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>
33 #include <asm/cpufeature.h>
34 #include <asm/exception.h>
35 #include <asm/debug-monitors.h>
37 #include <asm/sysreg.h>
38 #include <asm/system_misc.h>
39 #include <asm/pgtable.h>
40 #include <asm/tlbflush.h>
42 static const char *fault_name(unsigned int esr);
45 * Dump out the page tables associated with 'addr' in mm 'mm'.
47 void show_pte(struct mm_struct *mm, unsigned long addr)
54 pr_alert("pgd = %p\n", mm->pgd);
55 pgd = pgd_offset(mm, addr);
56 pr_alert("[%08lx] *pgd=%016llx", addr, pgd_val(*pgd));
63 if (pgd_none(*pgd) || pgd_bad(*pgd))
66 pud = pud_offset(pgd, addr);
67 printk(", *pud=%016llx", pud_val(*pud));
68 if (pud_none(*pud) || pud_bad(*pud))
71 pmd = pmd_offset(pud, addr);
72 printk(", *pmd=%016llx", pmd_val(*pmd));
73 if (pmd_none(*pmd) || pmd_bad(*pmd))
76 pte = pte_offset_map(pmd, addr);
77 printk(", *pte=%016llx", pte_val(*pte));
85 * The kernel tried to access some page that wasn't present.
87 static void __do_kernel_fault(struct mm_struct *mm, unsigned long addr,
88 unsigned int esr, struct pt_regs *regs)
91 * Are we prepared to handle this kernel fault?
93 if (fixup_exception(regs))
97 * No handler, we'll have to terminate things with extreme prejudice.
100 pr_alert("Unable to handle kernel %s at virtual address %08lx\n",
101 (addr < PAGE_SIZE) ? "NULL pointer dereference" :
102 "paging request", addr);
105 die("Oops", regs, esr);
111 * Something tried to access memory that isn't in our memory map. User mode
112 * accesses just cause a SIGSEGV
114 static void __do_user_fault(struct task_struct *tsk, unsigned long addr,
115 unsigned int esr, unsigned int sig, int code,
116 struct pt_regs *regs)
120 if (unhandled_signal(tsk, sig) && show_unhandled_signals_ratelimited()) {
121 pr_info("%s[%d]: unhandled %s (%d) at 0x%08lx, esr 0x%03x\n",
122 tsk->comm, task_pid_nr(tsk), fault_name(esr), sig,
124 show_pte(tsk->mm, addr);
128 tsk->thread.fault_address = addr;
129 tsk->thread.fault_code = esr;
133 si.si_addr = (void __user *)addr;
134 force_sig_info(sig, &si, tsk);
137 static void do_bad_area(unsigned long addr, unsigned int esr, struct pt_regs *regs)
139 struct task_struct *tsk = current;
140 struct mm_struct *mm = tsk->active_mm;
143 * If we are in kernel mode at this point, we have no context to
144 * handle this fault with.
147 __do_user_fault(tsk, addr, esr, SIGSEGV, SEGV_MAPERR, regs);
149 __do_kernel_fault(mm, addr, esr, regs);
152 #define VM_FAULT_BADMAP 0x010000
153 #define VM_FAULT_BADACCESS 0x020000
155 #define ESR_LNX_EXEC (1 << 24)
157 static int __do_page_fault(struct mm_struct *mm, unsigned long addr,
158 unsigned int mm_flags, unsigned long vm_flags,
159 struct task_struct *tsk)
161 struct vm_area_struct *vma;
164 vma = find_vma(mm, addr);
165 fault = VM_FAULT_BADMAP;
168 if (unlikely(vma->vm_start > addr))
172 * Ok, we have a good vm_area for this memory access, so we can handle
177 * Check that the permissions on the VMA allow for the fault which
178 * occurred. If we encountered a write or exec fault, we must have
179 * appropriate permissions, otherwise we allow any permission.
