4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/module.h>
9 #include <linux/utsname.h>
10 #include <linux/mman.h>
11 #include <linux/smp_lock.h>
12 #include <linux/notifier.h>
13 #include <linux/reboot.h>
14 #include <linux/prctl.h>
15 #include <linux/highuid.h>
17 #include <linux/resource.h>
18 #include <linux/kernel.h>
19 #include <linux/kexec.h>
20 #include <linux/workqueue.h>
21 #include <linux/capability.h>
22 #include <linux/device.h>
23 #include <linux/key.h>
24 #include <linux/times.h>
25 #include <linux/posix-timers.h>
26 #include <linux/security.h>
27 #include <linux/dcookies.h>
28 #include <linux/suspend.h>
29 #include <linux/tty.h>
30 #include <linux/signal.h>
31 #include <linux/cn_proc.h>
32 #include <linux/getcpu.h>
33 #include <linux/task_io_accounting_ops.h>
34 #include <linux/seccomp.h>
35 #include <linux/cpu.h>
37 #include <linux/compat.h>
38 #include <linux/syscalls.h>
39 #include <linux/kprobes.h>
40 #include <linux/user_namespace.h>
42 #include <asm/uaccess.h>
44 #include <asm/unistd.h>
46 #ifndef SET_UNALIGN_CTL
47 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
49 #ifndef GET_UNALIGN_CTL
50 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
53 # define SET_FPEMU_CTL(a,b) (-EINVAL)
56 # define GET_FPEMU_CTL(a,b) (-EINVAL)
59 # define SET_FPEXC_CTL(a,b) (-EINVAL)
62 # define GET_FPEXC_CTL(a,b) (-EINVAL)
65 # define GET_ENDIAN(a,b) (-EINVAL)
68 # define SET_ENDIAN(a,b) (-EINVAL)
72 * this is where the system-wide overflow UID and GID are defined, for
73 * architectures that now have 32-bit UID/GID but didn't in the past
76 int overflowuid = DEFAULT_OVERFLOWUID;
77 int overflowgid = DEFAULT_OVERFLOWGID;
80 EXPORT_SYMBOL(overflowuid);
81 EXPORT_SYMBOL(overflowgid);
85 * the same as above, but for filesystems which can only store a 16-bit
86 * UID and GID. as such, this is needed on all architectures
89 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
90 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
92 EXPORT_SYMBOL(fs_overflowuid);
93 EXPORT_SYMBOL(fs_overflowgid);
96 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
101 EXPORT_SYMBOL(cad_pid);
104 * If set, this is used for preparing the system to power off.
107 void (*pm_power_off_prepare)(void);
108 EXPORT_SYMBOL(pm_power_off_prepare);
111 * Notifier list for kernel code which wants to be called
112 * at shutdown. This is used to stop any idling DMA operations
116 static BLOCKING_NOTIFIER_HEAD(reboot_notifier_list);
119 * Notifier chain core routines. The exported routines below
120 * are layered on top of these, with appropriate locking added.
123 static int notifier_chain_register(struct notifier_block **nl,
124 struct notifier_block *n)
126 while ((*nl) != NULL) {
127 if (n->priority > (*nl)->priority)
132 rcu_assign_pointer(*nl, n);
136 static int notifier_chain_unregister(struct notifier_block **nl,
137 struct notifier_block *n)
139 while ((*nl) != NULL) {
141 rcu_assign_pointer(*nl, n->next);
150 * notifier_call_chain - Informs the registered notifiers about an event.
151 * @nl: Pointer to head of the blocking notifier chain
152 * @val: Value passed unmodified to notifier function
153 * @v: Pointer passed unmodified to notifier function
154 * @nr_to_call: Number of notifier functions to be called. Don't care
155 * value of this parameter is -1.
156 * @nr_calls: Records the number of notifications sent. Don't care
157 * value of this field is NULL.
158 * @returns: notifier_call_chain returns the value returned by the
159 * last notifier function called.
162 static int __kprobes notifier_call_chain(struct notifier_block **nl,
163 unsigned long val, void *v,
164 int nr_to_call, int *nr_calls)
166 int ret = NOTIFY_DONE;
167 struct notifier_block *nb, *next_nb;
169 nb = rcu_dereference(*nl);
171 while (nb && nr_to_call) {
172 next_nb = rcu_dereference(nb->next);
173 ret = nb->notifier_call(nb, val, v);
178 if ((ret & NOTIFY_STOP_MASK) == NOTIFY_STOP_MASK)
187 * Atomic notifier chain routines. Registration and unregistration
188 * use a spinlock, and call_chain is synchronized by RCU (no locks).
192 * atomic_notifier_chain_register - Add notifier to an atomic notifier chain
193 * @nh: Pointer to head of the atomic notifier chain
194 * @n: New entry in notifier chain
196 * Adds a notifier to an atomic notifier chain.
198 * Currently always returns zero.
201 int atomic_notifier_chain_register(struct atomic_notifier_head *nh,
202 struct notifier_block *n)
207 spin_lock_irqsave(&nh->lock, flags);
208 ret = notifier_chain_register(&nh->head, n);
209 spin_unlock_irqrestore(&nh->lock, flags);
213 EXPORT_SYMBOL_GPL(atomic_notifier_chain_register);
216 * atomic_notifier_chain_unregister - Remove notifier from an atomic notifier chain
217 * @nh: Pointer to head of the atomic notifier chain
218 * @n: Entry to remove from notifier chain
220 * Removes a notifier from an atomic notifier chain.
222 * Returns zero on success or %-ENOENT on failure.
224 int atomic_notifier_chain_unregister(struct atomic_notifier_head *nh,
225 struct notifier_block *n)
230 spin_lock_irqsave(&nh->lock, flags);
231 ret = notifier_chain_unregister(&nh->head, n);
232 spin_unlock_irqrestore(&nh->lock, flags);
237 EXPORT_SYMBOL_GPL(atomic_notifier_chain_unregister);
240 * __atomic_notifier_call_chain - Call functions in an atomic notifier chain
241 * @nh: Pointer to head of the atomic notifier chain
242 * @val: Value passed unmodified to notifier function
243 * @v: Pointer passed unmodified to notifier function
244 * @nr_to_call: See the comment for notifier_call_chain.
245 * @nr_calls: See the comment for notifier_call_chain.
247 * Calls each function in a notifier chain in turn. The functions
248 * run in an atomic context, so they must not block.
249 * This routine uses RCU to synchronize with changes to the chain.
251 * If the return value of the notifier can be and'ed
252 * with %NOTIFY_STOP_MASK then atomic_notifier_call_chain()
253 * will return immediately, with the return value of
254 * the notifier function which halted execution.
255 * Otherwise the return value is the return value
256 * of the last notifier function called.
