fault.c 13.9 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0-only
L
Linus Torvalds 已提交
2 3 4 5 6 7
/*
 *  linux/arch/arm/mm/fault.c
 *
 *  Copyright (C) 1995  Linus Torvalds
 *  Modifications for ARM processor (c) 1995-2004 Russell King
 */
8
#include <linux/extable.h>
L
Linus Torvalds 已提交
9 10
#include <linux/signal.h>
#include <linux/mm.h>
11
#include <linux/hardirq.h>
L
Linus Torvalds 已提交
12
#include <linux/init.h>
13
#include <linux/kprobes.h>
14
#include <linux/uaccess.h>
15
#include <linux/page-flags.h>
16
#include <linux/sched/signal.h>
17
#include <linux/sched/debug.h>
R
Russell King 已提交
18
#include <linux/highmem.h>
19
#include <linux/perf_event.h>
L
Linus Torvalds 已提交
20

21 22
#include <asm/system_misc.h>
#include <asm/system_info.h>
L
Linus Torvalds 已提交
23 24 25 26
#include <asm/tlbflush.h>

#include "fault.h"

27
#ifdef CONFIG_MMU
28

29 30 31 32
#if defined(CONFIG_SYS_SUPPORTS_HUGETLBFS) && !defined(CONFIG_ARM_LPAE)
#include <linux/hugetlb.h>
#endif

L
Linus Torvalds 已提交
33 34 35 36
/*
 * This is useful to dump out the page tables associated with
 * 'addr' in mm 'mm'.
 */
37
void show_pte(const char *lvl, struct mm_struct *mm, unsigned long addr)
L
Linus Torvalds 已提交
38 39 40 41 42 43 44
{
	pgd_t *pgd;

	if (!mm)
		mm = &init_mm;

	pgd = pgd_offset(mm, addr);
45
	printk("%s[%08lx] *pgd=%08llx", lvl, addr, (long long)pgd_val(*pgd));
L
Linus Torvalds 已提交
46 47

	do {
48
		p4d_t *p4d;
R
Russell King 已提交
49
		pud_t *pud;
L
Linus Torvalds 已提交
50 51 52
		pmd_t *pmd;
		pte_t *pte;

53 54
		p4d = p4d_offset(pgd, addr);
		if (p4d_none(*p4d))
L
Linus Torvalds 已提交
55 56
			break;

57
		if (p4d_bad(*p4d)) {
R
Russell King 已提交
58
			pr_cont("(bad)");
L
Linus Torvalds 已提交
59 60 61
			break;
		}

62
		pud = pud_offset(p4d, addr);
R
Russell King 已提交
63
		if (PTRS_PER_PUD != 1)
R
Russell King 已提交
64
			pr_cont(", *pud=%08llx", (long long)pud_val(*pud));
R
Russell King 已提交
65 66 67 68 69

		if (pud_none(*pud))
			break;

		if (pud_bad(*pud)) {
R
Russell King 已提交
70
			pr_cont("(bad)");
R
Russell King 已提交
71 72 73 74
			break;
		}

		pmd = pmd_offset(pud, addr);
75
		if (PTRS_PER_PMD != 1)
R
Russell King 已提交
76
			pr_cont(", *pmd=%08llx", (long long)pmd_val(*pmd));
L
Linus Torvalds 已提交
77 78 79 80 81

		if (pmd_none(*pmd))
			break;

		if (pmd_bad(*pmd)) {
82 83 84 85 86 87 88 89 90
#if !defined(CONFIG_ARM_LPAE) && defined(CONFIG_HUGETLBFS)
			if (pte_huge((pte_t)*pgd)) {
				pte_t huge_pte = huge_ptep_get((pte_t *)pgd);

				pr_alert("[%08lx] *ppgd=%08llx", addr,
					(long long)pmd_val(huge_pte));
				break;
			}
#endif
R
Russell King 已提交
91
			pr_cont("(bad)");
L
Linus Torvalds 已提交
92 93 94 95
			break;
		}

		/* We must not map this if we have highmem enabled */
96 97 98
		if (PageHighMem(pfn_to_page(pmd_val(*pmd) >> PAGE_SHIFT)))
			break;

