fault.c 33.1 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,
174 175
		     struct task_struct *tsk, struct vm_area_struct *vma,
		     int fault)
176
{
177
	unsigned lsb = 0;
178 179
	siginfo_t info;

Ingo Molnar's avatar
Ingo Molnar committed
180 181 182 183
	info.si_signo	= si_signo;
	info.si_errno	= 0;
	info.si_code	= si_code;
	info.si_addr	= (void __user *)address;
184 185 186 187 188
	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
189

190 191 192
	force_sig_info(si_signo, &info, tsk);
}

193 194 195 196 197
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)
198
{
199 200 201 202
	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
203

204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224
	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;

225
	if (!pmd_present(*pmd))
226
		set_pmd(pmd, *pmd_k);
227
	else
228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244
		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;

245
		spin_lock(&pgd_lock);
246
		list_for_each_entry(page, &pgd_list, lru) {
247
			spinlock_t *pgt_lock;
248
			pmd_t *ret;
249

250
			/* the pgt_lock only for Xen */
251 252 253 254 255 256 257
			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)
258 259
				break;
		}
260
		spin_unlock(&pgd_lock);
261 262 263 264 265 266 267 268
	}
}

/*
 * 32-bit:
 *
 *   Handle a fault on the vmalloc or module mapping area
 */
269
static noinline int vmalloc_fault(unsigned long address)
270 271 272 273 274 275 276 277 278
{
	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;

279 280
	WARN_ON_ONCE(in_nmi());

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

/*
 * 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)
{
308
#ifdef CONFIG_VM86
309 310
	unsigned long bit;

311
	if (!v8086_mode(regs) || !tsk->thread.vm86)
312 313 314 315
		return;

	bit = (address - 0xA0000) >> PAGE_SHIFT;
	if (bit < 32)
316 317
		tsk->thread.vm86->screen_bitmap |= 1 << bit;
#endif
318
}
Linus Torvalds's avatar
Linus Torvalds committed
319

Akinobu Mita's avatar
Akinobu Mita committed
320
static bool low_pfn(unsigned long pfn)
Linus Torvalds's avatar
Linus Torvalds committed
321
{
Akinobu Mita's avatar
Akinobu Mita committed
322 323
	return pfn < max_low_pfn;
}
324

Akinobu Mita's avatar
Akinobu Mita committed
325 326 327 328 329 330
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
331

332
#ifdef CONFIG_X86_PAE
Akinobu Mita's avatar
Akinobu Mita committed
333 334 335
	printk("*pdpt = %016Lx ", pgd_val(*pgd));
	if (!low_pfn(pgd_val(*pgd) >> PAGE_SHIFT) || !pgd_present(*pgd))
		goto out;
336
#endif
Akinobu Mita's avatar
Akinobu Mita committed
337 338
	pmd = pmd_offset(pud_offset(pgd, address), address);
	printk(KERN_CONT "*pde = %0*Lx ", sizeof(*pmd) * 2, (u64)pmd_val(*pmd));
339 340 341 342 343

	/*
	 * 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
344
	 * it's allocated already:
345
	 */
Akinobu Mita's avatar
Akinobu Mita committed
346 347
	if (!low_pfn(pmd_pfn(*pmd)) || !pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
348

Akinobu Mita's avatar
Akinobu Mita committed
349 350 351
	pte = pte_offset_kernel(pmd, address);
	printk("*pte = %0*Lx ", sizeof(*pte) * 2, (u64)pte_val(*pte));
out:
352
	printk("\n");
353 354 355 356 357 358
}

#else /* CONFIG_X86_64: */

void vmalloc_sync_all(void)
{
359
	sync_global_pgds(VMALLOC_START & PGDIR_MASK, VMALLOC_END, 0);
360 361 362 363 364 365 366 367 368
}

/*
 * 64-bit:
 *
 *   Handle a fault on the vmalloc area
 *
 * This assumes no large pages in there.
 */
369
static noinline int vmalloc_fault(unsigned long address)
370 371 372 373 374 375 376 377 378 379
{
	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;

380 381
	WARN_ON_ONCE(in_nmi());

382 383 384 385 386 387 388 389 390 391
	/*
	 * 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;

