fault.c 26.3 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1 2
/*
 *  Copyright (C) 1995  Linus Torvalds
Ingo Molnar's avatar
Ingo Molnar committed
3
 *  Copyright (C) 2001, 2002 Andi Kleen, SuSE Labs.
4
 *  Copyright (C) 2008-2009, Red Hat Inc., Ingo Molnar
Linus Torvalds's avatar
Linus Torvalds committed
5
 */
6 7 8 9 10 11 12
#include <linux/magic.h>		/* STACK_END_MAGIC		*/
#include <linux/sched.h>		/* test_thread_flag(), ...	*/
#include <linux/kdebug.h>		/* oops_begin/end, ...		*/
#include <linux/module.h>		/* search_exception_table	*/
#include <linux/bootmem.h>		/* max_low_pfn			*/
#include <linux/kprobes.h>		/* __kprobes, ...		*/
#include <linux/mmiotrace.h>		/* kmmio_handler, ...		*/
13
#include <linux/perf_counter.h>		/* perf_swcounter_event		*/
Ingo Molnar's avatar
Ingo Molnar committed
14

15 16
#include <asm/traps.h>			/* dotraplinkage, ...		*/
#include <asm/pgalloc.h>		/* pgd_*(), ...			*/
Vegard Nossum's avatar
Vegard Nossum committed
17
#include <asm/kmemcheck.h>		/* kmemcheck_*(), ...		*/
Linus Torvalds's avatar
Linus Torvalds committed
18

19
/*
Ingo Molnar's avatar
Ingo Molnar committed
20 21 22 23 24 25 26
 * 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
27
 */
Ingo Molnar's avatar
Ingo Molnar committed
28 29 30 31 32 33 34 35
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,
};
36

37
/*
38 39
 * Returns 0 if mmiotrace is disabled, or if the fault is not
 * handled by mmiotrace:
40
 */
41
static inline int kmmio_fault(struct pt_regs *regs, unsigned long addr)
42
{
43 44 45 46
	if (unlikely(is_kmmio_active()))
		if (kmmio_handler(regs, addr) == 1)
			return -1;
	return 0;
47 48
}

49
static inline int notify_page_fault(struct pt_regs *regs)
50
{
51 52 53
	int ret = 0;

	/* kprobe_running() needs smp_processor_id() */
54
	if (kprobes_built_in() && !user_mode_vm(regs)) {
55 56 57 58 59
		preempt_disable();
		if (kprobe_running() && kprobe_fault_handler(regs, 14))
			ret = 1;
		preempt_enable();
	}
60

61
	return ret;
62
}
63

64
/*
Ingo Molnar's avatar
Ingo Molnar committed
65 66 67 68 69 70
 * Prefetch quirks:
 *
 * 32-bit mode:
 *
 *   Sometimes AMD Athlon/Opteron CPUs report invalid exceptions on prefetch.
 *   Check that here and ignore it.
71
 *
Ingo Molnar's avatar
Ingo Molnar committed
72
 * 64-bit mode:
73
 *
Ingo Molnar's avatar
Ingo Molnar committed
74 75 76 77
 *   Sometimes the CPU reports invalid exceptions on prefetch.
 *   Check that here and ignore it.
 *
 * Opcode checker based on code by Richard Brunner.
78
 */
79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125
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
126 127
static int
is_prefetch(struct pt_regs *regs, unsigned long error_code, unsigned long addr)
128
{
Ingo Molnar's avatar
Ingo Molnar committed
129
	unsigned char *max_instr;
130
	unsigned char *instr;
131
	int prefetch = 0;
Linus Torvalds's avatar
Linus Torvalds committed
132

Ingo Molnar's avatar
Ingo Molnar committed
133 134 135 136
	/*
	 * If it was a exec (instruction fetch) fault on NX page, then
	 * do not ignore the fault:
	 */
137
	if (error_code & PF_INSTR)
Linus Torvalds's avatar
Linus Torvalds committed
138
		return 0;
139

140
	instr = (void *)convert_ip_to_linear(current, regs);
141
	max_instr = instr + 15;
Linus Torvalds's avatar
Linus Torvalds committed
142

