fault.c 32.7 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(), ...	*/
16
#include <linux/uaccess.h>		/* faulthandler_disabled()	*/
Ingo Molnar's avatar
Ingo Molnar committed
17

18 19
#include <asm/traps.h>			/* dotraplinkage, ...		*/
#include <asm/pgalloc.h>		/* pgd_*(), ...			*/
Vegard Nossum's avatar
Vegard Nossum committed
20
#include <asm/kmemcheck.h>		/* kmemcheck_*(), ...		*/
21 22
#include <asm/fixmap.h>			/* VSYSCALL_ADDR		*/
#include <asm/vsyscall.h>		/* emulate_vsyscall		*/
23
#include <asm/vm86.h>			/* struct vm86			*/
Linus Torvalds's avatar
Linus Torvalds committed
24

25 26 27
#define CREATE_TRACE_POINTS
#include <asm/trace/exceptions.h>

28
/*
Ingo Molnar's avatar
Ingo Molnar committed
29 30 31 32 33 34 35
 * 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
36
 *   bit 5 ==				1: protection keys block access
37
 */
Ingo Molnar's avatar
Ingo Molnar committed
38 39 40 41 42 43 44
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,
45
	PF_PK		=		1 << 5,
Ingo Molnar's avatar
Ingo Molnar committed
46
};
47

48
/*
49 50
 * Returns 0 if mmiotrace is disabled, or if the fault is not
 * handled by mmiotrace:
51
 */
52
static nokprobe_inline int
53
kmmio_fault(struct pt_regs *regs, unsigned long addr)
54
{
55 56 57 58
	if (unlikely(is_kmmio_active()))
		if (kmmio_handler(regs, addr) == 1)
			return -1;
	return 0;
59 60
}

61
static nokprobe_inline int kprobes_fault(struct pt_regs *regs)
62
{
63 64 65
	int ret = 0;

	/* kprobe_running() needs smp_processor_id() */
66
	if (kprobes_built_in() && !user_mode(regs)) {
67 68 69 70 71
		preempt_disable();
		if (kprobe_running() && kprobe_fault_handler(regs, 14))
			ret = 1;
		preempt_enable();
	}
72

73
	return ret;
74
}
75

76
/*
Ingo Molnar's avatar
Ingo Molnar committed
77 78 79 80 81 82
 * Prefetch quirks:
 *
 * 32-bit mode:
 *
 *   Sometimes AMD Athlon/Opteron CPUs report invalid exceptions on prefetch.
 *   Check that here and ignore it.
83
 *
Ingo Molnar's avatar
Ingo Molnar committed
84
 * 64-bit mode:
85
 *
Ingo Molnar's avatar
Ingo Molnar committed
86 87 88 89
 *   Sometimes the CPU reports invalid exceptions on prefetch.
 *   Check that here and ignore it.
 *
 * Opcode checker based on code by Richard Brunner.
90
 */
91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116
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.
		 */
117
		return (!user_mode(regs) || user_64bit_mode(regs));
118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137
#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
138 139
static int
is_prefetch(struct pt_regs *regs, unsigned long error_code, unsigned long addr)
140
{
Ingo Molnar's avatar
Ingo Molnar committed
141
	unsigned char *max_instr;
142
	unsigned char *instr;
143
	int prefetch = 0;
Linus Torvalds's avatar
Linus Torvalds committed
144

Ingo Molnar's avatar
Ingo Molnar committed
145 146 147 148
	/*
	 * If it was a exec (instruction fetch) fault on NX page, then
	 * do not ignore the fault:
	 */
149
	if (error_code & PF_INSTR)
Linus Torvalds's avatar
Linus Torvalds committed
150
		return 0;
151

152
	instr = (void *)convert_ip_to_linear(current, regs);
153
	max_instr = instr + 15;
Linus Torvalds's avatar
Linus Torvalds committed
154

155
	if (user_mode(regs) && instr >= (unsigned char *)TASK_SIZE_MAX)
Linus Torvalds's avatar
Linus Torvalds committed
156 157
		return 0;

158
	while (instr < max_instr) {
Ingo Molnar's avatar
Ingo Molnar committed
159
		unsigned char opcode;
Linus Torvalds's avatar
Linus Torvalds committed
160

161
		if (probe_kernel_address(instr, opcode))
162
			break;
Linus Torvalds's avatar
Linus Torvalds committed
163 164 165

		instr++;