259 int __kprobes __atomic_notifier_call_chain(struct atomic_notifier_head *nh,
260 unsigned long val, void *v,
261 int nr_to_call, int *nr_calls)
266 ret = notifier_call_chain(&nh->head, val, v, nr_to_call, nr_calls);
271 EXPORT_SYMBOL_GPL(__atomic_notifier_call_chain);
273 int __kprobes atomic_notifier_call_chain(struct atomic_notifier_head *nh,
274 unsigned long val, void *v)
276 return __atomic_notifier_call_chain(nh, val, v, -1, NULL);
279 EXPORT_SYMBOL_GPL(atomic_notifier_call_chain);
281 * Blocking notifier chain routines. All access to the chain is
282 * synchronized by an rwsem.
286 * blocking_notifier_chain_register - Add notifier to a blocking notifier chain
287 * @nh: Pointer to head of the blocking notifier chain
288 * @n: New entry in notifier chain
290 * Adds a notifier to a blocking notifier chain.
291 * Must be called in process context.
293 * Currently always returns zero.
296 int blocking_notifier_chain_register(struct blocking_notifier_head *nh,
297 struct notifier_block *n)
302 * This code gets used during boot-up, when task switching is
303 * not yet working and interrupts must remain disabled. At
304 * such times we must not call down_write().
306 if (unlikely(system_state == SYSTEM_BOOTING))
307 return notifier_chain_register(&nh->head, n);
309 down_write(&nh->rwsem);
310 ret = notifier_chain_register(&nh->head, n);
311 up_write(&nh->rwsem);
315 EXPORT_SYMBOL_GPL(blocking_notifier_chain_register);
318 * blocking_notifier_chain_unregister - Remove notifier from a blocking notifier chain
319 * @nh: Pointer to head of the blocking notifier chain
320 * @n: Entry to remove from notifier chain
322 * Removes a notifier from a blocking notifier chain.
323 * Must be called from process context.
325 * Returns zero on success or %-ENOENT on failure.
327 int blocking_notifier_chain_unregister(struct blocking_notifier_head *nh,
328 struct notifier_block *n)
333 * This code gets used during boot-up, when task switching is
334 * not yet working and interrupts must remain disabled. At
335 * such times we must not call down_write().
337 if (unlikely(system_state == SYSTEM_BOOTING))
338 return notifier_chain_unregister(&nh->head, n);
340 down_write(&nh->rwsem);
341 ret = notifier_chain_unregister(&nh->head, n);
342 up_write(&nh->rwsem);
346 EXPORT_SYMBOL_GPL(blocking_notifier_chain_unregister);
349 * __blocking_notifier_call_chain - Call functions in a blocking notifier chain
350 * @nh: Pointer to head of the blocking notifier chain
351 * @val: Value passed unmodified to notifier function
352 * @v: Pointer passed unmodified to notifier function
353 * @nr_to_call: See comment for notifier_call_chain.
354 * @nr_calls: See comment for notifier_call_chain.
356 * Calls each function in a notifier chain in turn. The functions
357 * run in a process context, so they are allowed to block.
359 * If the return value of the notifier can be and'ed
360 * with %NOTIFY_STOP_MASK then blocking_notifier_call_chain()
361 * will return immediately, with the return value of
362 * the notifier function which halted execution.
363 * Otherwise the return value is the return value
364 * of the last notifier function called.
367 int __blocking_notifier_call_chain(struct blocking_notifier_head *nh,
368 unsigned long val, void *v,
369 int nr_to_call, int *nr_calls)
371 int ret = NOTIFY_DONE;
374 * We check the head outside the lock, but if this access is
375 * racy then it does not matter what the result of the test
376 * is, we re-check the list after having taken the lock anyway:
378 if (rcu_dereference(nh->head)) {
379 down_read(&nh->rwsem);
380 ret = notifier_call_chain(&nh->head, val, v, nr_to_call,
386 EXPORT_SYMBOL_GPL(__blocking_notifier_call_chain);
388 int blocking_notifier_call_chain(struct blocking_notifier_head *nh,
389 unsigned long val, void *v)
391 return __blocking_notifier_call_chain(nh, val, v, -1, NULL);
393 EXPORT_SYMBOL_GPL(blocking_notifier_call_chain);
396 * Raw notifier chain routines. There is no protection;
397 * the caller must provide it. Use at your own risk!
401 * raw_notifier_chain_register - Add notifier to a raw notifier chain
402 * @nh: Pointer to head of the raw notifier chain
403 * @n: New entry in notifier chain
405 * Adds a notifier to a raw notifier chain.
406 * All locking must be provided by the caller.
408 * Currently always returns zero.
411 int raw_notifier_chain_register(struct raw_notifier_head *nh,
412 struct notifier_block *n)
414 return notifier_chain_register(&nh->head, n);
417 EXPORT_SYMBOL_GPL(raw_notifier_chain_register);
420 * raw_notifier_chain_unregister - Remove notifier from a raw notifier chain
421 * @nh: Pointer to head of the raw notifier chain
422 * @n: Entry to remove from notifier chain
424 * Removes a notifier from a raw notifier chain.
425 * All locking must be provided by the caller.
427 * Returns zero on success or %-ENOENT on failure.
429 int raw_notifier_chain_unregister(struct raw_notifier_head *nh,
430 struct notifier_block *n)
432 return notifier_chain_unregister(&nh->head, n);
435 EXPORT_SYMBOL_GPL(raw_notifier_chain_unregister);
438 * __raw_notifier_call_chain - Call functions in a raw notifier chain
439 * @nh: Pointer to head of the raw notifier chain
440 * @val: Value passed unmodified to notifier function
441 * @v: Pointer passed unmodified to notifier function
442 * @nr_to_call: See comment for notifier_call_chain.
443 * @nr_calls: See comment for notifier_call_chain
445 * Calls each function in a notifier chain in turn. The functions
446 * run in an undefined context.
447 * All locking must be provided by the caller.
449 * If the return value of the notifier can be and'ed
450 * with %NOTIFY_STOP_MASK then raw_notifier_call_chain()
451 * will return immediately, with the return value of
452 * the notifier function which halted execution.
453 * Otherwise the return value is the return value
454 * of the last notifier function called.
457 int __raw_notifier_call_chain(struct raw_notifier_head *nh,
458 unsigned long val, void *v,
459 int nr_to_call, int *nr_calls)
461 return notifier_call_chain(&nh->head, val, v, nr_to_call, nr_calls);
464 EXPORT_SYMBOL_GPL(__raw_notifier_call_chain);
466 int raw_notifier_call_chain(struct raw_notifier_head *nh,
467 unsigned long val, void *v)
469 return __raw_notifier_call_chain(nh, val, v, -1, NULL);
472 EXPORT_SYMBOL_GPL(raw_notifier_call_chain);
475 * SRCU notifier chain routines. Registration and unregistration
476 * use a mutex, and call_chain is synchronized by SRCU (no locks).
480 * srcu_notifier_chain_register - Add notifier to an SRCU notifier chain
481 * @nh: Pointer to head of the SRCU notifier chain
482 * @n: New entry in notifier chain
484 * Adds a notifier to an SRCU notifier chain.
485 * Must be called in process context.
487 * Currently always returns zero.
