fault.c 8.3 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332
  1. /*
  2. * This file is subject to the terms and conditions of the GNU General Public
  3. * License. See the file "COPYING" in the main directory of this archive
  4. * for more details.
  5. *
  6. * Copyright (C) 1995 - 2000 by Ralf Baechle
  7. */
  8. #include <linux/context_tracking.h>
  9. #include <linux/signal.h>
  10. #include <linux/sched.h>
  11. #include <linux/interrupt.h>
  12. #include <linux/kernel.h>
  13. #include <linux/errno.h>
  14. #include <linux/string.h>
  15. #include <linux/types.h>
  16. #include <linux/ptrace.h>
  17. #include <linux/ratelimit.h>
  18. #include <linux/mman.h>
  19. #include <linux/mm.h>
  20. #include <linux/smp.h>
  21. #include <linux/kprobes.h>
  22. #include <linux/perf_event.h>
  23. #include <linux/uaccess.h>
  24. #include <asm/branch.h>
  25. #include <asm/mmu_context.h>
  26. #include <asm/ptrace.h>
  27. #include <asm/highmem.h> /* For VMALLOC_END */
  28. #include <linux/kdebug.h>
  29. int show_unhandled_signals = 1;
  30. /*
  31. * This routine handles page faults. It determines the address,
  32. * and the problem, and then passes it off to one of the appropriate
  33. * routines.
  34. */
  35. static void __kprobes __do_page_fault(struct pt_regs *regs, unsigned long write,
  36. unsigned long address)
  37. {
  38. struct vm_area_struct * vma = NULL;
  39. struct task_struct *tsk = current;
  40. struct mm_struct *mm = tsk->mm;
  41. const int field = sizeof(unsigned long) * 2;
  42. int si_code;
  43. vm_fault_t fault;
  44. unsigned int flags = FAULT_FLAG_DEFAULT;
  45. static DEFINE_RATELIMIT_STATE(ratelimit_state, 5 * HZ, 10);
  46. #if 0
  47. printk("Cpu%d[%s:%d:%0*lx:%ld:%0*lx]\n", raw_smp_processor_id(),
  48. current->comm, current->pid, field, address, write,
  49. field, regs->cp0_epc);
  50. #endif
  51. #ifdef CONFIG_KPROBES
  52. /*
  53. * This is to notify the fault handler of the kprobes.
  54. */
  55. if (notify_die(DIE_PAGE_FAULT, "page fault", regs, -1,
  56. current->thread.trap_nr, SIGSEGV) == NOTIFY_STOP)
  57. return;
  58. #endif
  59. si_code = SEGV_MAPERR;
  60. /*
  61. * We fault-in kernel-space virtual memory on-demand. The
  62. * 'reference' page table is init_mm.pgd.
  63. *
  64. * NOTE! We MUST NOT take any locks for this case. We may
  65. * be in an interrupt or a critical region, and should
  66. * only copy the information from the master page table,
  67. * nothing more.
  68. */
  69. #ifdef CONFIG_64BIT
  70. # define VMALLOC_FAULT_TARGET no_context
  71. #else
  72. # define VMALLOC_FAULT_TARGET vmalloc_fault
  73. #endif
  74. if (unlikely(address >= VMALLOC_START && address <= VMALLOC_END))
  75. goto VMALLOC_FAULT_TARGET;
  76. #ifdef MODULE_START
  77. if (unlikely(address >= MODULE_START && address < MODULE_END))
  78. goto VMALLOC_FAULT_TARGET;
  79. #endif
  80. /*
  81. * If we're in an interrupt or have no user
  82. * context, we must not take the fault..
  83. */
  84. if (faulthandler_disabled() || !mm)
  85. goto bad_area_nosemaphore;
  86. if (user_mode(regs))
  87. flags |= FAULT_FLAG_USER;
  88. perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address);
  89. retry:
  90. mmap_read_lock(mm);
  91. vma = find_vma(mm, address);
  92. if (!vma)
  93. goto bad_area;
  94. if (vma->vm_start <= address)
  95. goto good_area;
  96. if (!(vma->vm_flags & VM_GROWSDOWN))
  97. goto bad_area;
  98. if (expand_stack(vma, address))
  99. goto bad_area;
  100. /*
  101. * Ok, we have a good vm_area for this memory access, so
  102. * we can handle it..
