Linux Kernel  3.7.1
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
fault.c
Go to the documentation of this file.
1 /*
2  * Based on arch/arm/mm/fault.c
3  *
4  * Copyright (C) 1995 Linus Torvalds
5  * Copyright (C) 1995-2004 Russell King
6  * Copyright (C) 2012 ARM Ltd.
7  *
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.
11  *
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.
16  *
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/>.
19  */
20 
21 #include <linux/module.h>
22 #include <linux/signal.h>
23 #include <linux/mm.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 
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>
38 
39 /*
40  * Dump out the page tables associated with 'addr' in mm 'mm'.
41  */
42 void show_pte(struct mm_struct *mm, unsigned long addr)
43 {
44  pgd_t *pgd;
45 
46  if (!mm)
47  mm = &init_mm;
48 
49  pr_alert("pgd = %p\n", mm->pgd);
50  pgd = pgd_offset(mm, addr);
51  pr_alert("[%08lx] *pgd=%016llx", addr, pgd_val(*pgd));
52 
53  do {
54  pud_t *pud;
55  pmd_t *pmd;
56  pte_t *pte;
57 
58  if (pgd_none_or_clear_bad(pgd))
59  break;
60 
61  pud = pud_offset(pgd, addr);
62  if (pud_none_or_clear_bad(pud))
63  break;
64 
65  pmd = pmd_offset(pud, addr);
66  printk(", *pmd=%016llx", pmd_val(*pmd));
67  if (pmd_none_or_clear_bad(pmd))
68  break;
69 
70  pte = pte_offset_map(pmd, addr);
71  printk(", *pte=%016llx", pte_val(*pte));
72  pte_unmap(pte);
73  } while(0);
74 
75  printk("\n");
76 }
77 
78 /*
79  * The kernel tried to access some page that wasn't present.
80  */
81 static void __do_kernel_fault(struct mm_struct *mm, unsigned long addr,
82  unsigned int esr, struct pt_regs *regs)
83 {
84  /*
85  * Are we prepared to handle this kernel fault?
86  */
87  if (fixup_exception(regs))
88  return;
89 
90  /*
91  * No handler, we'll have to terminate things with extreme prejudice.
92  */
93  bust_spinlocks(1);
94  pr_alert("Unable to handle kernel %s at virtual address %08lx\n",
95  (addr < PAGE_SIZE) ? "NULL pointer dereference" :
96  "paging request", addr);
97 
98  show_pte(mm, addr);
99  die("Oops", regs, esr);
100  bust_spinlocks(0);
101  do_exit(SIGKILL);
102 }
103 
104 /*
105  * Something tried to access memory that isn't in our memory map. User mode
106  * accesses just cause a SIGSEGV
107  */
108 static void __do_user_fault(struct task_struct *tsk, unsigned long addr,
109  unsigned int esr, unsigned int sig, int code,
110  struct pt_regs *regs)
111 {
112  struct siginfo si;
113 
115  pr_info("%s[%d]: unhandled page fault (%d) at 0x%08lx, code 0x%03x\n",
116  tsk->comm, task_pid_nr(tsk), sig, addr, esr);
117  show_pte(tsk->mm, addr);
118  show_regs(regs);
119  }
120 
121  tsk->thread.fault_address = addr;
122  si.si_signo = sig;
123  si.si_errno = 0;
124  si.si_code = code;
125  si.si_addr = (void __user *)addr;
126  force_sig_info(sig, &si, tsk);
127 }
128 
129 void do_bad_area(unsigned long addr, unsigned int esr, struct pt_regs *regs)
130 {
131  struct task_struct *tsk = current;
132  struct mm_struct *mm = tsk->active_mm;
133 
134  /*
135  * If we are in kernel mode at this point, we have no context to
136  * handle this fault with.
137  */
138  if (user_mode(regs))
139  __do_user_fault(tsk, addr, esr, SIGSEGV, SEGV_MAPERR, regs);
140  else
141  __do_kernel_fault(mm, addr, esr, regs);
142 }
143 
144 #define VM_FAULT_BADMAP 0x010000
145 #define VM_FAULT_BADACCESS 0x020000
146 
147 #define ESR_WRITE (1 << 6)
148 #define ESR_LNX_EXEC (1 << 24)
149 
150 /*
151  * Check that the permissions on the VMA allow for the fault which occurred.
152  * If we encountered a write fault, we must have write permission, otherwise
153  * we allow any permission.