490 int srcu_notifier_chain_register(struct srcu_notifier_head *nh,
491 struct notifier_block *n)
496 * This code gets used during boot-up, when task switching is
497 * not yet working and interrupts must remain disabled. At
498 * such times we must not call mutex_lock().
500 if (unlikely(system_state == SYSTEM_BOOTING))
501 return notifier_chain_register(&nh->head, n);
503 mutex_lock(&nh->mutex);
504 ret = notifier_chain_register(&nh->head, n);
505 mutex_unlock(&nh->mutex);
509 EXPORT_SYMBOL_GPL(srcu_notifier_chain_register);
512 * srcu_notifier_chain_unregister - Remove notifier from an SRCU notifier chain
513 * @nh: Pointer to head of the SRCU notifier chain
514 * @n: Entry to remove from notifier chain
516 * Removes a notifier from an SRCU notifier chain.
517 * Must be called from process context.
519 * Returns zero on success or %-ENOENT on failure.
521 int srcu_notifier_chain_unregister(struct srcu_notifier_head *nh,
522 struct notifier_block *n)
527 * This code gets used during boot-up, when task switching is
528 * not yet working and interrupts must remain disabled. At
529 * such times we must not call mutex_lock().
531 if (unlikely(system_state == SYSTEM_BOOTING))
532 return notifier_chain_unregister(&nh->head, n);
534 mutex_lock(&nh->mutex);
535 ret = notifier_chain_unregister(&nh->head, n);
536 mutex_unlock(&nh->mutex);
537 synchronize_srcu(&nh->srcu);
541 EXPORT_SYMBOL_GPL(srcu_notifier_chain_unregister);
544 * __srcu_notifier_call_chain - Call functions in an SRCU notifier chain
545 * @nh: Pointer to head of the SRCU notifier chain
546 * @val: Value passed unmodified to notifier function
547 * @v: Pointer passed unmodified to notifier function
548 * @nr_to_call: See comment for notifier_call_chain.
549 * @nr_calls: See comment for notifier_call_chain
551 * Calls each function in a notifier chain in turn. The functions
552 * run in a process context, so they are allowed to block.
554 * If the return value of the notifier can be and'ed
555 * with %NOTIFY_STOP_MASK then srcu_notifier_call_chain()
556 * will return immediately, with the return value of
557 * the notifier function which halted execution.
558 * Otherwise the return value is the return value
559 * of the last notifier function called.
562 int __srcu_notifier_call_chain(struct srcu_notifier_head *nh,
563 unsigned long val, void *v,
564 int nr_to_call, int *nr_calls)
569 idx = srcu_read_lock(&nh->srcu);
570 ret = notifier_call_chain(&nh->head, val, v, nr_to_call, nr_calls);
571 srcu_read_unlock(&nh->srcu, idx);
574 EXPORT_SYMBOL_GPL(__srcu_notifier_call_chain);
576 int srcu_notifier_call_chain(struct srcu_notifier_head *nh,
577 unsigned long val, void *v)
579 return __srcu_notifier_call_chain(nh, val, v, -1, NULL);
581 EXPORT_SYMBOL_GPL(srcu_notifier_call_chain);
584 * srcu_init_notifier_head - Initialize an SRCU notifier head
585 * @nh: Pointer to head of the srcu notifier chain
587 * Unlike other sorts of notifier heads, SRCU notifier heads require
588 * dynamic initialization. Be sure to call this routine before
589 * calling any of the other SRCU notifier routines for this head.
591 * If an SRCU notifier head is deallocated, it must first be cleaned
592 * up by calling srcu_cleanup_notifier_head(). Otherwise the head's
593 * per-cpu data (used by the SRCU mechanism) will leak.
596 void srcu_init_notifier_head(struct srcu_notifier_head *nh)
598 mutex_init(&nh->mutex);
599 if (init_srcu_struct(&nh->srcu) < 0)
604 EXPORT_SYMBOL_GPL(srcu_init_notifier_head);
607 * register_reboot_notifier - Register function to be called at reboot time
608 * @nb: Info about notifier function to be called
610 * Registers a function with the list of functions
611 * to be called at reboot time.
613 * Currently always returns zero, as blocking_notifier_chain_register()
614 * always returns zero.
617 int register_reboot_notifier(struct notifier_block * nb)
619 return blocking_notifier_chain_register(&reboot_notifier_list, nb);
622 EXPORT_SYMBOL(register_reboot_notifier);
625 * unregister_reboot_notifier - Unregister previously registered reboot notifier
626 * @nb: Hook to be unregistered
628 * Unregisters a previously registered reboot
631 * Returns zero on success, or %-ENOENT on failure.
634 int unregister_reboot_notifier(struct notifier_block * nb)
636 return blocking_notifier_chain_unregister(&reboot_notifier_list, nb);
639 EXPORT_SYMBOL(unregister_reboot_notifier);
641 static int set_one_prio(struct task_struct *p, int niceval, int error)
645 if (p->uid != current->euid &&
646 p->euid != current->euid && !capable(CAP_SYS_NICE)) {
650 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
654 no_nice = security_task_setnice(p, niceval);
661 set_user_nice(p, niceval);
666 asmlinkage long sys_setpriority(int which, int who, int niceval)
668 struct task_struct *g, *p;
669 struct user_struct *user;
673 if (which > PRIO_USER || which < PRIO_PROCESS)
676 /* normalize: avoid signed division (rounding problems) */
683 read_lock(&tasklist_lock);
687 p = find_task_by_pid(who);
691 error = set_one_prio(p, niceval, error);
695 pgrp = find_pid(who);
697 pgrp = task_pgrp(current);
698 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
699 error = set_one_prio(p, niceval, error);
700 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
703 user = current->user;
707 if ((who != current->uid) && !(user = find_user(who)))
708 goto out_unlock; /* No processes for this user */
712 error = set_one_prio(p, niceval, error);
713 while_each_thread(g, p);
714 if (who != current->uid)
715 free_uid(user); /* For find_user() */
719 read_unlock(&tasklist_lock);
725 * Ugh. To avoid negative return values, "getpriority()" will
726 * not return the normal nice-value, but a negated value that
727 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
728 * to stay compatible.
730 asmlinkage long sys_getpriority(int which, int who)
732 struct task_struct *g, *p;
733 struct user_struct *user;
734 long niceval, retval = -ESRCH;
737 if (which > PRIO_USER || which < PRIO_PROCESS)
740 read_lock(&tasklist_lock);
744 p = find_task_by_pid(who);
748 niceval = 20 - task_nice(p);
749 if (niceval > retval)
755 pgrp = find_pid(who);
757 pgrp = task_pgrp(current);
758 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
759 niceval = 20 - task_nice(p);
760 if (niceval > retval)
762 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
765 user = current->user;
769 if ((who != current->uid) && !(user = find_user(who)))
770 goto out_unlock; /* No processes for this user */
774 niceval = 20 - task_nice(p);
775 if (niceval > retval)
778 while_each_thread(g, p);
779 if (who != current->uid)
780 free_uid(user); /* for find_user() */
784 read_unlock(&tasklist_lock);
790 * emergency_restart - reboot the system
792 * Without shutting down any hardware or taking any locks
793 * reboot the system. This is called when we know we are in
794 * trouble so this is our best effort to reboot. This is
795 * safe to call in interrupt context.
