fault.c 28.8 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1 2
/*
 *  Copyright (C) 1995  Linus Torvalds
Ingo Molnar's avatar
Ingo Molnar committed
3
 *  Copyright (C) 2001, 2002 Andi Kleen, SuSE Labs.
4
 *  Copyright (C) 2008-2009, Red Hat Inc., Ingo Molnar
Linus Torvalds's avatar
Linus Torvalds committed
5
 */
6 7 8 9 10 11 12
#include <linux/magic.h>		/* STACK_END_MAGIC		*/
#include <linux/sched.h>		/* test_thread_flag(), ...	*/
#include <linux/kdebug.h>		/* oops_begin/end, ...		*/
#include <linux/module.h>		/* search_exception_table	*/
#include <linux/bootmem.h>		/* max_low_pfn			*/
#include <linux/kprobes.h>		/* __kprobes, ...		*/
#include <linux/mmiotrace.h>		/* kmmio_handler, ...		*/
13
#include <linux/perf_event.h>		/* perf_sw_event		*/
14
#include <linux/hugetlb.h>		/* hstate_index_to_shift	*/
15
#include <linux/prefetch.h>		/* prefetchw			*/
Ingo Molnar's avatar
Ingo Molnar committed
16

17 18
#include <asm/traps.h>			/* dotraplinkage, ...		*/
#include <asm/pgalloc.h>		/* pgd_*(), ...			*/
Vegard Nossum's avatar
Vegard Nossum committed
19
#include <asm/kmemcheck.h>		/* kmemcheck_*(), ...		*/
20
#include <asm/fixmap.h>			/* VSYSCALL_START		*/
Linus Torvalds's avatar
Linus Torvalds committed
21

22
/*
Ingo Molnar's avatar
Ingo Molnar committed
23 24 25 26 27 28 29
 * 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
30
 */
Ingo Molnar's avatar
Ingo Molnar committed
31 32 33 34 35 36 37 38
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,
};
39

40
/*
41 42
 * Returns 0 if mmiotrace is disabled, or if the fault is not
 * handled by mmiotrace:
43
 */
44 45
static inline int __kprobes
kmmio_fault(struct pt_regs *regs, unsigned long addr)
46
{
47 48 49 50
	if (unlikely(is_kmmio_active()))
		if (kmmio_handler(regs, addr) == 1)
			return -1;
	return 0;
51 52
}

53
static inline int __kprobes notify_page_fault(struct pt_regs *regs)
54
{
55 56 57
	int ret = 0;

	/* kprobe_running() needs smp_processor_id() */
58
	if (kprobes_built_in() && !user_mode_vm(regs)) {
59 60 61 62 63
		preempt_disable();
		if (kprobe_running() && kprobe_fault_handler(regs, 14))
			ret = 1;
		preempt_enable();
	}
64

65
	return ret;
66
}
67

68
/*
Ingo Molnar's avatar
Ingo Molnar committed
69 70 71 72 73 74
 * Prefetch quirks:
 *
 * 32-bit mode:
 *
 *   Sometimes AMD Athlon/Opteron CPUs report invalid exceptions on prefetch.
 *   Check that here and ignore it.
75
 *
Ingo Molnar's avatar
Ingo Molnar committed
76
 * 64-bit mode:
77
 *
Ingo Molnar's avatar
Ingo Molnar committed
78 79 80 81
 *   Sometimes the CPU reports invalid exceptions on prefetch.
 *   Check that here and ignore it.
 *
 * Opcode checker based on code by Richard Brunner.
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
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.
		 */
109
		return (!user_mode(regs) || user_64bit_mode(regs));
110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129
#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
130 131
static int
is_prefetch(struct pt_regs *regs, unsigned long error_code, unsigned long addr)
132
{
Ingo Molnar's avatar
Ingo Molnar committed
133
	unsigned char *max_instr;
134
	unsigned char *instr;
135
	int prefetch = 0;
Linus Torvalds's avatar
Linus Torvalds committed
136

