fault.c 32.3 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1 2
/*
 *  Copyright (C) 1995  Linus Torvalds
Ingo Molnar's avatar
Ingo Molnar committed
3
 *  Copyright (C) 2001, 2002 Andi Kleen, SuSE Labs.
4
 *  Copyright (C) 2008-2009, Red Hat Inc., Ingo Molnar
Linus Torvalds's avatar
Linus Torvalds committed
5
 */
6 7 8 9
#include <linux/sched.h>		/* test_thread_flag(), ...	*/
#include <linux/kdebug.h>		/* oops_begin/end, ...		*/
#include <linux/module.h>		/* search_exception_table	*/
#include <linux/bootmem.h>		/* max_low_pfn			*/
10
#include <linux/kprobes.h>		/* NOKPROBE_SYMBOL, ...		*/
11
#include <linux/mmiotrace.h>		/* kmmio_handler, ...		*/
12
#include <linux/perf_event.h>		/* perf_sw_event		*/
13
#include <linux/hugetlb.h>		/* hstate_index_to_shift	*/
14
#include <linux/prefetch.h>		/* prefetchw			*/
15
#include <linux/context_tracking.h>	/* exception_enter(), ...	*/
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 21
#include <asm/fixmap.h>			/* VSYSCALL_ADDR		*/
#include <asm/vsyscall.h>		/* emulate_vsyscall		*/
Linus Torvalds's avatar
Linus Torvalds committed
22

23 24 25
#define CREATE_TRACE_POINTS
#include <asm/trace/exceptions.h>

26
/*
Ingo Molnar's avatar
Ingo Molnar committed
27 28 29 30 31 32 33
 * 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
34
 */
Ingo Molnar's avatar
Ingo Molnar committed
35 36 37 38 39 40 41 42
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,
};
43

44
/*
45 46
 * Returns 0 if mmiotrace is disabled, or if the fault is not
 * handled by mmiotrace:
47
 */
48
static nokprobe_inline int
49
kmmio_fault(struct pt_regs *regs, unsigned long addr)
50
{
51 52 53 54
	if (unlikely(is_kmmio_active()))
		if (kmmio_handler(regs, addr) == 1)
			return -1;
	return 0;
55 56
}

57
static nokprobe_inline int kprobes_fault(struct pt_regs *regs)
58
{
59 60 61
	int ret = 0;

	/* kprobe_running() needs smp_processor_id() */
62
	if (kprobes_built_in() && !user_mode_vm(regs)) {
63 64 65 66 67
		preempt_disable();
		if (kprobe_running() && kprobe_fault_handler(regs, 14))
			ret = 1;
		preempt_enable();
	}
68

69
	return ret;
70
}
71

72
/*
Ingo Molnar's avatar
Ingo Molnar committed
73 74 75 76 77 78
 * Prefetch quirks:
 *
 * 32-bit mode:
 *
 *   Sometimes AMD Athlon/Opteron CPUs report invalid exceptions on prefetch.
 *   Check that here and ignore it.
79
 *
Ingo Molnar's avatar
Ingo Molnar committed
80
 * 64-bit mode:
81
 *
Ingo Molnar's avatar
Ingo Molnar committed
82 83 84 85
 *   Sometimes the CPU reports invalid exceptions on prefetch.
 *   Check that here and ignore it.
 *
 * Opcode checker based on code by Richard Brunner.
86
 */
87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112
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.
		 */
113
		return (!user_mode(regs) || user_64bit_mode(regs));
114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133
#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
134 135
static int
is_prefetch(struct pt_regs *regs, unsigned long error_code, unsigned long addr)
136
{
Ingo Molnar's avatar
Ingo Molnar committed
137
	unsigned char *max_instr;
138
	unsigned char *instr;
139
	int prefetch = 0;
Linus Torvalds's avatar
Linus Torvalds committed
140

