panic.c

   1// SPDX-License-Identifier: GPL-2.0-only
   2/*
   3 *  linux/kernel/panic.c
   4 *
   5 *  Copyright (C) 1991, 1992  Linus Torvalds
   6 */
   7
   8/*
   9 * This function is used through-out the kernel (including mm and fs)
  10 * to indicate a major problem.
  11 */
  12#include <linux/debug_locks.h>
  13#include <linux/sched/debug.h>
  14#include <linux/interrupt.h>
  15#include <linux/kgdb.h>
  16#include <linux/kmsg_dump.h>
  17#include <linux/kallsyms.h>
  18#include <linux/notifier.h>
  19#include <linux/vt_kern.h>
  20#include <linux/module.h>
  21#include <linux/random.h>
  22#include <linux/ftrace.h>
  23#include <linux/reboot.h>
  24#include <linux/delay.h>
  25#include <linux/kexec.h>
  26#include <linux/panic_notifier.h>
  27#include <linux/sched.h>
  28#include <linux/string_helpers.h>
  29#include <linux/sysrq.h>
  30#include <linux/init.h>
  31#include <linux/nmi.h>
  32#include <linux/console.h>
  33#include <linux/bug.h>
  34#include <linux/ratelimit.h>
  35#include <linux/debugfs.h>
  36#include <linux/sysfs.h>
  37#include <linux/context_tracking.h>
  38#include <linux/seq_buf.h>
  39#include <linux/sys_info.h>
  40#include <trace/events/error_report.h>
  41#include <asm/sections.h>
  42
  43#define PANIC_TIMER_STEP 100
  44#define PANIC_BLINK_SPD 18
  45
  46#ifdef CONFIG_SMP
  47/*
  48 * Should we dump all CPUs backtraces in an oops event?
  49 * Defaults to 0, can be changed via sysctl.
  50 */
  51static unsigned int __read_mostly sysctl_oops_all_cpu_backtrace;
  52#else
  53#define sysctl_oops_all_cpu_backtrace 0
  54#endif /* CONFIG_SMP */
  55
  56int panic_on_oops = IS_ENABLED(CONFIG_PANIC_ON_OOPS);
  57static unsigned long tainted_mask =
  58	IS_ENABLED(CONFIG_RANDSTRUCT) ? (1 << TAINT_RANDSTRUCT) : 0;
  59static int pause_on_oops;
  60static int pause_on_oops_flag;
  61static DEFINE_SPINLOCK(pause_on_oops_lock);
  62bool crash_kexec_post_notifiers;
  63int panic_on_warn __read_mostly;
  64unsigned long panic_on_taint;
  65bool panic_on_taint_nousertaint = false;
  66static unsigned int warn_limit __read_mostly;
  67static bool panic_console_replay;
  68
  69bool panic_triggering_all_cpu_backtrace;
  70static bool panic_this_cpu_backtrace_printed;
  71
  72int panic_timeout = CONFIG_PANIC_TIMEOUT;
  73EXPORT_SYMBOL_GPL(panic_timeout);
  74
  75unsigned long panic_print;
  76
  77ATOMIC_NOTIFIER_HEAD(panic_notifier_list);
  78
  79EXPORT_SYMBOL(panic_notifier_list);
  80
  81static void panic_print_deprecated(void)
  82{
  83	pr_info_once("Kernel: The 'panic_print' parameter is now deprecated. Please use 'panic_sys_info' and 'panic_console_replay' instead.\n");
  84}
  85
  86#ifdef CONFIG_SYSCTL
  87
  88/*
  89 * Taint values can only be increased
  90 * This means we can safely use a temporary.
  91 */
  92static int proc_taint(const struct ctl_table *table, int write,
  93			       void *buffer, size_t *lenp, loff_t *ppos)
  94{
  95	struct ctl_table t;
  96	unsigned long tmptaint = get_taint();
  97	int err;
  98
  99	if (write && !capable(CAP_SYS_ADMIN))
 100		return -EPERM;
 101
 102	t = *table;
 103	t.data = &tmptaint;
 104	err = proc_doulongvec_minmax(&t, write, buffer, lenp, ppos);
 105	if (err < 0)
 106		return err;
 107
 108	if (write) {
 109		int i;
 110
 111		/*
 112		 * If we are relying on panic_on_taint not producing
 113		 * false positives due to userspace input, bail out
 114		 * before setting the requested taint flags.