143
	if (user_mode(regs) && instr >= (unsigned char *)TASK_SIZE)
Linus Torvalds's avatar
Linus Torvalds committed
144 145
		return 0;

146
	while (instr < max_instr) {
Ingo Molnar's avatar
Ingo Molnar committed
147
		unsigned char opcode;
Linus Torvalds's avatar
Linus Torvalds committed
148

149
		if (probe_kernel_address(instr, opcode))
150
			break;
Linus Torvalds's avatar
Linus Torvalds committed
151 152 153

		instr++;

154
		if (!check_prefetch_opcode(regs, instr, opcode, &prefetch))
Linus Torvalds's avatar
Linus Torvalds committed
155 156 157 158 159
			break;
	}
	return prefetch;
}

Ingo Molnar's avatar
Ingo Molnar committed
160 161 162
static void
force_sig_info_fault(int si_signo, int si_code, unsigned long address,
		     struct task_struct *tsk)
163 164 165
{
	siginfo_t info;

Ingo Molnar's avatar
Ingo Molnar committed
166 167 168 169 170
	info.si_signo	= si_signo;
	info.si_errno	= 0;
	info.si_code	= si_code;
	info.si_addr	= (void __user *)address;

171 172 173
	force_sig_info(si_signo, &info, tsk);
}

174 175 176 177 178
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)
179
{
180 181 182 183
	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
184

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

206
	if (!pmd_present(*pmd))
207
		set_pmd(pmd, *pmd_k);
208
	else
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
		BUG_ON(pmd_page(*pmd) != pmd_page(*pmd_k));

	return pmd_k;
}

void vmalloc_sync_all(void)
{
	unsigned long address;

	if (SHARED_KERNEL_PMD)
		return;

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

		unsigned long flags;
		struct page *page;

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

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

	/* Make sure we are in vmalloc area: */
	if (!(address >= VMALLOC_START && address < VMALLOC_END))
		return -1;

	/*
	 * Synchronize this task's top level page-table
	 * with the 'reference' page table.
	 *
	 * Do _not_ use "current" here. We might be inside
	 * an interrupt in the middle of a task switch..
	 */
	pgd_paddr = read_cr3();
	pmd_k = vmalloc_sync_one(__va(pgd_paddr), address);
	if (!pmd_k)
		return -1;

	pte_k = pte_offset_kernel(pmd_k, address);
	if (!pte_present(*pte_k))
		return -1;

	return 0;
}

/*
 * Did it hit the DOS screen memory VA from vm86 mode?
 */
static inline void
check_v8086_mode(struct pt_regs *regs, unsigned long address,
		 struct task_struct *tsk)
{
	unsigned long bit;

	if (!v8086_mode(regs))
		return;

	bit = (address - 0xA0000) >> PAGE_SHIFT;
	if (bit < 32)
		tsk->thread.screen_bitmap |= 1 << bit;
286
}
Linus Torvalds's avatar
Linus Torvalds committed
287

Akinobu Mita's avatar
Akinobu Mita committed
288
static bool low_pfn(unsigned long pfn)
Linus Torvalds's avatar
Linus Torvalds committed
289
{
Akinobu Mita's avatar
Akinobu Mita committed
290 291
	return pfn < max_low_pfn;
}
292

Akinobu Mita's avatar
Akinobu Mita committed
293 294 295 296 297 298
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
299

300
#ifdef CONFIG_X86_PAE
Akinobu Mita's avatar
Akinobu Mita committed
301 302 303
	printk("*pdpt = %016Lx ", pgd_val(*pgd));
	if (!low_pfn(pgd_val(*pgd) >> PAGE_SHIFT) || !pgd_present(*pgd))
		goto out;
304
#endif
Akinobu Mita's avatar
Akinobu Mita committed
305 306
	pmd = pmd_offset(pud_offset(pgd, address), address);
	printk(KERN_CONT "*pde = %0*Lx ", sizeof(*pmd) * 2, (u64)pmd_val(*pmd));
307 308 309 310 311