Ingo Molnar's avatar
Ingo Molnar committed
141 142 143 144
	/*
	 * If it was a exec (instruction fetch) fault on NX page, then
	 * do not ignore the fault:
	 */
145
	if (error_code & PF_INSTR)
Linus Torvalds's avatar
Linus Torvalds committed
146
		return 0;
147

148
	instr = (void *)convert_ip_to_linear(current, regs);
149
	max_instr = instr + 15;
Linus Torvalds's avatar
Linus Torvalds committed
150

151
	if (user_mode(regs) && instr >= (unsigned char *)TASK_SIZE)
Linus Torvalds's avatar
Linus Torvalds committed
152 153
		return 0;

154
	while (instr < max_instr) {
Ingo Molnar's avatar
Ingo Molnar committed
155
		unsigned char opcode;
Linus Torvalds's avatar
Linus Torvalds committed
156

157
		if (probe_kernel_address(instr, opcode))
158
			break;
Linus Torvalds's avatar
Linus Torvalds committed
159 160 161

		instr++;

162
		if (!check_prefetch_opcode(regs, instr, opcode, &prefetch))
Linus Torvalds's avatar
Linus Torvalds committed
163 164 165 166 167
			break;
	}
	return prefetch;
}

Ingo Molnar's avatar
Ingo Molnar committed
168 169
static void
force_sig_info_fault(int si_signo, int si_code, unsigned long address,
170
		     struct task_struct *tsk, int fault)
171
{
172
	unsigned lsb = 0;
173 174
	siginfo_t info;

Ingo Molnar's avatar
Ingo Molnar committed
175 176 177 178
	info.si_signo	= si_signo;
	info.si_errno	= 0;
	info.si_code	= si_code;
	info.si_addr	= (void __user *)address;
179 180 181 182 183
	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
184

185 186 187
	force_sig_info(si_signo, &info, tsk);
}

188 189 190 191 192
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)
193
{
194 195 196 197
	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
198

199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219
	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;

220
	if (!pmd_present(*pmd))
221
		set_pmd(pmd, *pmd_k);
222
	else
223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239
		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;

240
		spin_lock(&pgd_lock);
241
		list_for_each_entry(page, &pgd_list, lru) {
242
			spinlock_t *pgt_lock;
243
			pmd_t *ret;
244

245
			/* the pgt_lock only for Xen */
246 247 248 249 250 251 252
			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)
253 254
				break;
		}
255
		spin_unlock(&pgd_lock);
256 257 258 259 260 261 262 263
	}
}

/*
 * 32-bit:
 *
 *   Handle a fault on the vmalloc or module mapping area
 */
264
static noinline int vmalloc_fault(unsigned long address)
265 266 267 268 269 270 271 272 273
{
	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;

274 275
	WARN_ON_ONCE(in_nmi());

276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293
	/*
	 * 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;
}
294
NOKPROBE_SYMBOL(vmalloc_fault);
295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310

/*
 * 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;
311
}
Linus Torvalds's avatar
Linus Torvalds committed
312

Akinobu Mita's avatar
Akinobu Mita committed
313
static bool low_pfn(unsigned long pfn)
Linus Torvalds's avatar
Linus Torvalds committed
314
{
Akinobu Mita's avatar
Akinobu Mita committed
315 316
	return pfn < max_low_pfn;
}
317

Akinobu Mita's avatar
Akinobu Mita committed
318 319 320 321 322 323
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
324

325
#ifdef CONFIG_X86_PAE
Akinobu Mita's avatar
Akinobu Mita committed
326 327 328
	printk("*pdpt = %016Lx ", pgd_val(*pgd));
	if (!low_pfn(pgd_val(*pgd) >> PAGE_SHIFT) || !pgd_present(*pgd))
		goto out;
329
#endif
Akinobu Mita's avatar
Akinobu Mita committed
330 331
	pmd = pmd_offset(pud_offset(pgd, address), address);
	printk(KERN_CONT "*pde = %0*Lx ", sizeof(*pmd) * 2, (u64)pmd_val(*pmd));
332 333 334 335 336

