fault.c 32.4 KB
Newer Older
L
Linus Torvalds 已提交
1 2
/*
 *  Copyright (C) 1995  Linus Torvalds
I
Ingo Molnar 已提交
3
 *  Copyright (C) 2001, 2002 Andi Kleen, SuSE Labs.
4
 *  Copyright (C) 2008-2009, Red Hat Inc., Ingo Molnar
L
Linus Torvalds 已提交
5
 */
6 7 8 9
#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			*/
10
#include <linux/kprobes.h>		/* NOKPROBE_SYMBOL, ...		*/
11
#include <linux/mmiotrace.h>		/* kmmio_handler, ...		*/
12
#include <linux/perf_event.h>		/* perf_sw_event		*/
13
#include <linux/hugetlb.h>		/* hstate_index_to_shift	*/
14
#include <linux/prefetch.h>		/* prefetchw			*/
15
#include <linux/context_tracking.h>	/* exception_enter(), ...	*/
16
#include <linux/uaccess.h>		/* faulthandler_disabled()	*/
I
Ingo Molnar 已提交
17

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

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

27
/*
I
Ingo Molnar 已提交
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
 */
I
Ingo Molnar 已提交
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
static nokprobe_inline int
50
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 nokprobe_inline int 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(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
/*
I
Ingo Molnar 已提交
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
 *
I
Ingo Molnar 已提交
81
 * 64-bit mode:
82
 *
I
Ingo Molnar 已提交
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;
	}
}

I
Ingo Molnar 已提交
135 136
static int
is_prefetch(struct pt_regs *regs, unsigned long error_code, unsigned long addr)
137
{
I
Ingo Molnar 已提交
138
	unsigned char *max_instr;
139
	unsigned char *instr;
140
	int prefetch = 0;
L
Linus Torvalds 已提交
141

I
Ingo Molnar 已提交
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)
L
Linus Torvalds 已提交
147
		return 0;
148

149
	instr = (void *)convert_ip_to_linear(current, regs);
150
	max_instr = instr + 15;
L
Linus Torvalds 已提交
151

152
	if (user_mode(regs) && instr >= (unsigned char *)TASK_SIZE_MAX)
L
Linus Torvalds 已提交
153 154
		return 0;

155
	while (instr < max_instr) {
I
Ingo Molnar 已提交
156
		unsigned char opcode;
L
Linus Torvalds 已提交
157

158
		if (probe_kernel_address(instr, opcode))
159
			break;
L
Linus Torvalds 已提交
160 161 162

		instr++;

163
		if (!check_prefetch_opcode(regs, instr, opcode, &prefetch))
L
Linus Torvalds 已提交
164 165 166 167 168
			break;
	}
	return prefetch;
}

I
Ingo Molnar 已提交
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;

I
Ingo Molnar 已提交
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;
I
Ingo Molnar 已提交
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;
I
Ingo Molnar 已提交
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;

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

A
Andrea Arcangeli 已提交
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;
		}
A
Andrea Arcangeli 已提交
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 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
	/*
	 * 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;
}
295
NOKPROBE_SYMBOL(vmalloc_fault);
296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311

/*
 * 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;
312
}
L
Linus Torvalds 已提交
313

A
Akinobu Mita 已提交
314
static bool low_pfn(unsigned long pfn)
L
Linus Torvalds 已提交
315
{
A
Akinobu Mita 已提交
316 317
	return pfn < max_low_pfn;
}
318

A
Akinobu Mita 已提交
319 320 321 322 323 324
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;
I
Ingo Molnar 已提交
325

326
#ifdef CONFIG_X86_PAE
A
Akinobu Mita 已提交
327 328 329
	printk("*pdpt = %016Lx ", pgd_val(*pgd));
	if (!low_pfn(pgd_val(*pgd) >> PAGE_SHIFT) || !pgd_present(*pgd))
		goto out;
330
#endif
A
Akinobu Mita 已提交
331 332
	pmd = pmd_offset(pud_offset(pgd, address), address);
	printk(KERN_CONT "*pde = %0*Lx ", sizeof(*pmd) * 2, (u64)pmd_val(*pmd));
333 334 335 336 337

