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>
35 #include <linux/compat.h>
36 #include <linux/syscalls.h>
37 #include <linux/kprobes.h>
38 #include <linux/user_namespace.h>
40 #include <asm/uaccess.h>
42 #include <asm/unistd.h>
44 #ifndef SET_UNALIGN_CTL
45 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
47 #ifndef GET_UNALIGN_CTL
48 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
51 # define SET_FPEMU_CTL(a,b) (-EINVAL)
54 # define GET_FPEMU_CTL(a,b) (-EINVAL)
57 # define SET_FPEXC_CTL(a,b) (-EINVAL)
60 # define GET_FPEXC_CTL(a,b) (-EINVAL)
63 # define GET_ENDIAN(a,b) (-EINVAL)
66 # define SET_ENDIAN(a,b) (-EINVAL)
70 * this is where the system-wide overflow UID and GID are defined, for
71 * architectures that now have 32-bit UID/GID but didn't in the past
74 int overflowuid = DEFAULT_OVERFLOWUID;
75 int overflowgid = DEFAULT_OVERFLOWGID;
78 EXPORT_SYMBOL(overflowuid);
79 EXPORT_SYMBOL(overflowgid);
83 * the same as above, but for filesystems which can only store a 16-bit
84 * UID and GID. as such, this is needed on all architectures
87 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
88 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
90 EXPORT_SYMBOL(fs_overflowuid);
91 EXPORT_SYMBOL(fs_overflowgid);
94 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
99 EXPORT_SYMBOL(cad_pid);
102 * Notifier list for kernel code which wants to be called
103 * at shutdown. This is used to stop any idling DMA operations
107 static BLOCKING_NOTIFIER_HEAD(reboot_notifier_list);
110 * Notifier chain core routines. The exported routines below
111 * are layered on top of these, with appropriate locking added.
114 static int notifier_chain_register(struct notifier_block **nl,
115 struct notifier_block *n)
117 while ((*nl) != NULL) {
118 if (n->priority > (*nl)->priority)
123 rcu_assign_pointer(*nl, n);
127 static int notifier_chain_unregister(struct notifier_block **nl,
128 struct notifier_block *n)
130 while ((*nl) != NULL) {
132 rcu_assign_pointer(*nl, n->next);
141 * notifier_call_chain - Informs the registered notifiers about an event.
142 * @nl: Pointer to head of the blocking notifier chain
143 * @val: Value passed unmodified to notifier function
144 * @v: Pointer passed unmodified to notifier function
145 * @nr_to_call: Number of notifier functions to be called. Don't care
146 * value of this parameter is -1.
147 * @nr_calls: Records the number of notifications sent. Don't care
148 * value of this field is NULL.
149 * @returns: notifier_call_chain returns the value returned by the
150 * last notifier function called.
153 static int __kprobes notifier_call_chain(struct notifier_block **nl,
154 unsigned long val, void *v,
155 int nr_to_call, int *nr_calls)
157 int ret = NOTIFY_DONE;
158 struct notifier_block *nb, *next_nb;
160 nb = rcu_dereference(*nl);
162 while (nb && nr_to_call) {
163 next_nb = rcu_dereference(nb->next);
164 ret = nb->notifier_call(nb, val, v);
169 if ((ret & NOTIFY_STOP_MASK) == NOTIFY_STOP_MASK)
178 * Atomic notifier chain routines. Registration and unregistration
179 * use a spinlock, and call_chain is synchronized by RCU (no locks).
183 * atomic_notifier_chain_register - Add notifier to an atomic notifier chain
184 * @nh: Pointer to head of the atomic notifier chain
185 * @n: New entry in notifier chain
187 * Adds a notifier to an atomic notifier chain.
189 * Currently always returns zero.
192 int atomic_notifier_chain_register(struct atomic_notifier_head *nh,
193 struct notifier_block *n)
198 spin_lock_irqsave(&nh->lock, flags);
199 ret = notifier_chain_register(&nh->head, n);
200 spin_unlock_irqrestore(&nh->lock, flags);
204 EXPORT_SYMBOL_GPL(atomic_notifier_chain_register);
207 * atomic_notifier_chain_unregister - Remove notifier from an atomic notifier chain
208 * @nh: Pointer to head of the atomic notifier chain
209 * @n: Entry to remove from notifier chain
211 * Removes a notifier from an atomic notifier chain.
213 * Returns zero on success or %-ENOENT on failure.
215 int atomic_notifier_chain_unregister(struct atomic_notifier_head *nh,
216 struct notifier_block *n)
221 spin_lock_irqsave(&nh->lock, flags);
222 ret = notifier_chain_unregister(&nh->head, n);
223 spin_unlock_irqrestore(&nh->lock, flags);
228 EXPORT_SYMBOL_GPL(atomic_notifier_chain_unregister);
231 * __atomic_notifier_call_chain - Call functions in an atomic notifier chain
232 * @nh: Pointer to head of the atomic notifier chain
233 * @val: Value passed unmodified to notifier function
234 * @v: Pointer passed unmodified to notifier function
235 * @nr_to_call: See the comment for notifier_call_chain.
236 * @nr_calls: See the comment for notifier_call_chain.
238 * Calls each function in a notifier chain in turn. The functions
239 * run in an atomic context, so they must not block.
240 * This routine uses RCU to synchronize with changes to the chain.
242 * If the return value of the notifier can be and'ed
243 * with %NOTIFY_STOP_MASK then atomic_notifier_call_chain()
244 * will return immediately, with the return value of
245 * the notifier function which halted execution.
246 * Otherwise the return value is the return value
247 * of the last notifier function called.
250 int __kprobes __atomic_notifier_call_chain(struct atomic_notifier_head *nh,
251 unsigned long val, void *v,
252 int nr_to_call, int *nr_calls)
257 ret = notifier_call_chain(&nh->head, val, v, nr_to_call, nr_calls);
262 EXPORT_SYMBOL_GPL(__atomic_notifier_call_chain);
264 int __kprobes atomic_notifier_call_chain(struct atomic_notifier_head *nh,
265 unsigned long val, void *v)
267 return __atomic_notifier_call_chain(nh, val, v, -1, NULL);
270 EXPORT_SYMBOL_GPL(atomic_notifier_call_chain);
272 * Blocking notifier chain routines. All access to the chain is
273 * synchronized by an rwsem.
277 * blocking_notifier_chain_register - Add notifier to a blocking notifier chain
278 * @nh: Pointer to head of the blocking notifier chain
279 * @n: New entry in notifier chain
281 * Adds a notifier to a blocking notifier chain.
282 * Must be called in process context.
284 * Currently always returns zero.
