blob: 0742c938dfa70d4554d9d6b4ef37482259a0c776 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/kernel/sys.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
Linus Torvalds1da177e2005-04-16 15:20:36 -07007#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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070015#include <linux/highuid.h>
16#include <linux/fs.h>
Eric W. Biedermandc009d92005-06-25 14:57:52 -070017#include <linux/kernel.h>
18#include <linux/kexec.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include <linux/workqueue.h>
Randy.Dunlapc59ede72006-01-11 12:17:46 -080020#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021#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>
Jesper Juhl7ed20e12005-05-01 08:59:14 -070029#include <linux/signal.h>
Matt Helsley9f460802005-11-07 00:59:16 -080030#include <linux/cn_proc.h>
Andi Kleen3cfc3482006-09-26 10:52:28 +020031#include <linux/getcpu.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070032
33#include <linux/compat.h>
34#include <linux/syscalls.h>
Keshavamurthy Anil S00d7c052005-12-12 00:37:33 -080035#include <linux/kprobes.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070036
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
Anton Blanchard651d7652006-06-07 16:10:19 +100059#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
Linus Torvalds1da177e2005-04-16 15:20:36 -070065
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;
Cedric Le Goater9ec52092006-10-02 02:19:00 -070095struct pid *cad_pid;
96EXPORT_SYMBOL(cad_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -070097
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
Alan Sterne041c682006-03-27 01:16:30 -0800104static BLOCKING_NOTIFIER_HEAD(reboot_notifier_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105
Alan Sterne041c682006-03-27 01:16:30 -0800106/*
107 * Notifier chain core routines. The exported routines below
108 * are layered on top of these, with appropriate locking added.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109 */
Alan Sterne041c682006-03-27 01:16:30 -0800110
111static int notifier_chain_register(struct notifier_block **nl,
112 struct notifier_block *n)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113{
Alan Sterne041c682006-03-27 01:16:30 -0800114 while ((*nl) != NULL) {
115 if (n->priority > (*nl)->priority)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116 break;
Alan Sterne041c682006-03-27 01:16:30 -0800117 nl = &((*nl)->next);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118 }
Alan Sterne041c682006-03-27 01:16:30 -0800119 n->next = *nl;
120 rcu_assign_pointer(*nl, n);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121 return 0;
122}
123
Alan Sterne041c682006-03-27 01:16:30 -0800124static int notifier_chain_unregister(struct notifier_block **nl,
125 struct notifier_block *n)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126{
Alan Sterne041c682006-03-27 01:16:30 -0800127 while ((*nl) != NULL) {
128 if ((*nl) == n) {
129 rcu_assign_pointer(*nl, n->next);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130 return 0;
131 }
Alan Sterne041c682006-03-27 01:16:30 -0800132 nl = &((*nl)->next);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134 return -ENOENT;
135}
136
Alan Sterne041c682006-03-27 01:16:30 -0800137static int __kprobes notifier_call_chain(struct notifier_block **nl,
138 unsigned long val, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139{
Alan Sterne041c682006-03-27 01:16:30 -0800140 int ret = NOTIFY_DONE;
Alan Sternbbb17472006-06-25 05:47:15 -0700141 struct notifier_block *nb, *next_nb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142
Alan Sterne041c682006-03-27 01:16:30 -0800143 nb = rcu_dereference(*nl);
144 while (nb) {
Alan Sternbbb17472006-06-25 05:47:15 -0700145 next_nb = rcu_dereference(nb->next);
Alan Sterne041c682006-03-27 01:16:30 -0800146 ret = nb->notifier_call(nb, val, v);
147 if ((ret & NOTIFY_STOP_MASK) == NOTIFY_STOP_MASK)
148 break;
Alan Sternbbb17472006-06-25 05:47:15 -0700149 nb = next_nb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150 }
151 return ret;
152}
153
Alan Sterne041c682006-03-27 01:16:30 -0800154/*
155 * Atomic notifier chain routines. Registration and unregistration
Alan Sterneabc0692006-10-04 02:17:04 -0700156 * use a spinlock, and call_chain is synchronized by RCU (no locks).
Alan Sterne041c682006-03-27 01:16:30 -0800157 */
158
159/**
160 * atomic_notifier_chain_register - Add notifier to an atomic notifier chain
161 * @nh: Pointer to head of the atomic notifier chain
162 * @n: New entry in notifier chain
163 *
164 * Adds a notifier to an atomic notifier chain.
165 *
166 * Currently always returns zero.
167 */
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
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800218 * with %NOTIFY_STOP_MASK then atomic_notifier_call_chain()
Alan Sterne041c682006-03-27 01:16:30 -0800219 * 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 */
224
bibo,maof2aa85a2006-10-02 02:17:34 -0700225int __kprobes atomic_notifier_call_chain(struct atomic_notifier_head *nh,
Alan Sterne041c682006-03-27 01:16:30 -0800226 unsigned long val, void *v)
227{
228 int ret;
229
230 rcu_read_lock();
231 ret = notifier_call_chain(&nh->head, val, v);
232 rcu_read_unlock();
233 return ret;
234}
235
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 */
242
243/**
244 * blocking_notifier_chain_register - Add notifier to a blocking notifier chain
245 * @nh: Pointer to head of the blocking notifier chain
246 * @n: New entry in notifier chain
247 *
248 * Adds a notifier to a blocking notifier chain.
249 * Must be called in process context.
250 *
251 * Currently always returns zero.
252 */
253
254int blocking_notifier_chain_register(struct blocking_notifier_head *nh,
255 struct notifier_block *n)
256{
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;
271}
272
273EXPORT_SYMBOL_GPL(blocking_notifier_chain_register);
274
275/**
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
309 * @val: Value passed unmodified to notifier function
310 * @v: Pointer passed unmodified to notifier function
311 *
312 * Calls each function in a notifier chain in turn. The functions
313 * run in a process context, so they are allowed to block.
314 *
315 * If the return value of the notifier can be and'ed
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800316 * with %NOTIFY_STOP_MASK then blocking_notifier_call_chain()
Alan Sterne041c682006-03-27 01:16:30 -0800317 * will return immediately, with the return value of
318 * the notifier function which halted execution.
319 * Otherwise the return value is the return value
320 * of the last notifier function called.
