projects/linux·blob master · refs

1468 lines · 36.1 KiB

1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 * linux/kernel/reboot.c
4 *
5 * Copyright (C) 2013 Linus Torvalds
6 */
7
8#define pr_fmt(fmt) "reboot: " fmt
9
10#include <linux/atomic.h>
11#include <linux/console.h>
12#include <linux/ctype.h>
13#include <linux/export.h>
14#include <linux/kexec.h>
15#include <linux/kmod.h>
16#include <linux/kmsg_dump.h>
17#include <linux/rcupdate.h>
18#include <linux/reboot.h>
19#include <linux/sched/signal.h>
20#include <linux/suspend.h>
21#include <linux/syscalls.h>
22#include <linux/syscore_ops.h>
23#include <linux/uaccess.h>
24
25/*
26 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
27 */
28
29static int C_A_D = 1;
30struct pid __rcu *cad_pid;
31
32#if defined(CONFIG_ARM)
33#define DEFAULT_REBOOT_MODE = REBOOT_HARD
34#else
35#define DEFAULT_REBOOT_MODE
36#endif
37enum reboot_mode reboot_mode DEFAULT_REBOOT_MODE;
38EXPORT_SYMBOL_GPL(reboot_mode);
39enum reboot_mode panic_reboot_mode = REBOOT_UNDEFINED;
40
41static enum hw_protection_action hw_protection_action = HWPROT_ACT_SHUTDOWN;
42
43/*
44 * This variable is used privately to keep track of whether or not
45 * reboot_type is still set to its default value (i.e., reboot= hasn't
46 * been set on the command line). This is needed so that we can
47 * suppress DMI scanning for reboot quirks. Without it, it's
48 * impossible to override a faulty reboot quirk without recompiling.
49 */
50int reboot_default = 1;
51int reboot_cpu;
52enum reboot_type reboot_type = BOOT_ACPI;
53int reboot_force;
54
55struct sys_off_handler {
56 struct notifier_block nb;
57 int (*sys_off_cb)(struct sys_off_data *data);
58 void *cb_data;
59 enum sys_off_mode mode;
60 bool blocking;
61 void *list;
62 struct device *dev;
63};
64
65/*
66 * This variable is used to indicate if a halt was initiated instead of a
67 * reboot when the reboot call was invoked with LINUX_REBOOT_CMD_POWER_OFF, but
68 * the system cannot be powered off. This allowes kernel_halt() to notify users
69 * of that.
70 */
71static bool poweroff_fallback_to_halt;
72
73/*
74 * Temporary stub that prevents linkage failure while we're in process
75 * of removing all uses of legacy pm_power_off() around the kernel.
76 */
77void __weak (*pm_power_off)(void);
78
79/*
80 * Notifier list for kernel code which wants to be called
81 * at shutdown. This is used to stop any idling DMA operations
82 * and the like.
83 */
84static BLOCKING_NOTIFIER_HEAD(reboot_notifier_list);
85
86/**
87 * emergency_restart - reboot the system
88 *
89 * Without shutting down any hardware or taking any locks
90 * reboot the system. This is called when we know we are in
91 * trouble so this is our best effort to reboot. This is
92 * safe to call in interrupt context.
93 */
94void emergency_restart(void)
95{
96 kmsg_dump(KMSG_DUMP_EMERG);
97 system_state = SYSTEM_RESTART;
98
99 /*
100 * NBCON kthreads won't be flushed and stopped properly in
101 * the emergency restart. Move them to the atomic mode.
102 */
103 nbcon_cpu_emergency_enter();
104 printk_trigger_flush();
105
106 machine_emergency_restart();
107}
108EXPORT_SYMBOL_GPL(emergency_restart);
109
110void kernel_restart_prepare(char *cmd)
111{
112 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
113 system_state = SYSTEM_RESTART;
114 usermodehelper_disable();
115 device_shutdown();
116}
117
118/**
119 * register_reboot_notifier - Register function to be called at reboot time
120 * @nb: Info about notifier function to be called
121 *
122 * Registers a function with the list of functions
123 * to be called at reboot time.
124 *
125 * Currently always returns zero, as blocking_notifier_chain_register()
126 * always returns zero.
127 */
128int register_reboot_notifier(struct notifier_block *nb)
129{
130 return blocking_notifier_chain_register(&reboot_notifier_list, nb);
131}
132EXPORT_SYMBOL(register_reboot_notifier);
133
134/**
135 * unregister_reboot_notifier - Unregister previously registered reboot notifier
136 * @nb: Hook to be unregistered
137 *
138 * Unregisters a previously registered reboot
139 * notifier function.
140 *
141 * Returns zero on success, or %-ENOENT on failure.
142 */
143int unregister_reboot_notifier(struct notifier_block *nb)
144{
145 return blocking_notifier_chain_unregister(&reboot_notifier_list, nb);
146}
147EXPORT_SYMBOL(unregister_reboot_notifier);
148
149static void devm_unregister_reboot_notifier(struct device *dev, void *res)
150{
151 WARN_ON(unregister_reboot_notifier(*(struct notifier_block **)res));
152}
153
154int devm_register_reboot_notifier(struct device *dev, struct notifier_block *nb)
155{
156 struct notifier_block **rcnb;
157 int ret;
158
159 rcnb = devres_alloc(devm_unregister_reboot_notifier,
160 sizeof(*rcnb), GFP_KERNEL);
161 if (!rcnb)
162 return -ENOMEM;
163
164 ret = register_reboot_notifier(nb);
165 if (!ret) {
166 *rcnb = nb;
167 devres_add(dev, rcnb);
168 } else {
169 devres_free(rcnb);
170 }
171
172 return ret;
173}
174EXPORT_SYMBOL(devm_register_reboot_notifier);
175
176/*
177 * Notifier list for kernel code which wants to be called
178 * to restart the system.
179 */
180static ATOMIC_NOTIFIER_HEAD(restart_handler_list);
181
182/**
183 * register_restart_handler - Register function to be called to reset
184 * the system
185 * @nb: Info about handler function to be called
186 * @nb->priority: Handler priority. Handlers should follow the
187 * following guidelines for setting priorities.