287 int blocking_notifier_chain_register(struct blocking_notifier_head *nh,
288 struct notifier_block *n)
293 * This code gets used during boot-up, when task switching is
294 * not yet working and interrupts must remain disabled. At
295 * such times we must not call down_write().
297 if (unlikely(system_state == SYSTEM_BOOTING))
298 return notifier_chain_register(&nh->head, n);
300 down_write(&nh->rwsem);
301 ret = notifier_chain_register(&nh->head, n);
302 up_write(&nh->rwsem);
306 EXPORT_SYMBOL_GPL(blocking_notifier_chain_register);
309 * blocking_notifier_chain_unregister - Remove notifier from a blocking notifier chain
310 * @nh: Pointer to head of the blocking notifier chain
311 * @n: Entry to remove from notifier chain
313 * Removes a notifier from a blocking notifier chain.
314 * Must be called from process context.
316 * Returns zero on success or %-ENOENT on failure.
318 int blocking_notifier_chain_unregister(struct blocking_notifier_head *nh,
319 struct notifier_block *n)
324 * This code gets used during boot-up, when task switching is
325 * not yet working and interrupts must remain disabled. At
326 * such times we must not call down_write().
328 if (unlikely(system_state == SYSTEM_BOOTING))
329 return notifier_chain_unregister(&nh->head, n);
331 down_write(&nh->rwsem);
332 ret = notifier_chain_unregister(&nh->head, n);
333 up_write(&nh->rwsem);
337 EXPORT_SYMBOL_GPL(blocking_notifier_chain_unregister);
340 * __blocking_notifier_call_chain - Call functions in a blocking notifier chain
341 * @nh: Pointer to head of the blocking notifier chain
342 * @val: Value passed unmodified to notifier function
343 * @v: Pointer passed unmodified to notifier function
344 * @nr_to_call: See comment for notifier_call_chain.
345 * @nr_calls: See comment for notifier_call_chain.
347 * Calls each function in a notifier chain in turn. The functions
348 * run in a process context, so they are allowed to block.
350 * If the return value of the notifier can be and'ed
351 * with %NOTIFY_STOP_MASK then blocking_notifier_call_chain()
352 * will return immediately, with the return value of
353 * the notifier function which halted execution.
354 * Otherwise the return value is the return value
355 * of the last notifier function called.
358 int __blocking_notifier_call_chain(struct blocking_notifier_head *nh,
359 unsigned long val, void *v,
360 int nr_to_call, int *nr_calls)
362 int ret = NOTIFY_DONE;
365 * We check the head outside the lock, but if this access is
366 * racy then it does not matter what the result of the test
367 * is, we re-check the list after having taken the lock anyway:
369 if (rcu_dereference(nh->head)) {
370 down_read(&nh->rwsem);
371 ret = notifier_call_chain(&nh->head, val, v, nr_to_call,
377 EXPORT_SYMBOL_GPL(__blocking_notifier_call_chain);
379 int blocking_notifier_call_chain(struct blocking_notifier_head *nh,
380 unsigned long val, void *v)
382 return __blocking_notifier_call_chain(nh, val, v, -1, NULL);
384 EXPORT_SYMBOL_GPL(blocking_notifier_call_chain);
387 * Raw notifier chain routines. There is no protection;
388 * the caller must provide it. Use at your own risk!
392 * raw_notifier_chain_register - Add notifier to a raw notifier chain
393 * @nh: Pointer to head of the raw notifier chain
394 * @n: New entry in notifier chain
396 * Adds a notifier to a raw notifier chain.
397 * All locking must be provided by the caller.
399 * Currently always returns zero.
402 int raw_notifier_chain_register(struct raw_notifier_head *nh,
403 struct notifier_block *n)
405 return notifier_chain_register(&nh->head, n);
408 EXPORT_SYMBOL_GPL(raw_notifier_chain_register);
411 * raw_notifier_chain_unregister - Remove notifier from a raw notifier chain
412 * @nh: Pointer to head of the raw notifier chain
413 * @n: Entry to remove from notifier chain
415 * Removes a notifier from a raw notifier chain.
416 * All locking must be provided by the caller.
418 * Returns zero on success or %-ENOENT on failure.
420 int raw_notifier_chain_unregister(struct raw_notifier_head *nh,
421 struct notifier_block *n)
423 return notifier_chain_unregister(&nh->head, n);
426 EXPORT_SYMBOL_GPL(raw_notifier_chain_unregister);
429 * __raw_notifier_call_chain - Call functions in a raw notifier chain
430 * @nh: Pointer to head of the raw notifier chain
431 * @val: Value passed unmodified to notifier function
432 * @v: Pointer passed unmodified to notifier function
433 * @nr_to_call: See comment for notifier_call_chain.
434 * @nr_calls: See comment for notifier_call_chain
436 * Calls each function in a notifier chain in turn. The functions
437 * run in an undefined context.
438 * All locking must be provided by the caller.
440 * If the return value of the notifier can be and'ed
441 * with %NOTIFY_STOP_MASK then raw_notifier_call_chain()
442 * will return immediately, with the return value of
443 * the notifier function which halted execution.
444 * Otherwise the return value is the return value
445 * of the last notifier function called.
448 int __raw_notifier_call_chain(struct raw_notifier_head *nh,
449 unsigned long val, void *v,
450 int nr_to_call, int *nr_calls)
452 return notifier_call_chain(&nh->head, val, v, nr_to_call, nr_calls);
455 EXPORT_SYMBOL_GPL(__raw_notifier_call_chain);
457 int raw_notifier_call_chain(struct raw_notifier_head *nh,
458 unsigned long val, void *v)
460 return __raw_notifier_call_chain(nh, val, v, -1, NULL);
463 EXPORT_SYMBOL_GPL(raw_notifier_call_chain);
466 * SRCU notifier chain routines. Registration and unregistration
467 * use a mutex, and call_chain is synchronized by SRCU (no locks).
471 * srcu_notifier_chain_register - Add notifier to an SRCU notifier chain
472 * @nh: Pointer to head of the SRCU notifier chain
473 * @n: New entry in notifier chain
475 * Adds a notifier to an SRCU notifier chain.
476 * Must be called in process context.
478 * Currently always returns zero.
481 int srcu_notifier_chain_register(struct srcu_notifier_head *nh,
482 struct notifier_block *n)
487 * This code gets used during boot-up, when task switching is
488 * not yet working and interrupts must remain disabled. At
489 * such times we must not call mutex_lock().