392
	if (pgd_none(*pgd)) {
393
		set_pgd(pgd, *pgd_ref);
394 395
		arch_flush_lazy_mmu_mode();
	} else {
396
		BUG_ON(pgd_page_vaddr(*pgd) != pgd_page_vaddr(*pgd_ref));
397
	}
398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435

	/*
	 * 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;
}
436
NOKPROBE_SYMBOL(vmalloc_fault);
437

438
#ifdef CONFIG_CPU_SUP_AMD
439
static const char errata93_warning[] =
440 441 442 443 444
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";
445
#endif
446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464

/*
 * 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
465 466
	pgd_t *base = __va(read_cr3() & PHYSICAL_PAGE_MASK);
	pgd_t *pgd = base + pgd_index(address);
Linus Torvalds's avatar
Linus Torvalds committed
467 468 469 470
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;

Ingo Molnar's avatar
Ingo Molnar committed
471 472 473
	if (bad_address(pgd))
		goto bad;

474
	printk("PGD %lx ", pgd_val(*pgd));
Ingo Molnar's avatar
Ingo Molnar committed
475 476 477

	if (!pgd_present(*pgd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
478

479
	pud = pud_offset(pgd, address);
Ingo Molnar's avatar
Ingo Molnar committed
480 481 482
	if (bad_address(pud))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
483
	printk("PUD %lx ", pud_val(*pud));
484
	if (!pud_present(*pud) || pud_large(*pud))
Ingo Molnar's avatar
Ingo Molnar committed
485
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
486 487

	pmd = pmd_offset(pud, address);
Ingo Molnar's avatar
Ingo Molnar committed
488 489 490
	if (bad_address(pmd))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
491
	printk("PMD %lx ", pmd_val(*pmd));
Ingo Molnar's avatar
Ingo Molnar committed
492 493
	if (!pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
494 495

	pte = pte_offset_kernel(pmd, address);
Ingo Molnar's avatar
Ingo Molnar committed
496 497 498
	if (bad_address(pte))
		goto bad;

499
	printk("PTE %lx", pte_val(*pte));
Ingo Molnar's avatar
Ingo Molnar committed
500
out:
Linus Torvalds's avatar
Linus Torvalds committed
501 502 503 504
	printk("\n");
	return;
bad:
	printk("BAD\n");
505 506
}

507
#endif /* CONFIG_X86_64 */
Linus Torvalds's avatar
Linus Torvalds committed
508

Ingo Molnar's avatar
Ingo Molnar committed
509 510 511 512 513 514 515 516 517 518 519 520 521
/*
 * 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.
522
 */
523
static int is_errata93(struct pt_regs *regs, unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
524
{
525 526 527 528 529
#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;

530
	if (address != regs->ip)
Linus Torvalds's avatar
Linus Torvalds committed
531
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
532

533
	if ((address >> 32) != 0)
Linus Torvalds's avatar
Linus Torvalds committed
534
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
535

Linus Torvalds's avatar
Linus Torvalds committed
536
	address |= 0xffffffffUL << 32;
537 538
	if ((address >= (u64)_stext && address <= (u64)_etext) ||
	    (address >= MODULES_VADDR && address <= MODULES_END)) {
539
		printk_once(errata93_warning);
540
		regs->ip = address;
Linus Torvalds's avatar
Linus Torvalds committed
541 542
		return 1;
	}
543
#endif
Linus Torvalds's avatar
Linus Torvalds committed
544
	return 0;
545
}
Linus Torvalds's avatar
Linus Torvalds committed
546

547
/*
Ingo Molnar's avatar
Ingo Molnar committed
548 549 550 551 552
 * 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
553 554 555 556 557
 * 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
558
	if ((regs->cs == __USER32_CS || (regs->cs & (1<<2))) && (address >> 32))
559 560 561 562 563
		return 1;
#endif
	return 0;
}

564 565 566 567
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
568

569
	/*
Ingo Molnar's avatar
Ingo Molnar committed
570
	 * Pentium F0 0F C7 C8 bug workaround:
571
	 */
572
	if (boot_cpu_has_bug(X86_BUG_F00F)) {
573 574 575 576 577 578 579 580 581 582 583
		nr = (address - idt_descr.address) >> 3;

