]> bbs.cooldavid.org Git - net-next-2.6.git/blame - kernel/sys.c
[NETLINK]: Directly return -EINTR from netlink_dump_start()
[net-next-2.6.git] / kernel / sys.c
CommitLineData
1da177e4
LT
1/*
2 * linux/kernel/sys.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
1da177e4
LT
7#include <linux/module.h>
8#include <linux/mm.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>
1da177e4
LT
15#include <linux/highuid.h>
16#include <linux/fs.h>
dc009d92
EB
17#include <linux/kernel.h>
18#include <linux/kexec.h>
1da177e4 19#include <linux/workqueue.h>
c59ede7b 20#include <linux/capability.h>
1da177e4
LT
21#include <linux/device.h>
22#include <linux/key.h>
23#include <linux/times.h>
24#include <linux/posix-timers.h>
25#include <linux/security.h>
26#include <linux/dcookies.h>
27#include <linux/suspend.h>
28#include <linux/tty.h>
7ed20e1a 29#include <linux/signal.h>
9f46080c 30#include <linux/cn_proc.h>
3cfc348b 31#include <linux/getcpu.h>
1da177e4
LT
32
33#include <linux/compat.h>
34#include <linux/syscalls.h>
00d7c05a 35#include <linux/kprobes.h>
1da177e4
LT
36
37#include <asm/uaccess.h>
38#include <asm/io.h>
39#include <asm/unistd.h>
40
41#ifndef SET_UNALIGN_CTL
42# define SET_UNALIGN_CTL(a,b) (-EINVAL)
43#endif
44#ifndef GET_UNALIGN_CTL
45# define GET_UNALIGN_CTL(a,b) (-EINVAL)
46#endif
47#ifndef SET_FPEMU_CTL
48# define SET_FPEMU_CTL(a,b) (-EINVAL)
49#endif
50#ifndef GET_FPEMU_CTL
51# define GET_FPEMU_CTL(a,b) (-EINVAL)
52#endif
53#ifndef SET_FPEXC_CTL
54# define SET_FPEXC_CTL(a,b) (-EINVAL)
55#endif
56#ifndef GET_FPEXC_CTL
57# define GET_FPEXC_CTL(a,b) (-EINVAL)
58#endif
651d765d
AB
59#ifndef GET_ENDIAN
60# define GET_ENDIAN(a,b) (-EINVAL)
61#endif
62#ifndef SET_ENDIAN
63# define SET_ENDIAN(a,b) (-EINVAL)
64#endif
1da177e4
LT
65
66/*
67 * this is where the system-wide overflow UID and GID are defined, for
68 * architectures that now have 32-bit UID/GID but didn't in the past
69 */
70
71int overflowuid = DEFAULT_OVERFLOWUID;
72int overflowgid = DEFAULT_OVERFLOWGID;
73
74#ifdef CONFIG_UID16
75EXPORT_SYMBOL(overflowuid);
76EXPORT_SYMBOL(overflowgid);
77#endif
78
79/*
80 * the same as above, but for filesystems which can only store a 16-bit
81 * UID and GID. as such, this is needed on all architectures
82 */
83
84int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
85int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
86
87EXPORT_SYMBOL(fs_overflowuid);
88EXPORT_SYMBOL(fs_overflowgid);
89
90/*
91 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
92 */
93
94int C_A_D = 1;
9ec52099
CLG
95struct pid *cad_pid;
96EXPORT_SYMBOL(cad_pid);
1da177e4
LT
97
98/*
99 * Notifier list for kernel code which wants to be called
100 * at shutdown. This is used to stop any idling DMA operations
101 * and the like.
102 */
103
e041c683
AS
104static BLOCKING_NOTIFIER_HEAD(reboot_notifier_list);
105
106/*
107 * Notifier chain core routines. The exported routines below
108 * are layered on top of these, with appropriate locking added.
109 */
110
111static int notifier_chain_register(struct notifier_block **nl,
112 struct notifier_block *n)
113{
114 while ((*nl) != NULL) {
115 if (n->priority > (*nl)->priority)
116 break;
117 nl = &((*nl)->next);
118 }
119 n->next = *nl;
120 rcu_assign_pointer(*nl, n);
121 return 0;
122}
123
124static int notifier_chain_unregister(struct notifier_block **nl,
125 struct notifier_block *n)
126{
127 while ((*nl) != NULL) {
128 if ((*nl) == n) {
129 rcu_assign_pointer(*nl, n->next);
130 return 0;
131 }
132 nl = &((*nl)->next);
133 }
134 return -ENOENT;
135}
136
137static int __kprobes notifier_call_chain(struct notifier_block **nl,
138 unsigned long val, void *v)
139{
140 int ret = NOTIFY_DONE;
bbb1747d 141 struct notifier_block *nb, *next_nb;
e041c683
AS
142
143 nb = rcu_dereference(*nl);
144 while (nb) {
bbb1747d 145 next_nb = rcu_dereference(nb->next);
e041c683
AS
146 ret = nb->notifier_call(nb, val, v);
147 if ((ret & NOTIFY_STOP_MASK) == NOTIFY_STOP_MASK)
148 break;
bbb1747d 149 nb = next_nb;
e041c683
AS
150 }
151 return ret;
152}
153
154/*
155 * Atomic notifier chain routines. Registration and unregistration
eabc0694 156 * use a spinlock, and call_chain is synchronized by RCU (no locks).
e041c683 157 */
1da177e4
LT
158
159/**
e041c683
AS
160 * atomic_notifier_chain_register - Add notifier to an atomic notifier chain
161 * @nh: Pointer to head of the atomic notifier chain
1da177e4
LT
162 * @n: New entry in notifier chain
163 *
e041c683 164 * Adds a notifier to an atomic notifier chain.
