fault.c 31.5 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.
4
 *  Copyright (C) 2008-2009, Red Hat Inc., Ingo Molnar
Linus Torvalds's avatar
Linus Torvalds committed
5
 */
6
7
8
9
10
11
12
#include <linux/magic.h>		/* STACK_END_MAGIC		*/
#include <linux/sched.h>		/* test_thread_flag(), ...	*/
#include <linux/kdebug.h>		/* oops_begin/end, ...		*/
#include <linux/module.h>		/* search_exception_table	*/
#include <linux/bootmem.h>		/* max_low_pfn			*/
#include <linux/kprobes.h>		/* __kprobes, ...		*/
#include <linux/mmiotrace.h>		/* kmmio_handler, ...		*/
13
#include <linux/perf_event.h>		/* perf_sw_event		*/
14
#include <linux/hugetlb.h>		/* hstate_index_to_shift	*/
15
#include <linux/prefetch.h>		/* prefetchw			*/
16
#include <linux/context_tracking.h>	/* exception_enter(), ...	*/
Ingo Molnar's avatar
Ingo Molnar committed
17

18
19
#include <asm/traps.h>			/* dotraplinkage, ...		*/
#include <asm/pgalloc.h>		/* pgd_*(), ...			*/
Vegard Nossum's avatar
Vegard Nossum committed
20
#include <asm/kmemcheck.h>		/* kmemcheck_*(), ...		*/
21
22
#include <asm/fixmap.h>			/* VSYSCALL_ADDR		*/
#include <asm/vsyscall.h>		/* emulate_vsyscall		*/
Linus Torvalds's avatar
Linus Torvalds committed
23

24
25
26
#define CREATE_TRACE_POINTS
#include <asm/trace/exceptions.h>

27
/*
Ingo Molnar's avatar
Ingo Molnar committed
28
29
30
31
32
33
34
 * 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
35
 */
Ingo Molnar's avatar
Ingo Molnar committed
36
37
38
39
40
41
42
43
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,
};
44

45
/*
46
47
 * Returns 0 if mmiotrace is disabled, or if the fault is not
 * handled by mmiotrace:
48
 */
49
50
static inline int __kprobes
kmmio_fault(struct pt_regs *regs, unsigned long addr)
51
{
52
53
54
55
	if (unlikely(is_kmmio_active()))
		if (kmmio_handler(regs, addr) == 1)
			return -1;
	return 0;
56
57
}

58
static inline int __kprobes kprobes_fault(struct pt_regs *regs)
59
{
60
61
62
	int ret = 0;

	/* kprobe_running() needs smp_processor_id() */
63
	if (kprobes_built_in() && !user_mode_vm(regs)) {
64
65
66
67
68
		preempt_disable();
		if (kprobe_running() && kprobe_fault_handler(regs, 14))
			ret = 1;
		preempt_enable();
	}
69

70
	return ret;
71
}
72

73
/*
Ingo Molnar's avatar
Ingo Molnar committed
74
75
76
77
78
79
 * Prefetch quirks:
 *
 * 32-bit mode:
 *
 *   Sometimes AMD Athlon/Opteron CPUs report invalid exceptions on prefetch.
 *   Check that here and ignore it.
80
 *
Ingo Molnar's avatar
Ingo Molnar committed
81
 * 64-bit mode:
82
 *
Ingo Molnar's avatar
Ingo Molnar committed
83
84
85
86
 *   Sometimes the CPU reports invalid exceptions on prefetch.
 *   Check that here and ignore it.
 *
 * Opcode checker based on code by Richard Brunner.
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
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.
		 */
114
		return (!user_mode(regs) || user_64bit_mode(regs));
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
#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
135
136
static int
is_prefetch(struct pt_regs *regs, unsigned long error_code, unsigned long addr)
137
{
Ingo Molnar's avatar
Ingo Molnar committed
138
	unsigned char *max_instr;
139
	unsigned char *instr;
140
	int prefetch = 0;
Linus Torvalds's avatar
Linus Torvalds committed
141

Ingo Molnar's avatar
Ingo Molnar committed
142
143
144
145
	/*
	 * If it was a exec (instruction fetch) fault on NX page, then
	 * do not ignore the fault:
	 */
146
	if (error_code & PF_INSTR)
Linus Torvalds's avatar
Linus Torvalds committed
147
		return 0;
148