166
		if (!check_prefetch_opcode(regs, instr, opcode, &prefetch))
Linus Torvalds's avatar
Linus Torvalds committed
167 168 169 170 171
			break;
	}
	return prefetch;
}

Ingo Molnar's avatar
Ingo Molnar committed
172 173
static void
force_sig_info_fault(int si_signo, int si_code, unsigned long address,
174
		     struct task_struct *tsk, int fault)
175
{
176
	unsigned lsb = 0;
177 178
	siginfo_t info;

Ingo Molnar's avatar
Ingo Molnar committed
179 180 181 182
	info.si_signo	= si_signo;
	info.si_errno	= 0;
	info.si_code	= si_code;
	info.si_addr	= (void __user *)address;
183 184 185 186 187
	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
188

189 190 191
	force_sig_info(si_signo, &info, tsk);
}

192 193 194 195 196
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)
197
{
198 199 200 201
	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
202

203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223
	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;

224
	if (!pmd_present(*pmd))
225
		set_pmd(pmd, *pmd_k);
226
	else
227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243
		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;

244
		spin_lock(&pgd_lock);
245
		list_for_each_entry(page, &pgd_list, lru) {
246
			spinlock_t *pgt_lock;
247
			pmd_t *ret;
248

249
			/* the pgt_lock only for Xen */
250 251 252 253 254 255 256
			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)
257 258
				break;
		}
259
		spin_unlock(&pgd_lock);
260 261 262 263 264 265 266 267
	}
}

/*
 * 32-bit:
 *
 *   Handle a fault on the vmalloc or module mapping area
 */
268
static noinline int vmalloc_fault(unsigned long address)
269 270 271 272 273 274 275 276 277
{
	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;

278 279
	WARN_ON_ONCE(in_nmi());

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

/*
 * 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)
{
307
#ifdef CONFIG_VM86
308 309
	unsigned long bit;

310
	if (!v8086_mode(regs) || !tsk->thread.vm86)
311 312 313 314
		return;

	bit = (address - 0xA0000) >> PAGE_SHIFT;
	if (bit < 32)
315 316
		tsk->thread.vm86->screen_bitmap |= 1 << bit;
#endif
317
}
Linus Torvalds's avatar
Linus Torvalds committed
318

Akinobu Mita's avatar
Akinobu Mita committed
319
static bool low_pfn(unsigned long pfn)
Linus Torvalds's avatar
Linus Torvalds committed
320
{
Akinobu Mita's avatar
Akinobu Mita committed
321 322
	return pfn < max_low_pfn;
}
323

Akinobu Mita's avatar
Akinobu Mita committed
324 325 326 327 328 329
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
330

331
#ifdef CONFIG_X86_PAE
Akinobu Mita's avatar
Akinobu Mita committed
332 333 334
	printk("*pdpt = %016Lx ", pgd_val(*pgd));
	if (!low_pfn(pgd_val(*pgd) >> PAGE_SHIFT) || !pgd_present(*pgd))
		goto out;
335
#endif
Akinobu Mita's avatar
Akinobu Mita committed
336 337
	pmd = pmd_offset(pud_offset(pgd, address), address);
	printk(KERN_CONT "*pde = %0*Lx ", sizeof(*pmd) * 2, (u64)pmd_val(*pmd));
338 339 340 341 342

	/*
	 * 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
343
	 * it's allocated already:
344
	 */
Akinobu Mita's avatar
Akinobu Mita committed
345 346
	if (!low_pfn(pmd_pfn(*pmd)) || !pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
347

Akinobu Mita's avatar
Akinobu Mita committed
348 349 350
	pte = pte_offset_kernel(pmd, address);
	printk("*pte = %0*Lx ", sizeof(*pte) * 2, (u64)pte_val(*pte));
out:
351
	printk("\n");
352 353 354 355 356 357
}

#else /* CONFIG_X86_64: */

void vmalloc_sync_all(void)
{
358
	sync_global_pgds(VMALLOC_START & PGDIR_MASK, VMALLOC_END, 0);
359 360 361 362 363 364 365 366 367
}

/*
 * 64-bit:
 *
 *   Handle a fault on the vmalloc area
 *
 * This assumes no large pages in there.
 */
368
static noinline int vmalloc_fault(unsigned long address)
369 370 371 372 373 374 375 376 377 378
{
	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;