  103. */
  104. good_area:
  105. si_code = SEGV_ACCERR;
  106. if (write) {
  107. if (!(vma->vm_flags & VM_WRITE))
  108. goto bad_area;
  109. flags |= FAULT_FLAG_WRITE;
  110. } else {
  111. if (cpu_has_rixi) {
  112. if (address == regs->cp0_epc && !(vma->vm_flags & VM_EXEC)) {
  113. #if 0
  114. pr_notice("Cpu%d[%s:%d:%0*lx:%ld:%0*lx] XI violation\n",
  115. raw_smp_processor_id(),
  116. current->comm, current->pid,
  117. field, address, write,
  118. field, regs->cp0_epc);
  119. #endif
  120. goto bad_area;
  121. }
  122. if (!(vma->vm_flags & VM_READ) &&
  123. exception_epc(regs) != address) {
  124. #if 0
  125. pr_notice("Cpu%d[%s:%d:%0*lx:%ld:%0*lx] RI violation\n",
  126. raw_smp_processor_id(),
  127. current->comm, current->pid,
  128. field, address, write,
  129. field, regs->cp0_epc);
  130. #endif
  131. goto bad_area;
  132. }
  133. } else {
  134. if (unlikely(!vma_is_accessible(vma)))
  135. goto bad_area;
  136. }
  137. }
  138. /*
  139. * If for any reason at all we couldn't handle the fault,
  140. * make sure we exit gracefully rather than endlessly redo
  141. * the fault.
  142. */
  143. fault = handle_mm_fault(vma, address, flags, regs);
  144. if (fault_signal_pending(fault, regs))
  145. return;
  146. if (unlikely(fault & VM_FAULT_ERROR)) {
  147. if (fault & VM_FAULT_OOM)
  148. goto out_of_memory;
  149. else if (fault & VM_FAULT_SIGSEGV)
  150. goto bad_area;
  151. else if (fault & VM_FAULT_SIGBUS)
  152. goto do_sigbus;
  153. BUG();
  154. }
  155. if (flags & FAULT_FLAG_ALLOW_RETRY) {
  156. if (fault & VM_FAULT_RETRY) {
  157. flags |= FAULT_FLAG_TRIED;
  158. /*
  159. * No need to mmap_read_unlock(mm) as we would
  160. * have already released it in __lock_page_or_retry
  161. * in mm/filemap.c.
  162. */
  163. goto retry;
  164. }
  165. }
  166. mmap_read_unlock(mm);
  167. return;
  168. /*
  169. * Something tried to access memory that isn't in our memory map..
  170. * Fix it, but check if it's kernel or user first..
  171. */
  172. bad_area:
  173. mmap_read_unlock(mm);
  174. bad_area_nosemaphore:
  175. /* User mode accesses just cause a SIGSEGV */
  176. if (user_mode(regs)) {
  177. tsk->thread.cp0_badvaddr = address;
  178. tsk->thread.error_code = write;
  179. if (show_unhandled_signals &&
  180. unhandled_signal(tsk, SIGSEGV) &&
  181. __ratelimit(&ratelimit_state)) {
  182. pr_info("do_page_fault(): sending SIGSEGV to %s for invalid %s %0*lx\n",
  183. tsk->comm,
  184. write ? "write access to" : "read access from",
  185. field, address);
  186. pr_info("epc = %0*lx in", field,
  187. (unsigned long) regs->cp0_epc);
  188. print_vma_addr(KERN_CONT " ", regs->cp0_epc);
  189. pr_cont("\n");
  190. pr_info("ra = %0*lx in", field,
  191. (unsigned long) regs->regs[31]);
  192. print_vma_addr(KERN_CONT " ", regs->regs[31]);
  193. pr_cont("\n");
  194. }
  195. current->thread.trap_nr = (regs->cp0_cause >> 2) & 0x1f;
  196. force_sig_fault(SIGSEGV, si_code, (void __user *)address);
  197. return;
  198. }
  199. no_context:
  200. /* Are we prepared to handle this kernel fault? */
  201. if (fixup_exception(regs)) {
  202. current->thread.cp0_baduaddr = address;
  203. return;
  204. }
  205. /*
  206. * Oops. The kernel tried to access some bad page. We'll have to
  207. * terminate things with extreme prejudice.