181 if (!(vma->vm_flags & vm_flags)) {
182 fault = VM_FAULT_BADACCESS;
186 return handle_mm_fault(mm, vma, addr & PAGE_MASK, mm_flags);
189 if (vma->vm_flags & VM_GROWSDOWN && !expand_stack(vma, addr))
195 static inline int permission_fault(unsigned int esr)
197 unsigned int ec = (esr & ESR_ELx_EC_MASK) >> ESR_ELx_EC_SHIFT;
198 unsigned int fsc_type = esr & ESR_ELx_FSC_TYPE;
200 return (ec == ESR_ELx_EC_DABT_CUR && fsc_type == ESR_ELx_FSC_PERM);
203 static int __kprobes do_page_fault(unsigned long addr, unsigned int esr,
204 struct pt_regs *regs)
206 struct task_struct *tsk;
207 struct mm_struct *mm;
208 int fault, sig, code;
209 unsigned long vm_flags = VM_READ | VM_WRITE | VM_EXEC;
210 unsigned int mm_flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
215 /* Enable interrupts if they were enabled in the parent context. */
216 if (interrupts_enabled(regs))
220 * If we're in an interrupt or have no user context, we must not take
223 if (faulthandler_disabled() || !mm)
227 mm_flags |= FAULT_FLAG_USER;
229 if (esr & ESR_LNX_EXEC) {
231 } else if ((esr & ESR_ELx_WNR) && !(esr & ESR_ELx_CM)) {
233 mm_flags |= FAULT_FLAG_WRITE;
236 if (permission_fault(esr) && (addr < USER_DS)) {
237 if (get_fs() == KERNEL_DS)
238 die("Accessing user space memory with fs=KERNEL_DS", regs, esr);
240 if (!search_exception_tables(regs->pc))
241 die("Accessing user space memory outside uaccess.h routines", regs, esr);
245 * As per x86, we may deadlock here. However, since the kernel only
246 * validly references user space from well defined areas of the code,
247 * we can bug out early if this is from code which shouldn't.
249 if (!down_read_trylock(&mm->mmap_sem)) {
250 if (!user_mode(regs) && !search_exception_tables(regs->pc))
253 down_read(&mm->mmap_sem);
256 * The above down_read_trylock() might have succeeded in which
257 * case, we'll have missed the might_sleep() from down_read().
260 #ifdef CONFIG_DEBUG_VM
261 if (!user_mode(regs) && !search_exception_tables(regs->pc))
266 fault = __do_page_fault(mm, addr, mm_flags, vm_flags, tsk);
269 * If we need to retry but a fatal signal is pending, handle the
270 * signal first. We do not need to release the mmap_sem because it
271 * would already be released in __lock_page_or_retry in mm/filemap.c.
273 if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current))
277 * Major/minor page fault accounting is only done on the initial
278 * attempt. If we go through a retry, it is extremely likely that the
279 * page will be found in page cache at that point.
282 perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr);
283 if (mm_flags & FAULT_FLAG_ALLOW_RETRY) {
284 if (fault & VM_FAULT_MAJOR) {
286 perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs,
290 perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs,
293 if (fault & VM_FAULT_RETRY) {
295 * Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk of
298 mm_flags &= ~FAULT_FLAG_ALLOW_RETRY;
299 mm_flags |= FAULT_FLAG_TRIED;
304 up_read(&mm->mmap_sem);
307 * Handle the "normal" case first - VM_FAULT_MAJOR
309 if (likely(!(fault & (VM_FAULT_ERROR | VM_FAULT_BADMAP |
310 VM_FAULT_BADACCESS))))
314 * If we are in kernel mode at this point, we have no context to
315 * handle this fault with.
317 if (!user_mode(regs))
320 if (fault & VM_FAULT_OOM) {
322 * We ran out of memory, call the OOM killer, and return to
323 * userspace (which will retry the fault, or kill us if we got
326 pagefault_out_of_memory();
330 if (fault & VM_FAULT_SIGBUS) {
332 * We had some memory, but were unable to successfully fix up
339 * Something tried to access memory that isn't in our memory
343 code = fault == VM_FAULT_BADACCESS ?
344 SEGV_ACCERR : SEGV_MAPERR;
347 __do_user_fault(tsk, addr, esr, sig, code, regs);
351 __do_kernel_fault(mm, addr, esr, regs);
356 * First Level Translation Fault Handler
358 * We enter here because the first level page table doesn't contain a valid
359 * entry for the address.
361 * If the address is in kernel space (>= TASK_SIZE), then we are probably
362 * faulting in the vmalloc() area.
364 * If the init_task's first level page tables contains the relevant entry, we
365 * copy the it to this task. If not, we send the process a signal, fixup the
366 * exception, or oops the kernel.
368 * NOTE! We MUST NOT take any locks for this case. We may be in an interrupt
369 * or a critical region, and should only copy the information from the master
370 * page table, nothing more.