379 380
	WARN_ON_ONCE(in_nmi());

381 382 383 384 385 386 387 388 389 390
	/*
	 * 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;

391
	if (pgd_none(*pgd)) {
392
		set_pgd(pgd, *pgd_ref);
393 394
		arch_flush_lazy_mmu_mode();
	} else {
395
		BUG_ON(pgd_page_vaddr(*pgd) != pgd_page_vaddr(*pgd_ref));
396
	}
397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434

	/*
	 * 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;
}
435
NOKPROBE_SYMBOL(vmalloc_fault);
436

437
#ifdef CONFIG_CPU_SUP_AMD
438
static const char errata93_warning[] =
439 440 441 442 443
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";
444
#endif
445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463

/*
 * 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
464 465
	pgd_t *base = __va(read_cr3() & PHYSICAL_PAGE_MASK);
	pgd_t *pgd = base + pgd_index(address);
Linus Torvalds's avatar
Linus Torvalds committed
466 467 468 469
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;

Ingo Molnar's avatar
Ingo Molnar committed
470 471 472
	if (bad_address(pgd))
		goto bad;

473
	printk("PGD %lx ", pgd_val(*pgd));
Ingo Molnar's avatar
Ingo Molnar committed
474 475 476

	if (!pgd_present(*pgd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
477

478
	pud = pud_offset(pgd, address);
Ingo Molnar's avatar
Ingo Molnar committed
479 480 481
	if (bad_address(pud))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
482
	printk("PUD %lx ", pud_val(*pud));
483
	if (!pud_present(*pud) || pud_large(*pud))
Ingo Molnar's avatar
Ingo Molnar committed
484
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
485 486

	pmd = pmd_offset(pud, address);
Ingo Molnar's avatar
Ingo Molnar committed
487 488 489
	if (bad_address(pmd))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
490
	printk("PMD %lx ", pmd_val(*pmd));
Ingo Molnar's avatar
Ingo Molnar committed
491 492
	if (!pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
493 494

	pte = pte_offset_kernel(pmd, address);
Ingo Molnar's avatar
Ingo Molnar committed
495 496 497
	if (bad_address(pte))
		goto bad;

498
	printk("PTE %lx", pte_val(*pte));
Ingo Molnar's avatar
Ingo Molnar committed
499
out:
Linus Torvalds's avatar
Linus Torvalds committed
500 501 502 503
	printk("\n");
	return;
bad:
	printk("BAD\n");
504 505
}

506
#endif /* CONFIG_X86_64 */
Linus Torvalds's avatar
Linus Torvalds committed
507

Ingo Molnar's avatar
Ingo Molnar committed
508 509 510 511 512 513 514 515 516 517 518 519 520
/*
 * 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.
521
 */
522
static int is_errata93(struct pt_regs *regs, unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
523
{
524 525 526 527 528
#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;

529
	if (address != regs->ip)
Linus Torvalds's avatar
Linus Torvalds committed
530
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
531

532
	if ((address >> 32) != 0)
Linus Torvalds's avatar
Linus Torvalds committed
533
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
534

Linus Torvalds's avatar
Linus Torvalds committed
535
	address |= 0xffffffffUL << 32;
536 537
	if ((address >= (u64)_stext && address <= (u64)_etext) ||
	    (address >= MODULES_VADDR && address <= MODULES_END)) {
538
		printk_once(errata93_warning);
539
		regs->ip = address;
Linus Torvalds's avatar
Linus Torvalds committed
540 541
		return 1;
	}
542
#endif
Linus Torvalds's avatar
Linus Torvalds committed
543
	return 0;
544
}
Linus Torvalds's avatar
Linus Torvalds committed
545

546
/*
Ingo Molnar's avatar
Ingo Molnar committed
547 548 549 550 551
 * 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
552 553 554 555 556
 * 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
557
	if ((regs->cs == __USER32_CS || (regs->cs & (1<<2))) && (address >> 32))
558 559 560 561 562
		return 1;
#endif
	return 0;
}

563 564 565 566
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
567

568
	/*
Ingo Molnar's avatar
Ingo Molnar committed
569
	 * Pentium F0 0F C7 C8 bug workaround:
570
	 */
571
	if (boot_cpu_has_bug(X86_BUG_F00F)) {
572 573 574 575 576 577 578 579 580 581 582
		nr = (address - idt_descr.address) >> 3;

