fault.c 26.5 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1 2
/*
 *  Copyright (C) 1995  Linus Torvalds
Ingo Molnar's avatar
Ingo Molnar committed
3
 *  Copyright (C) 2001, 2002 Andi Kleen, SuSE Labs.
Linus Torvalds's avatar
Linus Torvalds committed
4 5
 */
#include <linux/interrupt.h>
Ingo Molnar's avatar
Ingo Molnar committed
6 7
#include <linux/mmiotrace.h>
#include <linux/bootmem.h>
Linus Torvalds's avatar
Linus Torvalds committed
8
#include <linux/compiler.h>
Harvey Harrison's avatar
Harvey Harrison committed
9
#include <linux/highmem.h>
10
#include <linux/kprobes.h>
11
#include <linux/uaccess.h>
Ingo Molnar's avatar
Ingo Molnar committed
12 13 14 15 16 17 18
#include <linux/vmalloc.h>
#include <linux/vt_kern.h>
#include <linux/signal.h>
#include <linux/kernel.h>
#include <linux/ptrace.h>
#include <linux/string.h>
#include <linux/module.h>
19
#include <linux/kdebug.h>
Ingo Molnar's avatar
Ingo Molnar committed
20
#include <linux/errno.h>
21
#include <linux/magic.h>
Ingo Molnar's avatar
Ingo Molnar committed
22 23 24 25 26 27 28 29 30
#include <linux/sched.h>
#include <linux/types.h>
#include <linux/init.h>
#include <linux/mman.h>
#include <linux/tty.h>
#include <linux/smp.h>
#include <linux/mm.h>

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

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

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

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

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

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

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

81
	return ret;
82
}
83

84
/*
Ingo Molnar's avatar
Ingo Molnar committed
85 86 87 88 89 90
 * Prefetch quirks:
 *
 * 32-bit mode:
 *
 *   Sometimes AMD Athlon/Opteron CPUs report invalid exceptions on prefetch.
 *   Check that here and ignore it.
91
 *
Ingo Molnar's avatar
Ingo Molnar committed
92
 * 64-bit mode:
93
 *
Ingo Molnar's avatar
Ingo Molnar committed
94 95 96 97
 *   Sometimes the CPU reports invalid exceptions on prefetch.
 *   Check that here and ignore it.
 *
 * Opcode checker based on code by Richard Brunner.
98
 */
99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145
static inline int
check_prefetch_opcode(struct pt_regs *regs, unsigned char *instr,
		      unsigned char opcode, int *prefetch)
{
	unsigned char instr_hi = opcode & 0xf0;
	unsigned char instr_lo = opcode & 0x0f;

	switch (instr_hi) {
	case 0x20:
	case 0x30:
		/*
		 * Values 0x26,0x2E,0x36,0x3E are valid x86 prefixes.
		 * In X86_64 long mode, the CPU will signal invalid
		 * opcode if some of these prefixes are present so
		 * X86_64 will never get here anyway
		 */
		return ((instr_lo & 7) == 0x6);
#ifdef CONFIG_X86_64
	case 0x40:
		/*
		 * In AMD64 long mode 0x40..0x4F are valid REX prefixes
		 * Need to figure out under what instruction mode the
		 * instruction was issued. Could check the LDT for lm,
		 * but for now it's good enough to assume that long
		 * mode only uses well known segments or kernel.
		 */
		return (!user_mode(regs)) || (regs->cs == __USER_CS);
#endif
	case 0x60:
		/* 0x64 thru 0x67 are valid prefixes in all modes. */
		return (instr_lo & 0xC) == 0x4;
	case 0xF0:
		/* 0xF0, 0xF2, 0xF3 are valid prefixes in all modes. */
		return !instr_lo || (instr_lo>>1) == 1;
	case 0x00:
		/* Prefetch instruction is 0x0F0D or 0x0F18 */
		if (probe_kernel_address(instr, opcode))
			return 0;