	/*
	 * 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
I
Ingo Molnar 已提交
338
	 * it's allocated already:
339
	 */
A
Akinobu Mita 已提交
340 341
	if (!low_pfn(pmd_pfn(*pmd)) || !pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
342

A
Akinobu Mita 已提交
343 344 345
	pte = pte_offset_kernel(pmd, address);
	printk("*pte = %0*Lx ", sizeof(*pte) * 2, (u64)pte_val(*pte));
out:
346
	printk("\n");
347 348 349 350 351 352
}

#else /* CONFIG_X86_64: */

void vmalloc_sync_all(void)
{
353
	sync_global_pgds(VMALLOC_START & PGDIR_MASK, VMALLOC_END, 0);
354 355 356 357 358 359 360 361 362
}

/*
 * 64-bit:
 *
 *   Handle a fault on the vmalloc area
 *
 * This assumes no large pages in there.
 */
363
static noinline int vmalloc_fault(unsigned long address)
364 365 366 367 368 369 370 371 372 373
{
	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;

374 375
	WARN_ON_ONCE(in_nmi());

376 377 378 379 380 381 382 383 384 385
	/*
	 * 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;

386
	if (pgd_none(*pgd)) {
387
		set_pgd(pgd, *pgd_ref);
388 389
		arch_flush_lazy_mmu_mode();
	} else {
390
		BUG_ON(pgd_page_vaddr(*pgd) != pgd_page_vaddr(*pgd_ref));
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
NOKPROBE_SYMBOL(vmalloc_fault);
431

432
#ifdef CONFIG_CPU_SUP_AMD
433
static const char errata93_warning[] =
434 435 436 437 438
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";
439
#endif
440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458

/*
 * 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)
{
A
Akinobu Mita 已提交
459 460
	pgd_t *base = __va(read_cr3() & PHYSICAL_PAGE_MASK);
	pgd_t *pgd = base + pgd_index(address);
L
Linus Torvalds 已提交
461 462 463 464
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;

I
Ingo Molnar 已提交
465 466 467
	if (bad_address(pgd))
		goto bad;

468
	printk("PGD %lx ", pgd_val(*pgd));
I
Ingo Molnar 已提交
469 470 471

	if (!pgd_present(*pgd))
		goto out;
L
Linus Torvalds 已提交
472

473
	pud = pud_offset(pgd, address);
I
Ingo Molnar 已提交
474 475 476
	if (bad_address(pud))
		goto bad;

L
Linus Torvalds 已提交
477
	printk("PUD %lx ", pud_val(*pud));
478
	if (!pud_present(*pud) || pud_large(*pud))
I
Ingo Molnar 已提交
479
		goto out;
L
Linus Torvalds 已提交
480 481

	pmd = pmd_offset(pud, address);
I
Ingo Molnar 已提交
482 483 484
	if (bad_address(pmd))
		goto bad;

L
Linus Torvalds 已提交
485
	printk("PMD %lx ", pmd_val(*pmd));
I
Ingo Molnar 已提交
486 487
	if (!pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
L
Linus Torvalds 已提交
488 489

	pte = pte_offset_kernel(pmd, address);
I
Ingo Molnar 已提交
490 491 492
	if (bad_address(pte))
		goto bad;

493
	printk("PTE %lx", pte_val(*pte));
I
Ingo Molnar 已提交
494
out:
L
Linus Torvalds 已提交
495 496 497 498
	printk("\n");
	return;
bad:
	printk("BAD\n");
499 500
}

501
#endif /* CONFIG_X86_64 */
L
Linus Torvalds 已提交
502

I
Ingo Molnar 已提交
503 504 505 506 507 508 509 510 511 512 513 514 515
/*
 * 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.
516
 */
517
static int is_errata93(struct pt_regs *regs, unsigned long address)
L
Linus Torvalds 已提交
518
{
519 520 521 522 523
#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;

524
	if (address != regs->ip)
L
Linus Torvalds 已提交
525
		return 0;
I
Ingo Molnar 已提交
526