188 * 0: Restart handler of last resort,
189 * with limited restart capabilities
190 * 128: Default restart handler; use if no other
191 * restart handler is expected to be available,
192 * and/or if restart functionality is
193 * sufficient to restart the entire system
194 * 255: Highest priority restart handler, will
195 * preempt all other restart handlers
196 *
197 * Registers a function with code to be called to restart the
198 * system.
199 *
200 * Registered functions will be called from machine_restart as last
201 * step of the restart sequence (if the architecture specific
202 * machine_restart function calls do_kernel_restart - see below
203 * for details).
204 * Registered functions are expected to restart the system immediately.
205 * If more than one function is registered, the restart handler priority
206 * selects which function will be called first.
207 *
208 * Restart handlers are expected to be registered from non-architecture
209 * code, typically from drivers. A typical use case would be a system
210 * where restart functionality is provided through a watchdog. Multiple
211 * restart handlers may exist; for example, one restart handler might
212 * restart the entire system, while another only restarts the CPU.
213 * In such cases, the restart handler which only restarts part of the
214 * hardware is expected to register with low priority to ensure that
215 * it only runs if no other means to restart the system is available.
216 *
217 * Currently always returns zero, as atomic_notifier_chain_register()
218 * always returns zero.
219 */
220int register_restart_handler(struct notifier_block *nb)
221{
222 return atomic_notifier_chain_register(&restart_handler_list, nb);
223}
224EXPORT_SYMBOL(register_restart_handler);
225
226/**
227 * unregister_restart_handler - Unregister previously registered
228 * restart handler
229 * @nb: Hook to be unregistered
230 *
231 * Unregisters a previously registered restart handler function.
232 *
233 * Returns zero on success, or %-ENOENT on failure.
234 */
235int unregister_restart_handler(struct notifier_block *nb)
236{
237 return atomic_notifier_chain_unregister(&restart_handler_list, nb);
238}
239EXPORT_SYMBOL(unregister_restart_handler);
240
241/**
242 * do_kernel_restart - Execute kernel restart handler call chain
243 *
244 * @cmd: pointer to buffer containing command to execute for restart
245 * or %NULL
246 *
247 * Calls functions registered with register_restart_handler.
248 *
249 * Expected to be called from machine_restart as last step of the restart
250 * sequence.
251 *
252 * Restarts the system immediately if a restart handler function has been
253 * registered. Otherwise does nothing.
254 */
255void do_kernel_restart(char *cmd)
256{
257 atomic_notifier_call_chain(&restart_handler_list, reboot_mode, cmd);
258}
259
260void migrate_to_reboot_cpu(void)
261{
262 /* The boot cpu is always logical cpu 0 */
263 int cpu = reboot_cpu;
264
265 cpu_hotplug_disable();
266
267 /* Make certain the cpu I'm about to reboot on is online */
268 if (!cpu_online(cpu))
269 cpu = cpumask_first(cpu_online_mask);
270
271 /* Prevent races with other tasks migrating this task */
272 current->flags |= PF_NO_SETAFFINITY;
273
274 /* Make certain I only run on the appropriate processor */
275 set_cpus_allowed_ptr(current, cpumask_of(cpu));
276}
277
278/*
279 * Notifier list for kernel code which wants to be called
280 * to prepare system for restart.
281 */
282static BLOCKING_NOTIFIER_HEAD(restart_prep_handler_list);
283
284static void do_kernel_restart_prepare(void)
285{
286 blocking_notifier_call_chain(&restart_prep_handler_list, 0, NULL);
287}
288
289/**
290 * kernel_restart - reboot the system
291 * @cmd: pointer to buffer containing command to execute for restart
292 * or %NULL
293 *
294 * Shutdown everything and perform a clean reboot.
295 * This is not safe to call in interrupt context.
296 */
297void kernel_restart(char *cmd)
298{
299 kernel_restart_prepare(cmd);
300 do_kernel_restart_prepare();
301 migrate_to_reboot_cpu();
302 syscore_shutdown();
303 if (!cmd)
304 pr_emerg("Restarting system\n");
305 else
306 pr_emerg("Restarting system with command '%s'\n", cmd);
307 kmsg_dump(KMSG_DUMP_SHUTDOWN);
308 machine_restart(cmd);
309}
310EXPORT_SYMBOL_GPL(kernel_restart);
311
312static void kernel_shutdown_prepare(enum system_states state)
313{
314 blocking_notifier_call_chain(&reboot_notifier_list,
315 (state == SYSTEM_HALT) ? SYS_HALT : SYS_POWER_OFF, NULL);
316 system_state = state;
317 usermodehelper_disable();
318 device_shutdown();
319}
320/**
321 * kernel_halt - halt the system
322 *
323 * Shutdown everything and perform a clean system halt.
324 */
325void kernel_halt(void)
326{
327 kernel_shutdown_prepare(SYSTEM_HALT);
328 migrate_to_reboot_cpu();
329 syscore_shutdown();
330 if (poweroff_fallback_to_halt)
331 pr_emerg("Power off not available: System halted instead\n");
332 else
333 pr_emerg("System halted\n");
334 kmsg_dump(KMSG_DUMP_SHUTDOWN);
335 machine_halt();
336}
337EXPORT_SYMBOL_GPL(kernel_halt);
338
339/*
340 * Notifier list for kernel code which wants to be called
341 * to prepare system for power off.
342 */
343static BLOCKING_NOTIFIER_HEAD(power_off_prep_handler_list);
344
345/*
346 * Notifier list for kernel code which wants to be called
347 * to power off system.
