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;
110 /* set PTE_RDONLY if actual read-only or clean PTE */
111 if (!pte_write(entry) || !pte_sw_dirty(entry))
112 pte_val(entry) |= PTE_RDONLY;
115 * Setting the flags must be done atomically to avoid racing with the
116 * hardware update of the access/dirty state. The PTE_RDONLY bit must
117 * be set to the most permissive (lowest value) of *ptep and entry
118 * (calculated as: a & b == ~(~a | ~b)).
120 pte_val(entry) ^= PTE_RDONLY;
121 asm volatile("// ptep_set_access_flags\n"
122 " prfm pstl1strm, %2\n"
124 " eor %0, %0, %3 // negate PTE_RDONLY in *ptep\n"
125 " orr %0, %0, %4 // set flags\n"
126 " eor %0, %0, %3 // negate final PTE_RDONLY\n"
127 " stxr %w1, %0, %2\n"
129 : "=&r" (old_pteval), "=&r" (tmp), "+Q" (pte_val(*ptep))
130 : "L" (PTE_RDONLY), "r" (pte_val(entry)));
132 flush_tlb_fix_spurious_fault(vma, address);
138 * The kernel tried to access some page that wasn't present.
140 static void __do_kernel_fault(struct mm_struct *mm, unsigned long addr,
141 unsigned int esr, struct pt_regs *regs)
144 * Are we prepared to handle this kernel fault?
146 if (fixup_exception(regs))
150 * No handler, we'll have to terminate things with extreme prejudice.
153 pr_alert("Unable to handle kernel %s at virtual address %08lx\n",
154 (addr < PAGE_SIZE) ? "NULL pointer dereference" :
155 "paging request", addr);
158 die("Oops", regs, esr);
164 * Something tried to access memory that isn't in our memory map. User mode
165 * accesses just cause a SIGSEGV
167 static void __do_user_fault(struct task_struct *tsk, unsigned long addr,
168 unsigned int esr, unsigned int sig, int code,
169 struct pt_regs *regs)
173 if (unhandled_signal(tsk, sig) && show_unhandled_signals_ratelimited()) {
174 pr_info("%s[%d]: unhandled %s (%d) at 0x%08lx, esr 0x%03x\n",
175 tsk->comm, task_pid_nr(tsk), fault_name(esr), sig,
177 show_pte(tsk->mm, addr);
181 tsk->thread.fault_address = addr;
182 tsk->thread.fault_code = esr;
186 si.si_addr = (void __user *)addr;
187 force_sig_info(sig, &si, tsk);
190 static void do_bad_area(unsigned long addr, unsigned int esr, struct pt_regs *regs)
192 struct task_struct *tsk = current;
193 struct mm_struct *mm = tsk->active_mm;
196 * If we are in kernel mode at this point, we have no context to
197 * handle this fault with.
200 __do_user_fault(tsk, addr, esr, SIGSEGV, SEGV_MAPERR, regs);
202 __do_kernel_fault(mm, addr, esr, regs);
205 #define VM_FAULT_BADMAP 0x010000
206 #define VM_FAULT_BADACCESS 0x020000
208 #define ESR_LNX_EXEC (1 << 24)
210 static int __do_page_fault(struct mm_struct *mm, unsigned long addr,
211 unsigned int mm_flags, unsigned long vm_flags,
212 struct task_struct *tsk)
214 struct vm_area_struct *vma;
217 vma = find_vma(mm, addr);
218 fault = VM_FAULT_BADMAP;
221 if (unlikely(vma->vm_start > addr))
225 * Ok, we have a good vm_area for this memory access, so we can handle
230 * Check that the permissions on the VMA allow for the fault which
231 * occurred. If we encountered a write or exec fault, we must have
232 * appropriate permissions, otherwise we allow any permission.