149
	instr = (void *)convert_ip_to_linear(current, regs);
150
	max_instr = instr + 15;
Linus Torvalds's avatar
Linus Torvalds committed
151

152
	if (user_mode(regs) && instr >= (unsigned char *)TASK_SIZE)
Linus Torvalds's avatar
Linus Torvalds committed
153
154
		return 0;

155
	while (instr < max_instr) {
Ingo Molnar's avatar
Ingo Molnar committed
156
		unsigned char opcode;
Linus Torvalds's avatar
Linus Torvalds committed
157

158
		if (probe_kernel_address(instr, opcode))
159
			break;
Linus Torvalds's avatar
Linus Torvalds committed
160
161
162

		instr++;

163
		if (!check_prefetch_opcode(regs, instr, opcode, &prefetch))
Linus Torvalds's avatar
Linus Torvalds committed
164
165
166
167
168
			break;
	}
	return prefetch;
}

Ingo Molnar's avatar
Ingo Molnar committed
169
170
static void
force_sig_info_fault(int si_signo, int si_code, unsigned long address,
171
		     struct task_struct *tsk, int fault)
172
{
173
	unsigned lsb = 0;
174
175
	siginfo_t info;

Ingo Molnar's avatar
Ingo Molnar committed
176
177
178
179
	info.si_signo	= si_signo;
	info.si_errno	= 0;
	info.si_code	= si_code;
	info.si_addr	= (void __user *)address;
180
181
182
183
184
	if (fault & VM_FAULT_HWPOISON_LARGE)
		lsb = hstate_index_to_shift(VM_FAULT_GET_HINDEX(fault)); 
	if (fault & VM_FAULT_HWPOISON)
		lsb = PAGE_SHIFT;
	info.si_addr_lsb = lsb;
Ingo Molnar's avatar
Ingo Molnar committed
185

186
187
188
	force_sig_info(si_signo, &info, tsk);
}

189
190
191
192
193
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)
194
{
195
196
197
198
	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
199

200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
	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;

221
	if (!pmd_present(*pmd))
222
		set_pmd(pmd, *pmd_k);
223
	else
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
		BUG_ON(pmd_page(*pmd) != pmd_page(*pmd_k));

	return pmd_k;
}

void vmalloc_sync_all(void)
{
	unsigned long address;

	if (SHARED_KERNEL_PMD)
		return;

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

241
		spin_lock(&pgd_lock);
242
		list_for_each_entry(page, &pgd_list, lru) {
243
			spinlock_t *pgt_lock;
244
			pmd_t *ret;
245

246
			/* the pgt_lock only for Xen */
247
248
249
250
251
252
253
			pgt_lock = &pgd_page_get_mm(page)->page_table_lock;

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

			if (!ret)
254
255
				break;
		}
256
		spin_unlock(&pgd_lock);
257
258
259
260
261
262
263
264
	}
}

/*
 * 32-bit:
 *
 *   Handle a fault on the vmalloc or module mapping area
 */
265
static noinline __kprobes int vmalloc_fault(unsigned long address)
266
267
268
269
270
271
272
273
274
{
	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;

275
276
	WARN_ON_ONCE(in_nmi());

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
308
309
310
	/*
	 * 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;
311
}
Linus Torvalds's avatar
Linus Torvalds committed
312

Akinobu Mita's avatar
Akinobu Mita committed
313
static bool low_pfn(unsigned long pfn)
Linus Torvalds's avatar
Linus Torvalds committed
314
{
Akinobu Mita's avatar
Akinobu Mita committed
315
316
	return pfn < max_low_pfn;
}
317

Akinobu Mita's avatar
Akinobu Mita committed
318
319
320
321
322
323
static void dump_pagetable(unsigned long address)
{
	pgd_t *base = __va(read_cr3());
	pgd_t *pgd = &base[pgd_index(address)];
	pmd_t *pmd;
	pte_t *pte;
Ingo Molnar's avatar
Ingo Molnar committed
324