		*prefetch = (instr_lo == 0xF) &&
			(opcode == 0x0D || opcode == 0x18);
		return 0;
	default:
		return 0;
	}
}

Ingo Molnar's avatar
Ingo Molnar committed
146 147
static int
is_prefetch(struct pt_regs *regs, unsigned long error_code, unsigned long addr)
148
{
Ingo Molnar's avatar
Ingo Molnar committed
149
	unsigned char *max_instr;
150
	unsigned char *instr;
151
	int prefetch = 0;
Linus Torvalds's avatar
Linus Torvalds committed
152

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

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

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

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

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

		instr++;

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

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

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

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

194
#ifdef CONFIG_X86_64
195 196
static int bad_address(void *p)
{
Linus Torvalds's avatar
Linus Torvalds committed
197
	unsigned long dummy;
Ingo Molnar's avatar
Ingo Molnar committed
198

199
	return probe_kernel_address((unsigned long *)p, dummy);
200
}
201
#endif
Linus Torvalds's avatar
Linus Torvalds committed
202

203
static void dump_pagetable(unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
204
{
205 206 207 208 209
#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
210

211 212 213 214 215 216
#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
217
							& (PTRS_PER_PMD - 1)];
218 219 220 221 222 223 224 225 226 227 228
		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
229
	 * it's allocated already:
230 231 232 233
	 */
	if ((page >> PAGE_SHIFT) < max_low_pfn
	    && (page & _PAGE_PRESENT)
	    && !(page & _PAGE_PSE)) {
Ingo Molnar's avatar
Ingo Molnar committed
234

235 236
		page &= PAGE_MASK;
		page = ((__typeof__(page) *) __va(page))[(address >> PAGE_SHIFT)
Ingo Molnar's avatar
Ingo Molnar committed
237
							& (PTRS_PER_PTE - 1)];
238 239 240 241 242
		printk("*pte = %0*Lx ", sizeof(page)*2, (u64)page);
	}

	printk("\n");
#else /* CONFIG_X86_64 */
Linus Torvalds's avatar
Linus Torvalds committed
243 244 245 246 247
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;

248
	pgd = (pgd_t *)read_cr3();
Linus Torvalds's avatar
Linus Torvalds committed
249

250
	pgd = __va((unsigned long)pgd & PHYSICAL_PAGE_MASK);
Ingo Molnar's avatar
Ingo Molnar committed
251

Linus Torvalds's avatar
Linus Torvalds committed
252
	pgd += pgd_index(address);
Ingo Molnar's avatar
Ingo Molnar committed
253 254 255
	if (bad_address(pgd))
		goto bad;

256
	printk("PGD %lx ", pgd_val(*pgd));
Ingo Molnar's avatar
Ingo Molnar committed
257 258 259

	if (!pgd_present(*pgd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
260

261
	pud = pud_offset(pgd, address);
Ingo Molnar's avatar
Ingo Molnar committed
262 263 264
	if (bad_address(pud))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
265
	printk("PUD %lx ", pud_val(*pud));
266
	if (!pud_present(*pud) || pud_large(*pud))
Ingo Molnar's avatar
Ingo Molnar committed
267
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
268 269

	pmd = pmd_offset(pud, address);
Ingo Molnar's avatar
Ingo Molnar committed
270 271 272
	if (bad_address(pmd))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
273
	printk("PMD %lx ", pmd_val(*pmd));
Ingo Molnar's avatar
Ingo Molnar committed
274 275
	if (!pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
276 277

	pte = pte_offset_kernel(pmd, address);
Ingo Molnar's avatar
Ingo Molnar committed
278 279 280
	if (bad_address(pte))
		goto bad;

281
	printk("PTE %lx", pte_val(*pte));
Ingo Molnar's avatar
Ingo Molnar committed
282
out:
Linus Torvalds's avatar
Linus Torvalds committed
283 284 285 286
	printk("\n");
	return;
bad:
	printk("BAD\n");
287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317
#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
318

