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/exception.h>
34 #include <asm/debug-monitors.h>
35 #include <asm/system_misc.h>
36 #include <asm/pgtable.h>
37 #include <asm/tlbflush.h>
39 static const char *fault_name(unsigned int esr);
42 * Dump out the page tables associated with 'addr' in mm 'mm'.
44 void show_pte(struct mm_struct *mm, unsigned long addr)
51 pr_alert("pgd = %p\n", mm->pgd);
52 pgd = pgd_offset(mm, addr);
53 pr_alert("[%08lx] *pgd=%016llx", addr, pgd_val(*pgd));
60 if (pgd_none(*pgd) || pgd_bad(*pgd))
63 pud = pud_offset(pgd, addr);
64 if (pud_none(*pud) || pud_bad(*pud))
67 pmd = pmd_offset(pud, addr);
68 printk(", *pmd=%016llx", pmd_val(*pmd));
69 if (pmd_none(*pmd) || pmd_bad(*pmd))
72 pte = pte_offset_map(pmd, addr);
73 printk(", *pte=%016llx", pte_val(*pte));
81 * The kernel tried to access some page that wasn't present.
83 static void __do_kernel_fault(struct mm_struct *mm, unsigned long addr,
84 unsigned int esr, struct pt_regs *regs)
87 * Are we prepared to handle this kernel fault?
89 if (fixup_exception(regs))
93 * No handler, we'll have to terminate things with extreme prejudice.
96 pr_alert("Unable to handle kernel %s at virtual address %08lx\n",
97 (addr < PAGE_SIZE) ? "NULL pointer dereference" :
98 "paging request", addr);
101 die("Oops", regs, esr);
107 * Something tried to access memory that isn't in our memory map. User mode
108 * accesses just cause a SIGSEGV
110 static void __do_user_fault(struct task_struct *tsk, unsigned long addr,
111 unsigned int esr, unsigned int sig, int code,
112 struct pt_regs *regs)
116 if (show_unhandled_signals && unhandled_signal(tsk, sig) &&
117 printk_ratelimit()) {
118 pr_info("%s[%d]: unhandled %s (%d) at 0x%08lx, esr 0x%03x\n",
119 tsk->comm, task_pid_nr(tsk), fault_name(esr), sig,
121 show_pte(tsk->mm, addr);
125 tsk->thread.fault_address = addr;
129 si.si_addr = (void __user *)addr;
130 force_sig_info(sig, &si, tsk);
133 void do_bad_area(unsigned long addr, unsigned int esr, struct pt_regs *regs)
135 struct task_struct *tsk = current;
136 struct mm_struct *mm = tsk->active_mm;
139 * If we are in kernel mode at this point, we have no context to
140 * handle this fault with.
143 __do_user_fault(tsk, addr, esr, SIGSEGV, SEGV_MAPERR, regs);
145 __do_kernel_fault(mm, addr, esr, regs);
148 #define VM_FAULT_BADMAP 0x010000
149 #define VM_FAULT_BADACCESS 0x020000
151 #define ESR_WRITE (1 << 6)
152 #define ESR_CM (1 << 8)
153 #define ESR_LNX_EXEC (1 << 24)
155 static int __do_page_fault(struct mm_struct *mm, unsigned long addr,
156 unsigned int mm_flags, unsigned long vm_flags,
157 struct task_struct *tsk)
159 struct vm_area_struct *vma;
162 vma = find_vma(mm, addr);
163 fault = VM_FAULT_BADMAP;
166 if (unlikely(vma->vm_start > addr))
170 * Ok, we have a good vm_area for this memory access, so we can handle
175 * Check that the permissions on the VMA allow for the fault which
176 * occurred. If we encountered a write or exec fault, we must have
177 * appropriate permissions, otherwise we allow any permission.