325
#ifdef CONFIG_X86_PAE
Akinobu Mita's avatar
Akinobu Mita committed
326
327
328
	printk("*pdpt = %016Lx ", pgd_val(*pgd));
	if (!low_pfn(pgd_val(*pgd) >> PAGE_SHIFT) || !pgd_present(*pgd))
		goto out;
329
#endif
Akinobu Mita's avatar
Akinobu Mita committed
330
331
	pmd = pmd_offset(pud_offset(pgd, address), address);
	printk(KERN_CONT "*pde = %0*Lx ", sizeof(*pmd) * 2, (u64)pmd_val(*pmd));
332
333
334
335
336

	/*
	 * 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
337
	 * it's allocated already:
338
	 */
Akinobu Mita's avatar
Akinobu Mita committed
339
340
	if (!low_pfn(pmd_pfn(*pmd)) || !pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
341

Akinobu Mita's avatar
Akinobu Mita committed
342
343
344
	pte = pte_offset_kernel(pmd, address);
	printk("*pte = %0*Lx ", sizeof(*pte) * 2, (u64)pte_val(*pte));
out:
345
	printk("\n");
346
347
348
349
350
351
}

#else /* CONFIG_X86_64: */

void vmalloc_sync_all(void)
{
352
	sync_global_pgds(VMALLOC_START & PGDIR_MASK, VMALLOC_END);
353
354
355
356
357
358
359
360
361
}

/*
 * 64-bit:
 *
 *   Handle a fault on the vmalloc area
 *
 * This assumes no large pages in there.
 */
362
static noinline __kprobes int vmalloc_fault(unsigned long address)
363
364
365
366
367
368
369
370
371
372
{
	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;

373
374
	WARN_ON_ONCE(in_nmi());

375
376
377
378
379
380
381
382
383
384
	/*
	 * 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;

385
	if (pgd_none(*pgd)) {
386
		set_pgd(pgd, *pgd_ref);
387
388
		arch_flush_lazy_mmu_mode();
	} else {
389
		BUG_ON(pgd_page_vaddr(*pgd) != pgd_page_vaddr(*pgd_ref));
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

	/*
	 * 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;
}

430
#ifdef CONFIG_CPU_SUP_AMD
431
static const char errata93_warning[] =
432
433
434
435
436
KERN_ERR 
"******* Your BIOS seems to not contain a fix for K8 errata #93\n"
"******* Working around it, but it may cause SEGVs or burn power.\n"
"******* Please consider a BIOS update.\n"
"******* Disabling USB legacy in the BIOS may also help.\n";
437
#endif
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456

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

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

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

static void dump_pagetable(unsigned long address)
{
Akinobu Mita's avatar
Akinobu Mita committed
457
458
	pgd_t *base = __va(read_cr3() & PHYSICAL_PAGE_MASK);
	pgd_t *pgd = base + pgd_index(address);
Linus Torvalds's avatar
Linus Torvalds committed
459
460
461
462
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;

Ingo Molnar's avatar
Ingo Molnar committed
463
464
465
	if (bad_address(pgd))
		goto bad;

466
	printk("PGD %lx ", pgd_val(*pgd));
Ingo Molnar's avatar
Ingo Molnar committed
467
468
469

	if (!pgd_present(*pgd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
470

471
	pud = pud_offset(pgd, address);
Ingo Molnar's avatar
Ingo Molnar committed
472
473
474
	if (bad_address(pud))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
475
	printk("PUD %lx ", pud_val(*pud));
476
	if (!pud_present(*pud) || pud_large(*pud))
Ingo Molnar's avatar
Ingo Molnar committed
477
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
478
479

	pmd = pmd_offset(pud, address);
Ingo Molnar's avatar
Ingo Molnar committed
480
481
482
	if (bad_address(pmd))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
483
	printk("PMD %lx ", pmd_val(*pmd));
Ingo Molnar's avatar
Ingo Molnar committed
484
485
	if (!pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
486
487

	pte = pte_offset_kernel(pmd, address);
Ingo Molnar's avatar
Ingo Molnar committed
488
489
490
	if (bad_address(pte))
		goto bad;

491
	printk("PTE %lx", pte_val(*pte));
Ingo Molnar's avatar
Ingo Molnar committed
492
out:
Linus Torvalds's avatar
Linus Torvalds committed
493
494
495
496
	printk("\n");
	return;
bad:
	printk("BAD\n");
497
498
}

