panic.c 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515
  1. /*
  2. * linux/kernel/panic.c
  3. *
  4. * Copyright (C) 1991, 1992 Linus Torvalds
  5. */
  6. /*
  7. * This function is used through-out the kernel (including mm and fs)
  8. * to indicate a major problem.
  9. */
  10. #include <linux/debug_locks.h>
  11. #include <linux/interrupt.h>
  12. #include <linux/kmsg_dump.h>
  13. #include <linux/kallsyms.h>
  14. #include <linux/notifier.h>
  15. #include <linux/module.h>
  16. #include <linux/random.h>
  17. #include <linux/ftrace.h>
  18. #include <linux/reboot.h>
  19. #include <linux/delay.h>
  20. #include <linux/kexec.h>
  21. #include <linux/sched.h>
  22. #include <linux/sysrq.h>
  23. #include <linux/init.h>
  24. #include <linux/nmi.h>
  25. #define PANIC_TIMER_STEP 100
  26. #define PANIC_BLINK_SPD 18
  27. /* Machine specific panic information string */
  28. char *mach_panic_string;
  29. int panic_on_oops = CONFIG_PANIC_ON_OOPS_VALUE;
  30. static unsigned long tainted_mask;
  31. static int pause_on_oops;
  32. static int pause_on_oops_flag;
  33. static DEFINE_SPINLOCK(pause_on_oops_lock);
  34. static bool crash_kexec_post_notifiers;
  35. int panic_timeout = CONFIG_PANIC_TIMEOUT;
  36. EXPORT_SYMBOL_GPL(panic_timeout);
  37. ATOMIC_NOTIFIER_HEAD(panic_notifier_list);
  38. EXPORT_SYMBOL(panic_notifier_list);
  39. static long no_blink(int state)
  40. {
  41. return 0;
  42. }
  43. /* Returns how long it waited in ms */
  44. long (*panic_blink)(int state);
  45. EXPORT_SYMBOL(panic_blink);
  46. /*
  47. * Stop ourself in panic -- architecture code may override this
  48. */
  49. void __weak panic_smp_self_stop(void)
  50. {
  51. while (1)
  52. cpu_relax();
  53. }
  54. /**
  55. * panic - halt the system
  56. * @fmt: The text string to print
  57. *
  58. * Display a message, then perform cleanups.
  59. *
  60. * This function never returns.
  61. */
  62. void panic(const char *fmt, ...)
  63. {
  64. static DEFINE_SPINLOCK(panic_lock);
  65. static char buf[1024];
  66. va_list args;
  67. long i, i_next = 0;
  68. int state = 0;
  69. /*
  70. * Disable local interrupts. This will prevent panic_smp_self_stop
  71. * from deadlocking the first cpu that invokes the panic, since
  72. * there is nothing to prevent an interrupt handler (that runs
  73. * after the panic_lock is acquired) from invoking panic again.
  74. */
  75. local_irq_disable();
  76. /*
  77. * It's possible to come here directly from a panic-assertion and
  78. * not have preempt disabled. Some functions called from here want
  79. * preempt to be disabled. No point enabling it later though...
  80. *
  81. * Only one CPU is allowed to execute the panic code from here. For
  82. * multiple parallel invocations of panic, all other CPUs either
  83. * stop themself or will wait until they are stopped by the 1st CPU
  84. * with smp_send_stop().
  85. */
  86. if (!spin_trylock(&panic_lock))
  87. panic_smp_self_stop();
  88. console_verbose();
  89. bust_spinlocks(1);
  90. va_start(args, fmt);
  91. vsnprintf(buf, sizeof(buf), fmt, args);
  92. va_end(args);
  93. pr_emerg("Kernel panic - not syncing: %s\n", buf);
  94. #ifdef CONFIG_DEBUG_BUGVERBOSE
  95. /*
  96. * Avoid nested stack-dumping if a panic occurs during oops processing
  97. */
  98. if (!test_taint(TAINT_DIE) && oops_in_progress <= 1)
  99. dump_stack();
  100. #endif
  101. /*
  102. * If we have crashed and we have a crash kernel loaded let it handle
  103. * everything else.
  104. * If we want to run this after calling panic_notifiers, pass
  105. * the "crash_kexec_post_notifiers" option to the kernel.
  106. */
  107. if (!crash_kexec_post_notifiers)
  108. crash_kexec(NULL);
  109. /*
  110. * Note smp_send_stop is the usual smp shutdown function, which
  111. * unfortunately means it may not be hardened to work in a panic
  112. * situation.