319 320 321
	if (!pmd_present(*pmd)) {
		set_pmd(pmd, *pmd_k);
		arch_flush_lazy_mmu_mode();
Ingo Molnar's avatar
Ingo Molnar committed
322
	} else {
323
		BUG_ON(pmd_page(*pmd) != pmd_page(*pmd_k));
Ingo Molnar's avatar
Ingo Molnar committed
324 325
	}

326
	return pmd_k;
Linus Torvalds's avatar
Linus Torvalds committed
327 328
}

329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347
/*
 * 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;
}

#else /* CONFIG_X86_64: */

348
static const char errata93_warning[] =
Linus Torvalds's avatar
Linus Torvalds committed
349 350 351 352
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";
353 354 355 356 357 358 359 360 361 362

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

363
#endif
Linus Torvalds's avatar
Linus Torvalds committed
364

Ingo Molnar's avatar
Ingo Molnar committed
365 366 367 368 369 370 371 372 373 374 375 376 377
/*
 * 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.
378
 */
379
static int is_errata93(struct pt_regs *regs, unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
380
{
381
#ifdef CONFIG_X86_64
Ingo Molnar's avatar
Ingo Molnar committed
382 383
	static int once;

384
	if (address != regs->ip)
Linus Torvalds's avatar
Linus Torvalds committed
385
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
386

387
	if ((address >> 32) != 0)
Linus Torvalds's avatar
Linus Torvalds committed
388
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
389

Linus Torvalds's avatar
Linus Torvalds committed
390
	address |= 0xffffffffUL << 32;
391 392
	if ((address >= (u64)_stext && address <= (u64)_etext) ||
	    (address >= MODULES_VADDR && address <= MODULES_END)) {
Ingo Molnar's avatar
Ingo Molnar committed
393
		if (!once) {
394
			printk(errata93_warning);
Ingo Molnar's avatar
Ingo Molnar committed
395
			once = 1;
Linus Torvalds's avatar
Linus Torvalds committed
396
		}
397
		regs->ip = address;
Linus Torvalds's avatar
Linus Torvalds committed
398 399
		return 1;
	}
400
#endif
Linus Torvalds's avatar
Linus Torvalds committed
401
	return 0;
402
}
Linus Torvalds's avatar
Linus Torvalds committed
403

404
/*
Ingo Molnar's avatar
Ingo Molnar committed
405 406 407 408 409
 * 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
410 411 412 413 414
 * 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
415
	if ((regs->cs == __USER32_CS || (regs->cs & (1<<2))) && (address >> 32))
416 417 418 419 420
		return 1;
#endif
	return 0;
}

421 422 423 424
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
425

426
	/*
Ingo Molnar's avatar
Ingo Molnar committed
427
	 * Pentium F0 0F C7 C8 bug workaround:
428 429 430 431 432 433 434 435 436 437 438 439 440
	 */
	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
441 442 443
static void
show_fault_oops(struct pt_regs *regs, unsigned long error_code,
		unsigned long address)
444
{
445 446 447
#ifdef CONFIG_X86_32
	if (!oops_may_print())
		return;
448
#endif
449 450 451

#ifdef CONFIG_X86_PAE
	if (error_code & PF_INSTR) {
452
		unsigned int level;
Ingo Molnar's avatar
Ingo Molnar committed
453

454 455
		pte_t *pte = lookup_address(address, &level);

Ingo Molnar's avatar
Ingo Molnar committed
456
		if (pte && pte_present(*pte) && !pte_exec(*pte)) {
457 458
			printk(KERN_CRIT "kernel tried to execute "
				"NX-protected page - exploit attempt? "
459
				"(uid: %d)\n", current_uid());
Ingo Molnar's avatar
Ingo Molnar committed
460
		}
461 462 463
	}
#endif