L
Linus Torvalds 已提交
99
		pte = pte_offset_map(pmd, addr);
R
Russell King 已提交
100
		pr_cont(", *pte=%08llx", (long long)pte_val(*pte));
101
#ifndef CONFIG_ARM_LPAE
R
Russell King 已提交
102
		pr_cont(", *ppte=%08llx",
103
		       (long long)pte_val(pte[PTE_HWTABLE_PTRS]));
104
#endif
L
Linus Torvalds 已提交
105 106 107
		pte_unmap(pte);
	} while(0);

R
Russell King 已提交
108
	pr_cont("\n");
L
Linus Torvalds 已提交
109
}
110
#else					/* CONFIG_MMU */
111
void show_pte(const char *lvl, struct mm_struct *mm, unsigned long addr)
112 113
{ }
#endif					/* CONFIG_MMU */
L
Linus Torvalds 已提交
114

115 116 117 118 119 120 121 122 123 124 125 126 127 128 129
static void die_kernel_fault(const char *msg, struct mm_struct *mm,
			     unsigned long addr, unsigned int fsr,
			     struct pt_regs *regs)
{
	bust_spinlocks(1);
	pr_alert("8<--- cut here ---\n");
	pr_alert("Unable to handle kernel %s at virtual address %08lx\n",
		 msg, addr);

	show_pte(KERN_ALERT, mm, addr);
	die("Oops", regs, fsr);
	bust_spinlocks(0);
	do_exit(SIGKILL);
}

L
Linus Torvalds 已提交
130 131 132 133 134 135 136
/*
 * Oops.  The kernel tried to access some page that wasn't present.
 */
static void
__do_kernel_fault(struct mm_struct *mm, unsigned long addr, unsigned int fsr,
		  struct pt_regs *regs)
{
137
	const char *msg;
L
Linus Torvalds 已提交
138 139 140 141 142 143 144 145 146
	/*
	 * Are we prepared to handle this kernel fault?
	 */
	if (fixup_exception(regs))
		return;

	/*
	 * No handler, we'll have to terminate things with extreme prejudice.
	 */
147 148 149 150
	if (addr < PAGE_SIZE)
		msg = "NULL pointer dereference";
	else
		msg = "paging request";
L
Linus Torvalds 已提交
151

152
	die_kernel_fault(msg, mm, addr, fsr, regs);
L
Linus Torvalds 已提交
153 154 155 156 157 158 159
}

/*
 * Something tried to access memory that isn't in our memory map..
 * User mode accesses just cause a SIGSEGV
 */
static void
160 161
__do_user_fault(unsigned long addr, unsigned int fsr, unsigned int sig,
		int code, struct pt_regs *regs)
L
Linus Torvalds 已提交
162
{
163 164
	struct task_struct *tsk = current;

165 166 167
	if (addr > TASK_SIZE)
		harden_branch_predictor();

L
Linus Torvalds 已提交
168
#ifdef CONFIG_DEBUG_USER
169 170
	if (((user_debug & UDBG_SEGV) && (sig == SIGSEGV)) ||
	    ((user_debug & UDBG_BUS)  && (sig == SIGBUS))) {
171
		pr_err("8<--- cut here ---\n");
172
		pr_err("%s: unhandled page fault (%d) at 0x%08lx, code 0x%03x\n",
A
akpm@osdl.org 已提交
173
		       tsk->comm, sig, addr, fsr);
174
		show_pte(KERN_ERR, tsk->mm, addr);
L
Linus Torvalds 已提交
175 176 177
		show_regs(regs);
	}
#endif
178 179 180 181 182 183
#ifndef CONFIG_KUSER_HELPERS
	if ((sig == SIGSEGV) && ((addr & PAGE_MASK) == 0xffff0000))
		printk_ratelimited(KERN_DEBUG
				   "%s: CONFIG_KUSER_HELPERS disabled at 0x%08lx\n",
				   tsk->comm, addr);
#endif
L
Linus Torvalds 已提交
184 185 186 187

	tsk->thread.address = addr;
	tsk->thread.error_code = fsr;
	tsk->thread.trap_no = 14;
188
	force_sig_fault(sig, code, (void __user *)addr);
L
Linus Torvalds 已提交
189 190
}