Ingo Molnar's avatar
Ingo Molnar committed
137 138 139 140
	/*
	 * If it was a exec (instruction fetch) fault on NX page, then
	 * do not ignore the fault:
	 */
141
	if (error_code & PF_INSTR)
Linus Torvalds's avatar
Linus Torvalds committed
142
		return 0;
143

144
	instr = (void *)convert_ip_to_linear(current, regs);
145
	max_instr = instr + 15;
Linus Torvalds's avatar
Linus Torvalds committed
146

147
	if (user_mode(regs) && instr >= (unsigned char *)TASK_SIZE)
Linus Torvalds's avatar
Linus Torvalds committed
148 149
		return 0;

150
	while (instr < max_instr) {
Ingo Molnar's avatar
Ingo Molnar committed
151
		unsigned char opcode;
Linus Torvalds's avatar
Linus Torvalds committed
152

153
		if (probe_kernel_address(instr, opcode))
154
			break;
Linus Torvalds's avatar
Linus Torvalds committed
155 156 157

		instr++;

158
		if (!check_prefetch_opcode(regs, instr, opcode, &prefetch))
Linus Torvalds's avatar
Linus Torvalds committed
159 160 161 162 163
			break;
	}
	return prefetch;
}

Ingo Molnar's avatar
Ingo Molnar committed
164 165
static void
force_sig_info_fault(int si_signo, int si_code, unsigned long address,
166
		     struct task_struct *tsk, int fault)
167
{
168
	unsigned lsb = 0;
169 170
	siginfo_t info;

Ingo Molnar's avatar
Ingo Molnar committed
171 172 173 174
	info.si_signo	= si_signo;
	info.si_errno	= 0;
	info.si_code	= si_code;
	info.si_addr	= (void __user *)address;
175 176 177 178 179
	if (fault & VM_FAULT_HWPOISON_LARGE)
		lsb = hstate_index_to_shift(VM_FAULT_GET_HINDEX(fault)); 
	if (fault & VM_FAULT_HWPOISON)
		lsb = PAGE_SHIFT;
	info.si_addr_lsb = lsb;
Ingo Molnar's avatar
Ingo Molnar committed
180

181 182 183
	force_sig_info(si_signo, &info, tsk);
}

184 185 186 187 188
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)
189
{
190 191 192 193
	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
194

195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215
	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;

216
	if (!pmd_present(*pmd))
217
		set_pmd(pmd, *pmd_k);
218
	else
219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235
		BUG_ON(pmd_page(*pmd) != pmd_page(*pmd_k));

	return pmd_k;
}

void vmalloc_sync_all(void)
{
	unsigned long address;

	if (SHARED_KERNEL_PMD)
		return;

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

236
		spin_lock(&pgd_lock);
237
		list_for_each_entry(page, &pgd_list, lru) {
238
			spinlock_t *pgt_lock;
239
			pmd_t *ret;
240

241
			/* the pgt_lock only for Xen */
242 243 244 245 246 247 248
			pgt_lock = &pgd_page_get_mm(page)->page_table_lock;

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

			if (!ret)
249 250
				break;
		}
251
		spin_unlock(&pgd_lock);
252 253 254 255 256 257 258 259
	}
}

/*
 * 32-bit:
 *
 *   Handle a fault on the vmalloc or module mapping area
 */
260
static noinline __kprobes int vmalloc_fault(unsigned long address)
261 262 263 264 265 266 267 268 269
{
	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;

270 271
	WARN_ON_ONCE(in_nmi());

272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305
	/*
	 * 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;
306
}
Linus Torvalds's avatar
Linus Torvalds committed
307

Akinobu Mita's avatar
Akinobu Mita committed
308
static bool low_pfn(unsigned long pfn)
Linus Torvalds's avatar
Linus Torvalds committed
309
{
Akinobu Mita's avatar
Akinobu Mita committed
310 311
	return pfn < max_low_pfn;
}
312

Akinobu Mita's avatar
Akinobu Mita committed
313 314 315 316 317 318
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
319