 115		 */
 116		if (panic_on_taint_nousertaint && (tmptaint & panic_on_taint))
 117			return -EINVAL;
 118
 119		/*
 120		 * Poor man's atomic or. Not worth adding a primitive
 121		 * to everyone's atomic.h for this
 122		 */
 123		for (i = 0; i < TAINT_FLAGS_COUNT; i++)
 124			if ((1UL << i) & tmptaint)
 125				add_taint(i, LOCKDEP_STILL_OK);
 126	}
 127
 128	return err;
 129}
 130
 131static int sysctl_panic_print_handler(const struct ctl_table *table, int write,
 132			   void *buffer, size_t *lenp, loff_t *ppos)
 133{
 134	if (write)
 135		panic_print_deprecated();
 136	return proc_doulongvec_minmax(table, write, buffer, lenp, ppos);
 137}
 138
 139static const struct ctl_table kern_panic_table[] = {
 140#ifdef CONFIG_SMP
 141	{
 142		.procname       = "oops_all_cpu_backtrace",
 143		.data           = &sysctl_oops_all_cpu_backtrace,
 144		.maxlen         = sizeof(int),
 145		.mode           = 0644,
 146		.proc_handler   = proc_dointvec_minmax,
 147		.extra1         = SYSCTL_ZERO,
 148		.extra2         = SYSCTL_ONE,
 149	},
 150#endif
 151	{
 152		.procname	= "tainted",
 153		.maxlen		= sizeof(long),
 154		.mode		= 0644,
 155		.proc_handler	= proc_taint,
 156	},
 157	{
 158		.procname	= "panic",
 159		.data		= &panic_timeout,
 160		.maxlen		= sizeof(int),
 161		.mode		= 0644,
 162		.proc_handler	= proc_dointvec,
 163	},
 164	{
 165		.procname	= "panic_on_oops",
 166		.data		= &panic_on_oops,
 167		.maxlen		= sizeof(int),
 168		.mode		= 0644,
 169		.proc_handler	= proc_dointvec,
 170	},
 171	{
 172		.procname	= "panic_print",
 173		.data		= &panic_print,
 174		.maxlen		= sizeof(unsigned long),
 175		.mode		= 0644,
 176		.proc_handler	= sysctl_panic_print_handler,
 177	},
 178	{
 179		.procname	= "panic_on_warn",
 180		.data		= &panic_on_warn,
 181		.maxlen		= sizeof(int),
 182		.mode		= 0644,
 183		.proc_handler	= proc_dointvec_minmax,
 184		.extra1		= SYSCTL_ZERO,
 185		.extra2		= SYSCTL_ONE,
 186	},
 187	{
 188		.procname       = "warn_limit",
 189		.data           = &warn_limit,
 190		.maxlen         = sizeof(warn_limit),
 191		.mode           = 0644,
 192		.proc_handler   = proc_douintvec,
 193	},
 194#if (defined(CONFIG_X86_32) || defined(CONFIG_PARISC)) && \
 195	defined(CONFIG_DEBUG_STACKOVERFLOW)
 196	{
 197		.procname	= "panic_on_stackoverflow",
 198		.data		= &sysctl_panic_on_stackoverflow,
 199		.maxlen		= sizeof(int),
 200		.mode		= 0644,
 201		.proc_handler	= proc_dointvec,
 202	},
 203#endif
 204	{
 205		.procname	= "panic_sys_info",
 206		.data		= &panic_print,
 207		.maxlen         = sizeof(panic_print),
 208		.mode		= 0644,
 209		.proc_handler	= sysctl_sys_info_handler,
 210	},
 211};
 212
 213static __init int kernel_panic_sysctls_init(void)
 214{
 215	register_sysctl_init("kernel", kern_panic_table);
 216	return 0;
 217}
 218late_initcall(kernel_panic_sysctls_init);
 219#endif
 220
 221/* The format is "panic_sys_info=tasks,mem,locks,ftrace,..." */
 222static int __init setup_panic_sys_info(char *buf)
 223{
 224	/* There is no risk of race in kernel boot phase */
 225	panic_print = sys_info_parse_param(buf);
 226	return 1;
 227}
 228__setup("panic_sys_info=", setup_panic_sys_info);
 229
 230static atomic_t warn_count = ATOMIC_INIT(0);
 231
 232#ifdef CONFIG_SYSFS
 233static ssize_t warn_count_show(struct kobject *kobj, struct kobj_attribute *attr,
 234			       char *page)
 235{
 236	return sysfs_emit(page, "%d\n", atomic_read(&warn_count));
 237}
 238
 239static struct kobj_attribute warn_count_attr = __ATTR_RO(warn_count);
 240
 241static __init int kernel_panic_sysfs_init(void)
 242{
 243	sysfs_add_file_to_group(kernel_kobj, &warn_count_attr.attr, NULL);
 244	return 0;
 245}
 246late_initcall(kernel_panic_sysfs_init);
 247#endif
 248
 249static long no_blink(int state)
 250{
 251	return 0;
 252}
 253
 254/* Returns how long it waited in ms */
 255long (*panic_blink)(int state);
 256EXPORT_SYMBOL(panic_blink);
 257
 258/*
 259 * Stop ourself in panic -- architecture code may override this
 260 */
 261void __weak __noreturn panic_smp_self_stop(void)
 262{
 263	while (1)
 264		cpu_relax();
 265}
 266
 267/*
 268 * Stop ourselves in NMI context if another CPU has already panicked. Arch code
 269 * may override this to prepare for crash dumping, e.g. save regs info.
