 * 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
572
	if ((regs->cs == __USER32_CS || (regs->cs & (1<<2))) && (address >> 32))
573
574
575
576
577
		return 1;
#endif
	return 0;
}

578
579
580
581
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
582

583
	/*
Ingo Molnar's avatar
Ingo Molnar committed
584
	 * Pentium F0 0F C7 C8 bug workaround:
585
586
587
588
589
590
591
592
593
594
595
596
597
	 */
	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;
}

598
599
600
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
601
602
603
static void
show_fault_oops(struct pt_regs *regs, unsigned long error_code,
		unsigned long address)
604
{
605
606
607
608
	if (!oops_may_print())
		return;

	if (error_code & PF_INSTR) {
609
		unsigned int level;
Ingo Molnar's avatar
Ingo Molnar committed
610

611
612
		pte_t *pte = lookup_address(address, &level);

613
614
		if (pte && pte_present(*pte) && !pte_exec(*pte))
			printk(nx_warning, current_uid());
615
616
	}

617
	printk(KERN_ALERT "BUG: unable to handle kernel ");
618
	if (address < PAGE_SIZE)
619
		printk(KERN_CONT "NULL pointer dereference");
620
	else
621
		printk(KERN_CONT "paging request");
Ingo Molnar's avatar
Ingo Molnar committed
622

623
	printk(KERN_CONT " at %p\n", (void *) address);
624
	printk(KERN_ALERT "IP:");
625
	printk_address(regs->ip, 1);
Ingo Molnar's avatar
Ingo Molnar committed
626

627
628
629
	dump_pagetable(address);
}

Ingo Molnar's avatar
Ingo Molnar committed
630
631
632
static noinline void
pgtable_bad(struct pt_regs *regs, unsigned long error_code,
	    unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
633
{
Ingo Molnar's avatar
Ingo Molnar committed
634
635
636
637
638
639
640
	struct task_struct *tsk;
	unsigned long flags;
	int sig;

	flags = oops_begin();
	tsk = current;
	sig = SIGKILL;
641

Linus Torvalds's avatar
Linus Torvalds committed
642
	printk(KERN_ALERT "%s: Corrupted page table at address %lx\n",
643
	       tsk->comm, address);
Linus Torvalds's avatar
Linus Torvalds committed
644
	dump_pagetable(address);
Ingo Molnar's avatar
Ingo Molnar committed
645
646
647
648
649

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

650
	if (__die("Bad pagetable", regs, error_code))
651
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
652

653
	oops_end(flags, regs, sig);
Linus Torvalds's avatar
Linus Torvalds committed
654
655
}

Ingo Molnar's avatar
Ingo Molnar committed
656
657
658
static noinline void
no_context(struct pt_regs *regs, unsigned long error_code,
	   unsigned long address)
659
660
{
	struct task_struct *tsk = current;
661
662
	unsigned long *stackend;

663
664
665
666
667
#ifdef CONFIG_X86_64
	unsigned long flags;
	int sig;
#endif

Ingo Molnar's avatar
Ingo Molnar committed
668
	/* Are we prepared to handle this kernel fault? */
669
670
671
672
	if (fixup_exception(regs))
		return;

	/*
Ingo Molnar's avatar
Ingo Molnar committed
673
674
675
676
677
678
679
	 * 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:
680
	 *
Ingo Molnar's avatar
Ingo Molnar committed
681
	 *   Hall of shame of CPU/BIOS bugs.
682
683
684
685
686
687
688
689
690
	 */
	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
691
	 * terminate things with extreme prejudice:
692
693
694
695
696
697
698
699
700
	 */
#ifdef CONFIG_X86_32
	bust_spinlocks(1);
#else
	flags = oops_begin();
#endif

	show_fault_oops(regs, error_code, address);

Ingo Molnar's avatar
Ingo Molnar committed
701
	stackend = end_of_stack(tsk);
702
703
704
	if (*stackend != STACK_END_MAGIC)
		printk(KERN_ALERT "Thread overran stack, or stack corrupted\n");

705
706
707
708
709
710
711
712
713
714
715
716
	tsk->thread.cr2 = address;
	tsk->thread.trap_no = 14;
	tsk->thread.error_code = error_code;

#ifdef CONFIG_X86_32
	die("Oops", regs, error_code);
	bust_spinlocks(0);
	do_exit(SIGKILL);
#else
	sig = SIGKILL;
	if (__die("Oops", regs, error_code))
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
717