234 if (!(vma->vm_flags & vm_flags)) {
235 fault = VM_FAULT_BADACCESS;
239 return handle_mm_fault(mm, vma, addr & PAGE_MASK, mm_flags);
242 if (vma->vm_flags & VM_GROWSDOWN && !expand_stack(vma, addr))
248 static int __kprobes do_page_fault(unsigned long addr, unsigned int esr,
249 struct pt_regs *regs)
251 struct task_struct *tsk;
252 struct mm_struct *mm;
253 int fault, sig, code;
254 unsigned long vm_flags = VM_READ | VM_WRITE | VM_EXEC;
255 unsigned int mm_flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
260 /* Enable interrupts if they were enabled in the parent context. */
261 if (interrupts_enabled(regs))
265 * If we're in an interrupt or have no user context, we must not take
268 if (faulthandler_disabled() || !mm)
272 mm_flags |= FAULT_FLAG_USER;
274 if (esr & ESR_LNX_EXEC) {
276 } else if ((esr & ESR_ELx_WNR) && !(esr & ESR_ELx_CM)) {
278 mm_flags |= FAULT_FLAG_WRITE;
282 * PAN bit set implies the fault happened in kernel space, but not
283 * in the arch's user access functions.
285 if (IS_ENABLED(CONFIG_ARM64_PAN) && (regs->pstate & PSR_PAN_BIT))
289 * As per x86, we may deadlock here. However, since the kernel only
290 * validly references user space from well defined areas of the code,
291 * we can bug out early if this is from code which shouldn't.
293 if (!down_read_trylock(&mm->mmap_sem)) {
294 if (!user_mode(regs) && !search_exception_tables(regs->pc))
297 down_read(&mm->mmap_sem);
300 * The above down_read_trylock() might have succeeded in which
301 * case, we'll have missed the might_sleep() from down_read().
304 #ifdef CONFIG_DEBUG_VM
305 if (!user_mode(regs) && !search_exception_tables(regs->pc))
310 fault = __do_page_fault(mm, addr, mm_flags, vm_flags, tsk);
313 * If we need to retry but a fatal signal is pending, handle the
314 * signal first. We do not need to release the mmap_sem because it
315 * would already be released in __lock_page_or_retry in mm/filemap.c.
317 if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current)) {
318 if (!user_mode(regs))
324 * Major/minor page fault accounting is only done on the initial
325 * attempt. If we go through a retry, it is extremely likely that the
326 * page will be found in page cache at that point.
329 perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr);
330 if (mm_flags & FAULT_FLAG_ALLOW_RETRY) {
331 if (fault & VM_FAULT_MAJOR) {
333 perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs,
337 perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs,
340 if (fault & VM_FAULT_RETRY) {
342 * Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk of
345 mm_flags &= ~FAULT_FLAG_ALLOW_RETRY;
346 mm_flags |= FAULT_FLAG_TRIED;
351 up_read(&mm->mmap_sem);
354 * Handle the "normal" case first - VM_FAULT_MAJOR / VM_FAULT_MINOR
356 if (likely(!(fault & (VM_FAULT_ERROR | VM_FAULT_BADMAP |
357 VM_FAULT_BADACCESS))))
361 * If we are in kernel mode at this point, we have no context to
362 * handle this fault with.
364 if (!user_mode(regs))
367 if (fault & VM_FAULT_OOM) {
369 * We ran out of memory, call the OOM killer, and return to
370 * userspace (which will retry the fault, or kill us if we got
373 pagefault_out_of_memory();
377 if (fault & VM_FAULT_SIGBUS) {
379 * We had some memory, but were unable to successfully fix up
386 * Something tried to access memory that isn't in our memory
390 code = fault == VM_FAULT_BADACCESS ?
391 SEGV_ACCERR : SEGV_MAPERR;
394 __do_user_fault(tsk, addr, esr, sig, code, regs);
398 __do_kernel_fault(mm, addr, esr, regs);
403 * First Level Translation Fault Handler
405 * We enter here because the first level page table doesn't contain a valid
406 * entry for the address.
408 * If the address is in kernel space (>= TASK_SIZE), then we are probably
409 * faulting in the vmalloc() area.
411 * If the init_task's first level page tables contains the relevant entry, we
412 * copy the it to this task. If not, we send the process a signal, fixup the
413 * exception, or oops the kernel.
415 * NOTE! We MUST NOT take any locks for this case. We may be in an interrupt
416 * or a critical region, and should only copy the information from the master
417 * page table, nothing more.
419 static int __kprobes do_translation_fault(unsigned long addr,
421 struct pt_regs *regs)
423 if (addr < TASK_SIZE)
424 return do_page_fault(addr, esr, regs);
426 do_bad_area(addr, esr, regs);
431 * This abort handler always returns "fault".
