fault.c 36 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
#include <asm/cpufeature.h>		/* boot_cpu_has, ...		*/
19 20
#include <asm/traps.h>			/* dotraplinkage, ...		*/
#include <asm/pgalloc.h>		/* pgd_*(), ...			*/
Vegard Nossum's avatar
Vegard Nossum committed
21
#include <asm/kmemcheck.h>		/* kmemcheck_*(), ...		*/
22 23
#include <asm/fixmap.h>			/* VSYSCALL_ADDR		*/
#include <asm/vsyscall.h>		/* emulate_vsyscall		*/
24
#include <asm/vm86.h>			/* struct vm86			*/
25
#include <asm/mmu_context.h>		/* vma_pkey()			*/
Linus Torvalds's avatar
Linus Torvalds committed
26

27 28 29
#define CREATE_TRACE_POINTS
#include <asm/trace/exceptions.h>

30
/*
Ingo Molnar's avatar
Ingo Molnar committed
31 32 33 34 35 36 37
 * 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
38
 *   bit 5 ==				1: protection keys block access
39
 */
Ingo Molnar's avatar
Ingo Molnar committed
40 41 42 43 44 45 46
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,
47
	PF_PK		=		1 << 5,
Ingo Molnar's avatar
Ingo Molnar committed
48
};
49

50
/*
51 52
 * Returns 0 if mmiotrace is disabled, or if the fault is not
 * handled by mmiotrace:
53
 */
54
static nokprobe_inline int
55
kmmio_fault(struct pt_regs *regs, unsigned long addr)
56
{
57 58 59 60
	if (unlikely(is_kmmio_active()))
		if (kmmio_handler(regs, addr) == 1)
			return -1;
	return 0;
61 62
}

63
static nokprobe_inline int kprobes_fault(struct pt_regs *regs)
64
{
65 66 67
	int ret = 0;

	/* kprobe_running() needs smp_processor_id() */
68
	if (kprobes_built_in() && !user_mode(regs)) {
69 70 71 72 73
		preempt_disable();
		if (kprobe_running() && kprobe_fault_handler(regs, 14))
			ret = 1;
		preempt_enable();
	}
74

75
	return ret;
76
}
77

78
/*
Ingo Molnar's avatar
Ingo Molnar committed
79 80 81 82 83 84
 * Prefetch quirks:
 *
 * 32-bit mode:
 *
 *   Sometimes AMD Athlon/Opteron CPUs report invalid exceptions on prefetch.
 *   Check that here and ignore it.
85
 *
Ingo Molnar's avatar
Ingo Molnar committed
86
 * 64-bit mode:
87
 *
Ingo Molnar's avatar
Ingo Molnar committed
88 89 90 91
 *   Sometimes the CPU reports invalid exceptions on prefetch.
 *   Check that here and ignore it.
 *
 * Opcode checker based on code by Richard Brunner.
92
 */
93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118
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.
		 */
119
		return (!user_mode(regs) || user_64bit_mode(regs));
120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139
#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
140 141
static int
is_prefetch(struct pt_regs *regs, unsigned long error_code, unsigned long addr)
142
{
Ingo Molnar's avatar
Ingo Molnar committed
143
	unsigned char *max_instr;
144
	unsigned char *instr;
145
	int prefetch = 0;
Linus Torvalds's avatar
Linus Torvalds committed
146

Ingo Molnar's avatar
Ingo Molnar committed
147 148 149 150
	/*
	 * If it was a exec (instruction fetch) fault on NX page, then
	 * do not ignore the fault:
	 */
151
	if (error_code & PF_INSTR)
Linus Torvalds's avatar
Linus Torvalds committed
152
		return 0;
153

154
	instr = (void *)convert_ip_to_linear(current, regs);
155
	max_instr = instr + 15;
Linus Torvalds's avatar
Linus Torvalds committed
156

157
	if (user_mode(regs) && instr >= (unsigned char *)TASK_SIZE_MAX)
Linus Torvalds's avatar
Linus Torvalds committed
158 159
		return 0;