499
#endif /* CONFIG_X86_64 */
Linus Torvalds's avatar
Linus Torvalds committed
500

Ingo Molnar's avatar
Ingo Molnar committed
501
502
503
504
505
506
507
508
509
510
511
512
513
/*
 * 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.
514
 */
515
static int is_errata93(struct pt_regs *regs, unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
516
{
517
518
519
520
521
#if defined(CONFIG_X86_64) && defined(CONFIG_CPU_SUP_AMD)
	if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD
	    || boot_cpu_data.x86 != 0xf)
		return 0;

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

525
	if ((address >> 32) != 0)
Linus Torvalds's avatar
Linus Torvalds committed
526
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
527

Linus Torvalds's avatar
Linus Torvalds committed
528
	address |= 0xffffffffUL << 32;
529
530
	if ((address >= (u64)_stext && address <= (u64)_etext) ||
	    (address >= MODULES_VADDR && address <= MODULES_END)) {
531
		printk_once(errata93_warning);
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
	if (boot_cpu_has_bug(X86_BUG_F00F)) {
565
566
567
568
569
570
571
572
573
574
575
		nr = (address - idt_descr.address) >> 3;

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

576
577
static const char nx_warning[] = KERN_CRIT
"kernel tried to execute NX-protected page - exploit attempt? (uid: %d)\n";
578
579
static const char smep_warning[] = KERN_CRIT
"unable to execute userspace code (SMEP?) (uid: %d)\n";
580

Ingo Molnar's avatar
Ingo Molnar committed
581
582
583
static void
show_fault_oops(struct pt_regs *regs, unsigned long error_code,
		unsigned long address)
584
{
585
586
587
588
	if (!oops_may_print())
		return;

	if (error_code & PF_INSTR) {
589
		unsigned int level;
590
591
		pgd_t *pgd;
		pte_t *pte;
Ingo Molnar's avatar
Ingo Molnar committed
592

593
594
595
596
		pgd = __va(read_cr3() & PHYSICAL_PAGE_MASK);
		pgd += pgd_index(address);

		pte = lookup_address_in_pgd(pgd, address, &level);
597

598
		if (pte && pte_present(*pte) && !pte_exec(*pte))
599
			printk(nx_warning, from_kuid(&init_user_ns, current_uid()));
600
601
602
603
		if (pte && pte_present(*pte) && pte_exec(*pte) &&
				(pgd_flags(*pgd) & _PAGE_USER) &&
				(read_cr4() & X86_CR4_SMEP))
			printk(smep_warning, from_kuid(&init_user_ns, current_uid()));
604
605
	}

606
	printk(KERN_ALERT "BUG: unable to handle kernel ");
607
	if (address < PAGE_SIZE)
608
		printk(KERN_CONT "NULL pointer dereference");
609
	else
610
		printk(KERN_CONT "paging request");
Ingo Molnar's avatar
Ingo Molnar committed
611

612
	printk(KERN_CONT " at %p\n", (void *) address);
613
	printk(KERN_ALERT "IP:");
614
	printk_address(regs->ip);
Ingo Molnar's avatar
Ingo Molnar committed
615

616
617
618
	dump_pagetable(address);
}

Ingo Molnar's avatar
Ingo Molnar committed
619
620
621
static noinline void
pgtable_bad(struct pt_regs *regs, unsigned long error_code,
	    unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
622
{
Ingo Molnar's avatar
Ingo Molnar committed
623
624
625
626
627
628
629
	struct task_struct *tsk;
	unsigned long flags;
	int sig;

	flags = oops_begin();
	tsk = current;
	sig = SIGKILL;
630

Linus Torvalds's avatar
Linus Torvalds committed
631
	printk(KERN_ALERT "%s: Corrupted page table at address %lx\n",
632
	       tsk->comm, address);
Linus Torvalds's avatar
Linus Torvalds committed
633
	dump_pagetable(address);
Ingo Molnar's avatar
Ingo Molnar committed
634
635

	tsk->thread.cr2		= address;
636
	tsk->thread.trap_nr	= X86_TRAP_PF;
Ingo Molnar's avatar
Ingo Molnar committed
637
638
	tsk->thread.error_code	= error_code;

