2 * Based on arch/arm/mm/fault.c
4 * Copyright (C) 1995 Linus Torvalds
5 * Copyright (C) 1995-2004 Russell King
6 * Copyright (C) 2012 ARM Ltd.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 as
10 * published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
17 * You should have received a copy of the GNU General Public License
18 * along with this program. If not, see <http://www.gnu.org/licenses/>.
21 #include <linux/module.h>
22 #include <linux/signal.h>
24 #include <linux/hardirq.h>
25 #include <linux/init.h>
26 #include <linux/kprobes.h>
27 #include <linux/uaccess.h>
28 #include <linux/page-flags.h>
29 #include <linux/sched.h>
30 #include <linux/highmem.h>
31 #include <linux/perf_event.h>
32 #include <linux/preempt.h>
35 #include <asm/cpufeature.h>
36 #include <asm/exception.h>
37 #include <asm/debug-monitors.h>
39 #include <asm/sysreg.h>
40 #include <asm/system_misc.h>
41 #include <asm/pgtable.h>
42 #include <asm/tlbflush.h>
44 static const char *fault_name(unsigned int esr
);
47 * Dump out the page tables associated with 'addr' in mm 'mm'.
49 void show_pte(struct mm_struct
*mm
, unsigned long addr
)
56 pr_alert("pgd = %p\n", mm
->pgd
);
57 pgd
= pgd_offset(mm
, addr
);
58 pr_alert("[%08lx] *pgd=%016llx", addr
, pgd_val(*pgd
));
65 if (pgd_none(*pgd
) || pgd_bad(*pgd
))
68 pud
= pud_offset(pgd
, addr
);
69 pr_cont(", *pud=%016llx", pud_val(*pud
));
70 if (pud_none(*pud
) || pud_bad(*pud
))
73 pmd
= pmd_offset(pud
, addr
);
74 pr_cont(", *pmd=%016llx", pmd_val(*pmd
));
75 if (pmd_none(*pmd
) || pmd_bad(*pmd
))
78 pte
= pte_offset_map(pmd
, addr
);
79 pr_cont(", *pte=%016llx", pte_val(*pte
));
86 #ifdef CONFIG_ARM64_HW_AFDBM
88 * This function sets the access flags (dirty, accessed), as well as write
89 * permission, and only to a more permissive setting.
91 * It needs to cope with hardware update of the accessed/dirty state by other
92 * agents in the system and can safely skip the __sync_icache_dcache() call as,
93 * like set_pte_at(), the PTE is never changed from no-exec to exec here.
95 * Returns whether or not the PTE actually changed.
97 int ptep_set_access_flags(struct vm_area_struct
*vma
,
98 unsigned long address
, pte_t
*ptep
,
99 pte_t entry
, int dirty
)
104 if (pte_same(*ptep
, entry
))
107 /* only preserve the access flags and write permission */
108 pte_val(entry
) &= PTE_AF
| PTE_WRITE
| PTE_DIRTY
;
111 * PTE_RDONLY is cleared by default in the asm below, so set it in
112 * back if necessary (read-only or clean PTE).
114 if (!pte_write(entry
) || !pte_sw_dirty(entry
))
115 pte_val(entry
) |= PTE_RDONLY
;
118 * Setting the flags must be done atomically to avoid racing with the
119 * hardware update of the access/dirty state.
121 asm volatile("// ptep_set_access_flags\n"
122 " prfm pstl1strm, %2\n"
124 " and %0, %0, %3 // clear PTE_RDONLY\n"
125 " orr %0, %0, %4 // set flags\n"
126 " stxr %w1, %0, %2\n"
128 : "=&r" (old_pteval
), "=&r" (tmp
), "+Q" (pte_val(*ptep
))
129 : "L" (~PTE_RDONLY
), "r" (pte_val(entry
)));
131 flush_tlb_fix_spurious_fault(vma
, address
);
136 static bool is_el1_instruction_abort(unsigned int esr
)
138 return ESR_ELx_EC(esr
) == ESR_ELx_EC_IABT_CUR
;
142 * The kernel tried to access some page that wasn't present.