	/*
	 * 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
337
	 * it's allocated already:
338
	 */
Akinobu Mita's avatar
Akinobu Mita committed
339 340
	if (!low_pfn(pmd_pfn(*pmd)) || !pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
341

Akinobu Mita's avatar
Akinobu Mita committed
342 343 344
	pte = pte_offset_kernel(pmd, address);
	printk("*pte = %0*Lx ", sizeof(*pte) * 2, (u64)pte_val(*pte));
out:
345
	printk("\n");
346 347 348 349 350 351
}

#else /* CONFIG_X86_64: */

void vmalloc_sync_all(void)
{
352
	sync_global_pgds(VMALLOC_START & PGDIR_MASK, VMALLOC_END, 0);
353 354 355 356 357 358 359 360 361
}

/*
 * 64-bit:
 *
 *   Handle a fault on the vmalloc area
 *
 * This assumes no large pages in there.
 */
362
static noinline int vmalloc_fault(unsigned long address)
363 364 365 366 367 368 369 370 371 372
{
	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;

373 374
	WARN_ON_ONCE(in_nmi());

375 376 377 378 379 380 381 382 383 384
	/*
	 * 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;

385
	if (pgd_none(*pgd)) {
386
		set_pgd(pgd, *pgd_ref);
387 388
		arch_flush_lazy_mmu_mode();
	} else {
389
		BUG_ON(pgd_page_vaddr(*pgd) != pgd_page_vaddr(*pgd_ref));
390
	}
391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428

	/*
	 * 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;
}
429
NOKPROBE_SYMBOL(vmalloc_fault);
430

431
#ifdef CONFIG_CPU_SUP_AMD
432
static const char errata93_warning[] =
433 434 435 436 437
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";
438
#endif
439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457

/*
 * 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
458 459
	pgd_t *base = __va(read_cr3() & PHYSICAL_PAGE_MASK);
	pgd_t *pgd = base + pgd_index(address);
Linus Torvalds's avatar
Linus Torvalds committed
460 461 462 463
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;

Ingo Molnar's avatar
Ingo Molnar committed
464 465 466
	if (bad_address(pgd))
		goto bad;

467
	printk("PGD %lx ", pgd_val(*pgd));
Ingo Molnar's avatar
Ingo Molnar committed
468 469 470

	if (!pgd_present(*pgd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
471

472
	pud = pud_offset(pgd, address);
Ingo Molnar's avatar
Ingo Molnar committed
473 474 475
	if (bad_address(pud))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
476
	printk("PUD %lx ", pud_val(*pud));
477
	if (!pud_present(*pud) || pud_large(*pud))
Ingo Molnar's avatar
Ingo Molnar committed
478
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
479 480

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

Linus Torvalds's avatar
Linus Torvalds committed
484
	printk("PMD %lx ", pmd_val(*pmd));
Ingo Molnar's avatar
Ingo Molnar committed
485 486
	if (!pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
487 488

	pte = pte_offset_kernel(pmd, address);
Ingo Molnar's avatar
Ingo Molnar committed
489 490 491
	if (bad_address(pte))
		goto bad;

492
	printk("PTE %lx", pte_val(*pte));
Ingo Molnar's avatar
Ingo Molnar committed
493
out:
Linus Torvalds's avatar
Linus Torvalds committed
494 495 496 497
	printk("\n");
	return;
bad:
	printk("BAD\n");
498 499
}

500
#endif /* CONFIG_X86_64 */
Linus Torvalds's avatar
Linus Torvalds committed
501

Ingo Molnar's avatar
Ingo Molnar committed
502 503 504 505 506 507 508 509 510 511 512 513 514
/*
 * 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.
515
 */
516
static int is_errata93(struct pt_regs *regs, unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
517
{
518 519 520 521 522
#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;

523
	if (address != regs->ip)
Linus Torvalds's avatar
Linus Torvalds committed
524
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
525