  208. */
  209. bust_spinlocks(1);
  210. printk(KERN_ALERT "CPU %d Unable to handle kernel paging request at "
  211. "virtual address %0*lx, epc == %0*lx, ra == %0*lx\n",
  212. raw_smp_processor_id(), field, address, field, regs->cp0_epc,
  213. field, regs->regs[31]);
  214. die("Oops", regs);
  215. out_of_memory:
  216. /*
  217. * We ran out of memory, call the OOM killer, and return the userspace
  218. * (which will retry the fault, or kill us if we got oom-killed).
  219. */
  220. mmap_read_unlock(mm);
  221. if (!user_mode(regs))
  222. goto no_context;
  223. pagefault_out_of_memory();
  224. return;
  225. do_sigbus:
  226. mmap_read_unlock(mm);
  227. /* Kernel mode? Handle exceptions or die */
  228. if (!user_mode(regs))
  229. goto no_context;
  230. /*
  231. * Send a sigbus, regardless of whether we were in kernel
  232. * or user mode.
  233. */
  234. #if 0
  235. printk("do_page_fault() #3: sending SIGBUS to %s for "
  236. "invalid %s\n%0*lx (epc == %0*lx, ra == %0*lx)\n",
  237. tsk->comm,
  238. write ? "write access to" : "read access from",
  239. field, address,
  240. field, (unsigned long) regs->cp0_epc,
  241. field, (unsigned long) regs->regs[31]);
  242. #endif
  243. current->thread.trap_nr = (regs->cp0_cause >> 2) & 0x1f;
  244. tsk->thread.cp0_badvaddr = address;
  245. force_sig_fault(SIGBUS, BUS_ADRERR, (void __user *)address);
  246. return;
  247. #ifndef CONFIG_64BIT
  248. vmalloc_fault:
  249. {
  250. /*
  251. * Synchronize this task's top level page-table
  252. * with the 'reference' page table.
  253. *
  254. * Do _not_ use "tsk" here. We might be inside
  255. * an interrupt in the middle of a task switch..
  256. */
  257. int offset = pgd_index(address);
  258. pgd_t *pgd, *pgd_k;
  259. p4d_t *p4d, *p4d_k;
  260. pud_t *pud, *pud_k;
  261. pmd_t *pmd, *pmd_k;
  262. pte_t *pte_k;
  263. pgd = (pgd_t *) pgd_current[raw_smp_processor_id()] + offset;
  264. pgd_k = init_mm.pgd + offset;
  265. if (!pgd_present(*pgd_k))
  266. goto no_context;
  267. set_pgd(pgd, *pgd_k);
  268. p4d = p4d_offset(pgd, address);
  269. p4d_k = p4d_offset(pgd_k, address);
  270. if (!p4d_present(*p4d_k))
  271. goto no_context;
  272. pud = pud_offset(p4d, address);
  273. pud_k = pud_offset(p4d_k, address);
  274. if (!pud_present(*pud_k))
  275. goto no_context;
  276. pmd = pmd_offset(pud, address);
  277. pmd_k = pmd_offset(pud_k, address);
  278. if (!pmd_present(*pmd_k))
  279. goto no_context;
  280. set_pmd(pmd, *pmd_k);
  281. pte_k = pte_offset_kernel(pmd_k, address);
  282. if (!pte_present(*pte_k))
  283. goto no_context;
  284. return;
  285. }
  286. #endif
  287. }
  288. asmlinkage void __kprobes do_page_fault(struct pt_regs *regs,
  289. unsigned long write, unsigned long address)
  290. {
  291. enum ctx_state prev_state;
  292. prev_state = exception_enter();
  293. __do_page_fault(regs, write, address);
  294. exception_exit(prev_state);
  295. }