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

583 584
static const char nx_warning[] = KERN_CRIT
"kernel tried to execute NX-protected page - exploit attempt? (uid: %d)\n";
585 586
static const char smep_warning[] = KERN_CRIT
"unable to execute userspace code (SMEP?) (uid: %d)\n";
587

Ingo Molnar's avatar
Ingo Molnar committed
588 589 590
static void
show_fault_oops(struct pt_regs *regs, unsigned long error_code,
		unsigned long address)
591
{
592 593 594 595
	if (!oops_may_print())
		return;

	if (error_code & PF_INSTR) {
596
		unsigned int level;
597 598
		pgd_t *pgd;
		pte_t *pte;
Ingo Molnar's avatar
Ingo Molnar committed
599

600 601 602 603
		pgd = __va(read_cr3() & PHYSICAL_PAGE_MASK);
		pgd += pgd_index(address);

		pte = lookup_address_in_pgd(pgd, address, &level);
604

605
		if (pte && pte_present(*pte) && !pte_exec(*pte))
606
			printk(nx_warning, from_kuid(&init_user_ns, current_uid()));
607 608
		if (pte && pte_present(*pte) && pte_exec(*pte) &&
				(pgd_flags(*pgd) & _PAGE_USER) &&
609
				(__read_cr4() & X86_CR4_SMEP))
610
			printk(smep_warning, from_kuid(&init_user_ns, current_uid()));
611 612
	}

613
	printk(KERN_ALERT "BUG: unable to handle kernel ");
614
	if (address < PAGE_SIZE)
615
		printk(KERN_CONT "NULL pointer dereference");
616
	else
617
		printk(KERN_CONT "paging request");
Ingo Molnar's avatar
Ingo Molnar committed
618

619
	printk(KERN_CONT " at %p\n", (void *) address);
620
	printk(KERN_ALERT "IP:");
621
	printk_address(regs->ip);
Ingo Molnar's avatar
Ingo Molnar committed
622

623 624 625
	dump_pagetable(address);
}

Ingo Molnar's avatar
Ingo Molnar committed
626 627 628
static noinline void
pgtable_bad(struct pt_regs *regs, unsigned long error_code,
	    unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
629
{
Ingo Molnar's avatar
Ingo Molnar committed
630 631 632 633 634 635 636
	struct task_struct *tsk;
	unsigned long flags;
	int sig;

	flags = oops_begin();
	tsk = current;
	sig = SIGKILL;
637

Linus Torvalds's avatar
Linus Torvalds committed
638
	printk(KERN_ALERT "%s: Corrupted page table at address %lx\n",
639
	       tsk->comm, address);
Linus Torvalds's avatar
Linus Torvalds committed
640
	dump_pagetable(address);
Ingo Molnar's avatar
Ingo Molnar committed
641 642

	tsk->thread.cr2		= address;
643
	tsk->thread.trap_nr	= X86_TRAP_PF;
Ingo Molnar's avatar
Ingo Molnar committed
644 645
	tsk->thread.error_code	= error_code;

646
	if (__die("Bad pagetable", regs, error_code))
647
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
648

649
	oops_end(flags, regs, sig);
Linus Torvalds's avatar
Linus Torvalds committed
650 651
}

Ingo Molnar's avatar
Ingo Molnar committed
652 653
static noinline void
no_context(struct pt_regs *regs, unsigned long error_code,
654
	   unsigned long address, int signal, int si_code)
655 656 657 658 659
{
	struct task_struct *tsk = current;
	unsigned long flags;
	int sig;

Ingo Molnar's avatar
Ingo Molnar committed
660
	/* Are we prepared to handle this kernel fault? */
661
	if (fixup_exception(regs)) {
662 663 664 665 666 667 668 669 670 671 672 673 674 675
		/*
		 * 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.
		 */
676
		if (current_thread_info()->sig_on_uaccess_error && signal) {
677
			tsk->thread.trap_nr = X86_TRAP_PF;
678 679 680 681 682 683
			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);
		}
684 685 686 687

		/*
		 * Barring that, we can do the fixup and be happy.
		 */
688
		return;
689
	}
690 691

