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>
36 #include <linux/compat.h>
37 #include <linux/syscalls.h>
38 #include <linux/kprobes.h>
39 #include <linux/user_namespace.h>
41 #include <asm/uaccess.h>
43 #include <asm/unistd.h>
45 #ifndef SET_UNALIGN_CTL
46 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
48 #ifndef GET_UNALIGN_CTL
49 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
52 # define SET_FPEMU_CTL(a,b) (-EINVAL)
55 # define GET_FPEMU_CTL(a,b) (-EINVAL)
58 # define SET_FPEXC_CTL(a,b) (-EINVAL)
61 # define GET_FPEXC_CTL(a,b) (-EINVAL)
64 # define GET_ENDIAN(a,b) (-EINVAL)
67 # define SET_ENDIAN(a,b) (-EINVAL)
71 * this is where the system-wide overflow UID and GID are defined, for
72 * architectures that now have 32-bit UID/GID but didn't in the past
75 int overflowuid = DEFAULT_OVERFLOWUID;
76 int overflowgid = DEFAULT_OVERFLOWGID;
79 EXPORT_SYMBOL(overflowuid);
80 EXPORT_SYMBOL(overflowgid);
84 * the same as above, but for filesystems which can only store a 16-bit
85 * UID and GID. as such, this is needed on all architectures
88 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
89 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
91 EXPORT_SYMBOL(fs_overflowuid);
92 EXPORT_SYMBOL(fs_overflowgid);
95 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
100 EXPORT_SYMBOL(cad_pid);
103 * If set, this is used for preparing the system to power off.
106 void (*pm_power_off_prepare)(void);
107 EXPORT_SYMBOL(pm_power_off_prepare);
110 * Notifier list for kernel code which wants to be called
111 * at shutdown. This is used to stop any idling DMA operations
115 static BLOCKING_NOTIFIER_HEAD(reboot_notifier_list);
118 * Notifier chain core routines. The exported routines below
119 * are layered on top of these, with appropriate locking added.
122 static int notifier_chain_register(struct notifier_block **nl,
123 struct notifier_block *n)
125 while ((*nl) != NULL) {
126 if (n->priority > (*nl)->priority)
131 rcu_assign_pointer(*nl, n);
135 static int notifier_chain_unregister(struct notifier_block **nl,
136 struct notifier_block *n)
138 while ((*nl) != NULL) {
140 rcu_assign_pointer(*nl, n->next);
149 * notifier_call_chain - Informs the registered notifiers about an event.
150 * @nl: Pointer to head of the blocking notifier chain
151 * @val: Value passed unmodified to notifier function
152 * @v: Pointer passed unmodified to notifier function
153 * @nr_to_call: Number of notifier functions to be called. Don't care
154 * value of this parameter is -1.
155 * @nr_calls: Records the number of notifications sent. Don't care
156 * value of this field is NULL.
157 * @returns: notifier_call_chain returns the value returned by the
158 * last notifier function called.
161 static int __kprobes notifier_call_chain(struct notifier_block **nl,
162 unsigned long val, void *v,
163 int nr_to_call, int *nr_calls)
165 int ret = NOTIFY_DONE;
166 struct notifier_block *nb, *next_nb;
168 nb = rcu_dereference(*nl);
170 while (nb && nr_to_call) {
171 next_nb = rcu_dereference(nb->next);
172 ret = nb->notifier_call(nb, val, v);
177 if ((ret & NOTIFY_STOP_MASK) == NOTIFY_STOP_MASK)
186 * Atomic notifier chain routines. Registration and unregistration
187 * use a spinlock, and call_chain is synchronized by RCU (no locks).
191 * atomic_notifier_chain_register - Add notifier to an atomic notifier chain
192 * @nh: Pointer to head of the atomic notifier chain
193 * @n: New entry in notifier chain
195 * Adds a notifier to an atomic notifier chain.
197 * Currently always returns zero.
200 int atomic_notifier_chain_register(struct atomic_notifier_head *nh,
201 struct notifier_block *n)
206 spin_lock_irqsave(&nh->lock, flags);
207 ret = notifier_chain_register(&nh->head, n);
208 spin_unlock_irqrestore(&nh->lock, flags);
212 EXPORT_SYMBOL_GPL(atomic_notifier_chain_register);
215 * atomic_notifier_chain_unregister - Remove notifier from an atomic notifier chain
216 * @nh: Pointer to head of the atomic notifier chain
217 * @n: Entry to remove from notifier chain
219 * Removes a notifier from an atomic notifier chain.
221 * Returns zero on success or %-ENOENT on failure.
223 int atomic_notifier_chain_unregister(struct atomic_notifier_head *nh,
224 struct notifier_block *n)
229 spin_lock_irqsave(&nh->lock, flags);
230 ret = notifier_chain_unregister(&nh->head, n);
231 spin_unlock_irqrestore(&nh->lock, flags);
236 EXPORT_SYMBOL_GPL(atomic_notifier_chain_unregister);
239 * __atomic_notifier_call_chain - Call functions in an atomic notifier chain
240 * @nh: Pointer to head of the atomic notifier chain
241 * @val: Value passed unmodified to notifier function
242 * @v: Pointer passed unmodified to notifier function
243 * @nr_to_call: See the comment for notifier_call_chain.
244 * @nr_calls: See the comment for notifier_call_chain.
246 * Calls each function in a notifier chain in turn. The functions
247 * run in an atomic context, so they must not block.
248 * This routine uses RCU to synchronize with changes to the chain.
250 * If the return value of the notifier can be and'ed
251 * with %NOTIFY_STOP_MASK then atomic_notifier_call_chain()
252 * will return immediately, with the return value of
253 * the notifier function which halted execution.
254 * Otherwise the return value is the return value
255 * of the last notifier function called.
258 int __kprobes __atomic_notifier_call_chain(struct atomic_notifier_head *nh,
259 unsigned long val, void *v,
260 int nr_to_call, int *nr_calls)
265 ret = notifier_call_chain(&nh->head, val, v, nr_to_call, nr_calls);
270 EXPORT_SYMBOL_GPL(__atomic_notifier_call_chain);
272 int __kprobes atomic_notifier_call_chain(struct atomic_notifier_head *nh,
273 unsigned long val, void *v)
275 return __atomic_notifier_call_chain(nh, val, v, -1, NULL);
278 EXPORT_SYMBOL_GPL(atomic_notifier_call_chain);
280 * Blocking notifier chain routines. All access to the chain is
281 * synchronized by an rwsem.
285 * blocking_notifier_chain_register - Add notifier to a blocking notifier chain
286 * @nh: Pointer to head of the blocking notifier chain
287 * @n: New entry in notifier chain
289 * Adds a notifier to a blocking notifier chain.
290 * Must be called in process context.
292 * Currently always returns zero.
