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 31 int panic_on_oops = CONFIG_PANIC_ON_OOPS_VALUE; 32 static unsigned long tainted_mask; 33 static int pause_on_oops; 34 static int pause_on_oops_flag; 35 static DEFINE_SPINLOCK(pause_on_oops_lock); 36 bool crash_kexec_post_notifiers; 37 int panic_on_warn __read_mostly; 38 39 int panic_timeout = CONFIG_PANIC_TIMEOUT; 40 EXPORT_SYMBOL_GPL(panic_timeout); 41 42 ATOMIC_NOTIFIER_HEAD(panic_notifier_list); 43 44 EXPORT_SYMBOL(panic_notifier_list); 45 46 static long no_blink(int state) 47 { 48 return 0; 49 } 50 51 /* Returns how long it waited in ms */ 52 long (*panic_blink)(int state); 53 EXPORT_SYMBOL(panic_blink); 54 55 /* 56 * Stop ourself in panic -- architecture code may override this 57 */ 58 void __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 */ 72 void 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 if (crash_kexec_post_notifiers) 147 crash_kexec(NULL); 148 149 bust_spinlocks(0); 150 151 /* 152 * We may have ended up stopping the CPU holding the lock (in 153 * smp_send_stop()) while still having some valuable data in the console 154 * buffer. Try to acquire the lock then release it regardless of the 155 * result. The release will also print the buffers out. Locks debug 156 * should be disabled to avoid reporting bad unlock balance when 157 * panic() is not being callled from OOPS. 158 */ 159 debug_locks_off(); 160 console_trylock(); 161 console_unlock(); 162 163 if (!panic_blink) 164 panic_blink = no_blink; 165 166 if (panic_timeout > 0) { 167 /* 168 * Delay timeout seconds before rebooting the machine. 169 * We can't use the "normal" timers since we just panicked. 170 */ 171 pr_emerg("Rebooting in %d seconds..", panic_timeout); 172 173 for (i = 0; i < panic_timeout * 1000; i += PANIC_TIMER_STEP) { 174 touch_nmi_watchdog(); 175 if (i >= i_next) { 176 i += panic_blink(state ^= 1); 177 i_next = i + 3600 / PANIC_BLINK_SPD; 178 } 179 mdelay(PANIC_TIMER_STEP); 180 } 181 } 182 if (panic_timeout != 0) { 183 /* 184 * This will not be a clean reboot, with everything 185 * shutting down. But if there is a chance of 186 * rebooting the system it will be rebooted. 187 */ 188 emergency_restart(); 189 } 190 #ifdef __sparc__ 191 { 192 extern int stop_a_enabled; 193 /* Make sure the user can actually press Stop-A (L1-A) */ 194 stop_a_enabled = 1; 195 pr_emerg("Press Stop-A (L1-A) to return to the boot prom\n"); 196 } 197 #endif 198 #if defined(CONFIG_S390) 199 { 200 unsigned long caller; 201 202 caller = (unsigned long)__builtin_return_address(0); 203 disabled_wait(caller); 204 } 205 #endif 206 pr_emerg("---[ end Kernel panic - not syncing: %s\n", buf); 207 local_irq_enable(); 208 for (i = 0; ; i += PANIC_TIMER_STEP) { 209 touch_softlockup_watchdog(); 210 if (i >= i_next) { 211 i += panic_blink(state ^= 1); 212 i_next = i + 3600 / PANIC_BLINK_SPD; 213 } 214 mdelay(PANIC_TIMER_STEP); 215 } 216 } 217 218 EXPORT_SYMBOL(panic); 219 220 221 struct tnt { 222 u8 bit; 223 char true; 224 char false; 225 }; 226 227 static const struct tnt tnts[] = { 228 { TAINT_PROPRIETARY_MODULE, 'P', 'G' }, 229 { TAINT_FORCED_MODULE, 'F', ' ' }, 230 { TAINT_CPU_OUT_OF_SPEC, 'S', ' ' }, 231 { TAINT_FORCED_RMMOD, 'R', ' ' }, 232 { TAINT_MACHINE_CHECK, 'M', ' ' }, 233 { TAINT_BAD_PAGE, 'B', ' ' }, 234 { TAINT_USER, 'U', ' ' }, 235 { TAINT_DIE, 'D', ' ' }, 236 { TAINT_OVERRIDDEN_ACPI_TABLE, 'A', ' ' }, 237 { TAINT_WARN, 'W', ' ' }, 238 { TAINT_CRAP, 'C', ' ' }, 239 { TAINT_FIRMWARE_WORKAROUND, 'I', ' ' }, 240 { TAINT_OOT_MODULE, 'O', ' ' }, 241 { TAINT_UNSIGNED_MODULE, 'E', ' ' }, 242 { TAINT_SOFTLOCKUP, 'L', ' ' }, 243 { TAINT_LIVEPATCH, 'K', ' ' }, 244 }; 245 246 /** 247 * print_tainted - return a string to represent the kernel taint state. 248 * 249 * 'P' - Proprietary module has been loaded. 250 * 'F' - Module has been forcibly loaded. 