321 */
322
323int blocking_notifier_call_chain(struct blocking_notifier_head *nh,
324 unsigned long val, void *v)
325{
Ingo Molnar1b5180b2007-01-23 10:45:50 +0100326 int ret = NOTIFY_DONE;
Alan Sterne041c682006-03-27 01:16:30 -0800327
Ingo Molnar1b5180b2007-01-23 10:45:50 +0100328 /*
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 }
Alan Sterne041c682006-03-27 01:16:30 -0800338 return ret;
339}
340
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
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800396 * with %NOTIFY_STOP_MASK then raw_notifier_call_chain()
Alan Sterne041c682006-03-27 01:16:30 -0800397 * 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);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410
Alan Sterneabc0692006-10-04 02:17:04 -0700411/*
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
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800490 * with %NOTIFY_STOP_MASK then srcu_notifier_call_chain()
Alan Sterneabc0692006-10-04 02:17:04 -0700491 * 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);
Alan Sterne6a92012006-10-04 02:17:05 -0700527 if (init_srcu_struct(&nh->srcu) < 0)
528 BUG();
Alan Sterneabc0692006-10-04 02:17:04 -0700529 nh->head = NULL;
530}
531
532EXPORT_SYMBOL_GPL(srcu_init_notifier_head);
533
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534/**
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 *
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800541 * Currently always returns zero, as blocking_notifier_chain_register()
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 * always returns zero.
543 */
544
545int register_reboot_notifier(struct notifier_block * nb)
546{
Alan Sterne041c682006-03-27 01:16:30 -0800547 return blocking_notifier_chain_register(&reboot_notifier_list, nb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548}
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{
Alan Sterne041c682006-03-27 01:16:30 -0800564 return blocking_notifier_chain_unregister(&reboot_notifier_list, nb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565}
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 }
Matt Mackalle43379f2005-05-01 08:59:00 -0700578 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579 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;
Eric W. Biederman41487c62007-02-12 00:53:01 -0800599 struct pid *pgrp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600
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:
Eric W. Biederman41487c62007-02-12 00:53:01 -0800614 if (who)
615 p = find_task_by_pid(who);
616 else
617 p = current;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618 if (p)
619 error = set_one_prio(p, niceval, error);
620 break;
621 case PRIO_PGRP:
Eric W. Biederman41487c62007-02-12 00:53:01 -0800622 if (who)
623 pgrp = find_pid(who);
624 else
625 pgrp = task_pgrp(current);
626 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627 error = set_one_prio(p, niceval, error);
Eric W. Biederman41487c62007-02-12 00:53:01 -0800628 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629 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;
Eric W. Biederman41487c62007-02-12 00:53:01 -0800663 struct pid *pgrp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664
665 if (which > 2 || which < 0)
666 return -EINVAL;
667
668 read_lock(&tasklist_lock);
669 switch (which) {
670 case PRIO_PROCESS:
Eric W. Biederman41487c62007-02-12 00:53:01 -0800671 if (who)
672 p = find_task_by_pid(who);
673 else
674 p = current;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675 if (p) {
676 niceval = 20 - task_nice(p);
677 if (niceval > retval)
678 retval = niceval;
679 }
680 break;
681 case PRIO_PGRP:
Eric W. Biederman41487c62007-02-12 00:53:01 -0800682 if (who)
683 pgrp = find_pid(who);
684 else
685 pgrp = task_pgrp(current);
686 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687 niceval = 20 - task_nice(p);
688 if (niceval > retval)
689 retval = niceval;
Eric W. Biederman41487c62007-02-12 00:53:01 -0800690 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691 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
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700717/**
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 */
Eric W. Biederman7c903472005-07-26 11:29:55 -0600725void emergency_restart(void)
726{
727 machine_emergency_restart();
728}
729EXPORT_SYMBOL_GPL(emergency_restart);
730
Adrian Bunk83cc5ed2006-06-25 05:47:41 -0700731static void kernel_restart_prepare(char *cmd)
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600732{
Alan Sterne041c682006-03-27 01:16:30 -0800733 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600734 system_state = SYSTEM_RESTART;
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600735 device_shutdown();
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700736}
Randy Dunlap1e5d5332005-11-07 01:01:06 -0800737
738/**
739 * kernel_restart - reboot the system
740 * @cmd: pointer to buffer containing command to execute for restart
Randy Dunlapb8887e62005-11-07 01:01:07 -0800741 * or %NULL
Randy Dunlap1e5d5332005-11-07 01:01:06 -0800742 *
743 * Shutdown everything and perform a clean reboot.
744 * This is not safe to call in interrupt context.
745 */
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700746void kernel_restart(char *cmd)
747{
748 kernel_restart_prepare(cmd);
Cal Peake756184b2006-09-30 23:27:24 -0700749 if (!cmd)
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600750 printk(KERN_EMERG "Restarting system.\n");
Cal Peake756184b2006-09-30 23:27:24 -0700751 else
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600752 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600753 machine_restart(cmd);
754}
755EXPORT_SYMBOL_GPL(kernel_restart);
756
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700757/**
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 */
Adrian Bunk83cc5ed2006-06-25 05:47:41 -0700763static void kernel_kexec(void)
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600764{
765#ifdef CONFIG_KEXEC
766 struct kimage *image;
Al Viro4bb80892006-02-01 05:57:32 -0500767 image = xchg(&kexec_image, NULL);
Cal Peake756184b2006-09-30 23:27:24 -0700768 if (!image)
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600769 return;
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700770 kernel_restart_prepare(NULL);
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600771 printk(KERN_EMERG "Starting new kernel\n");
772 machine_shutdown();
773 machine_kexec(image);
774#endif
775}
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600776
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500777void kernel_shutdown_prepare(enum system_states state)
778{
Alan Sterne041c682006-03-27 01:16:30 -0800779 blocking_notifier_call_chain(&reboot_notifier_list,
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500780 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
781 system_state = state;
782 device_shutdown();
783}
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700784/**
785 * kernel_halt - halt the system
786 *
787 * Shutdown everything and perform a clean system halt.
788 */
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700789void kernel_halt(void)
790{
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500791 kernel_shutdown_prepare(SYSTEM_HALT);
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600792 printk(KERN_EMERG "System halted.\n");
793 machine_halt();
794}
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500795
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600796EXPORT_SYMBOL_GPL(kernel_halt);
797
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700798/**
799 * kernel_power_off - power_off the system
800 *
801 * Shutdown everything and perform a clean system power_off.