	/*
	 * We must not directly access the pte in the highpte
	 * case if the page table is located in highmem.
	 * And let's rather not kmap-atomic the pte, just in case
Ingo Molnar's avatar
Ingo Molnar committed
312
	 * it's allocated already:
313
	 */
Akinobu Mita's avatar
Akinobu Mita committed
314 315
	if (!low_pfn(pmd_pfn(*pmd)) || !pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
316

Akinobu Mita's avatar
Akinobu Mita committed
317 318 319
	pte = pte_offset_kernel(pmd, address);
	printk("*pte = %0*Lx ", sizeof(*pte) * 2, (u64)pte_val(*pte));
out:
320
	printk("\n");
321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423
}

#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[] =
424 425 426 427 428
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";
429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447

/*
 * 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
448 449
	pgd_t *base = __va(read_cr3() & PHYSICAL_PAGE_MASK);
	pgd_t *pgd = base + pgd_index(address);
Linus Torvalds's avatar
Linus Torvalds committed
450 451 452 453
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;

Ingo Molnar's avatar
Ingo Molnar committed
454 455 456
	if (bad_address(pgd))
		goto bad;

457
	printk("PGD %lx ", pgd_val(*pgd));
Ingo Molnar's avatar
Ingo Molnar committed
458 459 460

	if (!pgd_present(*pgd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
461

462
	pud = pud_offset(pgd, address);
Ingo Molnar's avatar
Ingo Molnar committed
463 464 465
	if (bad_address(pud))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
466
	printk("PUD %lx ", pud_val(*pud));
467
	if (!pud_present(*pud) || pud_large(*pud))
Ingo Molnar's avatar
Ingo Molnar committed
468
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
469 470

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

Linus Torvalds's avatar
Linus Torvalds committed
474
	printk("PMD %lx ", pmd_val(*pmd));
Ingo Molnar's avatar
Ingo Molnar committed
475 476
	if (!pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
477 478

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

482
	printk("PTE %lx", pte_val(*pte));
Ingo Molnar's avatar
Ingo Molnar committed
483
out:
Linus Torvalds's avatar
Linus Torvalds committed
484 485 486 487
	printk("\n");
	return;
bad:
	printk("BAD\n");
488 489
}

490
#endif /* CONFIG_X86_64 */
Linus Torvalds's avatar
Linus Torvalds committed
491

Ingo Molnar's avatar
Ingo Molnar committed
492 493 494 495 496 497 498 499 500 501 502 503 504
/*
 * 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.
505
 */
506
static int is_errata93(struct pt_regs *regs, unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
507
{
508
#ifdef CONFIG_X86_64
509
	if (address != regs->ip)
Linus Torvalds's avatar
Linus Torvalds committed
510
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
511

512
	if ((address >> 32) != 0)
Linus Torvalds's avatar
Linus Torvalds committed
513
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
514

Linus Torvalds's avatar
Linus Torvalds committed
515
	address |= 0xffffffffUL << 32;
516 517
	if ((address >= (u64)_stext && address <= (u64)_etext) ||
	    (address >= MODULES_VADDR && address <= MODULES_END)) {
518
		printk_once(errata93_warning);
519
		regs->ip = address;
Linus Torvalds's avatar
Linus Torvalds committed
520 521
		return 1;
	}
522
#endif
Linus Torvalds's avatar
Linus Torvalds committed
523
	return 0;
524
}
Linus Torvalds's avatar
Linus Torvalds committed
525

526
/*
Ingo Molnar's avatar
Ingo Molnar committed
527 528 529 530 531
 * 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
532 533 534 535 536
 * 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
537
	if ((regs->cs == __USER32_CS || (regs->cs & (1<<2))) && (address >> 32))
538 539 540 541 542
		return 1;
#endif
	return 0;
}

543 544 545 546
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
547

548
	/*
Ingo Molnar's avatar
Ingo Molnar committed
549
	 * Pentium F0 0F C7 C8 bug workaround:
550 551 552 553 554 555 556 557 558 559 560 561 562
	 */
	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;
}