295 int blocking_notifier_chain_register(struct blocking_notifier_head *nh,
296 struct notifier_block *n)
301 * This code gets used during boot-up, when task switching is
302 * not yet working and interrupts must remain disabled. At
303 * such times we must not call down_write().
305 if (unlikely(system_state == SYSTEM_BOOTING))
306 return notifier_chain_register(&nh->head, n);
308 down_write(&nh->rwsem);
309 ret = notifier_chain_register(&nh->head, n);
310 up_write(&nh->rwsem);
314 EXPORT_SYMBOL_GPL(blocking_notifier_chain_register);
317 * blocking_notifier_chain_unregister - Remove notifier from a blocking notifier chain
318 * @nh: Pointer to head of the blocking notifier chain
319 * @n: Entry to remove from notifier chain
321 * Removes a notifier from a blocking notifier chain.
322 * Must be called from process context.
324 * Returns zero on success or %-ENOENT on failure.
326 int blocking_notifier_chain_unregister(struct blocking_notifier_head *nh,
327 struct notifier_block *n)
332 * This code gets used during boot-up, when task switching is
333 * not yet working and interrupts must remain disabled. At
334 * such times we must not call down_write().
336 if (unlikely(system_state == SYSTEM_BOOTING))
337 return notifier_chain_unregister(&nh->head, n);
339 down_write(&nh->rwsem);
340 ret = notifier_chain_unregister(&nh->head, n);
341 up_write(&nh->rwsem);
345 EXPORT_SYMBOL_GPL(blocking_notifier_chain_unregister);
348 * __blocking_notifier_call_chain - Call functions in a blocking notifier chain
349 * @nh: Pointer to head of the blocking notifier chain
350 * @val: Value passed unmodified to notifier function
351 * @v: Pointer passed unmodified to notifier function
352 * @nr_to_call: See comment for notifier_call_chain.
353 * @nr_calls: See comment for notifier_call_chain.
355 * Calls each function in a notifier chain in turn. The functions
356 * run in a process context, so they are allowed to block.
358 * If the return value of the notifier can be and'ed
359 * with %NOTIFY_STOP_MASK then blocking_notifier_call_chain()
360 * will return immediately, with the return value of
361 * the notifier function which halted execution.
362 * Otherwise the return value is the return value
363 * of the last notifier function called.
366 int __blocking_notifier_call_chain(struct blocking_notifier_head *nh,
367 unsigned long val, void *v,
368 int nr_to_call, int *nr_calls)
370 int ret = NOTIFY_DONE;
373 * We check the head outside the lock, but if this access is
374 * racy then it does not matter what the result of the test
375 * is, we re-check the list after having taken the lock anyway:
377 if (rcu_dereference(nh->head)) {
378 down_read(&nh->rwsem);
379 ret = notifier_call_chain(&nh->head, val, v, nr_to_call,
385 EXPORT_SYMBOL_GPL(__blocking_notifier_call_chain);
387 int blocking_notifier_call_chain(struct blocking_notifier_head *nh,
388 unsigned long val, void *v)
390 return __blocking_notifier_call_chain(nh, val, v, -1, NULL);
392 EXPORT_SYMBOL_GPL(blocking_notifier_call_chain);
395 * Raw notifier chain routines. There is no protection;
396 * the caller must provide it. Use at your own risk!
400 * raw_notifier_chain_register - Add notifier to a raw notifier chain
401 * @nh: Pointer to head of the raw notifier chain
402 * @n: New entry in notifier chain
404 * Adds a notifier to a raw notifier chain.
405 * All locking must be provided by the caller.
407 * Currently always returns zero.
410 int raw_notifier_chain_register(struct raw_notifier_head *nh,
411 struct notifier_block *n)
413 return notifier_chain_register(&nh->head, n);
416 EXPORT_SYMBOL_GPL(raw_notifier_chain_register);
419 * raw_notifier_chain_unregister - Remove notifier from a raw notifier chain
420 * @nh: Pointer to head of the raw notifier chain
421 * @n: Entry to remove from notifier chain
423 * Removes a notifier from a raw notifier chain.
424 * All locking must be provided by the caller.
426 * Returns zero on success or %-ENOENT on failure.
428 int raw_notifier_chain_unregister(struct raw_notifier_head *nh,
429 struct notifier_block *n)
431 return notifier_chain_unregister(&nh->head, n);
434 EXPORT_SYMBOL_GPL(raw_notifier_chain_unregister);
437 * __raw_notifier_call_chain - Call functions in a raw notifier chain
438 * @nh: Pointer to head of the raw notifier chain
439 * @val: Value passed unmodified to notifier function
440 * @v: Pointer passed unmodified to notifier function
441 * @nr_to_call: See comment for notifier_call_chain.
442 * @nr_calls: See comment for notifier_call_chain
444 * Calls each function in a notifier chain in turn. The functions
445 * run in an undefined context.
446 * All locking must be provided by the caller.
448 * If the return value of the notifier can be and'ed
449 * with %NOTIFY_STOP_MASK then raw_notifier_call_chain()
450 * will return immediately, with the return value of
451 * the notifier function which halted execution.
452 * Otherwise the return value is the return value
453 * of the last notifier function called.
456 int __raw_notifier_call_chain(struct raw_notifier_head *nh,
457 unsigned long val, void *v,
458 int nr_to_call, int *nr_calls)
460 return notifier_call_chain(&nh->head, val, v, nr_to_call, nr_calls);
463 EXPORT_SYMBOL_GPL(__raw_notifier_call_chain);
465 int raw_notifier_call_chain(struct raw_notifier_head *nh,
466 unsigned long val, void *v)
468 return __raw_notifier_call_chain(nh, val, v, -1, NULL);
471 EXPORT_SYMBOL_GPL(raw_notifier_call_chain);
474 * SRCU notifier chain routines. Registration and unregistration
475 * use a mutex, and call_chain is synchronized by SRCU (no locks).
479 * srcu_notifier_chain_register - Add notifier to an SRCU notifier chain
480 * @nh: Pointer to head of the SRCU notifier chain
481 * @n: New entry in notifier chain
483 * Adds a notifier to an SRCU notifier chain.
484 * Must be called in process context.
486 * Currently always returns zero.
489 int srcu_notifier_chain_register(struct srcu_notifier_head *nh,
490 struct notifier_block *n)
495 * This code gets used during boot-up, when task switching is
496 * not yet working and interrupts must remain disabled. At
497 * such times we must not call mutex_lock().
499 if (unlikely(system_state == SYSTEM_BOOTING))
500 return notifier_chain_register(&nh->head, n);
502 mutex_lock(&nh->mutex);
503 ret = notifier_chain_register(&nh->head, n);
504 mutex_unlock(&nh->mutex);
508 EXPORT_SYMBOL_GPL(srcu_notifier_chain_register);
511 * srcu_notifier_chain_unregister - Remove notifier from an SRCU notifier chain
512 * @nh: Pointer to head of the SRCU notifier chain
513 * @n: Entry to remove from notifier chain
515 * Removes a notifier from an SRCU notifier chain.