348 */
349static ATOMIC_NOTIFIER_HEAD(power_off_handler_list);
350
351static int sys_off_notify(struct notifier_block *nb,
352 unsigned long mode, void *cmd)
353{
354 struct sys_off_handler *handler;
355 struct sys_off_data data = {};
356
357 handler = container_of(nb, struct sys_off_handler, nb);
358 data.cb_data = handler->cb_data;
359 data.mode = mode;
360 data.cmd = cmd;
361 data.dev = handler->dev;
362
363 return handler->sys_off_cb(&data);
364}
365
366static struct sys_off_handler platform_sys_off_handler;
367
368static struct sys_off_handler *alloc_sys_off_handler(int priority)
369{
370 struct sys_off_handler *handler;
371 gfp_t flags;
372
373 /*
374 * Platforms like m68k can't allocate sys_off handler dynamically
375 * at the early boot time because memory allocator isn't available yet.
376 */
377 if (priority == SYS_OFF_PRIO_PLATFORM) {
378 handler = &platform_sys_off_handler;
379 if (handler->cb_data)
380 return ERR_PTR(-EBUSY);
381 } else {
382 if (system_state > SYSTEM_RUNNING)
383 flags = GFP_ATOMIC;
384 else
385 flags = GFP_KERNEL;
386
387 handler = kzalloc_obj(*handler, flags);
388 if (!handler)
389 return ERR_PTR(-ENOMEM);
390 }
391
392 return handler;
393}
394
395static void free_sys_off_handler(struct sys_off_handler *handler)
396{
397 if (handler == &platform_sys_off_handler)
398 memset(handler, 0, sizeof(*handler));
399 else
400 kfree(handler);
401}
402
403/**
404 * register_sys_off_handler - Register sys-off handler
405 * @mode: Sys-off mode
406 * @priority: Handler priority
407 * @callback: Callback function
408 * @cb_data: Callback argument
409 *
410 * Registers system power-off or restart handler that will be invoked
411 * at the step corresponding to the given sys-off mode. Handler's callback
412 * should return NOTIFY_DONE to permit execution of the next handler in
413 * the call chain or NOTIFY_STOP to break the chain (in error case for
414 * example).
415 *
416 * Multiple handlers can be registered at the default priority level.
417 *
418 * Only one handler can be registered at the non-default priority level,
419 * otherwise ERR_PTR(-EBUSY) is returned.
420 *
421 * Returns a new instance of struct sys_off_handler on success, or
422 * an ERR_PTR()-encoded error code otherwise.
423 */
424struct sys_off_handler *
425register_sys_off_handler(enum sys_off_mode mode,
426 int priority,
427 int (*callback)(struct sys_off_data *data),
428 void *cb_data)
429{
430 struct sys_off_handler *handler;
431 int err;
432
433 handler = alloc_sys_off_handler(priority);
434 if (IS_ERR(handler))
435 return handler;
436
437 switch (mode) {
438 case SYS_OFF_MODE_POWER_OFF_PREPARE:
439 handler->list = &power_off_prep_handler_list;
440 handler->blocking = true;
441 break;
442
443 case SYS_OFF_MODE_POWER_OFF:
444 handler->list = &power_off_handler_list;
445 break;
446
447 case SYS_OFF_MODE_RESTART_PREPARE:
448 handler->list = &restart_prep_handler_list;
449 handler->blocking = true;
450 break;
451
452 case SYS_OFF_MODE_RESTART:
453 handler->list = &restart_handler_list;
454 break;
455
456 default:
457 free_sys_off_handler(handler);
458 return ERR_PTR(-EINVAL);
459 }
460
461 handler->nb.notifier_call = sys_off_notify;
462 handler->nb.priority = priority;
463 handler->sys_off_cb = callback;
464 handler->cb_data = cb_data;
465 handler->mode = mode;
466
467 if (handler->blocking) {
468 if (priority == SYS_OFF_PRIO_DEFAULT)
469 err = blocking_notifier_chain_register(handler->list,
470 &handler->nb);
471 else
472 err = blocking_notifier_chain_register_unique_prio(handler->list,
473 &handler->nb);
474 } else {
475 if (priority == SYS_OFF_PRIO_DEFAULT)
476 err = atomic_notifier_chain_register(handler->list,
477 &handler->nb);
478 else
479 err = atomic_notifier_chain_register_unique_prio(handler->list,
480 &handler->nb);
481 }
482
483 if (err) {
484 free_sys_off_handler(handler);
485 return ERR_PTR(err);
486 }
487
488 return handler;
489}
490EXPORT_SYMBOL_GPL(register_sys_off_handler);
491
492/**
493 * unregister_sys_off_handler - Unregister sys-off handler
494 * @handler: Sys-off handler
495 *
496 * Unregisters given sys-off handler.
497 */
498void unregister_sys_off_handler(struct sys_off_handler *handler)
499{
500 int err;
501
502 if (IS_ERR_OR_NULL(handler))
503 return;
504
505 if (handler->blocking)
506 err = blocking_notifier_chain_unregister(handler->list,
507 &handler->nb);
508 else
509 err = atomic_notifier_chain_unregister(handler->list,
510 &handler->nb);
511
512 /* sanity check, shall never happen */
513 WARN_ON(err);
514
515 free_sys_off_handler(handler);
516}
517EXPORT_SYMBOL_GPL(unregister_sys_off_handler);
518
519static void devm_unregister_sys_off_handler(void *data)
520{
521 struct sys_off_handler *handler = data;
522
523 unregister_sys_off_handler(handler);
524}
525
526/**
527 * devm_register_sys_off_handler - Register sys-off handler
528 * @dev: Device that registers handler
529 * @mode: Sys-off mode
530 * @priority: Handler priority
531 * @callback: Callback function
532 * @cb_data: Callback argument
533 *
534 * Registers resource-managed sys-off handler.
535 *
536 * Returns zero on success, or error code on failure.
537 */
538int devm_register_sys_off_handler(struct device *dev,
539 enum sys_off_mode mode,
540 int priority,
541 int (*callback)(struct sys_off_data *data),
542 void *cb_data)
543{
544 struct sys_off_handler *handler;
545
546 handler = register_sys_off_handler(mode, priority, callback, cb_data);
547 if (IS_ERR(handler))
548 return PTR_ERR(handler);
549 handler->dev = dev;
550
551 return devm_add_action_or_reset(dev, devm_unregister_sys_off_handler,
552 handler);
553}
554EXPORT_SYMBOL_GPL(devm_register_sys_off_handler);
555
556/**
557 * devm_register_power_off_handler - Register power-off handler
558 * @dev: Device that registers callback
559 * @callback: Callback function
560 * @cb_data: Callback's argument
561 *
562 * Registers resource-managed sys-off handler with a default priority
563 * and using power-off mode.