320
#ifdef CONFIG_X86_PAE
Akinobu Mita's avatar
Akinobu Mita committed
321 322 323
	printk("*pdpt = %016Lx ", pgd_val(*pgd));
	if (!low_pfn(pgd_val(*pgd) >> PAGE_SHIFT) || !pgd_present(*pgd))
		goto out;
324
#endif
Akinobu Mita's avatar
Akinobu Mita committed
325 326
	pmd = pmd_offset(pud_offset(pgd, address), address);
	printk(KERN_CONT "*pde = %0*Lx ", sizeof(*pmd) * 2, (u64)pmd_val(*pmd));
327 328 329 330 331

	/*
	 * 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
332
	 * it's allocated already:
333
	 */
Akinobu Mita's avatar
Akinobu Mita committed
334 335
	if (!low_pfn(pmd_pfn(*pmd)) || !pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
336

Akinobu Mita's avatar
Akinobu Mita committed
337 338 339
	pte = pte_offset_kernel(pmd, address);
	printk("*pte = %0*Lx ", sizeof(*pte) * 2, (u64)pte_val(*pte));
out:
340
	printk("\n");
341 342 343 344 345 346
}

#else /* CONFIG_X86_64: */

void vmalloc_sync_all(void)
{
347
	sync_global_pgds(VMALLOC_START & PGDIR_MASK, VMALLOC_END);
348 349 350 351 352 353 354 355 356
}

/*
 * 64-bit:
 *
 *   Handle a fault on the vmalloc area
 *
 * This assumes no large pages in there.
 */
357
static noinline __kprobes int vmalloc_fault(unsigned long address)
358 359 360 361 362 363 364 365 366 367
{
	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;

368 369
	WARN_ON_ONCE(in_nmi());

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
	/*
	 * 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;
}

423
#ifdef CONFIG_CPU_SUP_AMD
424
static const char errata93_warning[] =
425 426 427 428 429
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";
430
#endif
431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449

/*
 * 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
450 451
	pgd_t *base = __va(read_cr3() & PHYSICAL_PAGE_MASK);
	pgd_t *pgd = base + pgd_index(address);
Linus Torvalds's avatar
Linus Torvalds committed
452 453 454 455
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;

Ingo Molnar's avatar
Ingo Molnar committed
456 457 458
	if (bad_address(pgd))
		goto bad;

459
	printk("PGD %lx ", pgd_val(*pgd));
Ingo Molnar's avatar
Ingo Molnar committed
460 461 462

	if (!pgd_present(*pgd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
463

464
	pud = pud_offset(pgd, address);
Ingo Molnar's avatar
Ingo Molnar committed
465 466 467
	if (bad_address(pud))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
468
	printk("PUD %lx ", pud_val(*pud));
469
	if (!pud_present(*pud) || pud_large(*pud))
Ingo Molnar's avatar
Ingo Molnar committed
470
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
471 472

	pmd = pmd_offset(pud, address);
Ingo Molnar's avatar
Ingo Molnar committed
473 474 475
	if (bad_address(pmd))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
476
	printk("PMD %lx ", pmd_val(*pmd));
Ingo Molnar's avatar
Ingo Molnar committed
477 478
	if (!pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
479 480

	pte = pte_offset_kernel(pmd, address);
Ingo Molnar's avatar
Ingo Molnar committed
481 482 483
	if (bad_address(pte))
		goto bad;

484
	printk("PTE %lx", pte_val(*pte));
Ingo Molnar's avatar
Ingo Molnar committed
485
out:
Linus Torvalds's avatar
Linus Torvalds committed
486 487 488 489
	printk("\n");
	return;
bad:
	printk("BAD\n");
490 491
}

492
#endif /* CONFIG_X86_64 */
Linus Torvalds's avatar
Linus Torvalds committed
493

Ingo Molnar's avatar
Ingo Molnar committed
494 495 496 497 498 499 500 501 502 503 504 505 506
/*
 * 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.
507
 */
508
static int is_errata93(struct pt_regs *regs, unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
509
{
510 511 512 513 514
#if defined(CONFIG_X86_64) && defined(CONFIG_CPU_SUP_AMD)
	if (boot_cpu_data.x86_vendor != X86_VENDOR_AMD
	    || boot_cpu_data.x86 != 0xf)
		return 0;