154  */
155 static inline bool access_error(unsigned int esr, struct vm_area_struct *vma)
156 {
157  unsigned int mask = VM_READ | VM_WRITE | VM_EXEC;
158 
159  if (esr & ESR_WRITE)
160  mask = VM_WRITE;
161  if (esr & ESR_LNX_EXEC)
162  mask = VM_EXEC;
163 
164  return vma->vm_flags & mask ? false : true;
165 }
166 
167 static int __do_page_fault(struct mm_struct *mm, unsigned long addr,
168  unsigned int esr, unsigned int flags,
169  struct task_struct *tsk)
170 {
171  struct vm_area_struct *vma;
172  int fault;
173 
174  vma = find_vma(mm, addr);
175  fault = VM_FAULT_BADMAP;
176  if (unlikely(!vma))
177  goto out;
178  if (unlikely(vma->vm_start > addr))
179  goto check_stack;
180 
181  /*
182  * Ok, we have a good vm_area for this memory access, so we can handle
183  * it.
184  */
185 good_area:
186  if (access_error(esr, vma)) {
187  fault = VM_FAULT_BADACCESS;
188  goto out;
189  }
190 
191  return handle_mm_fault(mm, vma, addr & PAGE_MASK, flags);
192 
193 check_stack:
194  if (vma->vm_flags & VM_GROWSDOWN && !expand_stack(vma, addr))
195  goto good_area;
196 out:
197  return fault;
198 }
199 
200 static int __kprobes do_page_fault(unsigned long addr, unsigned int esr,
201  struct pt_regs *regs)
202 {
203  struct task_struct *tsk;
204  struct mm_struct *mm;
205  int fault, sig, code;
206  int write = esr & ESR_WRITE;
207  unsigned int flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE |
208  (write ? FAULT_FLAG_WRITE : 0);
209 
210  tsk = current;
211  mm = tsk->mm;
212 
213  /* Enable interrupts if they were enabled in the parent context. */
214  if (interrupts_enabled(regs))
216 
217  /*
218  * If we're in an interrupt or have no user context, we must not take
219  * the fault.
220  */
221  if (in_atomic() || !mm)
222  goto no_context;
223 
224  /*
225  * As per x86, we may deadlock here. However, since the kernel only
226  * validly references user space from well defined areas of the code,
227  * we can bug out early if this is from code which shouldn't.
228  */
229  if (!down_read_trylock(&mm->mmap_sem)) {
230  if (!user_mode(regs) && !search_exception_tables(regs->pc))
231  goto no_context;
232 retry:
233  down_read(&mm->mmap_sem);
234  } else {
235  /*
236  * The above down_read_trylock() might have succeeded in which
237  * case, we'll have missed the might_sleep() from down_read().
238  */
239  might_sleep();
240 #ifdef CONFIG_DEBUG_VM
241  if (!user_mode(regs) && !search_exception_tables(regs->pc))
242  goto no_context;
243 #endif
244  }
245 
246  fault = __do_page_fault(mm, addr, esr, flags, tsk);
247 
248  /*
249  * If we need to retry but a fatal signal is pending, handle the
250  * signal first. We do not need to release the mmap_sem because it
251  * would already be released in __lock_page_or_retry in mm/filemap.c.
252  */
253  if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current))
254  return 0;
255 
256  /*
257  * Major/minor page fault accounting is only done on the initial
258  * attempt. If we go through a retry, it is extremely likely that the
259  * page will be found in page cache at that point.
260  */
261 
262  perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr);
263  if (flags & FAULT_FLAG_ALLOW_RETRY) {
264  if (fault & VM_FAULT_MAJOR) {
265  tsk->maj_flt++;
266  perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs,
267  addr);
268  } else {
269  tsk->min_flt++;
270  perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs,
271  addr);
272  }
273  if (fault & VM_FAULT_RETRY) {
274  /*
275  * Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk of
276  * starvation.
277  */
278  flags &= ~FAULT_FLAG_ALLOW_RETRY;
279  goto retry;
280  }
281  }
282 
283  up_read(&mm->mmap_sem);
284 
285  /*
286  * Handle the "normal" case first - VM_FAULT_MAJOR / VM_FAULT_MINOR
287  */
288  if (likely(!(fault & (VM_FAULT_ERROR | VM_FAULT_BADMAP |
290  return 0;
291 
292  if (fault & VM_FAULT_OOM) {
293  /*
294  * We ran out of memory, call the OOM killer, and return to
295  * userspace (which will retry the fault, or kill us if we got
296  * oom-killed).
297  */
299  return 0;
300  }
301 
302  /*
303  * If we are in kernel mode at this point, we have no context to
304  * handle this fault with.
305  */
306  if (!user_mode(regs))
307  goto no_context;
308 
309  if (fault & VM_FAULT_SIGBUS) {
310  /*
311  * We had some memory, but were unable to successfully fix up
312  * this page fault.