527
	if ((address >> 32) != 0)
L
Linus Torvalds 已提交
528
		return 0;
I
Ingo Molnar 已提交
529

L
Linus Torvalds 已提交
530
	address |= 0xffffffffUL << 32;
531 532
	if ((address >= (u64)_stext && address <= (u64)_etext) ||
	    (address >= MODULES_VADDR && address <= MODULES_END)) {
533
		printk_once(errata93_warning);
534
		regs->ip = address;
L
Linus Torvalds 已提交
535 536
		return 1;
	}
537
#endif
L
Linus Torvalds 已提交
538
	return 0;
539
}
L
Linus Torvalds 已提交
540

541
/*
I
Ingo Molnar 已提交
542 543 544 545 546
 * 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
547 548 549 550 551
 * segment in LDT is compatibility mode.
 */
static int is_errata100(struct pt_regs *regs, unsigned long address)
{
#ifdef CONFIG_X86_64
I
Ingo Molnar 已提交
552
	if ((regs->cs == __USER32_CS || (regs->cs & (1<<2))) && (address >> 32))
553 554 555 556 557
		return 1;
#endif
	return 0;
}

558 559 560 561
static int is_f00f_bug(struct pt_regs *regs, unsigned long address)
{
#ifdef CONFIG_X86_F00F_BUG
	unsigned long nr;
I
Ingo Molnar 已提交
562

563
	/*
I
Ingo Molnar 已提交
564
	 * Pentium F0 0F C7 C8 bug workaround:
565
	 */
566
	if (boot_cpu_has_bug(X86_BUG_F00F)) {
567 568 569 570 571 572 573 574 575 576 577
		nr = (address - idt_descr.address) >> 3;

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

578 579
static const char nx_warning[] = KERN_CRIT
"kernel tried to execute NX-protected page - exploit attempt? (uid: %d)\n";
580 581
static const char smep_warning[] = KERN_CRIT
"unable to execute userspace code (SMEP?) (uid: %d)\n";
582

I
Ingo Molnar 已提交
583 584 585
static void
show_fault_oops(struct pt_regs *regs, unsigned long error_code,
		unsigned long address)
586
{
587 588 589 590
	if (!oops_may_print())
		return;

	if (error_code & PF_INSTR) {
591
		unsigned int level;
592 593
		pgd_t *pgd;
		pte_t *pte;
I
Ingo Molnar 已提交
594

595 596 597 598
		pgd = __va(read_cr3() & PHYSICAL_PAGE_MASK);
		pgd += pgd_index(address);

		pte = lookup_address_in_pgd(pgd, address, &level);
599

600
		if (pte && pte_present(*pte) && !pte_exec(*pte))
601
			printk(nx_warning, from_kuid(&init_user_ns, current_uid()));
602 603
		if (pte && pte_present(*pte) && pte_exec(*pte) &&
				(pgd_flags(*pgd) & _PAGE_USER) &&
604
				(__read_cr4() & X86_CR4_SMEP))
605
			printk(smep_warning, from_kuid(&init_user_ns, current_uid()));
606 607
	}

608
	printk(KERN_ALERT "BUG: unable to handle kernel ");
609
	if (address < PAGE_SIZE)
610
		printk(KERN_CONT "NULL pointer dereference");
611
	else
612
		printk(KERN_CONT "paging request");
I
Ingo Molnar 已提交
613

614
	printk(KERN_CONT " at %p\n", (void *) address);
615
	printk(KERN_ALERT "IP:");
616
	printk_address(regs->ip);
I
Ingo Molnar 已提交
617

618 619 620
	dump_pagetable(address);
}

I
Ingo Molnar 已提交
621 622 623
static noinline void
pgtable_bad(struct pt_regs *regs, unsigned long error_code,
	    unsigned long address)
L
Linus Torvalds 已提交
624
{
I
Ingo Molnar 已提交
625 626 627 628 629 630 631
	struct task_struct *tsk;
	unsigned long flags;
	int sig;

	flags = oops_begin();
	tsk = current;
	sig = SIGKILL;
632

L
Linus Torvalds 已提交
633
	printk(KERN_ALERT "%s: Corrupted page table at address %lx\n",
634
	       tsk->comm, address);
L
Linus Torvalds 已提交
635
	dump_pagetable(address);
I
Ingo Molnar 已提交
636 637

	tsk->thread.cr2		= address;
638
	tsk->thread.trap_nr	= X86_TRAP_PF;
I
Ingo Molnar 已提交
639 640
	tsk->thread.error_code	= error_code;