516 * Must be called from process context.
518 * Returns zero on success or %-ENOENT on failure.
520 int srcu_notifier_chain_unregister(struct srcu_notifier_head *nh,
521 struct notifier_block *n)
526 * This code gets used during boot-up, when task switching is
527 * not yet working and interrupts must remain disabled. At
528 * such times we must not call mutex_lock().
530 if (unlikely(system_state == SYSTEM_BOOTING))
531 return notifier_chain_unregister(&nh->head, n);
533 mutex_lock(&nh->mutex);
534 ret = notifier_chain_unregister(&nh->head, n);
535 mutex_unlock(&nh->mutex);
536 synchronize_srcu(&nh->srcu);
540 EXPORT_SYMBOL_GPL(srcu_notifier_chain_unregister);
543 * __srcu_notifier_call_chain - Call functions in an SRCU notifier chain
544 * @nh: Pointer to head of the SRCU notifier chain
545 * @val: Value passed unmodified to notifier function
546 * @v: Pointer passed unmodified to notifier function
547 * @nr_to_call: See comment for notifier_call_chain.
548 * @nr_calls: See comment for notifier_call_chain
550 * Calls each function in a notifier chain in turn. The functions
551 * run in a process context, so they are allowed to block.
553 * If the return value of the notifier can be and'ed
554 * with %NOTIFY_STOP_MASK then srcu_notifier_call_chain()
555 * will return immediately, with the return value of
556 * the notifier function which halted execution.
557 * Otherwise the return value is the return value
558 * of the last notifier function called.
561 int __srcu_notifier_call_chain(struct srcu_notifier_head *nh,
562 unsigned long val, void *v,
563 int nr_to_call, int *nr_calls)
568 idx = srcu_read_lock(&nh->srcu);
569 ret = notifier_call_chain(&nh->head, val, v, nr_to_call, nr_calls);
570 srcu_read_unlock(&nh->srcu, idx);
573 EXPORT_SYMBOL_GPL(__srcu_notifier_call_chain);
575 int srcu_notifier_call_chain(struct srcu_notifier_head *nh,
576 unsigned long val, void *v)
578 return __srcu_notifier_call_chain(nh, val, v, -1, NULL);
580 EXPORT_SYMBOL_GPL(srcu_notifier_call_chain);
583 * srcu_init_notifier_head - Initialize an SRCU notifier head
584 * @nh: Pointer to head of the srcu notifier chain
586 * Unlike other sorts of notifier heads, SRCU notifier heads require
587 * dynamic initialization. Be sure to call this routine before
588 * calling any of the other SRCU notifier routines for this head.
590 * If an SRCU notifier head is deallocated, it must first be cleaned
591 * up by calling srcu_cleanup_notifier_head(). Otherwise the head's
592 * per-cpu data (used by the SRCU mechanism) will leak.
595 void srcu_init_notifier_head(struct srcu_notifier_head *nh)
597 mutex_init(&nh->mutex);
598 if (init_srcu_struct(&nh->srcu) < 0)
603 EXPORT_SYMBOL_GPL(srcu_init_notifier_head);
606 * register_reboot_notifier - Register function to be called at reboot time
607 * @nb: Info about notifier function to be called
609 * Registers a function with the list of functions
610 * to be called at reboot time.
612 * Currently always returns zero, as blocking_notifier_chain_register()
613 * always returns zero.
616 int register_reboot_notifier(struct notifier_block * nb)
618 return blocking_notifier_chain_register(&reboot_notifier_list, nb);
621 EXPORT_SYMBOL(register_reboot_notifier);
624 * unregister_reboot_notifier - Unregister previously registered reboot notifier
625 * @nb: Hook to be unregistered
627 * Unregisters a previously registered reboot
630 * Returns zero on success, or %-ENOENT on failure.
633 int unregister_reboot_notifier(struct notifier_block * nb)
635 return blocking_notifier_chain_unregister(&reboot_notifier_list, nb);
638 EXPORT_SYMBOL(unregister_reboot_notifier);
640 static int set_one_prio(struct task_struct *p, int niceval, int error)
644 if (p->uid != current->euid &&
645 p->euid != current->euid && !capable(CAP_SYS_NICE)) {
649 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
653 no_nice = security_task_setnice(p, niceval);
660 set_user_nice(p, niceval);
665 asmlinkage long sys_setpriority(int which, int who, int niceval)
667 struct task_struct *g, *p;
668 struct user_struct *user;
672 if (which > PRIO_USER || which < PRIO_PROCESS)
675 /* normalize: avoid signed division (rounding problems) */
682 read_lock(&tasklist_lock);
686 p = find_task_by_pid(who);
690 error = set_one_prio(p, niceval, error);
694 pgrp = find_pid(who);
696 pgrp = task_pgrp(current);
697 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
698 error = set_one_prio(p, niceval, error);
699 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
702 user = current->user;
706 if ((who != current->uid) && !(user = find_user(who)))
707 goto out_unlock; /* No processes for this user */
711 error = set_one_prio(p, niceval, error);
712 while_each_thread(g, p);
713 if (who != current->uid)
714 free_uid(user); /* For find_user() */
718 read_unlock(&tasklist_lock);
724 * Ugh. To avoid negative return values, "getpriority()" will
725 * not return the normal nice-value, but a negated value that
726 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
727 * to stay compatible.
729 asmlinkage long sys_getpriority(int which, int who)
731 struct task_struct *g, *p;
732 struct user_struct *user;
733 long niceval, retval = -ESRCH;
736 if (which > PRIO_USER || which < PRIO_PROCESS)
739 read_lock(&tasklist_lock);
743 p = find_task_by_pid(who);
747 niceval = 20 - task_nice(p);
748 if (niceval > retval)
754 pgrp = find_pid(who);
756 pgrp = task_pgrp(current);
757 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
758 niceval = 20 - task_nice(p);
759 if (niceval > retval)
761 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
764 user = current->user;
768 if ((who != current->uid) && !(user = find_user(who)))
769 goto out_unlock; /* No processes for this user */
773 niceval = 20 - task_nice(p);
774 if (niceval > retval)
777 while_each_thread(g, p);
778 if (who != current->uid)
779 free_uid(user); /* for find_user() */
783 read_unlock(&tasklist_lock);
789 * emergency_restart - reboot the system
791 * Without shutting down any hardware or taking any locks
792 * reboot the system. This is called when we know we are in
793 * trouble so this is our best effort to reboot. This is
794 * safe to call in interrupt context.
796 void emergency_restart(void)
798 machine_emergency_restart();
800 EXPORT_SYMBOL_GPL(emergency_restart);
802 static void kernel_restart_prepare(char *cmd)
804 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
805 system_state = SYSTEM_RESTART;
811 * kernel_restart - reboot the system
812 * @cmd: pointer to buffer containing command to execute for restart
815 * Shutdown everything and perform a clean reboot.