639
	if (__die("Bad pagetable", regs, error_code))
640
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
641

642
	oops_end(flags, regs, sig);
Linus Torvalds's avatar
Linus Torvalds committed
643
644
}

Ingo Molnar's avatar
Ingo Molnar committed
645
646
static noinline void
no_context(struct pt_regs *regs, unsigned long error_code,
647
	   unsigned long address, int signal, int si_code)
648
649
{
	struct task_struct *tsk = current;
650
	unsigned long *stackend;
651
652
653
	unsigned long flags;
	int sig;

Ingo Molnar's avatar
Ingo Molnar committed
654
	/* Are we prepared to handle this kernel fault? */
655
	if (fixup_exception(regs)) {
656
657
658
659
660
661
662
663
664
665
666
667
668
669
		/*
		 * Any interrupt that takes a fault gets the fixup. This makes
		 * the below recursive fault logic only apply to a faults from
		 * task context.
		 */
		if (in_interrupt())
			return;

		/*
		 * Per the above we're !in_interrupt(), aka. task context.
		 *
		 * In this case we need to make sure we're not recursively
		 * faulting through the emulate_vsyscall() logic.
		 */
670
		if (current_thread_info()->sig_on_uaccess_error && signal) {
671
			tsk->thread.trap_nr = X86_TRAP_PF;
672
673
674
675
676
677
			tsk->thread.error_code = error_code | PF_USER;
			tsk->thread.cr2 = address;

			/* XXX: hwpoison faults will set the wrong code. */
			force_sig_info_fault(signal, si_code, address, tsk, 0);
		}
678
679
680
681

		/*
		 * Barring that, we can do the fixup and be happy.
		 */
682
		return;
683
	}
684
685

	/*
Ingo Molnar's avatar
Ingo Molnar committed
686
687
688
689
690
691
692
	 * 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:
693
	 *
Ingo Molnar's avatar
Ingo Molnar committed
694
	 *   Hall of shame of CPU/BIOS bugs.
695
696
697
698
699
700
701
702
703
	 */
	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
704
	 * terminate things with extreme prejudice:
705
706
707
708
709
	 */
	flags = oops_begin();

	show_fault_oops(regs, error_code, address);

Ingo Molnar's avatar
Ingo Molnar committed
710
	stackend = end_of_stack(tsk);
711
	if (tsk != &init_task && *stackend != STACK_END_MAGIC)
712
		printk(KERN_EMERG "Thread overran stack, or stack corrupted\n");
713

714
	tsk->thread.cr2		= address;
715
	tsk->thread.trap_nr	= X86_TRAP_PF;
716
	tsk->thread.error_code	= error_code;
717
718
719
720

	sig = SIGKILL;
	if (__die("Oops", regs, error_code))
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
721

722
	/* Executive summary in case the body of the oops scrolled away */
723
	printk(KERN_DEFAULT "CR2: %016lx\n", address);
Ingo Molnar's avatar
Ingo Molnar committed
724

725
726
727
	oops_end(flags, regs, sig);
}

Ingo Molnar's avatar
Ingo Molnar committed
728
729
730
731
732
733
734
735
736
737
738
739
740
741
/*
 * 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;

742
	printk("%s%s[%d]: segfault at %lx ip %p sp %p error %lx",
Ingo Molnar's avatar
Ingo Molnar committed
743
744
745
746
747
748
749
750
751
752
753
754
		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)
755
756
757
758
759
760
{
	struct task_struct *tsk = current;

	/* User mode accesses just cause a SIGSEGV */
	if (error_code & PF_USER) {
		/*
Ingo Molnar's avatar
Ingo Molnar committed
761
		 * It's possible to have interrupts off here:
762
763
764
765
766
		 */
		local_irq_enable();

		/*
		 * Valid to do another page fault here because this one came
Ingo Molnar's avatar
Ingo Molnar committed
767
		 * from user space:
768
769
770
771
772
773
774
		 */
		if (is_prefetch(regs, error_code, address))
			return;

		if (is_errata100(regs, address))
			return;