 270 */
 271void __weak __noreturn nmi_panic_self_stop(struct pt_regs *regs)
 272{
 273	panic_smp_self_stop();
 274}
 275
 276/*
 277 * Stop other CPUs in panic.  Architecture dependent code may override this
 278 * with more suitable version.  For example, if the architecture supports
 279 * crash dump, it should save registers of each stopped CPU and disable
 280 * per-CPU features such as virtualization extensions.
 281 */
 282void __weak crash_smp_send_stop(void)
 283{
 284	static int cpus_stopped;
 285
 286	/*
 287	 * This function can be called twice in panic path, but obviously
 288	 * we execute this only once.
 289	 */
 290	if (cpus_stopped)
 291		return;
 292
 293	/*
 294	 * Note smp_send_stop is the usual smp shutdown function, which
 295	 * unfortunately means it may not be hardened to work in a panic
 296	 * situation.
 297	 */
 298	smp_send_stop();
 299	cpus_stopped = 1;
 300}
 301
 302atomic_t panic_cpu = ATOMIC_INIT(PANIC_CPU_INVALID);
 303
 304bool panic_try_start(void)
 305{
 306	int old_cpu, this_cpu;
 307
 308	/*
 309	 * Only one CPU is allowed to execute the crash_kexec() code as with
 310	 * panic().  Otherwise parallel calls of panic() and crash_kexec()
 311	 * may stop each other.  To exclude them, we use panic_cpu here too.
 312	 */
 313	old_cpu = PANIC_CPU_INVALID;
 314	this_cpu = raw_smp_processor_id();
 315
 316	return atomic_try_cmpxchg(&panic_cpu, &old_cpu, this_cpu);
 317}
 318EXPORT_SYMBOL(panic_try_start);
 319
 320void panic_reset(void)
 321{
 322	atomic_set(&panic_cpu, PANIC_CPU_INVALID);
 323}
 324EXPORT_SYMBOL(panic_reset);
 325
 326bool panic_in_progress(void)
 327{
 328	return unlikely(atomic_read(&panic_cpu) != PANIC_CPU_INVALID);
 329}
 330EXPORT_SYMBOL(panic_in_progress);
 331
 332/* Return true if a panic is in progress on the current CPU. */
 333bool panic_on_this_cpu(void)
 334{
 335	/*
 336	 * We can use raw_smp_processor_id() here because it is impossible for
 337	 * the task to be migrated to the panic_cpu, or away from it. If
 338	 * panic_cpu has already been set, and we're not currently executing on
 339	 * that CPU, then we never will be.
 340	 */
 341	return unlikely(atomic_read(&panic_cpu) == raw_smp_processor_id());
 342}
 343EXPORT_SYMBOL(panic_on_this_cpu);
 344
 345/*
 346 * Return true if a panic is in progress on a remote CPU.
 347 *
 348 * On true, the local CPU should immediately release any printing resources
 349 * that may be needed by the panic CPU.
 350 */
 351bool panic_on_other_cpu(void)
 352{
 353	return (panic_in_progress() && !panic_on_this_cpu());
 354}
 355EXPORT_SYMBOL(panic_on_other_cpu);
 356
 357/*
 358 * A variant of panic() called from NMI context. We return if we've already
 359 * panicked on this CPU. If another CPU already panicked, loop in
 360 * nmi_panic_self_stop() which can provide architecture dependent code such
 361 * as saving register state for crash dump.
 362 */
 363void nmi_panic(struct pt_regs *regs, const char *msg)
 364{
 365	if (panic_try_start())
 366		panic("%s", msg);
 367	else if (panic_on_other_cpu())
 368		nmi_panic_self_stop(regs);
 369}
 370EXPORT_SYMBOL(nmi_panic);
 371
 372void check_panic_on_warn(const char *origin)
 373{
 374	unsigned int limit;
 375
 376	if (panic_on_warn)
 377		panic("%s: panic_on_warn set ...\n", origin);
 378
 379	limit = READ_ONCE(warn_limit);
 380	if (atomic_inc_return(&warn_count) >= limit && limit)
 381		panic("%s: system warned too often (kernel.warn_limit is %d)",
 382		      origin, limit);
 383}
 384
 385static void panic_trigger_all_cpu_backtrace(void)
 386{
 387	/* Temporary allow non-panic CPUs to write their backtraces. */
 388	panic_triggering_all_cpu_backtrace = true;
 389
 390	if (panic_this_cpu_backtrace_printed)
 391		trigger_allbutcpu_cpu_backtrace(raw_smp_processor_id());
 392	else
 393		trigger_all_cpu_backtrace();
 394
 395	panic_triggering_all_cpu_backtrace = false;
 396}
 397
 398/*
 399 * Helper that triggers the NMI backtrace (if set in panic_print)
 400 * and then performs the secondary CPUs shutdown - we cannot have
 401 * the NMI backtrace after the CPUs are off!