191
void do_bad_area(unsigned long addr, unsigned int fsr, struct pt_regs *regs)
L
Linus Torvalds 已提交
192
{
193 194 195
	struct task_struct *tsk = current;
	struct mm_struct *mm = tsk->active_mm;

L
Linus Torvalds 已提交
196 197 198 199 200
	/*
	 * If we are in kernel mode at this point, we
	 * have no context to handle this fault with.
	 */
	if (user_mode(regs))
201
		__do_user_fault(addr, fsr, SIGSEGV, SEGV_MAPERR, regs);
L
Linus Torvalds 已提交
202 203 204 205
	else
		__do_kernel_fault(mm, addr, fsr, regs);
}

206
#ifdef CONFIG_MMU
207 208
#define VM_FAULT_BADMAP		0x010000
#define VM_FAULT_BADACCESS	0x020000
L
Linus Torvalds 已提交
209

210
static vm_fault_t __kprobes
K
Kefeng Wang 已提交
211 212
__do_page_fault(struct mm_struct *mm, unsigned long addr, unsigned int flags,
		unsigned long vma_flags, struct pt_regs *regs)
L
Linus Torvalds 已提交
213
{
214
	struct vm_area_struct *vma = find_vma(mm, addr);
215
	if (unlikely(!vma))
216 217 218 219 220 221 222 223 224 225
		return VM_FAULT_BADMAP;

	if (unlikely(vma->vm_start > addr)) {
		if (!(vma->vm_flags & VM_GROWSDOWN))
			return VM_FAULT_BADMAP;
		if (addr < FIRST_USER_ADDRESS)
			return VM_FAULT_BADMAP;
		if (expand_stack(vma, addr))
			return VM_FAULT_BADMAP;
	}
L
Linus Torvalds 已提交
226 227

	/*
K
Kefeng Wang 已提交
228 229
	 * ok, we have a good vm_area for this memory access, check the
	 * permissions on the VMA allow for the fault which occurred.
L
Linus Torvalds 已提交
230
	 */
K
Kefeng Wang 已提交
231
	if (!(vma->vm_flags & vma_flags))
232
		return VM_FAULT_BADACCESS;
L
Linus Torvalds 已提交
233

234
	return handle_mm_fault(vma, addr & PAGE_MASK, flags, regs);
L
Linus Torvalds 已提交
235 236
}

237
static int __kprobes
L
Linus Torvalds 已提交
238 239
do_page_fault(unsigned long addr, unsigned int fsr, struct pt_regs *regs)
{
240
	struct mm_struct *mm = current->mm;
241 242
	int sig, code;
	vm_fault_t fault;
P
Peter Xu 已提交
243
	unsigned int flags = FAULT_FLAG_DEFAULT;
K
Kefeng Wang 已提交
244
	unsigned long vm_flags = VM_ACCESS_FLAGS;
L
Linus Torvalds 已提交
245

246
	if (kprobe_page_fault(regs, fsr))
247 248
		return 0;

L
Linus Torvalds 已提交
249

250 251 252 253
	/* Enable interrupts if they were enabled in the parent context. */
	if (interrupts_enabled(regs))
		local_irq_enable();

L
Linus Torvalds 已提交
254 255 256 257
	/*
	 * If we're in an interrupt or have no user
	 * context, we must not take the fault..
	 */
258
	if (faulthandler_disabled() || !mm)
L
Linus Torvalds 已提交
259 260
		goto no_context;

261 262
	if (user_mode(regs))
		flags |= FAULT_FLAG_USER;
K
Kefeng Wang 已提交
263 264

	if ((fsr & FSR_WRITE) && !(fsr & FSR_CM)) {
265
		flags |= FAULT_FLAG_WRITE;
K
Kefeng Wang 已提交
266 267 268 269 270
		vm_flags = VM_WRITE;
	}

	if (fsr & FSR_LNX_PF)
		vm_flags = VM_EXEC;
271

272 273
	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr);