144 static void __do_kernel_fault(struct mm_struct
*mm
, unsigned long addr
,
145 unsigned int esr
, struct pt_regs
*regs
)
148 * Are we prepared to handle this kernel fault?
149 * We are almost certainly not prepared to handle instruction faults.
151 if (!is_el1_instruction_abort(esr
) && fixup_exception(regs
))
155 * No handler, we'll have to terminate things with extreme prejudice.
158 pr_alert("Unable to handle kernel %s at virtual address %08lx\n",
159 (addr
< PAGE_SIZE
) ? "NULL pointer dereference" :
160 "paging request", addr
);
163 die("Oops", regs
, esr
);
169 * Something tried to access memory that isn't in our memory map. User mode
170 * accesses just cause a SIGSEGV
172 static void __do_user_fault(struct task_struct
*tsk
, unsigned long addr
,
173 unsigned int esr
, unsigned int sig
, int code
,
174 struct pt_regs
*regs
)
178 if (unhandled_signal(tsk
, sig
) && show_unhandled_signals_ratelimited()) {
179 pr_info("%s[%d]: unhandled %s (%d) at 0x%08lx, esr 0x%03x\n",
180 tsk
->comm
, task_pid_nr(tsk
), fault_name(esr
), sig
,
182 show_pte(tsk
->mm
, addr
);
186 tsk
->thread
.fault_address
= addr
;
187 tsk
->thread
.fault_code
= esr
;
191 si
.si_addr
= (void __user
*)addr
;
192 force_sig_info(sig
, &si
, tsk
);
195 static void do_bad_area(unsigned long addr
, unsigned int esr
, struct pt_regs
*regs
)
197 struct task_struct
*tsk
= current
;
198 struct mm_struct
*mm
= tsk
->active_mm
;
201 * If we are in kernel mode at this point, we have no context to
202 * handle this fault with.
205 __do_user_fault(tsk
, addr
, esr
, SIGSEGV
, SEGV_MAPERR
, regs
);
207 __do_kernel_fault(mm
, addr
, esr
, regs
);
210 #define VM_FAULT_BADMAP 0x010000
211 #define VM_FAULT_BADACCESS 0x020000
213 static int __do_page_fault(struct mm_struct
*mm
, unsigned long addr
,
214 unsigned int mm_flags
, unsigned long vm_flags
,
215 struct task_struct
*tsk
)
217 struct vm_area_struct
*vma
;
220 vma
= find_vma(mm
, addr
);
221 fault
= VM_FAULT_BADMAP
;
224 if (unlikely(vma
->vm_start
> addr
))
228 * Ok, we have a good vm_area for this memory access, so we can handle
233 * Check that the permissions on the VMA allow for the fault which
234 * occurred. If we encountered a write or exec fault, we must have
235 * appropriate permissions, otherwise we allow any permission.