802 */
Eric W. Biedermane4c94332005-09-22 21:43:45 -0700803void kernel_power_off(void)
804{
Alexey Starikovskiy729b4d42005-12-01 04:29:00 -0500805 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600806 printk(KERN_EMERG "Power down.\n");
807 machine_power_off();
808}
809EXPORT_SYMBOL_GPL(kernel_power_off);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810/*
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
Eric W. Biederman5e382912006-01-08 01:03:46 -0800834 /* 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
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840 lock_kernel();
841 switch (cmd) {
842 case LINUX_REBOOT_CMD_RESTART:
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600843 kernel_restart(NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844 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:
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600855 kernel_halt();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 unlock_kernel();
857 do_exit(0);
858 break;
859
860 case LINUX_REBOOT_CMD_POWER_OFF:
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600861 kernel_power_off();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 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
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600873 kernel_restart(buffer);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 break;
875
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700876 case LINUX_REBOOT_CMD_KEXEC:
Eric W. Biederman4a00ea12005-07-26 11:24:14 -0600877 kernel_kexec();
878 unlock_kernel();
879 return -EINVAL;
880
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881#ifdef CONFIG_SOFTWARE_SUSPEND
882 case LINUX_REBOOT_CMD_SW_SUSPEND:
883 {
Johannes Bergab3bfca2007-05-06 14:50:49 -0700884 int ret = pm_suspend(PM_SUSPEND_DISK);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885 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
David Howells65f27f32006-11-22 14:55:48 +0000898static void deferred_cad(struct work_struct *dummy)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899{
Eric W. Biedermanabcd9e52005-07-26 11:27:34 -0600900 kernel_restart(NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901}
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{
David Howells65f27f32006-11-22 14:55:48 +0000910 static DECLARE_WORK(cad_work, deferred_cad);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911
912 if (C_A_D)
913 schedule_work(&cad_work);
914 else
Cedric Le Goater9ec52092006-10-02 02:19:00 -0700915 kill_cad_pid(SIGINT, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916}
917
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918/*
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;
Cal Peake756184b2006-09-30 23:27:24 -0700962 else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 return -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964 }
Cal Peake756184b2006-09-30 23:27:24 -0700965 if (new_egid != old_egid) {
Alan Coxd6e71142005-06-23 00:09:43 -0700966 current->mm->dumpable = suid_dumpable;
akpm@osdl.orgd59dd462005-05-01 08:58:47 -0700967 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 }
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);
Matt Helsley9f460802005-11-07 00:59:16 -0800976 proc_id_connector(current, PROC_EVENT_GID);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 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
Cal Peake756184b2006-09-30 23:27:24 -0700994 if (capable(CAP_SETGID)) {
995 if (old_egid != gid) {
Alan Coxd6e71142005-06-23 00:09:43 -0700996 current->mm->dumpable = suid_dumpable;
akpm@osdl.orgd59dd462005-05-01 08:58:47 -0700997 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998 }
999 current->gid = current->egid = current->sgid = current->fsgid = gid;
Cal Peake756184b2006-09-30 23:27:24 -07001000 } else if ((gid == current->gid) || (gid == current->sgid)) {
1001 if (old_egid != gid) {
Alan Coxd6e71142005-06-23 00:09:43 -07001002 current->mm->dumpable = suid_dumpable;
akpm@osdl.orgd59dd462005-05-01 08:58:47 -07001003 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 }
1005 current->egid = current->fsgid = gid;
1006 }
1007 else
1008 return -EPERM;
1009
1010 key_fsgid_changed(current);
Matt Helsley9f460802005-11-07 00:59:16 -08001011 proc_id_connector(current, PROC_EVENT_GID);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012 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
Cal Peake756184b2006-09-30 23:27:24 -07001032 if (dumpclear) {
Alan Coxd6e71142005-06-23 00:09:43 -07001033 current->mm->dumpable = suid_dumpable;
akpm@osdl.orgd59dd462005-05-01 08:58:47 -07001034 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 }
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
Cal Peake756184b2006-09-30 23:27:24 -07001088 if (new_euid != old_euid) {
Alan Coxd6e71142005-06-23 00:09:43 -07001089 current->mm->dumpable = suid_dumpable;
akpm@osdl.orgd59dd462005-05-01 08:58:47 -07001090 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091 }
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);
Matt Helsley9f460802005-11-07 00:59:16 -08001099 proc_id_connector(current, PROC_EVENT_UID);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100
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;
David Rientjesa09c17a2006-12-06 20:40:18 -08001120 int old_ruid, old_suid, new_suid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121 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
David Rientjesa09c17a2006-12-06 20:40:18 -08001127 old_ruid = current->uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128 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
Cal Peake756184b2006-09-30 23:27:24 -07001138 if (old_euid != uid) {
Alan Coxd6e71142005-06-23 00:09:43 -07001139 current->mm->dumpable = suid_dumpable;
akpm@osdl.orgd59dd462005-05-01 08:58:47 -07001140 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141 }
1142 current->fsuid = current->euid = uid;
1143 current->suid = new_suid;
1144
1145 key_fsuid_changed(current);
Matt Helsley9f460802005-11-07 00:59:16 -08001146 proc_id_connector(current, PROC_EVENT_UID);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147
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) {
Cal Peake756184b2006-09-30 23:27:24 -07001183 if (euid != current->euid) {
Alan Coxd6e71142005-06-23 00:09:43 -07001184 current->mm->dumpable = suid_dumpable;
akpm@osdl.orgd59dd462005-05-01 08:58:47 -07001185 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186 }
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);
Matt Helsley9f460802005-11-07 00:59:16 -08001194 proc_id_connector(current, PROC_EVENT_UID);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195
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) {
Cal Peake756184b2006-09-30 23:27:24 -07001233 if (egid != current->egid) {
Alan Coxd6e71142005-06-23 00:09:43 -07001234 current->mm->dumpable = suid_dumpable;
akpm@osdl.orgd59dd462005-05-01 08:58:47 -07001235 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001236 }
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);
Matt Helsley9f460802005-11-07 00:59:16 -08001246 proc_id_connector(current, PROC_EVENT_GID);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247 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 ||
Cal Peake756184b2006-09-30 23:27:24 -07001278 capable(CAP_SETUID)) {
1279 if (uid != old_fsuid) {
Alan Coxd6e71142005-06-23 00:09:43 -07001280 current->mm->dumpable = suid_dumpable;
akpm@osdl.orgd59dd462005-05-01 08:58:47 -07001281 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282 }
1283 current->fsuid = uid;
1284 }
1285
1286 key_fsuid_changed(current);
Matt Helsley9f460802005-11-07 00:59:16 -08001287 proc_id_connector(current, PROC_EVENT_UID);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288
1289 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
1290
1291 return old_fsuid;
1292}
1293
1294/*
John Anthony Kazos Jrf42df9e2007-05-09 08:23:08 +02001295 * Samma på svenska..