491 if (unlikely(system_state == SYSTEM_BOOTING))
492 return notifier_chain_register(&nh->head, n);
494 mutex_lock(&nh->mutex);
495 ret = notifier_chain_register(&nh->head, n);
496 mutex_unlock(&nh->mutex);
500 EXPORT_SYMBOL_GPL(srcu_notifier_chain_register);
503 * srcu_notifier_chain_unregister - Remove notifier from an SRCU notifier chain
504 * @nh: Pointer to head of the SRCU notifier chain
505 * @n: Entry to remove from notifier chain
507 * Removes a notifier from an SRCU notifier chain.
508 * Must be called from process context.
510 * Returns zero on success or %-ENOENT on failure.
512 int srcu_notifier_chain_unregister(struct srcu_notifier_head *nh,
513 struct notifier_block *n)
518 * This code gets used during boot-up, when task switching is
519 * not yet working and interrupts must remain disabled. At
520 * such times we must not call mutex_lock().
522 if (unlikely(system_state == SYSTEM_BOOTING))
523 return notifier_chain_unregister(&nh->head, n);
525 mutex_lock(&nh->mutex);
526 ret = notifier_chain_unregister(&nh->head, n);
527 mutex_unlock(&nh->mutex);
528 synchronize_srcu(&nh->srcu);
532 EXPORT_SYMBOL_GPL(srcu_notifier_chain_unregister);
535 * __srcu_notifier_call_chain - Call functions in an SRCU notifier chain
536 * @nh: Pointer to head of the SRCU notifier chain
537 * @val: Value passed unmodified to notifier function
538 * @v: Pointer passed unmodified to notifier function
539 * @nr_to_call: See comment for notifier_call_chain.
540 * @nr_calls: See comment for notifier_call_chain
542 * Calls each function in a notifier chain in turn. The functions
543 * run in a process context, so they are allowed to block.
545 * If the return value of the notifier can be and'ed
546 * with %NOTIFY_STOP_MASK then srcu_notifier_call_chain()
547 * will return immediately, with the return value of
548 * the notifier function which halted execution.
549 * Otherwise the return value is the return value
550 * of the last notifier function called.
553 int __srcu_notifier_call_chain(struct srcu_notifier_head *nh,
554 unsigned long val, void *v,
555 int nr_to_call, int *nr_calls)
560 idx = srcu_read_lock(&nh->srcu);
561 ret = notifier_call_chain(&nh->head, val, v, nr_to_call, nr_calls);
562 srcu_read_unlock(&nh->srcu, idx);
565 EXPORT_SYMBOL_GPL(__srcu_notifier_call_chain);
567 int srcu_notifier_call_chain(struct srcu_notifier_head *nh,
568 unsigned long val, void *v)
570 return __srcu_notifier_call_chain(nh, val, v, -1, NULL);
572 EXPORT_SYMBOL_GPL(srcu_notifier_call_chain);
575 * srcu_init_notifier_head - Initialize an SRCU notifier head
576 * @nh: Pointer to head of the srcu notifier chain
578 * Unlike other sorts of notifier heads, SRCU notifier heads require
579 * dynamic initialization. Be sure to call this routine before
580 * calling any of the other SRCU notifier routines for this head.
582 * If an SRCU notifier head is deallocated, it must first be cleaned
583 * up by calling srcu_cleanup_notifier_head(). Otherwise the head's
584 * per-cpu data (used by the SRCU mechanism) will leak.
587 void srcu_init_notifier_head(struct srcu_notifier_head *nh)
589 mutex_init(&nh->mutex);
590 if (init_srcu_struct(&nh->srcu) < 0)
595 EXPORT_SYMBOL_GPL(srcu_init_notifier_head);
598 * register_reboot_notifier - Register function to be called at reboot time
599 * @nb: Info about notifier function to be called
601 * Registers a function with the list of functions
602 * to be called at reboot time.
604 * Currently always returns zero, as blocking_notifier_chain_register()
605 * always returns zero.
608 int register_reboot_notifier(struct notifier_block * nb)
610 return blocking_notifier_chain_register(&reboot_notifier_list, nb);
613 EXPORT_SYMBOL(register_reboot_notifier);
616 * unregister_reboot_notifier - Unregister previously registered reboot notifier
617 * @nb: Hook to be unregistered
619 * Unregisters a previously registered reboot
622 * Returns zero on success, or %-ENOENT on failure.
625 int unregister_reboot_notifier(struct notifier_block * nb)
627 return blocking_notifier_chain_unregister(&reboot_notifier_list, nb);
630 EXPORT_SYMBOL(unregister_reboot_notifier);
632 static int set_one_prio(struct task_struct *p, int niceval, int error)
636 if (p->uid != current->euid &&
637 p->euid != current->euid && !capable(CAP_SYS_NICE)) {
641 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
645 no_nice = security_task_setnice(p, niceval);
652 set_user_nice(p, niceval);
657 asmlinkage long sys_setpriority(int which, int who, int niceval)
659 struct task_struct *g, *p;
660 struct user_struct *user;
664 if (which > PRIO_USER || which < PRIO_PROCESS)
667 /* normalize: avoid signed division (rounding problems) */
674 read_lock(&tasklist_lock);
678 p = find_task_by_pid(who);
682 error = set_one_prio(p, niceval, error);
686 pgrp = find_pid(who);
688 pgrp = task_pgrp(current);
689 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
690 error = set_one_prio(p, niceval, error);
691 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
694 user = current->user;
698 if ((who != current->uid) && !(user = find_user(who)))
699 goto out_unlock; /* No processes for this user */
703 error = set_one_prio(p, niceval, error);
704 while_each_thread(g, p);
705 if (who != current->uid)
706 free_uid(user); /* For find_user() */
710 read_unlock(&tasklist_lock);
716 * Ugh. To avoid negative return values, "getpriority()" will
717 * not return the normal nice-value, but a negated value that
718 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
719 * to stay compatible.
721 asmlinkage long sys_getpriority(int which, int who)
723 struct task_struct *g, *p;
724 struct user_struct *user;
725 long niceval, retval = -ESRCH;
728 if (which > PRIO_USER || which < PRIO_PROCESS)
731 read_lock(&tasklist_lock);
735 p = find_task_by_pid(who);
739 niceval = 20 - task_nice(p);
740 if (niceval > retval)
746 pgrp = find_pid(who);
748 pgrp = task_pgrp(current);
749 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
750 niceval = 20 - task_nice(p);
751 if (niceval > retval)
753 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
756 user = current->user;
760 if ((who != current->uid) && !(user = find_user(who)))
761 goto out_unlock; /* No processes for this user */
765 niceval = 20 - task_nice(p);
766 if (niceval > retval)
769 while_each_thread(g, p);
770 if (who != current->uid)
771 free_uid(user); /* for find_user() */
775 read_unlock(&tasklist_lock);
781 * emergency_restart - reboot the system
783 * Without shutting down any hardware or taking any locks
784 * reboot the system. This is called when we know we are in
785 * trouble so this is our best effort to reboot. This is
786 * safe to call in interrupt context.