797 void emergency_restart(void)
799 machine_emergency_restart();
801 EXPORT_SYMBOL_GPL(emergency_restart);
803 static void kernel_restart_prepare(char *cmd)
805 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
806 system_state = SYSTEM_RESTART;
812 * kernel_restart - reboot the system
813 * @cmd: pointer to buffer containing command to execute for restart
816 * Shutdown everything and perform a clean reboot.
817 * This is not safe to call in interrupt context.
819 void kernel_restart(char *cmd)
821 kernel_restart_prepare(cmd);
823 printk(KERN_EMERG "Restarting system.\n");
825 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
826 machine_restart(cmd);
828 EXPORT_SYMBOL_GPL(kernel_restart);
831 * kernel_kexec - reboot the system
833 * Move into place and start executing a preloaded standalone
834 * executable. If nothing was preloaded return an error.
836 static void kernel_kexec(void)
839 struct kimage *image;
840 image = xchg(&kexec_image, NULL);
843 kernel_restart_prepare(NULL);
844 printk(KERN_EMERG "Starting new kernel\n");
846 machine_kexec(image);
850 void kernel_shutdown_prepare(enum system_states state)
852 blocking_notifier_call_chain(&reboot_notifier_list,
853 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
854 system_state = state;
858 * kernel_halt - halt the system
860 * Shutdown everything and perform a clean system halt.
862 void kernel_halt(void)
864 kernel_shutdown_prepare(SYSTEM_HALT);
866 printk(KERN_EMERG "System halted.\n");
870 EXPORT_SYMBOL_GPL(kernel_halt);
873 * kernel_power_off - power_off the system
875 * Shutdown everything and perform a clean system power_off.
877 void kernel_power_off(void)
879 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
880 if (pm_power_off_prepare)
881 pm_power_off_prepare();
882 disable_nonboot_cpus();
884 printk(KERN_EMERG "Power down.\n");
887 EXPORT_SYMBOL_GPL(kernel_power_off);
889 * Reboot system call: for obvious reasons only root may call it,
890 * and even root needs to set up some magic numbers in the registers
891 * so that some mistake won't make this reboot the whole machine.
892 * You can also set the meaning of the ctrl-alt-del-key here.
894 * reboot doesn't sync: do that yourself before calling this.
896 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
900 /* We only trust the superuser with rebooting the system. */
901 if (!capable(CAP_SYS_BOOT))
904 /* For safety, we require "magic" arguments. */
905 if (magic1 != LINUX_REBOOT_MAGIC1 ||
906 (magic2 != LINUX_REBOOT_MAGIC2 &&
907 magic2 != LINUX_REBOOT_MAGIC2A &&
908 magic2 != LINUX_REBOOT_MAGIC2B &&
909 magic2 != LINUX_REBOOT_MAGIC2C))
912 /* Instead of trying to make the power_off code look like
913 * halt when pm_power_off is not set do it the easy way.
915 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
916 cmd = LINUX_REBOOT_CMD_HALT;
920 case LINUX_REBOOT_CMD_RESTART:
921 kernel_restart(NULL);
924 case LINUX_REBOOT_CMD_CAD_ON:
928 case LINUX_REBOOT_CMD_CAD_OFF:
932 case LINUX_REBOOT_CMD_HALT:
938 case LINUX_REBOOT_CMD_POWER_OFF:
944 case LINUX_REBOOT_CMD_RESTART2:
945 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
949 buffer[sizeof(buffer) - 1] = '\0';
951 kernel_restart(buffer);
954 case LINUX_REBOOT_CMD_KEXEC:
959 #ifdef CONFIG_HIBERNATION
960 case LINUX_REBOOT_CMD_SW_SUSPEND:
962 int ret = hibernate();
976 static void deferred_cad(struct work_struct *dummy)
978 kernel_restart(NULL);
982 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
983 * As it's called within an interrupt, it may NOT sync: the only choice
984 * is whether to reboot at once, or just ignore the ctrl-alt-del.
986 void ctrl_alt_del(void)
988 static DECLARE_WORK(cad_work, deferred_cad);
991 schedule_work(&cad_work);
993 kill_cad_pid(SIGINT, 1);
997 * Unprivileged users may change the real gid to the effective gid
998 * or vice versa. (BSD-style)
1000 * If you set the real gid at all, or set the effective gid to a value not
1001 * equal to the real gid, then the saved gid is set to the new effective gid.
1003 * This makes it possible for a setgid program to completely drop its
1004 * privileges, which is often a useful assertion to make when you are doing
1005 * a security audit over a program.
1007 * The general idea is that a program which uses just setregid() will be
1008 * 100% compatible with BSD. A program which uses just setgid() will be
1009 * 100% compatible with POSIX with saved IDs.
1011 * SMP: There are not races, the GIDs are checked only by filesystem
1012 * operations (as far as semantic preservation is concerned).
1014 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
1016 int old_rgid = current->gid;
1017 int old_egid = current->egid;
1018 int new_rgid = old_rgid;
1019 int new_egid = old_egid;
1022 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
1026 if (rgid != (gid_t) -1) {
1027 if ((old_rgid == rgid) ||
1028 (current->egid==rgid) ||
1029 capable(CAP_SETGID))
1034 if (egid != (gid_t) -1) {
1035 if ((old_rgid == egid) ||
1036 (current->egid == egid) ||
1037 (current->sgid == egid) ||
1038 capable(CAP_SETGID))
1043 if (new_egid != old_egid) {
1044 set_dumpable(current->mm, suid_dumpable);
1047 if (rgid != (gid_t) -1 ||
1048 (egid != (gid_t) -1 && egid != old_rgid))
1049 current->sgid = new_egid;
1050 current->fsgid = new_egid;
1051 current->egid = new_egid;
1052 current->gid = new_rgid;
1053 key_fsgid_changed(current);
1054 proc_id_connector(current, PROC_EVENT_GID);
1059 * setgid() is implemented like SysV w/ SAVED_IDS
1061 * SMP: Same implicit races as above.