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

584 585
static const char nx_warning[] = KERN_CRIT
"kernel tried to execute NX-protected page - exploit attempt? (uid: %d)\n";
586 587
static const char smep_warning[] = KERN_CRIT
"unable to execute userspace code (SMEP?) (uid: %d)\n";
588

Ingo Molnar's avatar
Ingo Molnar committed
589 590 591
static void
show_fault_oops(struct pt_regs *regs, unsigned long error_code,
		unsigned long address)
592
{
593 594 595 596
	if (!oops_may_print())
		return;

	if (error_code & PF_INSTR) {
597
		unsigned int level;
598 599
		pgd_t *pgd;
		pte_t *pte;
Ingo Molnar's avatar
Ingo Molnar committed
600

601 602 603 604
		pgd = __va(read_cr3() & PHYSICAL_PAGE_MASK);
		pgd += pgd_index(address);

		pte = lookup_address_in_pgd(pgd, address, &level);
605

606
		if (pte && pte_present(*pte) && !pte_exec(*pte))
607
			printk(nx_warning, from_kuid(&init_user_ns, current_uid()));
608 609
		if (pte && pte_present(*pte) && pte_exec(*pte) &&
				(pgd_flags(*pgd) & _PAGE_USER) &&
610
				(__read_cr4() & X86_CR4_SMEP))
611
			printk(smep_warning, from_kuid(&init_user_ns, current_uid()));
612 613
	}

614
	printk(KERN_ALERT "BUG: unable to handle kernel ");
615
	if (address < PAGE_SIZE)
616
		printk(KERN_CONT "NULL pointer dereference");
617
	else
618
		printk(KERN_CONT "paging request");
Ingo Molnar's avatar
Ingo Molnar committed
619

620
	printk(KERN_CONT " at %p\n", (void *) address);
621
	printk(KERN_ALERT "IP:");
622
	printk_address(regs->ip);
Ingo Molnar's avatar
Ingo Molnar committed
623

624 625 626
	dump_pagetable(address);
}

Ingo Molnar's avatar
Ingo Molnar committed
627 628 629
static noinline void
pgtable_bad(struct pt_regs *regs, unsigned long error_code,
	    unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
630
{
Ingo Molnar's avatar
Ingo Molnar committed
631 632 633 634 635 636 637
	struct task_struct *tsk;
	unsigned long flags;
	int sig;

	flags = oops_begin();
	tsk = current;
	sig = SIGKILL;
638

Linus Torvalds's avatar
Linus Torvalds committed
639
	printk(KERN_ALERT "%s: Corrupted page table at address %lx\n",
640
	       tsk->comm, address);
Linus Torvalds's avatar
Linus Torvalds committed
641
	dump_pagetable(address);
Ingo Molnar's avatar
Ingo Molnar committed
642 643

	tsk->thread.cr2		= address;
644
	tsk->thread.trap_nr	= X86_TRAP_PF;
Ingo Molnar's avatar
Ingo Molnar committed
645 646
	tsk->thread.error_code	= error_code;

647
	if (__die("Bad pagetable", regs, error_code))
648
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
649

650
	oops_end(flags, regs, sig);
Linus Torvalds's avatar
Linus Torvalds committed
651 652
}

Ingo Molnar's avatar
Ingo Molnar committed
653 654
static noinline void
no_context(struct pt_regs *regs, unsigned long error_code,
655
	   unsigned long address, int signal, int si_code)
656 657 658 659
{
	struct task_struct *tsk = current;
	unsigned long flags;
	int sig;
660 661
	/* No context means no VMA to pass down */
	struct vm_area_struct *vma = NULL;
662

Ingo Molnar's avatar
Ingo Molnar committed
663
	/* Are we prepared to handle this kernel fault? */
664
	if (fixup_exception(regs)) {
665 666 667 668 669 670 671 672 673 674 675 676 677 678
		/*
		 * 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.
		 */
679
		if (current_thread_info()->sig_on_uaccess_error && signal) {
680
			tsk->thread.trap_nr = X86_TRAP_PF;
681 682 683 684
			tsk->thread.error_code = error_code | PF_USER;
			tsk->thread.cr2 = address;

			/* XXX: hwpoison faults will set the wrong code. */
685 686
			force_sig_info_fault(signal, si_code, address,
					     tsk, vma, 0);
687
		}
688 689 690 691