718
719
	/* 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
720

721
722
723
724
	oops_end(flags, regs, sig);
#endif
}

Ingo Molnar's avatar
Ingo Molnar committed
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
/*
 * 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)
752
753
754
755
756
757
{
	struct task_struct *tsk = current;

	/* User mode accesses just cause a SIGSEGV */
	if (error_code & PF_USER) {
		/*
Ingo Molnar's avatar
Ingo Molnar committed
758
		 * It's possible to have interrupts off here:
759
760
761
762
763
		 */
		local_irq_enable();

		/*
		 * Valid to do another page fault here because this one came
Ingo Molnar's avatar
Ingo Molnar committed
764
		 * from user space:
765
766
767
768
769
770
771
		 */
		if (is_prefetch(regs, error_code, address))
			return;

		if (is_errata100(regs, address))
			return;

Ingo Molnar's avatar
Ingo Molnar committed
772
773
774
775
776
777
778
		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;
779
780

		force_sig_info_fault(SIGSEGV, si_code, address, tsk);
Ingo Molnar's avatar
Ingo Molnar committed
781

782
783
784
785
786
787
788
789
790
		return;
	}

	if (is_f00f_bug(regs, address))
		return;

	no_context(regs, error_code, address);
}

Ingo Molnar's avatar
Ingo Molnar committed
791
792
793
static noinline void
bad_area_nosemaphore(struct pt_regs *regs, unsigned long error_code,
		     unsigned long address)
794
795
796
797
{
	__bad_area_nosemaphore(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
798
799
800
static void
__bad_area(struct pt_regs *regs, unsigned long error_code,
	   unsigned long address, int si_code)
801
802
803
804
805
806
807
808
809
810
811
812
{
	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
813
814
static noinline void
bad_area(struct pt_regs *regs, unsigned long error_code, unsigned long address)
815
816
817
818
{
	__bad_area(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
819
820
821
static noinline void
bad_area_access_error(struct pt_regs *regs, unsigned long error_code,
		      unsigned long address)
822
823
824
825
826
{
	__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
827
828
829
static void
out_of_memory(struct pt_regs *regs, unsigned long error_code,
	      unsigned long address)
830
831
832
{
	/*
	 * We ran out of memory, call the OOM killer, and return the userspace
Ingo Molnar's avatar
Ingo Molnar committed
833
	 * (which will retry the fault, or kill us if we got oom-killed):
834
835
	 */
	up_read(&current->mm->mmap_sem);
Ingo Molnar's avatar
Ingo Molnar committed
836

837
838
839
	pagefault_out_of_memory();
}

Ingo Molnar's avatar
Ingo Molnar committed
840
841
static void
do_sigbus(struct pt_regs *regs, unsigned long error_code, unsigned long address)
842
843
844
845
846
847
{
	struct task_struct *tsk = current;
	struct mm_struct *mm = tsk->mm;

	up_read(&mm->mmap_sem);

Ingo Molnar's avatar
Ingo Molnar committed
848
	/* Kernel mode? Handle exceptions or die: */
849
850
	if (!(error_code & PF_USER))
		no_context(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
851

852
#ifdef CONFIG_X86_32
Ingo Molnar's avatar
Ingo Molnar committed
853
	/* User space => ok to do another page fault: */
854
855
856
	if (is_prefetch(regs, error_code, address))
		return;
#endif
Ingo Molnar's avatar
Ingo Molnar committed
857
858
859
860
861

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

862
863
864
	force_sig_info_fault(SIGBUS, BUS_ADRERR, address, tsk);
}

Ingo Molnar's avatar
Ingo Molnar committed
865
866
867
static noinline void
mm_fault_error(struct pt_regs *regs, unsigned long error_code,
	       unsigned long address, unsigned int fault)
868
{
Ingo Molnar's avatar
Ingo Molnar committed
869
	if (fault & VM_FAULT_OOM) {
870
		out_of_memory(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
871
872
873
874
875
876
	} else {
		if (fault & VM_FAULT_SIGBUS)
			do_sigbus(regs, error_code, address);
		else
			BUG();
	}
877
878
}