	/*
Ingo Molnar's avatar
Ingo Molnar committed
692 693 694 695 696 697 698
	 * 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:
699
	 *
Ingo Molnar's avatar
Ingo Molnar committed
700
	 *   Hall of shame of CPU/BIOS bugs.
701 702 703 704 705 706 707 708 709
	 */
	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
710
	 * terminate things with extreme prejudice:
711 712 713 714 715
	 */
	flags = oops_begin();

	show_fault_oops(regs, error_code, address);

716
	if (task_stack_end_corrupted(tsk))
717
		printk(KERN_EMERG "Thread overran stack, or stack corrupted\n");
718

719
	tsk->thread.cr2		= address;
720
	tsk->thread.trap_nr	= X86_TRAP_PF;
721
	tsk->thread.error_code	= error_code;
722 723 724 725

	sig = SIGKILL;
	if (__die("Oops", regs, error_code))
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
726

727
	/* Executive summary in case the body of the oops scrolled away */
728
	printk(KERN_DEFAULT "CR2: %016lx\n", address);
Ingo Molnar's avatar
Ingo Molnar committed
729

730 731 732
	oops_end(flags, regs, sig);
}

Ingo Molnar's avatar
Ingo Molnar committed
733 734 735 736 737 738 739 740 741 742 743 744 745 746
/*
 * 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;

747
	printk("%s%s[%d]: segfault at %lx ip %p sp %p error %lx",
Ingo Molnar's avatar
Ingo Molnar committed
748 749 750 751 752 753 754 755 756 757 758 759
		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)
760 761 762 763 764 765
{
	struct task_struct *tsk = current;

	/* User mode accesses just cause a SIGSEGV */
	if (error_code & PF_USER) {
		/*
Ingo Molnar's avatar
Ingo Molnar committed
766
		 * It's possible to have interrupts off here:
767 768 769 770 771
		 */
		local_irq_enable();

		/*
		 * Valid to do another page fault here because this one came
Ingo Molnar's avatar
Ingo Molnar committed
772
		 * from user space:
773 774 775 776 777 778 779
		 */
		if (is_prefetch(regs, error_code, address))
			return;

		if (is_errata100(regs, address))
			return;

780 781 782 783 784 785
#ifdef CONFIG_X86_64
		/*
		 * Instruction fetch faults in the vsyscall page might need
		 * emulation.
		 */
		if (unlikely((error_code & PF_INSTR) &&
786
			     ((address & ~0xfff) == VSYSCALL_ADDR))) {
787 788 789 790
			if (emulate_vsyscall(regs, address))
				return;
		}
#endif
791 792 793
		/* Kernel addresses are always protection faults: */
		if (address >= TASK_SIZE)
			error_code |= PF_PROT;
794

795
		if (likely(show_unhandled_signals))
Ingo Molnar's avatar
Ingo Molnar committed
796 797 798
			show_signal_msg(regs, error_code, address, tsk);

		tsk->thread.cr2		= address;
799
		tsk->thread.error_code	= error_code;
800
		tsk->thread.trap_nr	= X86_TRAP_PF;
801

802
		force_sig_info_fault(SIGSEGV, si_code, address, tsk, 0);
Ingo Molnar's avatar
Ingo Molnar committed
803

804 805 806 807 808 809
		return;
	}

	if (is_f00f_bug(regs, address))
		return;

810
	no_context(regs, error_code, address, SIGSEGV, si_code);
811 812
}

Ingo Molnar's avatar
Ingo Molnar committed
813 814 815
static noinline void
bad_area_nosemaphore(struct pt_regs *regs, unsigned long error_code,
		     unsigned long address)
816 817 818 819
{
	__bad_area_nosemaphore(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
820 821 822
static void
__bad_area(struct pt_regs *regs, unsigned long error_code,
	   unsigned long address, int si_code)
823 824 825 826 827 828 829 830 831 832 833 834
{
	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
835 836
static noinline void
bad_area(struct pt_regs *regs, unsigned long error_code, unsigned long address)
837 838 839 840
{
	__bad_area(regs, error_code, address, SEGV_MAPERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
841 842 843
static noinline void
bad_area_access_error(struct pt_regs *regs, unsigned long error_code,
		      unsigned long address)
844 845 846 847
{
	__bad_area(regs, error_code, address, SEGV_ACCERR);
}