1da177e4
LT
165 *
166 * Currently always returns zero.
167 */
e041c683
AS
168
169int atomic_notifier_chain_register(struct atomic_notifier_head *nh,
170 struct notifier_block *n)
171{
172 unsigned long flags;
173 int ret;
174
175 spin_lock_irqsave(&nh->lock, flags);
176 ret = notifier_chain_register(&nh->head, n);
177 spin_unlock_irqrestore(&nh->lock, flags);
178 return ret;
179}
180
181EXPORT_SYMBOL_GPL(atomic_notifier_chain_register);
182
183/**
184 * atomic_notifier_chain_unregister - Remove notifier from an atomic notifier chain
185 * @nh: Pointer to head of the atomic notifier chain
186 * @n: Entry to remove from notifier chain
187 *
188 * Removes a notifier from an atomic notifier chain.
189 *
190 * Returns zero on success or %-ENOENT on failure.
191 */
192int atomic_notifier_chain_unregister(struct atomic_notifier_head *nh,
193 struct notifier_block *n)
194{
195 unsigned long flags;
196 int ret;
197
198 spin_lock_irqsave(&nh->lock, flags);
199 ret = notifier_chain_unregister(&nh->head, n);
200 spin_unlock_irqrestore(&nh->lock, flags);
201 synchronize_rcu();
202 return ret;
203}
204
205EXPORT_SYMBOL_GPL(atomic_notifier_chain_unregister);
206
207/**
208 * atomic_notifier_call_chain - Call functions in an atomic notifier chain
209 * @nh: Pointer to head of the atomic notifier chain
210 * @val: Value passed unmodified to notifier function
211 * @v: Pointer passed unmodified to notifier function
212 *
213 * Calls each function in a notifier chain in turn. The functions
214 * run in an atomic context, so they must not block.
215 * This routine uses RCU to synchronize with changes to the chain.
216 *
217 * If the return value of the notifier can be and'ed
72fd4a35 218 * with %NOTIFY_STOP_MASK then atomic_notifier_call_chain()
e041c683
AS
219 * will return immediately, with the return value of
220 * the notifier function which halted execution.
221 * Otherwise the return value is the return value
222 * of the last notifier function called.
223 */
1da177e4 224
f2aa85a0 225int __kprobes atomic_notifier_call_chain(struct atomic_notifier_head *nh,
e041c683 226 unsigned long val, void *v)
1da177e4 227{
e041c683
AS
228 int ret;
229
230 rcu_read_lock();
231 ret = notifier_call_chain(&nh->head, val, v);
232 rcu_read_unlock();
233 return ret;
1da177e4
LT
234}
235
e041c683
AS
236EXPORT_SYMBOL_GPL(atomic_notifier_call_chain);
237
238/*
239 * Blocking notifier chain routines. All access to the chain is
240 * synchronized by an rwsem.
241 */
1da177e4
LT
242
243/**
e041c683
AS
244 * blocking_notifier_chain_register - Add notifier to a blocking notifier chain
245 * @nh: Pointer to head of the blocking notifier chain
1da177e4
LT
246 * @n: New entry in notifier chain
247 *
e041c683
AS
248 * Adds a notifier to a blocking notifier chain.
249 * Must be called in process context.
1da177e4 250 *
e041c683 251 * Currently always returns zero.
1da177e4
LT
252 */
253
e041c683
AS
254int blocking_notifier_chain_register(struct blocking_notifier_head *nh,
255 struct notifier_block *n)
1da177e4 256{
e041c683
AS
257 int ret;
258
259 /*
260 * This code gets used during boot-up, when task switching is
261 * not yet working and interrupts must remain disabled. At
262 * such times we must not call down_write().
263 */
264 if (unlikely(system_state == SYSTEM_BOOTING))
265 return notifier_chain_register(&nh->head, n);
266
267 down_write(&nh->rwsem);
268 ret = notifier_chain_register(&nh->head, n);
269 up_write(&nh->rwsem);
270 return ret;
1da177e4
LT
271}
272
e041c683 273EXPORT_SYMBOL_GPL(blocking_notifier_chain_register);
1da177e4
LT
274
275/**
e041c683
AS
276 * blocking_notifier_chain_unregister - Remove notifier from a blocking notifier chain
277 * @nh: Pointer to head of the blocking notifier chain
278 * @n: Entry to remove from notifier chain
279 *
280 * Removes a notifier from a blocking notifier chain.
281 * Must be called from process context.
282 *
283 * Returns zero on success or %-ENOENT on failure.
284 */
285int blocking_notifier_chain_unregister(struct blocking_notifier_head *nh,
286 struct notifier_block *n)
287{
288 int ret;
289
290 /*
291 * This code gets used during boot-up, when task switching is
292 * not yet working and interrupts must remain disabled. At
293 * such times we must not call down_write().
294 */
295 if (unlikely(system_state == SYSTEM_BOOTING))
296 return notifier_chain_unregister(&nh->head, n);
297
298 down_write(&nh->rwsem);
299 ret = notifier_chain_unregister(&nh->head, n);
300 up_write(&nh->rwsem);
301 return ret;
302}
303
304EXPORT_SYMBOL_GPL(blocking_notifier_chain_unregister);
305
306/**
307 * blocking_notifier_call_chain - Call functions in a blocking notifier chain
308 * @nh: Pointer to head of the blocking notifier chain
1da177e4
LT
309 * @val: Value passed unmodified to notifier function
310 * @v: Pointer passed unmodified to notifier function
311 *
e041c683
AS
312 * Calls each function in a notifier chain in turn. The functions
313 * run in a process context, so they are allowed to block.