 402 */
 403static void panic_other_cpus_shutdown(bool crash_kexec)
 404{
 405	if (panic_print & SYS_INFO_ALL_BT)
 406		panic_trigger_all_cpu_backtrace();
 407
 408	/*
 409	 * Note that smp_send_stop() is the usual SMP shutdown function,
 410	 * which unfortunately may not be hardened to work in a panic
 411	 * situation. If we want to do crash dump after notifier calls
 412	 * and kmsg_dump, we will need architecture dependent extra
 413	 * bits in addition to stopping other CPUs, hence we rely on
 414	 * crash_smp_send_stop() for that.
 415	 */
 416	if (!crash_kexec)
 417		smp_send_stop();
 418	else
 419		crash_smp_send_stop();
 420}
 421
 422/**
 423 * vpanic - halt the system
 424 * @fmt: The text string to print
 425 * @args: Arguments for the format string
 426 *
 427 * Display a message, then perform cleanups. This function never returns.
 428 */
 429void vpanic(const char *fmt, va_list args)
 430{
 431	static char buf[1024];
 432	long i, i_next = 0, len;
 433	int state = 0;
 434	bool _crash_kexec_post_notifiers = crash_kexec_post_notifiers;
 435
 436	if (panic_on_warn) {
 437		/*
 438		 * This thread may hit another WARN() in the panic path.
 439		 * Resetting this prevents additional WARN() from panicking the
 440		 * system on this thread.  Other threads are blocked by the
 441		 * panic_mutex in panic().
 442		 */
 443		panic_on_warn = 0;
 444	}
 445
 446	/*
 447	 * Disable local interrupts. This will prevent panic_smp_self_stop
 448	 * from deadlocking the first cpu that invokes the panic, since
 449	 * there is nothing to prevent an interrupt handler (that runs
 450	 * after setting panic_cpu) from invoking panic() again.
 451	 */
 452	local_irq_disable();
 453	preempt_disable_notrace();
 454
 455	/*
 456	 * It's possible to come here directly from a panic-assertion and
 457	 * not have preempt disabled. Some functions called from here want
 458	 * preempt to be disabled. No point enabling it later though...
 459	 *
 460	 * Only one CPU is allowed to execute the panic code from here. For
 461	 * multiple parallel invocations of panic, all other CPUs either
 462	 * stop themself or will wait until they are stopped by the 1st CPU
 463	 * with smp_send_stop().
 464	 *
 465	 * cmpxchg success means this is the 1st CPU which comes here,
 466	 * so go ahead.
 467	 * `old_cpu == this_cpu' means we came from nmi_panic() which sets
 468	 * panic_cpu to this CPU.  In this case, this is also the 1st CPU.
 469	 */
 470	/* atomic_try_cmpxchg updates old_cpu on failure */
 471	if (panic_try_start()) {
 472		/* go ahead */
 473	} else if (panic_on_other_cpu())
 474		panic_smp_self_stop();
 475
 476	console_verbose();
 477	bust_spinlocks(1);
 478	len = vscnprintf(buf, sizeof(buf), fmt, args);
 479
 480	if (len && buf[len - 1] == '\n')
 481		buf[len - 1] = '\0';
 482
 483	pr_emerg("Kernel panic - not syncing: %s\n", buf);
 484	/*
 485	 * Avoid nested stack-dumping if a panic occurs during oops processing
 486	 */
 487	if (test_taint(TAINT_DIE) || oops_in_progress > 1) {
 488		panic_this_cpu_backtrace_printed = true;
 489	} else if (IS_ENABLED(CONFIG_DEBUG_BUGVERBOSE)) {
 490		dump_stack();
 491		panic_this_cpu_backtrace_printed = true;
 492	}
 493
 494	/*
 495	 * If kgdb is enabled, give it a chance to run before we stop all
 496	 * the other CPUs or else we won't be able to debug processes left
 497	 * running on them.
 498	 */
 499	kgdb_panic(buf);
 500
 501	/*
 502	 * If we have crashed and we have a crash kernel loaded let it handle
 503	 * everything else.
 504	 * If we want to run this after calling panic_notifiers, pass
 505	 * the "crash_kexec_post_notifiers" option to the kernel.
 506	 *
 507	 * Bypass the panic_cpu check and call __crash_kexec directly.
 508	 */
 509	if (!_crash_kexec_post_notifiers)
 510		__crash_kexec(NULL);
 511
 512	panic_other_cpus_shutdown(_crash_kexec_post_notifiers);
 513
 514	printk_legacy_allow_panic_sync();
 515
 516	/*
 517	 * Run any panic handlers, including those that might need to
 518	 * add information to the kmsg dump output.
 519	 */
 520	atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
 521
 522	sys_info(panic_print);
 523
 524	kmsg_dump_desc(KMSG_DUMP_PANIC, buf);
 525
 526	/*
 527	 * If you doubt kdump always works fine in any situation,
 528	 * "crash_kexec_post_notifiers" offers you a chance to run
 529	 * panic_notifiers and dumping kmsg before kdump.