274 275 276 277 278
	/*
	 * As per x86, we may deadlock here.  However, since the kernel only
	 * validly references user space from well defined areas of the code,
	 * we can bug out early if this is from code which shouldn't.
	 */
279
	if (!mmap_read_trylock(mm)) {
280 281
		if (!user_mode(regs) && !search_exception_tables(regs->ARM_pc))
			goto no_context;
282
retry:
283
		mmap_read_lock(mm);
284 285 286 287 288 289 290
	} else {
		/*
		 * The above down_read_trylock() might have succeeded in
		 * which case, we'll have missed the might_sleep() from
		 * down_read()
		 */
		might_sleep();
291 292 293 294 295
#ifdef CONFIG_DEBUG_VM
		if (!user_mode(regs) &&
		    !search_exception_tables(regs->ARM_pc))
			goto no_context;
#endif
296 297
	}

K
Kefeng Wang 已提交
298
	fault = __do_page_fault(mm, addr, flags, vm_flags, regs);
299 300

	/* If we need to retry but a fatal signal is pending, handle the
301
	 * signal first. We do not need to release the mmap_lock because
302 303
	 * it would already be released in __lock_page_or_retry in
	 * mm/filemap.c. */
P
Peter Xu 已提交
304
	if (fault_signal_pending(fault, regs)) {
305 306
		if (!user_mode(regs))
			goto no_context;
307
		return 0;
308
	}
309

310
	if (!(fault & VM_FAULT_ERROR) && flags & FAULT_FLAG_ALLOW_RETRY) {
311
		if (fault & VM_FAULT_RETRY) {
312
			flags |= FAULT_FLAG_TRIED;
313 314 315 316
			goto retry;
		}
	}

317
	mmap_read_unlock(mm);
318

L
Linus Torvalds 已提交
319
	/*
J
Jan Kara 已提交
320
	 * Handle the "normal" case first - VM_FAULT_MAJOR
L
Linus Torvalds 已提交
321
	 */
322
	if (likely(!(fault & (VM_FAULT_ERROR | VM_FAULT_BADMAP | VM_FAULT_BADACCESS))))
L
Linus Torvalds 已提交
323 324
		return 0;

325 326 327 328 329 330 331
	/*
	 * If we are in kernel mode at this point, we
	 * have no context to handle this fault with.
	 */
	if (!user_mode(regs))
		goto no_context;

332 333 334 335 336 337 338 339 340 341
	if (fault & VM_FAULT_OOM) {
		/*
		 * We ran out of memory, call the OOM killer, and return to
		 * userspace (which will retry the fault, or kill us if we
		 * got oom-killed)
		 */
		pagefault_out_of_memory();
		return 0;
	}

N
Nick Piggin 已提交
342
	if (fault & VM_FAULT_SIGBUS) {
A
akpm@osdl.org 已提交
343 344 345 346 347 348
		/*
		 * We had some memory, but were unable to
		 * successfully fix up this page fault.
		 */
		sig = SIGBUS;
		code = BUS_ADRERR;
N
Nick Piggin 已提交
349
	} else {
A
akpm@osdl.org 已提交
350 351 352 353 354 355 356
		/*
		 * Something tried to access memory that
		 * isn't in our memory map..
		 */
		sig = SIGSEGV;
		code = fault == VM_FAULT_BADACCESS ?
			SEGV_ACCERR : SEGV_MAPERR;
L
Linus Torvalds 已提交
357 358
	}

359
	__do_user_fault(addr, fsr, sig, code, regs);
A
akpm@osdl.org 已提交
360
	return 0;
L
Linus Torvalds 已提交
361 362 363 364 365

no_context:
	__do_kernel_fault(mm, addr, fsr, regs);
	return 0;
}
366 367 368 369 370 371 372
#else					/* CONFIG_MMU */
static int
do_page_fault(unsigned long addr, unsigned int fsr, struct pt_regs *regs)
{
	return 0;
}
#endif					/* CONFIG_MMU */
L
Linus Torvalds 已提交
373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390