160
	while (instr < max_instr) {
Ingo Molnar's avatar
Ingo Molnar committed
161
		unsigned char opcode;
Linus Torvalds's avatar
Linus Torvalds committed
162

163
		if (probe_kernel_address(instr, opcode))
164
			break;
Linus Torvalds's avatar
Linus Torvalds committed
165 166 167

		instr++;

168
		if (!check_prefetch_opcode(regs, instr, opcode, &prefetch))
Linus Torvalds's avatar
Linus Torvalds committed
169 170 171 172 173
			break;
	}
	return prefetch;
}

174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223
/*
 * A protection key fault means that the PKRU value did not allow
 * access to some PTE.  Userspace can figure out what PKRU was
 * from the XSAVE state, and this function fills out a field in
 * siginfo so userspace can discover which protection key was set
 * on the PTE.
 *
 * If we get here, we know that the hardware signaled a PF_PK
 * fault and that there was a VMA once we got in the fault
 * handler.  It does *not* guarantee that the VMA we find here
 * was the one that we faulted on.
 *
 * 1. T1   : mprotect_key(foo, PAGE_SIZE, pkey=4);
 * 2. T1   : set PKRU to deny access to pkey=4, touches page
 * 3. T1   : faults...
 * 4.    T2: mprotect_key(foo, PAGE_SIZE, pkey=5);
 * 5. T1   : enters fault handler, takes mmap_sem, etc...
 * 6. T1   : reaches here, sees vma_pkey(vma)=5, when we really
 *	     faulted on a pte with its pkey=4.
 */
static void fill_sig_info_pkey(int si_code, siginfo_t *info,
		struct vm_area_struct *vma)
{
	/* This is effectively an #ifdef */
	if (!boot_cpu_has(X86_FEATURE_OSPKE))
		return;

	/* Fault not from Protection Keys: nothing to do */
	if (si_code != SEGV_PKUERR)
		return;
	/*
	 * force_sig_info_fault() is called from a number of
	 * contexts, some of which have a VMA and some of which
	 * do not.  The PF_PK handing happens after we have a
	 * valid VMA, so we should never reach this without a
	 * valid VMA.
	 */
	if (!vma) {
		WARN_ONCE(1, "PKU fault with no VMA passed in");
		info->si_pkey = 0;
		return;
	}
	/*
	 * si_pkey should be thought of as a strong hint, but not
	 * absolutely guranteed to be 100% accurate because of
	 * the race explained above.
	 */
	info->si_pkey = vma_pkey(vma);
}

Ingo Molnar's avatar
Ingo Molnar committed
224 225
static void
force_sig_info_fault(int si_signo, int si_code, unsigned long address,
226 227
		     struct task_struct *tsk, struct vm_area_struct *vma,
		     int fault)
228
{
229
	unsigned lsb = 0;
230 231
	siginfo_t info;

Ingo Molnar's avatar
Ingo Molnar committed
232 233 234 235
	info.si_signo	= si_signo;
	info.si_errno	= 0;
	info.si_code	= si_code;
	info.si_addr	= (void __user *)address;
236 237 238 239 240
	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
241

242 243
	fill_sig_info_pkey(si_code, &info, vma);

244 245 246
	force_sig_info(si_signo, &info, tsk);
}

247 248 249 250 251
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)
252
{
253 254 255 256
	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
257

258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278
	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;

279
	if (!pmd_present(*pmd))
280
		set_pmd(pmd, *pmd_k);
281
	else
282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298
		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;

299
		spin_lock(&pgd_lock);
300
		list_for_each_entry(page, &pgd_list, lru) {
301
			spinlock_t *pgt_lock;
302
			pmd_t *ret;
303

304
			/* the pgt_lock only for Xen */
305 306 307 308 309 310 311
			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)
312 313
				break;
		}
314
		spin_unlock(&pgd_lock);
315 316 317 318 319 320 321 322
	}
}

/*
 * 32-bit:
 *
 *   Handle a fault on the vmalloc or module mapping area
 */
323
static noinline int vmalloc_fault(unsigned long address)
324 325 326 327 328 329 330 331 332
{
	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;

333 334
	WARN_ON_ONCE(in_nmi());