251 * 'S' - SMP with CPUs not designed for SMP. 252 * 'R' - User forced a module unload. 253 * 'M' - System experienced a machine check exception. 254 * 'B' - System has hit bad_page. 255 * 'U' - Userspace-defined naughtiness. 256 * 'D' - Kernel has oopsed before 257 * 'A' - ACPI table overridden. 258 * 'W' - Taint on warning. 259 * 'C' - modules from drivers/staging are loaded. 260 * 'I' - Working around severe firmware bug. 261 * 'O' - Out-of-tree module has been loaded. 262 * 'E' - Unsigned module has been loaded. 263 * 'L' - A soft lockup has previously occurred. 264 * 'K' - Kernel has been live patched. 265 * 266 * The string is overwritten by the next call to print_tainted(). 267 */ 268 const char *print_tainted(void) 269 { 270 static char buf[ARRAY_SIZE(tnts) + sizeof("Tainted: ")]; 271 272 if (tainted_mask) { 273 char *s; 274 int i; 275 276 s = buf + sprintf(buf, "Tainted: "); 277 for (i = 0; i < ARRAY_SIZE(tnts); i++) { 278 const struct tnt *t = &tnts[i]; 279 *s++ = test_bit(t->bit, &tainted_mask) ? 280 t->true : t->false; 281 } 282 *s = 0; 283 } else 284 snprintf(buf, sizeof(buf), "Not tainted"); 285 286 return buf; 287 } 288 289 int test_taint(unsigned flag) 290 { 291 return test_bit(flag, &tainted_mask); 292 } 293 EXPORT_SYMBOL(test_taint); 294 295 unsigned long get_taint(void) 296 { 297 return tainted_mask; 298 } 299 300 /** 301 * add_taint: add a taint flag if not already set. 302 * @flag: one of the TAINT_* constants. 303 * @lockdep_ok: whether lock debugging is still OK. 304 * 305 * If something bad has gone wrong, you'll want @lockdebug_ok = false, but for 306 * some notewortht-but-not-corrupting cases, it can be set to true. 307 */ 308 void add_taint(unsigned flag, enum lockdep_ok lockdep_ok) 309 { 310 if (lockdep_ok == LOCKDEP_NOW_UNRELIABLE && __debug_locks_off()) 311 pr_warn("Disabling lock debugging due to kernel taint\n"); 312 313 set_bit(flag, &tainted_mask); 314 } 315 EXPORT_SYMBOL(add_taint); 316 317 static void spin_msec(int msecs) 318 { 319 int i; 320 321 for (i = 0; i < msecs; i++) { 322 touch_nmi_watchdog(); 323 mdelay(1); 324 } 325 } 326 327 /* 328 * It just happens that oops_enter() and oops_exit() are identically 329 * implemented... 330 */ 331 static void do_oops_enter_exit(void) 332 { 333 unsigned long flags; 334 static int spin_counter; 335 336 if (!pause_on_oops) 337 return; 338 339 spin_lock_irqsave(&pause_on_oops_lock, flags); 340 if (pause_on_oops_flag == 0) { 341 /* This CPU may now print the oops message */ 342 pause_on_oops_flag = 1; 343 } else { 344 /* We need to stall this CPU */ 345 if (!spin_counter) { 346 /* This CPU gets to do the counting */ 347 spin_counter = pause_on_oops; 348 do { 349 spin_unlock(&pause_on_oops_lock); 350 spin_msec(MSEC_PER_SEC); 351 spin_lock(&pause_on_oops_lock); 352 } while (--spin_counter); 353 pause_on_oops_flag = 0; 354 } else { 355 /* This CPU waits for a different one */ 356 while (spin_counter) { 357 spin_unlock(&pause_on_oops_lock); 358 spin_msec(1); 359 spin_lock(&pause_on_oops_lock); 360 } 361 } 362 } 363 spin_unlock_irqrestore(&pause_on_oops_lock, flags); 364 } 365 366 /* 367 * Return true if the calling CPU is allowed to print oops-related info. 368 * This is a bit racy.. 369 */ 370 int oops_may_print(void) 371 { 372 return pause_on_oops_flag == 0; 373 } 374 375 /* 376 * Called when the architecture enters its oops handler, before it prints 377 * anything. If this is the first CPU to oops, and it's oopsing the first 378 * time then let it proceed. 379 * 380 * This is all enabled by the pause_on_oops kernel boot option. We do all 381 * this to ensure that oopses don't scroll off the screen. It has the 382 * side-effect of preventing later-oopsing CPUs from mucking up the display, 383 * too. 384 * 385 * It turns out that the CPU which is allowed to print ends up pausing for 386 * the right duration, whereas all the other CPUs pause for twice as long: 387 * once in oops_enter(), once in oops_exit(). 