fault.c 8.5 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335
  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_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
  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. retry:
  89. down_read(&mm->mmap_sem);
  90. vma = find_vma(mm, address);
  91. if (!vma)
  92. goto bad_area;
  93. if (vma->vm_start <= address)
  94. goto good_area;
  95. if (!(vma->vm_flags & VM_GROWSDOWN))
  96. goto bad_area;
  97. if (expand_stack(vma, address))
  98. goto bad_area;
  99. /*
  100. * Ok, we have a good vm_area for this memory access, so
  101. * we can handle it..
  102. */
  103. good_area:
  104. si_code = SEGV_ACCERR;
  105. if (write) {
  106. if (!(vma->vm_flags & VM_WRITE))
  107. goto bad_area;
  108. flags |= FAULT_FLAG_WRITE;
  109. } else {
  110. if (cpu_has_rixi) {
  111. if (address == regs->cp0_epc && !(vma->vm_flags & VM_EXEC)) {
  112. #if 0
  113. pr_notice("Cpu%d[%s:%d:%0*lx:%ld:%0*lx] XI violation\n",
  114. raw_smp_processor_id(),
  115. current->comm, current->pid,
  116. field, address, write,
  117. field, regs->cp0_epc);
  118. #endif
  119. goto bad_area;
  120. }
  121. if (!(vma->vm_flags & VM_READ) &&
  122. exception_epc(regs) != address) {
  123. #if 0
  124. pr_notice("Cpu%d[%s:%d:%0*lx:%ld:%0*lx] RI violation\n",
  125. raw_smp_processor_id(),
  126. current->comm, current->pid,
  127. field, address, write,
  128. field, regs->cp0_epc);
  129. #endif
  130. goto bad_area;
  131. }
  132. } else {
  133. if (!(vma->vm_flags & (VM_READ | VM_WRITE | VM_EXEC)))
  134. goto bad_area;
  135. }
  136. }
  137. /*
  138. * If for any reason at all we couldn't handle the fault,
  139. * make sure we exit gracefully rather than endlessly redo
  140. * the fault.
  141. */
  142. fault = handle_mm_fault(vma, address, flags);
  143. if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current))
  144. return;
  145. perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address);
  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_MAJOR) {
  157. perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1,
  158. regs, address);
  159. tsk->maj_flt++;
  160. } else {
  161. perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1,
  162. regs, address);
  163. tsk->min_flt++;
  164. }
  165. if (fault & VM_FAULT_RETRY) {
  166. flags &= ~FAULT_FLAG_ALLOW_RETRY;
  167. flags |= FAULT_FLAG_TRIED;
  168. /*
  169. * No need to up_read(&mm->mmap_sem) as we would
  170. * have already released it in __lock_page_or_retry
  171. * in mm/filemap.c.
  172. */
  173. goto retry;
  174. }
  175. }
  176. up_read(&mm->mmap_sem);
  177. return;
  178. /*
  179. * Something tried to access memory that isn't in our memory map..
  180. * Fix it, but check if it's kernel or user first..
  181. */
  182. bad_area:
  183. up_read(&mm->mmap_sem);
  184. bad_area_nosemaphore:
  185. /* User mode accesses just cause a SIGSEGV */
  186. if (user_mode(regs)) {
  187. tsk->thread.cp0_badvaddr = address;
  188. tsk->thread.error_code = write;
  189. if (show_unhandled_signals &&
  190. unhandled_signal(tsk, SIGSEGV) &&
  191. __ratelimit(&ratelimit_state)) {
  192. pr_info("do_page_fault(): sending SIGSEGV to %s for invalid %s %0*lx\n",
  193. tsk->comm,
  194. write ? "write access to" : "read access from",
  195. field, address);
  196. pr_info("epc = %0*lx in", field,
  197. (unsigned long) regs->cp0_epc);
  198. print_vma_addr(KERN_CONT " ", regs->cp0_epc);
  199. pr_cont("\n");
  200. pr_info("ra = %0*lx in", field,
  201. (unsigned long) regs->regs[31]);
  202. print_vma_addr(KERN_CONT " ", regs->regs[31]);
  203. pr_cont("\n");
  204. }
  205. current->thread.trap_nr = (regs->cp0_cause >> 2) & 0x1f;
  206. force_sig_fault(SIGSEGV, si_code, (void __user *)address, tsk);
  207. return;
  208. }
  209. no_context:
  210. /* Are we prepared to handle this kernel fault? */
  211. if (fixup_exception(regs)) {
  212. current->thread.cp0_baduaddr = address;
  213. return;
  214. }
  215. /*
  216. * Oops. The kernel tried to access some bad page. We'll have to
  217. * terminate things with extreme prejudice.