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296 */
1297asmlinkage long sys_setfsgid(gid_t gid)
1298{
1299 int old_fsgid;
1300
1301 old_fsgid = current->fsgid;
1302 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
1303 return old_fsgid;
1304
1305 if (gid == current->gid || gid == current->egid ||
1306 gid == current->sgid || gid == current->fsgid ||
Cal Peake756184b2006-09-30 23:27:24 -07001307 capable(CAP_SETGID)) {
1308 if (gid != old_fsgid) {
Alan Coxd6e71142005-06-23 00:09:43 -07001309 current->mm->dumpable = suid_dumpable;
akpm@osdl.orgd59dd462005-05-01 08:58:47 -07001310 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001311 }
1312 current->fsgid = gid;
1313 key_fsgid_changed(current);
Matt Helsley9f460802005-11-07 00:59:16 -08001314 proc_id_connector(current, PROC_EVENT_GID);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315 }
1316 return old_fsgid;
1317}
1318
1319asmlinkage long sys_times(struct tms __user * tbuf)
1320{
1321 /*
1322 * In the SMP world we might just be unlucky and have one of
1323 * the times increment as we use it. Since the value is an
1324 * atomically safe type this is just fine. Conceptually its
1325 * as if the syscall took an instant longer to occur.
1326 */
1327 if (tbuf) {
1328 struct tms tmp;
Oleg Nesterov35f5cad2006-03-28 16:11:19 -08001329 struct task_struct *tsk = current;
1330 struct task_struct *t;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331 cputime_t utime, stime, cutime, cstime;
1332
Oleg Nesterov7d7185c2006-03-28 16:11:21 -08001333 spin_lock_irq(&tsk->sighand->siglock);
Oleg Nesterov35f5cad2006-03-28 16:11:19 -08001334 utime = tsk->signal->utime;
1335 stime = tsk->signal->stime;
1336 t = tsk;
1337 do {
1338 utime = cputime_add(utime, t->utime);
1339 stime = cputime_add(stime, t->stime);
1340 t = next_thread(t);
1341 } while (t != tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342
Oleg Nesterov35f5cad2006-03-28 16:11:19 -08001343 cutime = tsk->signal->cutime;
1344 cstime = tsk->signal->cstime;
1345 spin_unlock_irq(&tsk->sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346
1347 tmp.tms_utime = cputime_to_clock_t(utime);
1348 tmp.tms_stime = cputime_to_clock_t(stime);
1349 tmp.tms_cutime = cputime_to_clock_t(cutime);
1350 tmp.tms_cstime = cputime_to_clock_t(cstime);
1351 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1352 return -EFAULT;
1353 }
1354 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1355}
1356
1357/*
1358 * This needs some heavy checking ...
1359 * I just haven't the stomach for it. I also don't fully
1360 * understand sessions/pgrp etc. Let somebody who does explain it.
1361 *
1362 * OK, I think I have the protection semantics right.... this is really
1363 * only important on a multi-user system anyway, to make sure one user
1364 * can't send a signal to a process owned by another. -TYT, 12/12/91
1365 *
1366 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1367 * LBT 04.03.94
1368 */
1369
1370asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1371{
1372 struct task_struct *p;
Oleg Nesterovee0acf92006-01-08 01:03:53 -08001373 struct task_struct *group_leader = current->group_leader;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374 int err = -EINVAL;
1375
1376 if (!pid)
Oleg Nesterovee0acf92006-01-08 01:03:53 -08001377 pid = group_leader->pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378 if (!pgid)
1379 pgid = pid;
1380 if (pgid < 0)
1381 return -EINVAL;
1382
1383 /* From this point forward we keep holding onto the tasklist lock
1384 * so that our parent does not change from under us. -DaveM
1385 */
1386 write_lock_irq(&tasklist_lock);
1387
1388 err = -ESRCH;
1389 p = find_task_by_pid(pid);
1390 if (!p)
1391 goto out;
1392
1393 err = -EINVAL;
1394 if (!thread_group_leader(p))
1395 goto out;
1396
Oleg Nesterovf7dd7952006-01-08 01:03:59 -08001397 if (p->real_parent == group_leader) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398 err = -EPERM;
Eric W. Biederman41487c62007-02-12 00:53:01 -08001399 if (task_session(p) != task_session(group_leader))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400 goto out;
1401 err = -EACCES;
1402 if (p->did_exec)
1403 goto out;
1404 } else {
1405 err = -ESRCH;
Oleg Nesterovee0acf92006-01-08 01:03:53 -08001406 if (p != group_leader)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407 goto out;
1408 }
1409
1410 err = -EPERM;
1411 if (p->signal->leader)
1412 goto out;
1413
1414 if (pgid != pid) {
Oleg Nesterovf020bc42006-12-08 02:38:02 -08001415 struct task_struct *g =
1416 find_task_by_pid_type(PIDTYPE_PGID, pgid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417
Eric W. Biederman41487c62007-02-12 00:53:01 -08001418 if (!g || task_session(g) != task_session(group_leader))
Oleg Nesterovf020bc42006-12-08 02:38:02 -08001419 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420 }
1421
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422 err = security_task_setpgid(p, pgid);
1423 if (err)
1424 goto out;
1425
1426 if (process_group(p) != pgid) {
1427 detach_pid(p, PIDTYPE_PGID);
1428 p->signal->pgrp = pgid;
1429 attach_pid(p, PIDTYPE_PGID, pgid);
1430 }
1431
1432 err = 0;
1433out:
1434 /* All paths lead to here, thus we are safe. -DaveM */
1435 write_unlock_irq(&tasklist_lock);
1436 return err;
1437}
1438
1439asmlinkage long sys_getpgid(pid_t pid)
1440{
Cal Peake756184b2006-09-30 23:27:24 -07001441 if (!pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442 return process_group(current);
Cal Peake756184b2006-09-30 23:27:24 -07001443 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444 int retval;
1445 struct task_struct *p;
1446
1447 read_lock(&tasklist_lock);
1448 p = find_task_by_pid(pid);
1449
1450 retval = -ESRCH;
1451 if (p) {
1452 retval = security_task_getpgid(p);
1453 if (!retval)
1454 retval = process_group(p);
1455 }
1456 read_unlock(&tasklist_lock);