1da177e4 314 *
e041c683 315 * If the return value of the notifier can be and'ed
72fd4a35 316 * with %NOTIFY_STOP_MASK then blocking_notifier_call_chain()
1da177e4
LT
317 * will return immediately, with the return value of
318 * the notifier function which halted execution.
e041c683 319 * Otherwise the return value is the return value
1da177e4
LT
320 * of the last notifier function called.
321 */
322
e041c683
AS
323int blocking_notifier_call_chain(struct blocking_notifier_head *nh,
324 unsigned long val, void *v)
1da177e4 325{
1b5180b6 326 int ret = NOTIFY_DONE;
e041c683 327
1b5180b6
IM
328 /*
329 * We check the head outside the lock, but if this access is
330 * racy then it does not matter what the result of the test
331 * is, we re-check the list after having taken the lock anyway:
332 */
333 if (rcu_dereference(nh->head)) {
334 down_read(&nh->rwsem);
335 ret = notifier_call_chain(&nh->head, val, v);
336 up_read(&nh->rwsem);
337 }
1da177e4
LT
338 return ret;
339}
340
e041c683
AS
341EXPORT_SYMBOL_GPL(blocking_notifier_call_chain);
342
343/*
344 * Raw notifier chain routines. There is no protection;
345 * the caller must provide it. Use at your own risk!
346 */
347
348/**
349 * raw_notifier_chain_register - Add notifier to a raw notifier chain
350 * @nh: Pointer to head of the raw notifier chain
351 * @n: New entry in notifier chain
352 *
353 * Adds a notifier to a raw notifier chain.
354 * All locking must be provided by the caller.
355 *
356 * Currently always returns zero.
357 */
358
359int raw_notifier_chain_register(struct raw_notifier_head *nh,
360 struct notifier_block *n)
361{
362 return notifier_chain_register(&nh->head, n);
363}
364
365EXPORT_SYMBOL_GPL(raw_notifier_chain_register);
366
367/**
368 * raw_notifier_chain_unregister - Remove notifier from a raw notifier chain
369 * @nh: Pointer to head of the raw notifier chain
370 * @n: Entry to remove from notifier chain
371 *
372 * Removes a notifier from a raw notifier chain.
373 * All locking must be provided by the caller.
374 *
375 * Returns zero on success or %-ENOENT on failure.
376 */
377int raw_notifier_chain_unregister(struct raw_notifier_head *nh,
378 struct notifier_block *n)
379{
380 return notifier_chain_unregister(&nh->head, n);
381}
382
383EXPORT_SYMBOL_GPL(raw_notifier_chain_unregister);
384
385/**
386 * raw_notifier_call_chain - Call functions in a raw notifier chain
387 * @nh: Pointer to head of the raw notifier chain
388 * @val: Value passed unmodified to notifier function
389 * @v: Pointer passed unmodified to notifier function
390 *
391 * Calls each function in a notifier chain in turn. The functions
392 * run in an undefined context.
393 * All locking must be provided by the caller.
394 *
395 * If the return value of the notifier can be and'ed
72fd4a35 396 * with %NOTIFY_STOP_MASK then raw_notifier_call_chain()
e041c683
AS
397 * will return immediately, with the return value of
398 * the notifier function which halted execution.
399 * Otherwise the return value is the return value
400 * of the last notifier function called.
401 */
402
403int raw_notifier_call_chain(struct raw_notifier_head *nh,
404 unsigned long val, void *v)
405{
406 return notifier_call_chain(&nh->head, val, v);
407}
408
409EXPORT_SYMBOL_GPL(raw_notifier_call_chain);
1da177e4 410
eabc0694
AS
411/*
412 * SRCU notifier chain routines. Registration and unregistration
413 * use a mutex, and call_chain is synchronized by SRCU (no locks).
414 */
415
416/**
417 * srcu_notifier_chain_register - Add notifier to an SRCU notifier chain
418 * @nh: Pointer to head of the SRCU notifier chain
419 * @n: New entry in notifier chain
420 *
421 * Adds a notifier to an SRCU notifier chain.
422 * Must be called in process context.
423 *
424 * Currently always returns zero.
425 */
426
427int srcu_notifier_chain_register(struct srcu_notifier_head *nh,
428 struct notifier_block *n)
429{
430 int ret;
431
432 /*
433 * This code gets used during boot-up, when task switching is
434 * not yet working and interrupts must remain disabled. At
435 * such times we must not call mutex_lock().
436 */
437 if (unlikely(system_state == SYSTEM_BOOTING))
438 return notifier_chain_register(&nh->head, n);
439
440 mutex_lock(&nh->mutex);
441 ret = notifier_chain_register(&nh->head, n);
442 mutex_unlock(&nh->mutex);
443 return ret;
444}
445
446EXPORT_SYMBOL_GPL(srcu_notifier_chain_register);
447
448/**
449 * srcu_notifier_chain_unregister - Remove notifier from an SRCU notifier chain
450 * @nh: Pointer to head of the SRCU notifier chain
451 * @n: Entry to remove from notifier chain
452 *
453 * Removes a notifier from an SRCU notifier chain.
454 * Must be called from process context.
455 *
456 * Returns zero on success or %-ENOENT on failure.