335 336 337 338 339 340 341 342 343 344 345 346
	/*
	 * 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;

347 348 349
	if (pmd_huge(*pmd_k))
		return 0;

350 351 352 353 354 355
	pte_k = pte_offset_kernel(pmd_k, address);
	if (!pte_present(*pte_k))
		return -1;

	return 0;
}
356
NOKPROBE_SYMBOL(vmalloc_fault);
357 358 359 360 361 362 363 364

/*
 * 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)
{
365
#ifdef CONFIG_VM86
366 367
	unsigned long bit;

368
	if (!v8086_mode(regs) || !tsk->thread.vm86)
369 370 371 372
		return;

	bit = (address - 0xA0000) >> PAGE_SHIFT;
	if (bit < 32)
373 374
		tsk->thread.vm86->screen_bitmap |= 1 << bit;
#endif
375
}
Linus Torvalds's avatar
Linus Torvalds committed
376

Akinobu Mita's avatar
Akinobu Mita committed
377
static bool low_pfn(unsigned long pfn)
Linus Torvalds's avatar
Linus Torvalds committed
378
{
Akinobu Mita's avatar
Akinobu Mita committed
379 380
	return pfn < max_low_pfn;
}
381

Akinobu Mita's avatar
Akinobu Mita committed
382 383 384 385 386 387
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
388

389
#ifdef CONFIG_X86_PAE
Akinobu Mita's avatar
Akinobu Mita committed
390 391 392
	printk("*pdpt = %016Lx ", pgd_val(*pgd));
	if (!low_pfn(pgd_val(*pgd) >> PAGE_SHIFT) || !pgd_present(*pgd))
		goto out;
393
#endif
Akinobu Mita's avatar
Akinobu Mita committed
394 395
	pmd = pmd_offset(pud_offset(pgd, address), address);
	printk(KERN_CONT "*pde = %0*Lx ", sizeof(*pmd) * 2, (u64)pmd_val(*pmd));
396 397 398 399 400

	/*
	 * 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
401
	 * it's allocated already:
402
	 */
Akinobu Mita's avatar
Akinobu Mita committed
403 404
	if (!low_pfn(pmd_pfn(*pmd)) || !pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
405

Akinobu Mita's avatar
Akinobu Mita committed
406 407 408
	pte = pte_offset_kernel(pmd, address);
	printk("*pte = %0*Lx ", sizeof(*pte) * 2, (u64)pte_val(*pte));
out:
409
	printk("\n");
410 411 412 413 414 415
}

#else /* CONFIG_X86_64: */

void vmalloc_sync_all(void)
{
416
	sync_global_pgds(VMALLOC_START & PGDIR_MASK, VMALLOC_END, 0);
417 418 419 420 421 422 423
}

/*
 * 64-bit:
 *
 *   Handle a fault on the vmalloc area
 */
424
static noinline int vmalloc_fault(unsigned long address)
425 426 427 428 429 430 431 432 433 434
{
	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;

435 436
	WARN_ON_ONCE(in_nmi());

437 438 439 440 441 442 443 444 445 446
	/*
	 * 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;

447
	if (pgd_none(*pgd)) {
448
		set_pgd(pgd, *pgd_ref);
449 450
		arch_flush_lazy_mmu_mode();
	} else {
451
		BUG_ON(pgd_page_vaddr(*pgd) != pgd_page_vaddr(*pgd_ref));
452
	}
453 454 455 456 457 458 459 460 461 462 463

	/*
	 * 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;

464
	if (pud_none(*pud) || pud_pfn(*pud) != pud_pfn(*pud_ref))
465 466
		BUG();

467 468 469
	if (pud_huge(*pud))
		return 0;

470 471 472 473 474
	pmd = pmd_offset(pud, address);
	pmd_ref = pmd_offset(pud_ref, address);
	if (pmd_none(*pmd_ref))
		return -1;

475
	if (pmd_none(*pmd) || pmd_pfn(*pmd) != pmd_pfn(*pmd_ref))
476 477
		BUG();

478 479 480
	if (pmd_huge(*pmd))
		return 0;