1457 return retval;
1458 }
1459}
1460
1461#ifdef __ARCH_WANT_SYS_GETPGRP
1462
1463asmlinkage long sys_getpgrp(void)
1464{
1465 /* SMP - assuming writes are word atomic this is fine */
1466 return process_group(current);
1467}
1468
1469#endif
1470
1471asmlinkage long sys_getsid(pid_t pid)
1472{
Cal Peake756184b2006-09-30 23:27:24 -07001473 if (!pid)
Cedric Le Goater937949d2006-12-08 02:37:54 -08001474 return process_session(current);
Cal Peake756184b2006-09-30 23:27:24 -07001475 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476 int retval;
1477 struct task_struct *p;
1478
1479 read_lock(&tasklist_lock);
1480 p = find_task_by_pid(pid);
1481
1482 retval = -ESRCH;
Cal Peake756184b2006-09-30 23:27:24 -07001483 if (p) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001484 retval = security_task_getsid(p);
1485 if (!retval)
Cedric Le Goater937949d2006-12-08 02:37:54 -08001486 retval = process_session(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487 }
1488 read_unlock(&tasklist_lock);
1489 return retval;
1490 }
1491}
1492
1493asmlinkage long sys_setsid(void)
1494{
Oren Laadane19f2472006-01-08 01:03:58 -08001495 struct task_struct *group_leader = current->group_leader;
Eric W. Biederman390e2ff2006-03-31 02:31:33 -08001496 pid_t session;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497 int err = -EPERM;
1498
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499 write_lock_irq(&tasklist_lock);
1500
Eric W. Biederman390e2ff2006-03-31 02:31:33 -08001501 /* Fail if I am already a session leader */
1502 if (group_leader->signal->leader)
1503 goto out;
1504
1505 session = group_leader->pid;
1506 /* Fail if a process group id already exists that equals the
1507 * proposed session id.
1508 *
1509 * Don't check if session id == 1 because kernel threads use this
1510 * session id and so the check will always fail and make it so
1511 * init cannot successfully call setsid.
1512 */
1513 if (session > 1 && find_task_by_pid_type(PIDTYPE_PGID, session))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514 goto out;
1515
Oren Laadane19f2472006-01-08 01:03:58 -08001516 group_leader->signal->leader = 1;
Eric W. Biederman390e2ff2006-03-31 02:31:33 -08001517 __set_special_pids(session, session);
Peter Zijlstra24ec8392006-12-08 02:36:04 -08001518
1519 spin_lock(&group_leader->sighand->siglock);
Oren Laadane19f2472006-01-08 01:03:58 -08001520 group_leader->signal->tty = NULL;
Peter Zijlstra24ec8392006-12-08 02:36:04 -08001521 spin_unlock(&group_leader->sighand->siglock);
1522
Oren Laadane19f2472006-01-08 01:03:58 -08001523 err = process_group(group_leader);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524out:
1525 write_unlock_irq(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526 return err;
1527}
1528
1529/*
1530 * Supplementary group IDs
1531 */
1532
1533/* init to 2 - one for init_task, one to ensure it is never freed */
1534struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1535
1536struct group_info *groups_alloc(int gidsetsize)
1537{
1538 struct group_info *group_info;
1539 int nblocks;
1540 int i;
1541
1542 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1543 /* Make sure we always allocate at least one indirect block pointer */
1544 nblocks = nblocks ? : 1;
1545 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1546 if (!group_info)
1547 return NULL;
1548 group_info->ngroups = gidsetsize;
1549 group_info->nblocks = nblocks;
1550 atomic_set(&group_info->usage, 1);
1551
Cal Peake756184b2006-09-30 23:27:24 -07001552 if (gidsetsize <= NGROUPS_SMALL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001553 group_info->blocks[0] = group_info->small_block;
Cal Peake756184b2006-09-30 23:27:24 -07001554 else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555 for (i = 0; i < nblocks; i++) {
1556 gid_t *b;
1557 b = (void *)__get_free_page(GFP_USER);
1558 if (!b)
1559 goto out_undo_partial_alloc;
1560 group_info->blocks[i] = b;
1561 }
1562 }
1563 return group_info;
1564
1565out_undo_partial_alloc:
1566 while (--i >= 0) {
1567 free_page((unsigned long)group_info->blocks[i]);
1568 }
1569 kfree(group_info);
1570 return NULL;
1571}
1572
1573EXPORT_SYMBOL(groups_alloc);
1574
1575void groups_free(struct group_info *group_info)
1576{
1577 if (group_info->blocks[0] != group_info->small_block) {
1578 int i;
1579 for (i = 0; i < group_info->nblocks; i++)
1580 free_page((unsigned long)group_info->blocks[i]);
1581 }
1582 kfree(group_info);
1583}
1584
1585EXPORT_SYMBOL(groups_free);
1586
1587/* export the group_info to a user-space array */
1588static int groups_to_user(gid_t __user *grouplist,
1589 struct group_info *group_info)
1590{
1591 int i;
1592 int count = group_info->ngroups;
1593
1594 for (i = 0; i < group_info->nblocks; i++) {
1595 int cp_count = min(NGROUPS_PER_BLOCK, count);
1596 int off = i * NGROUPS_PER_BLOCK;
1597 int len = cp_count * sizeof(*grouplist);
1598
1599 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1600 return -EFAULT;
1601
1602 count -= cp_count;
1603 }
1604 return 0;
1605}
1606
1607/* fill a group_info from a user-space array - it must be allocated already */
1608static int groups_from_user(struct group_info *group_info,
1609 gid_t __user *grouplist)
Cal Peake756184b2006-09-30 23:27:24 -07001610{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001611 int i;
1612 int count = group_info->ngroups;
1613
1614 for (i = 0; i < group_info->nblocks; i++) {
1615 int cp_count = min(NGROUPS_PER_BLOCK, count);
1616 int off = i * NGROUPS_PER_BLOCK;
1617 int len = cp_count * sizeof(*grouplist);
1618
1619 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1620 return -EFAULT;
1621
1622 count -= cp_count;
1623 }
1624 return 0;
1625}
1626
Domen Puncerebe8b542005-05-05 16:16:19 -07001627/* a simple Shell sort */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628static void groups_sort(struct group_info *group_info)