816 * This is not safe to call in interrupt context.
818 void kernel_restart(char *cmd)
820 kernel_restart_prepare(cmd);
822 printk(KERN_EMERG "Restarting system.\n");
824 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
825 machine_restart(cmd);
827 EXPORT_SYMBOL_GPL(kernel_restart);
830 * kernel_kexec - reboot the system
832 * Move into place and start executing a preloaded standalone
833 * executable. If nothing was preloaded return an error.
835 static void kernel_kexec(void)
838 struct kimage *image;
839 image = xchg(&kexec_image, NULL);
842 kernel_restart_prepare(NULL);
843 printk(KERN_EMERG "Starting new kernel\n");
845 machine_kexec(image);
849 void kernel_shutdown_prepare(enum system_states state)
851 blocking_notifier_call_chain(&reboot_notifier_list,
852 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
853 system_state = state;
857 * kernel_halt - halt the system
859 * Shutdown everything and perform a clean system halt.
861 void kernel_halt(void)
863 kernel_shutdown_prepare(SYSTEM_HALT);
865 printk(KERN_EMERG "System halted.\n");
869 EXPORT_SYMBOL_GPL(kernel_halt);
872 * kernel_power_off - power_off the system
874 * Shutdown everything and perform a clean system power_off.
876 void kernel_power_off(void)
878 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
879 if (pm_power_off_prepare)
880 pm_power_off_prepare();
882 printk(KERN_EMERG "Power down.\n");
885 EXPORT_SYMBOL_GPL(kernel_power_off);
887 * Reboot system call: for obvious reasons only root may call it,
888 * and even root needs to set up some magic numbers in the registers
889 * so that some mistake won't make this reboot the whole machine.
890 * You can also set the meaning of the ctrl-alt-del-key here.
892 * reboot doesn't sync: do that yourself before calling this.
894 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
898 /* We only trust the superuser with rebooting the system. */
899 if (!capable(CAP_SYS_BOOT))
902 /* For safety, we require "magic" arguments. */
903 if (magic1 != LINUX_REBOOT_MAGIC1 ||
904 (magic2 != LINUX_REBOOT_MAGIC2 &&
905 magic2 != LINUX_REBOOT_MAGIC2A &&
906 magic2 != LINUX_REBOOT_MAGIC2B &&
907 magic2 != LINUX_REBOOT_MAGIC2C))
910 /* Instead of trying to make the power_off code look like
911 * halt when pm_power_off is not set do it the easy way.
913 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
914 cmd = LINUX_REBOOT_CMD_HALT;
918 case LINUX_REBOOT_CMD_RESTART:
919 kernel_restart(NULL);
922 case LINUX_REBOOT_CMD_CAD_ON:
926 case LINUX_REBOOT_CMD_CAD_OFF:
930 case LINUX_REBOOT_CMD_HALT:
936 case LINUX_REBOOT_CMD_POWER_OFF:
942 case LINUX_REBOOT_CMD_RESTART2:
943 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
947 buffer[sizeof(buffer) - 1] = '\0';
949 kernel_restart(buffer);
952 case LINUX_REBOOT_CMD_KEXEC:
957 #ifdef CONFIG_HIBERNATION
958 case LINUX_REBOOT_CMD_SW_SUSPEND:
960 int ret = hibernate();
974 static void deferred_cad(struct work_struct *dummy)
976 kernel_restart(NULL);
980 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
981 * As it's called within an interrupt, it may NOT sync: the only choice
982 * is whether to reboot at once, or just ignore the ctrl-alt-del.
984 void ctrl_alt_del(void)
986 static DECLARE_WORK(cad_work, deferred_cad);
989 schedule_work(&cad_work);
991 kill_cad_pid(SIGINT, 1);
995 * Unprivileged users may change the real gid to the effective gid
996 * or vice versa. (BSD-style)
998 * If you set the real gid at all, or set the effective gid to a value not
999 * equal to the real gid, then the saved gid is set to the new effective gid.
1001 * This makes it possible for a setgid program to completely drop its
1002 * privileges, which is often a useful assertion to make when you are doing
1003 * a security audit over a program.
1005 * The general idea is that a program which uses just setregid() will be
1006 * 100% compatible with BSD. A program which uses just setgid() will be
1007 * 100% compatible with POSIX with saved IDs.
1009 * SMP: There are not races, the GIDs are checked only by filesystem
1010 * operations (as far as semantic preservation is concerned).
1012 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
1014 int old_rgid = current->gid;
1015 int old_egid = current->egid;
1016 int new_rgid = old_rgid;
1017 int new_egid = old_egid;
1020 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
1024 if (rgid != (gid_t) -1) {
1025 if ((old_rgid == rgid) ||
1026 (current->egid==rgid) ||
1027 capable(CAP_SETGID))
1032 if (egid != (gid_t) -1) {
1033 if ((old_rgid == egid) ||
1034 (current->egid == egid) ||
1035 (current->sgid == egid) ||
1036 capable(CAP_SETGID))
1041 if (new_egid != old_egid) {
1042 set_dumpable(current->mm, suid_dumpable);
1045 if (rgid != (gid_t) -1 ||
1046 (egid != (gid_t) -1 && egid != old_rgid))
1047 current->sgid = new_egid;
1048 current->fsgid = new_egid;
1049 current->egid = new_egid;
1050 current->gid = new_rgid;
1051 key_fsgid_changed(current);
1052 proc_id_connector(current, PROC_EVENT_GID);
1057 * setgid() is implemented like SysV w/ SAVED_IDS
1059 * SMP: Same implicit races as above.
1061 asmlinkage long sys_setgid(gid_t gid)
1063 int old_egid = current->egid;
1066 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
1070 if (capable(CAP_SETGID)) {
1071 if (old_egid != gid) {
1072 set_dumpable(current->mm, suid_dumpable);
1075 current->gid = current->egid = current->sgid = current->fsgid = gid;
1076 } else if ((gid == current->gid) || (gid == current->sgid)) {
1077 if (old_egid != gid) {
1078 set_dumpable(current->mm, suid_dumpable);
1081 current->egid = current->fsgid = gid;
1086 key_fsgid_changed(current);
1087 proc_id_connector(current, PROC_EVENT_GID);
1091 static int set_user(uid_t new_ruid, int dumpclear)
1093 struct user_struct *new_user;
1095 new_user = alloc_uid(current->nsproxy->user_ns, new_ruid);
1099 if (atomic_read(&new_user->processes) >=
1100 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
1101 new_user != current->nsproxy->user_ns->root_user) {
1106 switch_uid(new_user);
1109 set_dumpable(current->mm, suid_dumpable);
1112 current->uid = new_ruid;
1117 * Unprivileged users may change the real uid to the effective uid
1118 * or vice versa. (BSD-style)
1120 * If you set the real uid at all, or set the effective uid to a value not
1121 * equal to the real uid, then the saved uid is set to the new effective uid.
