1/* 2 * linux/kernel/panic.c 3 * 4 * Copyright (C) 1991, 1992 Linus Torvalds 5 */ 6 7/* 8 * This function is used through-out the kernel (including mm and fs) 9 * to indicate a major problem. 10 */ 11#include <linux/debug_locks.h> 12#include <linux/interrupt.h> 13#include <linux/kmsg_dump.h> 14#include <linux/kallsyms.h> 15#include <linux/notifier.h> 16#include <linux/module.h> 17#include <linux/random.h> 18#include <linux/ftrace.h> 19#include <linux/reboot.h> 20#include <linux/delay.h> 21#include <linux/kexec.h> 22#include <linux/sched.h> 23#include <linux/sysrq.h> 24#include <linux/init.h> 25#include <linux/nmi.h> 26#include <linux/console.h> 27 28#define PANIC_TIMER_STEP 100 29#define PANIC_BLINK_SPD 18 30 31int panic_on_oops = CONFIG_PANIC_ON_OOPS_VALUE; 32static unsigned long tainted_mask; 33static int pause_on_oops; 34static int pause_on_oops_flag; 35static DEFINE_SPINLOCK(pause_on_oops_lock); 36static bool crash_kexec_post_notifiers; 37int panic_on_warn __read_mostly; 38 39int panic_timeout = CONFIG_PANIC_TIMEOUT; 40EXPORT_SYMBOL_GPL(panic_timeout); 41 42ATOMIC_NOTIFIER_HEAD(panic_notifier_list); 43 44EXPORT_SYMBOL(panic_notifier_list); 45 46static long no_blink(int state) 47{ 48 return 0; 49} 50 51/* Returns how long it waited in ms */ 52long (*panic_blink)(int state); 53EXPORT_SYMBOL(panic_blink); 54 55/* 56 * Stop ourself in panic -- architecture code may override this 57 */ 58void __weak panic_smp_self_stop(void) 59{ 60 while (1) 61 cpu_relax(); 62} 63 64/** 65 * panic - halt the system 66 * @fmt: The text string to print 67 * 68 * Display a message, then perform cleanups. 69 * 70 * This function never returns. 71 */ 72void panic(const char *fmt, ...) 73{ 74 static DEFINE_SPINLOCK(panic_lock); 75 static char buf[1024]; 76 va_list args; 77 long i, i_next = 0; 78 int state = 0; 79 80 /* 81 * Disable local interrupts. This will prevent panic_smp_self_stop 82 * from deadlocking the first cpu that invokes the panic, since 83 * there is nothing to prevent an interrupt handler (that runs 84 * after the panic_lock is acquired) from invoking panic again. 85 */ 86 local_irq_disable(); 87 88 /* 89 * It's possible to come here directly from a panic-assertion and 90 * not have preempt disabled. Some functions called from here want 91 * preempt to be disabled. No point enabling it later though... 92 * 93 * Only one CPU is allowed to execute the panic code from here. For 94 * multiple parallel invocations of panic, all other CPUs either 95 * stop themself or will wait until they are stopped by the 1st CPU 96 * with smp_send_stop(). 97 */ 98 if (!spin_trylock(&panic_lock)) 99 panic_smp_self_stop(); 100 101 console_verbose(); 102 bust_spinlocks(1); 103 va_start(args, fmt); 104 vsnprintf(buf, sizeof(buf), fmt, args); 105 va_end(args); 106 pr_emerg("Kernel panic - not syncing: %s\n", buf); 107#ifdef CONFIG_DEBUG_BUGVERBOSE 108 /* 109 * Avoid nested stack-dumping if a panic occurs during oops processing 110 */ 111 if (!test_taint(TAINT_DIE) && oops_in_progress <= 1) 112 dump_stack(); 113#endif 114 115 /* 116 * If we have crashed and we have a crash kernel loaded let it handle 117 * everything else. 118 * If we want to run this after calling panic_notifiers, pass 119 * the "crash_kexec_post_notifiers" option to the kernel. 120 */ 121 if (!crash_kexec_post_notifiers) 122 crash_kexec(NULL); 123 124 /* 125 * Note smp_send_stop is the usual smp shutdown function, which 126 * unfortunately means it may not be hardened to work in a panic 127 * situation. 128 */ 129 smp_send_stop(); 130 131 /* 132 * Run any panic handlers, including those that might need to 133 * add information to the kmsg dump output. 