1629{
1630 int base, max, stride;
1631 int gidsetsize = group_info->ngroups;
1632
1633 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1634 ; /* nothing */
1635 stride /= 3;
1636
1637 while (stride) {
1638 max = gidsetsize - stride;
1639 for (base = 0; base < max; base++) {
1640 int left = base;
1641 int right = left + stride;
1642 gid_t tmp = GROUP_AT(group_info, right);
1643
1644 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1645 GROUP_AT(group_info, right) =
1646 GROUP_AT(group_info, left);
1647 right = left;
1648 left -= stride;
1649 }
1650 GROUP_AT(group_info, right) = tmp;
1651 }
1652 stride /= 3;
1653 }
1654}
1655
1656/* a simple bsearch */
David Howells3e301482005-06-23 22:00:56 -07001657int groups_search(struct group_info *group_info, gid_t grp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658{
Eric Dumazetd74beb9f2006-03-25 03:08:19 -08001659 unsigned int left, right;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660
1661 if (!group_info)
1662 return 0;
1663
1664 left = 0;
1665 right = group_info->ngroups;
1666 while (left < right) {
Eric Dumazetd74beb9f2006-03-25 03:08:19 -08001667 unsigned int mid = (left+right)/2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668 int cmp = grp - GROUP_AT(group_info, mid);
1669 if (cmp > 0)
1670 left = mid + 1;
1671 else if (cmp < 0)
1672 right = mid;
1673 else
1674 return 1;
1675 }
1676 return 0;
1677}
1678
1679/* validate and set current->group_info */
1680int set_current_groups(struct group_info *group_info)
1681{
1682 int retval;
1683 struct group_info *old_info;
1684
1685 retval = security_task_setgroups(group_info);
1686 if (retval)
1687 return retval;
1688
1689 groups_sort(group_info);
1690 get_group_info(group_info);
1691
1692 task_lock(current);
1693 old_info = current->group_info;
1694 current->group_info = group_info;
1695 task_unlock(current);
1696
1697 put_group_info(old_info);
1698
1699 return 0;
1700}
1701
1702EXPORT_SYMBOL(set_current_groups);
1703
1704asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1705{
1706 int i = 0;
1707
1708 /*
1709 * SMP: Nobody else can change our grouplist. Thus we are
1710 * safe.
1711 */
1712
1713 if (gidsetsize < 0)
1714 return -EINVAL;
1715
1716 /* no need to grab task_lock here; it cannot change */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001717 i = current->group_info->ngroups;
1718 if (gidsetsize) {
1719 if (i > gidsetsize) {
1720 i = -EINVAL;
1721 goto out;
1722 }
1723 if (groups_to_user(grouplist, current->group_info)) {
1724 i = -EFAULT;
1725 goto out;
1726 }
1727 }
1728out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729 return i;
1730}
1731
1732/*
1733 * SMP: Our groups are copy-on-write. We can set them safely
1734 * without another task interfering.
1735 */
1736
1737asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1738{
1739 struct group_info *group_info;
1740 int retval;
1741
1742 if (!capable(CAP_SETGID))
1743 return -EPERM;
1744 if ((unsigned)gidsetsize > NGROUPS_MAX)
1745 return -EINVAL;
1746
1747 group_info = groups_alloc(gidsetsize);
1748 if (!group_info)
1749 return -ENOMEM;
1750 retval = groups_from_user(group_info, grouplist);
1751 if (retval) {
1752 put_group_info(group_info);
1753 return retval;
1754 }
1755
1756 retval = set_current_groups(group_info);
1757 put_group_info(group_info);
1758
1759 return retval;
1760}
1761
1762/*
1763 * Check whether we're fsgid/egid or in the supplemental group..
1764 */
1765int in_group_p(gid_t grp)
1766{
1767 int retval = 1;
Cal Peake756184b2006-09-30 23:27:24 -07001768 if (grp != current->fsgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001769 retval = groups_search(current->group_info, grp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770 return retval;
1771}
1772
1773EXPORT_SYMBOL(in_group_p);
1774
1775int in_egroup_p(gid_t grp)
1776{
1777 int retval = 1;
Cal Peake756184b2006-09-30 23:27:24 -07001778 if (grp != current->egid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779 retval = groups_search(current->group_info, grp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780 return retval;
1781}
1782
1783EXPORT_SYMBOL(in_egroup_p);
1784
1785DECLARE_RWSEM(uts_sem);
1786
David S. Miller393b0722005-11-10 12:47:50 -08001787EXPORT_SYMBOL(uts_sem);
1788
Linus Torvalds1da177e2005-04-16 15:20:36 -07001789asmlinkage long sys_newuname(struct new_utsname __user * name)
1790{
1791 int errno = 0;
1792
1793 down_read(&uts_sem);
Serge E. Hallyne9ff3992006-10-02 02:18:11 -07001794 if (copy_to_user(name, utsname(), sizeof *name))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001795 errno = -EFAULT;
1796 up_read(&uts_sem);
1797 return errno;
1798}
1799
1800asmlinkage long sys_sethostname(char __user *name, int len)
1801{
1802 int errno;
1803 char tmp[__NEW_UTS_LEN];
1804
1805 if (!capable(CAP_SYS_ADMIN))
1806 return -EPERM;
1807 if (len < 0 || len > __NEW_UTS_LEN)
1808 return -EINVAL;
1809 down_write(&uts_sem);
1810 errno = -EFAULT;
1811 if (!copy_from_user(tmp, name, len)) {
Serge E. Hallyne9ff3992006-10-02 02:18:11 -07001812 memcpy(utsname()->nodename, tmp, len);
1813 utsname()->nodename[len] = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001814 errno = 0;
1815 }
1816 up_write(&uts_sem);
1817 return errno;
1818}
1819
1820#ifdef __ARCH_WANT_SYS_GETHOSTNAME
1821
1822asmlinkage long sys_gethostname(char __user *name, int len)
1823{
1824 int i, errno;
1825
1826 if (len < 0)
1827 return -EINVAL;
1828 down_read(&uts_sem);
Serge E. Hallyne9ff3992006-10-02 02:18:11 -07001829 i = 1 + strlen(utsname()->nodename);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830 if (i > len)
1831 i = len;
1832 errno = 0;
Serge E. Hallyne9ff3992006-10-02 02:18:11 -07001833 if (copy_to_user(name, utsname()->nodename, i))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001834 errno = -EFAULT;
1835 up_read(&uts_sem);