1063 asmlinkage long sys_setgid(gid_t gid)
1065 int old_egid = current->egid;
1068 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
1072 if (capable(CAP_SETGID)) {
1073 if (old_egid != gid) {
1074 set_dumpable(current->mm, suid_dumpable);
1077 current->gid = current->egid = current->sgid = current->fsgid = gid;
1078 } else if ((gid == current->gid) || (gid == current->sgid)) {
1079 if (old_egid != gid) {
1080 set_dumpable(current->mm, suid_dumpable);
1083 current->egid = current->fsgid = gid;
1088 key_fsgid_changed(current);
1089 proc_id_connector(current, PROC_EVENT_GID);
1093 static int set_user(uid_t new_ruid, int dumpclear)
1095 struct user_struct *new_user;
1097 new_user = alloc_uid(current->nsproxy->user_ns, new_ruid);
1101 if (atomic_read(&new_user->processes) >=
1102 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
1103 new_user != current->nsproxy->user_ns->root_user) {
1108 switch_uid(new_user);
1111 set_dumpable(current->mm, suid_dumpable);
1114 current->uid = new_ruid;
1119 * Unprivileged users may change the real uid to the effective uid
1120 * or vice versa. (BSD-style)
1122 * If you set the real uid at all, or set the effective uid to a value not
1123 * equal to the real uid, then the saved uid is set to the new effective uid.
1125 * This makes it possible for a setuid program to completely drop its
1126 * privileges, which is often a useful assertion to make when you are doing
1127 * a security audit over a program.
1129 * The general idea is that a program which uses just setreuid() will be
1130 * 100% compatible with BSD. A program which uses just setuid() will be
1131 * 100% compatible with POSIX with saved IDs.
1133 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
1135 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
1138 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
1142 new_ruid = old_ruid = current->uid;
1143 new_euid = old_euid = current->euid;
1144 old_suid = current->suid;
1146 if (ruid != (uid_t) -1) {
1148 if ((old_ruid != ruid) &&
1149 (current->euid != ruid) &&
1150 !capable(CAP_SETUID))
1154 if (euid != (uid_t) -1) {
1156 if ((old_ruid != euid) &&
1157 (current->euid != euid) &&
1158 (current->suid != euid) &&
1159 !capable(CAP_SETUID))
1163 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
1166 if (new_euid != old_euid) {
1167 set_dumpable(current->mm, suid_dumpable);
1170 current->fsuid = current->euid = new_euid;
1171 if (ruid != (uid_t) -1 ||
1172 (euid != (uid_t) -1 && euid != old_ruid))
1173 current->suid = current->euid;
1174 current->fsuid = current->euid;
1176 key_fsuid_changed(current);
1177 proc_id_connector(current, PROC_EVENT_UID);
1179 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
1185 * setuid() is implemented like SysV with SAVED_IDS
1187 * Note that SAVED_ID's is deficient in that a setuid root program
1188 * like sendmail, for example, cannot set its uid to be a normal
1189 * user and then switch back, because if you're root, setuid() sets
1190 * the saved uid too. If you don't like this, blame the bright people
1191 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
1192 * will allow a root program to temporarily drop privileges and be able to
1193 * regain them by swapping the real and effective uid.
1195 asmlinkage long sys_setuid(uid_t uid)
1197 int old_euid = current->euid;
1198 int old_ruid, old_suid, new_suid;
1201 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
1205 old_ruid = current->uid;
1206 old_suid = current->suid;
1207 new_suid = old_suid;
1209 if (capable(CAP_SETUID)) {
1210 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
1213 } else if ((uid != current->uid) && (uid != new_suid))
1216 if (old_euid != uid) {
1217 set_dumpable(current->mm, suid_dumpable);
1220 current->fsuid = current->euid = uid;
1221 current->suid = new_suid;
1223 key_fsuid_changed(current);
1224 proc_id_connector(current, PROC_EVENT_UID);
1226 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
1231 * This function implements a generic ability to update ruid, euid,
1232 * and suid. This allows you to implement the 4.4 compatible seteuid().
1234 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
1236 int old_ruid = current->uid;
1237 int old_euid = current->euid;
1238 int old_suid = current->suid;
1241 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
1245 if (!capable(CAP_SETUID)) {
1246 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
1247 (ruid != current->euid) && (ruid != current->suid))
1249 if ((euid != (uid_t) -1) && (euid != current->uid) &&
1250 (euid != current->euid) && (euid != current->suid))
1252 if ((suid != (uid_t) -1) && (suid != current->uid) &&
1253 (suid != current->euid) && (suid != current->suid))
1256 if (ruid != (uid_t) -1) {
1257 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
1260 if (euid != (uid_t) -1) {
1261 if (euid != current->euid) {
1262 set_dumpable(current->mm, suid_dumpable);
1265 current->euid = euid;
1267 current->fsuid = current->euid;
1268 if (suid != (uid_t) -1)
1269 current->suid = suid;
1271 key_fsuid_changed(current);
1272 proc_id_connector(current, PROC_EVENT_UID);
1274 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
1277 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
1281 if (!(retval = put_user(current->uid, ruid)) &&
1282 !(retval = put_user(current->euid, euid)))
1283 retval = put_user(current->suid, suid);
1289 * Same as above, but for rgid, egid, sgid.
1291 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
1295 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
1299 if (!capable(CAP_SETGID)) {
1300 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
1301 (rgid != current->egid) && (rgid != current->sgid))
1303 if ((egid != (gid_t) -1) && (egid != current->gid) &&
1304 (egid != current->egid) && (egid != current->sgid))
1306 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
1307 (sgid != current->egid) && (sgid != current->sgid))
1310 if (egid != (gid_t) -1) {
1311 if (egid != current->egid) {
1312 set_dumpable(current->mm, suid_dumpable);
1315 current->egid = egid;
1317 current->fsgid = current->egid;
1318 if (rgid != (gid_t) -1)
1319 current->gid = rgid;
1320 if (sgid != (gid_t) -1)
1321 current->sgid = sgid;
1323 key_fsgid_changed(current);
1324 proc_id_connector(current, PROC_EVENT_GID);
1328 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
1332 if (!(retval = put_user(current->gid, rgid)) &&
1333 !(retval = put_user(current->egid, egid)))
1334 retval = put_user(current->sgid, sgid);
1341 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
1342 * is used for "access()" and for the NFS daemon (letting nfsd stay at
1343 * whatever uid it wants to). It normally shadows "euid", except when
1344 * explicitly set by setfsuid() or for access..
1346 asmlinkage long sys_setfsuid(uid_t uid)
1350 old_fsuid = current->fsuid;
1351 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
1354 if (uid == current->uid || uid == current->euid ||
1355 uid == current->suid || uid == current->fsuid ||
1356 capable(CAP_SETUID)) {
1357 if (uid != old_fsuid) {
1358 set_dumpable(current->mm, suid_dumpable);
1361 current->fsuid = uid;
1364 key_fsuid_changed(current);
1365 proc_id_connector(current, PROC_EVENT_UID);
1367 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
1373 * Samma på svenska..
1375 asmlinkage long sys_setfsgid(gid_t gid)
1379 old_fsgid = current->fsgid;
1380 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
1383 if (gid == current->gid || gid == current->egid ||
1384 gid == current->sgid || gid == current->fsgid ||
1385 capable(CAP_SETGID)) {
1386 if (gid != old_fsgid) {
1387 set_dumpable(current->mm, suid_dumpable);
1390 current->fsgid = gid;
1391 key_fsgid_changed(current);
1392 proc_id_connector(current, PROC_EVENT_GID);
1397 asmlinkage long sys_times(struct tms __user * tbuf)
1400 * In the SMP world we might just be unlucky and have one of
1401 * the times increment as we use it. Since the value is an
1402 * atomically safe type this is just fine. Conceptually its
1403 * as if the syscall took an instant longer to occur.