134 */ 135 atomic_notifier_call_chain(&panic_notifier_list, 0, buf); 136 137 kmsg_dump(KMSG_DUMP_PANIC); 138 139 /* 140 * If you doubt kdump always works fine in any situation, 141 * "crash_kexec_post_notifiers" offers you a chance to run 142 * panic_notifiers and dumping kmsg before kdump. 143 * Note: since some panic_notifiers can make crashed kernel 144 * more unstable, it can increase risks of the kdump failure too. 145 */ 146 crash_kexec(NULL); 147 148 bust_spinlocks(0); 149 150 /* 151 * We may have ended up stopping the CPU holding the lock (in 152 * smp_send_stop()) while still having some valuable data in the console 153 * buffer. Try to acquire the lock then release it regardless of the 154 * result. The release will also print the buffers out. Locks debug 155 * should be disabled to avoid reporting bad unlock balance when 156 * panic() is not being callled from OOPS. 157 */ 158 debug_locks_off(); 159 console_flush_on_panic(); 160 161 if (!panic_blink) 162 panic_blink = no_blink; 163 164 if (panic_timeout > 0) { 165 /* 166 * Delay timeout seconds before rebooting the machine. 167 * We can't use the "normal" timers since we just panicked. 168 */ 169 pr_emerg("Rebooting in %d seconds..", panic_timeout); 170 171 for (i = 0; i < panic_timeout * 1000; i += PANIC_TIMER_STEP) { 172 touch_nmi_watchdog(); 173 if (i >= i_next) { 174 i += panic_blink(state ^= 1); 175 i_next = i + 3600 / PANIC_BLINK_SPD; 176 } 177 mdelay(PANIC_TIMER_STEP); 178 } 179 } 180 if (panic_timeout != 0) { 181 /* 182 * This will not be a clean reboot, with everything 183 * shutting down. But if there is a chance of 184 * rebooting the system it will be rebooted. 185 */ 186 emergency_restart(); 187 } 188#ifdef __sparc__ 189 { 190 extern int stop_a_enabled; 191 /* Make sure the user can actually press Stop-A (L1-A) */ 192 stop_a_enabled = 1; 193 pr_emerg("Press Stop-A (L1-A) to return to the boot prom\n"); 194 } 195#endif 196#if defined(CONFIG_S390) 197 { 198 unsigned long caller; 199 200 caller = (unsigned long)__builtin_return_address(0); 201 disabled_wait(caller); 202 } 203#endif 204 pr_emerg("---[ end Kernel panic - not syncing: %s\n", buf); 205 local_irq_enable(); 206 for (i = 0; ; i += PANIC_TIMER_STEP) { 207 touch_softlockup_watchdog(); 208 if (i >= i_next) { 209 i += panic_blink(state ^= 1); 210 i_next = i + 3600 / PANIC_BLINK_SPD; 211 } 212 mdelay(PANIC_TIMER_STEP); 213 } 214} 215 216EXPORT_SYMBOL(panic); 217 218 219struct tnt { 220 u8 bit; 221 char true; 222 char false; 223}; 224 225static const struct tnt tnts[] = { 226 { TAINT_PROPRIETARY_MODULE, 'P', 'G' }, 227 { TAINT_FORCED_MODULE, 'F', ' ' }, 228 { TAINT_CPU_OUT_OF_SPEC, 'S', ' ' }, 229 { TAINT_FORCED_RMMOD, 'R', ' ' }, 230 { TAINT_MACHINE_CHECK, 'M', ' ' }, 231 { TAINT_BAD_PAGE, 'B', ' ' }, 232 { TAINT_USER, 'U', ' ' }, 233 { TAINT_DIE, 'D', ' ' }, 234 { TAINT_OVERRIDDEN_ACPI_TABLE, 'A', ' ' }, 235 { TAINT_WARN, 'W', ' ' }, 236 { TAINT_CRAP, 'C', ' ' }, 237 { TAINT_FIRMWARE_WORKAROUND, 'I', ' ' }, 238 { TAINT_OOT_MODULE, 'O', ' ' }, 239 { TAINT_UNSIGNED_MODULE, 'E', ' ' }, 240 { TAINT_SOFTLOCKUP, 'L', ' ' }, 241 { TAINT_LIVEPATCH, 'K', ' ' }, 242}; 243 244/** 245 * print_tainted - return a string to represent the kernel taint state. 246 * 247 * 'P' - Proprietary module has been loaded. 248 * 'F' - Module has been forcibly loaded. 249 * 'S' - SMP with CPUs not designed for SMP. 250 * 'R' - User forced a module unload. 251 * 'M' - System experienced a machine check exception. 252 * 'B' - System has hit bad_page. 253 * 'U' - Userspace-defined naughtiness. 254 * 'D' - Kernel has oopsed before 255 * 'A' - ACPI table overridden. 