457 */
458int srcu_notifier_chain_unregister(struct srcu_notifier_head *nh,
459 struct notifier_block *n)
460{
461 int ret;
462
463 /*
464 * This code gets used during boot-up, when task switching is
465 * not yet working and interrupts must remain disabled. At
466 * such times we must not call mutex_lock().
467 */
468 if (unlikely(system_state == SYSTEM_BOOTING))
469 return notifier_chain_unregister(&nh->head, n);
470
471 mutex_lock(&nh->mutex);
472 ret = notifier_chain_unregister(&nh->head, n);
473 mutex_unlock(&nh->mutex);
474 synchronize_srcu(&nh->srcu);
475 return ret;
476}
477
478EXPORT_SYMBOL_GPL(srcu_notifier_chain_unregister);
479
480/**
481 * srcu_notifier_call_chain - Call functions in an SRCU notifier chain
482 * @nh: Pointer to head of the SRCU notifier chain
483 * @val: Value passed unmodified to notifier function
484 * @v: Pointer passed unmodified to notifier function
485 *
486 * Calls each function in a notifier chain in turn. The functions
487 * run in a process context, so they are allowed to block.
488 *
489 * If the return value of the notifier can be and'ed
72fd4a35 490 * with %NOTIFY_STOP_MASK then srcu_notifier_call_chain()
eabc0694
AS
491 * will return immediately, with the return value of
492 * the notifier function which halted execution.
493 * Otherwise the return value is the return value
494 * of the last notifier function called.
495 */
496
497int srcu_notifier_call_chain(struct srcu_notifier_head *nh,
498 unsigned long val, void *v)
499{
500 int ret;
501 int idx;
502
503 idx = srcu_read_lock(&nh->srcu);
504 ret = notifier_call_chain(&nh->head, val, v);
505 srcu_read_unlock(&nh->srcu, idx);
506 return ret;
507}
508
509EXPORT_SYMBOL_GPL(srcu_notifier_call_chain);
510
511/**
512 * srcu_init_notifier_head - Initialize an SRCU notifier head
513 * @nh: Pointer to head of the srcu notifier chain
514 *
515 * Unlike other sorts of notifier heads, SRCU notifier heads require
516 * dynamic initialization. Be sure to call this routine before
517 * calling any of the other SRCU notifier routines for this head.
518 *
519 * If an SRCU notifier head is deallocated, it must first be cleaned
520 * up by calling srcu_cleanup_notifier_head(). Otherwise the head's
521 * per-cpu data (used by the SRCU mechanism) will leak.
522 */
523
524void srcu_init_notifier_head(struct srcu_notifier_head *nh)
525{
526 mutex_init(&nh->mutex);
e6a92013
AS
527 if (init_srcu_struct(&nh->srcu) < 0)
528 BUG();
eabc0694
AS
529 nh->head = NULL;
530}
531
532EXPORT_SYMBOL_GPL(srcu_init_notifier_head);
533
1da177e4
LT
534/**
535 * register_reboot_notifier - Register function to be called at reboot time
536 * @nb: Info about notifier function to be called
537 *
538 * Registers a function with the list of functions
539 * to be called at reboot time.
540 *
72fd4a35 541 * Currently always returns zero, as blocking_notifier_chain_register()
1da177e4
LT
542 * always returns zero.
543 */
544
545int register_reboot_notifier(struct notifier_block * nb)
546{
e041c683 547 return blocking_notifier_chain_register(&reboot_notifier_list, nb);
1da177e4
LT
548}
549
550EXPORT_SYMBOL(register_reboot_notifier);
551
552/**
553 * unregister_reboot_notifier - Unregister previously registered reboot notifier
554 * @nb: Hook to be unregistered
555 *
556 * Unregisters a previously registered reboot
557 * notifier function.
558 *
559 * Returns zero on success, or %-ENOENT on failure.
560 */
561
562int unregister_reboot_notifier(struct notifier_block * nb)
563{
e041c683 564 return blocking_notifier_chain_unregister(&reboot_notifier_list, nb);
1da177e4
LT
565}
566
567EXPORT_SYMBOL(unregister_reboot_notifier);
568
569static int set_one_prio(struct task_struct *p, int niceval, int error)
570{
571 int no_nice;
572
573 if (p->uid != current->euid &&
574 p->euid != current->euid && !capable(CAP_SYS_NICE)) {
575 error = -EPERM;
576 goto out;
577 }
e43379f1 578 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
1da177e4
LT
579 error = -EACCES;
580 goto out;
581 }
582 no_nice = security_task_setnice(p, niceval);
583 if (no_nice) {
584 error = no_nice;
585 goto out;
586 }
587 if (error == -ESRCH)
588 error = 0;
589 set_user_nice(p, niceval);
590out:
591 return error;
592}
593
594asmlinkage long sys_setpriority(int which, int who, int niceval)
595{
596 struct task_struct *g, *p;
597 struct user_struct *user;
598 int error = -EINVAL;
41487c65 599 struct pid *pgrp;
1da177e4
LT
600
601 if (which > 2 || which < 0)
602 goto out;
603
604 /* normalize: avoid signed division (rounding problems) */
605 error = -ESRCH;
606 if (niceval < -20)
607 niceval = -20;
608 if (niceval > 19)
609 niceval = 19;
610
611 read_lock(&tasklist_lock);
612 switch (which) {
613 case PRIO_PROCESS:
41487c65
EB
614 if (who)
615 p = find_task_by_pid(who);
616 else
617 p = current;
1da177e4
LT
618 if (p)
619 error = set_one_prio(p, niceval, error);
620 break;
621 case PRIO_PGRP:
41487c65
EB
622 if (who)
623 pgrp = find_pid(who);
624 else
625 pgrp = task_pgrp(current);
626 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
1da177e4 627 error = set_one_prio(p, niceval, error);
41487c65 628 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
1da177e4
LT
629 break;
630 case PRIO_USER:
631 user = current->user;
632 if (!who)
633 who = current->uid;
634 else
635 if ((who != current->uid) && !(user = find_user(who)))
636 goto out_unlock; /* No processes for this user */
637
638 do_each_thread(g, p)
639 if (p->uid == who)
640 error = set_one_prio(p, niceval, error);
641 while_each_thread(g, p);
642 if (who != current->uid)
643 free_uid(user); /* For find_user() */
644 break;
645 }
646out_unlock:
647 read_unlock(&tasklist_lock);
648out:
649 return error;
650}
651
652/*
653 * Ugh. To avoid negative return values, "getpriority()" will
654 * not return the normal nice-value, but a negated value that
655 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
656 * to stay compatible.