1407 struct task_struct *tsk = current;
1408 struct task_struct *t;
1409 cputime_t utime, stime, cutime, cstime;
1411 spin_lock_irq(&tsk->sighand->siglock);
1412 utime = tsk->signal->utime;
1413 stime = tsk->signal->stime;
1416 utime = cputime_add(utime, t->utime);
1417 stime = cputime_add(stime, t->stime);
1421 cutime = tsk->signal->cutime;
1422 cstime = tsk->signal->cstime;
1423 spin_unlock_irq(&tsk->sighand->siglock);
1425 tmp.tms_utime = cputime_to_clock_t(utime);
1426 tmp.tms_stime = cputime_to_clock_t(stime);
1427 tmp.tms_cutime = cputime_to_clock_t(cutime);
1428 tmp.tms_cstime = cputime_to_clock_t(cstime);
1429 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1432 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1436 * This needs some heavy checking ...
1437 * I just haven't the stomach for it. I also don't fully
1438 * understand sessions/pgrp etc. Let somebody who does explain it.
1440 * OK, I think I have the protection semantics right.... this is really
1441 * only important on a multi-user system anyway, to make sure one user
1442 * can't send a signal to a process owned by another. -TYT, 12/12/91
1444 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1447 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1449 struct task_struct *p;
1450 struct task_struct *group_leader = current->group_leader;
1454 pid = group_leader->pid;
1460 /* From this point forward we keep holding onto the tasklist lock
1461 * so that our parent does not change from under us. -DaveM
1463 write_lock_irq(&tasklist_lock);
1466 p = find_task_by_pid(pid);
1471 if (!thread_group_leader(p))
1474 if (p->real_parent->tgid == group_leader->tgid) {
1476 if (task_session(p) != task_session(group_leader))
1483 if (p != group_leader)
1488 if (p->signal->leader)
1492 struct task_struct *g =
1493 find_task_by_pid_type(PIDTYPE_PGID, pgid);
1495 if (!g || task_session(g) != task_session(group_leader))
1499 err = security_task_setpgid(p, pgid);
1503 if (process_group(p) != pgid) {
1504 detach_pid(p, PIDTYPE_PGID);
1505 p->signal->pgrp = pgid;
1506 attach_pid(p, PIDTYPE_PGID, find_pid(pgid));
1511 /* All paths lead to here, thus we are safe. -DaveM */
1512 write_unlock_irq(&tasklist_lock);
1516 asmlinkage long sys_getpgid(pid_t pid)
1519 return process_group(current);
1522 struct task_struct *p;
1524 read_lock(&tasklist_lock);
1525 p = find_task_by_pid(pid);
1529 retval = security_task_getpgid(p);
1531 retval = process_group(p);
1533 read_unlock(&tasklist_lock);
1538 #ifdef __ARCH_WANT_SYS_GETPGRP
1540 asmlinkage long sys_getpgrp(void)
1542 /* SMP - assuming writes are word atomic this is fine */
1543 return process_group(current);
1548 asmlinkage long sys_getsid(pid_t pid)
1551 return process_session(current);
1554 struct task_struct *p;
1556 read_lock(&tasklist_lock);
1557 p = find_task_by_pid(pid);
1561 retval = security_task_getsid(p);
1563 retval = process_session(p);
1565 read_unlock(&tasklist_lock);
1570 asmlinkage long sys_setsid(void)
1572 struct task_struct *group_leader = current->group_leader;
1576 write_lock_irq(&tasklist_lock);
1578 /* Fail if I am already a session leader */
1579 if (group_leader->signal->leader)
1582 session = group_leader->pid;
1583 /* Fail if a process group id already exists that equals the
1584 * proposed session id.
1586 * Don't check if session id == 1 because kernel threads use this
1587 * session id and so the check will always fail and make it so
1588 * init cannot successfully call setsid.
1590 if (session > 1 && find_task_by_pid_type(PIDTYPE_PGID, session))
1593 group_leader->signal->leader = 1;
1594 __set_special_pids(session, session);
1596 spin_lock(&group_leader->sighand->siglock);
1597 group_leader->signal->tty = NULL;
1598 spin_unlock(&group_leader->sighand->siglock);
1600 err = process_group(group_leader);
1602 write_unlock_irq(&tasklist_lock);
1607 * Supplementary group IDs
1610 /* init to 2 - one for init_task, one to ensure it is never freed */
1611 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1613 struct group_info *groups_alloc(int gidsetsize)
1615 struct group_info *group_info;
1619 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1620 /* Make sure we always allocate at least one indirect block pointer */
1621 nblocks = nblocks ? : 1;
1622 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1625 group_info->ngroups = gidsetsize;
1626 group_info->nblocks = nblocks;
1627 atomic_set(&group_info->usage, 1);
1629 if (gidsetsize <= NGROUPS_SMALL)
1630 group_info->blocks[0] = group_info->small_block;
1632 for (i = 0; i < nblocks; i++) {
1634 b = (void *)__get_free_page(GFP_USER);
1636 goto out_undo_partial_alloc;
1637 group_info->blocks[i] = b;
1642 out_undo_partial_alloc:
1644 free_page((unsigned long)group_info->blocks[i]);
1650 EXPORT_SYMBOL(groups_alloc);
1652 void groups_free(struct group_info *group_info)
1654 if (group_info->blocks[0] != group_info->small_block) {
1656 for (i = 0; i < group_info->nblocks; i++)
1657 free_page((unsigned long)group_info->blocks[i]);
1662 EXPORT_SYMBOL(groups_free);
1664 /* export the group_info to a user-space array */
1665 static int groups_to_user(gid_t __user *grouplist,
1666 struct group_info *group_info)
1669 int count = group_info->ngroups;
1671 for (i = 0; i < group_info->nblocks; i++) {
1672 int cp_count = min(NGROUPS_PER_BLOCK, count);
1673 int off = i * NGROUPS_PER_BLOCK;
1674 int len = cp_count * sizeof(*grouplist);
1676 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1684 /* fill a group_info from a user-space array - it must be allocated already */
1685 static int groups_from_user(struct group_info *group_info,
1686 gid_t __user *grouplist)
1689 int count = group_info->ngroups;
1691 for (i = 0; i < group_info->nblocks; i++) {
1692 int cp_count = min(NGROUPS_PER_BLOCK, count);
1693 int off = i * NGROUPS_PER_BLOCK;
1694 int len = cp_count * sizeof(*grouplist);
1696 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1704 /* a simple Shell sort */
1705 static void groups_sort(struct group_info *group_info)
1707 int base, max, stride;
1708 int gidsetsize = group_info->ngroups;
1710 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1715 max = gidsetsize - stride;
1716 for (base = 0; base < max; base++) {
1718 int right = left + stride;
1719 gid_t tmp = GROUP_AT(group_info, right);
1721 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1722 GROUP_AT(group_info, right) =
1723 GROUP_AT(group_info, left);
1727 GROUP_AT(group_info, right) = tmp;
1733 /* a simple bsearch */
1734 int groups_search(struct group_info *group_info, gid_t grp)
1736 unsigned int left, right;
1742 right = group_info->ngroups;
1743 while (left < right) {
1744 unsigned int mid = (left+right)/2;
1745 int cmp = grp - GROUP_AT(group_info, mid);
1756 /* validate and set current->group_info */
1757 int set_current_groups(struct group_info *group_info)
1760 struct group_info *old_info;
1762 retval = security_task_setgroups(group_info);
1766 groups_sort(group_info);
1767 get_group_info(group_info);
1770 old_info = current->group_info;
1771 current->group_info = group_info;
1772 task_unlock(current);
1774 put_group_info(old_info);
1779 EXPORT_SYMBOL(set_current_groups);
1781 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1786 * SMP: Nobody else can change our grouplist. Thus we are
1793 /* no need to grab task_lock here; it cannot change */
1794 i = current->group_info->ngroups;
1796 if (i > gidsetsize) {
1800 if (groups_to_user(grouplist, current->group_info)) {
1810 * SMP: Our groups are copy-on-write. We can set them safely
1811 * without another task interfering.