641
	if (__die("Bad pagetable", regs, error_code))
642
		sig = 0;
I
Ingo Molnar 已提交
643

644
	oops_end(flags, regs, sig);
L
Linus Torvalds 已提交
645 646
}

I
Ingo Molnar 已提交
647 648
static noinline void
no_context(struct pt_regs *regs, unsigned long error_code,
649
	   unsigned long address, int signal, int si_code)
650 651 652 653 654
{
	struct task_struct *tsk = current;
	unsigned long flags;
	int sig;

I
Ingo Molnar 已提交
655
	/* Are we prepared to handle this kernel fault? */
656
	if (fixup_exception(regs)) {
657 658 659 660 661 662 663 664 665 666 667 668 669 670
		/*
		 * 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.
		 */
671
		if (current_thread_info()->sig_on_uaccess_error && signal) {
672
			tsk->thread.trap_nr = X86_TRAP_PF;
673 674 675 676 677 678
			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);
		}
679 680 681 682

		/*
		 * Barring that, we can do the fixup and be happy.
		 */
683
		return;
684
	}
685 686

	/*
I
Ingo Molnar 已提交
687 688 689 690 691 692 693
	 * 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:
694
	 *
I
Ingo Molnar 已提交
695
	 *   Hall of shame of CPU/BIOS bugs.
696 697 698 699 700 701 702 703 704
	 */
	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
I
Ingo Molnar 已提交
705
	 * terminate things with extreme prejudice:
706 707 708 709 710
	 */
	flags = oops_begin();

	show_fault_oops(regs, error_code, address);

711
	if (task_stack_end_corrupted(tsk))
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;
I
Ingo Molnar 已提交
721

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

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

I
Ingo Molnar 已提交
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",
I
Ingo Molnar 已提交
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) {
		/*
I
Ingo Molnar 已提交
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
I
Ingo Molnar 已提交
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))
I
Ingo Molnar 已提交
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);
I
Ingo Molnar 已提交
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
}

I
Ingo Molnar 已提交
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);
}

I
Ingo Molnar 已提交
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);
}

I
Ingo Molnar 已提交
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);
}

I
Ingo Molnar 已提交
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);
}

I
Ingo Molnar 已提交
843
static void
844 845
do_sigbus(struct pt_regs *regs, unsigned long error_code, unsigned long address,
	  unsigned int fault)
846 847
{
	struct task_struct *tsk = current;
848
	int code = BUS_ADRERR;
849

I
Ingo Molnar 已提交
850
	/* Kernel mode? Handle exceptions or die: */
851
	if (!(error_code & PF_USER)) {
852
		no_context(regs, error_code, address, SIGBUS, BUS_ADRERR);
853 854
		return;
	}
I
Ingo Molnar 已提交
855

856
	/* User-space => ok to do another page fault: */
857 858
	if (is_prefetch(regs, error_code, address))
		return;
I
Ingo Molnar 已提交
859 860 861

	tsk->thread.cr2		= address;
	tsk->thread.error_code	= error_code;
862
	tsk->thread.trap_nr	= X86_TRAP_PF;
I
Ingo Molnar 已提交
863

864
#ifdef CONFIG_MEMORY_FAILURE
865
	if (fault & (VM_FAULT_HWPOISON|VM_FAULT_HWPOISON_LARGE)) {
866 867 868 869 870 871
		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
872
	force_sig_info_fault(SIGBUS, code, address, tsk, fault);
873 874
}