433 static int do_bad(unsigned long addr, unsigned int esr, struct pt_regs *regs)
438 static struct fault_info {
439 int (*fn)(unsigned long addr, unsigned int esr, struct pt_regs *regs);
444 { do_bad, SIGBUS, 0, "ttbr address size fault" },
445 { do_bad, SIGBUS, 0, "level 1 address size fault" },
446 { do_bad, SIGBUS, 0, "level 2 address size fault" },
447 { do_bad, SIGBUS, 0, "level 3 address size fault" },
448 { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 0 translation fault" },
449 { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 1 translation fault" },
450 { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 2 translation fault" },
451 { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 3 translation fault" },
452 { do_bad, SIGBUS, 0, "unknown 8" },
453 { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 1 access flag fault" },
454 { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 2 access flag fault" },
455 { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 3 access flag fault" },
456 { do_bad, SIGBUS, 0, "unknown 12" },
457 { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 1 permission fault" },
458 { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 2 permission fault" },
459 { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 3 permission fault" },
460 { do_bad, SIGBUS, 0, "synchronous external abort" },
461 { do_bad, SIGBUS, 0, "unknown 17" },
462 { do_bad, SIGBUS, 0, "unknown 18" },
463 { do_bad, SIGBUS, 0, "unknown 19" },
464 { do_bad, SIGBUS, 0, "synchronous abort (translation table walk)" },
465 { do_bad, SIGBUS, 0, "synchronous abort (translation table walk)" },
466 { do_bad, SIGBUS, 0, "synchronous abort (translation table walk)" },
467 { do_bad, SIGBUS, 0, "synchronous abort (translation table walk)" },
468 { do_bad, SIGBUS, 0, "synchronous parity error" },
469 { do_bad, SIGBUS, 0, "unknown 25" },
470 { do_bad, SIGBUS, 0, "unknown 26" },
471 { do_bad, SIGBUS, 0, "unknown 27" },
472 { do_bad, SIGBUS, 0, "synchronous parity error (translation table walk)" },
473 { do_bad, SIGBUS, 0, "synchronous parity error (translation table walk)" },
474 { do_bad, SIGBUS, 0, "synchronous parity error (translation table walk)" },
475 { do_bad, SIGBUS, 0, "synchronous parity error (translation table walk)" },
476 { do_bad, SIGBUS, 0, "unknown 32" },
477 { do_bad, SIGBUS, BUS_ADRALN, "alignment fault" },
478 { do_bad, SIGBUS, 0, "unknown 34" },
479 { do_bad, SIGBUS, 0, "unknown 35" },
480 { do_bad, SIGBUS, 0, "unknown 36" },
481 { do_bad, SIGBUS, 0, "unknown 37" },
482 { do_bad, SIGBUS, 0, "unknown 38" },
483 { do_bad, SIGBUS, 0, "unknown 39" },
484 { do_bad, SIGBUS, 0, "unknown 40" },
485 { do_bad, SIGBUS, 0, "unknown 41" },
486 { do_bad, SIGBUS, 0, "unknown 42" },
487 { do_bad, SIGBUS, 0, "unknown 43" },
488 { do_bad, SIGBUS, 0, "unknown 44" },
489 { do_bad, SIGBUS, 0, "unknown 45" },
490 { do_bad, SIGBUS, 0, "unknown 46" },
491 { do_bad, SIGBUS, 0, "unknown 47" },
492 { do_bad, SIGBUS, 0, "TLB conflict abort" },
493 { do_bad, SIGBUS, 0, "unknown 49" },
494 { do_bad, SIGBUS, 0, "unknown 50" },
495 { do_bad, SIGBUS, 0, "unknown 51" },
496 { do_bad, SIGBUS, 0, "implementation fault (lockdown abort)" },
497 { do_bad, SIGBUS, 0, "implementation fault (unsupported exclusive)" },
498 { do_bad, SIGBUS, 0, "unknown 54" },
499 { do_bad, SIGBUS, 0, "unknown 55" },
500 { do_bad, SIGBUS, 0, "unknown 56" },
501 { do_bad, SIGBUS, 0, "unknown 57" },
502 { do_bad, SIGBUS, 0, "unknown 58" },
503 { do_bad, SIGBUS, 0, "unknown 59" },
504 { do_bad, SIGBUS, 0, "unknown 60" },
505 { do_bad, SIGBUS, 0, "section domain fault" },
506 { do_bad, SIGBUS, 0, "page domain fault" },
507 { do_bad, SIGBUS, 0, "unknown 63" },
510 static const char *fault_name(unsigned int esr)
512 const struct fault_info *inf = fault_info + (esr & 63);
517 * Dispatch a data abort to the relevant handler.
