fault.c 26.2 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1 2
/*
 *  Copyright (C) 1995  Linus Torvalds
Ingo Molnar's avatar
Ingo Molnar committed
3
 *  Copyright (C) 2001, 2002 Andi Kleen, SuSE Labs.
Linus Torvalds's avatar
Linus Torvalds committed
4 5
 */
#include <linux/interrupt.h>
Ingo Molnar's avatar
Ingo Molnar committed
6 7
#include <linux/mmiotrace.h>
#include <linux/bootmem.h>
Linus Torvalds's avatar
Linus Torvalds committed
8
#include <linux/compiler.h>
Harvey Harrison's avatar
Harvey Harrison committed
9
#include <linux/highmem.h>
10
#include <linux/kprobes.h>
11
#include <linux/uaccess.h>
Ingo Molnar's avatar
Ingo Molnar committed
12 13 14 15 16 17 18
#include <linux/vmalloc.h>
#include <linux/vt_kern.h>
#include <linux/signal.h>
#include <linux/kernel.h>
#include <linux/ptrace.h>
#include <linux/string.h>
#include <linux/module.h>
19
#include <linux/kdebug.h>
Ingo Molnar's avatar
Ingo Molnar committed
20
#include <linux/errno.h>
21
#include <linux/magic.h>
Ingo Molnar's avatar
Ingo Molnar committed
22 23 24 25 26 27 28 29 30
#include <linux/sched.h>
#include <linux/types.h>
#include <linux/init.h>
#include <linux/mman.h>
#include <linux/tty.h>
#include <linux/smp.h>
#include <linux/mm.h>

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

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

40
/*
Ingo Molnar's avatar
Ingo Molnar committed
41 42 43 44 45 46 47
 * Page fault error code bits:
 *
 *   bit 0 ==	 0: no page found	1: protection fault
 *   bit 1 ==	 0: read access		1: write access
 *   bit 2 ==	 0: kernel-mode access	1: user-mode access
 *   bit 3 ==				1: use of reserved bit detected
 *   bit 4 ==				1: fault was an instruction fetch
48
 */
Ingo Molnar's avatar
Ingo Molnar committed
49 50 51 52 53 54 55 56
enum x86_pf_error_code {

	PF_PROT		=		1 << 0,
	PF_WRITE	=		1 << 1,
	PF_USER		=		1 << 2,
	PF_RSVD		=		1 << 3,
	PF_INSTR	=		1 << 4,
};
57

58
static inline int kmmio_fault(struct pt_regs *regs, unsigned long addr)
59
{
60
#ifdef CONFIG_MMIOTRACE
61 62 63
	if (unlikely(is_kmmio_active()))
		if (kmmio_handler(regs, addr) == 1)
			return -1;
64
#endif
65
	return 0;
66 67
}

68
static inline int notify_page_fault(struct pt_regs *regs)
69
{
70
#ifdef CONFIG_KPROBES
71 72 73
	int ret = 0;

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

81 82 83 84
	return ret;
#else
	return 0;
#endif
85
}
86

87
/*
Ingo Molnar's avatar
Ingo Molnar committed
88 89 90 91 92 93
 * Prefetch quirks:
 *
 * 32-bit mode:
 *
 *   Sometimes AMD Athlon/Opteron CPUs report invalid exceptions on prefetch.
 *   Check that here and ignore it.
94
 *
Ingo Molnar's avatar
Ingo Molnar committed
95
 * 64-bit mode:
96
 *
Ingo Molnar's avatar
Ingo Molnar committed
97 98 99 100
 *   Sometimes the CPU reports invalid exceptions on prefetch.
 *   Check that here and ignore it.
 *
 * Opcode checker based on code by Richard Brunner.
101
 */
Ingo Molnar's avatar
Ingo Molnar committed
102 103
static int
is_prefetch(struct pt_regs *regs, unsigned long error_code, unsigned long addr)
104
{
Ingo Molnar's avatar
Ingo Molnar committed
105
	unsigned char *max_instr;
106
	unsigned char *instr;
Linus Torvalds's avatar
Linus Torvalds committed
107
	int scan_more = 1;
108
	int prefetch = 0;
Linus Torvalds's avatar
Linus Torvalds committed
109

Ingo Molnar's avatar
Ingo Molnar committed
110 111 112 113
	/*
	 * If it was a exec (instruction fetch) fault on NX page, then
	 * do not ignore the fault:
	 */
114
	if (error_code & PF_INSTR)
Linus Torvalds's avatar
Linus Torvalds committed
115
		return 0;
116