875
static noinline void
I
Ingo Molnar 已提交
876 877
mm_fault_error(struct pt_regs *regs, unsigned long error_code,
	       unsigned long address, unsigned int fault)
878
{
879 880 881
	if (fatal_signal_pending(current) && !(error_code & PF_USER)) {
		no_context(regs, error_code, address, 0, 0);
		return;
882 883
	}

I
Ingo Molnar 已提交
884
	if (fault & VM_FAULT_OOM) {
885 886
		/* Kernel mode? Handle exceptions or die: */
		if (!(error_code & PF_USER)) {
887 888
			no_context(regs, error_code, address,
				   SIGSEGV, SEGV_MAPERR);
889
			return;
890 891
		}

892 893 894 895 896 897
		/*
		 * 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();
I
Ingo Molnar 已提交
898
	} else {
899 900
		if (fault & (VM_FAULT_SIGBUS|VM_FAULT_HWPOISON|
			     VM_FAULT_HWPOISON_LARGE))
901
			do_sigbus(regs, error_code, address, fault);
902 903
		else if (fault & VM_FAULT_SIGSEGV)
			bad_area_nosemaphore(regs, error_code, address);
I
Ingo Molnar 已提交
904 905 906
		else
			BUG();
	}
907 908
}

909 910 911 912
static int spurious_fault_check(unsigned long error_code, pte_t *pte)
{
	if ((error_code & PF_WRITE) && !pte_write(*pte))
		return 0;
I
Ingo Molnar 已提交
913

914 915 916 917 918 919
	if ((error_code & PF_INSTR) && !pte_exec(*pte))
		return 0;

	return 1;
}

920
/*
I
Ingo Molnar 已提交
921 922 923 924 925 926 927 928
 * 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.
 *
929 930 931 932
 * Spurious faults may only occur if the TLB contains an entry with
 * fewer permission than the page table entry.  Non-present (P = 0)
 * and reserved bit (R = 1) faults are never spurious.
 *
933 934
 * There are no security implications to leaving a stale TLB when
 * increasing the permissions on a page.
935 936 937 938 939
 *
 * Returns non-zero if a spurious fault was handled, zero otherwise.
 *
 * See Intel Developer's Manual Vol 3 Section 4.10.4.3, bullet 3
 * (Optional Invalidation).
940
 */
941
static noinline int
I
Ingo Molnar 已提交
942
spurious_fault(unsigned long error_code, unsigned long address)
943 944 945 946 947
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;
948
	int ret;
949

950 951 952 953 954 955 956 957 958 959 960
	/*
	 * Only writes to RO or instruction fetches from NX may cause
	 * spurious faults.
	 *
	 * These could be from user or supervisor accesses but the TLB
	 * is only lazily flushed after a kernel mapping protection
	 * change, so user accesses are not expected to cause spurious
	 * faults.
	 */
	if (error_code != (PF_WRITE | PF_PROT)
	    && error_code != (PF_INSTR | PF_PROT))
961 962 963 964 965 966 967 968 969 970
		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;

971 972 973
	if (pud_large(*pud))
		return spurious_fault_check(error_code, (pte_t *) pud);

974 975 976 977
	pmd = pmd_offset(pud, address);
	if (!pmd_present(*pmd))
		return 0;

978 979 980
	if (pmd_large(*pmd))
		return spurious_fault_check(error_code, (pte_t *) pmd);

981
	pte = pte_offset_kernel(pmd, address);
982
	if (!pte_present(*pte))
983 984
		return 0;

985 986 987 988 989
	ret = spurious_fault_check(error_code, pte);
	if (!ret)
		return 0;

	/*
I
Ingo Molnar 已提交
990 991
	 * Make sure we have permissions in PMD.
	 * If not, then there's a bug in the page tables:
992 993 994
	 */
	ret = spurious_fault_check(error_code, (pte_t *) pmd);
	WARN_ONCE(!ret, "PMD has incorrect permission bits\n");
I
Ingo Molnar 已提交
995