526
	if ((address >> 32) != 0)
Linus Torvalds's avatar
Linus Torvalds committed
527
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
528

Linus Torvalds's avatar
Linus Torvalds committed
529
	address |= 0xffffffffUL << 32;
530 531
	if ((address >= (u64)_stext && address <= (u64)_etext) ||
	    (address >= MODULES_VADDR && address <= MODULES_END)) {
532
		printk_once(errata93_warning);
533
		regs->ip = address;
Linus Torvalds's avatar
Linus Torvalds committed
534 535
		return 1;
	}
536
#endif
Linus Torvalds's avatar
Linus Torvalds committed
537
	return 0;
538
}
Linus Torvalds's avatar
Linus Torvalds committed
539

540
/*
Ingo Molnar's avatar
Ingo Molnar committed
541 542 543 544 545
 * 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
546 547 548 549 550
 * 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
551
	if ((regs->cs == __USER32_CS || (regs->cs & (1<<2))) && (address >> 32))
552 553 554 555 556
		return 1;
#endif
	return 0;
}

557 558 559 560
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
561

562
	/*
Ingo Molnar's avatar
Ingo Molnar committed
563
	 * Pentium F0 0F C7 C8 bug workaround:
564
	 */
565
	if (boot_cpu_has_bug(X86_BUG_F00F)) {
566 567 568 569 570 571 572 573 574 575 576
		nr = (address - idt_descr.address) >> 3;

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

577 578
static const char nx_warning[] = KERN_CRIT
"kernel tried to execute NX-protected page - exploit attempt? (uid: %d)\n";
579 580
static const char smep_warning[] = KERN_CRIT
"unable to execute userspace code (SMEP?) (uid: %d)\n";
581

Ingo Molnar's avatar
Ingo Molnar committed
582 583 584
static void
show_fault_oops(struct pt_regs *regs, unsigned long error_code,
		unsigned long address)
585
{
586 587 588 589
	if (!oops_may_print())
		return;

	if (error_code & PF_INSTR) {
590
		unsigned int level;
591 592
		pgd_t *pgd;
		pte_t *pte;
Ingo Molnar's avatar
Ingo Molnar committed
593

594 595 596 597
		pgd = __va(read_cr3() & PHYSICAL_PAGE_MASK);
		pgd += pgd_index(address);

		pte = lookup_address_in_pgd(pgd, address, &level);
598

599
		if (pte && pte_present(*pte) && !pte_exec(*pte))
600
			printk(nx_warning, from_kuid(&init_user_ns, current_uid()));
601 602
		if (pte && pte_present(*pte) && pte_exec(*pte) &&
				(pgd_flags(*pgd) & _PAGE_USER) &&
603
				(__read_cr4() & X86_CR4_SMEP))
604
			printk(smep_warning, from_kuid(&init_user_ns, current_uid()));
605 606
	}

607
	printk(KERN_ALERT "BUG: unable to handle kernel ");
608
	if (address < PAGE_SIZE)
609
		printk(KERN_CONT "NULL pointer dereference");
610
	else
611
		printk(KERN_CONT "paging request");
Ingo Molnar's avatar
Ingo Molnar committed
612

613
	printk(KERN_CONT " at %p\n", (void *) address);
614
	printk(KERN_ALERT "IP:");
615
	printk_address(regs->ip);
Ingo Molnar's avatar
Ingo Molnar committed
616

617 618 619
	dump_pagetable(address);
}

Ingo Molnar's avatar
Ingo Molnar committed
620 621 622
static noinline void
pgtable_bad(struct pt_regs *regs, unsigned long error_code,
	    unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
623
{
Ingo Molnar's avatar
Ingo Molnar committed
624 625 626 627 628 629 630
	struct task_struct *tsk;
	unsigned long flags;
	int sig;

	flags = oops_begin();
	tsk = current;
	sig = SIGKILL;
631

Linus Torvalds's avatar
Linus Torvalds committed
632
	printk(KERN_ALERT "%s: Corrupted page table at address %lx\n",
633
	       tsk->comm, address);
Linus Torvalds's avatar
Linus Torvalds committed
634
	dump_pagetable(address);
Ingo Molnar's avatar
Ingo Molnar committed
635 636

	tsk->thread.cr2		= address;
637
	tsk->thread.trap_nr	= X86_TRAP_PF;
Ingo Molnar's avatar
Ingo Molnar committed
638 639
	tsk->thread.error_code	= error_code;