117
	instr = (unsigned char *)convert_ip_to_linear(current, regs);
118
	max_instr = instr + 15;
Linus Torvalds's avatar
Linus Torvalds committed
119

120
	if (user_mode(regs) && instr >= (unsigned char *)TASK_SIZE)
Linus Torvalds's avatar
Linus Torvalds committed
121 122
		return 0;

123
	while (scan_more && instr < max_instr) {
Linus Torvalds's avatar
Linus Torvalds committed
124 125
		unsigned char instr_hi;
		unsigned char instr_lo;
Ingo Molnar's avatar
Ingo Molnar committed
126
		unsigned char opcode;
Linus Torvalds's avatar
Linus Torvalds committed
127

128
		if (probe_kernel_address(instr, opcode))
129
			break;
Linus Torvalds's avatar
Linus Torvalds committed
130

131 132
		instr_hi = opcode & 0xf0;
		instr_lo = opcode & 0x0f;
Linus Torvalds's avatar
Linus Torvalds committed
133 134
		instr++;

135
		switch (instr_hi) {
Linus Torvalds's avatar
Linus Torvalds committed
136 137
		case 0x20:
		case 0x30:
138 139 140 141 142 143
			/*
			 * 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
			 */
Linus Torvalds's avatar
Linus Torvalds committed
144 145
			scan_more = ((instr_lo & 7) == 0x6);
			break;
146
#ifdef CONFIG_X86_64
Linus Torvalds's avatar
Linus Torvalds committed
147
		case 0x40:
148 149 150 151 152 153 154
			/*
			 * 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.
			 */
155
			scan_more = (!user_mode(regs)) || (regs->cs == __USER_CS);
Linus Torvalds's avatar
Linus Torvalds committed
156
			break;
157
#endif
Linus Torvalds's avatar
Linus Torvalds committed
158 159 160
		case 0x60:
			/* 0x64 thru 0x67 are valid prefixes in all modes. */
			scan_more = (instr_lo & 0xC) == 0x4;
161
			break;
Linus Torvalds's avatar
Linus Torvalds committed
162
		case 0xF0:
163
			/* 0xF0, 0xF2, 0xF3 are valid prefixes in all modes. */
Linus Torvalds's avatar
Linus Torvalds committed
164
			scan_more = !instr_lo || (instr_lo>>1) == 1;
165
			break;
Linus Torvalds's avatar
Linus Torvalds committed
166 167 168
		case 0x00:
			/* Prefetch instruction is 0x0F0D or 0x0F18 */
			scan_more = 0;
169

170
			if (probe_kernel_address(instr, opcode))
Linus Torvalds's avatar
Linus Torvalds committed
171 172 173
				break;
			prefetch = (instr_lo == 0xF) &&
				(opcode == 0x0D || opcode == 0x18);
174
			break;
Linus Torvalds's avatar
Linus Torvalds committed
175 176 177
		default:
			scan_more = 0;
			break;
178
		}
Linus Torvalds's avatar
Linus Torvalds committed
179 180 181 182
	}
	return prefetch;
}

Ingo Molnar's avatar
Ingo Molnar committed
183 184 185
static void
force_sig_info_fault(int si_signo, int si_code, unsigned long address,
		     struct task_struct *tsk)
186 187 188
{
	siginfo_t info;

Ingo Molnar's avatar
Ingo Molnar committed
189 190 191 192 193
	info.si_signo	= si_signo;
	info.si_errno	= 0;
	info.si_code	= si_code;
	info.si_addr	= (void __user *)address;

194 195 196
	force_sig_info(si_signo, &info, tsk);
}

197
#ifdef CONFIG_X86_64
198 199
static int bad_address(void *p)
{
Linus Torvalds's avatar
Linus Torvalds committed
200
	unsigned long dummy;
Ingo Molnar's avatar
Ingo Molnar committed
201

202
	return probe_kernel_address((unsigned long *)p, dummy);
203
}
204
#endif
Linus Torvalds's avatar
Linus Torvalds committed
205

206
static void dump_pagetable(unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
207
{
208 209 210 211 212
#ifdef CONFIG_X86_32
	__typeof__(pte_val(__pte(0))) page;

	page = read_cr3();
	page = ((__typeof__(page) *) __va(page))[address >> PGDIR_SHIFT];
Ingo Molnar's avatar
Ingo Molnar committed
213