996
	return ret;
997
}
998
NOKPROBE_SYMBOL(spurious_fault);
999

1000
int show_unhandled_signals = 1;
L
Linus Torvalds 已提交
1001

I
Ingo Molnar 已提交
1002
static inline int
M
Michel Lespinasse 已提交
1003
access_error(unsigned long error_code, struct vm_area_struct *vma)
1004
{
M
Michel Lespinasse 已提交
1005
	if (error_code & PF_WRITE) {
I
Ingo Molnar 已提交
1006
		/* write, present and write, not present: */
1007 1008
		if (unlikely(!(vma->vm_flags & VM_WRITE)))
			return 1;
I
Ingo Molnar 已提交
1009
		return 0;
1010 1011
	}

I
Ingo Molnar 已提交
1012 1013 1014 1015 1016 1017 1018 1019
	/* 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;

1020 1021 1022
	return 0;
}

1023 1024
static int fault_in_kernel_space(unsigned long address)
{
1025
	return address >= TASK_SIZE_MAX;
1026 1027
}

1028 1029
static inline bool smap_violation(int error_code, struct pt_regs *regs)
{
1030 1031 1032 1033 1034 1035
	if (!IS_ENABLED(CONFIG_X86_SMAP))
		return false;

	if (!static_cpu_has(X86_FEATURE_SMAP))
		return false;

1036 1037 1038
	if (error_code & PF_USER)
		return false;

1039
	if (!user_mode(regs) && (regs->flags & X86_EFLAGS_AC))
1040 1041 1042 1043 1044
		return false;

	return true;
}

L
Linus Torvalds 已提交
1045 1046 1047 1048
/*
 * This routine handles page faults.  It determines the address,
 * and the problem, and then passes it off to one of the appropriate
 * routines.
1049 1050 1051 1052
 *
 * 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.
L
Linus Torvalds 已提交
1053
 */
1054
static noinline void
1055 1056
__do_page_fault(struct pt_regs *regs, unsigned long error_code,
		unsigned long address)
L
Linus Torvalds 已提交
1057
{
I
Ingo Molnar 已提交
1058
	struct vm_area_struct *vma;
L
Linus Torvalds 已提交
1059 1060
	struct task_struct *tsk;
	struct mm_struct *mm;
1061
	int fault, major = 0;
1062
	unsigned int flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
L
Linus Torvalds 已提交
1063

1064 1065
	tsk = current;
	mm = tsk->mm;
I
Ingo Molnar 已提交
1066

V
Vegard Nossum 已提交
1067 1068 1069 1070 1071 1072
	/*
	 * 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);
1073
	prefetchw(&mm->mmap_sem);
V
Vegard Nossum 已提交
1074

1075
	if (unlikely(kmmio_fault(regs, address)))
1076
		return;
L
Linus Torvalds 已提交
1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088

	/*
	 * 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
1089
	 * protection error (error_code & 9) == 0.
L
Linus Torvalds 已提交
1090
	 */
1091
	if (unlikely(fault_in_kernel_space(address))) {
V
Vegard Nossum 已提交
1092 1093 1094 1095 1096 1097 1098
		if (!(error_code & (PF_RSVD | PF_USER | PF_PROT))) {
			if (vmalloc_fault(address) >= 0)
				return;

			if (kmemcheck_fault(regs, address, error_code))
				return;
		}
1099

I
Ingo Molnar 已提交
1100
		/* Can handle a stale RO->RW TLB: */
1101
		if (spurious_fault(error_code, address))
1102 1103
			return;

I
Ingo Molnar 已提交
1104
		/* kprobes don't want to hook the spurious faults: */
1105
		if (kprobes_fault(regs))
1106
			return;
1107 1108
		/*
		 * Don't take the mm semaphore here. If we fixup a prefetch
I
Ingo Molnar 已提交
1109
		 * fault we could otherwise deadlock:
1110
		 */
1111
		bad_area_nosemaphore(regs, error_code, address);
I
Ingo Molnar 已提交
1112

1113
		return;
1114 1115
	}

I
Ingo Molnar 已提交
1116
	/* kprobes don't want to hook the spurious faults: */
1117
	if (unlikely(kprobes_fault(regs)))
1118
		return;
1119

1120
	if (unlikely(error_code & PF_RSVD))
1121
		pgtable_bad(regs, error_code, address);
L
Linus Torvalds 已提交
1122