640
	if (__die("Bad pagetable", regs, error_code))
641
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
642

643
	oops_end(flags, regs, sig);
Linus Torvalds's avatar
Linus Torvalds committed
644 645
}

Ingo Molnar's avatar
Ingo Molnar committed
646 647
static noinline void
no_context(struct pt_regs *regs, unsigned long error_code,
648
	   unsigned long address, int signal, int si_code)
649 650 651 652 653
{
	struct task_struct *tsk = current;
	unsigned long flags;
	int sig;

Ingo Molnar's avatar
Ingo Molnar committed
654
	/* Are we prepared to handle this kernel fault? */
655
	if (fixup_exception(regs)) {
656 657 658 659 660 661 662 663 664 665 666 667 668 669
		/*
		 * Any interrupt that takes a fault gets the fixup. This makes
		 * the below recursive fault logic only apply to a faults from
		 * task context.
		 */
		if (in_interrupt())
			return;

		/*
		 * Per the above we're !in_interrupt(), aka. task context.
		 *
		 * In this case we need to make sure we're not recursively
		 * faulting through the emulate_vsyscall() logic.
		 */
670
		if (current_thread_info()->sig_on_uaccess_error && signal) {
671
			tsk->thread.trap_nr = X86_TRAP_PF;
672 673 674 675 676 677
			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);
		}
678 679 680 681

		/*
		 * Barring that, we can do the fixup and be happy.
		 */
682
		return;
683
	}
684 685

	/*
Ingo Molnar's avatar
Ingo Molnar committed
686 687 688 689 690 691 692
	 * 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:
693
	 *
Ingo Molnar's avatar
Ingo Molnar committed
694
	 *   Hall of shame of CPU/BIOS bugs.
695 696 697 698 699 700 701 702 703
	 */
	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
704
	 * terminate things with extreme prejudice:
705 706 707 708 709
	 */
	flags = oops_begin();

	show_fault_oops(regs, error_code, address);

710
	if (task_stack_end_corrupted(tsk))
711
		printk(KERN_EMERG "Thread overran stack, or stack corrupted\n");
712

713
	tsk->thread.cr2		= address;
714
	tsk->thread.trap_nr	= X86_TRAP_PF;
715
	tsk->thread.error_code	= error_code;
716 717 718 719

	sig = SIGKILL;
	if (__die("Oops", regs, error_code))
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
720

721
	/* Executive summary in case the body of the oops scrolled away */
722
	printk(KERN_DEFAULT "CR2: %016lx\n", address);
Ingo Molnar's avatar
Ingo Molnar committed
723

724 725 726
	oops_end(flags, regs, sig);
}

Ingo Molnar's avatar
Ingo Molnar committed
727 728 729 730 731 732 733 734 735 736 737 738 739 740
/*
 * 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;

741
	printk("%s%s[%d]: segfault at %lx ip %p sp %p error %lx",
Ingo Molnar's avatar
Ingo Molnar committed
742 743 744 745 746 747 748 749 750 751 752 753
		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)
754 755 756 757 758 759
{
	struct task_struct *tsk = current;

	/* User mode accesses just cause a SIGSEGV */
	if (error_code & PF_USER) {
		/*
Ingo Molnar's avatar
Ingo Molnar committed
760
		 * It's possible to have interrupts off here:
761 762 763 764 765
		 */
		local_irq_enable();