657 */
658asmlinkage long sys_getpriority(int which, int who)
659{
660 struct task_struct *g, *p;
661 struct user_struct *user;
662 long niceval, retval = -ESRCH;
41487c65 663 struct pid *pgrp;
1da177e4
LT
664
665 if (which > 2 || which < 0)
666 return -EINVAL;
667
668 read_lock(&tasklist_lock);
669 switch (which) {
670 case PRIO_PROCESS:
41487c65
EB
671 if (who)
672 p = find_task_by_pid(who);
673 else
674 p = current;
1da177e4
LT
675 if (p) {
676 niceval = 20 - task_nice(p);
677 if (niceval > retval)
678 retval = niceval;
679 }
680 break;
681 case PRIO_PGRP:
41487c65
EB
682 if (who)
683 pgrp = find_pid(who);
684 else
685 pgrp = task_pgrp(current);
686 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
1da177e4
LT
687 niceval = 20 - task_nice(p);
688 if (niceval > retval)
689 retval = niceval;
41487c65 690 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
1da177e4
LT
691 break;
692 case PRIO_USER:
693 user = current->user;
694 if (!who)
695 who = current->uid;
696 else
697 if ((who != current->uid) && !(user = find_user(who)))
698 goto out_unlock; /* No processes for this user */
699
700 do_each_thread(g, p)
701 if (p->uid == who) {
702 niceval = 20 - task_nice(p);
703 if (niceval > retval)
704 retval = niceval;
705 }
706 while_each_thread(g, p);
707 if (who != current->uid)
708 free_uid(user); /* for find_user() */
709 break;
710 }
711out_unlock:
712 read_unlock(&tasklist_lock);
713
714 return retval;
715}
716
e4c94330
EB
717/**
718 * emergency_restart - reboot the system
719 *
720 * Without shutting down any hardware or taking any locks
721 * reboot the system. This is called when we know we are in
722 * trouble so this is our best effort to reboot. This is
723 * safe to call in interrupt context.
724 */
7c903473
EB
725void emergency_restart(void)
726{
727 machine_emergency_restart();
728}
729EXPORT_SYMBOL_GPL(emergency_restart);
730
83cc5ed3 731static void kernel_restart_prepare(char *cmd)
4a00ea1e 732{
e041c683 733 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
4a00ea1e 734 system_state = SYSTEM_RESTART;
4a00ea1e 735 device_shutdown();
e4c94330 736}
1e5d5331
RD
737
738/**
739 * kernel_restart - reboot the system
740 * @cmd: pointer to buffer containing command to execute for restart
b8887e6e 741 * or %NULL
1e5d5331
RD
742 *
743 * Shutdown everything and perform a clean reboot.
744 * This is not safe to call in interrupt context.
745 */
e4c94330
EB
746void kernel_restart(char *cmd)
747{
748 kernel_restart_prepare(cmd);
756184b7 749 if (!cmd)
4a00ea1e 750 printk(KERN_EMERG "Restarting system.\n");
756184b7 751 else
4a00ea1e 752 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
4a00ea1e
EB
753 machine_restart(cmd);
754}
755EXPORT_SYMBOL_GPL(kernel_restart);
756
e4c94330
EB
757/**
758 * kernel_kexec - reboot the system
759 *
760 * Move into place and start executing a preloaded standalone
761 * executable. If nothing was preloaded return an error.
762 */
83cc5ed3 763static void kernel_kexec(void)
4a00ea1e
EB
764{
765#ifdef CONFIG_KEXEC
766 struct kimage *image;
4bb8089c 767 image = xchg(&kexec_image, NULL);
756184b7 768 if (!image)
4a00ea1e 769 return;
e4c94330 770 kernel_restart_prepare(NULL);
4a00ea1e
EB
771 printk(KERN_EMERG "Starting new kernel\n");
772 machine_shutdown();
773 machine_kexec(image);
774#endif
775}
4a00ea1e 776
729b4d4c
AS
777void kernel_shutdown_prepare(enum system_states state)
778{
e041c683 779 blocking_notifier_call_chain(&reboot_notifier_list,
729b4d4c
AS
780 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
781 system_state = state;
782 device_shutdown();
783}
e4c94330
EB
784/**
785 * kernel_halt - halt the system
786 *
787 * Shutdown everything and perform a clean system halt.
788 */
e4c94330
EB
789void kernel_halt(void)
790{
729b4d4c 791 kernel_shutdown_prepare(SYSTEM_HALT);
4a00ea1e
EB
792 printk(KERN_EMERG "System halted.\n");
793 machine_halt();
794}
729b4d4c 795
4a00ea1e
EB
796EXPORT_SYMBOL_GPL(kernel_halt);
797
e4c94330
EB
798/**
799 * kernel_power_off - power_off the system
800 *
801 * Shutdown everything and perform a clean system power_off.