  113. */
  114. smp_send_stop();
  115. /*
  116. * Run any panic handlers, including those that might need to
  117. * add information to the kmsg dump output.
  118. */
  119. atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
  120. kmsg_dump(KMSG_DUMP_PANIC);
  121. /*
  122. * If you doubt kdump always works fine in any situation,
  123. * "crash_kexec_post_notifiers" offers you a chance to run
  124. * panic_notifiers and dumping kmsg before kdump.
  125. * Note: since some panic_notifiers can make crashed kernel
  126. * more unstable, it can increase risks of the kdump failure too.
  127. */
  128. crash_kexec(NULL);
  129. bust_spinlocks(0);
  130. if (!panic_blink)
  131. panic_blink = no_blink;
  132. if (panic_timeout > 0) {
  133. /*
  134. * Delay timeout seconds before rebooting the machine.
  135. * We can't use the "normal" timers since we just panicked.
  136. */
  137. pr_emerg("Rebooting in %d seconds..", panic_timeout);
  138. for (i = 0; i < panic_timeout * 1000; i += PANIC_TIMER_STEP) {
  139. touch_nmi_watchdog();
  140. if (i >= i_next) {
  141. i += panic_blink(state ^= 1);
  142. i_next = i + 3600 / PANIC_BLINK_SPD;
  143. }
  144. mdelay(PANIC_TIMER_STEP);
  145. }
  146. }
  147. if (panic_timeout != 0) {
  148. /*
  149. * This will not be a clean reboot, with everything
  150. * shutting down. But if there is a chance of
  151. * rebooting the system it will be rebooted.
  152. */
  153. emergency_restart();
  154. }
  155. #ifdef __sparc__
  156. {
  157. extern int stop_a_enabled;
  158. /* Make sure the user can actually press Stop-A (L1-A) */
  159. stop_a_enabled = 1;
  160. pr_emerg("Press Stop-A (L1-A) to return to the boot prom\n");
  161. }
  162. #endif
  163. #if defined(CONFIG_S390)
  164. {
  165. unsigned long caller;
  166. caller = (unsigned long)__builtin_return_address(0);
  167. disabled_wait(caller);
  168. }
  169. #endif
  170. pr_emerg("---[ end Kernel panic - not syncing: %s\n", buf);
  171. local_irq_enable();
  172. for (i = 0; ; i += PANIC_TIMER_STEP) {
  173. touch_softlockup_watchdog();
  174. if (i >= i_next) {
  175. i += panic_blink(state ^= 1);
  176. i_next = i + 3600 / PANIC_BLINK_SPD;
  177. }
  178. mdelay(PANIC_TIMER_STEP);
  179. }
  180. }
  181. EXPORT_SYMBOL(panic);
  182. struct tnt {
  183. u8 bit;
  184. char true;
  185. char false;
  186. };
  187. static const struct tnt tnts[] = {
  188. { TAINT_PROPRIETARY_MODULE, 'P', 'G' },
  189. { TAINT_FORCED_MODULE, 'F', ' ' },
  190. { TAINT_CPU_OUT_OF_SPEC, 'S', ' ' },
  191. { TAINT_FORCED_RMMOD, 'R', ' ' },
  192. { TAINT_MACHINE_CHECK, 'M', ' ' },
  193. { TAINT_BAD_PAGE, 'B', ' ' },
  194. { TAINT_USER, 'U', ' ' },
  195. { TAINT_DIE, 'D', ' ' },
  196. { TAINT_OVERRIDDEN_ACPI_TABLE, 'A', ' ' },
  197. { TAINT_WARN, 'W', ' ' },
  198. { TAINT_CRAP, 'C', ' ' },
  199. { TAINT_FIRMWARE_WORKAROUND, 'I', ' ' },
  200. { TAINT_OOT_MODULE, 'O', ' ' },
  201. { TAINT_UNSIGNED_MODULE, 'E', ' ' },
  202. { TAINT_SOFTLOCKUP, 'L', ' ' },
  203. };
  204. /**
  205. * print_tainted - return a string to represent the kernel taint state.
  206. *
  207. * 'P' - Proprietary module has been loaded.
  208. * 'F' - Module has been forcibly loaded.
  209. * 'S' - SMP with CPUs not designed for SMP.
  210. * 'R' - User forced a module unload.
  211. * 'M' - System experienced a machine check exception.
  212. * 'B' - System has hit bad_page.
  213. * 'U' - Userspace-defined naughtiness.
  214. * 'D' - Kernel has oopsed before
  215. * 'A' - ACPI table overridden.