179 if (!(vma->vm_flags & vm_flags)) {
180 fault = VM_FAULT_BADACCESS;
184 return handle_mm_fault(mm, vma, addr & PAGE_MASK, mm_flags);
187 if (vma->vm_flags & VM_GROWSDOWN && !expand_stack(vma, addr))
193 static int __kprobes do_page_fault(unsigned long addr, unsigned int esr,
194 struct pt_regs *regs)
196 struct task_struct *tsk;
197 struct mm_struct *mm;
198 int fault, sig, code;
199 unsigned long vm_flags = VM_READ | VM_WRITE | VM_EXEC;
200 unsigned int mm_flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
202 if (esr & ESR_LNX_EXEC) {
204 } else if ((esr & ESR_WRITE) && !(esr & ESR_CM)) {
206 mm_flags |= FAULT_FLAG_WRITE;
212 /* Enable interrupts if they were enabled in the parent context. */
213 if (interrupts_enabled(regs))
217 * If we're in an interrupt or have no user context, we must not take
220 if (in_atomic() || !mm)
224 * As per x86, we may deadlock here. However, since the kernel only
225 * validly references user space from well defined areas of the code,
226 * we can bug out early if this is from code which shouldn't.
228 if (!down_read_trylock(&mm->mmap_sem)) {
229 if (!user_mode(regs) && !search_exception_tables(regs->pc))
232 down_read(&mm->mmap_sem);
235 * The above down_read_trylock() might have succeeded in which
236 * case, we'll have missed the might_sleep() from down_read().
239 #ifdef CONFIG_DEBUG_VM
240 if (!user_mode(regs) && !search_exception_tables(regs->pc))
245 fault = __do_page_fault(mm, addr, mm_flags, vm_flags, tsk);
248 * If we need to retry but a fatal signal is pending, handle the
249 * signal first. We do not need to release the mmap_sem because it
250 * would already be released in __lock_page_or_retry in mm/filemap.c.
252 if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current))
256 * Major/minor page fault accounting is only done on the initial
257 * attempt. If we go through a retry, it is extremely likely that the
258 * page will be found in page cache at that point.
261 perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr);
262 if (mm_flags & FAULT_FLAG_ALLOW_RETRY) {
263 if (fault & VM_FAULT_MAJOR) {
265 perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs,
269 perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs,
272 if (fault & VM_FAULT_RETRY) {
274 * Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk of
277 mm_flags &= ~FAULT_FLAG_ALLOW_RETRY;
282 up_read(&mm->mmap_sem);
285 * Handle the "normal" case first - VM_FAULT_MAJOR / VM_FAULT_MINOR
287 if (likely(!(fault & (VM_FAULT_ERROR | VM_FAULT_BADMAP |
288 VM_FAULT_BADACCESS))))
291 if (fault & VM_FAULT_OOM) {
293 * We ran out of memory, call the OOM killer, and return to
294 * userspace (which will retry the fault, or kill us if we got
297 pagefault_out_of_memory();
302 * If we are in kernel mode at this point, we have no context to
303 * handle this fault with.
305 if (!user_mode(regs))
308 if (fault & VM_FAULT_SIGBUS) {
310 * We had some memory, but were unable to successfully fix up
317 * Something tried to access memory that isn't in our memory
321 code = fault == VM_FAULT_BADACCESS ?
322 SEGV_ACCERR : SEGV_MAPERR;
325 __do_user_fault(tsk, addr, esr, sig, code, regs);
329 __do_kernel_fault(mm, addr, esr, regs);
334 * First Level Translation Fault Handler
336 * We enter here because the first level page table doesn't contain a valid
337 * entry for the address.
339 * If the address is in kernel space (>= TASK_SIZE), then we are probably
340 * faulting in the vmalloc() area.
342 * If the init_task's first level page tables contains the relevant entry, we
343 * copy the it to this task. If not, we send the process a signal, fixup the
344 * exception, or oops the kernel.
346 * NOTE! We MUST NOT take any locks for this case. We may be in an interrupt
347 * or a critical region, and should only copy the information from the master
348 * page table, nothing more.