775
776
777
778
779
780
#ifdef CONFIG_X86_64
		/*
		 * Instruction fetch faults in the vsyscall page might need
		 * emulation.
		 */
		if (unlikely((error_code & PF_INSTR) &&
781
			     ((address & ~0xfff) == VSYSCALL_ADDR))) {
782
783
784
785
			if (emulate_vsyscall(regs, address))
				return;
		}
#endif
786
787
788
		/* Kernel addresses are always protection faults: */
		if (address >= TASK_SIZE)
			error_code |= PF_PROT;
789

790
		if (likely(show_unhandled_signals))
Ingo Molnar's avatar
Ingo Molnar committed
791
792
793
			show_signal_msg(regs, error_code, address, tsk);

		tsk->thread.cr2		= address;
794
		tsk->thread.error_code	= error_code;
795
		tsk->thread.trap_nr	= X86_TRAP_PF;
796

797
		force_sig_info_fault(SIGSEGV, si_code, address, tsk, 0);
Ingo Molnar's avatar
Ingo Molnar committed
798

799
800
801
802
803
804
		return;
	}

	if (is_f00f_bug(regs, address))
		return;

805
	no_context(regs, error_code, address, SIGSEGV, si_code);
806
807
}

Ingo Molnar's avatar
Ingo Molnar committed
808
809
810
static noinline void
bad_area_nosemaphore(struct pt_regs *regs, unsigned long error_code,
		     unsigned long address)
811
812
813
814
{
	__bad_area_nosemaphore(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
815
816
817
static void
__bad_area(struct pt_regs *regs, unsigned long error_code,
	   unsigned long address, int si_code)
818
819
820
821
822
823
824
825
826
827
828
829
{
	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
830
831
static noinline void
bad_area(struct pt_regs *regs, unsigned long error_code, unsigned long address)
832
833
834
835
{
	__bad_area(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
836
837
838
static noinline void
bad_area_access_error(struct pt_regs *regs, unsigned long error_code,
		      unsigned long address)
839
840
841
842
{
	__bad_area(regs, error_code, address, SEGV_ACCERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
843
static void
844
845
do_sigbus(struct pt_regs *regs, unsigned long error_code, unsigned long address,
	  unsigned int fault)
846
847
848
{
	struct task_struct *tsk = current;
	struct mm_struct *mm = tsk->mm;
849
	int code = BUS_ADRERR;
850
851
852

	up_read(&mm->mmap_sem);

Ingo Molnar's avatar
Ingo Molnar committed
853
	/* Kernel mode? Handle exceptions or die: */
854
	if (!(error_code & PF_USER)) {
855
		no_context(regs, error_code, address, SIGBUS, BUS_ADRERR);
856
857
		return;
	}
Ingo Molnar's avatar
Ingo Molnar committed
858

859
	/* User-space => ok to do another page fault: */
860
861
	if (is_prefetch(regs, error_code, address))
		return;
Ingo Molnar's avatar
Ingo Molnar committed
862
863
864

	tsk->thread.cr2		= address;
	tsk->thread.error_code	= error_code;
865
	tsk->thread.trap_nr	= X86_TRAP_PF;
Ingo Molnar's avatar
Ingo Molnar committed
866

867
#ifdef CONFIG_MEMORY_FAILURE
868
	if (fault & (VM_FAULT_HWPOISON|VM_FAULT_HWPOISON_LARGE)) {
869
870
871
872
873
874
		printk(KERN_ERR
	"MCE: Killing %s:%d due to hardware memory corruption fault at %lx\n",
			tsk->comm, tsk->pid, address);
		code = BUS_MCEERR_AR;
	}
#endif
875
	force_sig_info_fault(SIGBUS, code, address, tsk, fault);
876
877
}

878
static noinline void
Ingo Molnar's avatar
Ingo Molnar committed
879
880
mm_fault_error(struct pt_regs *regs, unsigned long error_code,
	       unsigned long address, unsigned int fault)
881
{
882
883
884
885
	if (fatal_signal_pending(current) && !(error_code & PF_USER)) {
		up_read(&current->mm->mmap_sem);
		no_context(regs, error_code, address, 0, 0);
		return;
886
887
	}

Ingo Molnar's avatar
Ingo Molnar committed
888
	if (fault & VM_FAULT_OOM) {
889
890
891
		/* Kernel mode? Handle exceptions or die: */
		if (!(error_code & PF_USER)) {
			up_read(&current->mm->mmap_sem);
892
893
			no_context(regs, error_code, address,
				   SIGSEGV, SEGV_MAPERR);
894
			return;
895
896
		}