/*
 * First Level Translation Fault Handler
 *
 * We enter here because the first level page table doesn't contain
 * a valid entry for the address.
 *
 * If the address is in kernel space (>= TASK_SIZE), then we are
 * probably faulting in the vmalloc() area.
 *
 * If the init_task's first level page tables contains the relevant
 * entry, we copy the it to this task.  If not, we send the process
 * a signal, fixup the exception, or oops the kernel.
 *
 * NOTE! We MUST NOT take any locks for this case. We may be in an
 * interrupt or a critical region, and should only copy the information
 * from the master page table, nothing more.
 */
391
#ifdef CONFIG_MMU
392
static int __kprobes
L
Linus Torvalds 已提交
393 394 395 396 397
do_translation_fault(unsigned long addr, unsigned int fsr,
		     struct pt_regs *regs)
{
	unsigned int index;
	pgd_t *pgd, *pgd_k;
398
	p4d_t *p4d, *p4d_k;
R
Russell King 已提交
399
	pud_t *pud, *pud_k;
L
Linus Torvalds 已提交
400 401 402 403 404
	pmd_t *pmd, *pmd_k;

	if (addr < TASK_SIZE)
		return do_page_fault(addr, fsr, regs);

405 406 407
	if (user_mode(regs))
		goto bad_area;

L
Linus Torvalds 已提交
408 409 410 411 412
	index = pgd_index(addr);

	pgd = cpu_get_pgd() + index;
	pgd_k = init_mm.pgd + index;

413 414 415 416
	p4d = p4d_offset(pgd, addr);
	p4d_k = p4d_offset(pgd_k, addr);

	if (p4d_none(*p4d_k))
L
Linus Torvalds 已提交
417
		goto bad_area;
418 419
	if (!p4d_present(*p4d))
		set_p4d(p4d, *p4d_k);
L
Linus Torvalds 已提交
420

421 422
	pud = pud_offset(p4d, addr);
	pud_k = pud_offset(p4d_k, addr);
R
Russell King 已提交
423 424 425 426 427 428 429 430

	if (pud_none(*pud_k))
		goto bad_area;
	if (!pud_present(*pud))
		set_pud(pud, *pud_k);

	pmd = pmd_offset(pud, addr);
	pmd_k = pmd_offset(pud_k, addr);
L
Linus Torvalds 已提交
431

432 433 434 435 436 437
#ifdef CONFIG_ARM_LPAE
	/*
	 * Only one hardware entry per PMD with LPAE.
	 */
	index = 0;
#else
438 439 440 441 442 443 444 445 446
	/*
	 * On ARM one Linux PGD entry contains two hardware entries (see page
	 * tables layout in pgtable.h). We normally guarantee that we always
	 * fill both L1 entries. But create_mapping() doesn't follow the rule.
	 * It can create inidividual L1 entries, so here we have to call
	 * pmd_none() check for the entry really corresponded to address, not
	 * for the first of pair.
	 */
	index = (addr >> SECTION_SHIFT) & 1;
447
#endif
448
	if (pmd_none(pmd_k[index]))
L
Linus Torvalds 已提交
449 450 451 452 453 454
		goto bad_area;

	copy_pmd(pmd, pmd_k);
	return 0;

bad_area:
455
	do_bad_area(addr, fsr, regs);
L
Linus Torvalds 已提交
456 457
	return 0;
}
458 459 460 461 462 463 464 465
#else					/* CONFIG_MMU */
static int
do_translation_fault(unsigned long addr, unsigned int fsr,
		     struct pt_regs *regs)
{
	return 0;
}
#endif					/* CONFIG_MMU */
L
Linus Torvalds 已提交
466 467 468 469 470 471 472 473 474 475

/*
 * This abort handler always returns "fault".
 */
static int
do_bad(unsigned long addr, unsigned int fsr, struct pt_regs *regs)
{
	return 1;
}

476
struct fsr_info {
L
Linus Torvalds 已提交
477 478
	int	(*fn)(unsigned long addr, unsigned int fsr, struct pt_regs *regs);
	int	sig;
479
	int	code;
L
Linus Torvalds 已提交
480 481 482
	const char *name;
};

483
/* FSR definition */
484 485 486
#ifdef CONFIG_ARM_LPAE
#include "fsr-3level.c"
#else
487
#include "fsr-2level.c"
488
#endif
489

L
Linus Torvalds 已提交
490 491
void __init
hook_fault_code(int nr, int (*fn)(unsigned long, unsigned int, struct pt_regs *),
492
		int sig, int code, const char *name)
L
Linus Torvalds 已提交
493
{
494 495 496 497 498 499 500
	if (nr < 0 || nr >= ARRAY_SIZE(fsr_info))
		BUG();

	fsr_info[nr].fn   = fn;
	fsr_info[nr].sig  = sig;
	fsr_info[nr].code = code;
	fsr_info[nr].name = name;
L
Linus Torvalds 已提交
501 502 503 504 505
}