350 static int __kprobes do_translation_fault(unsigned long addr,
352 struct pt_regs *regs)
354 if (addr < TASK_SIZE)
355 return do_page_fault(addr, esr, regs);
357 do_bad_area(addr, esr, regs);
362 * Some section permission faults need to be handled gracefully. They can
363 * happen due to a __{get,put}_user during an oops.
365 static int do_sect_fault(unsigned long addr, unsigned int esr,
366 struct pt_regs *regs)
368 do_bad_area(addr, esr, regs);
373 * This abort handler always returns "fault".
375 static int do_bad(unsigned long addr, unsigned int esr, struct pt_regs *regs)
380 static struct fault_info {
381 int (*fn)(unsigned long addr, unsigned int esr, struct pt_regs *regs);
386 { do_bad, SIGBUS, 0, "ttbr address size fault" },
387 { do_bad, SIGBUS, 0, "level 1 address size fault" },
388 { do_bad, SIGBUS, 0, "level 2 address size fault" },
389 { do_bad, SIGBUS, 0, "level 3 address size fault" },
390 { do_translation_fault, SIGSEGV, SEGV_MAPERR, "input address range fault" },
391 { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 1 translation fault" },
392 { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 2 translation fault" },
393 { do_page_fault, SIGSEGV, SEGV_MAPERR, "level 3 translation fault" },
394 { do_bad, SIGBUS, 0, "reserved access flag fault" },
395 { do_bad, SIGSEGV, SEGV_ACCERR, "level 1 access flag fault" },
396 { do_bad, SIGSEGV, SEGV_ACCERR, "level 2 access flag fault" },
397 { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 3 access flag fault" },
398 { do_bad, SIGBUS, 0, "reserved permission fault" },
399 { do_bad, SIGSEGV, SEGV_ACCERR, "level 1 permission fault" },
400 { do_sect_fault, SIGSEGV, SEGV_ACCERR, "level 2 permission fault" },
401 { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 3 permission fault" },
402 { do_bad, SIGBUS, 0, "synchronous external abort" },
403 { do_bad, SIGBUS, 0, "asynchronous external abort" },
404 { do_bad, SIGBUS, 0, "unknown 18" },
405 { do_bad, SIGBUS, 0, "unknown 19" },
406 { do_bad, SIGBUS, 0, "synchronous abort (translation table walk)" },
407 { do_bad, SIGBUS, 0, "synchronous abort (translation table walk)" },
408 { do_bad, SIGBUS, 0, "synchronous abort (translation table walk)" },
409 { do_bad, SIGBUS, 0, "synchronous abort (translation table walk)" },
410 { do_bad, SIGBUS, 0, "synchronous parity error" },
411 { do_bad, SIGBUS, 0, "asynchronous parity error" },
412 { do_bad, SIGBUS, 0, "unknown 26" },
413 { do_bad, SIGBUS, 0, "unknown 27" },
414 { do_bad, SIGBUS, 0, "synchronous parity error (translation table walk" },
415 { do_bad, SIGBUS, 0, "synchronous parity error (translation table walk" },
416 { do_bad, SIGBUS, 0, "synchronous parity error (translation table walk" },
417 { do_bad, SIGBUS, 0, "synchronous parity error (translation table walk" },
418 { do_bad, SIGBUS, 0, "unknown 32" },
419 { do_bad, SIGBUS, BUS_ADRALN, "alignment fault" },
420 { do_bad, SIGBUS, 0, "debug event" },
421 { do_bad, SIGBUS, 0, "unknown 35" },
422 { do_bad, SIGBUS, 0, "unknown 36" },
423 { do_bad, SIGBUS, 0, "unknown 37" },
424 { do_bad, SIGBUS, 0, "unknown 38" },
425 { do_bad, SIGBUS, 0, "unknown 39" },
426 { do_bad, SIGBUS, 0, "unknown 40" },