879
880
881
882
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
883

884
885
886
887
888
889
	if ((error_code & PF_INSTR) && !pte_exec(*pte))
		return 0;

	return 1;
}

890
/*
Ingo Molnar's avatar
Ingo Molnar committed
891
892
893
894
895
896
897
898
 * 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.
 *
899
900
901
 * There are no security implications to leaving a stale TLB when
 * increasing the permissions on a page.
 */
Ingo Molnar's avatar
Ingo Molnar committed
902
903
static noinline int
spurious_fault(unsigned long error_code, unsigned long address)
904
905
906
907
908
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;
909
	int ret;
910
911
912
913
914
915
916
917
918
919
920
921
922

	/* 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;

923
924
925
	if (pud_large(*pud))
		return spurious_fault_check(error_code, (pte_t *) pud);

926
927
928
929
	pmd = pmd_offset(pud, address);
	if (!pmd_present(*pmd))
		return 0;

930
931
932
	if (pmd_large(*pmd))
		return spurious_fault_check(error_code, (pte_t *) pmd);

933
934
935
936
	pte = pte_offset_kernel(pmd, address);
	if (!pte_present(*pte))
		return 0;

937
938
939
940
941
	ret = spurious_fault_check(error_code, pte);
	if (!ret)
		return 0;

	/*
Ingo Molnar's avatar
Ingo Molnar committed
942
943
	 * Make sure we have permissions in PMD.
	 * If not, then there's a bug in the page tables:
944
945
946
	 */
	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
947

948
	return ret;
949
950
}

951
int show_unhandled_signals = 1;
Linus Torvalds's avatar
Linus Torvalds committed
952

Ingo Molnar's avatar
Ingo Molnar committed
953
954
static inline int
access_error(unsigned long error_code, int write, struct vm_area_struct *vma)
955
956
{
	if (write) {
Ingo Molnar's avatar
Ingo Molnar committed
957
		/* write, present and write, not present: */
958
959
		if (unlikely(!(vma->vm_flags & VM_WRITE)))
			return 1;
Ingo Molnar's avatar
Ingo Molnar committed
960
		return 0;
961
962
	}

Ingo Molnar's avatar
Ingo Molnar committed
963
964
965
966
967
968
969
970
	/* 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;

971
972
973
	return 0;
}

974
975
976
977
static int fault_in_kernel_space(unsigned long address)
{
#ifdef CONFIG_X86_32
	return address >= TASK_SIZE;
Ingo Molnar's avatar
Ingo Molnar committed
978
#else
979
	return address >= TASK_SIZE64;
Ingo Molnar's avatar
Ingo Molnar committed
980
#endif
981
982
}

Linus Torvalds's avatar
Linus Torvalds committed
983
984
985
986
987
/*
 * This routine handles page faults.  It determines the address,
 * and the problem, and then passes it off to one of the appropriate
 * routines.
 */
988
989
990
991
#ifdef CONFIG_X86_64
asmlinkage
#endif
void __kprobes do_page_fault(struct pt_regs *regs, unsigned long error_code)
Linus Torvalds's avatar
Linus Torvalds committed
992
{
Ingo Molnar's avatar
Ingo Molnar committed
993
	struct vm_area_struct *vma;
Linus Torvalds's avatar
Linus Torvalds committed
994
	struct task_struct *tsk;
Ingo Molnar's avatar
Ingo Molnar committed
995
	unsigned long address;
Linus Torvalds's avatar
Linus Torvalds committed
996
	struct mm_struct *mm;
997
	int write;
998
	int fault;
Linus Torvalds's avatar
Linus Torvalds committed
999

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

1003
1004
	prefetchw(&mm->mmap_sem);

Ingo Molnar's avatar
Ingo Molnar committed
1005
	/* Get the faulting address: */
1006
	address = read_cr2();
Linus Torvalds's avatar
Linus Torvalds committed
1007

1008
	if (unlikely(kmmio_fault(regs, address)))
1009
		return;
Linus Torvalds's avatar
Linus Torvalds committed
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021

	/*
	 * 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
1022
	 * protection error (error_code & 9) == 0.
Linus Torvalds's avatar
Linus Torvalds committed
1023
	 */
1024
	if (unlikely(fault_in_kernel_space(address))) {
1025
1026
1027
		if (!(error_code & (PF_RSVD|PF_USER|PF_PROT)) &&
		    vmalloc_fault(address) >= 0)
			return;
1028