464
	printk(KERN_ALERT "BUG: unable to handle kernel ");
465
	if (address < PAGE_SIZE)
466
		printk(KERN_CONT "NULL pointer dereference");
467
	else
468
		printk(KERN_CONT "paging request");
Ingo Molnar's avatar
Ingo Molnar committed
469

470
	printk(KERN_CONT " at %p\n", (void *) address);
471
	printk(KERN_ALERT "IP:");
472
	printk_address(regs->ip, 1);
Ingo Molnar's avatar
Ingo Molnar committed
473

474 475 476
	dump_pagetable(address);
}

Ingo Molnar's avatar
Ingo Molnar committed
477 478 479
static noinline void
pgtable_bad(struct pt_regs *regs, unsigned long error_code,
	    unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
480
{
Ingo Molnar's avatar
Ingo Molnar committed
481 482 483 484 485 486 487
	struct task_struct *tsk;
	unsigned long flags;
	int sig;

	flags = oops_begin();
	tsk = current;
	sig = SIGKILL;
488

Linus Torvalds's avatar
Linus Torvalds committed
489
	printk(KERN_ALERT "%s: Corrupted page table at address %lx\n",
490
	       tsk->comm, address);
Linus Torvalds's avatar
Linus Torvalds committed
491
	dump_pagetable(address);
Ingo Molnar's avatar
Ingo Molnar committed
492 493 494 495 496

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

497
	if (__die("Bad pagetable", regs, error_code))
498
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
499

500
	oops_end(flags, regs, sig);
Linus Torvalds's avatar
Linus Torvalds committed
501 502
}

Ingo Molnar's avatar
Ingo Molnar committed
503 504 505
static noinline void
no_context(struct pt_regs *regs, unsigned long error_code,
	   unsigned long address)
506 507
{
	struct task_struct *tsk = current;
508 509
	unsigned long *stackend;

510 511 512 513 514
#ifdef CONFIG_X86_64
	unsigned long flags;
	int sig;
#endif

Ingo Molnar's avatar
Ingo Molnar committed
515
	/* Are we prepared to handle this kernel fault? */
516 517 518 519
	if (fixup_exception(regs))
		return;

	/*
Ingo Molnar's avatar
Ingo Molnar committed
520 521 522 523 524 525 526
	 * 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:
527
	 *
Ingo Molnar's avatar
Ingo Molnar committed
528
	 *   Hall of shame of CPU/BIOS bugs.
529 530 531 532 533 534 535 536 537
	 */
	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
538
	 * terminate things with extreme prejudice:
539 540 541 542 543 544 545 546 547
	 */
#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
548
	stackend = end_of_stack(tsk);
549 550 551
	if (*stackend != STACK_END_MAGIC)
		printk(KERN_ALERT "Thread overran stack, or stack corrupted\n");

552 553 554 555 556 557 558 559 560 561 562 563
	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
564

565 566
	/* 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
567

568 569 570 571
	oops_end(flags, regs, sig);
#endif
}

Ingo Molnar's avatar
Ingo Molnar committed
572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598
/*
 * 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)
599 600 601 602 603 604
{
	struct task_struct *tsk = current;

	/* User mode accesses just cause a SIGSEGV */
	if (error_code & PF_USER) {
		/*
Ingo Molnar's avatar
Ingo Molnar committed
605
		 * It's possible to have interrupts off here:
606 607 608 609 610
		 */
		local_irq_enable();

		/*
		 * Valid to do another page fault here because this one came
Ingo Molnar's avatar
Ingo Molnar committed
611
		 * from user space:
612 613 614 615 616 617 618
		 */
		if (is_prefetch(regs, error_code, address))
			return;

		if (is_errata100(regs, address))
			return;

Ingo Molnar's avatar
Ingo Molnar committed
619 620 621 622 623 624 625
		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;
626 627

		force_sig_info_fault(SIGSEGV, si_code, address, tsk);
Ingo Molnar's avatar
Ingo Molnar committed
628