1123 1124 1125
	if (unlikely(smap_violation(error_code, regs))) {
		bad_area_nosemaphore(regs, error_code, address);
		return;
1126 1127
	}

L
Linus Torvalds 已提交
1128
	/*
I
Ingo Molnar 已提交
1129
	 * If we're in an interrupt, have no user context or are running
1130
	 * in a region with pagefaults disabled then we must not take the fault
L
Linus Torvalds 已提交
1131
	 */
1132
	if (unlikely(faulthandler_disabled() || !mm)) {
1133 1134 1135
		bad_area_nosemaphore(regs, error_code, address);
		return;
	}
L
Linus Torvalds 已提交
1136

1137 1138 1139 1140 1141 1142 1143
	/*
	 * 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:
	 */
1144
	if (user_mode(regs)) {
1145 1146 1147 1148 1149 1150 1151 1152 1153 1154
		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);

1155 1156 1157
	if (error_code & PF_WRITE)
		flags |= FAULT_FLAG_WRITE;

I
Ingo Molnar 已提交
1158 1159
	/*
	 * When running in the kernel we expect faults to occur only to
I
Ingo Molnar 已提交
1160 1161 1162 1163 1164 1165 1166
	 * 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.
L
Linus Torvalds 已提交
1167 1168
	 *
	 * As the vast majority of faults will be valid we will only perform
I
Ingo Molnar 已提交
1169 1170 1171 1172
	 * 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:
L
Linus Torvalds 已提交
1173
	 */
1174
	if (unlikely(!down_read_trylock(&mm->mmap_sem))) {
1175
		if ((error_code & PF_USER) == 0 &&
1176 1177 1178 1179
		    !search_exception_tables(regs->ip)) {
			bad_area_nosemaphore(regs, error_code, address);
			return;
		}
1180
retry:
L
Linus Torvalds 已提交
1181
		down_read(&mm->mmap_sem);
1182 1183
	} else {
		/*
I
Ingo Molnar 已提交
1184 1185 1186
		 * The above down_read_trylock() might have succeeded in
		 * which case we'll have missed the might_sleep() from
		 * down_read():
1187 1188
		 */
		might_sleep();
L
Linus Torvalds 已提交
1189 1190 1191
	}

	vma = find_vma(mm, address);
1192 1193 1194 1195 1196
	if (unlikely(!vma)) {
		bad_area(regs, error_code, address);
		return;
	}
	if (likely(vma->vm_start <= address))
L
Linus Torvalds 已提交
1197
		goto good_area;
1198 1199 1200 1201
	if (unlikely(!(vma->vm_flags & VM_GROWSDOWN))) {
		bad_area(regs, error_code, address);
		return;
	}
1202
	if (error_code & PF_USER) {
1203 1204 1205
		/*
		 * Accessing the stack below %sp is always a bug.
		 * The large cushion allows instructions like enter
I
Ingo Molnar 已提交
1206
		 * and pusha to work. ("enter $65535, $31" pushes
1207
		 * 32 pointers and then decrements %sp by 65535.)
1208
		 */
1209 1210 1211 1212
		if (unlikely(address + 65536 + 32 * sizeof(unsigned long) < regs->sp)) {
			bad_area(regs, error_code, address);
			return;
		}
L
Linus Torvalds 已提交
1213
	}
1214 1215 1216 1217 1218 1219 1220 1221 1222
	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..
	 */
L
Linus Torvalds 已提交
1223
good_area:
M
Michel Lespinasse 已提交
1224
	if (unlikely(access_error(error_code, vma))) {
1225 1226
		bad_area_access_error(regs, error_code, address);
		return;
L
Linus Torvalds 已提交
1227 1228 1229 1230 1231
	}

	/*
	 * If for any reason at all we couldn't handle the fault,
	 * make sure we exit gracefully rather than endlessly redo
1232 1233
	 * the fault.  Since we never set FAULT_FLAG_RETRY_NOWAIT, if
	 * we get VM_FAULT_RETRY back, the mmap_sem has been unlocked.
L
Linus Torvalds 已提交
1234
	 */
1235
	fault = handle_mm_fault(mm, vma, address, flags);
1236
	major |= fault & VM_FAULT_MAJOR;
I
Ingo Molnar 已提交
1237