388 */ 389 void oops_enter(void) 390 { 391 tracing_off(); 392 /* can't trust the integrity of the kernel anymore: */ 393 debug_locks_off(); 394 do_oops_enter_exit(); 395 } 396 397 /* 398 * 64-bit random ID for oopses: 399 */ 400 static u64 oops_id; 401 402 static int init_oops_id(void) 403 { 404 if (!oops_id) 405 get_random_bytes(&oops_id, sizeof(oops_id)); 406 else 407 oops_id++; 408 409 return 0; 410 } 411 late_initcall(init_oops_id); 412 413 void print_oops_end_marker(void) 414 { 415 init_oops_id(); 416 pr_warn("---[ end trace %016llx ]---\n", (unsigned long long)oops_id); 417 } 418 419 /* 420 * Called when the architecture exits its oops handler, after printing 421 * everything. 422 */ 423 void oops_exit(void) 424 { 425 do_oops_enter_exit(); 426 print_oops_end_marker(); 427 kmsg_dump(KMSG_DUMP_OOPS); 428 } 429 430 #ifdef WANT_WARN_ON_SLOWPATH 431 struct slowpath_args { 432 const char *fmt; 433 va_list args; 434 }; 435 436 static void warn_slowpath_common(const char *file, int line, void *caller, 437 unsigned taint, struct slowpath_args *args) 438 { 439 disable_trace_on_warning(); 440 441 pr_warn("------------[ cut here ]------------\n"); 442 pr_warn("WARNING: CPU: %d PID: %d at %s:%d %pS()\n", 443 raw_smp_processor_id(), current->pid, file, line, caller); 444 445 if (args) 446 vprintk(args->fmt, args->args); 447 448 if (panic_on_warn) { 449 /* 450 * This thread may hit another WARN() in the panic path. 451 * Resetting this prevents additional WARN() from panicking the 452 * system on this thread. Other threads are blocked by the 453 * panic_mutex in panic(). 454 */ 455 panic_on_warn = 0; 456 panic("panic_on_warn set ...\n"); 457 } 458 459 print_modules(); 460 dump_stack(); 461 print_oops_end_marker(); 462 /* Just a warning, don't kill lockdep. */ 463 add_taint(taint, LOCKDEP_STILL_OK); 464 } 465 466 void warn_slowpath_fmt(const char *file, int line, const char *fmt, ...) 467 { 468 struct slowpath_args args; 469 470 args.fmt = fmt; 471 va_start(args.args, fmt); 472 warn_slowpath_common(file, line, __builtin_return_address(0), 473 TAINT_WARN, &args); 474 va_end(args.args); 475 } 476 EXPORT_SYMBOL(warn_slowpath_fmt); 477 478 void warn_slowpath_fmt_taint(const char *file, int line, 479 unsigned taint, const char *fmt, ...) 480 { 481 struct slowpath_args args; 482 483 args.fmt = fmt; 484 va_start(args.args, fmt); 485 warn_slowpath_common(file, line, __builtin_return_address(0), 486 taint, &args); 487 va_end(args.args); 488 } 489 EXPORT_SYMBOL(warn_slowpath_fmt_taint); 490 491 void warn_slowpath_null(const char *file, int line) 492 { 493 warn_slowpath_common(file, line, __builtin_return_address(0), 494 TAINT_WARN, NULL); 495 } 496 EXPORT_SYMBOL(warn_slowpath_null); 497 #endif 498 499 #ifdef CONFIG_CC_STACKPROTECTOR 500 501 /* 502 * Called when gcc's -fstack-protector feature is used, and 503 * gcc detects corruption of the on-stack canary value 504 */ 505 __visible void __stack_chk_fail(void) 506 { 507 panic("stack-protector: Kernel stack is corrupted in: %p\n", 508 __builtin_return_address(0)); 509 } 510 EXPORT_SYMBOL(__stack_chk_fail); 511 512 #endif 513 514 core_param(panic, panic_timeout, int, 0644); 515 core_param(pause_on_oops, pause_on_oops, int, 0644); 516 core_param(panic_on_warn, panic_on_warn, int, 0644); 517 518 static int __init setup_crash_kexec_post_notifiers(char *s) 519 { 520 crash_kexec_post_notifiers = true; 521 return 0; 522 } 523 early_param("crash_kexec_post_notifiers", setup_crash_kexec_post_notifiers); 524 525 static int __init oops_setup(char *s) 526 { 527 if (!s) 528 return -EINVAL; 529 if (!strcmp(s, "panic")) 530 panic_on_oops = 1; 531 return 0; 532 } 533 early_param("oops", oops_setup); 534