629 630 631 632 633 634 635 636 637
		return;
	}

	if (is_f00f_bug(regs, address))
		return;

	no_context(regs, error_code, address);
}

Ingo Molnar's avatar
Ingo Molnar committed
638 639 640
static noinline void
bad_area_nosemaphore(struct pt_regs *regs, unsigned long error_code,
		     unsigned long address)
641 642 643 644
{
	__bad_area_nosemaphore(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
645 646 647
static void
__bad_area(struct pt_regs *regs, unsigned long error_code,
	   unsigned long address, int si_code)
648 649 650 651 652 653 654 655 656 657 658 659
{
	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
660 661
static noinline void
bad_area(struct pt_regs *regs, unsigned long error_code, unsigned long address)
662 663 664 665
{
	__bad_area(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
666 667 668
static noinline void
bad_area_access_error(struct pt_regs *regs, unsigned long error_code,
		      unsigned long address)
669 670 671 672 673
{
	__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
674 675 676
static void
out_of_memory(struct pt_regs *regs, unsigned long error_code,
	      unsigned long address)
677 678 679
{
	/*
	 * We ran out of memory, call the OOM killer, and return the userspace
Ingo Molnar's avatar
Ingo Molnar committed
680
	 * (which will retry the fault, or kill us if we got oom-killed):
681 682
	 */
	up_read(&current->mm->mmap_sem);
Ingo Molnar's avatar
Ingo Molnar committed
683

684 685 686
	pagefault_out_of_memory();
}

Ingo Molnar's avatar
Ingo Molnar committed
687 688
static void
do_sigbus(struct pt_regs *regs, unsigned long error_code, unsigned long address)
689 690 691 692 693 694
{
	struct task_struct *tsk = current;
	struct mm_struct *mm = tsk->mm;

	up_read(&mm->mmap_sem);

Ingo Molnar's avatar
Ingo Molnar committed
695
	/* Kernel mode? Handle exceptions or die: */
696 697
	if (!(error_code & PF_USER))
		no_context(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
698

699
#ifdef CONFIG_X86_32
Ingo Molnar's avatar
Ingo Molnar committed
700
	/* User space => ok to do another page fault: */
701 702 703
	if (is_prefetch(regs, error_code, address))
		return;
#endif
Ingo Molnar's avatar
Ingo Molnar committed
704 705 706 707 708

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

709 710 711
	force_sig_info_fault(SIGBUS, BUS_ADRERR, address, tsk);
}

Ingo Molnar's avatar
Ingo Molnar committed
712 713 714
static noinline void
mm_fault_error(struct pt_regs *regs, unsigned long error_code,
	       unsigned long address, unsigned int fault)
715
{
Ingo Molnar's avatar
Ingo Molnar committed
716
	if (fault & VM_FAULT_OOM) {
717
		out_of_memory(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
718 719 720 721 722 723
	} else {
		if (fault & VM_FAULT_SIGBUS)
			do_sigbus(regs, error_code, address);
		else
			BUG();
	}
724 725
}

726 727 728 729
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
730

731 732 733 734 735 736
	if ((error_code & PF_INSTR) && !pte_exec(*pte))
		return 0;

	return 1;
}

737
/*
Ingo Molnar's avatar
Ingo Molnar committed
738 739 740 741 742 743 744 745
 * 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.
 *
746 747 748
 * There are no security implications to leaving a stale TLB when
 * increasing the permissions on a page.
 */
Ingo Molnar's avatar
Ingo Molnar committed
749 750
static noinline int
spurious_fault(unsigned long error_code, unsigned long address)
751 752 753 754 755
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;
756
	int ret;
757 758 759 760 761 762 763 764 765 766 767 768 769

	/* 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;

770 771 772
	if (pud_large(*pud))
		return spurious_fault_check(error_code, (pte_t *) pud);

773 774 775 776
	pmd = pmd_offset(pud, address);
	if (!pmd_present(*pmd))
		return 0;