1123 * This makes it possible for a setuid program to completely drop its
1124 * privileges, which is often a useful assertion to make when you are doing
1125 * a security audit over a program.
1127 * The general idea is that a program which uses just setreuid() will be
1128 * 100% compatible with BSD. A program which uses just setuid() will be
1129 * 100% compatible with POSIX with saved IDs.
1131 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
1133 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
1136 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
1140 new_ruid = old_ruid = current->uid;
1141 new_euid = old_euid = current->euid;
1142 old_suid = current->suid;
1144 if (ruid != (uid_t) -1) {
1146 if ((old_ruid != ruid) &&
1147 (current->euid != ruid) &&
1148 !capable(CAP_SETUID))
1152 if (euid != (uid_t) -1) {
1154 if ((old_ruid != euid) &&
1155 (current->euid != euid) &&
1156 (current->suid != euid) &&
1157 !capable(CAP_SETUID))
1161 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
1164 if (new_euid != old_euid) {
1165 set_dumpable(current->mm, suid_dumpable);
1168 current->fsuid = current->euid = new_euid;
1169 if (ruid != (uid_t) -1 ||
1170 (euid != (uid_t) -1 && euid != old_ruid))
1171 current->suid = current->euid;
1172 current->fsuid = current->euid;
1174 key_fsuid_changed(current);
1175 proc_id_connector(current, PROC_EVENT_UID);
1177 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
1183 * setuid() is implemented like SysV with SAVED_IDS
1185 * Note that SAVED_ID's is deficient in that a setuid root program
1186 * like sendmail, for example, cannot set its uid to be a normal
1187 * user and then switch back, because if you're root, setuid() sets
1188 * the saved uid too. If you don't like this, blame the bright people
1189 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
1190 * will allow a root program to temporarily drop privileges and be able to
1191 * regain them by swapping the real and effective uid.
1193 asmlinkage long sys_setuid(uid_t uid)
1195 int old_euid = current->euid;
1196 int old_ruid, old_suid, new_suid;
1199 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
1203 old_ruid = current->uid;
1204 old_suid = current->suid;
1205 new_suid = old_suid;
1207 if (capable(CAP_SETUID)) {
1208 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
1211 } else if ((uid != current->uid) && (uid != new_suid))
1214 if (old_euid != uid) {
1215 set_dumpable(current->mm, suid_dumpable);
1218 current->fsuid = current->euid = uid;
1219 current->suid = new_suid;
1221 key_fsuid_changed(current);
1222 proc_id_connector(current, PROC_EVENT_UID);
1224 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
1229 * This function implements a generic ability to update ruid, euid,
1230 * and suid. This allows you to implement the 4.4 compatible seteuid().
1232 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
1234 int old_ruid = current->uid;
1235 int old_euid = current->euid;
1236 int old_suid = current->suid;
1239 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
1243 if (!capable(CAP_SETUID)) {
1244 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
1245 (ruid != current->euid) && (ruid != current->suid))
1247 if ((euid != (uid_t) -1) && (euid != current->uid) &&
1248 (euid != current->euid) && (euid != current->suid))
1250 if ((suid != (uid_t) -1) && (suid != current->uid) &&
1251 (suid != current->euid) && (suid != current->suid))
1254 if (ruid != (uid_t) -1) {
1255 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
1258 if (euid != (uid_t) -1) {
1259 if (euid != current->euid) {
1260 set_dumpable(current->mm, suid_dumpable);
1263 current->euid = euid;
1265 current->fsuid = current->euid;
1266 if (suid != (uid_t) -1)
1267 current->suid = suid;
1269 key_fsuid_changed(current);
1270 proc_id_connector(current, PROC_EVENT_UID);
1272 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
1275 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
1279 if (!(retval = put_user(current->uid, ruid)) &&
1280 !(retval = put_user(current->euid, euid)))
1281 retval = put_user(current->suid, suid);
1287 * Same as above, but for rgid, egid, sgid.
1289 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
1293 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
1297 if (!capable(CAP_SETGID)) {
1298 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
1299 (rgid != current->egid) && (rgid != current->sgid))
1301 if ((egid != (gid_t) -1) && (egid != current->gid) &&
1302 (egid != current->egid) && (egid != current->sgid))
1304 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
1305 (sgid != current->egid) && (sgid != current->sgid))
1308 if (egid != (gid_t) -1) {
1309 if (egid != current->egid) {
1310 set_dumpable(current->mm, suid_dumpable);
1313 current->egid = egid;
1315 current->fsgid = current->egid;
1316 if (rgid != (gid_t) -1)
1317 current->gid = rgid;
1318 if (sgid != (gid_t) -1)
1319 current->sgid = sgid;
1321 key_fsgid_changed(current);
1322 proc_id_connector(current, PROC_EVENT_GID);
1326 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
1330 if (!(retval = put_user(current->gid, rgid)) &&
1331 !(retval = put_user(current->egid, egid)))
1332 retval = put_user(current->sgid, sgid);
1339 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
1340 * is used for "access()" and for the NFS daemon (letting nfsd stay at
1341 * whatever uid it wants to). It normally shadows "euid", except when
1342 * explicitly set by setfsuid() or for access..
1344 asmlinkage long sys_setfsuid(uid_t uid)
1348 old_fsuid = current->fsuid;
1349 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
1352 if (uid == current->uid || uid == current->euid ||
1353 uid == current->suid || uid == current->fsuid ||
1354 capable(CAP_SETUID)) {
1355 if (uid != old_fsuid) {
1356 set_dumpable(current->mm, suid_dumpable);
1359 current->fsuid = uid;
1362 key_fsuid_changed(current);
1363 proc_id_connector(current, PROC_EVENT_UID);
1365 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
1371 * Samma på svenska..
1373 asmlinkage long sys_setfsgid(gid_t gid)
1377 old_fsgid = current->fsgid;
1378 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
1381 if (gid == current->gid || gid == current->egid ||
1382 gid == current->sgid || gid == current->fsgid ||
1383 capable(CAP_SETGID)) {
1384 if (gid != old_fsgid) {
1385 set_dumpable(current->mm, suid_dumpable);
1388 current->fsgid = gid;
1389 key_fsgid_changed(current);
1390 proc_id_connector(current, PROC_EVENT_GID);
1395 asmlinkage long sys_times(struct tms __user * tbuf)
1398 * In the SMP world we might just be unlucky and have one of
1399 * the times increment as we use it. Since the value is an
1400 * atomically safe type this is just fine. Conceptually its
1401 * as if the syscall took an instant longer to occur.