		/*
		 * Valid to do another page fault here because this one came
Ingo Molnar's avatar
Ingo Molnar committed
766
		 * from user space:
767 768 769 770 771 772 773
		 */
		if (is_prefetch(regs, error_code, address))
			return;

		if (is_errata100(regs, address))
			return;

774 775 776 777 778 779
#ifdef CONFIG_X86_64
		/*
		 * Instruction fetch faults in the vsyscall page might need
		 * emulation.
		 */
		if (unlikely((error_code & PF_INSTR) &&
780
			     ((address & ~0xfff) == VSYSCALL_ADDR))) {
781 782 783 784
			if (emulate_vsyscall(regs, address))
				return;
		}
#endif
785 786 787
		/* Kernel addresses are always protection faults: */
		if (address >= TASK_SIZE)
			error_code |= PF_PROT;
788

789
		if (likely(show_unhandled_signals))
Ingo Molnar's avatar
Ingo Molnar committed
790 791 792
			show_signal_msg(regs, error_code, address, tsk);

		tsk->thread.cr2		= address;
793
		tsk->thread.error_code	= error_code;
794
		tsk->thread.trap_nr	= X86_TRAP_PF;
795

796
		force_sig_info_fault(SIGSEGV, si_code, address, tsk, 0);
Ingo Molnar's avatar
Ingo Molnar committed
797

798 799 800 801 802 803
		return;
	}

	if (is_f00f_bug(regs, address))
		return;

804
	no_context(regs, error_code, address, SIGSEGV, si_code);
805 806
}

Ingo Molnar's avatar
Ingo Molnar committed
807 808 809
static noinline void
bad_area_nosemaphore(struct pt_regs *regs, unsigned long error_code,
		     unsigned long address)
810 811 812 813
{
	__bad_area_nosemaphore(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
814 815 816
static void
__bad_area(struct pt_regs *regs, unsigned long error_code,
	   unsigned long address, int si_code)
817 818 819 820 821 822 823 824 825 826 827 828
{
	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
829 830
static noinline void
bad_area(struct pt_regs *regs, unsigned long error_code, unsigned long address)
831 832 833 834
{
	__bad_area(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
835 836 837
static noinline void
bad_area_access_error(struct pt_regs *regs, unsigned long error_code,
		      unsigned long address)
838 839 840 841
{
	__bad_area(regs, error_code, address, SEGV_ACCERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
842
static void
843 844
do_sigbus(struct pt_regs *regs, unsigned long error_code, unsigned long address,
	  unsigned int fault)
845 846
{
	struct task_struct *tsk = current;
847
	int code = BUS_ADRERR;
848

Ingo Molnar's avatar
Ingo Molnar committed
849
	/* Kernel mode? Handle exceptions or die: */
850
	if (!(error_code & PF_USER)) {
851
		no_context(regs, error_code, address, SIGBUS, BUS_ADRERR);
852 853
		return;
	}
Ingo Molnar's avatar
Ingo Molnar committed
854

855
	/* User-space => ok to do another page fault: */
856 857
	if (is_prefetch(regs, error_code, address))
		return;
Ingo Molnar's avatar
Ingo Molnar committed
858 859 860

	tsk->thread.cr2		= address;
	tsk->thread.error_code	= error_code;
861
	tsk->thread.trap_nr	= X86_TRAP_PF;
Ingo Molnar's avatar
Ingo Molnar committed
862

863
#ifdef CONFIG_MEMORY_FAILURE
864
	if (fault & (VM_FAULT_HWPOISON|VM_FAULT_HWPOISON_LARGE)) {
865 866 867 868 869 870
		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
871
	force_sig_info_fault(SIGBUS, code, address, tsk, fault);
872 873
}

874
static noinline void
Ingo Molnar's avatar
Ingo Molnar committed
875 876
mm_fault_error(struct pt_regs *regs, unsigned long error_code,
	       unsigned long address, unsigned int fault)
877
{
878 879 880
	if (fatal_signal_pending(current) && !(error_code & PF_USER)) {
		no_context(regs, error_code, address, 0, 0);
		return;
881 882
	}