1238
	/*
1239 1240 1241
	 * If we need to retry the mmap_sem has already been released,
	 * and if there is a fatal signal pending there is no guarantee
	 * that we made any progress. Handle this case first.
1242
	 */
1243 1244 1245 1246 1247 1248 1249 1250 1251 1252
	if (unlikely(fault & VM_FAULT_RETRY)) {
		/* Retry at most once */
		if (flags & FAULT_FLAG_ALLOW_RETRY) {
			flags &= ~FAULT_FLAG_ALLOW_RETRY;
			flags |= FAULT_FLAG_TRIED;
			if (!fatal_signal_pending(tsk))
				goto retry;
		}

		/* User mode? Just return to handle the fatal exception */
1253
		if (flags & FAULT_FLAG_USER)
1254 1255 1256 1257
			return;

		/* Not returning to user mode? Handle exceptions or die: */
		no_context(regs, error_code, address, SIGBUS, BUS_ADRERR);
1258
		return;
1259
	}
1260

1261
	up_read(&mm->mmap_sem);
1262 1263 1264
	if (unlikely(fault & VM_FAULT_ERROR)) {
		mm_fault_error(regs, error_code, address, fault);
		return;
1265 1266
	}

1267
	/*
1268 1269
	 * Major/minor page fault accounting. If any of the events
	 * returned VM_FAULT_MAJOR, we account it as a major fault.
1270
	 */
1271 1272 1273 1274 1275 1276
	if (major) {
		tsk->maj_flt++;
		perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs, address);
	} else {
		tsk->min_flt++;
		perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs, address);
1277
	}
1278

1279
	check_v8086_mode(regs, address, tsk);
L
Linus Torvalds 已提交
1280
}
1281
NOKPROBE_SYMBOL(__do_page_fault);
1282

1283
dotraplinkage void notrace
1284 1285
do_page_fault(struct pt_regs *regs, unsigned long error_code)
{
1286
	unsigned long address = read_cr2(); /* Get the faulting address */
1287
	enum ctx_state prev_state;
1288 1289 1290 1291 1292 1293 1294 1295

	/*
	 * We must have this function tagged with __kprobes, notrace and call
	 * read_cr2() before calling anything else. To avoid calling any kind
	 * of tracing machinery before we've observed the CR2 value.
	 *
	 * exception_{enter,exit}() contain all sorts of tracepoints.
	 */
1296 1297

	prev_state = exception_enter();
1298
	__do_page_fault(regs, error_code, address);
1299
	exception_exit(prev_state);
1300
}
1301
NOKPROBE_SYMBOL(do_page_fault);
1302

1303
#ifdef CONFIG_TRACING
1304 1305 1306
static nokprobe_inline void
trace_page_fault_entries(unsigned long address, struct pt_regs *regs,
			 unsigned long error_code)
1307 1308
{
	if (user_mode(regs))
1309
		trace_page_fault_user(address, regs, error_code);
1310
	else
1311
		trace_page_fault_kernel(address, regs, error_code);
1312 1313
}

1314
dotraplinkage void notrace
1315 1316
trace_do_page_fault(struct pt_regs *regs, unsigned long error_code)
{
1317 1318 1319 1320 1321 1322 1323
	/*
	 * The exception_enter and tracepoint processing could
	 * trigger another page faults (user space callchain
	 * reading) and destroy the original cr2 value, so read
	 * the faulting address now.
	 */
	unsigned long address = read_cr2();
1324
	enum ctx_state prev_state;
1325 1326

	prev_state = exception_enter();
1327
	trace_page_fault_entries(address, regs, error_code);
1328
	__do_page_fault(regs, error_code, address);
1329 1330
	exception_exit(prev_state);
}
1331
NOKPROBE_SYMBOL(trace_do_page_fault);
1332
#endif /* CONFIG_TRACING */