481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496
	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;
}
497
NOKPROBE_SYMBOL(vmalloc_fault);
498

499
#ifdef CONFIG_CPU_SUP_AMD
500
static const char errata93_warning[] =
501 502 503 504 505
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";
506
#endif
507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525

/*
 * 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
526 527
	pgd_t *base = __va(read_cr3() & PHYSICAL_PAGE_MASK);
	pgd_t *pgd = base + pgd_index(address);
Linus Torvalds's avatar
Linus Torvalds committed
528 529 530 531
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;

Ingo Molnar's avatar
Ingo Molnar committed
532 533 534
	if (bad_address(pgd))
		goto bad;

535
	printk("PGD %lx ", pgd_val(*pgd));
Ingo Molnar's avatar
Ingo Molnar committed
536 537 538

	if (!pgd_present(*pgd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
539

540
	pud = pud_offset(pgd, address);
Ingo Molnar's avatar
Ingo Molnar committed
541 542 543
	if (bad_address(pud))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
544
	printk("PUD %lx ", pud_val(*pud));
545
	if (!pud_present(*pud) || pud_large(*pud))
Ingo Molnar's avatar
Ingo Molnar committed
546
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
547 548

	pmd = pmd_offset(pud, address);
Ingo Molnar's avatar
Ingo Molnar committed
549 550 551
	if (bad_address(pmd))
		goto bad;

Linus Torvalds's avatar
Linus Torvalds committed
552
	printk("PMD %lx ", pmd_val(*pmd));
Ingo Molnar's avatar
Ingo Molnar committed
553 554
	if (!pmd_present(*pmd) || pmd_large(*pmd))
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
555 556

	pte = pte_offset_kernel(pmd, address);
Ingo Molnar's avatar
Ingo Molnar committed
557 558 559
	if (bad_address(pte))
		goto bad;

560
	printk("PTE %lx", pte_val(*pte));
Ingo Molnar's avatar
Ingo Molnar committed
561
out:
Linus Torvalds's avatar
Linus Torvalds committed
562 563 564 565
	printk("\n");
	return;
bad:
	printk("BAD\n");
566 567
}

568
#endif /* CONFIG_X86_64 */
Linus Torvalds's avatar
Linus Torvalds committed
569

Ingo Molnar's avatar
Ingo Molnar committed
570 571 572 573 574 575 576 577 578 579 580 581 582
/*
 * 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.
583
 */
584
static int is_errata93(struct pt_regs *regs, unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
585
{
586 587 588 589 590
#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;

591
	if (address != regs->ip)
Linus Torvalds's avatar
Linus Torvalds committed
592
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
593

594
	if ((address >> 32) != 0)
Linus Torvalds's avatar
Linus Torvalds committed
595
		return 0;
Ingo Molnar's avatar
Ingo Molnar committed
596

Linus Torvalds's avatar
Linus Torvalds committed
597
	address |= 0xffffffffUL << 32;
598 599
	if ((address >= (u64)_stext && address <= (u64)_etext) ||
	    (address >= MODULES_VADDR && address <= MODULES_END)) {
600
		printk_once(errata93_warning);
601
		regs->ip = address;
Linus Torvalds's avatar
Linus Torvalds committed
602 603
		return 1;
	}
604
#endif
Linus Torvalds's avatar
Linus Torvalds committed
605
	return 0;
606
}
Linus Torvalds's avatar
Linus Torvalds committed
607

608
/*
Ingo Molnar's avatar
Ingo Molnar committed
609 610 611 612 613
 * 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
614 615 616 617 618
 * 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
619
	if ((regs->cs == __USER32_CS || (regs->cs & (1<<2))) && (address >> 32))
620 621 622 623 624
		return 1;
#endif
	return 0;
}

625 626 627 628
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
629

630
	/*
Ingo Molnar's avatar
Ingo Molnar committed
631
	 * Pentium F0 0F C7 C8 bug workaround:
632
	 */
633
	if (boot_cpu_has_bug(X86_BUG_F00F)) {
634 635 636 637 638 639 640 641 642 643 644
		nr = (address - idt_descr.address) >> 3;