788 void emergency_restart(void)
790 machine_emergency_restart();
792 EXPORT_SYMBOL_GPL(emergency_restart);
794 static void kernel_restart_prepare(char *cmd)
796 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
797 system_state = SYSTEM_RESTART;
802 * kernel_restart - reboot the system
803 * @cmd: pointer to buffer containing command to execute for restart
806 * Shutdown everything and perform a clean reboot.
807 * This is not safe to call in interrupt context.
809 void kernel_restart(char *cmd)
811 kernel_restart_prepare(cmd);
813 printk(KERN_EMERG "Restarting system.\n");
815 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
816 machine_restart(cmd);
818 EXPORT_SYMBOL_GPL(kernel_restart);
821 * kernel_kexec - reboot the system
823 * Move into place and start executing a preloaded standalone
824 * executable. If nothing was preloaded return an error.
826 static void kernel_kexec(void)
829 struct kimage *image;
830 image = xchg(&kexec_image, NULL);
833 kernel_restart_prepare(NULL);
834 printk(KERN_EMERG "Starting new kernel\n");
836 machine_kexec(image);
840 void kernel_shutdown_prepare(enum system_states state)
842 blocking_notifier_call_chain(&reboot_notifier_list,
843 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
844 system_state = state;
848 * kernel_halt - halt the system
850 * Shutdown everything and perform a clean system halt.
852 void kernel_halt(void)
854 kernel_shutdown_prepare(SYSTEM_HALT);
855 printk(KERN_EMERG "System halted.\n");
859 EXPORT_SYMBOL_GPL(kernel_halt);
862 * kernel_power_off - power_off the system
864 * Shutdown everything and perform a clean system power_off.
866 void kernel_power_off(void)
868 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
869 printk(KERN_EMERG "Power down.\n");
872 EXPORT_SYMBOL_GPL(kernel_power_off);
874 * Reboot system call: for obvious reasons only root may call it,
875 * and even root needs to set up some magic numbers in the registers
876 * so that some mistake won't make this reboot the whole machine.
877 * You can also set the meaning of the ctrl-alt-del-key here.
879 * reboot doesn't sync: do that yourself before calling this.
881 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
885 /* We only trust the superuser with rebooting the system. */
886 if (!capable(CAP_SYS_BOOT))
889 /* For safety, we require "magic" arguments. */
890 if (magic1 != LINUX_REBOOT_MAGIC1 ||
891 (magic2 != LINUX_REBOOT_MAGIC2 &&
892 magic2 != LINUX_REBOOT_MAGIC2A &&
893 magic2 != LINUX_REBOOT_MAGIC2B &&
894 magic2 != LINUX_REBOOT_MAGIC2C))
897 /* Instead of trying to make the power_off code look like
898 * halt when pm_power_off is not set do it the easy way.
900 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
901 cmd = LINUX_REBOOT_CMD_HALT;
905 case LINUX_REBOOT_CMD_RESTART:
906 kernel_restart(NULL);
909 case LINUX_REBOOT_CMD_CAD_ON:
913 case LINUX_REBOOT_CMD_CAD_OFF:
917 case LINUX_REBOOT_CMD_HALT:
923 case LINUX_REBOOT_CMD_POWER_OFF:
929 case LINUX_REBOOT_CMD_RESTART2:
930 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
934 buffer[sizeof(buffer) - 1] = '\0';
936 kernel_restart(buffer);
939 case LINUX_REBOOT_CMD_KEXEC:
944 #ifdef CONFIG_SOFTWARE_SUSPEND
945 case LINUX_REBOOT_CMD_SW_SUSPEND:
947 int ret = hibernate();
961 static void deferred_cad(struct work_struct *dummy)
963 kernel_restart(NULL);
967 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
968 * As it's called within an interrupt, it may NOT sync: the only choice
969 * is whether to reboot at once, or just ignore the ctrl-alt-del.
971 void ctrl_alt_del(void)
973 static DECLARE_WORK(cad_work, deferred_cad);
976 schedule_work(&cad_work);
978 kill_cad_pid(SIGINT, 1);
982 * Unprivileged users may change the real gid to the effective gid
983 * or vice versa. (BSD-style)
985 * If you set the real gid at all, or set the effective gid to a value not
986 * equal to the real gid, then the saved gid is set to the new effective gid.
988 * This makes it possible for a setgid program to completely drop its
989 * privileges, which is often a useful assertion to make when you are doing
990 * a security audit over a program.
992 * The general idea is that a program which uses just setregid() will be
993 * 100% compatible with BSD. A program which uses just setgid() will be
994 * 100% compatible with POSIX with saved IDs.
996 * SMP: There are not races, the GIDs are checked only by filesystem
997 * operations (as far as semantic preservation is concerned).
999 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
1001 int old_rgid = current->gid;
1002 int old_egid = current->egid;
1003 int new_rgid = old_rgid;
1004 int new_egid = old_egid;
1007 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
1011 if (rgid != (gid_t) -1) {
1012 if ((old_rgid == rgid) ||
1013 (current->egid==rgid) ||
1014 capable(CAP_SETGID))
1019 if (egid != (gid_t) -1) {
1020 if ((old_rgid == egid) ||
1021 (current->egid == egid) ||
1022 (current->sgid == egid) ||
1023 capable(CAP_SETGID))
1028 if (new_egid != old_egid) {
1029 current->mm->dumpable = suid_dumpable;
1032 if (rgid != (gid_t) -1 ||
1033 (egid != (gid_t) -1 && egid != old_rgid))
1034 current->sgid = new_egid;
1035 current->fsgid = new_egid;
1036 current->egid = new_egid;
1037 current->gid = new_rgid;
1038 key_fsgid_changed(current);
1039 proc_id_connector(current, PROC_EVENT_GID);
1044 * setgid() is implemented like SysV w/ SAVED_IDS
1046 * SMP: Same implicit races as above.
1048 asmlinkage long sys_setgid(gid_t gid)
1050 int old_egid = current->egid;
1053 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
1057 if (capable(CAP_SETGID)) {
1058 if (old_egid != gid) {
1059 current->mm->dumpable = suid_dumpable;
1062 current->gid = current->egid = current->sgid = current->fsgid = gid;
1063 } else if ((gid == current->gid) || (gid == current->sgid)) {
1064 if (old_egid != gid) {
1065 current->mm->dumpable = suid_dumpable;
1068 current->egid = current->fsgid = gid;
1073 key_fsgid_changed(current);
1074 proc_id_connector(current, PROC_EVENT_GID);
1078 static int set_user(uid_t new_ruid, int dumpclear)
1080 struct user_struct *new_user;
1082 new_user = alloc_uid(current->nsproxy->user_ns, new_ruid);
1086 if (atomic_read(&new_user->processes) >=
1087 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
1088 new_user != current->nsproxy->user_ns->root_user) {
1093 switch_uid(new_user);
1096 current->mm->dumpable = suid_dumpable;
1099 current->uid = new_ruid;
1104 * Unprivileged users may change the real uid to the effective uid
1105 * or vice versa. (BSD-style)
1107 * If you set the real uid at all, or set the effective uid to a value not
1108 * equal to the real uid, then the saved uid is set to the new effective uid.