214 215 216 217 218 219
#ifdef CONFIG_X86_PAE
	printk("*pdpt = %016Lx ", page);
	if ((page >> PAGE_SHIFT) < max_low_pfn
	    && page & _PAGE_PRESENT) {
		page &= PAGE_MASK;
		page = ((__typeof__(page) *) __va(page))[(address >> PMD_SHIFT)
Ingo Molnar's avatar
Ingo Molnar committed
220
							& (PTRS_PER_PMD - 1)];
221 222 223 224 225 226 227 228 229 230 231
		printk(KERN_CONT "*pde = %016Lx ", page);
		page &= ~_PAGE_NX;
	}
#else
	printk("*pde = %08lx ", page);
#endif

	/*
	 * We must not directly access the pte in the highpte
	 * case if the page table is located in highmem.
	 * And let's rather not kmap-atomic the pte, just in case
Ingo Molnar's avatar
Ingo Molnar committed
232
	 * it's allocated already:
233 234 235 236
	 */
	if ((page >> PAGE_SHIFT) < max_low_pfn
	    && (page & _PAGE_PRESENT)
	    && !(page & _PAGE_PSE)) {
Ingo Molnar's avatar
Ingo Molnar committed
237

238 239
		page &= PAGE_MASK;
		page = ((__typeof__(page) *) __va(page))[(address >> PAGE_SHIFT)
Ingo Molnar's avatar
Ingo Molnar committed
240
							& (PTRS_PER_PTE - 1)];
241 242 243 244 245
		printk("*pte = %0*Lx ", sizeof(page)*2, (u64)page);
	}

	printk("\n");
#else /* CONFIG_X86_64 */
Linus Torvalds's avatar
Linus Torvalds committed
246 247 248 249 250
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;

251
	pgd = (pgd_t *)read_cr3();
Linus Torvalds's avatar
Linus Torvalds committed
252

253
	pgd = __va((unsigned long)pgd & PHYSICAL_PAGE_MASK);
Ingo Molnar's avatar
Ingo Molnar committed
254

Linus Torvalds's avatar
Linus Torvalds committed
255
	pgd += pgd_index(address);
Ingo Molnar's avatar
Ingo Molnar committed
256 257 258
	if (bad_address(pgd))
		goto bad;

259
	printk("PGD %lx ", pgd_val(*pgd));
Ingo Molnar's avatar
Ingo Molnar committed
260 261 262

	if (!pgd_present(*pgd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
263

264
	pud = pud_offset(pgd, address);
Ingo Molnar's avatar
Ingo Molnar committed
265 266 267
	if (bad_address(pud))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
268
	printk("PUD %lx ", pud_val(*pud));
269
	if (!pud_present(*pud) || pud_large(*pud))
Ingo Molnar's avatar
Ingo Molnar committed
270
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
271 272

	pmd = pmd_offset(pud, address);
Ingo Molnar's avatar
Ingo Molnar committed
273 274 275
	if (bad_address(pmd))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
276
	printk("PMD %lx ", pmd_val(*pmd));
Ingo Molnar's avatar
Ingo Molnar committed
277 278
	if (!pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
279 280

	pte = pte_offset_kernel(pmd, address);
Ingo Molnar's avatar
Ingo Molnar committed
281 282 283
	if (bad_address(pte))
		goto bad;

284
	printk("PTE %lx", pte_val(*pte));
Ingo Molnar's avatar
Ingo Molnar committed
285
out:
Linus Torvalds's avatar
Linus Torvalds committed
286 287 288 289
	printk("\n");
	return;
bad:
	printk("BAD\n");
290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320
#endif
}

#ifdef CONFIG_X86_32
static inline pmd_t *vmalloc_sync_one(pgd_t *pgd, unsigned long address)
{
	unsigned index = pgd_index(address);
	pgd_t *pgd_k;
	pud_t *pud, *pud_k;
	pmd_t *pmd, *pmd_k;

	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;
Ingo Molnar's avatar
Ingo Molnar committed
321

322 323 324
	if (!pmd_present(*pmd)) {
		set_pmd(pmd, *pmd_k);
		arch_flush_lazy_mmu_mode();
Ingo Molnar's avatar
Ingo Molnar committed
325
	} else {
326
		BUG_ON(pmd_page(*pmd) != pmd_page(*pmd_k));
Ingo Molnar's avatar
Ingo Molnar committed
327 328
	}

329
	return pmd_k;
Linus Torvalds's avatar
Linus Torvalds committed
330
}
331
#endif
Linus Torvalds's avatar
Linus Torvalds committed
332