1836 return errno;
1837}
1838
1839#endif
1840
1841/*
1842 * Only setdomainname; getdomainname can be implemented by calling
1843 * uname()
1844 */
1845asmlinkage long sys_setdomainname(char __user *name, int len)
1846{
1847 int errno;
1848 char tmp[__NEW_UTS_LEN];
1849
1850 if (!capable(CAP_SYS_ADMIN))
1851 return -EPERM;
1852 if (len < 0 || len > __NEW_UTS_LEN)
1853 return -EINVAL;
1854
1855 down_write(&uts_sem);
1856 errno = -EFAULT;
1857 if (!copy_from_user(tmp, name, len)) {
Serge E. Hallyne9ff3992006-10-02 02:18:11 -07001858 memcpy(utsname()->domainname, tmp, len);
1859 utsname()->domainname[len] = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860 errno = 0;
1861 }
1862 up_write(&uts_sem);
1863 return errno;
1864}
1865
1866asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1867{
1868 if (resource >= RLIM_NLIMITS)
1869 return -EINVAL;
1870 else {
1871 struct rlimit value;
1872 task_lock(current->group_leader);
1873 value = current->signal->rlim[resource];
1874 task_unlock(current->group_leader);
1875 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1876 }
1877}
1878
1879#ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1880
1881/*
1882 * Back compatibility for getrlimit. Needed for some apps.
1883 */
1884
1885asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1886{
1887 struct rlimit x;
1888 if (resource >= RLIM_NLIMITS)
1889 return -EINVAL;
1890
1891 task_lock(current->group_leader);
1892 x = current->signal->rlim[resource];
1893 task_unlock(current->group_leader);
Cal Peake756184b2006-09-30 23:27:24 -07001894 if (x.rlim_cur > 0x7FFFFFFF)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001895 x.rlim_cur = 0x7FFFFFFF;
Cal Peake756184b2006-09-30 23:27:24 -07001896 if (x.rlim_max > 0x7FFFFFFF)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001897 x.rlim_max = 0x7FFFFFFF;
1898 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1899}
1900
1901#endif
1902
1903asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1904{
1905 struct rlimit new_rlim, *old_rlim;
Andrew Mortonec9e16b2006-03-24 03:18:34 -08001906 unsigned long it_prof_secs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001907 int retval;
1908
1909 if (resource >= RLIM_NLIMITS)
1910 return -EINVAL;
Andrew Mortonec9e16b2006-03-24 03:18:34 -08001911 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001912 return -EFAULT;
Andrew Mortonec9e16b2006-03-24 03:18:34 -08001913 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1914 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001915 old_rlim = current->signal->rlim + resource;
1916 if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1917 !capable(CAP_SYS_RESOURCE))
1918 return -EPERM;
1919 if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
Andrew Mortonec9e16b2006-03-24 03:18:34 -08001920 return -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001921
1922 retval = security_task_setrlimit(resource, &new_rlim);
1923 if (retval)
1924 return retval;
1925
Tom Alsberg9926e4c2007-05-08 00:30:31 -07001926 if (resource == RLIMIT_CPU && new_rlim.rlim_cur == 0) {
1927 /*
1928 * The caller is asking for an immediate RLIMIT_CPU
1929 * expiry. But we use the zero value to mean "it was
1930 * never set". So let's cheat and make it one second
1931 * instead
1932 */
1933 new_rlim.rlim_cur = 1;
1934 }
1935
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936 task_lock(current->group_leader);
1937 *old_rlim = new_rlim;
1938 task_unlock(current->group_leader);
1939
Andrew Mortonec9e16b2006-03-24 03:18:34 -08001940 if (resource != RLIMIT_CPU)
1941 goto out;
Andrew Mortond3561f72006-03-24 03:18:36 -08001942
1943 /*
1944 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1945 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1946 * very long-standing error, and fixing it now risks breakage of
1947 * applications, so we live with it
1948 */
Andrew Mortonec9e16b2006-03-24 03:18:34 -08001949 if (new_rlim.rlim_cur == RLIM_INFINITY)
1950 goto out;
1951
1952 it_prof_secs = cputime_to_secs(current->signal->it_prof_expires);
1953 if (it_prof_secs == 0 || new_rlim.rlim_cur <= it_prof_secs) {
Andrew Mortone0661112006-03-24 03:18:35 -08001954 unsigned long rlim_cur = new_rlim.rlim_cur;
1955 cputime_t cputime;
Andrew Mortonec9e16b2006-03-24 03:18:34 -08001956
Andrew Mortone0661112006-03-24 03:18:35 -08001957 cputime = secs_to_cputime(rlim_cur);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958 read_lock(&tasklist_lock);
1959 spin_lock_irq(&current->sighand->siglock);
Andrew Mortonec9e16b2006-03-24 03:18:34 -08001960 set_process_cpu_timer(current, CPUCLOCK_PROF, &cputime, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001961 spin_unlock_irq(&current->sighand->siglock);
1962 read_unlock(&tasklist_lock);
1963 }
Andrew Mortonec9e16b2006-03-24 03:18:34 -08001964out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001965 return 0;
1966}
1967
1968/*
1969 * It would make sense to put struct rusage in the task_struct,
1970 * except that would make the task_struct be *really big*. After
1971 * task_struct gets moved into malloc'ed memory, it would
1972 * make sense to do this. It will make moving the rest of the information
1973 * a lot simpler! (Which we're not doing right now because we're not
1974 * measuring them yet).
1975 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001976 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1977 * races with threads incrementing their own counters. But since word
1978 * reads are atomic, we either get new values or old values and we don't
1979 * care which for the sums. We always take the siglock to protect reading
1980 * the c* fields from p->signal from races with exit.c updating those
1981 * fields when reaping, so a sample either gets all the additions of a
1982 * given child after it's reaped, or none so this sample is before reaping.