1110 * This makes it possible for a setuid program to completely drop its
1111 * privileges, which is often a useful assertion to make when you are doing
1112 * a security audit over a program.
1114 * The general idea is that a program which uses just setreuid() will be
1115 * 100% compatible with BSD. A program which uses just setuid() will be
1116 * 100% compatible with POSIX with saved IDs.
1118 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
1120 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
1123 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
1127 new_ruid = old_ruid = current->uid;
1128 new_euid = old_euid = current->euid;
1129 old_suid = current->suid;
1131 if (ruid != (uid_t) -1) {
1133 if ((old_ruid != ruid) &&
1134 (current->euid != ruid) &&
1135 !capable(CAP_SETUID))
1139 if (euid != (uid_t) -1) {
1141 if ((old_ruid != euid) &&
1142 (current->euid != euid) &&
1143 (current->suid != euid) &&
1144 !capable(CAP_SETUID))
1148 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
1151 if (new_euid != old_euid) {
1152 current->mm->dumpable = suid_dumpable;
1155 current->fsuid = current->euid = new_euid;
1156 if (ruid != (uid_t) -1 ||
1157 (euid != (uid_t) -1 && euid != old_ruid))
1158 current->suid = current->euid;
1159 current->fsuid = current->euid;
1161 key_fsuid_changed(current);
1162 proc_id_connector(current, PROC_EVENT_UID);
1164 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
1170 * setuid() is implemented like SysV with SAVED_IDS
1172 * Note that SAVED_ID's is deficient in that a setuid root program
1173 * like sendmail, for example, cannot set its uid to be a normal
1174 * user and then switch back, because if you're root, setuid() sets
1175 * the saved uid too. If you don't like this, blame the bright people
1176 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
1177 * will allow a root program to temporarily drop privileges and be able to
1178 * regain them by swapping the real and effective uid.
1180 asmlinkage long sys_setuid(uid_t uid)
1182 int old_euid = current->euid;
1183 int old_ruid, old_suid, new_suid;
1186 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
1190 old_ruid = current->uid;
1191 old_suid = current->suid;
1192 new_suid = old_suid;
1194 if (capable(CAP_SETUID)) {
1195 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
1198 } else if ((uid != current->uid) && (uid != new_suid))
1201 if (old_euid != uid) {
1202 current->mm->dumpable = suid_dumpable;
1205 current->fsuid = current->euid = uid;
1206 current->suid = new_suid;
1208 key_fsuid_changed(current);
1209 proc_id_connector(current, PROC_EVENT_UID);
1211 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
1216 * This function implements a generic ability to update ruid, euid,
1217 * and suid. This allows you to implement the 4.4 compatible seteuid().
1219 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
1221 int old_ruid = current->uid;
1222 int old_euid = current->euid;
1223 int old_suid = current->suid;
1226 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
1230 if (!capable(CAP_SETUID)) {
1231 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
1232 (ruid != current->euid) && (ruid != current->suid))
1234 if ((euid != (uid_t) -1) && (euid != current->uid) &&
1235 (euid != current->euid) && (euid != current->suid))
1237 if ((suid != (uid_t) -1) && (suid != current->uid) &&
1238 (suid != current->euid) && (suid != current->suid))
1241 if (ruid != (uid_t) -1) {
1242 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
1245 if (euid != (uid_t) -1) {
1246 if (euid != current->euid) {
1247 current->mm->dumpable = suid_dumpable;
1250 current->euid = euid;
1252 current->fsuid = current->euid;
1253 if (suid != (uid_t) -1)
1254 current->suid = suid;
1256 key_fsuid_changed(current);
1257 proc_id_connector(current, PROC_EVENT_UID);
1259 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
1262 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
1266 if (!(retval = put_user(current->uid, ruid)) &&
1267 !(retval = put_user(current->euid, euid)))
1268 retval = put_user(current->suid, suid);
1274 * Same as above, but for rgid, egid, sgid.
1276 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
1280 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
1284 if (!capable(CAP_SETGID)) {
1285 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
1286 (rgid != current->egid) && (rgid != current->sgid))
1288 if ((egid != (gid_t) -1) && (egid != current->gid) &&
1289 (egid != current->egid) && (egid != current->sgid))
1291 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
1292 (sgid != current->egid) && (sgid != current->sgid))
1295 if (egid != (gid_t) -1) {
1296 if (egid != current->egid) {
1297 current->mm->dumpable = suid_dumpable;
1300 current->egid = egid;
1302 current->fsgid = current->egid;
1303 if (rgid != (gid_t) -1)
1304 current->gid = rgid;
1305 if (sgid != (gid_t) -1)
1306 current->sgid = sgid;
1308 key_fsgid_changed(current);
1309 proc_id_connector(current, PROC_EVENT_GID);
1313 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
1317 if (!(retval = put_user(current->gid, rgid)) &&
1318 !(retval = put_user(current->egid, egid)))
1319 retval = put_user(current->sgid, sgid);
1326 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
1327 * is used for "access()" and for the NFS daemon (letting nfsd stay at
1328 * whatever uid it wants to). It normally shadows "euid", except when
1329 * explicitly set by setfsuid() or for access..
1331 asmlinkage long sys_setfsuid(uid_t uid)
1335 old_fsuid = current->fsuid;
1336 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
1339 if (uid == current->uid || uid == current->euid ||
1340 uid == current->suid || uid == current->fsuid ||
1341 capable(CAP_SETUID)) {
1342 if (uid != old_fsuid) {
1343 current->mm->dumpable = suid_dumpable;
1346 current->fsuid = uid;
1349 key_fsuid_changed(current);
1350 proc_id_connector(current, PROC_EVENT_UID);
1352 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
1358 * Samma på svenska..