563 564 565
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
566 567 568
static void
show_fault_oops(struct pt_regs *regs, unsigned long error_code,
		unsigned long address)
569
{
570 571 572 573
	if (!oops_may_print())
		return;

	if (error_code & PF_INSTR) {
574
		unsigned int level;
Ingo Molnar's avatar
Ingo Molnar committed
575

576 577
		pte_t *pte = lookup_address(address, &level);

578 579
		if (pte && pte_present(*pte) && !pte_exec(*pte))
			printk(nx_warning, current_uid());
580 581
	}

582
	printk(KERN_ALERT "BUG: unable to handle kernel ");
583
	if (address < PAGE_SIZE)
584
		printk(KERN_CONT "NULL pointer dereference");
585
	else
586
		printk(KERN_CONT "paging request");
Ingo Molnar's avatar
Ingo Molnar committed
587

588
	printk(KERN_CONT " at %p\n", (void *) address);
589
	printk(KERN_ALERT "IP:");
590
	printk_address(regs->ip, 1);
Ingo Molnar's avatar
Ingo Molnar committed
591

592 593 594
	dump_pagetable(address);
}

Ingo Molnar's avatar
Ingo Molnar committed
595 596 597
static noinline void
pgtable_bad(struct pt_regs *regs, unsigned long error_code,
	    unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
598
{
Ingo Molnar's avatar
Ingo Molnar committed
599 600 601 602 603 604 605
	struct task_struct *tsk;
	unsigned long flags;
	int sig;

	flags = oops_begin();
	tsk = current;
	sig = SIGKILL;
606

Linus Torvalds's avatar
Linus Torvalds committed
607
	printk(KERN_ALERT "%s: Corrupted page table at address %lx\n",
608
	       tsk->comm, address);
Linus Torvalds's avatar
Linus Torvalds committed
609
	dump_pagetable(address);
Ingo Molnar's avatar
Ingo Molnar committed
610 611 612 613 614

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

615
	if (__die("Bad pagetable", regs, error_code))
616
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
617

618
	oops_end(flags, regs, sig);
Linus Torvalds's avatar
Linus Torvalds committed
619 620
}

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

Ingo Molnar's avatar
Ingo Molnar committed
630
	/* Are we prepared to handle this kernel fault? */
631 632 633 634
	if (fixup_exception(regs))
		return;

	/*
Ingo Molnar's avatar
Ingo Molnar committed
635 636 637 638 639 640 641
	 * 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:
642
	 *
Ingo Molnar's avatar
Ingo Molnar committed
643
	 *   Hall of shame of CPU/BIOS bugs.
644 645 646 647 648 649 650 651 652
	 */
	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
653
	 * terminate things with extreme prejudice:
654 655 656 657 658
	 */
	flags = oops_begin();

	show_fault_oops(regs, error_code, address);

Ingo Molnar's avatar
Ingo Molnar committed
659
	stackend = end_of_stack(tsk);
660 661 662
	if (*stackend != STACK_END_MAGIC)
		printk(KERN_ALERT "Thread overran stack, or stack corrupted\n");

663 664 665
	tsk->thread.cr2		= address;
	tsk->thread.trap_no	= 14;
	tsk->thread.error_code	= error_code;
666 667 668 669

	sig = SIGKILL;
	if (__die("Oops", regs, error_code))
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
670

671 672
	/* 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
673

674 675 676
	oops_end(flags, regs, sig);
}

Ingo Molnar's avatar
Ingo Molnar committed
677 678 679 680 681 682 683 684 685 686 687 688 689 690
/*
 * 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;

691
	printk("%s%s[%d]: segfault at %lx ip %p sp %p error %lx",
Ingo Molnar's avatar
Ingo Molnar committed
692 693 694 695 696 697 698 699 700 701 702 703
		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)
704 705 706 707 708 709
{
	struct task_struct *tsk = current;

	/* User mode accesses just cause a SIGSEGV */
	if (error_code & PF_USER) {
		/*
Ingo Molnar's avatar
Ingo Molnar committed
710
		 * It's possible to have interrupts off here:
711 712 713 714 715
		 */
		local_irq_enable();