519 asmlinkage void __exception do_mem_abort(unsigned long addr, unsigned int esr,
520 struct pt_regs *regs)
522 const struct fault_info *inf = fault_info + (esr & 63);
525 if (!inf->fn(addr, esr, regs))
528 pr_alert("Unhandled fault: %s (0x%08x) at 0x%016lx\n",
529 inf->name, esr, addr);
531 info.si_signo = inf->sig;
533 info.si_code = inf->code;
534 info.si_addr = (void __user *)addr;
535 arm64_notify_die("", regs, &info, esr);
539 * Handle stack alignment exceptions.
541 asmlinkage void __exception do_sp_pc_abort(unsigned long addr,
543 struct pt_regs *regs)
546 struct task_struct *tsk = current;
548 if (show_unhandled_signals && unhandled_signal(tsk, SIGBUS))
549 pr_info_ratelimited("%s[%d]: %s exception: pc=%p sp=%p\n",
550 tsk->comm, task_pid_nr(tsk),
551 esr_get_class_string(esr), (void *)regs->pc,
554 info.si_signo = SIGBUS;
556 info.si_code = BUS_ADRALN;
557 info.si_addr = (void __user *)addr;
558 arm64_notify_die("Oops - SP/PC alignment exception", regs, &info, esr);
561 int __init early_brk64(unsigned long addr, unsigned int esr,
562 struct pt_regs *regs);
565 * __refdata because early_brk64 is __init, but the reference to it is
566 * clobbered at arch_initcall time.
567 * See traps.c and debug-monitors.c:debug_traps_init().
569 static struct fault_info __refdata debug_fault_info[] = {
570 { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware breakpoint" },
571 { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware single-step" },
572 { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware watchpoint" },
573 { do_bad, SIGBUS, 0, "unknown 3" },
574 { do_bad, SIGTRAP, TRAP_BRKPT, "aarch32 BKPT" },
575 { do_bad, SIGTRAP, 0, "aarch32 vector catch" },
576 { early_brk64, SIGTRAP, TRAP_BRKPT, "aarch64 BRK" },
577 { do_bad, SIGBUS, 0, "unknown 7" },
580 void __init hook_debug_fault_code(int nr,
581 int (*fn)(unsigned long, unsigned int, struct pt_regs *),
582 int sig, int code, const char *name)
584 BUG_ON(nr < 0 || nr >= ARRAY_SIZE(debug_fault_info));
586 debug_fault_info[nr].fn = fn;
587 debug_fault_info[nr].sig = sig;
588 debug_fault_info[nr].code = code;
589 debug_fault_info[nr].name = name;
592 asmlinkage int __exception do_debug_exception(unsigned long addr_if_watchpoint,
594 struct pt_regs *regs)
596 const struct fault_info *inf = debug_fault_info + DBG_ESR_EVT(esr);
597 unsigned long pc = instruction_pointer(regs);
602 * Tell lockdep we disabled irqs in entry.S. Do nothing if they were
603 * already disabled to preserve the last enabled/disabled addresses.
605 if (interrupts_enabled(regs))
606 trace_hardirqs_off();
608 if (!inf->fn(addr_if_watchpoint, esr, regs)) {
611 pr_alert("Unhandled debug exception: %s (0x%08x) at 0x%016lx\n",
614 info.si_signo = inf->sig;
616 info.si_code = inf->code;
617 info.si_addr = (void __user *)pc;
618 arm64_notify_die("", regs, &info, 0);
622 if (interrupts_enabled(regs))
628 #ifdef CONFIG_ARM64_PAN
629 int cpu_enable_pan(void *__unused)
632 * We modify PSTATE. This won't work from irq context as the PSTATE
633 * is discarded once we return from the exception.
635 WARN_ON_ONCE(in_interrupt());
637 config_sctlr_el1(SCTLR_EL1_SPAN, 0);
638 asm(SET_PSTATE_PAN(1));
641 #endif /* CONFIG_ARM64_PAN */