237 if (!(vma
->vm_flags
& vm_flags
)) {
238 fault
= VM_FAULT_BADACCESS
;
242 return handle_mm_fault(mm
, vma
, addr
& PAGE_MASK
, mm_flags
);
245 if (vma
->vm_flags
& VM_GROWSDOWN
&& !expand_stack(vma
, addr
))
251 static inline bool is_permission_fault(unsigned int esr
, struct pt_regs
*regs
)
253 unsigned int ec
= ESR_ELx_EC(esr
);
254 unsigned int fsc_type
= esr
& ESR_ELx_FSC_TYPE
;
256 if (ec
!= ESR_ELx_EC_DABT_CUR
&& ec
!= ESR_ELx_EC_IABT_CUR
)
259 if (system_uses_ttbr0_pan())
260 return fsc_type
== ESR_ELx_FSC_FAULT
&&
261 (regs
->pstate
& PSR_PAN_BIT
);
263 return fsc_type
== ESR_ELx_FSC_PERM
;
266 static bool is_el0_instruction_abort(unsigned int esr
)
268 return ESR_ELx_EC(esr
) == ESR_ELx_EC_IABT_LOW
;
271 static int __kprobes
do_page_fault(unsigned long addr
, unsigned int esr
,
272 struct pt_regs
*regs
)
274 struct task_struct
*tsk
;
275 struct mm_struct
*mm
;
276 int fault
, sig
, code
;
277 unsigned long vm_flags
= VM_READ
| VM_WRITE
| VM_EXEC
;
278 unsigned int mm_flags
= FAULT_FLAG_ALLOW_RETRY
| FAULT_FLAG_KILLABLE
;
283 /* Enable interrupts if they were enabled in the parent context. */
284 if (interrupts_enabled(regs
))
288 * If we're in an interrupt or have no user context, we must not take
291 if (faulthandler_disabled() || !mm
)
295 mm_flags
|= FAULT_FLAG_USER
;
297 if (is_el0_instruction_abort(esr
)) {
299 } else if ((esr
& ESR_ELx_WNR
) && !(esr
& ESR_ELx_CM
)) {
301 mm_flags
|= FAULT_FLAG_WRITE
;
304 if (addr
< USER_DS
&& is_permission_fault(esr
, regs
)) {
305 /* regs->orig_addr_limit may be 0 if we entered from EL0 */
306 if (regs
->orig_addr_limit
== KERNEL_DS
)
307 die("Accessing user space memory with fs=KERNEL_DS", regs
, esr
);
309 if (is_el1_instruction_abort(esr
))
310 die("Attempting to execute userspace memory", regs
, esr
);
312 if (!search_exception_tables(regs
->pc
))
313 die("Accessing user space memory outside uaccess.h routines", regs
, esr
);
317 * As per x86, we may deadlock here. However, since the kernel only
318 * validly references user space from well defined areas of the code,
319 * we can bug out early if this is from code which shouldn't.
321 if (!down_read_trylock(&mm
->mmap_sem
)) {
322 if (!user_mode(regs
) && !search_exception_tables(regs
->pc
))
325 down_read(&mm
->mmap_sem
);
328 * The above down_read_trylock() might have succeeded in which
329 * case, we'll have missed the might_sleep() from down_read().
332 #ifdef CONFIG_DEBUG_VM
333 if (!user_mode(regs
) && !search_exception_tables(regs
->pc
))
338 fault
= __do_page_fault(mm
, addr
, mm_flags
, vm_flags
, tsk
);
341 * If we need to retry but a fatal signal is pending, handle the
342 * signal first. We do not need to release the mmap_sem because it
343 * would already be released in __lock_page_or_retry in mm/filemap.c.
345 if ((fault
& VM_FAULT_RETRY
) && fatal_signal_pending(current
)) {
346 if (!user_mode(regs
))
352 * Major/minor page fault accounting is only done on the initial
353 * attempt. If we go through a retry, it is extremely likely that the
354 * page will be found in page cache at that point.
357 perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS
, 1, regs
, addr
);
358 if (mm_flags
& FAULT_FLAG_ALLOW_RETRY
) {
359 if (fault
& VM_FAULT_MAJOR
) {
361 perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ
, 1, regs
,
365 perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN
, 1, regs
,
368 if (fault
& VM_FAULT_RETRY
) {
370 * Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk of
373 mm_flags
&= ~FAULT_FLAG_ALLOW_RETRY
;
374 mm_flags
|= FAULT_FLAG_TRIED
;
379 up_read(&mm
->mmap_sem
);
382 * Handle the "normal" case first - VM_FAULT_MAJOR / VM_FAULT_MINOR
384 if (likely(!(fault
& (VM_FAULT_ERROR
| VM_FAULT_BADMAP
|
385 VM_FAULT_BADACCESS
))))
389 * If we are in kernel mode at this point, we have no context to
390 * handle this fault with.