		/*
		 * Valid to do another page fault here because this one came
Ingo Molnar's avatar
Ingo Molnar committed
716
		 * from user space:
717 718 719 720 721 722 723
		 */
		if (is_prefetch(regs, error_code, address))
			return;

		if (is_errata100(regs, address))
			return;

Ingo Molnar's avatar
Ingo Molnar committed
724 725 726 727 728 729 730
		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;
731 732

		force_sig_info_fault(SIGSEGV, si_code, address, tsk);
Ingo Molnar's avatar
Ingo Molnar committed
733

734 735 736 737 738 739 740 741 742
		return;
	}

	if (is_f00f_bug(regs, address))
		return;

	no_context(regs, error_code, address);
}

Ingo Molnar's avatar
Ingo Molnar committed
743 744 745
static noinline void
bad_area_nosemaphore(struct pt_regs *regs, unsigned long error_code,
		     unsigned long address)
746 747 748 749
{
	__bad_area_nosemaphore(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
750 751 752
static void
__bad_area(struct pt_regs *regs, unsigned long error_code,
	   unsigned long address, int si_code)
753 754 755 756 757 758 759 760 761 762 763 764
{
	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
765 766
static noinline void
bad_area(struct pt_regs *regs, unsigned long error_code, unsigned long address)
767 768 769 770
{
	__bad_area(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
771 772 773
static noinline void
bad_area_access_error(struct pt_regs *regs, unsigned long error_code,
		      unsigned long address)
774 775 776 777 778
{
	__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
779 780 781
static void
out_of_memory(struct pt_regs *regs, unsigned long error_code,
	      unsigned long address)
782 783 784
{
	/*
	 * We ran out of memory, call the OOM killer, and return the userspace
Ingo Molnar's avatar
Ingo Molnar committed
785
	 * (which will retry the fault, or kill us if we got oom-killed):
786 787
	 */
	up_read(&current->mm->mmap_sem);
Ingo Molnar's avatar
Ingo Molnar committed
788

789 790 791
	pagefault_out_of_memory();
}

Ingo Molnar's avatar
Ingo Molnar committed
792 793
static void
do_sigbus(struct pt_regs *regs, unsigned long error_code, unsigned long address)
794 795 796 797 798 799
{
	struct task_struct *tsk = current;
	struct mm_struct *mm = tsk->mm;

	up_read(&mm->mmap_sem);

Ingo Molnar's avatar
Ingo Molnar committed
800
	/* Kernel mode? Handle exceptions or die: */
801 802
	if (!(error_code & PF_USER))
		no_context(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
803

804
	/* User-space => ok to do another page fault: */
805 806
	if (is_prefetch(regs, error_code, address))
		return;
Ingo Molnar's avatar
Ingo Molnar committed
807 808 809 810 811

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

812 813 814
	force_sig_info_fault(SIGBUS, BUS_ADRERR, address, tsk);
}

Ingo Molnar's avatar
Ingo Molnar committed
815 816 817
static noinline void
mm_fault_error(struct pt_regs *regs, unsigned long error_code,
	       unsigned long address, unsigned int fault)
818
{
Ingo Molnar's avatar
Ingo Molnar committed
819
	if (fault & VM_FAULT_OOM) {
820
		out_of_memory(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
821 822 823 824 825 826
	} else {
		if (fault & VM_FAULT_SIGBUS)
			do_sigbus(regs, error_code, address);
		else
			BUG();
	}
827 828
}

829 830 831 832
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
833

834 835 836 837 838 839
	if ((error_code & PF_INSTR) && !pte_exec(*pte))
		return 0;

	return 1;
}

840
/*
Ingo Molnar's avatar
Ingo Molnar committed
841 842 843 844 845 846 847 848
 * 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.
 *
849 850 851
 * There are no security implications to leaving a stale TLB when
 * increasing the permissions on a page.
 */
Ingo Molnar's avatar
Ingo Molnar committed
852 853
static noinline int
spurious_fault(unsigned long error_code, unsigned long address)
854 855 856 857 858
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;
859
	int ret;
860 861 862 863 864 865 866 867 868 869 870 871 872

	/* 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;

873 874 875
	if (pud_large(*pud))
		return spurious_fault_check(error_code, (pte_t *) pud);