1814 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1816 struct group_info *group_info;
1819 if (!capable(CAP_SETGID))
1821 if ((unsigned)gidsetsize > NGROUPS_MAX)
1824 group_info = groups_alloc(gidsetsize);
1827 retval = groups_from_user(group_info, grouplist);
1829 put_group_info(group_info);
1833 retval = set_current_groups(group_info);
1834 put_group_info(group_info);
1840 * Check whether we're fsgid/egid or in the supplemental group..
1842 int in_group_p(gid_t grp)
1845 if (grp != current->fsgid)
1846 retval = groups_search(current->group_info, grp);
1850 EXPORT_SYMBOL(in_group_p);
1852 int in_egroup_p(gid_t grp)
1855 if (grp != current->egid)
1856 retval = groups_search(current->group_info, grp);
1860 EXPORT_SYMBOL(in_egroup_p);
1862 DECLARE_RWSEM(uts_sem);
1864 EXPORT_SYMBOL(uts_sem);
1866 asmlinkage long sys_newuname(struct new_utsname __user * name)
1870 down_read(&uts_sem);
1871 if (copy_to_user(name, utsname(), sizeof *name))
1877 asmlinkage long sys_sethostname(char __user *name, int len)
1880 char tmp[__NEW_UTS_LEN];
1882 if (!capable(CAP_SYS_ADMIN))
1884 if (len < 0 || len > __NEW_UTS_LEN)
1886 down_write(&uts_sem);
1888 if (!copy_from_user(tmp, name, len)) {
1889 memcpy(utsname()->nodename, tmp, len);
1890 utsname()->nodename[len] = 0;
1897 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1899 asmlinkage long sys_gethostname(char __user *name, int len)
1905 down_read(&uts_sem);
1906 i = 1 + strlen(utsname()->nodename);
1910 if (copy_to_user(name, utsname()->nodename, i))
1919 * Only setdomainname; getdomainname can be implemented by calling
1922 asmlinkage long sys_setdomainname(char __user *name, int len)
1925 char tmp[__NEW_UTS_LEN];
1927 if (!capable(CAP_SYS_ADMIN))
1929 if (len < 0 || len > __NEW_UTS_LEN)
1932 down_write(&uts_sem);
1934 if (!copy_from_user(tmp, name, len)) {
1935 memcpy(utsname()->domainname, tmp, len);
1936 utsname()->domainname[len] = 0;
1943 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1945 if (resource >= RLIM_NLIMITS)
1948 struct rlimit value;
1949 task_lock(current->group_leader);
1950 value = current->signal->rlim[resource];
1951 task_unlock(current->group_leader);
1952 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1956 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1959 * Back compatibility for getrlimit. Needed for some apps.
1962 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1965 if (resource >= RLIM_NLIMITS)
1968 task_lock(current->group_leader);
1969 x = current->signal->rlim[resource];
1970 task_unlock(current->group_leader);
1971 if (x.rlim_cur > 0x7FFFFFFF)
1972 x.rlim_cur = 0x7FFFFFFF;
1973 if (x.rlim_max > 0x7FFFFFFF)
1974 x.rlim_max = 0x7FFFFFFF;
1975 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1980 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1982 struct rlimit new_rlim, *old_rlim;
1983 unsigned long it_prof_secs;
1986 if (resource >= RLIM_NLIMITS)
1988 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1990 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1992 old_rlim = current->signal->rlim + resource;
1993 if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1994 !capable(CAP_SYS_RESOURCE))
1996 if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1999 retval = security_task_setrlimit(resource, &new_rlim);
2003 if (resource == RLIMIT_CPU && new_rlim.rlim_cur == 0) {
2005 * The caller is asking for an immediate RLIMIT_CPU
2006 * expiry. But we use the zero value to mean "it was
2007 * never set". So let's cheat and make it one second
2010 new_rlim.rlim_cur = 1;
2013 task_lock(current->group_leader);
2014 *old_rlim = new_rlim;
2015 task_unlock(current->group_leader);
2017 if (resource != RLIMIT_CPU)
2021 * RLIMIT_CPU handling. Note that the kernel fails to return an error
2022 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
2023 * very long-standing error, and fixing it now risks breakage of
2024 * applications, so we live with it
2026 if (new_rlim.rlim_cur == RLIM_INFINITY)
2029 it_prof_secs = cputime_to_secs(current->signal->it_prof_expires);
2030 if (it_prof_secs == 0 || new_rlim.rlim_cur <= it_prof_secs) {
2031 unsigned long rlim_cur = new_rlim.rlim_cur;
2034 cputime = secs_to_cputime(rlim_cur);
2035 read_lock(&tasklist_lock);
2036 spin_lock_irq(¤t->sighand->siglock);
2037 set_process_cpu_timer(current, CPUCLOCK_PROF, &cputime, NULL);
2038 spin_unlock_irq(¤t->sighand->siglock);
2039 read_unlock(&tasklist_lock);
2046 * It would make sense to put struct rusage in the task_struct,
2047 * except that would make the task_struct be *really big*. After
2048 * task_struct gets moved into malloc'ed memory, it would
2049 * make sense to do this. It will make moving the rest of the information
2050 * a lot simpler! (Which we're not doing right now because we're not
2051 * measuring them yet).
2053 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
2054 * races with threads incrementing their own counters. But since word
2055 * reads are atomic, we either get new values or old values and we don't
2056 * care which for the sums. We always take the siglock to protect reading
2057 * the c* fields from p->signal from races with exit.c updating those
2058 * fields when reaping, so a sample either gets all the additions of a
2059 * given child after it's reaped, or none so this sample is before reaping.