392 if (!user_mode(regs
))
395 if (fault
& VM_FAULT_OOM
) {
397 * We ran out of memory, call the OOM killer, and return to
398 * userspace (which will retry the fault, or kill us if we got
401 pagefault_out_of_memory();
405 if (fault
& VM_FAULT_SIGBUS
) {
407 * We had some memory, but were unable to successfully fix up
414 * Something tried to access memory that isn't in our memory
418 code
= fault
== VM_FAULT_BADACCESS
?
419 SEGV_ACCERR
: SEGV_MAPERR
;
422 __do_user_fault(tsk
, addr
, esr
, sig
, code
, regs
);
426 __do_kernel_fault(mm
, addr
, esr
, regs
);
431 * First Level Translation Fault Handler
433 * We enter here because the first level page table doesn't contain a valid
434 * entry for the address.
436 * If the address is in kernel space (>= TASK_SIZE), then we are probably
437 * faulting in the vmalloc() area.
439 * If the init_task's first level page tables contains the relevant entry, we
440 * copy the it to this task. If not, we send the process a signal, fixup the
441 * exception, or oops the kernel.
443 * NOTE! We MUST NOT take any locks for this case. We may be in an interrupt
444 * or a critical region, and should only copy the information from the master
445 * page table, nothing more.
447 static int __kprobes
do_translation_fault(unsigned long addr
,
449 struct pt_regs
*regs
)
451 if (addr
< TASK_SIZE
)
452 return do_page_fault(addr
, esr
, regs
);
454 do_bad_area(addr
, esr
, regs
);
459 * This abort handler always returns "fault".
461 static int do_bad(unsigned long addr
, unsigned int esr
, struct pt_regs
*regs
)
466 static const struct fault_info
{
467 int (*fn
)(unsigned long addr
, unsigned int esr
, struct pt_regs
*regs
);
472 { do_bad
, SIGBUS
, 0, "ttbr address size fault" },
473 { do_bad
, SIGBUS
, 0, "level 1 address size fault" },
474 { do_bad
, SIGBUS
, 0, "level 2 address size fault" },
475 { do_bad
, SIGBUS
, 0, "level 3 address size fault" },
476 { do_translation_fault
, SIGSEGV
, SEGV_MAPERR
, "level 0 translation fault" },
477 { do_translation_fault
, SIGSEGV
, SEGV_MAPERR
, "level 1 translation fault" },
478 { do_translation_fault
, SIGSEGV
, SEGV_MAPERR
, "level 2 translation fault" },
479 { do_page_fault
, SIGSEGV
, SEGV_MAPERR
, "level 3 translation fault" },
480 { do_bad
, SIGBUS
, 0, "unknown 8" },
481 { do_page_fault
, SIGSEGV
, SEGV_ACCERR
, "level 1 access flag fault" },
482 { do_page_fault
, SIGSEGV
, SEGV_ACCERR
, "level 2 access flag fault" },
483 { do_page_fault
, SIGSEGV
, SEGV_ACCERR
, "level 3 access flag fault" },
484 { do_bad
, SIGBUS
, 0, "unknown 12" },
485 { do_page_fault
, SIGSEGV
, SEGV_ACCERR
, "level 1 permission fault" },
486 { do_page_fault
, SIGSEGV
, SEGV_ACCERR
, "level 2 permission fault" },
487 { do_page_fault
, SIGSEGV
, SEGV_ACCERR
, "level 3 permission fault" },
488 { do_bad
, SIGBUS
, 0, "synchronous external abort" },
489 { do_bad
, SIGBUS
, 0, "unknown 17" },
490 { do_bad
, SIGBUS
, 0, "unknown 18" },
491 { do_bad
, SIGBUS
, 0, "unknown 19" },