333
#ifdef CONFIG_X86_64
334
static const char errata93_warning[] =
Linus Torvalds's avatar
Linus Torvalds committed
335 336 337 338
KERN_ERR "******* Your BIOS seems to not contain a fix for K8 errata #93\n"
KERN_ERR "******* Working around it, but it may cause SEGVs or burn power.\n"
KERN_ERR "******* Please consider a BIOS update.\n"
KERN_ERR "******* Disabling USB legacy in the BIOS may also help.\n";
339
#endif
Linus Torvalds's avatar
Linus Torvalds committed
340

Ingo Molnar's avatar
Ingo Molnar committed
341 342 343 344 345 346 347 348 349 350 351 352 353
/*
 * 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.
354
 */
355
static int is_errata93(struct pt_regs *regs, unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
356
{
357
#ifdef CONFIG_X86_64
Ingo Molnar's avatar
Ingo Molnar committed
358 359
	static int once;

360
	if (address != regs->ip)
Linus Torvalds's avatar
Linus Torvalds committed
361
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
362

363
	if ((address >> 32) != 0)
Linus Torvalds's avatar
Linus Torvalds committed
364
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
365

Linus Torvalds's avatar
Linus Torvalds committed
366
	address |= 0xffffffffUL << 32;
367 368
	if ((address >= (u64)_stext && address <= (u64)_etext) ||
	    (address >= MODULES_VADDR && address <= MODULES_END)) {
Ingo Molnar's avatar
Ingo Molnar committed
369
		if (!once) {
370
			printk(errata93_warning);
Ingo Molnar's avatar
Ingo Molnar committed
371
			once = 1;
Linus Torvalds's avatar
Linus Torvalds committed
372
		}
373
		regs->ip = address;
Linus Torvalds's avatar
Linus Torvalds committed
374 375
		return 1;
	}
376
#endif
Linus Torvalds's avatar
Linus Torvalds committed
377
	return 0;
378
}
Linus Torvalds's avatar
Linus Torvalds committed
379

380
/*
Ingo Molnar's avatar
Ingo Molnar committed
381 382 383 384 385
 * 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
386 387 388 389 390
 * 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
391
	if ((regs->cs == __USER32_CS || (regs->cs & (1<<2))) && (address >> 32))
392 393 394 395 396
		return 1;
#endif
	return 0;
}

397 398 399 400
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
401

402
	/*
Ingo Molnar's avatar
Ingo Molnar committed
403
	 * Pentium F0 0F C7 C8 bug workaround:
404 405 406 407 408 409 410 411 412 413 414 415 416
	 */
	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;
}

Ingo Molnar's avatar
Ingo Molnar committed
417 418 419
static void
show_fault_oops(struct pt_regs *regs, unsigned long error_code,
		unsigned long address)
420
{
421 422 423
#ifdef CONFIG_X86_32
	if (!oops_may_print())
		return;
424
#endif
425 426 427

#ifdef CONFIG_X86_PAE
	if (error_code & PF_INSTR) {
428
		unsigned int level;
Ingo Molnar's avatar
Ingo Molnar committed
429

430 431
		pte_t *pte = lookup_address(address, &level);

Ingo Molnar's avatar
Ingo Molnar committed
432
		if (pte && pte_present(*pte) && !pte_exec(*pte)) {
433 434
			printk(KERN_CRIT "kernel tried to execute "
				"NX-protected page - exploit attempt? "
435
				"(uid: %d)\n", current_uid());
Ingo Molnar's avatar
Ingo Molnar committed
436
		}
437 438 439
	}
#endif

440
	printk(KERN_ALERT "BUG: unable to handle kernel ");
441
	if (address < PAGE_SIZE)
442
		printk(KERN_CONT "NULL pointer dereference");
443
	else
444
		printk(KERN_CONT "paging request");
Ingo Molnar's avatar
Ingo Molnar committed
445

446
	printk(KERN_CONT " at %p\n", (void *) address);
447
	printk(KERN_ALERT "IP:");
448
	printk_address(regs->ip, 1);
Ingo Molnar's avatar
Ingo Molnar committed
449