1360 asmlinkage long sys_setfsgid(gid_t gid)
1364 old_fsgid = current->fsgid;
1365 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
1368 if (gid == current->gid || gid == current->egid ||
1369 gid == current->sgid || gid == current->fsgid ||
1370 capable(CAP_SETGID)) {
1371 if (gid != old_fsgid) {
1372 current->mm->dumpable = suid_dumpable;
1375 current->fsgid = gid;
1376 key_fsgid_changed(current);
1377 proc_id_connector(current, PROC_EVENT_GID);
1382 asmlinkage long sys_times(struct tms __user * tbuf)
1385 * In the SMP world we might just be unlucky and have one of
1386 * the times increment as we use it. Since the value is an
1387 * atomically safe type this is just fine. Conceptually its
1388 * as if the syscall took an instant longer to occur.
1392 struct task_struct *tsk = current;
1393 struct task_struct *t;
1394 cputime_t utime, stime, cutime, cstime;
1396 spin_lock_irq(&tsk->sighand->siglock);
1397 utime = tsk->signal->utime;
1398 stime = tsk->signal->stime;
1401 utime = cputime_add(utime, t->utime);
1402 stime = cputime_add(stime, t->stime);
1406 cutime = tsk->signal->cutime;
1407 cstime = tsk->signal->cstime;
1408 spin_unlock_irq(&tsk->sighand->siglock);
1410 tmp.tms_utime = cputime_to_clock_t(utime);
1411 tmp.tms_stime = cputime_to_clock_t(stime);
1412 tmp.tms_cutime = cputime_to_clock_t(cutime);
1413 tmp.tms_cstime = cputime_to_clock_t(cstime);
1414 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1417 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1421 * This needs some heavy checking ...
1422 * I just haven't the stomach for it. I also don't fully
1423 * understand sessions/pgrp etc. Let somebody who does explain it.
1425 * OK, I think I have the protection semantics right.... this is really
1426 * only important on a multi-user system anyway, to make sure one user
1427 * can't send a signal to a process owned by another. -TYT, 12/12/91
1429 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1433 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1435 struct task_struct *p;
1436 struct task_struct *group_leader = current->group_leader;
1440 pid = group_leader->pid;
1446 /* From this point forward we keep holding onto the tasklist lock
1447 * so that our parent does not change from under us. -DaveM
1449 write_lock_irq(&tasklist_lock);
1452 p = find_task_by_pid(pid);
1457 if (!thread_group_leader(p))
1460 if (p->real_parent == group_leader) {
1462 if (task_session(p) != task_session(group_leader))
1469 if (p != group_leader)
1474 if (p->signal->leader)
1478 struct task_struct *g =
1479 find_task_by_pid_type(PIDTYPE_PGID, pgid);
1481 if (!g || task_session(g) != task_session(group_leader))
1485 err = security_task_setpgid(p, pgid);
1489 if (process_group(p) != pgid) {
1490 detach_pid(p, PIDTYPE_PGID);
1491 p->signal->pgrp = pgid;
1492 attach_pid(p, PIDTYPE_PGID, find_pid(pgid));
1497 /* All paths lead to here, thus we are safe. -DaveM */
1498 write_unlock_irq(&tasklist_lock);
1502 asmlinkage long sys_getpgid(pid_t pid)
1505 return process_group(current);
1508 struct task_struct *p;
1510 read_lock(&tasklist_lock);
1511 p = find_task_by_pid(pid);
1515 retval = security_task_getpgid(p);
1517 retval = process_group(p);
1519 read_unlock(&tasklist_lock);
1524 #ifdef __ARCH_WANT_SYS_GETPGRP
1526 asmlinkage long sys_getpgrp(void)
1528 /* SMP - assuming writes are word atomic this is fine */
1529 return process_group(current);
1534 asmlinkage long sys_getsid(pid_t pid)
1537 return process_session(current);
1540 struct task_struct *p;
1542 read_lock(&tasklist_lock);
1543 p = find_task_by_pid(pid);
1547 retval = security_task_getsid(p);
1549 retval = process_session(p);
1551 read_unlock(&tasklist_lock);
1556 asmlinkage long sys_setsid(void)
1558 struct task_struct *group_leader = current->group_leader;
1562 write_lock_irq(&tasklist_lock);
1564 /* Fail if I am already a session leader */
1565 if (group_leader->signal->leader)
1568 session = group_leader->pid;
1569 /* Fail if a process group id already exists that equals the
1570 * proposed session id.
1572 * Don't check if session id == 1 because kernel threads use this
1573 * session id and so the check will always fail and make it so
1574 * init cannot successfully call setsid.
1576 if (session > 1 && find_task_by_pid_type(PIDTYPE_PGID, session))
1579 group_leader->signal->leader = 1;
1580 __set_special_pids(session, session);
1582 spin_lock(&group_leader->sighand->siglock);
1583 group_leader->signal->tty = NULL;
1584 spin_unlock(&group_leader->sighand->siglock);
1586 err = process_group(group_leader);
1588 write_unlock_irq(&tasklist_lock);
1593 * Supplementary group IDs
1596 /* init to 2 - one for init_task, one to ensure it is never freed */
1597 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1599 struct group_info *groups_alloc(int gidsetsize)
1601 struct group_info *group_info;
1605 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1606 /* Make sure we always allocate at least one indirect block pointer */
1607 nblocks = nblocks ? : 1;
1608 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1611 group_info->ngroups = gidsetsize;
1612 group_info->nblocks = nblocks;
1613 atomic_set(&group_info->usage, 1);
1615 if (gidsetsize <= NGROUPS_SMALL)
1616 group_info->blocks[0] = group_info->small_block;
1618 for (i = 0; i < nblocks; i++) {
1620 b = (void *)__get_free_page(GFP_USER);
1622 goto out_undo_partial_alloc;
1623 group_info->blocks[i] = b;
1628 out_undo_partial_alloc:
1630 free_page((unsigned long)group_info->blocks[i]);
1636 EXPORT_SYMBOL(groups_alloc);
1638 void groups_free(struct group_info *group_info)
1640 if (group_info->blocks[0] != group_info->small_block) {
1642 for (i = 0; i < group_info->nblocks; i++)
1643 free_page((unsigned long)group_info->blocks[i]);
1648 EXPORT_SYMBOL(groups_free);
1650 /* export the group_info to a user-space array */
1651 static int groups_to_user(gid_t __user *grouplist,
1652 struct group_info *group_info)
1655 int count = group_info->ngroups;
1657 for (i = 0; i < group_info->nblocks; i++) {
1658 int cp_count = min(NGROUPS_PER_BLOCK, count);
1659 int off = i * NGROUPS_PER_BLOCK;
1660 int len = cp_count * sizeof(*grouplist);
1662 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1670 /* fill a group_info from a user-space array - it must be allocated already */
1671 static int groups_from_user(struct group_info *group_info,
1672 gid_t __user *grouplist)
1675 int count = group_info->ngroups;
1677 for (i = 0; i < group_info->nblocks; i++) {
1678 int cp_count = min(NGROUPS_PER_BLOCK, count);
1679 int off = i * NGROUPS_PER_BLOCK;
1680 int len = cp_count * sizeof(*grouplist);
1682 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1690 /* a simple Shell sort */
1691 static void groups_sort(struct group_info *group_info)
1693 int base, max, stride;
1694 int gidsetsize = group_info->ngroups;
1696 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1701 max = gidsetsize - stride;
1702 for (base = 0; base < max; base++) {
1704 int right = left + stride;
1705 gid_t tmp = GROUP_AT(group_info, right);
1707 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1708 GROUP_AT(group_info, right) =
1709 GROUP_AT(group_info, left);
1713 GROUP_AT(group_info, right) = tmp;
1719 /* a simple bsearch */
1720 int groups_search(struct group_info *group_info, gid_t grp)
1722 unsigned int left, right;
1728 right = group_info->ngroups;
1729 while (left < right) {
1730 unsigned int mid = (left+right)/2;
1731 int cmp = grp - GROUP_AT(group_info, mid);
1742 /* validate and set current->group_info */
1743 int set_current_groups(struct group_info *group_info)
1746 struct group_info *old_info;
1748 retval = security_task_setgroups(group_info);
1752 groups_sort(group_info);
1753 get_group_info(group_info);
1756 old_info = current->group_info;
1757 current->group_info = group_info;
1758 task_unlock(current);
1760 put_group_info(old_info);
1765 EXPORT_SYMBOL(set_current_groups);
1767 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1772 * SMP: Nobody else can change our grouplist. Thus we are
1779 /* no need to grab task_lock here; it cannot change */
1780 i = current->group_info->ngroups;
1782 if (i > gidsetsize) {
1786 if (groups_to_user(grouplist, current->group_info)) {
1796 * SMP: Our groups are copy-on-write. We can set them safely
1797 * without another task interfering.