Ingo Molnar's avatar
Ingo Molnar committed
883
	if (fault & VM_FAULT_OOM) {
884 885
		/* Kernel mode? Handle exceptions or die: */
		if (!(error_code & PF_USER)) {
886 887
			no_context(regs, error_code, address,
				   SIGSEGV, SEGV_MAPERR);
888
			return;
889 890
		}

891 892 893 894 895 896
		/*
		 * We ran out of memory, call the OOM killer, and return the
		 * userspace (which will retry the fault, or kill us if we got
		 * oom-killed):
		 */
		pagefault_out_of_memory();
Ingo Molnar's avatar
Ingo Molnar committed
897
	} else {
898 899
		if (fault & (VM_FAULT_SIGBUS|VM_FAULT_HWPOISON|
			     VM_FAULT_HWPOISON_LARGE))
900
			do_sigbus(regs, error_code, address, fault);
901 902
		else if (fault & VM_FAULT_SIGSEGV)
			bad_area_nosemaphore(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
903 904 905
		else
			BUG();
	}
906 907
}

908 909 910 911
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
912

913 914 915 916 917 918
	if ((error_code & PF_INSTR) && !pte_exec(*pte))
		return 0;

	return 1;
}

919
/*
Ingo Molnar's avatar
Ingo Molnar committed
920 921 922 923 924 925 926 927
 * 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.
 *
928 929 930 931
 * Spurious faults may only occur if the TLB contains an entry with
 * fewer permission than the page table entry.  Non-present (P = 0)
 * and reserved bit (R = 1) faults are never spurious.
 *
932 933
 * There are no security implications to leaving a stale TLB when
 * increasing the permissions on a page.
934 935 936 937 938
 *
 * Returns non-zero if a spurious fault was handled, zero otherwise.
 *
 * See Intel Developer's Manual Vol 3 Section 4.10.4.3, bullet 3
 * (Optional Invalidation).
939
 */
940
static noinline int
Ingo Molnar's avatar
Ingo Molnar committed
941
spurious_fault(unsigned long error_code, unsigned long address)
942 943 944 945 946
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;
947
	int ret;
948

949 950 951 952 953 954 955 956 957 958 959
	/*
	 * Only writes to RO or instruction fetches from NX may cause
	 * spurious faults.
	 *
	 * These could be from user or supervisor accesses but the TLB
	 * is only lazily flushed after a kernel mapping protection
	 * change, so user accesses are not expected to cause spurious
	 * faults.
	 */
	if (error_code != (PF_WRITE | PF_PROT)
	    && error_code != (PF_INSTR | PF_PROT))
960 961 962 963 964 965 966 967 968 969
		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;

970 971 972
	if (pud_large(*pud))
		return spurious_fault_check(error_code, (pte_t *) pud);

973 974 975 976
	pmd = pmd_offset(pud, address);
	if (!pmd_present(*pmd))
		return 0;

977 978 979
	if (pmd_large(*pmd))
		return spurious_fault_check(error_code, (pte_t *) pmd);

980
	pte = pte_offset_kernel(pmd, address);
981
	if (!pte_present(*pte))
982 983
		return 0;

984 985 986 987 988
	ret = spurious_fault_check(error_code, pte);
	if (!ret)
		return 0;

	/*
Ingo Molnar's avatar
Ingo Molnar committed
989 990
	 * Make sure we have permissions in PMD.
	 * If not, then there's a bug in the page tables:
991 992 993
	 */
	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
994

995
	return ret;
996
}
997
NOKPROBE_SYMBOL(spurious_fault);
998

999
int show_unhandled_signals = 1;
Linus Torvalds's avatar
Linus Torvalds committed
1000

Ingo Molnar's avatar
Ingo Molnar committed
1001
static inline int
1002
access_error(unsigned long error_code, struct vm_area_struct *vma)
1003
{