450 451 452
	dump_pagetable(address);
}

453
#ifdef CONFIG_X86_64
Ingo Molnar's avatar
Ingo Molnar committed
454 455 456
static noinline void
pgtable_bad(struct pt_regs *regs, unsigned long error_code,
	    unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
457
{
Ingo Molnar's avatar
Ingo Molnar committed
458 459 460 461 462 463 464
	struct task_struct *tsk;
	unsigned long flags;
	int sig;

	flags = oops_begin();
	tsk = current;
	sig = SIGKILL;
465

Linus Torvalds's avatar
Linus Torvalds committed
466
	printk(KERN_ALERT "%s: Corrupted page table at address %lx\n",
467
	       tsk->comm, address);
Linus Torvalds's avatar
Linus Torvalds committed
468
	dump_pagetable(address);
Ingo Molnar's avatar
Ingo Molnar committed
469 470 471 472 473

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

474
	if (__die("Bad pagetable", regs, error_code))
475
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
476

477
	oops_end(flags, regs, sig);
Linus Torvalds's avatar
Linus Torvalds committed
478
}
479
#endif
Linus Torvalds's avatar
Linus Torvalds committed
480

Ingo Molnar's avatar
Ingo Molnar committed
481 482 483
static noinline void
no_context(struct pt_regs *regs, unsigned long error_code,
	   unsigned long address)
484 485
{
	struct task_struct *tsk = current;
486 487
	unsigned long *stackend;

488 489 490 491 492
#ifdef CONFIG_X86_64
	unsigned long flags;
	int sig;
#endif

Ingo Molnar's avatar
Ingo Molnar committed
493
	/* Are we prepared to handle this kernel fault? */
494 495 496 497
	if (fixup_exception(regs))
		return;

	/*
Ingo Molnar's avatar
Ingo Molnar committed
498 499 500 501 502 503 504
	 * 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:
505
	 *
Ingo Molnar's avatar
Ingo Molnar committed
506
	 *   Hall of shame of CPU/BIOS bugs.
507 508 509 510 511 512 513 514 515
	 */
	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
516
	 * terminate things with extreme prejudice:
517 518 519 520 521 522 523 524 525
	 */
#ifdef CONFIG_X86_32
	bust_spinlocks(1);
#else
	flags = oops_begin();
#endif

	show_fault_oops(regs, error_code, address);

Ingo Molnar's avatar
Ingo Molnar committed
526
	stackend = end_of_stack(tsk);
527 528 529
	if (*stackend != STACK_END_MAGIC)
		printk(KERN_ALERT "Thread overran stack, or stack corrupted\n");

530 531 532 533 534 535 536 537 538 539 540 541
	tsk->thread.cr2 = address;
	tsk->thread.trap_no = 14;
	tsk->thread.error_code = error_code;

#ifdef CONFIG_X86_32
	die("Oops", regs, error_code);
	bust_spinlocks(0);
	do_exit(SIGKILL);
#else
	sig = SIGKILL;
	if (__die("Oops", regs, error_code))
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
542

543 544
	/* 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
545

546 547 548 549
	oops_end(flags, regs, sig);
#endif
}

Ingo Molnar's avatar
Ingo Molnar committed
550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576
/*
 * Print out info about fatal segfaults, if the show_unhandled_signals
 * sysctl is set:
 */
static inline void
show_signal_msg(struct pt_regs *regs, unsigned long error_code,
		unsigned long address, struct task_struct *tsk)
{
	if (!unhandled_signal(tsk, SIGSEGV))
		return;

	if (!printk_ratelimit())
		return;

	printk(KERN_CONT "%s%s[%d]: segfault at %lx ip %p sp %p error %lx",
		task_pid_nr(tsk) > 1 ? KERN_INFO : KERN_EMERG,
		tsk->comm, task_pid_nr(tsk), address,
		(void *)regs->ip, (void *)regs->sp, error_code);

	print_vma_addr(KERN_CONT " in ", regs->ip);

	printk(KERN_CONT "\n");
}

static void
__bad_area_nosemaphore(struct pt_regs *regs, unsigned long error_code,
		       unsigned long address, int si_code)
577 578 579 580 581 582
{
	struct task_struct *tsk = current;

	/* User mode accesses just cause a SIGSEGV */
	if (error_code & PF_USER) {
		/*
Ingo Molnar's avatar
Ingo Molnar committed
583
		 * It's possible to have interrupts off here:
584 585 586 587 588
		 */
		local_irq_enable();

		/*
		 * Valid to do another page fault here because this one came
Ingo Molnar's avatar
Ingo Molnar committed
589
		 * from user space:
590 591 592 593 594 595 596
		 */
		if (is_prefetch(regs, error_code, address))
			return;

		if (is_errata100(regs, address))
			return;