1405 struct task_struct *tsk = current;
1406 struct task_struct *t;
1407 cputime_t utime, stime, cutime, cstime;
1409 spin_lock_irq(&tsk->sighand->siglock);
1410 utime = tsk->signal->utime;
1411 stime = tsk->signal->stime;
1414 utime = cputime_add(utime, t->utime);
1415 stime = cputime_add(stime, t->stime);
1419 cutime = tsk->signal->cutime;
1420 cstime = tsk->signal->cstime;
1421 spin_unlock_irq(&tsk->sighand->siglock);
1423 tmp.tms_utime = cputime_to_clock_t(utime);
1424 tmp.tms_stime = cputime_to_clock_t(stime);
1425 tmp.tms_cutime = cputime_to_clock_t(cutime);
1426 tmp.tms_cstime = cputime_to_clock_t(cstime);
1427 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1430 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1434 * This needs some heavy checking ...
1435 * I just haven't the stomach for it. I also don't fully
1436 * understand sessions/pgrp etc. Let somebody who does explain it.
1438 * OK, I think I have the protection semantics right.... this is really
1439 * only important on a multi-user system anyway, to make sure one user
1440 * can't send a signal to a process owned by another. -TYT, 12/12/91
1442 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1445 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1447 struct task_struct *p;
1448 struct task_struct *group_leader = current->group_leader;
1452 pid = group_leader->pid;
1458 /* From this point forward we keep holding onto the tasklist lock
1459 * so that our parent does not change from under us. -DaveM
1461 write_lock_irq(&tasklist_lock);
1464 p = find_task_by_pid(pid);
1469 if (!thread_group_leader(p))
1472 if (p->real_parent->tgid == group_leader->tgid) {
1474 if (task_session(p) != task_session(group_leader))
1481 if (p != group_leader)
1486 if (p->signal->leader)
1490 struct task_struct *g =
1491 find_task_by_pid_type(PIDTYPE_PGID, pgid);
1493 if (!g || task_session(g) != task_session(group_leader))
1497 err = security_task_setpgid(p, pgid);
1501 if (process_group(p) != pgid) {
1502 detach_pid(p, PIDTYPE_PGID);
1503 p->signal->pgrp = pgid;
1504 attach_pid(p, PIDTYPE_PGID, find_pid(pgid));
1509 /* All paths lead to here, thus we are safe. -DaveM */
1510 write_unlock_irq(&tasklist_lock);
1514 asmlinkage long sys_getpgid(pid_t pid)
1517 return process_group(current);
1520 struct task_struct *p;
1522 read_lock(&tasklist_lock);
1523 p = find_task_by_pid(pid);
1527 retval = security_task_getpgid(p);
1529 retval = process_group(p);
1531 read_unlock(&tasklist_lock);
1536 #ifdef __ARCH_WANT_SYS_GETPGRP
1538 asmlinkage long sys_getpgrp(void)
1540 /* SMP - assuming writes are word atomic this is fine */
1541 return process_group(current);
1546 asmlinkage long sys_getsid(pid_t pid)
1549 return process_session(current);
1552 struct task_struct *p;
1554 read_lock(&tasklist_lock);
1555 p = find_task_by_pid(pid);
1559 retval = security_task_getsid(p);
1561 retval = process_session(p);
1563 read_unlock(&tasklist_lock);
1568 asmlinkage long sys_setsid(void)
1570 struct task_struct *group_leader = current->group_leader;
1574 write_lock_irq(&tasklist_lock);
1576 /* Fail if I am already a session leader */
1577 if (group_leader->signal->leader)
1580 session = group_leader->pid;
1581 /* Fail if a process group id already exists that equals the
1582 * proposed session id.
1584 * Don't check if session id == 1 because kernel threads use this
1585 * session id and so the check will always fail and make it so
1586 * init cannot successfully call setsid.
1588 if (session > 1 && find_task_by_pid_type(PIDTYPE_PGID, session))
1591 group_leader->signal->leader = 1;
1592 __set_special_pids(session, session);
1594 spin_lock(&group_leader->sighand->siglock);
1595 group_leader->signal->tty = NULL;
1596 spin_unlock(&group_leader->sighand->siglock);
1598 err = process_group(group_leader);
1600 write_unlock_irq(&tasklist_lock);
1605 * Supplementary group IDs
1608 /* init to 2 - one for init_task, one to ensure it is never freed */
1609 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1611 struct group_info *groups_alloc(int gidsetsize)
1613 struct group_info *group_info;
1617 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1618 /* Make sure we always allocate at least one indirect block pointer */
1619 nblocks = nblocks ? : 1;
1620 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1623 group_info->ngroups = gidsetsize;
1624 group_info->nblocks = nblocks;
1625 atomic_set(&group_info->usage, 1);
1627 if (gidsetsize <= NGROUPS_SMALL)
1628 group_info->blocks[0] = group_info->small_block;
1630 for (i = 0; i < nblocks; i++) {
1632 b = (void *)__get_free_page(GFP_USER);
1634 goto out_undo_partial_alloc;
1635 group_info->blocks[i] = b;
1640 out_undo_partial_alloc:
1642 free_page((unsigned long)group_info->blocks[i]);
1648 EXPORT_SYMBOL(groups_alloc);
1650 void groups_free(struct group_info *group_info)
1652 if (group_info->blocks[0] != group_info->small_block) {
1654 for (i = 0; i < group_info->nblocks; i++)
1655 free_page((unsigned long)group_info->blocks[i]);
1660 EXPORT_SYMBOL(groups_free);
1662 /* export the group_info to a user-space array */
1663 static int groups_to_user(gid_t __user *grouplist,
1664 struct group_info *group_info)
1667 int count = group_info->ngroups;
1669 for (i = 0; i < group_info->nblocks; i++) {
1670 int cp_count = min(NGROUPS_PER_BLOCK, count);
1671 int off = i * NGROUPS_PER_BLOCK;
1672 int len = cp_count * sizeof(*grouplist);
1674 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1682 /* fill a group_info from a user-space array - it must be allocated already */
1683 static int groups_from_user(struct group_info *group_info,
1684 gid_t __user *grouplist)
1687 int count = group_info->ngroups;
1689 for (i = 0; i < group_info->nblocks; i++) {
1690 int cp_count = min(NGROUPS_PER_BLOCK, count);
1691 int off = i * NGROUPS_PER_BLOCK;
1692 int len = cp_count * sizeof(*grouplist);
1694 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1702 /* a simple Shell sort */
1703 static void groups_sort(struct group_info *group_info)
1705 int base, max, stride;
1706 int gidsetsize = group_info->ngroups;
1708 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1713 max = gidsetsize - stride;
1714 for (base = 0; base < max; base++) {
1716 int right = left + stride;
1717 gid_t tmp = GROUP_AT(group_info, right);
1719 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1720 GROUP_AT(group_info, right) =
1721 GROUP_AT(group_info, left);
1725 GROUP_AT(group_info, right) = tmp;
1731 /* a simple bsearch */
1732 int groups_search(struct group_info *group_info, gid_t grp)
1734 unsigned int left, right;
1740 right = group_info->ngroups;
1741 while (left < right) {
1742 unsigned int mid = (left+right)/2;
1743 int cmp = grp - GROUP_AT(group_info, mid);
1754 /* validate and set current->group_info */
1755 int set_current_groups(struct group_info *group_info)
1758 struct group_info *old_info;
1760 retval = security_task_setgroups(group_info);
1764 groups_sort(group_info);
1765 get_group_info(group_info);
1768 old_info = current->group_info;
1769 current->group_info = group_info;
1770 task_unlock(current);
1772 put_group_info(old_info);
1777 EXPORT_SYMBOL(set_current_groups);
1779 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1784 * SMP: Nobody else can change our grouplist. Thus we are
1791 /* no need to grab task_lock here; it cannot change */
1792 i = current->group_info->ngroups;
1794 if (i > gidsetsize) {
1798 if (groups_to_user(grouplist, current->group_info)) {
1808 * SMP: Our groups are copy-on-write. We can set them safely
1809 * without another task interfering.