515
	if (address != regs->ip)
Linus Torvalds's avatar
Linus Torvalds committed
516
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
517

518
	if ((address >> 32) != 0)
Linus Torvalds's avatar
Linus Torvalds committed
519
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
520

Linus Torvalds's avatar
Linus Torvalds committed
521
	address |= 0xffffffffUL << 32;
522 523
	if ((address >= (u64)_stext && address <= (u64)_etext) ||
	    (address >= MODULES_VADDR && address <= MODULES_END)) {
524
		printk_once(errata93_warning);
525
		regs->ip = address;
Linus Torvalds's avatar
Linus Torvalds committed
526 527
		return 1;
	}
528
#endif
Linus Torvalds's avatar
Linus Torvalds committed
529
	return 0;
530
}
Linus Torvalds's avatar
Linus Torvalds committed
531

532
/*
Ingo Molnar's avatar
Ingo Molnar committed
533 534 535 536 537
 * 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
538 539 540 541 542
 * 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
543
	if ((regs->cs == __USER32_CS || (regs->cs & (1<<2))) && (address >> 32))
544 545 546 547 548
		return 1;
#endif
	return 0;
}

549 550 551 552
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
553

554
	/*
Ingo Molnar's avatar
Ingo Molnar committed
555
	 * Pentium F0 0F C7 C8 bug workaround:
556 557 558 559 560 561 562 563 564 565 566 567 568
	 */
	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;
}

569 570 571
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
572 573 574
static void
show_fault_oops(struct pt_regs *regs, unsigned long error_code,
		unsigned long address)
575
{
576 577 578 579
	if (!oops_may_print())
		return;

	if (error_code & PF_INSTR) {
580
		unsigned int level;
Ingo Molnar's avatar
Ingo Molnar committed
581

582 583
		pte_t *pte = lookup_address(address, &level);

584 585
		if (pte && pte_present(*pte) && !pte_exec(*pte))
			printk(nx_warning, current_uid());
586 587
	}

588
	printk(KERN_ALERT "BUG: unable to handle kernel ");
589
	if (address < PAGE_SIZE)
590
		printk(KERN_CONT "NULL pointer dereference");
591
	else
592
		printk(KERN_CONT "paging request");
Ingo Molnar's avatar
Ingo Molnar committed
593

594
	printk(KERN_CONT " at %p\n", (void *) address);
595
	printk(KERN_ALERT "IP:");
596
	printk_address(regs->ip, 1);
Ingo Molnar's avatar
Ingo Molnar committed
597

598 599 600
	dump_pagetable(address);
}

Ingo Molnar's avatar
Ingo Molnar committed
601 602 603
static noinline void
pgtable_bad(struct pt_regs *regs, unsigned long error_code,
	    unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
604
{
Ingo Molnar's avatar
Ingo Molnar committed
605 606 607 608 609 610 611
	struct task_struct *tsk;
	unsigned long flags;
	int sig;

	flags = oops_begin();
	tsk = current;
	sig = SIGKILL;
612

Linus Torvalds's avatar
Linus Torvalds committed
613
	printk(KERN_ALERT "%s: Corrupted page table at address %lx\n",
614
	       tsk->comm, address);
Linus Torvalds's avatar
Linus Torvalds committed
615
	dump_pagetable(address);
Ingo Molnar's avatar
Ingo Molnar committed
616 617 618 619 620

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

621
	if (__die("Bad pagetable", regs, error_code))
622
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
623

624
	oops_end(flags, regs, sig);
Linus Torvalds's avatar
Linus Torvalds committed
625 626
}

Ingo Molnar's avatar
Ingo Molnar committed
627 628
static noinline void
no_context(struct pt_regs *regs, unsigned long error_code,
629
	   unsigned long address, int signal, int si_code)
630 631
{
	struct task_struct *tsk = current;
632
	unsigned long *stackend;
633 634 635
	unsigned long flags;
	int sig;

Ingo Molnar's avatar
Ingo Molnar committed
636
	/* Are we prepared to handle this kernel fault? */
637 638 639 640 641 642 643 644 645
	if (fixup_exception(regs)) {
		if (current_thread_info()->sig_on_uaccess_error && signal) {
			tsk->thread.trap_no = 14;
			tsk->thread.error_code = error_code | PF_USER;
			tsk->thread.cr2 = address;