876 877 878 879
	pmd = pmd_offset(pud, address);
	if (!pmd_present(*pmd))
		return 0;

880 881 882
	if (pmd_large(*pmd))
		return spurious_fault_check(error_code, (pte_t *) pmd);

883 884 885 886
	pte = pte_offset_kernel(pmd, address);
	if (!pte_present(*pte))
		return 0;

887 888 889 890 891
	ret = spurious_fault_check(error_code, pte);
	if (!ret)
		return 0;

	/*
Ingo Molnar's avatar
Ingo Molnar committed
892 893
	 * Make sure we have permissions in PMD.
	 * If not, then there's a bug in the page tables:
894 895 896
	 */
	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
897

898
	return ret;
899 900
}

901
int show_unhandled_signals = 1;
Linus Torvalds's avatar
Linus Torvalds committed
902

Ingo Molnar's avatar
Ingo Molnar committed
903 904
static inline int
access_error(unsigned long error_code, int write, struct vm_area_struct *vma)
905 906
{
	if (write) {
Ingo Molnar's avatar
Ingo Molnar committed
907
		/* write, present and write, not present: */
908 909
		if (unlikely(!(vma->vm_flags & VM_WRITE)))
			return 1;
Ingo Molnar's avatar
Ingo Molnar committed
910
		return 0;
911 912
	}

Ingo Molnar's avatar
Ingo Molnar committed
913 914 915 916 917 918 919 920
	/* 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;

921 922 923
	return 0;
}

924 925
static int fault_in_kernel_space(unsigned long address)
{
926
	return address >= TASK_SIZE_MAX;
927 928
}

Linus Torvalds's avatar
Linus Torvalds committed
929 930 931 932 933
/*
 * This routine handles page faults.  It determines the address,
 * and the problem, and then passes it off to one of the appropriate
 * routines.
 */
934 935
dotraplinkage void __kprobes
do_page_fault(struct pt_regs *regs, unsigned long error_code)
Linus Torvalds's avatar
Linus Torvalds committed
936
{
Ingo Molnar's avatar
Ingo Molnar committed
937
	struct vm_area_struct *vma;
Linus Torvalds's avatar
Linus Torvalds committed
938
	struct task_struct *tsk;
Ingo Molnar's avatar
Ingo Molnar committed
939
	unsigned long address;
Linus Torvalds's avatar
Linus Torvalds committed
940
	struct mm_struct *mm;
941
	int write;
942
	int fault;
Linus Torvalds's avatar
Linus Torvalds committed
943

944 945
	tsk = current;
	mm = tsk->mm;
Ingo Molnar's avatar
Ingo Molnar committed
946 947

	/* Get the faulting address: */
948
	address = read_cr2();
Linus Torvalds's avatar
Linus Torvalds committed
949

Vegard Nossum's avatar
Vegard Nossum committed
950 951 952 953 954 955
	/*
	 * 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);
956
	prefetchw(&mm->mmap_sem);
Vegard Nossum's avatar
Vegard Nossum committed
957

958
	if (unlikely(kmmio_fault(regs, address)))
959
		return;
Linus Torvalds's avatar
Linus Torvalds committed
960 961 962 963 964 965 966 967 968 969 970 971

	/*
	 * 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
972
	 * protection error (error_code & 9) == 0.
Linus Torvalds's avatar
Linus Torvalds committed
973
	 */
974
	if (unlikely(fault_in_kernel_space(address))) {
Vegard Nossum's avatar
Vegard Nossum committed
975 976 977 978 979 980 981
		if (!(error_code & (PF_RSVD | PF_USER | PF_PROT))) {
			if (vmalloc_fault(address) >= 0)
				return;

			if (kmemcheck_fault(regs, address, error_code))
				return;
		}
982

Ingo Molnar's avatar
Ingo Molnar committed
983
		/* Can handle a stale RO->RW TLB: */
984
		if (spurious_fault(error_code, address))
985 986
			return;