/*
 * Dispatch a data abort to the relevant handler.
 */
506
asmlinkage void
L
Linus Torvalds 已提交
507 508
do_DataAbort(unsigned long addr, unsigned int fsr, struct pt_regs *regs)
{
509
	const struct fsr_info *inf = fsr_info + fsr_fs(fsr);
L
Linus Torvalds 已提交
510

511
	if (!inf->fn(addr, fsr & ~FSR_LNX_PF, regs))
L
Linus Torvalds 已提交
512 513
		return;

514
	pr_alert("8<--- cut here ---\n");
R
Russell King 已提交
515
	pr_alert("Unhandled fault: %s (0x%03x) at 0x%08lx\n",
L
Linus Torvalds 已提交
516
		inf->name, fsr, addr);
517
	show_pte(KERN_ALERT, current->mm, addr);
518

519 520
	arm_notify_die("", regs, inf->sig, inf->code, (void __user *)addr,
		       fsr, 0);
L
Linus Torvalds 已提交
521 522
}

523 524 525 526 527 528 529 530 531 532 533 534 535
void __init
hook_ifault_code(int nr, int (*fn)(unsigned long, unsigned int, struct pt_regs *),
		 int sig, int code, const char *name)
{
	if (nr < 0 || nr >= ARRAY_SIZE(ifsr_info))
		BUG();

	ifsr_info[nr].fn   = fn;
	ifsr_info[nr].sig  = sig;
	ifsr_info[nr].code = code;
	ifsr_info[nr].name = name;
}

536
asmlinkage void
537
do_PrefetchAbort(unsigned long addr, unsigned int ifsr, struct pt_regs *regs)
L
Linus Torvalds 已提交
538
{
539 540 541 542 543
	const struct fsr_info *inf = ifsr_info + fsr_fs(ifsr);

	if (!inf->fn(addr, ifsr | FSR_LNX_PF, regs))
		return;

R
Russell King 已提交
544
	pr_alert("Unhandled prefetch abort: %s (0x%03x) at 0x%08lx\n",
545 546
		inf->name, ifsr, addr);

547 548
	arm_notify_die("", regs, inf->sig, inf->code, (void __user *)addr,
		       ifsr, 0);
L
Linus Torvalds 已提交
549 550
}

551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567
/*
 * Abort handler to be used only during first unmasking of asynchronous aborts
 * on the boot CPU. This makes sure that the machine will not die if the
 * firmware/bootloader left an imprecise abort pending for us to trip over.
 */
static int __init early_abort_handler(unsigned long addr, unsigned int fsr,
				      struct pt_regs *regs)
{
	pr_warn("Hit pending asynchronous external abort (FSR=0x%08x) during "
		"first unmask, this is most likely caused by a "
		"firmware/bootloader bug.\n", fsr);

	return 0;
}

void __init early_abt_enable(void)
{
568
	fsr_info[FSR_FS_AEA].fn = early_abort_handler;
569
	local_abt_enable();
570
	fsr_info[FSR_FS_AEA].fn = do_bad;
571 572
}

573
#ifndef CONFIG_ARM_LPAE
574 575 576 577 578 579 580
static int __init exceptions_init(void)
{
	if (cpu_architecture() >= CPU_ARCH_ARMv6) {
		hook_fault_code(4, do_translation_fault, SIGSEGV, SEGV_MAPERR,
				"I-cache maintenance fault");
	}

581 582 583 584 585 586 587 588 589 590 591
	if (cpu_architecture() >= CPU_ARCH_ARMv7) {
		/*
		 * TODO: Access flag faults introduced in ARMv6K.
		 * Runtime check for 'K' extension is needed
		 */
		hook_fault_code(3, do_bad, SIGSEGV, SEGV_MAPERR,
				"section access flag fault");
		hook_fault_code(6, do_bad, SIGSEGV, SEGV_MAPERR,
				"section access flag fault");
	}

592 593 594 595
	return 0;
}

arch_initcall(exceptions_init);
596
#endif