			/* XXX: hwpoison faults will set the wrong code. */
			force_sig_info_fault(signal, si_code, address, tsk, 0);
		}
646
		return;
647
	}
648 649

	/*
Ingo Molnar's avatar
Ingo Molnar committed
650 651 652 653 654 655 656
	 * 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:
657
	 *
Ingo Molnar's avatar
Ingo Molnar committed
658
	 *   Hall of shame of CPU/BIOS bugs.
659 660 661 662 663 664 665 666 667
	 */
	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
668
	 * terminate things with extreme prejudice:
669 670 671 672 673
	 */
	flags = oops_begin();

	show_fault_oops(regs, error_code, address);

Ingo Molnar's avatar
Ingo Molnar committed
674
	stackend = end_of_stack(tsk);
675
	if (tsk != &init_task && *stackend != STACK_END_MAGIC)
676 677
		printk(KERN_ALERT "Thread overran stack, or stack corrupted\n");

678 679 680
	tsk->thread.cr2		= address;
	tsk->thread.trap_no	= 14;
	tsk->thread.error_code	= error_code;
681 682 683 684

	sig = SIGKILL;
	if (__die("Oops", regs, error_code))
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
685

686 687
	/* 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
688

689 690 691
	oops_end(flags, regs, sig);
}

Ingo Molnar's avatar
Ingo Molnar committed
692 693 694 695 696 697 698 699 700 701 702 703 704 705
/*
 * 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;

706
	printk("%s%s[%d]: segfault at %lx ip %p sp %p error %lx",
Ingo Molnar's avatar
Ingo Molnar committed
707 708 709 710 711 712 713 714 715 716 717 718
		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)
719 720 721 722 723 724
{
	struct task_struct *tsk = current;

	/* User mode accesses just cause a SIGSEGV */
	if (error_code & PF_USER) {
		/*
Ingo Molnar's avatar
Ingo Molnar committed
725
		 * It's possible to have interrupts off here:
726 727 728 729 730
		 */
		local_irq_enable();

		/*
		 * Valid to do another page fault here because this one came
Ingo Molnar's avatar
Ingo Molnar committed
731
		 * from user space:
732 733 734 735 736 737 738
		 */
		if (is_prefetch(regs, error_code, address))
			return;

		if (is_errata100(regs, address))
			return;

739 740 741 742 743 744 745 746 747 748 749 750
#ifdef CONFIG_X86_64
		/*
		 * Instruction fetch faults in the vsyscall page might need
		 * emulation.
		 */
		if (unlikely((error_code & PF_INSTR) &&
			     ((address & ~0xfff) == VSYSCALL_START))) {
			if (emulate_vsyscall(regs, address))
				return;
		}
#endif

Ingo Molnar's avatar
Ingo Molnar committed
751 752 753 754 755 756 757
		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;
758

759
		force_sig_info_fault(SIGSEGV, si_code, address, tsk, 0);
Ingo Molnar's avatar
Ingo Molnar committed
760

761 762 763 764 765 766
		return;
	}

	if (is_f00f_bug(regs, address))
		return;

767
	no_context(regs, error_code, address, SIGSEGV, si_code);
768 769
}

Ingo Molnar's avatar
Ingo Molnar committed
770 771 772
static noinline void
bad_area_nosemaphore(struct pt_regs *regs, unsigned long error_code,
		     unsigned long address)
773 774 775 776
{
	__bad_area_nosemaphore(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
777 778 779
static void
__bad_area(struct pt_regs *regs, unsigned long error_code,
	   unsigned long address, int si_code)
780 781 782 783 784 785 786 787 788 789 790 791
{
	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
792 793
static noinline void
bad_area(struct pt_regs *regs, unsigned long error_code, unsigned long address)
794 795 796 797
{
	__bad_area(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
798 799 800
static noinline void
bad_area_access_error(struct pt_regs *regs, unsigned long error_code,
		      unsigned long address)
801 802 803 804 805
{
	__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
806 807 808
static void
out_of_memory(struct pt_regs *regs, unsigned long error_code,
	      unsigned long address)
809 810 811
{
	/*
	 * We ran out of memory, call the OOM killer, and return the userspace
Ingo Molnar's avatar
Ingo Molnar committed
812
	 * (which will retry the fault, or kill us if we got oom-killed):
813 814
	 */
	up_read(&current->mm->mmap_sem);
Ingo Molnar's avatar
Ingo Molnar committed
815