802 */
e4c94330
EB
803void kernel_power_off(void)
804{
729b4d4c 805 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
4a00ea1e
EB
806 printk(KERN_EMERG "Power down.\n");
807 machine_power_off();
808}
809EXPORT_SYMBOL_GPL(kernel_power_off);
1da177e4
LT
810/*
811 * Reboot system call: for obvious reasons only root may call it,
812 * and even root needs to set up some magic numbers in the registers
813 * so that some mistake won't make this reboot the whole machine.
814 * You can also set the meaning of the ctrl-alt-del-key here.
815 *
816 * reboot doesn't sync: do that yourself before calling this.
817 */
818asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
819{
820 char buffer[256];
821
822 /* We only trust the superuser with rebooting the system. */
823 if (!capable(CAP_SYS_BOOT))
824 return -EPERM;
825
826 /* For safety, we require "magic" arguments. */
827 if (magic1 != LINUX_REBOOT_MAGIC1 ||
828 (magic2 != LINUX_REBOOT_MAGIC2 &&
829 magic2 != LINUX_REBOOT_MAGIC2A &&
830 magic2 != LINUX_REBOOT_MAGIC2B &&
831 magic2 != LINUX_REBOOT_MAGIC2C))
832 return -EINVAL;
833
5e38291d
EB
834 /* Instead of trying to make the power_off code look like
835 * halt when pm_power_off is not set do it the easy way.
836 */
837 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
838 cmd = LINUX_REBOOT_CMD_HALT;
839
1da177e4
LT
840 lock_kernel();
841 switch (cmd) {
842 case LINUX_REBOOT_CMD_RESTART:
4a00ea1e 843 kernel_restart(NULL);
1da177e4
LT
844 break;
845
846 case LINUX_REBOOT_CMD_CAD_ON:
847 C_A_D = 1;
848 break;
849
850 case LINUX_REBOOT_CMD_CAD_OFF:
851 C_A_D = 0;
852 break;
853
854 case LINUX_REBOOT_CMD_HALT:
4a00ea1e 855 kernel_halt();
1da177e4
LT
856 unlock_kernel();
857 do_exit(0);
858 break;
859
860 case LINUX_REBOOT_CMD_POWER_OFF:
4a00ea1e 861 kernel_power_off();
1da177e4
LT
862 unlock_kernel();
863 do_exit(0);
864 break;
865
866 case LINUX_REBOOT_CMD_RESTART2:
867 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
868 unlock_kernel();
869 return -EFAULT;
870 }
871 buffer[sizeof(buffer) - 1] = '\0';
872
4a00ea1e 873 kernel_restart(buffer);
1da177e4
LT
874 break;
875
dc009d92 876 case LINUX_REBOOT_CMD_KEXEC:
4a00ea1e
EB
877 kernel_kexec();
878 unlock_kernel();
879 return -EINVAL;
880
1da177e4
LT
881#ifdef CONFIG_SOFTWARE_SUSPEND
882 case LINUX_REBOOT_CMD_SW_SUSPEND:
883 {
884 int ret = software_suspend();
885 unlock_kernel();
886 return ret;
887 }
888#endif
889
890 default:
891 unlock_kernel();
892 return -EINVAL;
893 }
894 unlock_kernel();
895 return 0;
896}
897
65f27f38 898static void deferred_cad(struct work_struct *dummy)
1da177e4 899{
abcd9e51 900 kernel_restart(NULL);
1da177e4
LT
901}
902
903/*
904 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
905 * As it's called within an interrupt, it may NOT sync: the only choice
906 * is whether to reboot at once, or just ignore the ctrl-alt-del.
907 */
908void ctrl_alt_del(void)
909{
65f27f38 910 static DECLARE_WORK(cad_work, deferred_cad);
1da177e4
LT
911
912 if (C_A_D)
913 schedule_work(&cad_work);
914 else
9ec52099 915 kill_cad_pid(SIGINT, 1);
1da177e4
LT
916}
917
1da177e4
LT
918/*
919 * Unprivileged users may change the real gid to the effective gid
920 * or vice versa. (BSD-style)
921 *
922 * If you set the real gid at all, or set the effective gid to a value not
923 * equal to the real gid, then the saved gid is set to the new effective gid.
924 *
925 * This makes it possible for a setgid program to completely drop its
926 * privileges, which is often a useful assertion to make when you are doing
927 * a security audit over a program.
928 *
929 * The general idea is that a program which uses just setregid() will be
930 * 100% compatible with BSD. A program which uses just setgid() will be
931 * 100% compatible with POSIX with saved IDs.
932 *
933 * SMP: There are not races, the GIDs are checked only by filesystem
934 * operations (as far as semantic preservation is concerned).