  216. * 'W' - Taint on warning.
  217. * 'C' - modules from drivers/staging are loaded.
  218. * 'I' - Working around severe firmware bug.
  219. * 'O' - Out-of-tree module has been loaded.
  220. * 'E' - Unsigned module has been loaded.
  221. * 'L' - A soft lockup has previously occurred.
  222. *
  223. * The string is overwritten by the next call to print_tainted().
  224. */
  225. const char *print_tainted(void)
  226. {
  227. static char buf[ARRAY_SIZE(tnts) + sizeof("Tainted: ")];
  228. if (tainted_mask) {
  229. char *s;
  230. int i;
  231. s = buf + sprintf(buf, "Tainted: ");
  232. for (i = 0; i < ARRAY_SIZE(tnts); i++) {
  233. const struct tnt *t = &tnts[i];
  234. *s++ = test_bit(t->bit, &tainted_mask) ?
  235. t->true : t->false;
  236. }
  237. *s = 0;
  238. } else
  239. snprintf(buf, sizeof(buf), "Not tainted");
  240. return buf;
  241. }
  242. int test_taint(unsigned flag)
  243. {
  244. return test_bit(flag, &tainted_mask);
  245. }
  246. EXPORT_SYMBOL(test_taint);
  247. unsigned long get_taint(void)
  248. {
  249. return tainted_mask;
  250. }
  251. /**
  252. * add_taint: add a taint flag if not already set.
  253. * @flag: one of the TAINT_* constants.
  254. * @lockdep_ok: whether lock debugging is still OK.
  255. *
  256. * If something bad has gone wrong, you'll want @lockdebug_ok = false, but for
  257. * some notewortht-but-not-corrupting cases, it can be set to true.
  258. */
  259. void add_taint(unsigned flag, enum lockdep_ok lockdep_ok)
  260. {
  261. if (lockdep_ok == LOCKDEP_NOW_UNRELIABLE && __debug_locks_off())
  262. pr_warn("Disabling lock debugging due to kernel taint\n");
  263. set_bit(flag, &tainted_mask);
  264. }
  265. EXPORT_SYMBOL(add_taint);
  266. static void spin_msec(int msecs)
  267. {
  268. int i;
  269. for (i = 0; i < msecs; i++) {
  270. touch_nmi_watchdog();
  271. mdelay(1);
  272. }
  273. }
  274. /*
  275. * It just happens that oops_enter() and oops_exit() are identically
  276. * implemented...
  277. */
  278. static void do_oops_enter_exit(void)
  279. {
  280. unsigned long flags;
  281. static int spin_counter;
  282. if (!pause_on_oops)
  283. return;
  284. spin_lock_irqsave(&pause_on_oops_lock, flags);
  285. if (pause_on_oops_flag == 0) {
  286. /* This CPU may now print the oops message */
  287. pause_on_oops_flag = 1;
  288. } else {
  289. /* We need to stall this CPU */
  290. if (!spin_counter) {
  291. /* This CPU gets to do the counting */
  292. spin_counter = pause_on_oops;
  293. do {
  294. spin_unlock(&pause_on_oops_lock);
  295. spin_msec(MSEC_PER_SEC);
  296. spin_lock(&pause_on_oops_lock);
  297. } while (--spin_counter);
  298. pause_on_oops_flag = 0;
  299. } else {
  300. /* This CPU waits for a different one */
  301. while (spin_counter) {
  302. spin_unlock(&pause_on_oops_lock);
  303. spin_msec(1);
  304. spin_lock(&pause_on_oops_lock);
  305. }
  306. }
  307. }
  308. spin_unlock_irqrestore(&pause_on_oops_lock, flags);
  309. }
  310. /*
  311. * Return true if the calling CPU is allowed to print oops-related info.
  312. * This is a bit racy..
  313. */
  314. int oops_may_print(void)
  315. {
  316. return pause_on_oops_flag == 0;
  317. }
  318. /*
  319. * Called when the architecture enters its oops handler, before it prints
  320. * anything. If this is the first CPU to oops, and it's oopsing the first
  321. * time then let it proceed.
  322. *
  323. * This is all enabled by the pause_on_oops kernel boot option. We do all
  324. * this to ensure that oopses don't scroll off the screen. It has the
  325. * side-effect of preventing later-oopsing CPUs from mucking up the display,
  326. * too.
  327. *
  328. * It turns out that the CPU which is allowed to print ends up pausing for
  329. * the right duration, whereas all the other CPUs pause for twice as long:
  330. * once in oops_enter(), once in oops_exit().