564 *
565 * Returns zero on success, or error code on failure.
566 */
567int devm_register_power_off_handler(struct device *dev,
568 int (*callback)(struct sys_off_data *data),
569 void *cb_data)
570{
571 return devm_register_sys_off_handler(dev,
572 SYS_OFF_MODE_POWER_OFF,
573 SYS_OFF_PRIO_DEFAULT,
574 callback, cb_data);
575}
576EXPORT_SYMBOL_GPL(devm_register_power_off_handler);
577
578/**
579 * devm_register_restart_handler - Register restart handler
580 * @dev: Device that registers callback
581 * @callback: Callback function
582 * @cb_data: Callback's argument
583 *
584 * Registers resource-managed sys-off handler with a default priority
585 * and using restart mode.
586 *
587 * Returns zero on success, or error code on failure.
588 */
589int devm_register_restart_handler(struct device *dev,
590 int (*callback)(struct sys_off_data *data),
591 void *cb_data)
592{
593 return devm_register_sys_off_handler(dev,
594 SYS_OFF_MODE_RESTART,
595 SYS_OFF_PRIO_DEFAULT,
596 callback, cb_data);
597}
598EXPORT_SYMBOL_GPL(devm_register_restart_handler);
599
600static struct sys_off_handler *platform_power_off_handler;
601
602static int platform_power_off_notify(struct sys_off_data *data)
603{
604 void (*platform_power_power_off_cb)(void) = data->cb_data;
605
606 platform_power_power_off_cb();
607
608 return NOTIFY_DONE;
609}
610
611/**
612 * register_platform_power_off - Register platform-level power-off callback
613 * @power_off: Power-off callback
614 *
615 * Registers power-off callback that will be called as last step
616 * of the power-off sequence. This callback is expected to be invoked
617 * for the last resort. Only one platform power-off callback is allowed
618 * to be registered at a time.
619 *
620 * Returns zero on success, or error code on failure.
621 */
622int register_platform_power_off(void (*power_off)(void))
623{
624 struct sys_off_handler *handler;
625
626 handler = register_sys_off_handler(SYS_OFF_MODE_POWER_OFF,
627 SYS_OFF_PRIO_PLATFORM,
628 platform_power_off_notify,
629 power_off);
630 if (IS_ERR(handler))
631 return PTR_ERR(handler);
632
633 platform_power_off_handler = handler;
634
635 return 0;
636}
637EXPORT_SYMBOL_GPL(register_platform_power_off);
638
639/**
640 * unregister_platform_power_off - Unregister platform-level power-off callback
641 * @power_off: Power-off callback
642 *
643 * Unregisters previously registered platform power-off callback.
644 */
645void unregister_platform_power_off(void (*power_off)(void))
646{
647 if (platform_power_off_handler &&
648 platform_power_off_handler->cb_data == power_off) {
649 unregister_sys_off_handler(platform_power_off_handler);
650 platform_power_off_handler = NULL;
651 }
652}
653EXPORT_SYMBOL_GPL(unregister_platform_power_off);
654
655static int legacy_pm_power_off(struct sys_off_data *data)
656{
657 if (pm_power_off)
658 pm_power_off();
659
660 return NOTIFY_DONE;
661}
662
663static void do_kernel_power_off_prepare(void)
664{
665 blocking_notifier_call_chain(&power_off_prep_handler_list, 0, NULL);
666}
667
668/**
669 * do_kernel_power_off - Execute kernel power-off handler call chain
670 *
671 * Expected to be called as last step of the power-off sequence.
672 *
673 * Powers off the system immediately if a power-off handler function has
674 * been registered. Otherwise does nothing.
675 */
676void do_kernel_power_off(void)
677{
678 struct sys_off_handler *sys_off = NULL;
679
680 /*
681 * Register sys-off handlers for legacy PM callback. This allows
682 * legacy PM callbacks temporary co-exist with the new sys-off API.
683 *
684 * TODO: Remove legacy handlers once all legacy PM users will be
685 * switched to the sys-off based APIs.
686 */
687 if (pm_power_off)
688 sys_off = register_sys_off_handler(SYS_OFF_MODE_POWER_OFF,
689 SYS_OFF_PRIO_DEFAULT,
690 legacy_pm_power_off, NULL);
691
692 atomic_notifier_call_chain(&power_off_handler_list, 0, NULL);
693
694 unregister_sys_off_handler(sys_off);
695}
696
697/**
698 * kernel_can_power_off - check whether system can be powered off
699 *
700 * Returns true if power-off handler is registered and system can be
701 * powered off, false otherwise.
702 */
703bool kernel_can_power_off(void)
704{
705 return !atomic_notifier_call_chain_is_empty(&power_off_handler_list) ||
706 pm_power_off;
707}
708EXPORT_SYMBOL_GPL(kernel_can_power_off);
709
710/**
711 * kernel_power_off - power_off the system
712 *
713 * Shutdown everything and perform a clean system power_off.
714 */
715void kernel_power_off(void)
716{
717 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
718 do_kernel_power_off_prepare();
719 migrate_to_reboot_cpu();
720 syscore_shutdown();
721 pr_emerg("Power down\n");
722 pr_flush(1000, true);
723 kmsg_dump(KMSG_DUMP_SHUTDOWN);
724 machine_power_off();
725}
726EXPORT_SYMBOL_GPL(kernel_power_off);
727
728DEFINE_MUTEX(system_transition_mutex);
729
730/*
731 * Reboot system call: for obvious reasons only root may call it,
732 * and even root needs to set up some magic numbers in the registers
733 * so that some mistake won't make this reboot the whole machine.
734 * You can also set the meaning of the ctrl-alt-del-key here.
735 *
736 * reboot doesn't sync: do that yourself before calling this.
737 */
738SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
739 void __user *, arg)
740{
741 struct pid_namespace *pid_ns = task_active_pid_ns(current);
742 char buffer[256];
743 int ret = 0;
744
745 /* We only trust the superuser with rebooting the system. */
746 if (!ns_capable(pid_ns->user_ns, CAP_SYS_BOOT))
747 return -EPERM;
748
749 /* For safety, we require "magic" arguments. */
750 if (magic1 != LINUX_REBOOT_MAGIC1 ||
751 (magic2 != LINUX_REBOOT_MAGIC2 &&
752 magic2 != LINUX_REBOOT_MAGIC2A &&
753 magic2 != LINUX_REBOOT_MAGIC2B &&
754 magic2 != LINUX_REBOOT_MAGIC2C))
755 return -EINVAL;
756
757 /*
758 * If pid namespaces are enabled and the current task is in a child
759 * pid_namespace, the command is handled by reboot_pid_ns() which will
760 * call do_exit().
761 */
762 ret = reboot_pid_ns(pid_ns, cmd);
763 if (ret)
764 return ret;
765
766 /* Instead of trying to make the power_off code look like
767 * halt when pm_power_off is not set do it the easy way.
768 */
769 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !kernel_can_power_off()) {
770 poweroff_fallback_to_halt = true;
771 cmd = LINUX_REBOOT_CMD_HALT;
772 }
773
774 mutex_lock(&system_transition_mutex);
775 switch (cmd) {
776 case LINUX_REBOOT_CMD_RESTART:
777 kernel_restart(NULL);
778 break;
779
780 case LINUX_REBOOT_CMD_CAD_ON:
781 C_A_D = 1;
782 break;
783
784 case LINUX_REBOOT_CMD_CAD_OFF:
785 C_A_D = 0;
786 break;
787
788 case LINUX_REBOOT_CMD_HALT:
789 kernel_halt();
790 do_exit(0);
791
792 case LINUX_REBOOT_CMD_POWER_OFF:
793 kernel_power_off();
794 do_exit(0);
795 break;
796
797 case LINUX_REBOOT_CMD_RESTART2:
798 ret = strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1);
799 if (ret < 0) {
800 ret = -EFAULT;
801 break;
802 }
803 buffer[sizeof(buffer) - 1] = '\0';
804
805 kernel_restart(buffer);
806 break;
807
808#ifdef CONFIG_KEXEC_CORE
809 case LINUX_REBOOT_CMD_KEXEC:
810 ret = kernel_kexec();
811 break;
812#endif
813
814#ifdef CONFIG_HIBERNATION
815 case LINUX_REBOOT_CMD_SW_SUSPEND:
816 ret = hibernate();
817 break;
818#endif
819
820 default:
821 ret = -EINVAL;
822 break;
823 }
824 mutex_unlock(&system_transition_mutex);
825 return ret;
826}
827
828static void deferred_cad(struct work_struct *dummy)
829{
830 kernel_restart(NULL);
831}
832
833/*
834 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
835 * As it's called within an interrupt, it may NOT sync: the only choice
836 * is whether to reboot at once, or just ignore the ctrl-alt-del.
837 */
838void ctrl_alt_del(void)
839{
840 static DECLARE_WORK(cad_work, deferred_cad);
841
842 if (C_A_D)
843 schedule_work(&cad_work);
844 else
845 kill_cad_pid(SIGINT, 1);
846}
847
848#define POWEROFF_CMD_PATH_LEN 256
849static char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
850static const char reboot_cmd[] = "/sbin/reboot";
851
852static int run_cmd(const char *cmd)
853{
854 char **argv;
855 static char *envp[] = {
856 "HOME=/",
857 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
858 NULL
859 };
860 int ret;
861 argv = argv_split(GFP_KERNEL, cmd, NULL);
862 if (argv) {
863 ret = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_EXEC);
864 argv_free(argv);
865 } else {
866 ret = -ENOMEM;
867 }
868
869 return ret;
870}
871
872static int __orderly_reboot(void)
873{
874 int ret;
875
876 ret = run_cmd(reboot_cmd);
877
878 if (ret) {
879 pr_warn("Failed to start orderly reboot: forcing the issue\n");
880 emergency_sync();
881 kernel_restart(NULL);
882 }
883
884 return ret;
885}
886
887static int __orderly_poweroff(bool force)
888{
889 int ret;
890
891 ret = run_cmd(poweroff_cmd);
892
893 if (ret && force) {
894 pr_warn("Failed to start orderly shutdown: forcing the issue\n");
895
896 /*
897 * I guess this should try to kick off some daemon to sync and
898 * poweroff asap. Or not even bother syncing if we're doing an
899 * emergency shutdown?
900 */
901 emergency_sync();
902 kernel_power_off();
903 }
904
905 return ret;
906}
907
908static bool poweroff_force;
909
910static void poweroff_work_func(struct work_struct *work)
911{
912 __orderly_poweroff(poweroff_force);
913}
914
915static DECLARE_WORK(poweroff_work, poweroff_work_func);
916
917/**
918 * orderly_poweroff - Trigger an orderly system poweroff
919 * @force: force poweroff if command execution fails
920 *
921 * This may be called from any context to trigger a system shutdown.
922 * If the orderly shutdown fails, it will force an immediate shutdown.
923 */
924void orderly_poweroff(bool force)
925{
926 if (force) /* do not override the pending "true" */
927 poweroff_force = true;
928 schedule_work(&poweroff_work);
929}
930EXPORT_SYMBOL_GPL(orderly_poweroff);
931
932static void reboot_work_func(struct work_struct *work)
933{
934 __orderly_reboot();
935}
936
937static DECLARE_WORK(reboot_work, reboot_work_func);
938
939/**
940 * orderly_reboot - Trigger an orderly system reboot
941 *
942 * This may be called from any context to trigger a system reboot.
943 * If the orderly reboot fails, it will force an immediate reboot.
944 */
945void orderly_reboot(void)
946{
947 schedule_work(&reboot_work);
948}
949EXPORT_SYMBOL_GPL(orderly_reboot);
950
951static const char *hw_protection_action_str(enum hw_protection_action action)
952{
953 switch (action) {
954 case HWPROT_ACT_SHUTDOWN:
955 return "shutdown";
956 case HWPROT_ACT_REBOOT:
957 return "reboot";
958 default:
959 return "undefined";
960 }
961}
962
963static enum hw_protection_action hw_failure_emergency_action;
964
965/**
966 * hw_failure_emergency_action_func - emergency action work after a known delay
967 * @work: work_struct associated with the emergency action function
968 *
969 * This function is called in very critical situations to force
970 * a kernel poweroff or reboot after a configurable timeout value.
971 */
972static void hw_failure_emergency_action_func(struct work_struct *work)
973{
974 const char *action_str = hw_protection_action_str(hw_failure_emergency_action);
975
976 pr_emerg("Hardware protection timed-out. Trying forced %s\n",
977 action_str);
978
979 /*
980 * We have reached here after the emergency action waiting period has
981 * expired. This means orderly_poweroff/reboot has not been able to
982 * shut off the system for some reason.
983 *
984 * Try to shut off the system immediately if possible
985 */
986
987 if (hw_failure_emergency_action == HWPROT_ACT_REBOOT)
988 kernel_restart(NULL);
989 else
990 kernel_power_off();
991
992 /*
993 * Worst of the worst case trigger emergency restart
994 */
995 pr_emerg("Hardware protection %s failed. Trying emergency restart\n",
996 action_str);
997 emergency_restart();
998}
999
1000static DECLARE_DELAYED_WORK(hw_failure_emergency_action_work,
1001 hw_failure_emergency_action_func);
1002
1003/**
1004 * hw_failure_emergency_schedule - Schedule an emergency system shutdown or reboot
1005 *
1006 * @action: The hardware protection action to be taken
1007 * @action_delay_ms: Time in milliseconds to elapse before triggering action
1008 *
1009 * This may be called from any critical situation to trigger a system shutdown
1010 * or reboot after a given period of time.
1011 * If time is negative this is not scheduled.
1012 */
1013static void hw_failure_emergency_schedule(enum hw_protection_action action,
1014 int action_delay_ms)
1015{
1016 if (action_delay_ms <= 0)
1017 return;
1018 hw_failure_emergency_action = action;
1019 schedule_delayed_work(&hw_failure_emergency_action_work,
1020 msecs_to_jiffies(action_delay_ms));
1021}
1022
1023/**
1024 * __hw_protection_trigger - Trigger an emergency system shutdown or reboot
1025 *
1026 * @reason: Reason of emergency shutdown or reboot to be printed.
1027 * @ms_until_forced: Time to wait for orderly shutdown or reboot before
1028 * triggering it. Negative value disables the forced
1029 * shutdown or reboot.
1030 * @action: The hardware protection action to be taken.
1031 *
1032 * Initiate an emergency system shutdown or reboot in order to protect
1033 * hardware from further damage. Usage examples include a thermal protection.
1034 * NOTE: The request is ignored if protection shutdown or reboot is already
1035 * pending even if the previous request has given a large timeout for forced
1036 * shutdown/reboot.
1037 */
1038void __hw_protection_trigger(const char *reason, int ms_until_forced,
1039 enum hw_protection_action action)
1040{
1041 static atomic_t allow_proceed = ATOMIC_INIT(1);
1042
1043 if (action == HWPROT_ACT_DEFAULT)
1044 action = hw_protection_action;
1045
1046 pr_emerg("HARDWARE PROTECTION %s (%s)\n",
1047 hw_protection_action_str(action), reason);
1048
1049 /* Shutdown should be initiated only once. */
1050 if (!atomic_dec_and_test(&allow_proceed))
1051 return;
1052
1053 /*
1054 * Queue a backup emergency shutdown in the event of
1055 * orderly_poweroff failure
1056 */
1057 hw_failure_emergency_schedule(action, ms_until_forced);
1058 if (action == HWPROT_ACT_REBOOT)
1059 orderly_reboot();
1060 else
1061 orderly_poweroff(true);
1062}
1063EXPORT_SYMBOL_GPL(__hw_protection_trigger);
1064
1065static bool hw_protection_action_parse(const char *str,
1066 enum hw_protection_action *action)
1067{
1068 if (sysfs_streq(str, "shutdown"))
1069 *action = HWPROT_ACT_SHUTDOWN;
1070 else if (sysfs_streq(str, "reboot"))
1071 *action = HWPROT_ACT_REBOOT;
1072 else
1073 return false;
1074
1075 return true;
1076}
1077
1078static int __init hw_protection_setup(char *str)
1079{
1080 hw_protection_action_parse(str, &hw_protection_action);
1081 return 1;
1082}
1083__setup("hw_protection=", hw_protection_setup);
1084
1085#ifdef CONFIG_SYSFS
1086static ssize_t hw_protection_show(struct kobject *kobj,
1087 struct kobj_attribute *attr, char *buf)
1088{
1089 return sysfs_emit(buf, "%s\n",
1090 hw_protection_action_str(hw_protection_action));
1091}
1092static ssize_t hw_protection_store(struct kobject *kobj,
1093 struct kobj_attribute *attr, const char *buf,
1094 size_t count)
1095{
1096 if (!capable(CAP_SYS_ADMIN))
1097 return -EPERM;
1098
1099 if (!hw_protection_action_parse(buf, &hw_protection_action))
1100 return -EINVAL;
1101
1102 return count;
1103}
1104static struct kobj_attribute hw_protection_attr = __ATTR_RW(hw_protection);
1105#endif
1106
1107static int __init reboot_setup(char *str)
1108{
1109 for (;;) {
1110 enum reboot_mode *mode;
1111
1112 /*
1113 * Having anything passed on the command line via
1114 * reboot= will cause us to disable DMI checking
1115 * below.
1116 */
1117 reboot_default = 0;
1118
1119 if (!strncmp(str, "panic_", 6)) {
1120 mode = &panic_reboot_mode;
1121 str += 6;
1122 } else {
1123 mode = &reboot_mode;
1124 }
1125
1126 switch (*str) {
1127 case 'w':
1128 *mode = REBOOT_WARM;
1129 break;
1130
1131 case 'c':
1132 *mode = REBOOT_COLD;
1133 break;
1134
1135 case 'h':
1136 *mode = REBOOT_HARD;
1137 break;
1138
1139 case 's':
1140 /*
1141 * reboot_cpu is s[mp]#### with #### being the processor
1142 * to be used for rebooting. Skip 's' or 'smp' prefix.
1143 */
1144 str += str[1] == 'm' && str[2] == 'p' ? 3 : 1;
1145
1146 if (isdigit(str[0])) {
1147 int cpu = simple_strtoul(str, NULL, 0);
1148
1149 if (cpu >= num_possible_cpus()) {
1150 pr_err("Ignoring the CPU number in reboot= option. "
1151 "CPU %d exceeds possible cpu number %d\n",
1152 cpu, num_possible_cpus());
1153 break;
1154 }
1155 reboot_cpu = cpu;
1156 } else
1157 *mode = REBOOT_SOFT;
1158 break;
1159
1160 case 'g':
1161 *mode = REBOOT_GPIO;
1162 break;
1163
1164 case 'b':
1165 case 'a':
1166 case 'k':
1167 case 't':
1168 case 'e':
1169 case 'p':
1170 reboot_type = *str;
1171 break;
1172
1173 case 'f':
1174 reboot_force = 1;
1175 break;
1176 }
1177
1178 str = strchr(str, ',');
1179 if (str)
1180 str++;
1181 else
1182 break;
1183 }
1184 return 1;
1185}
1186__setup("reboot=", reboot_setup);
1187
1188#ifdef CONFIG_SYSFS
1189
1190#define REBOOT_COLD_STR "cold"
1191#define REBOOT_WARM_STR "warm"
1192#define REBOOT_HARD_STR "hard"
1193#define REBOOT_SOFT_STR "soft"
1194#define REBOOT_GPIO_STR "gpio"
1195#define REBOOT_UNDEFINED_STR "undefined"
1196
1197#define BOOT_TRIPLE_STR "triple"
1198#define BOOT_KBD_STR "kbd"
1199#define BOOT_BIOS_STR "bios"
1200#define BOOT_ACPI_STR "acpi"
1201#define BOOT_EFI_STR "efi"
1202#define BOOT_PCI_STR "pci"
1203
1204static ssize_t mode_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
1205{
1206 const char *val;
1207
1208 switch (reboot_mode) {
1209 case REBOOT_COLD:
1210 val = REBOOT_COLD_STR;
1211 break;
1212 case REBOOT_WARM:
1213 val = REBOOT_WARM_STR;
1214 break;
1215 case REBOOT_HARD:
1216 val = REBOOT_HARD_STR;
1217 break;
1218 case REBOOT_SOFT:
1219 val = REBOOT_SOFT_STR;
1220 break;
1221 case REBOOT_GPIO:
1222 val = REBOOT_GPIO_STR;
1223 break;
1224 default:
1225 val = REBOOT_UNDEFINED_STR;
1226 }
1227
1228 return sysfs_emit(buf, "%s\n", val);
1229}
1230static ssize_t mode_store(struct kobject *kobj, struct kobj_attribute *attr,
1231 const char *buf, size_t count)
1232{
1233 if (!capable(CAP_SYS_BOOT))
1234 return -EPERM;
1235
1236 if (!strncmp(buf, REBOOT_COLD_STR, strlen(REBOOT_COLD_STR)))
1237 reboot_mode = REBOOT_COLD;
1238 else if (!strncmp(buf, REBOOT_WARM_STR, strlen(REBOOT_WARM_STR)))
1239 reboot_mode = REBOOT_WARM;
1240 else if (!strncmp(buf, REBOOT_HARD_STR, strlen(REBOOT_HARD_STR)))
1241 reboot_mode = REBOOT_HARD;
1242 else if (!strncmp(buf, REBOOT_SOFT_STR, strlen(REBOOT_SOFT_STR)))
1243 reboot_mode = REBOOT_SOFT;
1244 else if (!strncmp(buf, REBOOT_GPIO_STR, strlen(REBOOT_GPIO_STR)))
1245 reboot_mode = REBOOT_GPIO;
1246 else
1247 return -EINVAL;
1248
1249 reboot_default = 0;
1250
1251 return count;
1252}
1253static struct kobj_attribute reboot_mode_attr = __ATTR_RW(mode);
1254
1255#ifdef CONFIG_X86
1256static ssize_t force_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
1257{
1258 return sysfs_emit(buf, "%d\n", reboot_force);
1259}
1260static ssize_t force_store(struct kobject *kobj, struct kobj_attribute *attr,
1261 const char *buf, size_t count)
1262{
1263 bool res;
1264
1265 if (!capable(CAP_SYS_BOOT))
1266 return -EPERM;
1267
1268 if (kstrtobool(buf, &res))
1269 return -EINVAL;
1270
1271 reboot_default = 0;
1272 reboot_force = res;
1273
1274 return count;
1275}
1276static struct kobj_attribute reboot_force_attr = __ATTR_RW(force);
1277
1278static ssize_t type_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
1279{
1280 const char *val;
1281
1282 switch (reboot_type) {
1283 case BOOT_TRIPLE:
1284 val = BOOT_TRIPLE_STR;
1285 break;
1286 case BOOT_KBD:
1287 val = BOOT_KBD_STR;
1288 break;
1289 case BOOT_BIOS:
1290 val = BOOT_BIOS_STR;
1291 break;
1292 case BOOT_ACPI:
1293 val = BOOT_ACPI_STR;
1294 break;
1295 case BOOT_EFI:
1296 val = BOOT_EFI_STR;
1297 break;
1298 case BOOT_CF9_FORCE:
1299 val = BOOT_PCI_STR;