 530	 * Note: since some panic_notifiers can make crashed kernel
 531	 * more unstable, it can increase risks of the kdump failure too.
 532	 *
 533	 * Bypass the panic_cpu check and call __crash_kexec directly.
 534	 */
 535	if (_crash_kexec_post_notifiers)
 536		__crash_kexec(NULL);
 537
 538	console_unblank();
 539
 540	/*
 541	 * We may have ended up stopping the CPU holding the lock (in
 542	 * smp_send_stop()) while still having some valuable data in the console
 543	 * buffer.  Try to acquire the lock then release it regardless of the
 544	 * result.  The release will also print the buffers out.  Locks debug
 545	 * should be disabled to avoid reporting bad unlock balance when
 546	 * panic() is not being callled from OOPS.
 547	 */
 548	debug_locks_off();
 549	console_flush_on_panic(CONSOLE_FLUSH_PENDING);
 550
 551	if ((panic_print & SYS_INFO_PANIC_CONSOLE_REPLAY) ||
 552		panic_console_replay)
 553		console_flush_on_panic(CONSOLE_REPLAY_ALL);
 554
 555	if (!panic_blink)
 556		panic_blink = no_blink;
 557
 558	if (panic_timeout > 0) {
 559		/*
 560		 * Delay timeout seconds before rebooting the machine.
 561		 * We can't use the "normal" timers since we just panicked.
 562		 */
 563		pr_emerg("Rebooting in %d seconds..\n", panic_timeout);
 564
 565		for (i = 0; i < panic_timeout * 1000; i += PANIC_TIMER_STEP) {
 566			touch_nmi_watchdog();
 567			if (i >= i_next) {
 568				i += panic_blink(state ^= 1);
 569				i_next = i + 3600 / PANIC_BLINK_SPD;
 570			}
 571			mdelay(PANIC_TIMER_STEP);
 572		}
 573	}
 574	if (panic_timeout != 0) {
 575		/*
 576		 * This will not be a clean reboot, with everything
 577		 * shutting down.  But if there is a chance of
 578		 * rebooting the system it will be rebooted.
 579		 */
 580		if (panic_reboot_mode != REBOOT_UNDEFINED)
 581			reboot_mode = panic_reboot_mode;
 582		emergency_restart();
 583	}
 584#ifdef __sparc__
 585	{
 586		extern int stop_a_enabled;
 587		/* Make sure the user can actually press Stop-A (L1-A) */
 588		stop_a_enabled = 1;
 589		pr_emerg("Press Stop-A (L1-A) from sun keyboard or send break\n"
 590			 "twice on console to return to the boot prom\n");
 591	}
 592#endif
 593#if defined(CONFIG_S390)
 594	disabled_wait();
 595#endif
 596	pr_emerg("---[ end Kernel panic - not syncing: %s ]---\n", buf);
 597
 598	/* Do not scroll important messages printed above */
 599	suppress_printk = 1;
 600
 601	/*
 602	 * The final messages may not have been printed if in a context that
 603	 * defers printing (such as NMI) and irq_work is not available.
 604	 * Explicitly flush the kernel log buffer one last time.
 605	 */
 606	console_flush_on_panic(CONSOLE_FLUSH_PENDING);
 607	nbcon_atomic_flush_unsafe();
 608
 609	local_irq_enable();
 610	for (i = 0; ; i += PANIC_TIMER_STEP) {
 611		touch_softlockup_watchdog();
 612		if (i >= i_next) {
 613			i += panic_blink(state ^= 1);
 614			i_next = i + 3600 / PANIC_BLINK_SPD;
 615		}
 616		mdelay(PANIC_TIMER_STEP);
 617	}
 618}
 619EXPORT_SYMBOL(vpanic);
 620
 621/* Identical to vpanic(), except it takes variadic arguments instead of va_list */
 622void panic(const char *fmt, ...)
 623{
 624	va_list args;
 625
 626	va_start(args, fmt);
 627	vpanic(fmt, args);
 628	va_end(args);
 629}
 630EXPORT_SYMBOL(panic);
 631
 632#define TAINT_FLAG(taint, _c_true, _c_false)				\
 633	[ TAINT_##taint ] = {						\
 634		.c_true = _c_true, .c_false = _c_false,			\
 635		.desc = #taint,						\
 636	}
 637
 638/*
 639 * NOTE: if you modify the taint_flags or TAINT_FLAGS_COUNT,
 640 * please also modify tools/debugging/kernel-chktaint and
 641 * Documentation/admin-guide/tainted-kernels.rst, including its
 642 * small shell script that prints the TAINT_FLAGS_COUNT bits of
 643 * /proc/sys/kernel/tainted.