313  */
314  sig = SIGBUS;
315  code = BUS_ADRERR;
316  } else {
317  /*
318  * Something tried to access memory that isn't in our memory
319  * map.
320  */
321  sig = SIGSEGV;
322  code = fault == VM_FAULT_BADACCESS ?
324  }
325 
326  __do_user_fault(tsk, addr, esr, sig, code, regs);
327  return 0;
328 
329 no_context:
330  __do_kernel_fault(mm, addr, esr, regs);
331  return 0;
332 }
333 
334 /*
335  * First Level Translation Fault Handler
336  *
337  * We enter here because the first level page table doesn't contain a valid
338  * entry for the address.
339  *
340  * If the address is in kernel space (>= TASK_SIZE), then we are probably
341  * faulting in the vmalloc() area.
342  *
343  * If the init_task's first level page tables contains the relevant entry, we
344  * copy the it to this task. If not, we send the process a signal, fixup the
345  * exception, or oops the kernel.
346  *
347  * NOTE! We MUST NOT take any locks for this case. We may be in an interrupt
348  * or a critical region, and should only copy the information from the master
349  * page table, nothing more.
350  */
351 static int __kprobes do_translation_fault(unsigned long addr,
352  unsigned int esr,
353  struct pt_regs *regs)
354 {
355  if (addr < TASK_SIZE)
356  return do_page_fault(addr, esr, regs);
357 
358  do_bad_area(addr, esr, regs);
359  return 0;
360 }
361 
362 /*
363  * Some section permission faults need to be handled gracefully. They can
364  * happen due to a __{get,put}_user during an oops.
365  */
366 static int do_sect_fault(unsigned long addr, unsigned int esr,
367  struct pt_regs *regs)
368 {
369  do_bad_area(addr, esr, regs);
370  return 0;
371 }
372 
373 /*
374  * This abort handler always returns "fault".
375  */
376 static int do_bad(unsigned long addr, unsigned int esr, struct pt_regs *regs)
377 {
378  return 1;
379 }
380 
381 static struct fault_info {
382  int (*fn)(unsigned long addr, unsigned int esr, struct pt_regs *regs);
383  int sig;
384  int code;
385  const char *name;
386 } fault_info[] = {
387  { do_bad, SIGBUS, 0, "ttbr address size fault" },
388  { do_bad, SIGBUS, 0, "level 1 address size fault" },
389  { do_bad, SIGBUS, 0, "level 2 address size fault" },
390  { do_bad, SIGBUS, 0, "level 3 address size fault" },
391  { do_translation_fault, SIGSEGV, SEGV_MAPERR, "input address range fault" },
392  { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 1 translation fault" },
393  { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 2 translation fault" },
394  { do_page_fault, SIGSEGV, SEGV_MAPERR, "level 3 translation fault" },
395  { do_bad, SIGBUS, 0, "reserved access flag fault" },
396  { do_bad, SIGSEGV, SEGV_ACCERR, "level 1 access flag fault" },
397  { do_bad, SIGSEGV, SEGV_ACCERR, "level 2 access flag fault" },
398  { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 3 access flag fault" },
399  { do_bad, SIGBUS, 0, "reserved permission fault" },
400  { do_bad, SIGSEGV, SEGV_ACCERR, "level 1 permission fault" },
401  { do_sect_fault, SIGSEGV, SEGV_ACCERR, "level 2 permission fault" },
402  { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 3 permission fault" },
403  { do_bad, SIGBUS, 0, "synchronous external abort" },
404  { do_bad, SIGBUS, 0, "asynchronous external abort" },
405  { do_bad, SIGBUS, 0, "unknown 18" },
406  { do_bad, SIGBUS, 0, "unknown 19" },
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 abort (translation table walk)" },
411  { do_bad, SIGBUS, 0, "synchronous parity error" },
412  { do_bad, SIGBUS, 0, "asynchronous parity error" },
413  { do_bad, SIGBUS, 0, "unknown 26" },
414  { do_bad, SIGBUS, 0, "unknown 27" },
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, "synchronous parity error (translation table walk" },
419  { do_bad, SIGBUS, 0, "unknown 32" },