935 */
936asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
937{
938 int old_rgid = current->gid;
939 int old_egid = current->egid;
940 int new_rgid = old_rgid;
941 int new_egid = old_egid;
942 int retval;
943
944 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
945 if (retval)
946 return retval;
947
948 if (rgid != (gid_t) -1) {
949 if ((old_rgid == rgid) ||
950 (current->egid==rgid) ||
951 capable(CAP_SETGID))
952 new_rgid = rgid;
953 else
954 return -EPERM;
955 }
956 if (egid != (gid_t) -1) {
957 if ((old_rgid == egid) ||
958 (current->egid == egid) ||
959 (current->sgid == egid) ||
960 capable(CAP_SETGID))
961 new_egid = egid;
756184b7 962 else
1da177e4 963 return -EPERM;
1da177e4 964 }
756184b7 965 if (new_egid != old_egid) {
d6e71144 966 current->mm->dumpable = suid_dumpable;
d59dd462 967 smp_wmb();
1da177e4
LT
968 }
969 if (rgid != (gid_t) -1 ||
970 (egid != (gid_t) -1 && egid != old_rgid))
971 current->sgid = new_egid;
972 current->fsgid = new_egid;
973 current->egid = new_egid;
974 current->gid = new_rgid;
975 key_fsgid_changed(current);
9f46080c 976 proc_id_connector(current, PROC_EVENT_GID);
1da177e4
LT
977 return 0;
978}
979
980/*
981 * setgid() is implemented like SysV w/ SAVED_IDS
982 *
983 * SMP: Same implicit races as above.
984 */
985asmlinkage long sys_setgid(gid_t gid)
986{
987 int old_egid = current->egid;
988 int retval;
989
990 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
991 if (retval)
992 return retval;
993
756184b7
CP
994 if (capable(CAP_SETGID)) {
995 if (old_egid != gid) {
d6e71144 996 current->mm->dumpable = suid_dumpable;
d59dd462 997 smp_wmb();
1da177e4
LT
998 }
999 current->gid = current->egid = current->sgid = current->fsgid = gid;
756184b7
CP
1000 } else if ((gid == current->gid) || (gid == current->sgid)) {
1001 if (old_egid != gid) {
d6e71144 1002 current->mm->dumpable = suid_dumpable;
d59dd462 1003 smp_wmb();
1da177e4
LT
1004 }
1005 current->egid = current->fsgid = gid;
1006 }
1007 else
1008 return -EPERM;
1009
1010 key_fsgid_changed(current);
9f46080c 1011 proc_id_connector(current, PROC_EVENT_GID);
1da177e4
LT
1012 return 0;
1013}
1014
1015static int set_user(uid_t new_ruid, int dumpclear)
1016{
1017 struct user_struct *new_user;
1018
1019 new_user = alloc_uid(new_ruid);
1020 if (!new_user)
1021 return -EAGAIN;
1022
1023 if (atomic_read(&new_user->processes) >=
1024 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
1025 new_user != &root_user) {
1026 free_uid(new_user);
1027 return -EAGAIN;
1028 }
1029
1030 switch_uid(new_user);
1031
756184b7 1032 if (dumpclear) {
d6e71144 1033 current->mm->dumpable = suid_dumpable;
d59dd462 1034 smp_wmb();
1da177e4
LT
1035 }
1036 current->uid = new_ruid;
1037 return 0;
1038}
1039
1040/*
1041 * Unprivileged users may change the real uid to the effective uid
1042 * or vice versa. (BSD-style)
1043 *
1044 * If you set the real uid at all, or set the effective uid to a value not
1045 * equal to the real uid, then the saved uid is set to the new effective uid.
1046 *
1047 * This makes it possible for a setuid program to completely drop its
1048 * privileges, which is often a useful assertion to make when you are doing
1049 * a security audit over a program.
1050 *
1051 * The general idea is that a program which uses just setreuid() will be
1052 * 100% compatible with BSD. A program which uses just setuid() will be
1053 * 100% compatible with POSIX with saved IDs.
1054 */
1055asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
1056{
1057 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
1058 int retval;
1059
1060 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
1061 if (retval)
1062 return retval;
1063
1064 new_ruid = old_ruid = current->uid;
1065 new_euid = old_euid = current->euid;
1066 old_suid = current->suid;
1067
1068 if (ruid != (uid_t) -1) {
1069 new_ruid = ruid;
1070 if ((old_ruid != ruid) &&
1071 (current->euid != ruid) &&
1072 !capable(CAP_SETUID))
1073 return -EPERM;
1074 }
1075
1076 if (euid != (uid_t) -1) {
1077 new_euid = euid;
1078 if ((old_ruid != euid) &&
1079 (current->euid != euid) &&
1080 (current->suid != euid) &&
1081 !capable(CAP_SETUID))
1082 return -EPERM;
1083 }
1084
1085 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
1086 return -EAGAIN;
1087
756184b7 1088 if (new_euid != old_euid) {
d6e71144 1089 current->mm->dumpable = suid_dumpable;
d59dd462 1090 smp_wmb();
1da177e4
LT
1091 }
1092 current->fsuid = current->euid = new_euid;
1093 if (ruid != (uid_t) -1 ||
1094 (euid != (uid_t) -1 && euid != old_ruid))
1095 current->suid = current->euid;
1096 current->fsuid = current->euid;
1097
1098 key_fsuid_changed(current);
9f46080c 1099 proc_id_connector(current, PROC_EVENT_UID);
1da177e4
LT
1100
1101 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
1102}
1103
1104
1105
1106/*
1107 * setuid() is implemented like SysV with SAVED_IDS
1108 *
1109 * Note that SAVED_ID's is deficient in that a setuid root program
1110 * like sendmail, for example, cannot set its uid to be a normal
1111 * user and then switch back, because if you're root, setuid() sets
1112 * the saved uid too. If you don't like this, blame the bright people
1113 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
1114 * will allow a root program to temporarily drop privileges and be able to
1115 * regain them by swapping the real and effective uid.