Ingo Molnar's avatar
Ingo Molnar committed
987
		/* kprobes don't want to hook the spurious faults: */
988 989
		if (notify_page_fault(regs))
			return;
990 991
		/*
		 * Don't take the mm semaphore here. If we fixup a prefetch
Ingo Molnar's avatar
Ingo Molnar committed
992
		 * fault we could otherwise deadlock:
993
		 */
994
		bad_area_nosemaphore(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
995

996
		return;
997 998
	}

Ingo Molnar's avatar
Ingo Molnar committed
999
	/* kprobes don't want to hook the spurious faults: */
1000
	if (unlikely(notify_page_fault(regs)))
1001
		return;
1002
	/*
1003 1004 1005 1006
	 * 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
1007
	 * potential system fault or CPU buglet:
1008
	 */
1009 1010 1011
	if (user_mode_vm(regs)) {
		local_irq_enable();
		error_code |= PF_USER;
Ingo Molnar's avatar
Ingo Molnar committed
1012 1013 1014 1015
	} else {
		if (regs->flags & X86_EFLAGS_IF)
			local_irq_enable();
	}
1016

1017
	if (unlikely(error_code & PF_RSVD))
1018
		pgtable_bad(regs, error_code, address);
Linus Torvalds's avatar
Linus Torvalds committed
1019

1020
	perf_swcounter_event(PERF_COUNT_SW_PAGE_FAULTS, 1, 0, regs, address);
1021

Linus Torvalds's avatar
Linus Torvalds committed
1022
	/*
Ingo Molnar's avatar
Ingo Molnar committed
1023 1024
	 * 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
1025
	 */
1026 1027 1028 1029
	if (unlikely(in_atomic() || !mm)) {
		bad_area_nosemaphore(regs, error_code, address);
		return;
	}
Linus Torvalds's avatar
Linus Torvalds committed
1030

1031 1032
	/*
	 * When running in the kernel we expect faults to occur only to
Ingo Molnar's avatar
Ingo Molnar committed
1033 1034 1035 1036 1037 1038 1039
	 * 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
1040 1041
	 *
	 * As the vast majority of faults will be valid we will only perform
Ingo Molnar's avatar
Ingo Molnar committed
1042 1043 1044 1045
	 * 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
1046
	 */
1047
	if (unlikely(!down_read_trylock(&mm->mmap_sem))) {
1048
		if ((error_code & PF_USER) == 0 &&
1049 1050 1051 1052
		    !search_exception_tables(regs->ip)) {
			bad_area_nosemaphore(regs, error_code, address);
			return;
		}
Linus Torvalds's avatar
Linus Torvalds committed
1053
		down_read(&mm->mmap_sem);
1054 1055
	} else {
		/*
Ingo Molnar's avatar
Ingo Molnar committed
1056 1057 1058
		 * The above down_read_trylock() might have succeeded in
		 * which case we'll have missed the might_sleep() from
		 * down_read():
1059 1060
		 */
		might_sleep();
Linus Torvalds's avatar
Linus Torvalds committed
1061 1062 1063
	}

	vma = find_vma(mm, address);
1064 1065 1066 1067 1068
	if (unlikely(!vma)) {
		bad_area(regs, error_code, address);
		return;
	}
	if (likely(vma->vm_start <= address))
Linus Torvalds's avatar
Linus Torvalds committed
1069
		goto good_area;
1070 1071 1072 1073
	if (unlikely(!(vma->vm_flags & VM_GROWSDOWN))) {
		bad_area(regs, error_code, address);
		return;
	}
1074
	if (error_code & PF_USER) {
1075 1076 1077
		/*
		 * Accessing the stack below %sp is always a bug.
		 * The large cushion allows instructions like enter
Ingo Molnar's avatar
Ingo Molnar committed
1078
		 * and pusha to work. ("enter $65535, $31" pushes
1079
		 * 32 pointers and then decrements %sp by 65535.)
1080
		 */
1081 1082 1083 1084
		if (unlikely(address + 65536 + 32 * sizeof(unsigned long) < regs->sp)) {
			bad_area(regs, error_code, address);
			return;
		}
Linus Torvalds's avatar
Linus Torvalds committed
1085
	}
1086 1087 1088 1089 1090 1091 1092 1093 1094
	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
1095
good_area: