xref: /linux/kernel/reboot.c (revision a9fc2304972b1db28b88af8203dffef23e1e92ba)
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/ctype.h>
12 #include <linux/export.h>
13 #include <linux/kexec.h>
14 #include <linux/kmod.h>
15 #include <linux/kmsg_dump.h>
16 #include <linux/reboot.h>
17 #include <linux/suspend.h>
18 #include <linux/syscalls.h>
19 #include <linux/syscore_ops.h>
20 #include <linux/uaccess.h>
21 
22 /*
23  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
24  */
25 
26 static int C_A_D = 1;
27 struct pid *cad_pid;
28 EXPORT_SYMBOL(cad_pid);
29 
30 #if defined(CONFIG_ARM)
31 #define DEFAULT_REBOOT_MODE		= REBOOT_HARD
32 #else
33 #define DEFAULT_REBOOT_MODE
34 #endif
35 enum reboot_mode reboot_mode DEFAULT_REBOOT_MODE;
36 EXPORT_SYMBOL_GPL(reboot_mode);
37 enum reboot_mode panic_reboot_mode = REBOOT_UNDEFINED;
38 
39 /*
40  * This variable is used privately to keep track of whether or not
41  * reboot_type is still set to its default value (i.e., reboot= hasn't
42  * been set on the command line).  This is needed so that we can
43  * suppress DMI scanning for reboot quirks.  Without it, it's
44  * impossible to override a faulty reboot quirk without recompiling.
45  */
46 int reboot_default = 1;
47 int reboot_cpu;
48 enum reboot_type reboot_type = BOOT_ACPI;
49 int reboot_force;
50 
51 struct sys_off_handler {
52 	struct notifier_block nb;
53 	int (*sys_off_cb)(struct sys_off_data *data);
54 	void *cb_data;
55 	enum sys_off_mode mode;
56 	bool blocking;
57 	void *list;
58 	struct device *dev;
59 };
60 
61 /*
62  * This variable is used to indicate if a halt was initiated instead of a
63  * reboot when the reboot call was invoked with LINUX_REBOOT_CMD_POWER_OFF, but
64  * the system cannot be powered off. This allowes kernel_halt() to notify users
65  * of that.
66  */
67 static bool poweroff_fallback_to_halt;
68 
69 /*
70  * Temporary stub that prevents linkage failure while we're in process
71  * of removing all uses of legacy pm_power_off() around the kernel.
72  */
73 void __weak (*pm_power_off)(void);
74 
75 /*
76  *	Notifier list for kernel code which wants to be called
77  *	at shutdown. This is used to stop any idling DMA operations
78  *	and the like.
79  */
80 static BLOCKING_NOTIFIER_HEAD(reboot_notifier_list);
81 
82 /**
83  *	emergency_restart - reboot the system
84  *
85  *	Without shutting down any hardware or taking any locks
86  *	reboot the system.  This is called when we know we are in
87  *	trouble so this is our best effort to reboot.  This is
88  *	safe to call in interrupt context.
89  */
90 void emergency_restart(void)
91 {
92 	kmsg_dump(KMSG_DUMP_EMERG);
93 	system_state = SYSTEM_RESTART;
94 	machine_emergency_restart();
95 }
96 EXPORT_SYMBOL_GPL(emergency_restart);
97 
98 void kernel_restart_prepare(char *cmd)
99 {
100 	blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
101 	system_state = SYSTEM_RESTART;
102 	usermodehelper_disable();
103 	device_shutdown();
104 }
105 
106 /**
107  *	register_reboot_notifier - Register function to be called at reboot time
108  *	@nb: Info about notifier function to be called
109  *
110  *	Registers a function with the list of functions
111  *	to be called at reboot time.
112  *
113  *	Currently always returns zero, as blocking_notifier_chain_register()
114  *	always returns zero.
115  */
116 int register_reboot_notifier(struct notifier_block *nb)
117 {
118 	return blocking_notifier_chain_register(&reboot_notifier_list, nb);
119 }
120 EXPORT_SYMBOL(register_reboot_notifier);
121 
122 /**
123  *	unregister_reboot_notifier - Unregister previously registered reboot notifier
124  *	@nb: Hook to be unregistered
125  *
126  *	Unregisters a previously registered reboot
127  *	notifier function.
128  *
129  *	Returns zero on success, or %-ENOENT on failure.
130  */
131 int unregister_reboot_notifier(struct notifier_block *nb)
132 {
133 	return blocking_notifier_chain_unregister(&reboot_notifier_list, nb);
134 }
135 EXPORT_SYMBOL(unregister_reboot_notifier);
136 
137 static void devm_unregister_reboot_notifier(struct device *dev, void *res)
138 {
139 	WARN_ON(unregister_reboot_notifier(*(struct notifier_block **)res));
140 }
141 
142 int devm_register_reboot_notifier(struct device *dev, struct notifier_block *nb)
143 {
144 	struct notifier_block **rcnb;
145 	int ret;
146 
147 	rcnb = devres_alloc(devm_unregister_reboot_notifier,
148 			    sizeof(*rcnb), GFP_KERNEL);
149 	if (!rcnb)
150 		return -ENOMEM;
151 
152 	ret = register_reboot_notifier(nb);
153 	if (!ret) {
154 		*rcnb = nb;
155 		devres_add(dev, rcnb);
156 	} else {
157 		devres_free(rcnb);
158 	}
159 
160 	return ret;
161 }
162 EXPORT_SYMBOL(devm_register_reboot_notifier);
163 
164 /*
165  *	Notifier list for kernel code which wants to be called
166  *	to restart the system.
167  */
168 static ATOMIC_NOTIFIER_HEAD(restart_handler_list);
169 
170 /**
171  *	register_restart_handler - Register function to be called to reset
172  *				   the system
173  *	@nb: Info about handler function to be called
174  *	@nb->priority:	Handler priority. Handlers should follow the
175  *			following guidelines for setting priorities.
176  *			0:	Restart handler of last resort,
177  *				with limited restart capabilities
178  *			128:	Default restart handler; use if no other
179  *				restart handler is expected to be available,
180  *				and/or if restart functionality is
181  *				sufficient to restart the entire system
182  *			255:	Highest priority restart handler, will
183  *				preempt all other restart handlers
184  *
185  *	Registers a function with code to be called to restart the
186  *	system.
187  *
188  *	Registered functions will be called from machine_restart as last
189  *	step of the restart sequence (if the architecture specific
190  *	machine_restart function calls do_kernel_restart - see below
191  *	for details).
192  *	Registered functions are expected to restart the system immediately.
193  *	If more than one function is registered, the restart handler priority
194  *	selects which function will be called first.
195  *
196  *	Restart handlers are expected to be registered from non-architecture
197  *	code, typically from drivers. A typical use case would be a system
198  *	where restart functionality is provided through a watchdog. Multiple
199  *	restart handlers may exist; for example, one restart handler might
200  *	restart the entire system, while another only restarts the CPU.
201  *	In such cases, the restart handler which only restarts part of the
202  *	hardware is expected to register with low priority to ensure that
203  *	it only runs if no other means to restart the system is available.
204  *
205  *	Currently always returns zero, as atomic_notifier_chain_register()
206  *	always returns zero.
207  */
208 int register_restart_handler(struct notifier_block *nb)
209 {
210 	return atomic_notifier_chain_register(&restart_handler_list, nb);
211 }
212 EXPORT_SYMBOL(register_restart_handler);
213 
214 /**
215  *	unregister_restart_handler - Unregister previously registered
216  *				     restart handler
217  *	@nb: Hook to be unregistered
218  *
219  *	Unregisters a previously registered restart handler function.
220  *
221  *	Returns zero on success, or %-ENOENT on failure.
222  */
223 int unregister_restart_handler(struct notifier_block *nb)
224 {
225 	return atomic_notifier_chain_unregister(&restart_handler_list, nb);
226 }
227 EXPORT_SYMBOL(unregister_restart_handler);
228 
229 /**
230  *	do_kernel_restart - Execute kernel restart handler call chain
231  *
232  *	Calls functions registered with register_restart_handler.
233  *
234  *	Expected to be called from machine_restart as last step of the restart
235  *	sequence.
236  *
237  *	Restarts the system immediately if a restart handler function has been
238  *	registered. Otherwise does nothing.
239  */
240 void do_kernel_restart(char *cmd)
241 {
242 	atomic_notifier_call_chain(&restart_handler_list, reboot_mode, cmd);
243 }
244 
245 void migrate_to_reboot_cpu(void)
246 {
247 	/* The boot cpu is always logical cpu 0 */
248 	int cpu = reboot_cpu;
249 
250 	cpu_hotplug_disable();
251 
252 	/* Make certain the cpu I'm about to reboot on is online */
253 	if (!cpu_online(cpu))
254 		cpu = cpumask_first(cpu_online_mask);
255 
256 	/* Prevent races with other tasks migrating this task */
257 	current->flags |= PF_NO_SETAFFINITY;
258 
259 	/* Make certain I only run on the appropriate processor */
260 	set_cpus_allowed_ptr(current, cpumask_of(cpu));
261 }
262 
263 /*
264  *	Notifier list for kernel code which wants to be called
265  *	to prepare system for restart.
266  */
267 static BLOCKING_NOTIFIER_HEAD(restart_prep_handler_list);
268 
269 static void do_kernel_restart_prepare(void)
270 {
271 	blocking_notifier_call_chain(&restart_prep_handler_list, 0, NULL);
272 }
273 
274 /**
275  *	kernel_restart - reboot the system
276  *	@cmd: pointer to buffer containing command to execute for restart
277  *		or %NULL
278  *
279  *	Shutdown everything and perform a clean reboot.
280  *	This is not safe to call in interrupt context.
281  */
282 void kernel_restart(char *cmd)
283 {
284 	kernel_restart_prepare(cmd);
285 	do_kernel_restart_prepare();
286 	migrate_to_reboot_cpu();
287 	syscore_shutdown();
288 	if (!cmd)
289 		pr_emerg("Restarting system\n");
290 	else
291 		pr_emerg("Restarting system with command '%s'\n", cmd);
292 	kmsg_dump(KMSG_DUMP_SHUTDOWN);
293 	machine_restart(cmd);
294 }
295 EXPORT_SYMBOL_GPL(kernel_restart);
296 
297 static void kernel_shutdown_prepare(enum system_states state)
298 {
299 	blocking_notifier_call_chain(&reboot_notifier_list,
300 		(state == SYSTEM_HALT) ? SYS_HALT : SYS_POWER_OFF, NULL);
301 	system_state = state;
302 	usermodehelper_disable();
303 	device_shutdown();
304 }
305 /**
306  *	kernel_halt - halt the system
307  *
308  *	Shutdown everything and perform a clean system halt.
309  */
310 void kernel_halt(void)
311 {
312 	kernel_shutdown_prepare(SYSTEM_HALT);
313 	migrate_to_reboot_cpu();
314 	syscore_shutdown();
315 	if (poweroff_fallback_to_halt)
316 		pr_emerg("Power off not available: System halted instead\n");
317 	else
318 		pr_emerg("System halted\n");
319 	kmsg_dump(KMSG_DUMP_SHUTDOWN);
320 	machine_halt();
321 }
322 EXPORT_SYMBOL_GPL(kernel_halt);
323 
324 /*
325  *	Notifier list for kernel code which wants to be called
326  *	to prepare system for power off.
327  */
328 static BLOCKING_NOTIFIER_HEAD(power_off_prep_handler_list);
329 
330 /*
331  *	Notifier list for kernel code which wants to be called
332  *	to power off system.
333  */
334 static ATOMIC_NOTIFIER_HEAD(power_off_handler_list);
335 
336 static int sys_off_notify(struct notifier_block *nb,
337 			  unsigned long mode, void *cmd)
338 {
339 	struct sys_off_handler *handler;
340 	struct sys_off_data data = {};
341 
342 	handler = container_of(nb, struct sys_off_handler, nb);
343 	data.cb_data = handler->cb_data;
344 	data.mode = mode;
345 	data.cmd = cmd;
346 	data.dev = handler->dev;
347 
348 	return handler->sys_off_cb(&data);
349 }
350 
351 static struct sys_off_handler platform_sys_off_handler;
352 
353 static struct sys_off_handler *alloc_sys_off_handler(int priority)
354 {
355 	struct sys_off_handler *handler;
356 	gfp_t flags;
357 
358 	/*
359 	 * Platforms like m68k can't allocate sys_off handler dynamically
360 	 * at the early boot time because memory allocator isn't available yet.
361 	 */
362 	if (priority == SYS_OFF_PRIO_PLATFORM) {
363 		handler = &platform_sys_off_handler;
364 		if (handler->cb_data)
365 			return ERR_PTR(-EBUSY);
366 	} else {
367 		if (system_state > SYSTEM_RUNNING)
368 			flags = GFP_ATOMIC;
369 		else
370 			flags = GFP_KERNEL;
371 
372 		handler = kzalloc(sizeof(*handler), flags);
373 		if (!handler)
374 			return ERR_PTR(-ENOMEM);
375 	}
376 
377 	return handler;
378 }
379 
380 static void free_sys_off_handler(struct sys_off_handler *handler)
381 {
382 	if (handler == &platform_sys_off_handler)
383 		memset(handler, 0, sizeof(*handler));
384 	else
385 		kfree(handler);
386 }
387 
388 /**
389  *	register_sys_off_handler - Register sys-off handler
390  *	@mode: Sys-off mode
391  *	@priority: Handler priority
392  *	@callback: Callback function
393  *	@cb_data: Callback argument
394  *
395  *	Registers system power-off or restart handler that will be invoked
396  *	at the step corresponding to the given sys-off mode. Handler's callback
397  *	should return NOTIFY_DONE to permit execution of the next handler in
398  *	the call chain or NOTIFY_STOP to break the chain (in error case for
399  *	example).
400  *
401  *	Multiple handlers can be registered at the default priority level.
402  *
403  *	Only one handler can be registered at the non-default priority level,
404  *	otherwise ERR_PTR(-EBUSY) is returned.
405  *
406  *	Returns a new instance of struct sys_off_handler on success, or
407  *	an ERR_PTR()-encoded error code otherwise.
408  */
409 struct sys_off_handler *
410 register_sys_off_handler(enum sys_off_mode mode,
411 			 int priority,
412 			 int (*callback)(struct sys_off_data *data),
413 			 void *cb_data)
414 {
415 	struct sys_off_handler *handler;
416 	int err;
417 
418 	handler = alloc_sys_off_handler(priority);
419 	if (IS_ERR(handler))
420 		return handler;
421 
422 	switch (mode) {
423 	case SYS_OFF_MODE_POWER_OFF_PREPARE:
424 		handler->list = &power_off_prep_handler_list;
425 		handler->blocking = true;
426 		break;
427 
428 	case SYS_OFF_MODE_POWER_OFF:
429 		handler->list = &power_off_handler_list;
430 		break;
431 
432 	case SYS_OFF_MODE_RESTART_PREPARE:
433 		handler->list = &restart_prep_handler_list;
434 		handler->blocking = true;
435 		break;
436 
437 	case SYS_OFF_MODE_RESTART:
438 		handler->list = &restart_handler_list;
439 		break;
440 
441 	default:
442 		free_sys_off_handler(handler);
443 		return ERR_PTR(-EINVAL);
444 	}
445 
446 	handler->nb.notifier_call = sys_off_notify;
447 	handler->nb.priority = priority;
448 	handler->sys_off_cb = callback;
449 	handler->cb_data = cb_data;
450 	handler->mode = mode;
451 
452 	if (handler->blocking) {
453 		if (priority == SYS_OFF_PRIO_DEFAULT)
454 			err = blocking_notifier_chain_register(handler->list,
455 							       &handler->nb);
456 		else
457 			err = blocking_notifier_chain_register_unique_prio(handler->list,
458 									   &handler->nb);
459 	} else {
460 		if (priority == SYS_OFF_PRIO_DEFAULT)
461 			err = atomic_notifier_chain_register(handler->list,
462 							     &handler->nb);
463 		else
464 			err = atomic_notifier_chain_register_unique_prio(handler->list,
465 									 &handler->nb);
466 	}
467 
468 	if (err) {
469 		free_sys_off_handler(handler);
470 		return ERR_PTR(err);
471 	}
472 
473 	return handler;
474 }
475 EXPORT_SYMBOL_GPL(register_sys_off_handler);
476 
477 /**
478  *	unregister_sys_off_handler - Unregister sys-off handler
479  *	@handler: Sys-off handler
480  *
481  *	Unregisters given sys-off handler.
482  */
483 void unregister_sys_off_handler(struct sys_off_handler *handler)
484 {
485 	int err;
486 
487 	if (IS_ERR_OR_NULL(handler))
488 		return;
489 
490 	if (handler->blocking)
491 		err = blocking_notifier_chain_unregister(handler->list,
492 							 &handler->nb);
493 	else
494 		err = atomic_notifier_chain_unregister(handler->list,
495 						       &handler->nb);
496 
497 	/* sanity check, shall never happen */
498 	WARN_ON(err);
499 
500 	free_sys_off_handler(handler);
501 }
502 EXPORT_SYMBOL_GPL(unregister_sys_off_handler);
503 
504 static void devm_unregister_sys_off_handler(void *data)
505 {
506 	struct sys_off_handler *handler = data;
507 
508 	unregister_sys_off_handler(handler);
509 }
510 
511 /**
512  *	devm_register_sys_off_handler - Register sys-off handler
513  *	@dev: Device that registers handler
514  *	@mode: Sys-off mode
515  *	@priority: Handler priority
516  *	@callback: Callback function
517  *	@cb_data: Callback argument
518  *
519  *	Registers resource-managed sys-off handler.
520  *
521  *	Returns zero on success, or error code on failure.
522  */
523 int devm_register_sys_off_handler(struct device *dev,
524 				  enum sys_off_mode mode,
525 				  int priority,
526 				  int (*callback)(struct sys_off_data *data),
527 				  void *cb_data)
528 {
529 	struct sys_off_handler *handler;
530 
531 	handler = register_sys_off_handler(mode, priority, callback, cb_data);
532 	if (IS_ERR(handler))
533 		return PTR_ERR(handler);
534 	handler->dev = dev;
535 
536 	return devm_add_action_or_reset(dev, devm_unregister_sys_off_handler,
537 					handler);
538 }
539 EXPORT_SYMBOL_GPL(devm_register_sys_off_handler);
540 
541 /**
542  *	devm_register_power_off_handler - Register power-off handler
543  *	@dev: Device that registers callback
544  *	@callback: Callback function
545  *	@cb_data: Callback's argument
546  *
547  *	Registers resource-managed sys-off handler with a default priority
548  *	and using power-off mode.
549  *
550  *	Returns zero on success, or error code on failure.
551  */
552 int devm_register_power_off_handler(struct device *dev,
553 				    int (*callback)(struct sys_off_data *data),
554 				    void *cb_data)
555 {
556 	return devm_register_sys_off_handler(dev,
557 					     SYS_OFF_MODE_POWER_OFF,
558 					     SYS_OFF_PRIO_DEFAULT,
559 					     callback, cb_data);
560 }
561 EXPORT_SYMBOL_GPL(devm_register_power_off_handler);
562 
563 /**
564  *	devm_register_restart_handler - Register restart handler
565  *	@dev: Device that registers callback
566  *	@callback: Callback function
567  *	@cb_data: Callback's argument
568  *
569  *	Registers resource-managed sys-off handler with a default priority
570  *	and using restart mode.
571  *
572  *	Returns zero on success, or error code on failure.
573  */
574 int devm_register_restart_handler(struct device *dev,
575 				  int (*callback)(struct sys_off_data *data),
576 				  void *cb_data)
577 {
578 	return devm_register_sys_off_handler(dev,
579 					     SYS_OFF_MODE_RESTART,
580 					     SYS_OFF_PRIO_DEFAULT,
581 					     callback, cb_data);
582 }
583 EXPORT_SYMBOL_GPL(devm_register_restart_handler);
584 
585 static struct sys_off_handler *platform_power_off_handler;
586 
587 static int platform_power_off_notify(struct sys_off_data *data)
588 {
589 	void (*platform_power_power_off_cb)(void) = data->cb_data;
590 
591 	platform_power_power_off_cb();
592 
593 	return NOTIFY_DONE;
594 }
595 
596 /**
597  *	register_platform_power_off - Register platform-level power-off callback
598  *	@power_off: Power-off callback
599  *
600  *	Registers power-off callback that will be called as last step
601  *	of the power-off sequence. This callback is expected to be invoked
602  *	for the last resort. Only one platform power-off callback is allowed
603  *	to be registered at a time.
604  *
605  *	Returns zero on success, or error code on failure.
606  */
607 int register_platform_power_off(void (*power_off)(void))
608 {
609 	struct sys_off_handler *handler;
610 
611 	handler = register_sys_off_handler(SYS_OFF_MODE_POWER_OFF,
612 					   SYS_OFF_PRIO_PLATFORM,
613 					   platform_power_off_notify,
614 					   power_off);
615 	if (IS_ERR(handler))
616 		return PTR_ERR(handler);
617 
618 	platform_power_off_handler = handler;
619 
620 	return 0;
621 }
622 EXPORT_SYMBOL_GPL(register_platform_power_off);
623 
624 /**
625  *	unregister_platform_power_off - Unregister platform-level power-off callback
626  *	@power_off: Power-off callback
627  *
628  *	Unregisters previously registered platform power-off callback.
629  */
630 void unregister_platform_power_off(void (*power_off)(void))
631 {
632 	if (platform_power_off_handler &&
633 	    platform_power_off_handler->cb_data == power_off) {
634 		unregister_sys_off_handler(platform_power_off_handler);
635 		platform_power_off_handler = NULL;
636 	}
637 }
638 EXPORT_SYMBOL_GPL(unregister_platform_power_off);
639 
640 static int legacy_pm_power_off(struct sys_off_data *data)
641 {
642 	if (pm_power_off)
643 		pm_power_off();
644 
645 	return NOTIFY_DONE;
646 }
647 
648 static void do_kernel_power_off_prepare(void)
649 {
650 	blocking_notifier_call_chain(&power_off_prep_handler_list, 0, NULL);
651 }
652 
653 /**
654  *	do_kernel_power_off - Execute kernel power-off handler call chain
655  *
656  *	Expected to be called as last step of the power-off sequence.
657  *
658  *	Powers off the system immediately if a power-off handler function has
659  *	been registered. Otherwise does nothing.
660  */
661 void do_kernel_power_off(void)
662 {
663 	struct sys_off_handler *sys_off = NULL;
664 
665 	/*
666 	 * Register sys-off handlers for legacy PM callback. This allows
667 	 * legacy PM callbacks temporary co-exist with the new sys-off API.
668 	 *
669 	 * TODO: Remove legacy handlers once all legacy PM users will be
670 	 *       switched to the sys-off based APIs.
671 	 */
672 	if (pm_power_off)
673 		sys_off = register_sys_off_handler(SYS_OFF_MODE_POWER_OFF,
674 						   SYS_OFF_PRIO_DEFAULT,
675 						   legacy_pm_power_off, NULL);
676 
677 	atomic_notifier_call_chain(&power_off_handler_list, 0, NULL);
678 
679 	unregister_sys_off_handler(sys_off);
680 }
681 
682 /**
683  *	kernel_can_power_off - check whether system can be powered off
684  *
685  *	Returns true if power-off handler is registered and system can be
686  *	powered off, false otherwise.
687  */
688 bool kernel_can_power_off(void)
689 {
690 	return !atomic_notifier_call_chain_is_empty(&power_off_handler_list) ||
691 		pm_power_off;
692 }
693 EXPORT_SYMBOL_GPL(kernel_can_power_off);
694 
695 /**
696  *	kernel_power_off - power_off the system
697  *
698  *	Shutdown everything and perform a clean system power_off.
699  */
700 void kernel_power_off(void)
701 {
702 	kernel_shutdown_prepare(SYSTEM_POWER_OFF);
703 	do_kernel_power_off_prepare();
704 	migrate_to_reboot_cpu();
705 	syscore_shutdown();
706 	pr_emerg("Power down\n");
707 	pr_flush(1000, true);
708 	kmsg_dump(KMSG_DUMP_SHUTDOWN);
709 	machine_power_off();
710 }
711 EXPORT_SYMBOL_GPL(kernel_power_off);
712 
713 DEFINE_MUTEX(system_transition_mutex);
714 
715 /*
716  * Reboot system call: for obvious reasons only root may call it,
717  * and even root needs to set up some magic numbers in the registers
718  * so that some mistake won't make this reboot the whole machine.
719  * You can also set the meaning of the ctrl-alt-del-key here.
720  *
721  * reboot doesn't sync: do that yourself before calling this.
722  */
723 SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
724 		void __user *, arg)
725 {
726 	struct pid_namespace *pid_ns = task_active_pid_ns(current);
727 	char buffer[256];
728 	int ret = 0;
729 
730 	/* We only trust the superuser with rebooting the system. */
731 	if (!ns_capable(pid_ns->user_ns, CAP_SYS_BOOT))
732 		return -EPERM;
733 
734 	/* For safety, we require "magic" arguments. */
735 	if (magic1 != LINUX_REBOOT_MAGIC1 ||
736 			(magic2 != LINUX_REBOOT_MAGIC2 &&
737 			magic2 != LINUX_REBOOT_MAGIC2A &&
738 			magic2 != LINUX_REBOOT_MAGIC2B &&
739 			magic2 != LINUX_REBOOT_MAGIC2C))
740 		return -EINVAL;
741 
742 	/*
743 	 * If pid namespaces are enabled and the current task is in a child
744 	 * pid_namespace, the command is handled by reboot_pid_ns() which will
745 	 * call do_exit().
746 	 */
747 	ret = reboot_pid_ns(pid_ns, cmd);
748 	if (ret)
749 		return ret;
750 
751 	/* Instead of trying to make the power_off code look like
752 	 * halt when pm_power_off is not set do it the easy way.
753 	 */
754 	if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !kernel_can_power_off()) {
755 		poweroff_fallback_to_halt = true;
756 		cmd = LINUX_REBOOT_CMD_HALT;
757 	}
758 
759 	mutex_lock(&system_transition_mutex);
760 	switch (cmd) {
761 	case LINUX_REBOOT_CMD_RESTART:
762 		kernel_restart(NULL);
763 		break;
764 
765 	case LINUX_REBOOT_CMD_CAD_ON:
766 		C_A_D = 1;
767 		break;
768 
769 	case LINUX_REBOOT_CMD_CAD_OFF:
770 		C_A_D = 0;
771 		break;
772 
773 	case LINUX_REBOOT_CMD_HALT:
774 		kernel_halt();
775 		do_exit(0);
776 
777 	case LINUX_REBOOT_CMD_POWER_OFF:
778 		kernel_power_off();
779 		do_exit(0);
780 		break;
781 
782 	case LINUX_REBOOT_CMD_RESTART2:
783 		ret = strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1);
784 		if (ret < 0) {
785 			ret = -EFAULT;
786 			break;
787 		}
788 		buffer[sizeof(buffer) - 1] = '\0';
789 
790 		kernel_restart(buffer);
791 		break;
792 
793 #ifdef CONFIG_KEXEC_CORE
794 	case LINUX_REBOOT_CMD_KEXEC:
795 		ret = kernel_kexec();
796 		break;
797 #endif
798 
799 #ifdef CONFIG_HIBERNATION
800 	case LINUX_REBOOT_CMD_SW_SUSPEND:
801 		ret = hibernate();
802 		break;
803 #endif
804 
805 	default:
806 		ret = -EINVAL;
807 		break;
808 	}
809 	mutex_unlock(&system_transition_mutex);
810 	return ret;
811 }
812 
813 static void deferred_cad(struct work_struct *dummy)
814 {
815 	kernel_restart(NULL);
816 }
817 
818 /*
819  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
820  * As it's called within an interrupt, it may NOT sync: the only choice
821  * is whether to reboot at once, or just ignore the ctrl-alt-del.
822  */
823 void ctrl_alt_del(void)
824 {
825 	static DECLARE_WORK(cad_work, deferred_cad);
826 
827 	if (C_A_D)
828 		schedule_work(&cad_work);
829 	else
830 		kill_cad_pid(SIGINT, 1);
831 }
832 
833 #define POWEROFF_CMD_PATH_LEN  256
834 static char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
835 static const char reboot_cmd[] = "/sbin/reboot";
836 
837 static int run_cmd(const char *cmd)
838 {
839 	char **argv;
840 	static char *envp[] = {
841 		"HOME=/",
842 		"PATH=/sbin:/bin:/usr/sbin:/usr/bin",
843 		NULL
844 	};
845 	int ret;
846 	argv = argv_split(GFP_KERNEL, cmd, NULL);
847 	if (argv) {
848 		ret = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_EXEC);
849 		argv_free(argv);
850 	} else {
851 		ret = -ENOMEM;
852 	}
853 
854 	return ret;
855 }
856 
857 static int __orderly_reboot(void)
858 {
859 	int ret;
860 
861 	ret = run_cmd(reboot_cmd);
862 
863 	if (ret) {
864 		pr_warn("Failed to start orderly reboot: forcing the issue\n");
865 		emergency_sync();
866 		kernel_restart(NULL);
867 	}
868 
869 	return ret;
870 }
871 
872 static int __orderly_poweroff(bool force)
873 {
874 	int ret;
875 
876 	ret = run_cmd(poweroff_cmd);
877 
878 	if (ret && force) {
879 		pr_warn("Failed to start orderly shutdown: forcing the issue\n");
880 
881 		/*
882 		 * I guess this should try to kick off some daemon to sync and
883 		 * poweroff asap.  Or not even bother syncing if we're doing an
884 		 * emergency shutdown?
885 		 */
886 		emergency_sync();
887 		kernel_power_off();
888 	}
889 
890 	return ret;
891 }
892 
893 static bool poweroff_force;
894 
895 static void poweroff_work_func(struct work_struct *work)
896 {
897 	__orderly_poweroff(poweroff_force);
898 }
899 
900 static DECLARE_WORK(poweroff_work, poweroff_work_func);
901 
902 /**
903  * orderly_poweroff - Trigger an orderly system poweroff
904  * @force: force poweroff if command execution fails
905  *
906  * This may be called from any context to trigger a system shutdown.
907  * If the orderly shutdown fails, it will force an immediate shutdown.
908  */
909 void orderly_poweroff(bool force)
910 {
911 	if (force) /* do not override the pending "true" */
912 		poweroff_force = true;
913 	schedule_work(&poweroff_work);
914 }
915 EXPORT_SYMBOL_GPL(orderly_poweroff);
916 
917 static void reboot_work_func(struct work_struct *work)
918 {
919 	__orderly_reboot();
920 }
921 
922 static DECLARE_WORK(reboot_work, reboot_work_func);
923 
924 /**
925  * orderly_reboot - Trigger an orderly system reboot
926  *
927  * This may be called from any context to trigger a system reboot.
928  * If the orderly reboot fails, it will force an immediate reboot.
929  */
930 void orderly_reboot(void)
931 {
932 	schedule_work(&reboot_work);
933 }
934 EXPORT_SYMBOL_GPL(orderly_reboot);
935 
936 /**
937  * hw_failure_emergency_poweroff_func - emergency poweroff work after a known delay
938  * @work: work_struct associated with the emergency poweroff function
939  *
940  * This function is called in very critical situations to force
941  * a kernel poweroff after a configurable timeout value.
942  */
943 static void hw_failure_emergency_poweroff_func(struct work_struct *work)
944 {
945 	/*
946 	 * We have reached here after the emergency shutdown waiting period has
947 	 * expired. This means orderly_poweroff has not been able to shut off
948 	 * the system for some reason.
949 	 *
950 	 * Try to shut down the system immediately using kernel_power_off
951 	 * if populated
952 	 */
953 	pr_emerg("Hardware protection timed-out. Trying forced poweroff\n");
954 	kernel_power_off();
955 
956 	/*
957 	 * Worst of the worst case trigger emergency restart
958 	 */
959 	pr_emerg("Hardware protection shutdown failed. Trying emergency restart\n");
960 	emergency_restart();
961 }
962 
963 static DECLARE_DELAYED_WORK(hw_failure_emergency_poweroff_work,
964 			    hw_failure_emergency_poweroff_func);
965 
966 /**
967  * hw_failure_emergency_poweroff - Trigger an emergency system poweroff
968  *
969  * This may be called from any critical situation to trigger a system shutdown
970  * after a given period of time. If time is negative this is not scheduled.
971  */
972 static void hw_failure_emergency_poweroff(int poweroff_delay_ms)
973 {
974 	if (poweroff_delay_ms <= 0)
975 		return;
976 	schedule_delayed_work(&hw_failure_emergency_poweroff_work,
977 			      msecs_to_jiffies(poweroff_delay_ms));
978 }
979 
980 /**
981  * __hw_protection_shutdown - Trigger an emergency system shutdown or reboot
982  *
983  * @reason:		Reason of emergency shutdown or reboot to be printed.
984  * @ms_until_forced:	Time to wait for orderly shutdown or reboot before
985  *			triggering it. Negative value disables the forced
986  *			shutdown or reboot.
987  * @shutdown:		If true, indicates that a shutdown will happen
988  *			after the critical tempeature is reached.
989  *			If false, indicates that a reboot will happen
990  *			after the critical tempeature is reached.
991  *
992  * Initiate an emergency system shutdown or reboot in order to protect
993  * hardware from further damage. Usage examples include a thermal protection.
994  * NOTE: The request is ignored if protection shutdown or reboot is already
995  * pending even if the previous request has given a large timeout for forced
996  * shutdown/reboot.
997  */
998 void __hw_protection_shutdown(const char *reason, int ms_until_forced, bool shutdown)
999 {
1000 	static atomic_t allow_proceed = ATOMIC_INIT(1);
1001 
1002 	pr_emerg("HARDWARE PROTECTION shutdown (%s)\n", reason);
1003 
1004 	/* Shutdown should be initiated only once. */
1005 	if (!atomic_dec_and_test(&allow_proceed))
1006 		return;
1007 
1008 	/*
1009 	 * Queue a backup emergency shutdown in the event of
1010 	 * orderly_poweroff failure
1011 	 */
1012 	hw_failure_emergency_poweroff(ms_until_forced);
1013 	if (shutdown)
1014 		orderly_poweroff(true);
1015 	else
1016 		orderly_reboot();
1017 }
1018 EXPORT_SYMBOL_GPL(__hw_protection_shutdown);
1019 
1020 static int __init reboot_setup(char *str)
1021 {
1022 	for (;;) {
1023 		enum reboot_mode *mode;
1024 
1025 		/*
1026 		 * Having anything passed on the command line via
1027 		 * reboot= will cause us to disable DMI checking
1028 		 * below.
1029 		 */
1030 		reboot_default = 0;
1031 
1032 		if (!strncmp(str, "panic_", 6)) {
1033 			mode = &panic_reboot_mode;
1034 			str += 6;
1035 		} else {
1036 			mode = &reboot_mode;
1037 		}
1038 
1039 		switch (*str) {
1040 		case 'w':
1041 			*mode = REBOOT_WARM;
1042 			break;
1043 
1044 		case 'c':
1045 			*mode = REBOOT_COLD;
1046 			break;
1047 
1048 		case 'h':
1049 			*mode = REBOOT_HARD;
1050 			break;
1051 
1052 		case 's':
1053 			/*
1054 			 * reboot_cpu is s[mp]#### with #### being the processor
1055 			 * to be used for rebooting. Skip 's' or 'smp' prefix.
1056 			 */
1057 			str += str[1] == 'm' && str[2] == 'p' ? 3 : 1;
1058 
1059 			if (isdigit(str[0])) {
1060 				int cpu = simple_strtoul(str, NULL, 0);
1061 
1062 				if (cpu >= num_possible_cpus()) {
1063 					pr_err("Ignoring the CPU number in reboot= option. "
1064 					"CPU %d exceeds possible cpu number %d\n",
1065 					cpu, num_possible_cpus());
1066 					break;
1067 				}
1068 				reboot_cpu = cpu;
1069 			} else
1070 				*mode = REBOOT_SOFT;
1071 			break;
1072 
1073 		case 'g':
1074 			*mode = REBOOT_GPIO;
1075 			break;
1076 
1077 		case 'b':
1078 		case 'a':
1079 		case 'k':
1080 		case 't':
1081 		case 'e':
1082 		case 'p':
1083 			reboot_type = *str;
1084 			break;
1085 
1086 		case 'f':
1087 			reboot_force = 1;
1088 			break;
1089 		}
1090 
1091 		str = strchr(str, ',');
1092 		if (str)
1093 			str++;
1094 		else
1095 			break;
1096 	}
1097 	return 1;
1098 }
1099 __setup("reboot=", reboot_setup);
1100 
1101 #ifdef CONFIG_SYSFS
1102 
1103 #define REBOOT_COLD_STR		"cold"
1104 #define REBOOT_WARM_STR		"warm"
1105 #define REBOOT_HARD_STR		"hard"
1106 #define REBOOT_SOFT_STR		"soft"
1107 #define REBOOT_GPIO_STR		"gpio"
1108 #define REBOOT_UNDEFINED_STR	"undefined"
1109 
1110 #define BOOT_TRIPLE_STR		"triple"
1111 #define BOOT_KBD_STR		"kbd"
1112 #define BOOT_BIOS_STR		"bios"
1113 #define BOOT_ACPI_STR		"acpi"
1114 #define BOOT_EFI_STR		"efi"
1115 #define BOOT_PCI_STR		"pci"
1116 
1117 static ssize_t mode_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
1118 {
1119 	const char *val;
1120 
1121 	switch (reboot_mode) {
1122 	case REBOOT_COLD:
1123 		val = REBOOT_COLD_STR;
1124 		break;
1125 	case REBOOT_WARM:
1126 		val = REBOOT_WARM_STR;
1127 		break;
1128 	case REBOOT_HARD:
1129 		val = REBOOT_HARD_STR;
1130 		break;
1131 	case REBOOT_SOFT:
1132 		val = REBOOT_SOFT_STR;
1133 		break;
1134 	case REBOOT_GPIO:
1135 		val = REBOOT_GPIO_STR;
1136 		break;
1137 	default:
1138 		val = REBOOT_UNDEFINED_STR;
1139 	}
1140 
1141 	return sysfs_emit(buf, "%s\n", val);
1142 }
1143 static ssize_t mode_store(struct kobject *kobj, struct kobj_attribute *attr,
1144 			  const char *buf, size_t count)
1145 {
1146 	if (!capable(CAP_SYS_BOOT))
1147 		return -EPERM;
1148 
1149 	if (!strncmp(buf, REBOOT_COLD_STR, strlen(REBOOT_COLD_STR)))
1150 		reboot_mode = REBOOT_COLD;
1151 	else if (!strncmp(buf, REBOOT_WARM_STR, strlen(REBOOT_WARM_STR)))
1152 		reboot_mode = REBOOT_WARM;
1153 	else if (!strncmp(buf, REBOOT_HARD_STR, strlen(REBOOT_HARD_STR)))
1154 		reboot_mode = REBOOT_HARD;
1155 	else if (!strncmp(buf, REBOOT_SOFT_STR, strlen(REBOOT_SOFT_STR)))
1156 		reboot_mode = REBOOT_SOFT;
1157 	else if (!strncmp(buf, REBOOT_GPIO_STR, strlen(REBOOT_GPIO_STR)))
1158 		reboot_mode = REBOOT_GPIO;
1159 	else
1160 		return -EINVAL;
1161 
1162 	reboot_default = 0;
1163 
1164 	return count;
1165 }
1166 static struct kobj_attribute reboot_mode_attr = __ATTR_RW(mode);
1167 
1168 #ifdef CONFIG_X86
1169 static ssize_t force_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
1170 {
1171 	return sysfs_emit(buf, "%d\n", reboot_force);
1172 }
1173 static ssize_t force_store(struct kobject *kobj, struct kobj_attribute *attr,
1174 			  const char *buf, size_t count)
1175 {
1176 	bool res;
1177 
1178 	if (!capable(CAP_SYS_BOOT))
1179 		return -EPERM;
1180 
1181 	if (kstrtobool(buf, &res))
1182 		return -EINVAL;
1183 
1184 	reboot_default = 0;
1185 	reboot_force = res;
1186 
1187 	return count;
1188 }
1189 static struct kobj_attribute reboot_force_attr = __ATTR_RW(force);
1190 
1191 static ssize_t type_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
1192 {
1193 	const char *val;
1194 
1195 	switch (reboot_type) {
1196 	case BOOT_TRIPLE:
1197 		val = BOOT_TRIPLE_STR;
1198 		break;
1199 	case BOOT_KBD:
1200 		val = BOOT_KBD_STR;
1201 		break;
1202 	case BOOT_BIOS:
1203 		val = BOOT_BIOS_STR;
1204 		break;
1205 	case BOOT_ACPI:
1206 		val = BOOT_ACPI_STR;
1207 		break;
1208 	case BOOT_EFI:
1209 		val = BOOT_EFI_STR;
1210 		break;
1211 	case BOOT_CF9_FORCE:
1212 		val = BOOT_PCI_STR;
1213 		break;
1214 	default:
1215 		val = REBOOT_UNDEFINED_STR;
1216 	}
1217 
1218 	return sysfs_emit(buf, "%s\n", val);
1219 }
1220 static ssize_t type_store(struct kobject *kobj, struct kobj_attribute *attr,
1221 			  const char *buf, size_t count)
1222 {
1223 	if (!capable(CAP_SYS_BOOT))
1224 		return -EPERM;
1225 
1226 	if (!strncmp(buf, BOOT_TRIPLE_STR, strlen(BOOT_TRIPLE_STR)))
1227 		reboot_type = BOOT_TRIPLE;
1228 	else if (!strncmp(buf, BOOT_KBD_STR, strlen(BOOT_KBD_STR)))
1229 		reboot_type = BOOT_KBD;
1230 	else if (!strncmp(buf, BOOT_BIOS_STR, strlen(BOOT_BIOS_STR)))
1231 		reboot_type = BOOT_BIOS;
1232 	else if (!strncmp(buf, BOOT_ACPI_STR, strlen(BOOT_ACPI_STR)))
1233 		reboot_type = BOOT_ACPI;
1234 	else if (!strncmp(buf, BOOT_EFI_STR, strlen(BOOT_EFI_STR)))
1235 		reboot_type = BOOT_EFI;
1236 	else if (!strncmp(buf, BOOT_PCI_STR, strlen(BOOT_PCI_STR)))
1237 		reboot_type = BOOT_CF9_FORCE;
1238 	else
1239 		return -EINVAL;
1240 
1241 	reboot_default = 0;
1242 
1243 	return count;
1244 }
1245 static struct kobj_attribute reboot_type_attr = __ATTR_RW(type);
1246 #endif
1247 
1248 #ifdef CONFIG_SMP
1249 static ssize_t cpu_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
1250 {
1251 	return sysfs_emit(buf, "%d\n", reboot_cpu);
1252 }
1253 static ssize_t cpu_store(struct kobject *kobj, struct kobj_attribute *attr,
1254 			  const char *buf, size_t count)
1255 {
1256 	unsigned int cpunum;
1257 	int rc;
1258 
1259 	if (!capable(CAP_SYS_BOOT))
1260 		return -EPERM;
1261 
1262 	rc = kstrtouint(buf, 0, &cpunum);
1263 
1264 	if (rc)
1265 		return rc;
1266 
1267 	if (cpunum >= num_possible_cpus())
1268 		return -ERANGE;
1269 
1270 	reboot_default = 0;
1271 	reboot_cpu = cpunum;
1272 
1273 	return count;
1274 }
1275 static struct kobj_attribute reboot_cpu_attr = __ATTR_RW(cpu);
1276 #endif
1277 
1278 static struct attribute *reboot_attrs[] = {
1279 	&reboot_mode_attr.attr,
1280 #ifdef CONFIG_X86
1281 	&reboot_force_attr.attr,
1282 	&reboot_type_attr.attr,
1283 #endif
1284 #ifdef CONFIG_SMP
1285 	&reboot_cpu_attr.attr,
1286 #endif
1287 	NULL,
1288 };
1289 
1290 #ifdef CONFIG_SYSCTL
1291 static const struct ctl_table kern_reboot_table[] = {
1292 	{
1293 		.procname       = "poweroff_cmd",
1294 		.data           = &poweroff_cmd,
1295 		.maxlen         = POWEROFF_CMD_PATH_LEN,
1296 		.mode           = 0644,
1297 		.proc_handler   = proc_dostring,
1298 	},
1299 	{
1300 		.procname       = "ctrl-alt-del",
1301 		.data           = &C_A_D,
1302 		.maxlen         = sizeof(int),
1303 		.mode           = 0644,
1304 		.proc_handler   = proc_dointvec,
1305 	},
1306 };
1307 
1308 static void __init kernel_reboot_sysctls_init(void)
1309 {
1310 	register_sysctl_init("kernel", kern_reboot_table);
1311 }
1312 #else
1313 #define kernel_reboot_sysctls_init() do { } while (0)
1314 #endif /* CONFIG_SYSCTL */
1315 
1316 static const struct attribute_group reboot_attr_group = {
1317 	.attrs = reboot_attrs,
1318 };
1319 
1320 static int __init reboot_ksysfs_init(void)
1321 {
1322 	struct kobject *reboot_kobj;
1323 	int ret;
1324 
1325 	reboot_kobj = kobject_create_and_add("reboot", kernel_kobj);
1326 	if (!reboot_kobj)
1327 		return -ENOMEM;
1328 
1329 	ret = sysfs_create_group(reboot_kobj, &reboot_attr_group);
1330 	if (ret) {
1331 		kobject_put(reboot_kobj);
1332 		return ret;
1333 	}
1334 
1335 	kernel_reboot_sysctls_init();
1336 
1337 	return 0;
1338 }
1339 late_initcall(reboot_ksysfs_init);
1340 
1341 #endif
1342