492 { do_bad
, SIGBUS
, 0, "synchronous abort (translation table walk)" },
493 { do_bad
, SIGBUS
, 0, "synchronous abort (translation table walk)" },
494 { do_bad
, SIGBUS
, 0, "synchronous abort (translation table walk)" },
495 { do_bad
, SIGBUS
, 0, "synchronous abort (translation table walk)" },
496 { do_bad
, SIGBUS
, 0, "synchronous parity error" },
497 { do_bad
, SIGBUS
, 0, "unknown 25" },
498 { do_bad
, SIGBUS
, 0, "unknown 26" },
499 { do_bad
, SIGBUS
, 0, "unknown 27" },
500 { do_bad
, SIGBUS
, 0, "synchronous parity error (translation table walk)" },
501 { do_bad
, SIGBUS
, 0, "synchronous parity error (translation table walk)" },
502 { do_bad
, SIGBUS
, 0, "synchronous parity error (translation table walk)" },
503 { do_bad
, SIGBUS
, 0, "synchronous parity error (translation table walk)" },
504 { do_bad
, SIGBUS
, 0, "unknown 32" },
505 { do_bad
, SIGBUS
, BUS_ADRALN
, "alignment fault" },
506 { do_bad
, SIGBUS
, 0, "unknown 34" },
507 { do_bad
, SIGBUS
, 0, "unknown 35" },
508 { do_bad
, SIGBUS
, 0, "unknown 36" },
509 { do_bad
, SIGBUS
, 0, "unknown 37" },
510 { do_bad
, SIGBUS
, 0, "unknown 38" },
511 { do_bad
, SIGBUS
, 0, "unknown 39" },
512 { do_bad
, SIGBUS
, 0, "unknown 40" },
513 { do_bad
, SIGBUS
, 0, "unknown 41" },
514 { do_bad
, SIGBUS
, 0, "unknown 42" },
515 { do_bad
, SIGBUS
, 0, "unknown 43" },
516 { do_bad
, SIGBUS
, 0, "unknown 44" },
517 { do_bad
, SIGBUS
, 0, "unknown 45" },
518 { do_bad
, SIGBUS
, 0, "unknown 46" },
519 { do_bad
, SIGBUS
, 0, "unknown 47" },
520 { do_bad
, SIGBUS
, 0, "TLB conflict abort" },
521 { do_bad
, SIGBUS
, 0, "unknown 49" },
522 { do_bad
, SIGBUS
, 0, "unknown 50" },
523 { do_bad
, SIGBUS
, 0, "unknown 51" },
524 { do_bad
, SIGBUS
, 0, "implementation fault (lockdown abort)" },
525 { do_bad
, SIGBUS
, 0, "implementation fault (unsupported exclusive)" },
526 { do_bad
, SIGBUS
, 0, "unknown 54" },
527 { do_bad
, SIGBUS
, 0, "unknown 55" },
528 { do_bad
, SIGBUS
, 0, "unknown 56" },
529 { do_bad
, SIGBUS
, 0, "unknown 57" },
530 { do_bad
, SIGBUS
, 0, "unknown 58" },
531 { do_bad
, SIGBUS
, 0, "unknown 59" },
532 { do_bad
, SIGBUS
, 0, "unknown 60" },
533 { do_bad
, SIGBUS
, 0, "section domain fault" },
534 { do_bad
, SIGBUS
, 0, "page domain fault" },
535 { do_bad
, SIGBUS
, 0, "unknown 63" },
538 static const char *fault_name(unsigned int esr
)
540 const struct fault_info
*inf
= fault_info
+ (esr
& 63);
545 * Dispatch a data abort to the relevant handler.
547 asmlinkage
void __exception
do_mem_abort(unsigned long addr
, unsigned int esr
,
548 struct pt_regs
*regs
)
550 const struct fault_info
*inf
= fault_info
+ (esr
& 63);
553 if (!inf
->fn(addr
, esr
, regs
))
556 pr_alert("Unhandled fault: %s (0x%08x) at 0x%016lx\n",
557 inf
->name
, esr
, addr
);
559 info
.si_signo
= inf
->sig
;
561 info
.si_code
= inf
->code
;
562 info
.si_addr
= (void __user
*)addr
;
563 arm64_notify_die("", regs
, &info
, esr
);
567 * Handle stack alignment exceptions.