897
898
899
900
901
902
903
904
		up_read(&current->mm->mmap_sem);

		/*
		 * We ran out of memory, call the OOM killer, and return the
		 * userspace (which will retry the fault, or kill us if we got
		 * oom-killed):
		 */
		pagefault_out_of_memory();
Ingo Molnar's avatar
Ingo Molnar committed
905
	} else {
906
907
		if (fault & (VM_FAULT_SIGBUS|VM_FAULT_HWPOISON|
			     VM_FAULT_HWPOISON_LARGE))
908
			do_sigbus(regs, error_code, address, fault);
Ingo Molnar's avatar
Ingo Molnar committed
909
910
911
		else
			BUG();
	}
912
913
}

914
915
916
917
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
918

919
920
921
922
923
924
	if ((error_code & PF_INSTR) && !pte_exec(*pte))
		return 0;

	return 1;
}

925
/*
Ingo Molnar's avatar
Ingo Molnar committed
926
927
928
929
930
931
932
933
 * 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.
 *
934
935
936
 * There are no security implications to leaving a stale TLB when
 * increasing the permissions on a page.
 */
937
static noinline __kprobes int
Ingo Molnar's avatar
Ingo Molnar committed
938
spurious_fault(unsigned long error_code, unsigned long address)
939
940
941
942
943
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;
944
	int ret;
945
946
947
948
949
950
951
952
953
954
955
956
957

	/* 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;

958
959
960
	if (pud_large(*pud))
		return spurious_fault_check(error_code, (pte_t *) pud);

961
962
963
964
	pmd = pmd_offset(pud, address);
	if (!pmd_present(*pmd))
		return 0;

965
966
967
	if (pmd_large(*pmd))
		return spurious_fault_check(error_code, (pte_t *) pmd);

968
	pte = pte_offset_kernel(pmd, address);
969
	if (!pte_present(*pte))
970
971
		return 0;

972
973
974
975
976
	ret = spurious_fault_check(error_code, pte);
	if (!ret)
		return 0;

	/*
Ingo Molnar's avatar
Ingo Molnar committed
977
978
	 * Make sure we have permissions in PMD.
	 * If not, then there's a bug in the page tables:
979
980
981
	 */
	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
982

983
	return ret;
984
985
}

986
int show_unhandled_signals = 1;
Linus Torvalds's avatar
Linus Torvalds committed
987

Ingo Molnar's avatar
Ingo Molnar committed
988
static inline int
989
access_error(unsigned long error_code, struct vm_area_struct *vma)
990
{
991
	if (error_code & PF_WRITE) {
Ingo Molnar's avatar
Ingo Molnar committed
992
		/* write, present and write, not present: */
993
994
		if (unlikely(!(vma->vm_flags & VM_WRITE)))
			return 1;
Ingo Molnar's avatar
Ingo Molnar committed
995
		return 0;
996
997
	}

Ingo Molnar's avatar
Ingo Molnar committed
998
999
1000
1001
1002
1003
1004
1005
	/* 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;

1006
1007
1008
	return 0;
}

1009
1010
static int fault_in_kernel_space(unsigned long address)
{
1011
	return address >= TASK_SIZE_MAX;
1012
1013
}

1014
1015
static inline bool smap_violation(int error_code, struct pt_regs *regs)
{
1016
1017
1018
1019
1020
1021
	if (!IS_ENABLED(CONFIG_X86_SMAP))
		return false;

	if (!static_cpu_has(X86_FEATURE_SMAP))
		return false;