		/*
		 * Barring that, we can do the fixup and be happy.
		 */
692
		return;
693
	}
694 695

	/*
Ingo Molnar's avatar
Ingo Molnar committed
696 697 698 699 700 701 702
	 * 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:
703
	 *
Ingo Molnar's avatar
Ingo Molnar committed
704
	 *   Hall of shame of CPU/BIOS bugs.
705 706 707 708 709 710 711 712 713
	 */
	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
714
	 * terminate things with extreme prejudice:
715 716 717 718 719
	 */
	flags = oops_begin();

	show_fault_oops(regs, error_code, address);

720
	if (task_stack_end_corrupted(tsk))
721
		printk(KERN_EMERG "Thread overran stack, or stack corrupted\n");
722

723
	tsk->thread.cr2		= address;
724
	tsk->thread.trap_nr	= X86_TRAP_PF;
725
	tsk->thread.error_code	= error_code;
726 727 728 729

	sig = SIGKILL;
	if (__die("Oops", regs, error_code))
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
730

731
	/* Executive summary in case the body of the oops scrolled away */
732
	printk(KERN_DEFAULT "CR2: %016lx\n", address);
Ingo Molnar's avatar
Ingo Molnar committed
733

734 735 736
	oops_end(flags, regs, sig);
}

Ingo Molnar's avatar
Ingo Molnar committed
737 738 739 740 741 742 743 744 745 746 747 748 749 750
/*
 * 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;

751
	printk("%s%s[%d]: segfault at %lx ip %p sp %p error %lx",
Ingo Molnar's avatar
Ingo Molnar committed
752 753 754 755 756 757 758 759 760 761 762
		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,
763 764
		       unsigned long address, struct vm_area_struct *vma,
		       int si_code)
765 766 767 768 769 770
{
	struct task_struct *tsk = current;

	/* User mode accesses just cause a SIGSEGV */
	if (error_code & PF_USER) {
		/*
Ingo Molnar's avatar
Ingo Molnar committed
771
		 * It's possible to have interrupts off here:
772 773 774 775 776
		 */
		local_irq_enable();

		/*
		 * Valid to do another page fault here because this one came
Ingo Molnar's avatar
Ingo Molnar committed
777
		 * from user space:
778 779 780 781 782 783 784
		 */
		if (is_prefetch(regs, error_code, address))
			return;

		if (is_errata100(regs, address))
			return;

785 786 787 788 789 790
#ifdef CONFIG_X86_64
		/*
		 * Instruction fetch faults in the vsyscall page might need
		 * emulation.
		 */
		if (unlikely((error_code & PF_INSTR) &&
791
			     ((address & ~0xfff) == VSYSCALL_ADDR))) {
792 793 794 795
			if (emulate_vsyscall(regs, address))
				return;
		}
#endif
796 797 798
		/* Kernel addresses are always protection faults: */
		if (address >= TASK_SIZE)
			error_code |= PF_PROT;
799

800
		if (likely(show_unhandled_signals))
Ingo Molnar's avatar
Ingo Molnar committed
801 802 803
			show_signal_msg(regs, error_code, address, tsk);

		tsk->thread.cr2		= address;
804
		tsk->thread.error_code	= error_code;
805
		tsk->thread.trap_nr	= X86_TRAP_PF;
806

807
		force_sig_info_fault(SIGSEGV, si_code, address, tsk, vma, 0);
Ingo Molnar's avatar
Ingo Molnar committed
808

809 810 811 812 813 814
		return;
	}

	if (is_f00f_bug(regs, address))
		return;

815
	no_context(regs, error_code, address, SIGSEGV, si_code);
816 817
}

Ingo Molnar's avatar
Ingo Molnar committed
818 819
static noinline void
bad_area_nosemaphore(struct pt_regs *regs, unsigned long error_code,
820
		     unsigned long address, struct vm_area_struct *vma)
821
{
822
	__bad_area_nosemaphore(regs, error_code, address, vma, SEGV_MAPERR);
823 824
}

Ingo Molnar's avatar
Ingo Molnar committed
825 826
static void
__bad_area(struct pt_regs *regs, unsigned long error_code,
827
	   unsigned long address,  struct vm_area_struct *vma, int si_code)
828 829 830 831 832 833 834 835 836
{
	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);