1116 */
1117asmlinkage long sys_setuid(uid_t uid)
1118{
1119 int old_euid = current->euid;
a09c17a6 1120 int old_ruid, old_suid, new_suid;
1da177e4
LT
1121 int retval;
1122
1123 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
1124 if (retval)
1125 return retval;
1126
a09c17a6 1127 old_ruid = current->uid;
1da177e4
LT
1128 old_suid = current->suid;
1129 new_suid = old_suid;
1130
1131 if (capable(CAP_SETUID)) {
1132 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
1133 return -EAGAIN;
1134 new_suid = uid;
1135 } else if ((uid != current->uid) && (uid != new_suid))
1136 return -EPERM;
1137
756184b7 1138 if (old_euid != uid) {
d6e71144 1139 current->mm->dumpable = suid_dumpable;
d59dd462 1140 smp_wmb();
1da177e4
LT
1141 }
1142 current->fsuid = current->euid = uid;
1143 current->suid = new_suid;
1144
1145 key_fsuid_changed(current);
9f46080c 1146 proc_id_connector(current, PROC_EVENT_UID);
1da177e4
LT
1147
1148 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
1149}
1150
1151
1152/*
1153 * This function implements a generic ability to update ruid, euid,
1154 * and suid. This allows you to implement the 4.4 compatible seteuid().
1155 */
1156asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
1157{
1158 int old_ruid = current->uid;
1159 int old_euid = current->euid;
1160 int old_suid = current->suid;
1161 int retval;
1162
1163 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
1164 if (retval)
1165 return retval;
1166
1167 if (!capable(CAP_SETUID)) {
1168 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
1169 (ruid != current->euid) && (ruid != current->suid))
1170 return -EPERM;
1171 if ((euid != (uid_t) -1) && (euid != current->uid) &&
1172 (euid != current->euid) && (euid != current->suid))
1173 return -EPERM;
1174 if ((suid != (uid_t) -1) && (suid != current->uid) &&
1175 (suid != current->euid) && (suid != current->suid))
1176 return -EPERM;
1177 }
1178 if (ruid != (uid_t) -1) {
1179 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
1180 return -EAGAIN;
1181 }
1182 if (euid != (uid_t) -1) {
756184b7 1183 if (euid != current->euid) {
d6e71144 1184 current->mm->dumpable = suid_dumpable;
d59dd462 1185 smp_wmb();
1da177e4
LT
1186 }
1187 current->euid = euid;
1188 }
1189 current->fsuid = current->euid;
1190 if (suid != (uid_t) -1)
1191 current->suid = suid;
1192
1193 key_fsuid_changed(current);
9f46080c 1194 proc_id_connector(current, PROC_EVENT_UID);
1da177e4
LT
1195
1196 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
1197}
1198
1199asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
1200{
1201 int retval;
1202
1203 if (!(retval = put_user(current->uid, ruid)) &&
1204 !(retval = put_user(current->euid, euid)))
1205 retval = put_user(current->suid, suid);
1206
1207 return retval;
1208}
1209
1210/*
1211 * Same as above, but for rgid, egid, sgid.
1212 */
1213asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
1214{
1215 int retval;
1216
1217 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
1218 if (retval)
1219 return retval;
1220
1221 if (!capable(CAP_SETGID)) {
1222 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
1223 (rgid != current->egid) && (rgid != current->sgid))
1224 return -EPERM;
1225 if ((egid != (gid_t) -1) && (egid != current->gid) &&
1226 (egid != current->egid) && (egid != current->sgid))
1227 return -EPERM;
1228 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
1229 (sgid != current->egid) && (sgid != current->sgid))
1230 return -EPERM;
1231 }
1232 if (egid != (gid_t) -1) {
756184b7 1233 if (egid != current->egid) {
d6e71144 1234 current->mm->dumpable = suid_dumpable;
d59dd462 1235 smp_wmb();
1da177e4
LT
1236 }
1237 current->egid = egid;
1238 }
1239 current->fsgid = current->egid;
1240 if (rgid != (gid_t) -1)
1241 current->gid = rgid;
1242 if (sgid != (gid_t) -1)
1243 current->sgid = sgid;
1244
1245 key_fsgid_changed(current);
9f46080c 1246 proc_id_connector(current, PROC_EVENT_GID);
1da177e4
LT
1247 return 0;
1248}
1249
1250asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
1251{
1252 int retval;
1253
1254 if (!(retval = put_user(current->gid, rgid)) &&
1255 !(retval = put_user(current->egid, egid)))
1256 retval = put_user(current->sgid, sgid);
1257
1258 return retval;
1259}
1260
1261
1262/*
1263 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
1264 * is used for "access()" and for the NFS daemon (letting nfsd stay at
1265 * whatever uid it wants to). It normally shadows "euid", except when
1266 * explicitly set by setfsuid() or for access..
1267 */
1268asmlinkage long sys_setfsuid(uid_t uid)
1269{
1270 int old_fsuid;
1271
1272 old_fsuid = current->fsuid;
1273 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
1274 return old_fsuid;
1275
1276 if (uid == current->uid || uid == current->euid ||
1277 uid == current->suid || uid == current->fsuid ||
756184b7
CP
1278 capable(CAP_SETUID)) {
1279 if (uid != old_fsuid) {
d6e71144 1280 current->mm->dumpable = suid_dumpable;
d59dd462 1281 smp_wmb();
1da177e4
LT
1282 }
1283 current->fsuid = uid;
1284 }
1285
1286 key_fsuid_changed(current);
9f46080c 1287 proc_id_connector(current, PROC_EVENT_UID);
1da177e4
LT
1288
1289 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
1290
1291 return old_fsuid;
1292}
1293
1294/*
1295