Ingo Molnar's avatar
Ingo Molnar committed
597 598 599 600 601 602 603
		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;
604 605

		force_sig_info_fault(SIGSEGV, si_code, address, tsk);
Ingo Molnar's avatar
Ingo Molnar committed
606

607 608 609 610 611 612 613 614 615
		return;
	}

	if (is_f00f_bug(regs, address))
		return;

	no_context(regs, error_code, address);
}

Ingo Molnar's avatar
Ingo Molnar committed
616 617 618
static noinline void
bad_area_nosemaphore(struct pt_regs *regs, unsigned long error_code,
		     unsigned long address)
619 620 621 622
{
	__bad_area_nosemaphore(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
623 624 625
static void
__bad_area(struct pt_regs *regs, unsigned long error_code,
	   unsigned long address, int si_code)
626 627 628 629 630 631 632 633 634 635 636 637
{
	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
638 639
static noinline void
bad_area(struct pt_regs *regs, unsigned long error_code, unsigned long address)
640 641 642 643
{
	__bad_area(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
644 645 646
static noinline void
bad_area_access_error(struct pt_regs *regs, unsigned long error_code,
		      unsigned long address)
647 648 649 650 651
{
	__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
652 653 654
static void
out_of_memory(struct pt_regs *regs, unsigned long error_code,
	      unsigned long address)
655 656 657
{
	/*
	 * We ran out of memory, call the OOM killer, and return the userspace
Ingo Molnar's avatar
Ingo Molnar committed
658
	 * (which will retry the fault, or kill us if we got oom-killed):
659 660
	 */
	up_read(&current->mm->mmap_sem);
Ingo Molnar's avatar
Ingo Molnar committed
661

662 663 664
	pagefault_out_of_memory();
}

Ingo Molnar's avatar
Ingo Molnar committed
665 666
static void
do_sigbus(struct pt_regs *regs, unsigned long error_code, unsigned long address)
667 668 669 670 671 672
{
	struct task_struct *tsk = current;
	struct mm_struct *mm = tsk->mm;

	up_read(&mm->mmap_sem);

Ingo Molnar's avatar
Ingo Molnar committed
673
	/* Kernel mode? Handle exceptions or die: */
674 675
	if (!(error_code & PF_USER))
		no_context(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
676

677
#ifdef CONFIG_X86_32
Ingo Molnar's avatar
Ingo Molnar committed
678
	/* User space => ok to do another page fault: */
679 680 681
	if (is_prefetch(regs, error_code, address))
		return;
#endif
Ingo Molnar's avatar
Ingo Molnar committed
682 683 684 685 686

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

687 688 689
	force_sig_info_fault(SIGBUS, BUS_ADRERR, address, tsk);
}

Ingo Molnar's avatar
Ingo Molnar committed
690 691 692
static noinline void
mm_fault_error(struct pt_regs *regs, unsigned long error_code,
	       unsigned long address, unsigned int fault)
693
{
Ingo Molnar's avatar
Ingo Molnar committed
694
	if (fault & VM_FAULT_OOM) {
695
		out_of_memory(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
696 697 698 699 700 701
	} else {
		if (fault & VM_FAULT_SIGBUS)
			do_sigbus(regs, error_code, address);
		else
			BUG();
	}
702 703
}

704 705 706 707
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
708

709 710 711 712 713 714
	if ((error_code & PF_INSTR) && !pte_exec(*pte))
		return 0;

	return 1;
}

715
/*
Ingo Molnar's avatar
Ingo Molnar committed
716 717 718 719 720 721 722 723
 * 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.
 *
724 725 726
 * There are no security implications to leaving a stale TLB when
 * increasing the permissions on a page.
 */
Ingo Molnar's avatar
Ingo Molnar committed
727 728
static noinline int
spurious_fault(unsigned long error_code, unsigned long address)
729 730 731 732 733
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;
734
	int ret;
735 736 737 738 739 740 741 742 743 744 745 746 747

	/* 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;

748 749 750
	if (pud_large(*pud))
		return spurious_fault_check(error_code, (pte_t *) pud);

751 752 753 754
	pmd = pmd_offset(pud, address);
	if (!pmd_present(*pmd))
		return 0;

755 756 757
	if (pmd_large(*pmd))
		return spurious_fault_check(error_code, (pte_t *) pmd);

758 759 760 761
	pte = pte_offset_kernel(pmd, address);
	if (!pte_present(*pte))
		return 0;

762 763 764 765 766
	ret = spurious_fault_check(error_code, pte);
	if (!ret)
		return 0;