569 asmlinkage
void __exception
do_sp_pc_abort(unsigned long addr
,
571 struct pt_regs
*regs
)
574 struct task_struct
*tsk
= current
;
576 if (show_unhandled_signals
&& unhandled_signal(tsk
, SIGBUS
))
577 pr_info_ratelimited("%s[%d]: %s exception: pc=%p sp=%p\n",
578 tsk
->comm
, task_pid_nr(tsk
),
579 esr_get_class_string(esr
), (void *)regs
->pc
,
582 info
.si_signo
= SIGBUS
;
584 info
.si_code
= BUS_ADRALN
;
585 info
.si_addr
= (void __user
*)addr
;
586 arm64_notify_die("Oops - SP/PC alignment exception", regs
, &info
, esr
);
589 int __init
early_brk64(unsigned long addr
, unsigned int esr
,
590 struct pt_regs
*regs
);
593 * __refdata because early_brk64 is __init, but the reference to it is
594 * clobbered at arch_initcall time.
595 * See traps.c and debug-monitors.c:debug_traps_init().
597 static struct fault_info __refdata debug_fault_info
[] = {
598 { do_bad
, SIGTRAP
, TRAP_HWBKPT
, "hardware breakpoint" },
599 { do_bad
, SIGTRAP
, TRAP_HWBKPT
, "hardware single-step" },
600 { do_bad
, SIGTRAP
, TRAP_HWBKPT
, "hardware watchpoint" },
601 { do_bad
, SIGBUS
, 0, "unknown 3" },
602 { do_bad
, SIGTRAP
, TRAP_BRKPT
, "aarch32 BKPT" },
603 { do_bad
, SIGTRAP
, 0, "aarch32 vector catch" },
604 { early_brk64
, SIGTRAP
, TRAP_BRKPT
, "aarch64 BRK" },
605 { do_bad
, SIGBUS
, 0, "unknown 7" },
608 void __init
hook_debug_fault_code(int nr
,
609 int (*fn
)(unsigned long, unsigned int, struct pt_regs
*),
610 int sig
, int code
, const char *name
)
612 BUG_ON(nr
< 0 || nr
>= ARRAY_SIZE(debug_fault_info
));
614 debug_fault_info
[nr
].fn
= fn
;
615 debug_fault_info
[nr
].sig
= sig
;
616 debug_fault_info
[nr
].code
= code
;
617 debug_fault_info
[nr
].name
= name
;
620 asmlinkage
int __exception
do_debug_exception(unsigned long addr
,
622 struct pt_regs
*regs
)
624 const struct fault_info
*inf
= debug_fault_info
+ DBG_ESR_EVT(esr
);
627 if (!inf
->fn(addr
, esr
, regs
))
630 pr_alert("Unhandled debug exception: %s (0x%08x) at 0x%016lx\n",
631 inf
->name
, esr
, addr
);
633 info
.si_signo
= inf
->sig
;
635 info
.si_code
= inf
->code
;
636 info
.si_addr
= (void __user
*)addr
;
637 arm64_notify_die("", regs
, &info
, 0);
642 #ifdef CONFIG_ARM64_PAN
643 int cpu_enable_pan(void *__unused
)
646 * We modify PSTATE. This won't work from irq context as the PSTATE
647 * is discarded once we return from the exception.
649 WARN_ON_ONCE(in_interrupt());
651 config_sctlr_el1(SCTLR_EL1_SPAN
, 0);
652 asm(SET_PSTATE_PAN(1));
655 #endif /* CONFIG_ARM64_PAN */
657 #ifdef CONFIG_ARM64_UAO
659 * Kernel threads have fs=KERNEL_DS by default, and don't need to call
660 * set_fs(), devtmpfs in particular relies on this behaviour.
661 * We need to enable the feature at runtime (instead of adding it to
662 * PSR_MODE_EL1h) as the feature may not be implemented by the cpu.
664 int cpu_enable_uao(void *__unused
)
666 asm(SET_PSTATE_UAO(1));
669 #endif /* CONFIG_ARM64_UAO */