837
	__bad_area_nosemaphore(regs, error_code, address, vma, si_code);
838 839
}

Ingo Molnar's avatar
Ingo Molnar committed
840 841
static noinline void
bad_area(struct pt_regs *regs, unsigned long error_code, unsigned long address)
842
{
843
	__bad_area(regs, error_code, address, NULL, SEGV_MAPERR);
844 845
}

Ingo Molnar's avatar
Ingo Molnar committed
846 847
static noinline void
bad_area_access_error(struct pt_regs *regs, unsigned long error_code,
848
		      unsigned long address, struct vm_area_struct *vma)
849
{
850
	__bad_area(regs, error_code, address, vma, SEGV_ACCERR);
851 852
}

Ingo Molnar's avatar
Ingo Molnar committed
853
static void
854
do_sigbus(struct pt_regs *regs, unsigned long error_code, unsigned long address,
855
	  struct vm_area_struct *vma, unsigned int fault)
856 857
{
	struct task_struct *tsk = current;
858
	int code = BUS_ADRERR;
859

Ingo Molnar's avatar
Ingo Molnar committed
860
	/* Kernel mode? Handle exceptions or die: */
861
	if (!(error_code & PF_USER)) {
862
		no_context(regs, error_code, address, SIGBUS, BUS_ADRERR);
863 864
		return;
	}
Ingo Molnar's avatar
Ingo Molnar committed
865

866
	/* User-space => ok to do another page fault: */
867 868
	if (is_prefetch(regs, error_code, address))
		return;
Ingo Molnar's avatar
Ingo Molnar committed
869 870 871

	tsk->thread.cr2		= address;
	tsk->thread.error_code	= error_code;
872
	tsk->thread.trap_nr	= X86_TRAP_PF;
Ingo Molnar's avatar
Ingo Molnar committed
873

874
#ifdef CONFIG_MEMORY_FAILURE
875
	if (fault & (VM_FAULT_HWPOISON|VM_FAULT_HWPOISON_LARGE)) {
876 877 878 879 880 881
		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
882
	force_sig_info_fault(SIGBUS, code, address, tsk, vma, fault);
883 884
}

885
static noinline void
Ingo Molnar's avatar
Ingo Molnar committed
886
mm_fault_error(struct pt_regs *regs, unsigned long error_code,
887 888
	       unsigned long address, struct vm_area_struct *vma,
	       unsigned int fault)
889
{
890 891 892
	if (fatal_signal_pending(current) && !(error_code & PF_USER)) {
		no_context(regs, error_code, address, 0, 0);
		return;
893 894
	}

Ingo Molnar's avatar
Ingo Molnar committed
895
	if (fault & VM_FAULT_OOM) {
896 897
		/* Kernel mode? Handle exceptions or die: */
		if (!(error_code & PF_USER)) {
898 899
			no_context(regs, error_code, address,
				   SIGSEGV, SEGV_MAPERR);
900
			return;
901 902
		}

903 904 905 906 907 908
		/*
		 * 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
909
	} else {
910 911
		if (fault & (VM_FAULT_SIGBUS|VM_FAULT_HWPOISON|
			     VM_FAULT_HWPOISON_LARGE))
912
			do_sigbus(regs, error_code, address, vma, fault);
913
		else if (fault & VM_FAULT_SIGSEGV)
914
			bad_area_nosemaphore(regs, error_code, address, vma);
Ingo Molnar's avatar
Ingo Molnar committed
915 916 917
		else
			BUG();
	}
918 919
}

920 921 922 923
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
924

925 926
	if ((error_code & PF_INSTR) && !pte_exec(*pte))
		return 0;
927 928 929 930 931 932
	/*
	 * 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;
933 934 935 936

	return 1;
}

937
/*
Ingo Molnar's avatar
Ingo Molnar committed
938 939 940 941 942 943 944 945
 * 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.
 *
946 947 948 949
 * 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.
 *
950 951
 * There are no security implications to leaving a stale TLB when
 * increasing the permissions on a page.
952 953 954 955 956
 *
 * 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).
957
 */
958
static noinline int
Ingo Molnar's avatar
Ingo Molnar committed
959
spurious_fault(unsigned long error_code, unsigned long address)
960 961 962 963 964
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;
965
	int ret;
966