1812 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1814 struct group_info *group_info;
1817 if (!capable(CAP_SETGID))
1819 if ((unsigned)gidsetsize > NGROUPS_MAX)
1822 group_info = groups_alloc(gidsetsize);
1825 retval = groups_from_user(group_info, grouplist);
1827 put_group_info(group_info);
1831 retval = set_current_groups(group_info);
1832 put_group_info(group_info);
1838 * Check whether we're fsgid/egid or in the supplemental group..
1840 int in_group_p(gid_t grp)
1843 if (grp != current->fsgid)
1844 retval = groups_search(current->group_info, grp);
1848 EXPORT_SYMBOL(in_group_p);
1850 int in_egroup_p(gid_t grp)
1853 if (grp != current->egid)
1854 retval = groups_search(current->group_info, grp);
1858 EXPORT_SYMBOL(in_egroup_p);
1860 DECLARE_RWSEM(uts_sem);
1862 EXPORT_SYMBOL(uts_sem);
1864 asmlinkage long sys_newuname(struct new_utsname __user * name)
1868 down_read(&uts_sem);
1869 if (copy_to_user(name, utsname(), sizeof *name))
1875 asmlinkage long sys_sethostname(char __user *name, int len)
1878 char tmp[__NEW_UTS_LEN];
1880 if (!capable(CAP_SYS_ADMIN))
1882 if (len < 0 || len > __NEW_UTS_LEN)
1884 down_write(&uts_sem);
1886 if (!copy_from_user(tmp, name, len)) {
1887 memcpy(utsname()->nodename, tmp, len);
1888 utsname()->nodename[len] = 0;
1895 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1897 asmlinkage long sys_gethostname(char __user *name, int len)
1903 down_read(&uts_sem);
1904 i = 1 + strlen(utsname()->nodename);
1908 if (copy_to_user(name, utsname()->nodename, i))
1917 * Only setdomainname; getdomainname can be implemented by calling
1920 asmlinkage long sys_setdomainname(char __user *name, int len)
1923 char tmp[__NEW_UTS_LEN];
1925 if (!capable(CAP_SYS_ADMIN))
1927 if (len < 0 || len > __NEW_UTS_LEN)
1930 down_write(&uts_sem);
1932 if (!copy_from_user(tmp, name, len)) {
1933 memcpy(utsname()->domainname, tmp, len);
1934 utsname()->domainname[len] = 0;
1941 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1943 if (resource >= RLIM_NLIMITS)
1946 struct rlimit value;
1947 task_lock(current->group_leader);
1948 value = current->signal->rlim[resource];
1949 task_unlock(current->group_leader);
1950 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1954 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1957 * Back compatibility for getrlimit. Needed for some apps.
1960 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1963 if (resource >= RLIM_NLIMITS)
1966 task_lock(current->group_leader);
1967 x = current->signal->rlim[resource];
1968 task_unlock(current->group_leader);
1969 if (x.rlim_cur > 0x7FFFFFFF)
1970 x.rlim_cur = 0x7FFFFFFF;
1971 if (x.rlim_max > 0x7FFFFFFF)
1972 x.rlim_max = 0x7FFFFFFF;
1973 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1978 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1980 struct rlimit new_rlim, *old_rlim;
1981 unsigned long it_prof_secs;
1984 if (resource >= RLIM_NLIMITS)
1986 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1988 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1990 old_rlim = current->signal->rlim + resource;
1991 if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1992 !capable(CAP_SYS_RESOURCE))
1994 if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1997 retval = security_task_setrlimit(resource, &new_rlim);
2001 if (resource == RLIMIT_CPU && new_rlim.rlim_cur == 0) {
2003 * The caller is asking for an immediate RLIMIT_CPU
2004 * expiry. But we use the zero value to mean "it was
2005 * never set". So let's cheat and make it one second
2008 new_rlim.rlim_cur = 1;
2011 task_lock(current->group_leader);
2012 *old_rlim = new_rlim;
2013 task_unlock(current->group_leader);
2015 if (resource != RLIMIT_CPU)
2019 * RLIMIT_CPU handling. Note that the kernel fails to return an error
2020 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
2021 * very long-standing error, and fixing it now risks breakage of
2022 * applications, so we live with it
2024 if (new_rlim.rlim_cur == RLIM_INFINITY)
2027 it_prof_secs = cputime_to_secs(current->signal->it_prof_expires);
2028 if (it_prof_secs == 0 || new_rlim.rlim_cur <= it_prof_secs) {
2029 unsigned long rlim_cur = new_rlim.rlim_cur;
2032 cputime = secs_to_cputime(rlim_cur);
2033 read_lock(&tasklist_lock);
2034 spin_lock_irq(¤t->sighand->siglock);
2035 set_process_cpu_timer(current, CPUCLOCK_PROF, &cputime, NULL);
2036 spin_unlock_irq(¤t->sighand->siglock);
2037 read_unlock(&tasklist_lock);
2044 * It would make sense to put struct rusage in the task_struct,
2045 * except that would make the task_struct be *really big*. After
2046 * task_struct gets moved into malloc'ed memory, it would
2047 * make sense to do this. It will make moving the rest of the information
2048 * a lot simpler! (Which we're not doing right now because we're not
2049 * measuring them yet).
2051 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
2052 * races with threads incrementing their own counters. But since word
2053 * reads are atomic, we either get new values or old values and we don't
2054 * care which for the sums. We always take the siglock to protect reading
2055 * the c* fields from p->signal from races with exit.c updating those
2056 * fields when reaping, so a sample either gets all the additions of a
2057 * given child after it's reaped, or none so this sample is before reaping.