816 817 818
	pagefault_out_of_memory();
}

Ingo Molnar's avatar
Ingo Molnar committed
819
static void
820 821
do_sigbus(struct pt_regs *regs, unsigned long error_code, unsigned long address,
	  unsigned int fault)
822 823 824
{
	struct task_struct *tsk = current;
	struct mm_struct *mm = tsk->mm;
825
	int code = BUS_ADRERR;
826 827 828

	up_read(&mm->mmap_sem);

Ingo Molnar's avatar
Ingo Molnar committed
829
	/* Kernel mode? Handle exceptions or die: */
830
	if (!(error_code & PF_USER)) {
831
		no_context(regs, error_code, address, SIGBUS, BUS_ADRERR);
832 833
		return;
	}
Ingo Molnar's avatar
Ingo Molnar committed
834

835
	/* User-space => ok to do another page fault: */
836 837
	if (is_prefetch(regs, error_code, address))
		return;
Ingo Molnar's avatar
Ingo Molnar committed
838 839 840 841 842

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

843
#ifdef CONFIG_MEMORY_FAILURE
844
	if (fault & (VM_FAULT_HWPOISON|VM_FAULT_HWPOISON_LARGE)) {
845 846 847 848 849 850
		printk(KERN_ERR
	"MCE: Killing %s:%d due to hardware memory corruption fault at %lx\n",
			tsk->comm, tsk->pid, address);
		code = BUS_MCEERR_AR;
	}
#endif
851
	force_sig_info_fault(SIGBUS, code, address, tsk, fault);
852 853
}

854
static noinline int
Ingo Molnar's avatar
Ingo Molnar committed
855 856
mm_fault_error(struct pt_regs *regs, unsigned long error_code,
	       unsigned long address, unsigned int fault)
857
{
858 859 860 861 862 863 864 865
	/*
	 * Pagefault was interrupted by SIGKILL. We have no reason to
	 * continue pagefault.
	 */
	if (fatal_signal_pending(current)) {
		if (!(fault & VM_FAULT_RETRY))
			up_read(&current->mm->mmap_sem);
		if (!(error_code & PF_USER))
866
			no_context(regs, error_code, address, 0, 0);
867 868 869 870 871
		return 1;
	}
	if (!(fault & VM_FAULT_ERROR))
		return 0;

Ingo Molnar's avatar
Ingo Molnar committed
872
	if (fault & VM_FAULT_OOM) {
873 874 875
		/* Kernel mode? Handle exceptions or die: */
		if (!(error_code & PF_USER)) {
			up_read(&current->mm->mmap_sem);
876 877
			no_context(regs, error_code, address,
				   SIGSEGV, SEGV_MAPERR);
878
			return 1;
879 880
		}

881
		out_of_memory(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
882
	} else {
883 884
		if (fault & (VM_FAULT_SIGBUS|VM_FAULT_HWPOISON|
			     VM_FAULT_HWPOISON_LARGE))
885
			do_sigbus(regs, error_code, address, fault);
Ingo Molnar's avatar
Ingo Molnar committed
886 887 888
		else
			BUG();
	}
889
	return 1;
890 891
}

892 893 894 895
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
896

897 898 899 900 901 902
	if ((error_code & PF_INSTR) && !pte_exec(*pte))
		return 0;

	return 1;
}

903
/*
Ingo Molnar's avatar
Ingo Molnar committed
904 905 906 907 908 909 910 911
 * 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.
 *
912 913 914
 * There are no security implications to leaving a stale TLB when
 * increasing the permissions on a page.
 */
915
static noinline __kprobes int
Ingo Molnar's avatar
Ingo Molnar committed
916
spurious_fault(unsigned long error_code, unsigned long address)
917 918 919 920 921
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;
922
	int ret;
923 924 925 926 927 928 929 930 931 932 933 934 935