  218. */
  219. bust_spinlocks(1);
  220. printk(KERN_ALERT "CPU %d Unable to handle kernel paging request at "
  221. "virtual address %0*lx, epc == %0*lx, ra == %0*lx\n",
  222. raw_smp_processor_id(), field, address, field, regs->cp0_epc,
  223. field, regs->regs[31]);
  224. die("Oops", regs);
  225. out_of_memory:
  226. /*
  227. * We ran out of memory, call the OOM killer, and return the userspace
  228. * (which will retry the fault, or kill us if we got oom-killed).
  229. */
  230. up_read(&mm->mmap_sem);
  231. if (!user_mode(regs))
  232. goto no_context;
  233. pagefault_out_of_memory();
  234. return;
  235. do_sigbus:
  236. up_read(&mm->mmap_sem);
  237. /* Kernel mode? Handle exceptions or die */
  238. if (!user_mode(regs))
  239. goto no_context;
  240. /*
  241. * Send a sigbus, regardless of whether we were in kernel
  242. * or user mode.
  243. */
  244. #if 0
  245. printk("do_page_fault() #3: sending SIGBUS to %s for "
  246. "invalid %s\n%0*lx (epc == %0*lx, ra == %0*lx)\n",
  247. tsk->comm,
  248. write ? "write access to" : "read access from",
  249. field, address,
  250. field, (unsigned long) regs->cp0_epc,
  251. field, (unsigned long) regs->regs[31]);
  252. #endif
  253. current->thread.trap_nr = (regs->cp0_cause >> 2) & 0x1f;
  254. tsk->thread.cp0_badvaddr = address;
  255. force_sig_fault(SIGBUS, BUS_ADRERR, (void __user *)address, tsk);
  256. return;
  257. #ifndef CONFIG_64BIT
  258. vmalloc_fault:
  259. {
  260. /*
  261. * Synchronize this task's top level page-table
  262. * with the 'reference' page table.
  263. *
  264. * Do _not_ use "tsk" here. We might be inside
  265. * an interrupt in the middle of a task switch..
  266. */
  267. int offset = __pgd_offset(address);
  268. pgd_t *pgd, *pgd_k;
  269. pud_t *pud, *pud_k;
  270. pmd_t *pmd, *pmd_k;
  271. pte_t *pte_k;
  272. pgd = (pgd_t *) pgd_current[raw_smp_processor_id()] + offset;
  273. pgd_k = init_mm.pgd + offset;
  274. if (!pgd_present(*pgd_k))
  275. goto no_context;
  276. set_pgd(pgd, *pgd_k);
  277. pud = pud_offset(pgd, address);
  278. pud_k = pud_offset(pgd_k, address);
  279. if (!pud_present(*pud_k))
  280. goto no_context;
  281. pmd = pmd_offset(pud, address);
  282. pmd_k = pmd_offset(pud_k, address);
  283. if (!pmd_present(*pmd_k))
  284. goto no_context;
  285. set_pmd(pmd, *pmd_k);
  286. pte_k = pte_offset_kernel(pmd_k, address);
  287. if (!pte_present(*pte_k))
  288. goto no_context;
  289. return;
  290. }
  291. #endif
  292. }
  293. asmlinkage void __kprobes do_page_fault(struct pt_regs *regs,
  294. unsigned long write, unsigned long address)
  295. {
  296. enum ctx_state prev_state;
  297. prev_state = exception_enter();
  298. __do_page_fault(regs, write, address);
  299. exception_exit(prev_state);
  300. }