 644 */
 645const struct taint_flag taint_flags[TAINT_FLAGS_COUNT] = {
 646	TAINT_FLAG(PROPRIETARY_MODULE,		'P', 'G'),
 647	TAINT_FLAG(FORCED_MODULE,		'F', ' '),
 648	TAINT_FLAG(CPU_OUT_OF_SPEC,		'S', ' '),
 649	TAINT_FLAG(FORCED_RMMOD,		'R', ' '),
 650	TAINT_FLAG(MACHINE_CHECK,		'M', ' '),
 651	TAINT_FLAG(BAD_PAGE,			'B', ' '),
 652	TAINT_FLAG(USER,			'U', ' '),
 653	TAINT_FLAG(DIE,				'D', ' '),
 654	TAINT_FLAG(OVERRIDDEN_ACPI_TABLE,	'A', ' '),
 655	TAINT_FLAG(WARN,			'W', ' '),
 656	TAINT_FLAG(CRAP,			'C', ' '),
 657	TAINT_FLAG(FIRMWARE_WORKAROUND,		'I', ' '),
 658	TAINT_FLAG(OOT_MODULE,			'O', ' '),
 659	TAINT_FLAG(UNSIGNED_MODULE,		'E', ' '),
 660	TAINT_FLAG(SOFTLOCKUP,			'L', ' '),
 661	TAINT_FLAG(LIVEPATCH,			'K', ' '),
 662	TAINT_FLAG(AUX,				'X', ' '),
 663	TAINT_FLAG(RANDSTRUCT,			'T', ' '),
 664	TAINT_FLAG(TEST,			'N', ' '),
 665	TAINT_FLAG(FWCTL,			'J', ' '),
 666};
 667
 668#undef TAINT_FLAG
 669
 670static void print_tainted_seq(struct seq_buf *s, bool verbose)
 671{
 672	const char *sep = "";
 673	int i;
 674
 675	if (!tainted_mask) {
 676		seq_buf_puts(s, "Not tainted");
 677		return;
 678	}
 679
 680	seq_buf_printf(s, "Tainted: ");
 681	for (i = 0; i < TAINT_FLAGS_COUNT; i++) {
 682		const struct taint_flag *t = &taint_flags[i];
 683		bool is_set = test_bit(i, &tainted_mask);
 684		char c = is_set ? t->c_true : t->c_false;
 685
 686		if (verbose) {
 687			if (is_set) {
 688				seq_buf_printf(s, "%s[%c]=%s", sep, c, t->desc);
 689				sep = ", ";
 690			}
 691		} else {
 692			seq_buf_putc(s, c);
 693		}
 694	}
 695}
 696
 697static const char *_print_tainted(bool verbose)
 698{
 699	/* FIXME: what should the size be? */
 700	static char buf[sizeof(taint_flags)];
 701	struct seq_buf s;
 702
 703	BUILD_BUG_ON(ARRAY_SIZE(taint_flags) != TAINT_FLAGS_COUNT);
 704
 705	seq_buf_init(&s, buf, sizeof(buf));
 706
 707	print_tainted_seq(&s, verbose);
 708
 709	return seq_buf_str(&s);
 710}
 711
 712/**
 713 * print_tainted - return a string to represent the kernel taint state.
 714 *
 715 * For individual taint flag meanings, see Documentation/admin-guide/sysctl/kernel.rst
 716 *
 717 * The string is overwritten by the next call to print_tainted(),
 718 * but is always NULL terminated.
 719 */
 720const char *print_tainted(void)
 721{
 722	return _print_tainted(false);
 723}
 724
 725/**
 726 * print_tainted_verbose - A more verbose version of print_tainted()
 727 */
 728const char *print_tainted_verbose(void)
 729{
 730	return _print_tainted(true);
 731}
 732
 733int test_taint(unsigned flag)
 734{
 735	return test_bit(flag, &tainted_mask);
 736}
 737EXPORT_SYMBOL(test_taint);
 738
 739unsigned long get_taint(void)
 740{
 741	return tainted_mask;
 742}
 743
 744/**
 745 * add_taint: add a taint flag if not already set.
 746 * @flag: one of the TAINT_* constants.
 747 * @lockdep_ok: whether lock debugging is still OK.
 748 *
 749 * If something bad has gone wrong, you'll want @lockdebug_ok = false, but for
 750 * some notewortht-but-not-corrupting cases, it can be set to true.
 751 */
 752void add_taint(unsigned flag, enum lockdep_ok lockdep_ok)
 753{
 754	if (lockdep_ok == LOCKDEP_NOW_UNRELIABLE && __debug_locks_off())
 755		pr_warn("Disabling lock debugging due to kernel taint\n");
 756
 757	set_bit(flag, &tainted_mask);
 758
 759	if (tainted_mask & panic_on_taint) {
 760		panic_on_taint = 0;
 761		panic("panic_on_taint set ...");
 762	}
 763}
 764EXPORT_SYMBOL(add_taint);
 765
 766static void spin_msec(int msecs)
 767{
 768	int i;
 769
 770	for (i = 0; i < msecs; i++) {
 771		touch_nmi_watchdog();
 772		mdelay(1);
 773	}
 774}
 775
 776/*
 777 * It just happens that oops_enter() and oops_exit() are identically
 778 * implemented...