420  { do_bad, SIGBUS, BUS_ADRALN, "alignment fault" },
421  { do_bad, SIGBUS, 0, "debug event" },
422  { do_bad, SIGBUS, 0, "unknown 35" },
423  { do_bad, SIGBUS, 0, "unknown 36" },
424  { do_bad, SIGBUS, 0, "unknown 37" },
425  { do_bad, SIGBUS, 0, "unknown 38" },
426  { do_bad, SIGBUS, 0, "unknown 39" },
427  { do_bad, SIGBUS, 0, "unknown 40" },
428  { do_bad, SIGBUS, 0, "unknown 41" },
429  { do_bad, SIGBUS, 0, "unknown 42" },
430  { do_bad, SIGBUS, 0, "unknown 43" },
431  { do_bad, SIGBUS, 0, "unknown 44" },
432  { do_bad, SIGBUS, 0, "unknown 45" },
433  { do_bad, SIGBUS, 0, "unknown 46" },
434  { do_bad, SIGBUS, 0, "unknown 47" },
435  { do_bad, SIGBUS, 0, "unknown 48" },
436  { do_bad, SIGBUS, 0, "unknown 49" },
437  { do_bad, SIGBUS, 0, "unknown 50" },
438  { do_bad, SIGBUS, 0, "unknown 51" },
439  { do_bad, SIGBUS, 0, "implementation fault (lockdown abort)" },
440  { do_bad, SIGBUS, 0, "unknown 53" },
441  { do_bad, SIGBUS, 0, "unknown 54" },
442  { do_bad, SIGBUS, 0, "unknown 55" },
443  { do_bad, SIGBUS, 0, "unknown 56" },
444  { do_bad, SIGBUS, 0, "unknown 57" },
445  { do_bad, SIGBUS, 0, "implementation fault (coprocessor abort)" },
446  { do_bad, SIGBUS, 0, "unknown 59" },
447  { do_bad, SIGBUS, 0, "unknown 60" },
448  { do_bad, SIGBUS, 0, "unknown 61" },
449  { do_bad, SIGBUS, 0, "unknown 62" },
450  { do_bad, SIGBUS, 0, "unknown 63" },
451 };
452 
453 /*
454  * Dispatch a data abort to the relevant handler.
455  */
456 asmlinkage void __exception do_mem_abort(unsigned long addr, unsigned int esr,
457  struct pt_regs *regs)
458 {
459  const struct fault_info *inf = fault_info + (esr & 63);
460  struct siginfo info;
461 
462  if (!inf->fn(addr, esr, regs))
463  return;
464 
465  pr_alert("Unhandled fault: %s (0x%08x) at 0x%016lx\n",
466  inf->name, esr, addr);
467 
468  info.si_signo = inf->sig;
469  info.si_errno = 0;
470  info.si_code = inf->code;
471  info.si_addr = (void __user *)addr;
472  arm64_notify_die("", regs, &info, esr);
473 }
474 
475 /*
476  * Handle stack alignment exceptions.
477  */
478 asmlinkage void __exception do_sp_pc_abort(unsigned long addr,
479  unsigned int esr,
480  struct pt_regs *regs)
481 {
482  struct siginfo info;
483 
484  info.si_signo = SIGBUS;
485  info.si_errno = 0;
486  info.si_code = BUS_ADRALN;
487  info.si_addr = (void __user *)addr;
488  arm64_notify_die("", regs, &info, esr);
489 }
490 
491 static struct fault_info debug_fault_info[] = {
492  { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware breakpoint" },
493  { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware single-step" },
494  { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware watchpoint" },
495  { do_bad, SIGBUS, 0, "unknown 3" },
496  { do_bad, SIGTRAP, TRAP_BRKPT, "aarch32 BKPT" },
497  { do_bad, SIGTRAP, 0, "aarch32 vector catch" },
498  { do_bad, SIGTRAP, TRAP_BRKPT, "aarch64 BRK" },
499  { do_bad, SIGBUS, 0, "unknown 7" },
500 };
501 
503  int (*fn)(unsigned long, unsigned int, struct pt_regs *),
504  int sig, int code, const char *name)
505 {
506  BUG_ON(nr < 0 || nr >= ARRAY_SIZE(debug_fault_info));
507 
508  debug_fault_info[nr].fn = fn;
509  debug_fault_info[nr].sig = sig;
510  debug_fault_info[nr].code = code;
511  debug_fault_info[nr].name = name;
512 }
513 
515  unsigned int esr,
516  struct pt_regs *regs)
517 {
518  const struct fault_info *inf = debug_fault_info + DBG_ESR_EVT(esr);
519  struct siginfo info;
520 
521  if (!inf->fn(addr, esr, regs))
522  return 1;
523 
524  pr_alert("Unhandled debug exception: %s (0x%08x) at 0x%016lx\n",
525  inf->name, esr, addr);
526 
527  info.si_signo = inf->sig;
528  info.si_errno = 0;
529  info.si_code = inf->code;
530  info.si_addr = (void __user *)addr;
531  arm64_notify_die("", regs, &info, esr);
532 
533  return 0;
534 }