777 778 779
	if (pmd_large(*pmd))
		return spurious_fault_check(error_code, (pte_t *) pmd);

780 781 782 783
	pte = pte_offset_kernel(pmd, address);
	if (!pte_present(*pte))
		return 0;

784 785 786 787 788
	ret = spurious_fault_check(error_code, pte);
	if (!ret)
		return 0;

	/*
Ingo Molnar's avatar
Ingo Molnar committed
789 790
	 * Make sure we have permissions in PMD.
	 * If not, then there's a bug in the page tables:
791 792 793
	 */
	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
794

795
	return ret;
796 797
}

Linus Torvalds's avatar
Linus Torvalds committed
798
/*
Ingo Molnar's avatar
Ingo Molnar committed
799 800 801
 * 32-bit:
 *
 *   Handle a fault on the vmalloc or module mapping area
802
 *
Ingo Molnar's avatar
Ingo Molnar committed
803 804 805
 * 64-bit:
 *
 *   Handle a fault on the vmalloc area
806 807
 *
 * This assumes no large pages in there.
Linus Torvalds's avatar
Linus Torvalds committed
808
 */
809
static noinline int vmalloc_fault(unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
810
{
811 812 813 814
#ifdef CONFIG_X86_32
	unsigned long pgd_paddr;
	pmd_t *pmd_k;
	pte_t *pte_k;
815

Ingo Molnar's avatar
Ingo Molnar committed
816
	/* Make sure we are in vmalloc area: */
817 818 819
	if (!(address >= VMALLOC_START && address < VMALLOC_END))
		return -1;

820 821 822 823 824 825 826 827 828 829 830
	/*
	 * 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
831

832 833 834
	pte_k = pte_offset_kernel(pmd_k, address);
	if (!pte_present(*pte_k))
		return -1;
Ingo Molnar's avatar
Ingo Molnar committed
835

836 837
	return 0;
#else
Linus Torvalds's avatar
Linus Torvalds committed
838 839 840 841 842
	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
843
	/* Make sure we are in vmalloc area: */
844 845 846
	if (!(address >= VMALLOC_START && address < VMALLOC_END))
		return -1;

Ingo Molnar's avatar
Ingo Molnar committed
847 848 849 850 851
	/*
	 * Copy kernel mappings over when needed. This can also
	 * happen within a race in page table update. In the later
	 * case just flush:
	 */
852
	pgd = pgd_offset(current->active_mm, address);
Linus Torvalds's avatar
Linus Torvalds committed
853 854 855
	pgd_ref = pgd_offset_k(address);
	if (pgd_none(*pgd_ref))
		return -1;
Ingo Molnar's avatar
Ingo Molnar committed
856

Linus Torvalds's avatar
Linus Torvalds committed
857 858
	if (pgd_none(*pgd))
		set_pgd(pgd, *pgd_ref);
859
	else
860
		BUG_ON(pgd_page_vaddr(*pgd) != pgd_page_vaddr(*pgd_ref));
Linus Torvalds's avatar
Linus Torvalds committed
861

Ingo Molnar's avatar
Ingo Molnar committed
862 863 864 865
	/*
	 * Below here mismatches are bugs because these lower tables
	 * are shared:
	 */
Linus Torvalds's avatar
Linus Torvalds committed
866 867 868 869 870

	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
871

872
	if (pud_none(*pud) || pud_page_vaddr(*pud) != pud_page_vaddr(*pud_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
875 876 877 878
	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
879

Linus Torvalds's avatar
Linus Torvalds committed
880 881
	if (pmd_none(*pmd) || pmd_page(*pmd) != pmd_page(*pmd_ref))
		BUG();
Ingo Molnar's avatar
Ingo Molnar committed
882

Linus Torvalds's avatar
Linus Torvalds committed
883 884 885
	pte_ref = pte_offset_kernel(pmd_ref, address);
	if (!pte_present(*pte_ref))
		return -1;
Ingo Molnar's avatar
Ingo Molnar committed
886