372 static int __kprobes do_translation_fault(unsigned long addr,
374 struct pt_regs *regs)
376 if (addr < TASK_SIZE)
377 return do_page_fault(addr, esr, regs);
379 do_bad_area(addr, esr, regs);
383 static int do_alignment_fault(unsigned long addr, unsigned int esr,
384 struct pt_regs *regs)
386 do_bad_area(addr, esr, regs);
391 * This abort handler always returns "fault".
393 static int do_bad(unsigned long addr, unsigned int esr, struct pt_regs *regs)
398 static struct fault_info {
399 int (*fn)(unsigned long addr, unsigned int esr, struct pt_regs *regs);
404 { do_bad, SIGBUS, 0, "ttbr address size fault" },
405 { do_bad, SIGBUS, 0, "level 1 address size fault" },
406 { do_bad, SIGBUS, 0, "level 2 address size fault" },
407 { do_bad, SIGBUS, 0, "level 3 address size fault" },
408 { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 0 translation fault" },
409 { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 1 translation fault" },
410 { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 2 translation fault" },
411 { do_page_fault, SIGSEGV, SEGV_MAPERR, "level 3 translation fault" },
412 { do_bad, SIGBUS, 0, "unknown 8" },
413 { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 1 access flag fault" },
414 { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 2 access flag fault" },
415 { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 3 access flag fault" },
416 { do_bad, SIGBUS, 0, "unknown 12" },
417 { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 1 permission fault" },
418 { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 2 permission fault" },
419 { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 3 permission fault" },
420 { do_bad, SIGBUS, 0, "synchronous external abort" },
421 { do_bad, SIGBUS, 0, "unknown 17" },
422 { do_bad, SIGBUS, 0, "unknown 18" },
423 { do_bad, SIGBUS, 0, "unknown 19" },
424 { do_bad, SIGBUS, 0, "synchronous abort (translation table walk)" },
425 { do_bad, SIGBUS, 0, "synchronous abort (translation table walk)" },
426 { do_bad, SIGBUS, 0, "synchronous abort (translation table walk)" },
427 { do_bad, SIGBUS, 0, "synchronous abort (translation table walk)" },
428 { do_bad, SIGBUS, 0, "synchronous parity error" },
429 { do_bad, SIGBUS, 0, "unknown 25" },
430 { do_bad, SIGBUS, 0, "unknown 26" },
431 { do_bad, SIGBUS, 0, "unknown 27" },
432 { do_bad, SIGBUS, 0, "synchronous parity error (translation table walk)" },
433 { do_bad, SIGBUS, 0, "synchronous parity error (translation table walk)" },
434 { do_bad, SIGBUS, 0, "synchronous parity error (translation table walk)" },
435 { do_bad, SIGBUS, 0, "synchronous parity error (translation table walk)" },
436 { do_bad, SIGBUS, 0, "unknown 32" },
437 { do_alignment_fault, SIGBUS, BUS_ADRALN, "alignment fault" },
438 { do_bad, SIGBUS, 0, "unknown 34" },
439 { do_bad, SIGBUS, 0, "unknown 35" },
440 { do_bad, SIGBUS, 0, "unknown 36" },
441 { do_bad, SIGBUS, 0, "unknown 37" },
442 { do_bad, SIGBUS, 0, "unknown 38" },
443 { do_bad, SIGBUS, 0, "unknown 39" },
444 { do_bad, SIGBUS, 0, "unknown 40" },
445 { do_bad, SIGBUS, 0, "unknown 41" },
446 { do_bad, SIGBUS, 0, "unknown 42" },
447 { do_bad, SIGBUS, 0, "unknown 43" },
448 { do_bad, SIGBUS, 0, "unknown 44" },
449 { do_bad, SIGBUS, 0, "unknown 45" },
450 { do_bad, SIGBUS, 0, "unknown 46" },
451 { do_bad, SIGBUS, 0, "unknown 47" },
452 { do_bad, SIGBUS, 0, "TLB conflict abort" },
453 { do_bad, SIGBUS, 0, "unknown 49" },
454 { do_bad, SIGBUS, 0, "unknown 50" },
455 { do_bad, SIGBUS, 0, "unknown 51" },
456 { do_bad, SIGBUS, 0, "implementation fault (lockdown abort)" },
457 { do_bad, SIGBUS, 0, "implementation fault (unsupported exclusive)" },
458 { do_bad, SIGBUS, 0, "unknown 54" },
459 { do_bad, SIGBUS, 0, "unknown 55" },
460 { do_bad, SIGBUS, 0, "unknown 56" },
461 { do_bad, SIGBUS, 0, "unknown 57" },
462 { do_bad, SIGBUS, 0, "unknown 58" },
463 { do_bad, SIGBUS, 0, "unknown 59" },
464 { do_bad, SIGBUS, 0, "unknown 60" },
465 { do_bad, SIGBUS, 0, "section domain fault" },
466 { do_bad, SIGBUS, 0, "page domain fault" },
467 { do_bad, SIGBUS, 0, "unknown 63" },
470 static const char *fault_name(unsigned int esr)
472 const struct fault_info *inf = fault_info + (esr & 63);
477 * Dispatch a data abort to the relevant handler.