	/* 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;

936 937 938
	if (pud_large(*pud))
		return spurious_fault_check(error_code, (pte_t *) pud);

939 940 941 942
	pmd = pmd_offset(pud, address);
	if (!pmd_present(*pmd))
		return 0;

943 944 945
	if (pmd_large(*pmd))
		return spurious_fault_check(error_code, (pte_t *) pmd);

946 947 948 949 950 951
	/*
	 * Note: don't use pte_present() here, since it returns true
	 * if the _PAGE_PROTNONE bit is set.  However, this aliases the
	 * _PAGE_GLOBAL bit, which for kernel pages give false positives
	 * when CONFIG_DEBUG_PAGEALLOC is used.
	 */
952
	pte = pte_offset_kernel(pmd, address);
953
	if (!(pte_flags(*pte) & _PAGE_PRESENT))
954 955
		return 0;

956 957 958 959 960
	ret = spurious_fault_check(error_code, pte);
	if (!ret)
		return 0;

	/*
Ingo Molnar's avatar
Ingo Molnar committed
961 962
	 * Make sure we have permissions in PMD.
	 * If not, then there's a bug in the page tables:
963 964 965
	 */
	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
966

967
	return ret;
968 969
}

970
int show_unhandled_signals = 1;
Linus Torvalds's avatar
Linus Torvalds committed
971

Ingo Molnar's avatar
Ingo Molnar committed
972
static inline int
973
access_error(unsigned long error_code, struct vm_area_struct *vma)
974
{
975
	if (error_code & PF_WRITE) {
Ingo Molnar's avatar
Ingo Molnar committed
976
		/* write, present and write, not present: */
977 978
		if (unlikely(!(vma->vm_flags & VM_WRITE)))
			return 1;
Ingo Molnar's avatar
Ingo Molnar committed
979
		return 0;
980 981
	}

Ingo Molnar's avatar
Ingo Molnar committed
982 983 984 985 986 987 988 989
	/* 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;

990 991 992
	return 0;
}

993 994
static int fault_in_kernel_space(unsigned long address)
{
995
	return address >= TASK_SIZE_MAX;
996 997
}

Linus Torvalds's avatar
Linus Torvalds committed
998 999 1000 1001 1002
/*
 * This routine handles page faults.  It determines the address,
 * and the problem, and then passes it off to one of the appropriate
 * routines.
 */
1003 1004
dotraplinkage void __kprobes
do_page_fault(struct pt_regs *regs, unsigned long error_code)
Linus Torvalds's avatar
Linus Torvalds committed
1005
{
Ingo Molnar's avatar
Ingo Molnar committed
1006
	struct vm_area_struct *vma;
Linus Torvalds's avatar
Linus Torvalds committed
1007
	struct task_struct *tsk;
Ingo Molnar's avatar
Ingo Molnar committed
1008
	unsigned long address;
Linus Torvalds's avatar
Linus Torvalds committed
1009
	struct mm_struct *mm;
1010
	int fault;
1011
	int write = error_code & PF_WRITE;
1012
	unsigned int flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE |
1013
					(write ? FAULT_FLAG_WRITE : 0);
Linus Torvalds's avatar
Linus Torvalds committed
1014

1015 1016
	tsk = current;
	mm = tsk->mm;
Ingo Molnar's avatar
Ingo Molnar committed
1017 1018

	/* Get the faulting address: */
1019
	address = read_cr2();
Linus Torvalds's avatar
Linus Torvalds committed
1020

Vegard Nossum's avatar
Vegard Nossum committed
1021 1022 1023 1024 1025 1026
	/*
	 * 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);
1027
	prefetchw(&mm->mmap_sem);
Vegard Nossum's avatar
Vegard Nossum committed
1028

1029
	if (unlikely(kmmio_fault(regs, address)))
1030
		return;
Linus Torvalds's avatar
Linus Torvalds committed
1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042

	/*
	 * 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