256 * 'W' - Taint on warning. 257 * 'C' - modules from drivers/staging are loaded. 258 * 'I' - Working around severe firmware bug. 259 * 'O' - Out-of-tree module has been loaded. 260 * 'E' - Unsigned module has been loaded. 261 * 'L' - A soft lockup has previously occurred. 262 * 'K' - Kernel has been live patched. 263 * 264 * The string is overwritten by the next call to print_tainted(). 265 */ 266const char *print_tainted(void) 267{ 268 static char buf[ARRAY_SIZE(tnts) + sizeof("Tainted: ")]; 269 270 if (tainted_mask) { 271 char *s; 272 int i; 273 274 s = buf + sprintf(buf, "Tainted: "); 275 for (i = 0; i < ARRAY_SIZE(tnts); i++) { 276 const struct tnt *t = &tnts[i]; 277 *s++ = test_bit(t->bit, &tainted_mask) ? 278 t->true : t->false; 279 } 280 *s = 0; 281 } else 282 snprintf(buf, sizeof(buf), "Not tainted"); 283 284 return buf; 285} 286 287int test_taint(unsigned flag) 288{ 289 return test_bit(flag, &tainted_mask); 290} 291EXPORT_SYMBOL(test_taint); 292 293unsigned long get_taint(void) 294{ 295 return tainted_mask; 296} 297 298/** 299 * add_taint: add a taint flag if not already set. 300 * @flag: one of the TAINT_* constants. 301 * @lockdep_ok: whether lock debugging is still OK. 302 * 303 * If something bad has gone wrong, you'll want @lockdebug_ok = false, but for 304 * some notewortht-but-not-corrupting cases, it can be set to true. 305 */ 306void add_taint(unsigned flag, enum lockdep_ok lockdep_ok) 307{ 308 if (lockdep_ok == LOCKDEP_NOW_UNRELIABLE && __debug_locks_off()) 309 pr_warn("Disabling lock debugging due to kernel taint\n"); 310 311 set_bit(flag, &tainted_mask); 312} 313EXPORT_SYMBOL(add_taint); 314 315static void spin_msec(int msecs) 316{ 317 int i; 318 319 for (i = 0; i < msecs; i++) { 320 touch_nmi_watchdog(); 321 mdelay(1); 322 } 323} 324 325/* 326 * It just happens that oops_enter() and oops_exit() are identically 327 * implemented... 328 */ 329static void do_oops_enter_exit(void) 330{ 331 unsigned long flags; 332 static int spin_counter; 333 334 if (!pause_on_oops) 335 return; 336 337 spin_lock_irqsave(&pause_on_oops_lock, flags); 338 if (pause_on_oops_flag == 0) { 339 /* This CPU may now print the oops message */ 340 pause_on_oops_flag = 1; 341 } else { 342 /* We need to stall this CPU */ 343 if (!spin_counter) { 344 /* This CPU gets to do the counting */ 345 spin_counter = pause_on_oops; 346 do { 347 spin_unlock(&pause_on_oops_lock); 348 spin_msec(MSEC_PER_SEC); 349 spin_lock(&pause_on_oops_lock); 350 } while (--spin_counter); 351 pause_on_oops_flag = 0; 352 } else { 353 /* This CPU waits for a different one */ 354 while (spin_counter) { 355 spin_unlock(&pause_on_oops_lock); 356 spin_msec(1); 357 spin_lock(&pause_on_oops_lock); 358 } 359 } 360 } 361 spin_unlock_irqrestore(&pause_on_oops_lock, flags); 362} 363 364/* 365 * Return true if the calling CPU is allowed to print oops-related info. 366 * This is a bit racy.. 367 */ 368int oops_may_print(void) 369{ 370 return pause_on_oops_flag == 0; 371} 372 373/* 374 * Called when the architecture enters its oops handler, before it prints 375 * anything. If this is the first CPU to oops, and it's oopsing the first 376 * time then let it proceed. 377 * 378 * This is all enabled by the pause_on_oops kernel boot option. We do all 379 * this to ensure that oopses don't scroll off the screen. It has the 380 * side-effect of preventing later-oopsing CPUs from mucking up the display, 381 * too. 382 * 383 * It turns out that the CPU which is allowed to print ends up pausing for 384 * the right duration, whereas all the other CPUs pause for twice as long: 385 * once in oops_enter(), once in oops_exit(). 