 779 */
 780static void do_oops_enter_exit(void)
 781{
 782	unsigned long flags;
 783	static int spin_counter;
 784
 785	if (!pause_on_oops)
 786		return;
 787
 788	spin_lock_irqsave(&pause_on_oops_lock, flags);
 789	if (pause_on_oops_flag == 0) {
 790		/* This CPU may now print the oops message */
 791		pause_on_oops_flag = 1;
 792	} else {
 793		/* We need to stall this CPU */
 794		if (!spin_counter) {
 795			/* This CPU gets to do the counting */
 796			spin_counter = pause_on_oops;
 797			do {
 798				spin_unlock(&pause_on_oops_lock);
 799				spin_msec(MSEC_PER_SEC);
 800				spin_lock(&pause_on_oops_lock);
 801			} while (--spin_counter);
 802			pause_on_oops_flag = 0;
 803		} else {
 804			/* This CPU waits for a different one */
 805			while (spin_counter) {
 806				spin_unlock(&pause_on_oops_lock);
 807				spin_msec(1);
 808				spin_lock(&pause_on_oops_lock);
 809			}
 810		}
 811	}
 812	spin_unlock_irqrestore(&pause_on_oops_lock, flags);
 813}
 814
 815/*
 816 * Return true if the calling CPU is allowed to print oops-related info.
 817 * This is a bit racy..
 818 */
 819bool oops_may_print(void)
 820{
 821	return pause_on_oops_flag == 0;
 822}
 823
 824/*
 825 * Called when the architecture enters its oops handler, before it prints
 826 * anything.  If this is the first CPU to oops, and it's oopsing the first
 827 * time then let it proceed.
 828 *
 829 * This is all enabled by the pause_on_oops kernel boot option.  We do all
 830 * this to ensure that oopses don't scroll off the screen.  It has the
 831 * side-effect of preventing later-oopsing CPUs from mucking up the display,
 832 * too.
 833 *
 834 * It turns out that the CPU which is allowed to print ends up pausing for
 835 * the right duration, whereas all the other CPUs pause for twice as long:
 836 * once in oops_enter(), once in oops_exit().
 837 */
 838void oops_enter(void)
 839{
 840	nbcon_cpu_emergency_enter();
 841	tracing_off();
 842	/* can't trust the integrity of the kernel anymore: */
 843	debug_locks_off();
 844	do_oops_enter_exit();
 845
 846	if (sysctl_oops_all_cpu_backtrace)
 847		trigger_all_cpu_backtrace();
 848}
 849
 850static void print_oops_end_marker(void)
 851{
 852	pr_warn("---[ end trace %016llx ]---\n", 0ULL);
 853}
 854
 855/*
 856 * Called when the architecture exits its oops handler, after printing
 857 * everything.
 858 */
 859void oops_exit(void)
 860{
 861	do_oops_enter_exit();
 862	print_oops_end_marker();
 863	nbcon_cpu_emergency_exit();
 864	kmsg_dump(KMSG_DUMP_OOPS);
 865}
 866
 867struct warn_args {
 868	const char *fmt;
 869	va_list args;
 870};
 871
 872void __warn(const char *file, int line, void *caller, unsigned taint,
 873	    struct pt_regs *regs, struct warn_args *args)
 874{
 875	nbcon_cpu_emergency_enter();
 876
 877	disable_trace_on_warning();
 878
 879	if (file) {
 880		pr_warn("WARNING: %s:%d at %pS, CPU#%d: %s/%d\n",
 881			file, line, caller,
 882			raw_smp_processor_id(), current->comm, current->pid);
 883	} else {
 884		pr_warn("WARNING: at %pS, CPU#%d: %s/%d\n",
 885			caller,
 886			raw_smp_processor_id(), current->comm, current->pid);
 887	}
 888
 889#pragma GCC diagnostic push
 890#ifndef __clang__
 891#pragma GCC diagnostic ignored "-Wsuggest-attribute=format"
 892#endif
 893	if (args)
 894		vprintk(args->fmt, args->args);
 895#pragma GCC diagnostic pop
 896
 897	print_modules();
 898
 899	if (regs)
 900		show_regs(regs);
 901
 902	check_panic_on_warn("kernel");
 903
 904	if (!regs)
 905		dump_stack();
 906
 907	print_irqtrace_events(current);
 908
 909	print_oops_end_marker();
 910	trace_error_report_end(ERROR_DETECTOR_WARN, (unsigned long)caller);
 911
 912	/* Just a warning, don't kill lockdep. */
 913	add_taint(taint, LOCKDEP_STILL_OK);
 914
 915	nbcon_cpu_emergency_exit();
 916}
 917
 918#ifdef CONFIG_BUG
 919#ifndef __WARN_FLAGS
 920void warn_slowpath_fmt(const char *file, int line, unsigned taint,
 921		       const char *fmt, ...)