479 asmlinkage void __exception do_mem_abort(unsigned long addr, unsigned int esr,
480 struct pt_regs *regs)
482 const struct fault_info *inf = fault_info + (esr & 63);
485 if (!inf->fn(addr, esr, regs))
488 pr_alert("Unhandled fault: %s (0x%08x) at 0x%016lx\n",
489 inf->name, esr, addr);
491 info.si_signo = inf->sig;
493 info.si_code = inf->code;
494 info.si_addr = (void __user *)addr;
495 arm64_notify_die("", regs, &info, esr);
499 * Handle stack alignment exceptions.
501 asmlinkage void __exception do_sp_pc_abort(unsigned long addr,
503 struct pt_regs *regs)
506 struct task_struct *tsk = current;
508 if (show_unhandled_signals && unhandled_signal(tsk, SIGBUS))
509 pr_info_ratelimited("%s[%d]: %s exception: pc=%p sp=%p\n",
510 tsk->comm, task_pid_nr(tsk),
511 esr_get_class_string(esr), (void *)regs->pc,
514 info.si_signo = SIGBUS;
516 info.si_code = BUS_ADRALN;
517 info.si_addr = (void __user *)addr;
518 arm64_notify_die("Oops - SP/PC alignment exception", regs, &info, esr);
521 int __init early_brk64(unsigned long addr, unsigned int esr,
522 struct pt_regs *regs);
525 * __refdata because early_brk64 is __init, but the reference to it is
526 * clobbered at arch_initcall time.
527 * See traps.c and debug-monitors.c:debug_traps_init().
529 static struct fault_info __refdata debug_fault_info[] = {
530 { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware breakpoint" },
531 { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware single-step" },
532 { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware watchpoint" },
533 { do_bad, SIGBUS, 0, "unknown 3" },
534 { do_bad, SIGTRAP, TRAP_BRKPT, "aarch32 BKPT" },
535 { do_bad, SIGTRAP, 0, "aarch32 vector catch" },
536 { early_brk64, SIGTRAP, TRAP_BRKPT, "aarch64 BRK" },
537 { do_bad, SIGBUS, 0, "unknown 7" },
540 void __init hook_debug_fault_code(int nr,
541 int (*fn)(unsigned long, unsigned int, struct pt_regs *),
542 int sig, int code, const char *name)
544 BUG_ON(nr < 0 || nr >= ARRAY_SIZE(debug_fault_info));
546 debug_fault_info[nr].fn = fn;
547 debug_fault_info[nr].sig = sig;
548 debug_fault_info[nr].code = code;
549 debug_fault_info[nr].name = name;
552 asmlinkage int __exception do_debug_exception(unsigned long addr,
554 struct pt_regs *regs)
556 const struct fault_info *inf = debug_fault_info + DBG_ESR_EVT(esr);
559 if (!inf->fn(addr, esr, regs))
562 pr_alert("Unhandled debug exception: %s (0x%08x) at 0x%016lx\n",
563 inf->name, esr, addr);
565 info.si_signo = inf->sig;
567 info.si_code = inf->code;
568 info.si_addr = (void __user *)addr;
569 arm64_notify_die("", regs, &info, 0);
574 #ifdef CONFIG_ARM64_PAN
575 void cpu_enable_pan(void *__unused)
577 config_sctlr_el1(SCTLR_EL1_SPAN, 0);
579 #endif /* CONFIG_ARM64_PAN */
581 #ifdef CONFIG_ARM64_UAO
583 * Kernel threads have fs=KERNEL_DS by default, and don't need to call
584 * set_fs(), devtmpfs in particular relies on this behaviour.
585 * We need to enable the feature at runtime (instead of adding it to
586 * PSR_MODE_EL1h) as the feature may not be implemented by the cpu.
588 void cpu_enable_uao(void *__unused)
590 asm(SET_PSTATE_UAO(1));
592 #endif /* CONFIG_ARM64_UAO */