386 */ 387void oops_enter(void) 388{ 389 tracing_off(); 390 /* can't trust the integrity of the kernel anymore: */ 391 debug_locks_off(); 392 do_oops_enter_exit(); 393} 394 395/* 396 * 64-bit random ID for oopses: 397 */ 398static u64 oops_id; 399 400static int init_oops_id(void) 401{ 402 if (!oops_id) 403 get_random_bytes(&oops_id, sizeof(oops_id)); 404 else 405 oops_id++; 406 407 return 0; 408} 409late_initcall(init_oops_id); 410 411void print_oops_end_marker(void) 412{ 413 init_oops_id(); 414 pr_warn("---[ end trace %016llx ]---\n", (unsigned long long)oops_id); 415} 416 417/* 418 * Called when the architecture exits its oops handler, after printing 419 * everything. 420 */ 421void oops_exit(void) 422{ 423 do_oops_enter_exit(); 424 print_oops_end_marker(); 425 kmsg_dump(KMSG_DUMP_OOPS); 426} 427 428#ifdef WANT_WARN_ON_SLOWPATH 429struct slowpath_args { 430 const char *fmt; 431 va_list args; 432}; 433 434static void warn_slowpath_common(const char *file, int line, void *caller, 435 unsigned taint, struct slowpath_args *args) 436{ 437 disable_trace_on_warning(); 438 439 pr_warn("------------[ cut here ]------------\n"); 440 pr_warn("WARNING: CPU: %d PID: %d at %s:%d %pS()\n", 441 raw_smp_processor_id(), current->pid, file, line, caller); 442 443 if (args) 444 vprintk(args->fmt, args->args); 445 446 if (panic_on_warn) { 447 /* 448 * This thread may hit another WARN() in the panic path. 449 * Resetting this prevents additional WARN() from panicking the 450 * system on this thread. Other threads are blocked by the 451 * panic_mutex in panic(). 452 */ 453 panic_on_warn = 0; 454 panic("panic_on_warn set ...\n"); 455 } 456 457 print_modules(); 458 dump_stack(); 459 print_oops_end_marker(); 460 /* Just a warning, don't kill lockdep. */ 461 add_taint(taint, LOCKDEP_STILL_OK); 462} 463 464void warn_slowpath_fmt(const char *file, int line, const char *fmt, ...) 465{ 466 struct slowpath_args args; 467 468 args.fmt = fmt; 469 va_start(args.args, fmt); 470 warn_slowpath_common(file, line, __builtin_return_address(0), 471 TAINT_WARN, &args); 472 va_end(args.args); 473} 474EXPORT_SYMBOL(warn_slowpath_fmt); 475 476void warn_slowpath_fmt_taint(const char *file, int line, 477 unsigned taint, const char *fmt, ...) 478{ 479 struct slowpath_args args; 480 481 args.fmt = fmt; 482 va_start(args.args, fmt); 483 warn_slowpath_common(file, line, __builtin_return_address(0), 484 taint, &args); 485 va_end(args.args); 486} 487EXPORT_SYMBOL(warn_slowpath_fmt_taint); 488 489void warn_slowpath_null(const char *file, int line) 490{ 491 warn_slowpath_common(file, line, __builtin_return_address(0), 492 TAINT_WARN, NULL); 493} 494EXPORT_SYMBOL(warn_slowpath_null); 495#endif 496 497#ifdef CONFIG_CC_STACKPROTECTOR 498 499/* 500 * Called when gcc's -fstack-protector feature is used, and 501 * gcc detects corruption of the on-stack canary value 502 */ 503__visible void __stack_chk_fail(void) 504{ 505 panic("stack-protector: Kernel stack is corrupted in: %p\n", 506 __builtin_return_address(0)); 507} 508EXPORT_SYMBOL(__stack_chk_fail); 509 510#endif 511 512core_param(panic, panic_timeout, int, 0644); 513core_param(pause_on_oops, pause_on_oops, int, 0644); 514core_param(panic_on_warn, panic_on_warn, int, 0644); 515 516static int __init setup_crash_kexec_post_notifiers(char *s) 517{ 518 crash_kexec_post_notifiers = true; 519 return 0; 520} 521early_param("crash_kexec_post_notifiers", setup_crash_kexec_post_notifiers); 522 523static int __init oops_setup(char *s) 524{ 525 if (!s) 526 return -EINVAL; 527 if (!strcmp(s, "panic")) 528 panic_on_oops = 1; 529 return 0; 530} 531early_param("oops", oops_setup); 532