Linus Torvalds's avatar
Linus Torvalds committed
887
	pte = pte_offset_kernel(pmd, address);
Ingo Molnar's avatar
Ingo Molnar committed
888 889 890 891 892 893

	/*
	 * Don't use pte_page here, because the mappings can point
	 * outside mem_map, and the NUMA hash lookup cannot handle
	 * that:
	 */
894
	if (!pte_present(*pte) || pte_pfn(*pte) != pte_pfn(*pte_ref))
Linus Torvalds's avatar
Linus Torvalds committed
895
		BUG();
Ingo Molnar's avatar
Ingo Molnar committed
896

Linus Torvalds's avatar
Linus Torvalds committed
897
	return 0;
898
#endif
Linus Torvalds's avatar
Linus Torvalds committed
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;
}

924 925 926 927
static int fault_in_kernel_space(unsigned long address)
{
#ifdef CONFIG_X86_32
	return address >= TASK_SIZE;
Ingo Molnar's avatar
Ingo Molnar committed
928
#else
929
	return address >= TASK_SIZE64;
Ingo Molnar's avatar
Ingo Molnar committed
930
#endif
931 932
}

Linus Torvalds's avatar
Linus Torvalds committed
933 934 935 936 937
/*
 * This routine handles page faults.  It determines the address,
 * and the problem, and then passes it off to one of the appropriate
 * routines.
 */
938 939 940 941
#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
942
{
Ingo Molnar's avatar
Ingo Molnar committed
943
	struct vm_area_struct *vma;
Linus Torvalds's avatar
Linus Torvalds committed
944
	struct task_struct *tsk;
Ingo Molnar's avatar
Ingo Molnar committed
945
	unsigned long address;
Linus Torvalds's avatar
Linus Torvalds committed
946
	struct mm_struct *mm;
947
	int write;
948
	int fault;
Linus Torvalds's avatar
Linus Torvalds committed
949

950 951
	tsk = current;
	mm = tsk->mm;
Ingo Molnar's avatar
Ingo Molnar committed
952

953 954
	prefetchw(&mm->mmap_sem);

Ingo Molnar's avatar
Ingo Molnar committed
955
	/* Get the faulting address: */
956
	address = read_cr2();
Linus Torvalds's avatar
Linus Torvalds 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))) {
975 976 977
		if (!(error_code & (PF_RSVD|PF_USER|PF_PROT)) &&
		    vmalloc_fault(address) >= 0)
			return;
978

Ingo Molnar's avatar
Ingo Molnar committed
979
		/* Can handle a stale RO->RW TLB: */
980
		if (spurious_fault(error_code, address))
981 982
			return;

Ingo Molnar's avatar
Ingo Molnar committed
983
		/* kprobes don't want to hook the spurious faults: */
984 985
		if (notify_page_fault(regs))
			return;
986 987
		/*
		 * Don't take the mm semaphore here. If we fixup a prefetch
Ingo Molnar's avatar
Ingo Molnar committed
988
		 * fault we could otherwise deadlock:
989
		 */
990
		bad_area_nosemaphore(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
991

992
		return;
993 994
	}

Ingo Molnar's avatar
Ingo Molnar committed
995
	/* kprobes don't want to hook the spurious faults: */
996
	if (unlikely(notify_page_fault(regs)))
997
		return;
998
	/*
999 1000 1001 1002
	 * 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
1003
	 * potential system fault or CPU buglet:
1004
	 */
1005 1006 1007
	if (user_mode_vm(regs)) {
		local_irq_enable();
		error_code |= PF_USER;
Ingo Molnar's avatar
Ingo Molnar committed
1008 1009 1010 1011
	} else {
		if (regs->flags & X86_EFLAGS_IF)
			local_irq_enable();
	}
1012

1013
	if (unlikely(error_code & PF_RSVD))