 922{
 923	bool rcu = warn_rcu_enter();
 924	struct warn_args args;
 925
 926	pr_warn(CUT_HERE);
 927
 928	if (!fmt) {
 929		__warn(file, line, __builtin_return_address(0), taint,
 930		       NULL, NULL);
 931		warn_rcu_exit(rcu);
 932		return;
 933	}
 934
 935	args.fmt = fmt;
 936	va_start(args.args, fmt);
 937	__warn(file, line, __builtin_return_address(0), taint, NULL, &args);
 938	va_end(args.args);
 939	warn_rcu_exit(rcu);
 940}
 941EXPORT_SYMBOL(warn_slowpath_fmt);
 942#else
 943void __warn_printk(const char *fmt, ...)
 944{
 945	bool rcu = warn_rcu_enter();
 946	va_list args;
 947
 948	pr_warn(CUT_HERE);
 949
 950	va_start(args, fmt);
 951	vprintk(fmt, args);
 952	va_end(args);
 953	warn_rcu_exit(rcu);
 954}
 955EXPORT_SYMBOL(__warn_printk);
 956#endif
 957
 958/* Support resetting WARN*_ONCE state */
 959
 960static int clear_warn_once_set(void *data, u64 val)
 961{
 962	generic_bug_clear_once();
 963	memset(__start_once, 0, __end_once - __start_once);
 964	return 0;
 965}
 966
 967DEFINE_DEBUGFS_ATTRIBUTE(clear_warn_once_fops, NULL, clear_warn_once_set,
 968			 "%lld\n");
 969
 970static __init int register_warn_debugfs(void)
 971{
 972	/* Don't care about failure */
 973	debugfs_create_file_unsafe("clear_warn_once", 0200, NULL, NULL,
 974				   &clear_warn_once_fops);
 975	return 0;
 976}
 977
 978device_initcall(register_warn_debugfs);
 979#endif
 980
 981#ifdef CONFIG_STACKPROTECTOR
 982
 983/*
 984 * Called when gcc's -fstack-protector feature is used, and
 985 * gcc detects corruption of the on-stack canary value
 986 */
 987__visible noinstr void __stack_chk_fail(void)
 988{
 989	unsigned long flags;
 990
 991	instrumentation_begin();
 992	flags = user_access_save();
 993
 994	panic("stack-protector: Kernel stack is corrupted in: %pB",
 995		__builtin_return_address(0));
 996
 997	user_access_restore(flags);
 998	instrumentation_end();
 999}
1000EXPORT_SYMBOL(__stack_chk_fail);
1001
1002#endif
1003
1004core_param(panic, panic_timeout, int, 0644);
1005core_param(pause_on_oops, pause_on_oops, int, 0644);
1006core_param(panic_on_warn, panic_on_warn, int, 0644);
1007core_param(crash_kexec_post_notifiers, crash_kexec_post_notifiers, bool, 0644);
1008core_param(panic_console_replay, panic_console_replay, bool, 0644);
1009
1010static int panic_print_set(const char *val, const struct kernel_param *kp)
1011{
1012	panic_print_deprecated();
1013	return  param_set_ulong(val, kp);
1014}
1015
1016static int panic_print_get(char *val, const struct kernel_param *kp)
1017{
1018	return  param_get_ulong(val, kp);
1019}
1020
1021static const struct kernel_param_ops panic_print_ops = {
1022	.set	= panic_print_set,
1023	.get	= panic_print_get,
1024};
1025__core_param_cb(panic_print, &panic_print_ops, &panic_print, 0644);
1026
1027static int __init oops_setup(char *s)
1028{
1029	if (!s)
1030		return -EINVAL;
1031	if (!strcmp(s, "panic"))
1032		panic_on_oops = 1;
1033	return 0;
1034}
1035early_param("oops", oops_setup);
1036
1037static int __init panic_on_taint_setup(char *s)
1038{
1039	char *taint_str;
1040
1041	if (!s)
1042		return -EINVAL;
1043
1044	taint_str = strsep(&s, ",");
1045	if (kstrtoul(taint_str, 16, &panic_on_taint))
1046		return -EINVAL;
1047
1048	/* make sure panic_on_taint doesn't hold out-of-range TAINT flags */
1049	panic_on_taint &= TAINT_FLAGS_MAX;
1050
1051	if (!panic_on_taint)
1052		return -EINVAL;
1053
1054	if (s && !strcmp(s, "nousertaint"))
1055		panic_on_taint_nousertaint = true;
1056
1057	pr_info("panic_on_taint: bitmask=0x%lx nousertaint_mode=%s\n",
1058		panic_on_taint, str_enabled_disabled(panic_on_taint_nousertaint));
1059
1060	return 0;
1061}
1062early_param("panic_on_taint", panic_on_taint_setup);