2062 * We need to take the siglock for CHILDEREN, SELF and BOTH
2063 * for the cases current multithreaded, non-current single threaded
2064 * non-current multithreaded. Thread traversal is now safe with
2066 * Strictly speaking, we donot need to take the siglock if we are current and
2067 * single threaded, as no one else can take our signal_struct away, no one
2068 * else can reap the children to update signal->c* counters, and no one else
2069 * can race with the signal-> fields. If we do not take any lock, the
2070 * signal-> fields could be read out of order while another thread was just
2071 * exiting. So we should place a read memory barrier when we avoid the lock.
2072 * On the writer side, write memory barrier is implied in __exit_signal
2073 * as __exit_signal releases the siglock spinlock after updating the signal->
2074 * fields. But we don't do this yet to keep things simple.
2078 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
2080 struct task_struct *t;
2081 unsigned long flags;
2082 cputime_t utime, stime;
2084 memset((char *) r, 0, sizeof *r);
2085 utime = stime = cputime_zero;
2088 if (!lock_task_sighand(p, &flags)) {
2095 case RUSAGE_CHILDREN:
2096 utime = p->signal->cutime;
2097 stime = p->signal->cstime;
2098 r->ru_nvcsw = p->signal->cnvcsw;
2099 r->ru_nivcsw = p->signal->cnivcsw;
2100 r->ru_minflt = p->signal->cmin_flt;
2101 r->ru_majflt = p->signal->cmaj_flt;
2102 r->ru_inblock = p->signal->cinblock;
2103 r->ru_oublock = p->signal->coublock;
2105 if (who == RUSAGE_CHILDREN)
2109 utime = cputime_add(utime, p->signal->utime);
2110 stime = cputime_add(stime, p->signal->stime);
2111 r->ru_nvcsw += p->signal->nvcsw;
2112 r->ru_nivcsw += p->signal->nivcsw;
2113 r->ru_minflt += p->signal->min_flt;
2114 r->ru_majflt += p->signal->maj_flt;
2115 r->ru_inblock += p->signal->inblock;
2116 r->ru_oublock += p->signal->oublock;
2119 utime = cputime_add(utime, t->utime);
2120 stime = cputime_add(stime, t->stime);
2121 r->ru_nvcsw += t->nvcsw;
2122 r->ru_nivcsw += t->nivcsw;
2123 r->ru_minflt += t->min_flt;
2124 r->ru_majflt += t->maj_flt;
2125 r->ru_inblock += task_io_get_inblock(t);
2126 r->ru_oublock += task_io_get_oublock(t);
2135 unlock_task_sighand(p, &flags);
2138 cputime_to_timeval(utime, &r->ru_utime);
2139 cputime_to_timeval(stime, &r->ru_stime);
2142 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
2145 k_getrusage(p, who, &r);
2146 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
2149 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
2151 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
2153 return getrusage(current, who, ru);
2156 asmlinkage long sys_umask(int mask)
2158 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
2162 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
2163 unsigned long arg4, unsigned long arg5)
2167 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2172 case PR_SET_PDEATHSIG:
2173 if (!valid_signal(arg2)) {
2177 current->pdeath_signal = arg2;
2179 case PR_GET_PDEATHSIG:
2180 error = put_user(current->pdeath_signal, (int __user *)arg2);
2182 case PR_GET_DUMPABLE:
2183 error = get_dumpable(current->mm);
2185 case PR_SET_DUMPABLE:
2186 if (arg2 < 0 || arg2 > 1) {
2190 set_dumpable(current->mm, arg2);
2193 case PR_SET_UNALIGN:
2194 error = SET_UNALIGN_CTL(current, arg2);
2196 case PR_GET_UNALIGN:
2197 error = GET_UNALIGN_CTL(current, arg2);
2200 error = SET_FPEMU_CTL(current, arg2);
2203 error = GET_FPEMU_CTL(current, arg2);
2206 error = SET_FPEXC_CTL(current, arg2);
2209 error = GET_FPEXC_CTL(current, arg2);
2212 error = PR_TIMING_STATISTICAL;
2215 if (arg2 == PR_TIMING_STATISTICAL)
2221 case PR_GET_KEEPCAPS:
2222 if (current->keep_capabilities)
2225 case PR_SET_KEEPCAPS:
2226 if (arg2 != 0 && arg2 != 1) {
2230 current->keep_capabilities = arg2;
2233 struct task_struct *me = current;
2234 unsigned char ncomm[sizeof(me->comm)];
2236 ncomm[sizeof(me->comm)-1] = 0;
2237 if (strncpy_from_user(ncomm, (char __user *)arg2,
2238 sizeof(me->comm)-1) < 0)
2240 set_task_comm(me, ncomm);
2244 struct task_struct *me = current;
2245 unsigned char tcomm[sizeof(me->comm)];
2247 get_task_comm(tcomm, me);
2248 if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))
2253 error = GET_ENDIAN(current, arg2);
2256 error = SET_ENDIAN(current, arg2);
2259 case PR_GET_SECCOMP:
2260 error = prctl_get_seccomp();
2262 case PR_SET_SECCOMP:
2263 error = prctl_set_seccomp(arg2);
2273 asmlinkage long sys_getcpu(unsigned __user *cpup, unsigned __user *nodep,
2274 struct getcpu_cache __user *cache)
2277 int cpu = raw_smp_processor_id();
2279 err |= put_user(cpu, cpup);
2281 err |= put_user(cpu_to_node(cpu), nodep);
2284 * The cache is not needed for this implementation,
2285 * but make sure user programs pass something
2286 * valid. vsyscall implementations can instead make
2287 * good use of the cache. Only use t0 and t1 because
2288 * these are available in both 32bit and 64bit ABI (no
2289 * need for a compat_getcpu). 32bit has enough
2292 unsigned long t0, t1;
2293 get_user(t0, &cache->blob[0]);
2294 get_user(t1, &cache->blob[1]);
2297 put_user(t0, &cache->blob[0]);
2298 put_user(t1, &cache->blob[1]);
2300 return err ? -EFAULT : 0;
2303 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
2305 static void argv_cleanup(char **argv, char **envp)
2311 * orderly_poweroff - Trigger an orderly system poweroff
2312 * @force: force poweroff if command execution fails
2314 * This may be called from any context to trigger a system shutdown.
2315 * If the orderly shutdown fails, it will force an immediate shutdown.
2317 int orderly_poweroff(bool force)
2320 char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
2321 static char *envp[] = {
2323 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
2327 struct subprocess_info *info;
2330 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
2331 __func__, poweroff_cmd);
2335 info = call_usermodehelper_setup(argv[0], argv, envp);
2341 call_usermodehelper_setcleanup(info, argv_cleanup);
2343 ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
2347 printk(KERN_WARNING "Failed to start orderly shutdown: "
2348 "forcing the issue\n");
2350 /* I guess this should try to kick off some daemon to
2351 sync and poweroff asap. Or not even bother syncing
2352 if we're doing an emergency shutdown? */
2359 EXPORT_SYMBOL_GPL(orderly_poweroff);