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

645 646
static const char nx_warning[] = KERN_CRIT
"kernel tried to execute NX-protected page - exploit attempt? (uid: %d)\n";
647 648
static const char smep_warning[] = KERN_CRIT
"unable to execute userspace code (SMEP?) (uid: %d)\n";
649

Ingo Molnar's avatar
Ingo Molnar committed
650 651 652
static void
show_fault_oops(struct pt_regs *regs, unsigned long error_code,
		unsigned long address)
653
{
654 655 656 657
	if (!oops_may_print())
		return;

	if (error_code & PF_INSTR) {
658
		unsigned int level;
659 660
		pgd_t *pgd;
		pte_t *pte;
Ingo Molnar's avatar
Ingo Molnar committed
661

662 663 664 665
		pgd = __va(read_cr3() & PHYSICAL_PAGE_MASK);
		pgd += pgd_index(address);

		pte = lookup_address_in_pgd(pgd, address, &level);
666

667
		if (pte && pte_present(*pte) && !pte_exec(*pte))
668
			printk(nx_warning, from_kuid(&init_user_ns, current_uid()));
669 670
		if (pte && pte_present(*pte) && pte_exec(*pte) &&
				(pgd_flags(*pgd) & _PAGE_USER) &&
671
				(__read_cr4() & X86_CR4_SMEP))
672
			printk(smep_warning, from_kuid(&init_user_ns, current_uid()));
673 674
	}

675
	printk(KERN_ALERT "BUG: unable to handle kernel ");
676
	if (address < PAGE_SIZE)
677
		printk(KERN_CONT "NULL pointer dereference");
678
	else
679
		printk(KERN_CONT "paging request");
Ingo Molnar's avatar
Ingo Molnar committed
680

681
	printk(KERN_CONT " at %p\n", (void *) address);
682
	printk(KERN_ALERT "IP:");
683
	printk_address(regs->ip);
Ingo Molnar's avatar
Ingo Molnar committed
684

685 686 687
	dump_pagetable(address);
}

Ingo Molnar's avatar
Ingo Molnar committed
688 689 690
static noinline void
pgtable_bad(struct pt_regs *regs, unsigned long error_code,
	    unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
691
{
Ingo Molnar's avatar
Ingo Molnar committed
692 693 694 695 696 697 698
	struct task_struct *tsk;
	unsigned long flags;
	int sig;

	flags = oops_begin();
	tsk = current;
	sig = SIGKILL;
699

Linus Torvalds's avatar
Linus Torvalds committed
700
	printk(KERN_ALERT "%s: Corrupted page table at address %lx\n",
701
	       tsk->comm, address);
Linus Torvalds's avatar
Linus Torvalds committed
702
	dump_pagetable(address);
Ingo Molnar's avatar
Ingo Molnar committed
703 704

	tsk->thread.cr2		= address;
705
	tsk->thread.trap_nr	= X86_TRAP_PF;
Ingo Molnar's avatar
Ingo Molnar committed
706 707
	tsk->thread.error_code	= error_code;

708
	if (__die("Bad pagetable", regs, error_code))
709
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
710

711
	oops_end(flags, regs, sig);
Linus Torvalds's avatar
Linus Torvalds committed
712 713
}

Ingo Molnar's avatar
Ingo Molnar committed
714 715
static noinline void
no_context(struct pt_regs *regs, unsigned long error_code,
716
	   unsigned long address, int signal, int si_code)
717 718 719 720
{
	struct task_struct *tsk = current;
	unsigned long flags;
	int sig;
721 722
	/* No context means no VMA to pass down */
	struct vm_area_struct *vma = NULL;
723

Ingo Molnar's avatar
Ingo Molnar committed
724
	/* Are we prepared to handle this kernel fault? */
725
	if (fixup_exception(regs, X86_TRAP_PF)) {
726 727 728 729 730 731 732 733 734 735 736 737 738 739
		/*
		 * 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.
		 */
740
		if (current_thread_info()->sig_on_uaccess_error && signal) {
741
			tsk->thread.trap_nr = X86_TRAP_PF;
742 743 744 745
			tsk->thread.error_code = error_code | PF_USER;
			tsk->thread.cr2 = address;