Ingo Molnar's avatar
Ingo Molnar committed
848
static void
849 850
do_sigbus(struct pt_regs *regs, unsigned long error_code, unsigned long address,
	  unsigned int fault)
851 852
{
	struct task_struct *tsk = current;
853
	int code = BUS_ADRERR;
854

Ingo Molnar's avatar
Ingo Molnar committed
855
	/* Kernel mode? Handle exceptions or die: */
856
	if (!(error_code & PF_USER)) {
857
		no_context(regs, error_code, address, SIGBUS, BUS_ADRERR);
858 859
		return;
	}
Ingo Molnar's avatar
Ingo Molnar committed
860

861
	/* User-space => ok to do another page fault: */
862 863
	if (is_prefetch(regs, error_code, address))
		return;
Ingo Molnar's avatar
Ingo Molnar committed
864 865 866

	tsk->thread.cr2		= address;
	tsk->thread.error_code	= error_code;
867
	tsk->thread.trap_nr	= X86_TRAP_PF;
Ingo Molnar's avatar
Ingo Molnar committed
868

869
#ifdef CONFIG_MEMORY_FAILURE
870
	if (fault & (VM_FAULT_HWPOISON|VM_FAULT_HWPOISON_LARGE)) {
871 872 873 874 875 876
		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
877
	force_sig_info_fault(SIGBUS, code, address, tsk, fault);
878 879
}

880
static noinline void
Ingo Molnar's avatar
Ingo Molnar committed
881 882
mm_fault_error(struct pt_regs *regs, unsigned long error_code,
	       unsigned long address, unsigned int fault)
883
{
884 885 886
	if (fatal_signal_pending(current) && !(error_code & PF_USER)) {
		no_context(regs, error_code, address, 0, 0);
		return;
887 888
	}

Ingo Molnar's avatar
Ingo Molnar committed
889
	if (fault & VM_FAULT_OOM) {
890 891
		/* Kernel mode? Handle exceptions or die: */
		if (!(error_code & PF_USER)) {
892 893
			no_context(regs, error_code, address,
				   SIGSEGV, SEGV_MAPERR);
894
			return;
895 896
		}

897 898 899 900 901 902
		/*
		 * 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
903
	} else {
904 905
		if (fault & (VM_FAULT_SIGBUS|VM_FAULT_HWPOISON|
			     VM_FAULT_HWPOISON_LARGE))
906
			do_sigbus(regs, error_code, address, fault);
907 908
		else if (fault & VM_FAULT_SIGSEGV)
			bad_area_nosemaphore(regs, error_code, address);
Ingo Molnar's avatar
Ingo Molnar committed
909 910 911
		else
			BUG();
	}
912 913
}

914 915 916 917
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
918

919 920
	if ((error_code & PF_INSTR) && !pte_exec(*pte))
		return 0;
921 922 923 924 925 926
	/*
	 * Note: We do not do lazy flushing on protection key
	 * changes, so no spurious fault will ever set PF_PK.
	 */
	if ((error_code & PF_PK))
		return 1;
927 928 929 930

	return 1;
}

931
/*
Ingo Molnar's avatar
Ingo Molnar committed
932 933 934 935 936 937 938 939
 * 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.
 *
940 941 942 943
 * 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.
 *
944 945
 * There are no security implications to leaving a stale TLB when
 * increasing the permissions on a page.
946 947 948 949 950
 *
 * 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).
951
 */
952
static noinline int
Ingo Molnar's avatar
Ingo Molnar committed
953
spurious_fault(unsigned long error_code, unsigned long address)
954 955 956 957 958
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;
959
	int ret;
960

961 962 963 964 965 966 967 968 969 970 971
	/*
	 * 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))
972 973 974 975 976 977 978 979 980 981
		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;

982 983 984
	if (pud_large(*pud))
		return spurious_fault_check(error_code, (pte_t *) pud);

985 986 987 988
	pmd = pmd_offset(pud, address);
	if (!pmd_present(*pmd))
		return 0;

989 990 991
	if (pmd_large(*pmd))
		return spurious_fault_check(error_code, (pte_t *) pmd);

992
	pte = pte_offset_kernel(pmd, address);
993
	if (!pte_present(*pte))
994 995
		return 0;

996 997 998 999 1000
	ret = spurious_fault_check(error_code, pte);
	if (!ret)
		return 0;

	/*
Ingo Molnar's avatar
Ingo Molnar committed
1