  331. */
  332. void oops_enter(void)
  333. {
  334. tracing_off();
  335. /* can't trust the integrity of the kernel anymore: */
  336. debug_locks_off();
  337. do_oops_enter_exit();
  338. }
  339. /*
  340. * 64-bit random ID for oopses:
  341. */
  342. static u64 oops_id;
  343. static int init_oops_id(void)
  344. {
  345. if (!oops_id)
  346. get_random_bytes(&oops_id, sizeof(oops_id));
  347. else
  348. oops_id++;
  349. return 0;
  350. }
  351. late_initcall(init_oops_id);
  352. void print_oops_end_marker(void)
  353. {
  354. init_oops_id();
  355. if (mach_panic_string)
  356. printk(KERN_WARNING "Board Information: %s\n",
  357. mach_panic_string);
  358. pr_warn("---[ end trace %016llx ]---\n", (unsigned long long)oops_id);
  359. }
  360. /*
  361. * Called when the architecture exits its oops handler, after printing
  362. * everything.
  363. */
  364. void oops_exit(void)
  365. {
  366. do_oops_enter_exit();
  367. print_oops_end_marker();
  368. kmsg_dump(KMSG_DUMP_OOPS);
  369. }
  370. #ifdef WANT_WARN_ON_SLOWPATH
  371. struct slowpath_args {
  372. const char *fmt;
  373. va_list args;
  374. };
  375. static void warn_slowpath_common(const char *file, int line, void *caller,
  376. unsigned taint, struct slowpath_args *args)
  377. {
  378. disable_trace_on_warning();
  379. pr_warn("------------[ cut here ]------------\n");
  380. pr_warn("WARNING: CPU: %d PID: %d at %s:%d %pS()\n",
  381. raw_smp_processor_id(), current->pid, file, line, caller);
  382. if (args)
  383. vprintk(args->fmt, args->args);
  384. print_modules();
  385. dump_stack();
  386. print_oops_end_marker();
  387. /* Just a warning, don't kill lockdep. */
  388. add_taint(taint, LOCKDEP_STILL_OK);
  389. }
  390. void warn_slowpath_fmt(const char *file, int line, const char *fmt, ...)
  391. {
  392. struct slowpath_args args;
  393. args.fmt = fmt;
  394. va_start(args.args, fmt);
  395. warn_slowpath_common(file, line, __builtin_return_address(0),
  396. TAINT_WARN, &args);
  397. va_end(args.args);
  398. }
  399. EXPORT_SYMBOL(warn_slowpath_fmt);
  400. void warn_slowpath_fmt_taint(const char *file, int line,
  401. unsigned taint, const char *fmt, ...)
  402. {
  403. struct slowpath_args args;
  404. args.fmt = fmt;
  405. va_start(args.args, fmt);
  406. warn_slowpath_common(file, line, __builtin_return_address(0),
  407. taint, &args);
  408. va_end(args.args);
  409. }
  410. EXPORT_SYMBOL(warn_slowpath_fmt_taint);
  411. void warn_slowpath_null(const char *file, int line)
  412. {
  413. warn_slowpath_common(file, line, __builtin_return_address(0),
  414. TAINT_WARN, NULL);
  415. }
  416. EXPORT_SYMBOL(warn_slowpath_null);
  417. #endif
  418. #ifdef CONFIG_CC_STACKPROTECTOR
  419. /*
  420. * Called when gcc's -fstack-protector feature is used, and
  421. * gcc detects corruption of the on-stack canary value
  422. */
  423. __visible void __stack_chk_fail(void)
  424. {
  425. /*
  426. panic("stack-protector: Kernel stack is corrupted in: %p\n",
  427. __builtin_return_address(0));
  428. */
  429. BUG();
  430. }
  431. EXPORT_SYMBOL(__stack_chk_fail);
  432. #endif
  433. core_param(panic, panic_timeout, int, 0644);
  434. core_param(pause_on_oops, pause_on_oops, int, 0644);
  435. static int __init setup_crash_kexec_post_notifiers(char *s)
  436. {
  437. crash_kexec_post_notifiers = true;
  438. return 0;
  439. }
  440. early_param("crash_kexec_post_notifiers", setup_crash_kexec_post_notifiers);
  441. static int __init oops_setup(char *s)
  442. {
  443. if (!s)
  444. return -EINVAL;
  445. if (!strcmp(s, "panic"))
  446. panic_on_oops = 1;
  447. return 0;
  448. }
  449. early_param("oops", oops_setup);