1800 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1802 struct group_info *group_info;
1805 if (!capable(CAP_SETGID))
1807 if ((unsigned)gidsetsize > NGROUPS_MAX)
1810 group_info = groups_alloc(gidsetsize);
1813 retval = groups_from_user(group_info, grouplist);
1815 put_group_info(group_info);
1819 retval = set_current_groups(group_info);
1820 put_group_info(group_info);
1826 * Check whether we're fsgid/egid or in the supplemental group..
1828 int in_group_p(gid_t grp)
1831 if (grp != current->fsgid)
1832 retval = groups_search(current->group_info, grp);
1836 EXPORT_SYMBOL(in_group_p);
1838 int in_egroup_p(gid_t grp)
1841 if (grp != current->egid)
1842 retval = groups_search(current->group_info, grp);
1846 EXPORT_SYMBOL(in_egroup_p);
1848 DECLARE_RWSEM(uts_sem);
1850 EXPORT_SYMBOL(uts_sem);
1852 asmlinkage long sys_newuname(struct new_utsname __user * name)
1856 down_read(&uts_sem);
1857 if (copy_to_user(name, utsname(), sizeof *name))
1863 asmlinkage long sys_sethostname(char __user *name, int len)
1866 char tmp[__NEW_UTS_LEN];
1868 if (!capable(CAP_SYS_ADMIN))
1870 if (len < 0 || len > __NEW_UTS_LEN)
1872 down_write(&uts_sem);
1874 if (!copy_from_user(tmp, name, len)) {
1875 memcpy(utsname()->nodename, tmp, len);
1876 utsname()->nodename[len] = 0;
1883 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1885 asmlinkage long sys_gethostname(char __user *name, int len)
1891 down_read(&uts_sem);
1892 i = 1 + strlen(utsname()->nodename);
1896 if (copy_to_user(name, utsname()->nodename, i))
1905 * Only setdomainname; getdomainname can be implemented by calling
1908 asmlinkage long sys_setdomainname(char __user *name, int len)
1911 char tmp[__NEW_UTS_LEN];
1913 if (!capable(CAP_SYS_ADMIN))
1915 if (len < 0 || len > __NEW_UTS_LEN)
1918 down_write(&uts_sem);
1920 if (!copy_from_user(tmp, name, len)) {
1921 memcpy(utsname()->domainname, tmp, len);
1922 utsname()->domainname[len] = 0;
1929 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1931 if (resource >= RLIM_NLIMITS)
1934 struct rlimit value;
1935 task_lock(current->group_leader);
1936 value = current->signal->rlim[resource];
1937 task_unlock(current->group_leader);
1938 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1942 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1945 * Back compatibility for getrlimit. Needed for some apps.
1948 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1951 if (resource >= RLIM_NLIMITS)
1954 task_lock(current->group_leader);
1955 x = current->signal->rlim[resource];
1956 task_unlock(current->group_leader);
1957 if (x.rlim_cur > 0x7FFFFFFF)
1958 x.rlim_cur = 0x7FFFFFFF;
1959 if (x.rlim_max > 0x7FFFFFFF)
1960 x.rlim_max = 0x7FFFFFFF;
1961 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1966 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1968 struct rlimit new_rlim, *old_rlim;
1969 unsigned long it_prof_secs;
1972 if (resource >= RLIM_NLIMITS)
1974 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1976 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1978 old_rlim = current->signal->rlim + resource;
1979 if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1980 !capable(CAP_SYS_RESOURCE))
1982 if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1985 retval = security_task_setrlimit(resource, &new_rlim);
1989 if (resource == RLIMIT_CPU && new_rlim.rlim_cur == 0) {
1991 * The caller is asking for an immediate RLIMIT_CPU
1992 * expiry. But we use the zero value to mean "it was
1993 * never set". So let's cheat and make it one second
1996 new_rlim.rlim_cur = 1;
1999 task_lock(current->group_leader);
2000 *old_rlim = new_rlim;
2001 task_unlock(current->group_leader);
2003 if (resource != RLIMIT_CPU)
2007 * RLIMIT_CPU handling. Note that the kernel fails to return an error
2008 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
2009 * very long-standing error, and fixing it now risks breakage of
2010 * applications, so we live with it
2012 if (new_rlim.rlim_cur == RLIM_INFINITY)
2015 it_prof_secs = cputime_to_secs(current->signal->it_prof_expires);
2016 if (it_prof_secs == 0 || new_rlim.rlim_cur <= it_prof_secs) {
2017 unsigned long rlim_cur = new_rlim.rlim_cur;
2020 cputime = secs_to_cputime(rlim_cur);
2021 read_lock(&tasklist_lock);
2022 spin_lock_irq(¤t->sighand->siglock);
2023 set_process_cpu_timer(current, CPUCLOCK_PROF, &cputime, NULL);
2024 spin_unlock_irq(¤t->sighand->siglock);
2025 read_unlock(&tasklist_lock);
2032 * It would make sense to put struct rusage in the task_struct,
2033 * except that would make the task_struct be *really big*. After
2034 * task_struct gets moved into malloc'ed memory, it would
2035 * make sense to do this. It will make moving the rest of the information
2036 * a lot simpler! (Which we're not doing right now because we're not
2037 * measuring them yet).