2060 * We need to take the siglock for CHILDEREN, SELF and BOTH
2061 * for the cases current multithreaded, non-current single threaded
2062 * non-current multithreaded. Thread traversal is now safe with
2064 * Strictly speaking, we donot need to take the siglock if we are current and
2065 * single threaded, as no one else can take our signal_struct away, no one
2066 * else can reap the children to update signal->c* counters, and no one else
2067 * can race with the signal-> fields. If we do not take any lock, the
2068 * signal-> fields could be read out of order while another thread was just
2069 * exiting. So we should place a read memory barrier when we avoid the lock.
2070 * On the writer side, write memory barrier is implied in __exit_signal
2071 * as __exit_signal releases the siglock spinlock after updating the signal->
2072 * fields. But we don't do this yet to keep things simple.
2076 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
2078 struct task_struct *t;
2079 unsigned long flags;
2080 cputime_t utime, stime;
2082 memset((char *) r, 0, sizeof *r);
2083 utime = stime = cputime_zero;
2086 if (!lock_task_sighand(p, &flags)) {
2093 case RUSAGE_CHILDREN:
2094 utime = p->signal->cutime;
2095 stime = p->signal->cstime;
2096 r->ru_nvcsw = p->signal->cnvcsw;
2097 r->ru_nivcsw = p->signal->cnivcsw;
2098 r->ru_minflt = p->signal->cmin_flt;
2099 r->ru_majflt = p->signal->cmaj_flt;
2100 r->ru_inblock = p->signal->cinblock;
2101 r->ru_oublock = p->signal->coublock;
2103 if (who == RUSAGE_CHILDREN)
2107 utime = cputime_add(utime, p->signal->utime);
2108 stime = cputime_add(stime, p->signal->stime);
2109 r->ru_nvcsw += p->signal->nvcsw;
2110 r->ru_nivcsw += p->signal->nivcsw;
2111 r->ru_minflt += p->signal->min_flt;
2112 r->ru_majflt += p->signal->maj_flt;
2113 r->ru_inblock += p->signal->inblock;
2114 r->ru_oublock += p->signal->oublock;
2117 utime = cputime_add(utime, t->utime);
2118 stime = cputime_add(stime, t->stime);
2119 r->ru_nvcsw += t->nvcsw;
2120 r->ru_nivcsw += t->nivcsw;
2121 r->ru_minflt += t->min_flt;
2122 r->ru_majflt += t->maj_flt;
2123 r->ru_inblock += task_io_get_inblock(t);
2124 r->ru_oublock += task_io_get_oublock(t);
2133 unlock_task_sighand(p, &flags);
2136 cputime_to_timeval(utime, &r->ru_utime);
2137 cputime_to_timeval(stime, &r->ru_stime);
2140 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
2143 k_getrusage(p, who, &r);
2144 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
2147 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
2149 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
2151 return getrusage(current, who, ru);
2154 asmlinkage long sys_umask(int mask)
2156 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
2160 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
2161 unsigned long arg4, unsigned long arg5)
2165 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2170 case PR_SET_PDEATHSIG:
2171 if (!valid_signal(arg2)) {
2175 current->pdeath_signal = arg2;
2177 case PR_GET_PDEATHSIG:
2178 error = put_user(current->pdeath_signal, (int __user *)arg2);
2180 case PR_GET_DUMPABLE:
2181 error = get_dumpable(current->mm);
2183 case PR_SET_DUMPABLE:
2184 if (arg2 < 0 || arg2 > 1) {
2188 set_dumpable(current->mm, arg2);
2191 case PR_SET_UNALIGN:
2192 error = SET_UNALIGN_CTL(current, arg2);
2194 case PR_GET_UNALIGN:
2195 error = GET_UNALIGN_CTL(current, arg2);
2198 error = SET_FPEMU_CTL(current, arg2);
2201 error = GET_FPEMU_CTL(current, arg2);
2204 error = SET_FPEXC_CTL(current, arg2);
2207 error = GET_FPEXC_CTL(current, arg2);
2210 error = PR_TIMING_STATISTICAL;
2213 if (arg2 == PR_TIMING_STATISTICAL)
2219 case PR_GET_KEEPCAPS:
2220 if (current->keep_capabilities)
2223 case PR_SET_KEEPCAPS:
2224 if (arg2 != 0 && arg2 != 1) {
2228 current->keep_capabilities = arg2;
2231 struct task_struct *me = current;
2232 unsigned char ncomm[sizeof(me->comm)];
2234 ncomm[sizeof(me->comm)-1] = 0;
2235 if (strncpy_from_user(ncomm, (char __user *)arg2,
2236 sizeof(me->comm)-1) < 0)
2238 set_task_comm(me, ncomm);
2242 struct task_struct *me = current;
2243 unsigned char tcomm[sizeof(me->comm)];
2245 get_task_comm(tcomm, me);
2246 if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))
2251 error = GET_ENDIAN(current, arg2);
2254 error = SET_ENDIAN(current, arg2);
2257 case PR_GET_SECCOMP:
2258 error = prctl_get_seccomp();
2260 case PR_SET_SECCOMP:
2261 error = prctl_set_seccomp(arg2);
2271 asmlinkage long sys_getcpu(unsigned __user *cpup, unsigned __user *nodep,
2272 struct getcpu_cache __user *cache)
2275 int cpu = raw_smp_processor_id();
2277 err |= put_user(cpu, cpup);
2279 err |= put_user(cpu_to_node(cpu), nodep);
2282 * The cache is not needed for this implementation,
2283 * but make sure user programs pass something
2284 * valid. vsyscall implementations can instead make
2285 * good use of the cache. Only use t0 and t1 because
2286 * these are available in both 32bit and 64bit ABI (no
2287 * need for a compat_getcpu). 32bit has enough
2290 unsigned long t0, t1;
2291 get_user(t0, &cache->blob[0]);
2292 get_user(t1, &cache->blob[1]);
2295 put_user(t0, &cache->blob[0]);
2296 put_user(t1, &cache->blob[1]);
2298 return err ? -EFAULT : 0;
2301 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
2303 static void argv_cleanup(char **argv, char **envp)
2309 * orderly_poweroff - Trigger an orderly system poweroff
2310 * @force: force poweroff if command execution fails
2312 * This may be called from any context to trigger a system shutdown.
2313 * If the orderly shutdown fails, it will force an immediate shutdown.
2315 int orderly_poweroff(bool force)
2318 char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
2319 static char *envp[] = {
2321 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
2325 struct subprocess_info *info;
2328 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
2329 __func__, poweroff_cmd);
2333 info = call_usermodehelper_setup(argv[0], argv, envp);
2339 call_usermodehelper_setcleanup(info, argv_cleanup);
2341 ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
2345 printk(KERN_WARNING "Failed to start orderly shutdown: "
2346 "forcing the issue\n");
2348 /* I guess this should try to kick off some daemon to
2349 sync and poweroff asap. Or not even bother syncing
2350 if we're doing an emergency shutdown? */
2357 EXPORT_SYMBOL_GPL(orderly_poweroff);