427 { do_bad, SIGBUS, 0, "unknown 41" },
428 { do_bad, SIGBUS, 0, "unknown 42" },
429 { do_bad, SIGBUS, 0, "unknown 43" },
430 { do_bad, SIGBUS, 0, "unknown 44" },
431 { do_bad, SIGBUS, 0, "unknown 45" },
432 { do_bad, SIGBUS, 0, "unknown 46" },
433 { do_bad, SIGBUS, 0, "unknown 47" },
434 { do_bad, SIGBUS, 0, "unknown 48" },
435 { do_bad, SIGBUS, 0, "unknown 49" },
436 { do_bad, SIGBUS, 0, "unknown 50" },
437 { do_bad, SIGBUS, 0, "unknown 51" },
438 { do_bad, SIGBUS, 0, "implementation fault (lockdown abort)" },
439 { do_bad, SIGBUS, 0, "unknown 53" },
440 { do_bad, SIGBUS, 0, "unknown 54" },
441 { do_bad, SIGBUS, 0, "unknown 55" },
442 { do_bad, SIGBUS, 0, "unknown 56" },
443 { do_bad, SIGBUS, 0, "unknown 57" },
444 { do_bad, SIGBUS, 0, "implementation fault (coprocessor abort)" },
445 { do_bad, SIGBUS, 0, "unknown 59" },
446 { do_bad, SIGBUS, 0, "unknown 60" },
447 { do_bad, SIGBUS, 0, "unknown 61" },
448 { do_bad, SIGBUS, 0, "unknown 62" },
449 { do_bad, SIGBUS, 0, "unknown 63" },
452 static const char *fault_name(unsigned int esr)
454 const struct fault_info *inf = fault_info + (esr & 63);
459 * Dispatch a data abort to the relevant handler.
461 asmlinkage void __exception do_mem_abort(unsigned long addr, unsigned int esr,
462 struct pt_regs *regs)
464 const struct fault_info *inf = fault_info + (esr & 63);
467 if (!inf->fn(addr, esr, regs))
470 pr_alert("Unhandled fault: %s (0x%08x) at 0x%016lx\n",
471 inf->name, esr, addr);
473 info.si_signo = inf->sig;
475 info.si_code = inf->code;
476 info.si_addr = (void __user *)addr;
477 arm64_notify_die("", regs, &info, esr);
481 * Handle stack alignment exceptions.
483 asmlinkage void __exception do_sp_pc_abort(unsigned long addr,
485 struct pt_regs *regs)
489 info.si_signo = SIGBUS;
491 info.si_code = BUS_ADRALN;
492 info.si_addr = (void __user *)addr;
493 arm64_notify_die("", regs, &info, esr);
496 static struct fault_info debug_fault_info[] = {
497 { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware breakpoint" },
498 { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware single-step" },
499 { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware watchpoint" },
500 { do_bad, SIGBUS, 0, "unknown 3" },
501 { do_bad, SIGTRAP, TRAP_BRKPT, "aarch32 BKPT" },
502 { do_bad, SIGTRAP, 0, "aarch32 vector catch" },
503 { do_bad, SIGTRAP, TRAP_BRKPT, "aarch64 BRK" },
504 { do_bad, SIGBUS, 0, "unknown 7" },
507 void __init hook_debug_fault_code(int nr,
508 int (*fn)(unsigned long, unsigned int, struct pt_regs *),
509 int sig, int code, const char *name)
511 BUG_ON(nr < 0 || nr >= ARRAY_SIZE(debug_fault_info));
513 debug_fault_info[nr].fn = fn;
514 debug_fault_info[nr].sig = sig;
515 debug_fault_info[nr].code = code;
516 debug_fault_info[nr].name = name;
519 asmlinkage int __exception do_debug_exception(unsigned long addr,
521 struct pt_regs *regs)
523 const struct fault_info *inf = debug_fault_info + DBG_ESR_EVT(esr);
526 if (!inf->fn(addr, esr, regs))
529 pr_alert("Unhandled debug exception: %s (0x%08x) at 0x%016lx\n",
530 inf->name, esr, addr);
532 info.si_signo = inf->sig;
534 info.si_code = inf->code;
535 info.si_addr = (void __user *)addr;
536 arm64_notify_die("", regs, &info, esr);