1300 break;
1301 default:
1302 val = REBOOT_UNDEFINED_STR;
1303 }
1304
1305 return sysfs_emit(buf, "%s\n", val);
1306}
1307static ssize_t type_store(struct kobject *kobj, struct kobj_attribute *attr,
1308 const char *buf, size_t count)
1309{
1310 if (!capable(CAP_SYS_BOOT))
1311 return -EPERM;
1312
1313 if (!strncmp(buf, BOOT_TRIPLE_STR, strlen(BOOT_TRIPLE_STR)))
1314 reboot_type = BOOT_TRIPLE;
1315 else if (!strncmp(buf, BOOT_KBD_STR, strlen(BOOT_KBD_STR)))
1316 reboot_type = BOOT_KBD;
1317 else if (!strncmp(buf, BOOT_BIOS_STR, strlen(BOOT_BIOS_STR)))
1318 reboot_type = BOOT_BIOS;
1319 else if (!strncmp(buf, BOOT_ACPI_STR, strlen(BOOT_ACPI_STR)))
1320 reboot_type = BOOT_ACPI;
1321 else if (!strncmp(buf, BOOT_EFI_STR, strlen(BOOT_EFI_STR)))
1322 reboot_type = BOOT_EFI;
1323 else if (!strncmp(buf, BOOT_PCI_STR, strlen(BOOT_PCI_STR)))
1324 reboot_type = BOOT_CF9_FORCE;
1325 else
1326 return -EINVAL;
1327
1328 reboot_default = 0;
1329
1330 return count;
1331}
1332static struct kobj_attribute reboot_type_attr = __ATTR_RW(type);
1333#endif
1334
1335#ifdef CONFIG_SMP
1336static ssize_t cpu_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
1337{
1338 return sysfs_emit(buf, "%d\n", reboot_cpu);
1339}
1340static ssize_t cpu_store(struct kobject *kobj, struct kobj_attribute *attr,
1341 const char *buf, size_t count)
1342{
1343 unsigned int cpunum;
1344 int rc;
1345
1346 if (!capable(CAP_SYS_BOOT))
1347 return -EPERM;
1348
1349 rc = kstrtouint(buf, 0, &cpunum);
1350
1351 if (rc)
1352 return rc;
1353
1354 if (cpunum >= num_possible_cpus())
1355 return -ERANGE;
1356
1357 reboot_default = 0;
1358 reboot_cpu = cpunum;
1359
1360 return count;
1361}
1362static struct kobj_attribute reboot_cpu_attr = __ATTR_RW(cpu);
1363#endif
1364
1365static struct attribute *reboot_attrs[] = {
1366 &hw_protection_attr.attr,
1367 &reboot_mode_attr.attr,
1368#ifdef CONFIG_X86
1369 &reboot_force_attr.attr,
1370 &reboot_type_attr.attr,
1371#endif
1372#ifdef CONFIG_SMP
1373 &reboot_cpu_attr.attr,
1374#endif
1375 NULL,
1376};
1377
1378#ifdef CONFIG_SYSCTL
1379static int proc_do_cad_pid(const struct ctl_table *table, int write, void *buffer,
1380 size_t *lenp, loff_t *ppos)
1381{
1382 struct ctl_table tmp_table = *table;
1383 struct pid *new_pid;
1384 struct pid *old_pid;
1385 pid_t tmp_pid;
1386 int r;
1387
1388 rcu_read_lock();
1389 tmp_pid = pid_vnr(rcu_dereference(cad_pid));
1390 rcu_read_unlock();
1391
1392 tmp_table.data = &tmp_pid;
1393
1394 r = proc_dointvec(&tmp_table, write, buffer, lenp, ppos);
1395 if (r || !write)
1396 return r;
1397
1398 new_pid = find_get_pid(tmp_pid);
1399 if (!new_pid)
1400 return -ESRCH;
1401
1402 old_pid = unrcu_pointer(xchg(&cad_pid, RCU_INITIALIZER(new_pid)));
1403 /*
1404 * Wait for cad_pid readers before put_pid(). We cannot use
1405 * call_rcu() here because free_pid() already owns pid->rcu.
1406 */
1407 synchronize_rcu();
1408 put_pid(old_pid);
1409 return 0;
1410}
1411
1412static const struct ctl_table kern_reboot_table[] = {
1413 {
1414 .procname = "poweroff_cmd",
1415 .data = &poweroff_cmd,
1416 .maxlen = POWEROFF_CMD_PATH_LEN,
1417 .mode = 0644,
1418 .proc_handler = proc_dostring,
1419 },
1420 {
1421 .procname = "ctrl-alt-del",
1422 .data = &C_A_D,
1423 .maxlen = sizeof(int),
1424 .mode = 0644,
1425 .proc_handler = proc_dointvec,
1426 },
1427 {
1428 .procname = "cad_pid",
1429 .maxlen = sizeof(int),
1430 .mode = 0600,
1431 .proc_handler = proc_do_cad_pid,
1432 },
1433};
1434
1435static void __init kernel_reboot_sysctls_init(void)
1436{
1437 register_sysctl_init("kernel", kern_reboot_table);
1438}
1439#else
1440#define kernel_reboot_sysctls_init() do { } while (0)
1441#endif /* CONFIG_SYSCTL */
1442
1443static const struct attribute_group reboot_attr_group = {
1444 .attrs = reboot_attrs,
1445};
1446
1447static int __init reboot_ksysfs_init(void)
1448{
1449 struct kobject *reboot_kobj;
1450 int ret;
1451
1452 reboot_kobj = kobject_create_and_add("reboot", kernel_kobj);
1453 if (!reboot_kobj)
1454 return -ENOMEM;
1455
1456 ret = sysfs_create_group(reboot_kobj, &reboot_attr_group);
1457 if (ret) {
1458 kobject_put(reboot_kobj);
1459 return ret;
1460 }
1461
1462 kernel_reboot_sysctls_init();
1463
1464 return 0;
1465}
1466late_initcall(reboot_ksysfs_init);
1467
1468#endif