	/*
Ingo Molnar's avatar
Ingo Molnar committed
767 768
	 * Make sure we have permissions in PMD.
	 * If not, then there's a bug in the page tables:
769 770 771
	 */
	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
772

773
	return ret;
774 775
}

Linus Torvalds's avatar
Linus Torvalds committed
776
/*
Ingo Molnar's avatar
Ingo Molnar committed
777 778 779
 * 32-bit:
 *
 *   Handle a fault on the vmalloc or module mapping area
780
 *
Ingo Molnar's avatar
Ingo Molnar committed
781 782 783
 * 64-bit:
 *
 *   Handle a fault on the vmalloc area
784 785
 *
 * This assumes no large pages in there.
Linus Torvalds's avatar
Linus Torvalds committed
786
 */
787
static noinline int vmalloc_fault(unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
788
{
789 790 791 792
#ifdef CONFIG_X86_32
	unsigned long pgd_paddr;
	pmd_t *pmd_k;
	pte_t *pte_k;
793

Ingo Molnar's avatar
Ingo Molnar committed
794
	/* Make sure we are in vmalloc area: */
795 796 797
	if (!(address >= VMALLOC_START && address < VMALLOC_END))
		return -1;

798 799 800 801 802 803 804 805 806 807 808
	/*
	 * 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;
Ingo Molnar's avatar
Ingo Molnar committed
809

810 811 812
	pte_k = pte_offset_kernel(pmd_k, address);
	if (!pte_present(*pte_k))
		return -1;
Ingo Molnar's avatar
Ingo Molnar committed
813

814 815
	return 0;
#else
Linus Torvalds's avatar
Linus Torvalds committed
816 817 818 819 820
	pgd_t *pgd, *pgd_ref;
	pud_t *pud, *pud_ref;
	pmd_t *pmd, *pmd_ref;
	pte_t *pte, *pte_ref;

Ingo Molnar's avatar
Ingo Molnar committed
821
	/* Make sure we are in vmalloc area: */
822 823 824
	if (!(address >= VMALLOC_START && address < VMALLOC_END))
		return -1;

Ingo Molnar's avatar
Ingo Molnar committed
825 826 827 828 829
	/*
	 * Copy kernel mappings over when needed. This can also
	 * happen within a race in page table update. In the later
	 * case just flush:
	 */
830
	pgd = pgd_offset(current->active_mm, address);
Linus Torvalds's avatar
Linus Torvalds committed
831 832 833
	pgd_ref = pgd_offset_k(address);
	if (pgd_none(*pgd_ref))
		return -1;
Ingo Molnar's avatar
Ingo Molnar committed
834

Linus Torvalds's avatar
Linus Torvalds committed
835 836
	if (pgd_none(*pgd))
		set_pgd(pgd, *pgd_ref);
837
	else
838
		BUG_ON(pgd_page_vaddr(*pgd) != pgd_page_vaddr(*pgd_ref));
Linus Torvalds's avatar
Linus Torvalds committed
839

Ingo Molnar's avatar
Ingo Molnar committed
840 841 842 843
	/*
	 * Below here mismatches are bugs because these lower tables
	 * are shared:
	 */
Linus Torvalds's avatar
Linus Torvalds committed
844 845 846 847 848

	pud = pud_offset(pgd, address);
	pud_ref = pud_offset(pgd_ref, address);
	if (pud_none(*pud_ref))
		return -1;
Ingo Molnar's avatar
Ingo Molnar committed
849

850
	if (pud_none(*pud) || pud_page_vaddr(*pud) != pud_page_vaddr(*pud_ref))
Linus Torvalds's avatar
Linus Torvalds committed
851
		BUG();
Ingo Molnar's avatar
Ingo Molnar committed
852

Linus Torvalds's avatar
Linus Torvalds committed
853 854 855 856
	pmd = pmd_offset(pud, address);
	pmd_ref = pmd_offset(pud_ref, address);
	if (pmd_none(*pmd_ref))
		return -1;
Ingo Molnar's avatar
Ingo Molnar committed
857

Linus Torvalds's avatar
Linus Torvalds committed
858 859
	if (pmd_none(*pmd) || pmd_page(*pmd) != pmd_page(*pmd_ref))
		BUG();
Ingo Molnar's avatar
Ingo Molnar committed
860

Linus Torvalds's avatar
Linus Torvalds committed
861 862 863
	pte_ref = pte_offset_kernel(pmd_ref, address);
	if (!pte_present(*pte_ref))
		return -1;
Ingo Molnar's avatar
Ingo Molnar committed
864