1022
1023
1024
1025
1026
1027
1028
1029
1030
	if (error_code & PF_USER)
		return false;

	if (!user_mode_vm(regs) && (regs->flags & X86_EFLAGS_AC))
		return false;

	return true;
}

Linus Torvalds's avatar
Linus Torvalds committed
1031
1032
1033
1034
/*
 * This routine handles page faults.  It determines the address,
 * and the problem, and then passes it off to one of the appropriate
 * routines.
1035
1036
1037
1038
 *
 * This function must have noinline because both callers
 * {,trace_}do_page_fault() have notrace on. Having this an actual function
 * guarantees there's a function trace entry.
Linus Torvalds's avatar
Linus Torvalds committed
1039
 */
1040
static void __kprobes noinline
1041
1042
__do_page_fault(struct pt_regs *regs, unsigned long error_code,
		unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
1043
{
Ingo Molnar's avatar
Ingo Molnar committed
1044
	struct vm_area_struct *vma;
Linus Torvalds's avatar
Linus Torvalds committed
1045
1046
	struct task_struct *tsk;
	struct mm_struct *mm;
1047
	int fault;
1048
	unsigned int flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
Linus Torvalds's avatar
Linus Torvalds committed
1049

1050
1051
	tsk = current;
	mm = tsk->mm;
Ingo Molnar's avatar
Ingo Molnar committed
1052

Vegard Nossum's avatar
Vegard Nossum committed
1053
1054
1055
1056
1057
1058
	/*
	 * Detect and handle instructions that would cause a page fault for
	 * both a tracked kernel page and a userspace page.
	 */
	if (kmemcheck_active(regs))
		kmemcheck_hide(regs);
1059
	prefetchw(&mm->mmap_sem);
Vegard Nossum's avatar
Vegard Nossum committed
1060

1061
	if (unlikely(kmmio_fault(regs, address)))
1062
		return;
Linus Torvalds's avatar
Linus Torvalds committed
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074

	/*
	 * 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
1075
	 * protection error (error_code & 9) == 0.
Linus Torvalds's avatar
Linus Torvalds committed
1076
	 */
1077
	if (unlikely(fault_in_kernel_space(address))) {
Vegard Nossum's avatar
Vegard Nossum committed
1078
1079
1080
1081
1082
1083
1084
		if (!(error_code & (PF_RSVD | PF_USER | PF_PROT))) {
			if (vmalloc_fault(address) >= 0)
				return;

			if (kmemcheck_fault(regs, address, error_code))
				return;
		}
1085

Ingo Molnar's avatar
Ingo Molnar committed
1086
		/* Can handle a stale RO->RW TLB: */
1087
		if (spurious_fault(error_code, address))
1088
1089
			return;

Ingo Molnar's avatar
Ingo Molnar committed
1090
		/* kprobes don't want to hook the spurious faults: */
1091
		if (kprobes_fault(regs))
1092
			return;
1093
1094
		/*
		 * Don't take the mm semaphore here. If we fixup a prefetch
Ingo Molnar's avatar
Ingo Molnar committed
1095
		 * fault we could otherwise deadlock:
1096
		 */
1097
		bad_area_nosemaphore(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
1098

1099
		return;
1100
1101
	}

Ingo Molnar's avatar
Ingo Molnar committed
1102
	/* kprobes don't want to hook the spurious faults: */
1103
	if (unlikely(kprobes_fault(regs)))
1104
		return;
1105

1106
	if (unlikely(error_code & PF_RSVD))
1107
		pgtable_bad(regs, error_code, address);
Linus Torvalds's avatar
Linus Torvalds committed
1108

1109
1110
1111
	if (unlikely(smap_violation(error_code, regs))) {
		bad_area_nosemaphore(regs, error_code, address);
		return;
1112
1113
	}

Linus Torvalds's avatar
Linus Torvalds committed
1114
	/*
Ingo Molnar's avatar
Ingo Molnar committed
1115
1116
	 * 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
1117
	 */
1118
1119
1120
1121
	if (unlikely(in_atomic() || !mm)) {
		bad_area_nosemaphore(regs, error_code, address);
		return;
	}
Linus Torvalds's avatar
Linus Torvalds committed
1122

1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
	/*
	 * 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
	 * potential system fault or CPU buglet:
	 */
	if (user_mode_vm(regs)) {
		local_irq_enable();
		error_code |= PF_USER;
		flags |= FAULT_FLAG_USER;
	} else {
		if (regs->flags & X86_EFLAGS_IF)
			local_irq_enable();
	}

	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address);

1141
1142
1143
	if (error_code & PF_WRITE)
		flags |= FAULT_FLAG_WRITE;

1144
1145
	/*
	 * When running in the kernel we expect faults to occur only to
Ingo Molnar's avatar
Ingo Molnar committed
1146
1147
1148
1149
1150