			/* XXX: hwpoison faults will set the wrong code. */
746 747
			force_sig_info_fault(signal, si_code, address,
					     tsk, vma, 0);
748
		}
749 750 751 752

		/*
		 * Barring that, we can do the fixup and be happy.
		 */
753
		return;
754
	}
755 756

	/*
Ingo Molnar's avatar
Ingo Molnar committed
757 758 759 760 761 762 763
	 * 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:
764
	 *
Ingo Molnar's avatar
Ingo Molnar committed
765
	 *   Hall of shame of CPU/BIOS bugs.
766 767 768 769 770 771 772 773 774
	 */
	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
775
	 * terminate things with extreme prejudice:
776 777 778 779 780
	 */
	flags = oops_begin();

	show_fault_oops(regs, error_code, address);

781
	if (task_stack_end_corrupted(tsk))
782
		printk(KERN_EMERG "Thread overran stack, or stack corrupted\n");
783

784
	tsk->thread.cr2		= address;
785
	tsk->thread.trap_nr	= X86_TRAP_PF;
786
	tsk->thread.error_code	= error_code;
787 788 789 790

	sig = SIGKILL;
	if (__die("Oops", regs, error_code))
		sig = 0;
Ingo Molnar's avatar
Ingo Molnar committed
791

792
	/* Executive summary in case the body of the oops scrolled away */
793
	printk(KERN_DEFAULT "CR2: %016lx\n", address);
Ingo Molnar's avatar
Ingo Molnar committed
794

795 796 797
	oops_end(flags, regs, sig);
}

Ingo Molnar's avatar
Ingo Molnar committed
798 799 800 801 802 803 804 805 806 807 808 809 810 811
/*
 * 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;

812
	printk("%s%s[%d]: segfault at %lx ip %p sp %p error %lx",
Ingo Molnar's avatar
Ingo Molnar committed
813 814 815 816 817 818 819 820 821 822 823
		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,
824 825
		       unsigned long address, struct vm_area_struct *vma,
		       int si_code)
826 827 828 829 830 831
{
	struct task_struct *tsk = current;

	/* User mode accesses just cause a SIGSEGV */
	if (error_code & PF_USER) {
		/*
Ingo Molnar's avatar
Ingo Molnar committed
832
		 * It's possible to have interrupts off here:
833 834 835 836 837
		 */
		local_irq_enable();

		/*
		 * Valid to do another page fault here because this one came
Ingo Molnar's avatar
Ingo Molnar committed
838
		 * from user space:
839 840 841 842 843 844 845
		 */
		if (is_prefetch(regs, error_code, address))
			return;

		if (is_errata100(regs, address))
			return;

846 847 848 849 850 851
#ifdef CONFIG_X86_64
		/*
		 * Instruction fetch faults in the vsyscall page might need
		 * emulation.
		 */
		if (unlikely((error_code & PF_INSTR) &&
852
			     ((address & ~0xfff) == VSYSCALL_ADDR))) {
853 854 855 856
			if (emulate_vsyscall(regs, address))
				return;
		}
#endif
857 858 859
		/* Kernel addresses are always protection faults: */
		if (address >= TASK_SIZE)
			error_code |= PF_PROT;
860

861
		if (likely(show_unhandled_signals))
Ingo Molnar's avatar
Ingo Molnar committed
862 863 864
			show_signal_msg(regs, error_code, address, tsk);

		tsk->thread.cr2		= address;
865
		tsk->thread.error_code	= error_code;
866
		tsk->thread.trap_nr	= X86_TRAP_PF;
867

868
		force_sig_info_fault(SIGSEGV, si_code, address, tsk, vma, 0);
Ingo Molnar's avatar
Ingo Molnar committed
869

870 871 872 873 874 875
		return;
	}

	if (is_f00f_bug(regs, address))
		return;