Linus Torvalds's avatar
Linus Torvalds committed
865
	pte = pte_offset_kernel(pmd, address);
Ingo Molnar's avatar
Ingo Molnar committed
866 867 868 869 870 871

	/*
	 * Don't use pte_page here, because the mappings can point
	 * outside mem_map, and the NUMA hash lookup cannot handle
	 * that:
	 */
872
	if (!pte_present(*pte) || pte_pfn(*pte) != pte_pfn(*pte_ref))
Linus Torvalds's avatar
Linus Torvalds committed
873
		BUG();
Ingo Molnar's avatar
Ingo Molnar committed
874

Linus Torvalds's avatar
Linus Torvalds committed
875
	return 0;
876
#endif
Linus Torvalds's avatar
Linus Torvalds committed
877 878
}

879
int show_unhandled_signals = 1;
Linus Torvalds's avatar
Linus Torvalds committed
880

Ingo Molnar's avatar
Ingo Molnar committed
881 882
static inline int
access_error(unsigned long error_code, int write, struct vm_area_struct *vma)
883 884
{
	if (write) {
Ingo Molnar's avatar
Ingo Molnar committed
885
		/* write, present and write, not present: */
886 887
		if (unlikely(!(vma->vm_flags & VM_WRITE)))
			return 1;
Ingo Molnar's avatar
Ingo Molnar committed
888
		return 0;
889 890
	}

Ingo Molnar's avatar
Ingo Molnar committed
891 892 893 894 895 896 897 898
	/* 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;

899 900 901
	return 0;
}

902 903 904 905
static int fault_in_kernel_space(unsigned long address)
{
#ifdef CONFIG_X86_32
	return address >= TASK_SIZE;
Ingo Molnar's avatar
Ingo Molnar committed
906
#else
907
	return address >= TASK_SIZE64;
Ingo Molnar's avatar
Ingo Molnar committed
908
#endif
909 910
}

Linus Torvalds's avatar
Linus Torvalds committed
911 912 913 914 915
/*
 * This routine handles page faults.  It determines the address,
 * and the problem, and then passes it off to one of the appropriate
 * routines.
 */
916 917 918 919
#ifdef CONFIG_X86_64
asmlinkage
#endif
void __kprobes do_page_fault(struct pt_regs *regs, unsigned long error_code)
Linus Torvalds's avatar
Linus Torvalds committed
920
{
Ingo Molnar's avatar
Ingo Molnar committed
921
	struct vm_area_struct *vma;
Linus Torvalds's avatar
Linus Torvalds committed
922
	struct task_struct *tsk;
Ingo Molnar's avatar
Ingo Molnar committed
923
	unsigned long address;
Linus Torvalds's avatar
Linus Torvalds committed
924
	struct mm_struct *mm;
925
	int write;
926
	int fault;
Linus Torvalds's avatar
Linus Torvalds committed
927

928 929
	tsk = current;
	mm = tsk->mm;
Ingo Molnar's avatar
Ingo Molnar committed
930

931 932
	prefetchw(&mm->mmap_sem);

Ingo Molnar's avatar
Ingo Molnar committed
933
	/* Get the faulting address: */
934
	address = read_cr2();
Linus Torvalds's avatar
Linus Torvalds committed
935

936
	if (unlikely(kmmio_fault(regs, address)))
937
		return;
Linus Torvalds's avatar
Linus Torvalds committed
938 939 940 941 942 943 944 945 946 947 948 949

	/*
	 * 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
950
	 * protection error (error_code & 9) == 0.
Linus Torvalds's avatar
Linus Torvalds committed
951
	 */
952
	if (unlikely(fault_in_kernel_space(address))) {
953 954 955
		if (!(error_code & (PF_RSVD|PF_USER|PF_PROT)) &&
		    vmalloc_fault(address) >= 0)
			return;
956

Ingo Molnar's avatar
Ingo Molnar committed
957
		/* Can handle a stale RO->RW TLB: */
958
		if (spurious_fault(error_code, address))
959 960
			return;

Ingo Molnar's avatar
Ingo Molnar committed
961
		/* kprobes don't want to hook the spurious faults: */
962 963
		if (notify_page_fault(regs))
			return;
964 965
		/*
		 * Don't take the mm semaphore here. If we fixup a prefetch
Ingo Molnar's avatar
Ingo Molnar committed
966
		 * fault we could otherwise deadlock:
967
		 */
968
		bad_area_nosemaphore(regs, error_code, address);