2039 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
2040 * races with threads incrementing their own counters. But since word
2041 * reads are atomic, we either get new values or old values and we don't
2042 * care which for the sums. We always take the siglock to protect reading
2043 * the c* fields from p->signal from races with exit.c updating those
2044 * fields when reaping, so a sample either gets all the additions of a
2045 * given child after it's reaped, or none so this sample is before reaping.
2048 * We need to take the siglock for CHILDEREN, SELF and BOTH
2049 * for the cases current multithreaded, non-current single threaded
2050 * non-current multithreaded. Thread traversal is now safe with
2052 * Strictly speaking, we donot need to take the siglock if we are current and
2053 * single threaded, as no one else can take our signal_struct away, no one
2054 * else can reap the children to update signal->c* counters, and no one else
2055 * can race with the signal-> fields. If we do not take any lock, the
2056 * signal-> fields could be read out of order while another thread was just
2057 * exiting. So we should place a read memory barrier when we avoid the lock.
2058 * On the writer side, write memory barrier is implied in __exit_signal
2059 * as __exit_signal releases the siglock spinlock after updating the signal->
2060 * fields. But we don't do this yet to keep things simple.
2064 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
2066 struct task_struct *t;
2067 unsigned long flags;
2068 cputime_t utime, stime;
2070 memset((char *) r, 0, sizeof *r);
2071 utime = stime = cputime_zero;
2074 if (!lock_task_sighand(p, &flags)) {
2081 case RUSAGE_CHILDREN:
2082 utime = p->signal->cutime;
2083 stime = p->signal->cstime;
2084 r->ru_nvcsw = p->signal->cnvcsw;
2085 r->ru_nivcsw = p->signal->cnivcsw;
2086 r->ru_minflt = p->signal->cmin_flt;
2087 r->ru_majflt = p->signal->cmaj_flt;
2088 r->ru_inblock = p->signal->cinblock;
2089 r->ru_oublock = p->signal->coublock;
2091 if (who == RUSAGE_CHILDREN)
2095 utime = cputime_add(utime, p->signal->utime);
2096 stime = cputime_add(stime, p->signal->stime);
2097 r->ru_nvcsw += p->signal->nvcsw;
2098 r->ru_nivcsw += p->signal->nivcsw;
2099 r->ru_minflt += p->signal->min_flt;
2100 r->ru_majflt += p->signal->maj_flt;
2101 r->ru_inblock += p->signal->inblock;
2102 r->ru_oublock += p->signal->oublock;
2105 utime = cputime_add(utime, t->utime);
2106 stime = cputime_add(stime, t->stime);
2107 r->ru_nvcsw += t->nvcsw;
2108 r->ru_nivcsw += t->nivcsw;
2109 r->ru_minflt += t->min_flt;
2110 r->ru_majflt += t->maj_flt;
2111 r->ru_inblock += task_io_get_inblock(t);
2112 r->ru_oublock += task_io_get_oublock(t);
2121 unlock_task_sighand(p, &flags);
2124 cputime_to_timeval(utime, &r->ru_utime);
2125 cputime_to_timeval(stime, &r->ru_stime);
2128 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
2131 k_getrusage(p, who, &r);
2132 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
2135 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
2137 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
2139 return getrusage(current, who, ru);
2142 asmlinkage long sys_umask(int mask)
2144 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
2148 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
2149 unsigned long arg4, unsigned long arg5)
2153 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2158 case PR_SET_PDEATHSIG:
2159 if (!valid_signal(arg2)) {
2163 current->pdeath_signal = arg2;
2165 case PR_GET_PDEATHSIG:
2166 error = put_user(current->pdeath_signal, (int __user *)arg2);
2168 case PR_GET_DUMPABLE:
2169 error = current->mm->dumpable;
2171 case PR_SET_DUMPABLE:
2172 if (arg2 < 0 || arg2 > 1) {
2176 current->mm->dumpable = arg2;
2179 case PR_SET_UNALIGN:
2180 error = SET_UNALIGN_CTL(current, arg2);
2182 case PR_GET_UNALIGN:
2183 error = GET_UNALIGN_CTL(current, arg2);
2186 error = SET_FPEMU_CTL(current, arg2);
2189 error = GET_FPEMU_CTL(current, arg2);
2192 error = SET_FPEXC_CTL(current, arg2);
2195 error = GET_FPEXC_CTL(current, arg2);
2198 error = PR_TIMING_STATISTICAL;
2201 if (arg2 == PR_TIMING_STATISTICAL)
2207 case PR_GET_KEEPCAPS:
2208 if (current->keep_capabilities)
2211 case PR_SET_KEEPCAPS:
2212 if (arg2 != 0 && arg2 != 1) {
2216 current->keep_capabilities = arg2;
2219 struct task_struct *me = current;
2220 unsigned char ncomm[sizeof(me->comm)];
2222 ncomm[sizeof(me->comm)-1] = 0;
2223 if (strncpy_from_user(ncomm, (char __user *)arg2,
2224 sizeof(me->comm)-1) < 0)
2226 set_task_comm(me, ncomm);
2230 struct task_struct *me = current;
2231 unsigned char tcomm[sizeof(me->comm)];
2233 get_task_comm(tcomm, me);
2234 if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))
2239 error = GET_ENDIAN(current, arg2);
2242 error = SET_ENDIAN(current, arg2);
2252 asmlinkage long sys_getcpu(unsigned __user *cpup, unsigned __user *nodep,
2253 struct getcpu_cache __user *cache)
2256 int cpu = raw_smp_processor_id();
2258 err |= put_user(cpu, cpup);
2260 err |= put_user(cpu_to_node(cpu), nodep);
2263 * The cache is not needed for this implementation,
2264 * but make sure user programs pass something
2265 * valid. vsyscall implementations can instead make
2266 * good use of the cache. Only use t0 and t1 because
2267 * these are available in both 32bit and 64bit ABI (no
2268 * need for a compat_getcpu). 32bit has enough
2271 unsigned long t0, t1;
2272 get_user(t0, &cache->blob[0]);
2273 get_user(t1, &cache->blob[1]);
2276 put_user(t0, &cache->blob[0]);
2277 put_user(t1, &cache->blob[1]);
2279 return err ? -EFAULT : 0;