876
	no_context(regs, error_code, address, SIGSEGV, si_code);
877 878
}

Ingo Molnar's avatar
Ingo Molnar committed
879 880
static noinline void
bad_area_nosemaphore(struct pt_regs *regs, unsigned long error_code,
881
		     unsigned long address, struct vm_area_struct *vma)
882
{
883
	__bad_area_nosemaphore(regs, error_code, address, vma, SEGV_MAPERR);
884 885
}

Ingo Molnar's avatar
Ingo Molnar committed
886 887
static void
__bad_area(struct pt_regs *regs, unsigned long error_code,
888
	   unsigned long address,  struct vm_area_struct *vma, int si_code)
889 890 891 892 893 894 895 896 897
{
	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);

898
	__bad_area_nosemaphore(regs, error_code, address, vma, si_code);
899 900
}

Ingo Molnar's avatar
Ingo Molnar committed
901 902
static noinline void
bad_area(struct pt_regs *regs, unsigned long error_code, unsigned long address)
903
{
904
	__bad_area(regs, error_code, address, NULL, SEGV_MAPERR);
905 906
}

907 908 909
static inline bool bad_area_access_from_pkeys(unsigned long error_code,
		struct vm_area_struct *vma)
{
910 911 912
	/* This code is always called on the current mm */
	bool foreign = false;

913 914 915 916
	if (!boot_cpu_has(X86_FEATURE_OSPKE))
		return false;
	if (error_code & PF_PK)
		return true;
917
	/* this checks permission keys on the VMA: */
918 919
	if (!arch_vma_access_permitted(vma, (error_code & PF_WRITE),
				(error_code & PF_INSTR), foreign))
920
		return true;
921
	return false;
922 923
}

Ingo Molnar's avatar
Ingo Molnar committed
924 925
static noinline void
bad_area_access_error(struct pt_regs *regs, unsigned long error_code,
926
		      unsigned long address, struct vm_area_struct *vma)
927
{
928 929 930 931 932
	/*
	 * This OSPKE check is not strictly necessary at runtime.
	 * But, doing it this way allows compiler optimizations
	 * if pkeys are compiled out.
	 */
933
	if (bad_area_access_from_pkeys(error_code, vma))
934 935 936
		__bad_area(regs, error_code, address, vma, SEGV_PKUERR);
	else
		__bad_area(regs, error_code, address, vma, SEGV_ACCERR);
937 938
}

Ingo Molnar's avatar
Ingo Molnar committed
939
static void
940
do_sigbus(struct pt_regs *regs, unsigned long error_code, unsigned long address,
941
	  struct vm_area_struct *vma, unsigned int fault)
942 943
{
	struct task_struct *tsk = current;
944
	int code = BUS_ADRERR;
945

Ingo Molnar's avatar
Ingo Molnar committed
946
	/* Kernel mode? Handle exceptions or die: */
947
	if (!(error_code & PF_USER)) {
948
		no_context(regs, error_code, address, SIGBUS, BUS_ADRERR);
949 950
		return;
	}
Ingo Molnar's avatar
Ingo Molnar committed
951

952
	/* User-space => ok to do another page fault: */
953 954
	if (is_prefetch(regs, error_code, address))
		return;
Ingo Molnar's avatar
Ingo Molnar committed
955 956 957

	tsk->thread.cr2		= address;
	tsk->thread.error_code	= error_code;
958
	tsk->thread.trap_nr	= X86_TRAP_PF;
Ingo Molnar's avatar
Ingo Molnar committed
959

960
#ifdef CONFIG_MEMORY_FAILURE
961
	if (fault & (VM_FAULT_HWPOISON|VM_FAULT_HWPOISON_LARGE)) {
962 963 964 965 966 967
		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
968
	force_sig_info_fault(SIGBUS, code, address, tsk, vma, fault);
969 970
}

971
static noinline void
Ingo Molnar's avatar
Ingo Molnar committed
972
mm_fault_error(struct pt_regs *regs, unsigned long error_code,
973 974
	       unsigned long address, struct vm_area_struct *vma,
	       unsigned int fault)
975
{
976 977 978
	if (fatal_signal_pending(current) && !(error_code & PF_USER)) {
		no_context(regs, error_code, address, 0, 0);
		return;
979 980
	}

Ingo Molnar's avatar
Ingo Molnar committed