Ingo Molnar's avatar
Ingo Molnar committed
1029
		/* Can handle a stale RO->RW TLB: */
1030
		if (spurious_fault(error_code, address))
1031
1032
			return;

Ingo Molnar's avatar
Ingo Molnar committed
1033
		/* kprobes don't want to hook the spurious faults: */
1034
1035
		if (notify_page_fault(regs))
			return;
1036
1037
		/*
		 * Don't take the mm semaphore here. If we fixup a prefetch
Ingo Molnar's avatar
Ingo Molnar committed
1038
		 * fault we could otherwise deadlock:
1039
		 */
1040
		bad_area_nosemaphore(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
1041

1042
		return;
1043
1044
	}

Ingo Molnar's avatar
Ingo Molnar committed
1045
	/* kprobes don't want to hook the spurious faults: */
1046
	if (unlikely(notify_page_fault(regs)))
1047
		return;
1048
	/*
1049
1050
1051
1052
	 * 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
1053
	 * potential system fault or CPU buglet:
1054
	 */
1055
1056
1057
	if (user_mode_vm(regs)) {
		local_irq_enable();
		error_code |= PF_USER;
Ingo Molnar's avatar
Ingo Molnar committed
1058
1059
1060
1061
	} else {
		if (regs->flags & X86_EFLAGS_IF)
			local_irq_enable();
	}
1062

1063
	if (unlikely(error_code & PF_RSVD))
1064
		pgtable_bad(regs, error_code, address);
Linus Torvalds's avatar
Linus Torvalds committed
1065
1066

	/*
Ingo Molnar's avatar
Ingo Molnar committed
1067
1068
	 * 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
1069
	 */
1070
1071
1072
1073
	if (unlikely(in_atomic() || !mm)) {
		bad_area_nosemaphore(regs, error_code, address);
		return;
	}
Linus Torvalds's avatar
Linus Torvalds committed
1074

1075
1076
	/*
	 * When running in the kernel we expect faults to occur only to
Ingo Molnar's avatar
Ingo Molnar committed
1077
1078
1079
1080
1081
1082
1083
	 * 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
1084
1085
	 *
	 * As the vast majority of faults will be valid we will only perform
Ingo Molnar's avatar
Ingo Molnar committed
1086
1087
1088
1089
	 * 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
1090
	 */
1091
	if (unlikely(!down_read_trylock(&mm->mmap_sem))) {
1092
		if ((error_code & PF_USER) == 0 &&
1093
1094
1095
1096
		    !search_exception_tables(regs->ip)) {
			bad_area_nosemaphore(regs, error_code, address);
			return;
		}
Linus Torvalds's avatar
Linus Torvalds committed
1097
		down_read(&mm->mmap_sem);
1098
1099
	} else {
		/*
Ingo Molnar's avatar
Ingo Molnar committed
1100
1101
1102
		 * The above down_read_trylock() might have succeeded in
		 * which case we'll have missed the might_sleep() from
		 * down_read():
1103
1104
		 */
		might_sleep();
Linus Torvalds's avatar
Linus Torvalds committed
1105
1106
1107
	}

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

1142
1143
1144
	if (unlikely(access_error(error_code, write, vma))) {
		bad_area_access_error(regs, error_code, address);
		return;
Linus Torvalds's avatar
Linus Torvalds committed
1145
1146
1147
1148
1149
	}

	/*
	 * 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
1150
	 * the fault:
Linus Torvalds's avatar
Linus Torvalds committed
1151
	 */
Nick Piggin's avatar
Nick Piggin committed
1152
	fault = handle_mm_fault(mm, vma, address, write);
Ingo Molnar's avatar
Ingo Molnar committed
1153

Nick Piggin's avatar
Nick Piggin committed
1154
	if (unlikely(fault & VM_FAULT_ERROR)) {
1155
1156
		mm_fault_error(regs, error_code, address, fault);
		return;
Linus Torvalds's avatar
Linus Torvalds committed
1157
	}
Ingo Molnar's avatar
Ingo Molnar committed
1158

Nick Piggin's avatar
Nick Piggin committed
1159
1160
1161
1162
	if (fault & VM_FAULT_MAJOR)
		tsk->maj_flt++;
	else
		tsk->min_flt++;
1163

1164
1165
	check_v8086_mode(regs, address, tsk);

Linus Torvalds's avatar
Linus Torvalds committed
1166
1167
	up_read(&mm->mmap_sem);
}