Ravikiran G Thirumalai2dd0ebc2006-03-23 03:00:13 -08001983 *
Ravikiran G Thirumalaide047c12006-06-22 14:47:26 -07001984 * Locking:
1985 * We need to take the siglock for CHILDEREN, SELF and BOTH
1986 * for the cases current multithreaded, non-current single threaded
1987 * non-current multithreaded. Thread traversal is now safe with
1988 * the siglock held.
1989 * Strictly speaking, we donot need to take the siglock if we are current and
1990 * single threaded, as no one else can take our signal_struct away, no one
1991 * else can reap the children to update signal->c* counters, and no one else
1992 * can race with the signal-> fields. If we do not take any lock, the
1993 * signal-> fields could be read out of order while another thread was just
1994 * exiting. So we should place a read memory barrier when we avoid the lock.
1995 * On the writer side, write memory barrier is implied in __exit_signal
1996 * as __exit_signal releases the siglock spinlock after updating the signal->
1997 * fields. But we don't do this yet to keep things simple.
Ravikiran G Thirumalai2dd0ebc2006-03-23 03:00:13 -08001998 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001999 */
2000
2001static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
2002{
2003 struct task_struct *t;
2004 unsigned long flags;
2005 cputime_t utime, stime;
2006
2007 memset((char *) r, 0, sizeof *r);
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08002008 utime = stime = cputime_zero;
2009
Ravikiran G Thirumalaide047c12006-06-22 14:47:26 -07002010 rcu_read_lock();
2011 if (!lock_task_sighand(p, &flags)) {
2012 rcu_read_unlock();
2013 return;
2014 }
Ravikiran G Thirumalai2dd0ebc2006-03-23 03:00:13 -08002015
Linus Torvalds1da177e2005-04-16 15:20:36 -07002016 switch (who) {
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08002017 case RUSAGE_BOTH:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002018 case RUSAGE_CHILDREN:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002019 utime = p->signal->cutime;
2020 stime = p->signal->cstime;
2021 r->ru_nvcsw = p->signal->cnvcsw;
2022 r->ru_nivcsw = p->signal->cnivcsw;
2023 r->ru_minflt = p->signal->cmin_flt;
2024 r->ru_majflt = p->signal->cmaj_flt;
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08002025
2026 if (who == RUSAGE_CHILDREN)
2027 break;
2028
Linus Torvalds1da177e2005-04-16 15:20:36 -07002029 case RUSAGE_SELF:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002030 utime = cputime_add(utime, p->signal->utime);
2031 stime = cputime_add(stime, p->signal->stime);
2032 r->ru_nvcsw += p->signal->nvcsw;
2033 r->ru_nivcsw += p->signal->nivcsw;
2034 r->ru_minflt += p->signal->min_flt;
2035 r->ru_majflt += p->signal->maj_flt;
2036 t = p;
2037 do {
2038 utime = cputime_add(utime, t->utime);
2039 stime = cputime_add(stime, t->stime);
2040 r->ru_nvcsw += t->nvcsw;
2041 r->ru_nivcsw += t->nivcsw;
2042 r->ru_minflt += t->min_flt;
2043 r->ru_majflt += t->maj_flt;
2044 t = next_thread(t);
2045 } while (t != p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002046 break;
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08002047
Linus Torvalds1da177e2005-04-16 15:20:36 -07002048 default:
2049 BUG();
2050 }
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08002051
Ravikiran G Thirumalaide047c12006-06-22 14:47:26 -07002052 unlock_task_sighand(p, &flags);
2053 rcu_read_unlock();
2054
Oleg Nesterov0f59cc42006-01-08 01:05:15 -08002055 cputime_to_timeval(utime, &r->ru_utime);
2056 cputime_to_timeval(stime, &r->ru_stime);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002057}
2058
2059int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
2060{
2061 struct rusage r;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002062 k_getrusage(p, who, &r);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002063 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
2064}
2065
2066asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
2067{
2068 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
2069 return -EINVAL;
2070 return getrusage(current, who, ru);
2071}
2072
2073asmlinkage long sys_umask(int mask)
2074{
2075 mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
2076 return mask;
2077}
2078
2079asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
2080 unsigned long arg4, unsigned long arg5)
2081{
2082 long error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002083
2084 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2085 if (error)
2086 return error;
2087
2088 switch (option) {
2089 case PR_SET_PDEATHSIG:
Jesper Juhl0730ded2005-09-06 15:17:37 -07002090 if (!valid_signal(arg2)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002091 error = -EINVAL;
2092 break;
2093 }
Jesper Juhl0730ded2005-09-06 15:17:37 -07002094 current->pdeath_signal = arg2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095 break;
2096 case PR_GET_PDEATHSIG:
2097 error = put_user(current->pdeath_signal, (int __user *)arg2);
2098 break;
2099 case PR_GET_DUMPABLE:
Michael Kerrisk2030c0f2005-09-16 19:28:02 -07002100 error = current->mm->dumpable;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101 break;
2102 case PR_SET_DUMPABLE:
Marcel Holtmannabf75a52006-07-12 13:12:00 +02002103 if (arg2 < 0 || arg2 > 1) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002104 error = -EINVAL;
2105 break;
2106 }
2107 current->mm->dumpable = arg2;
2108 break;
2109
2110 case PR_SET_UNALIGN:
2111 error = SET_UNALIGN_CTL(current, arg2);
2112 break;
2113 case PR_GET_UNALIGN:
2114 error = GET_UNALIGN_CTL(current, arg2);
2115 break;
2116 case PR_SET_FPEMU:
2117 error = SET_FPEMU_CTL(current, arg2);
2118 break;
2119 case PR_GET_FPEMU:
2120 error = GET_FPEMU_CTL(current, arg2);
2121 break;
2122 case PR_SET_FPEXC:
2123 error = SET_FPEXC_CTL(current, arg2);
2124 break;
2125 case PR_GET_FPEXC:
2126 error = GET_FPEXC_CTL(current, arg2);
2127 break;
2128 case PR_GET_TIMING:
2129 error = PR_TIMING_STATISTICAL;
2130 break;
2131 case PR_SET_TIMING:
2132 if (arg2 == PR_TIMING_STATISTICAL)
2133 error = 0;
2134 else
2135 error = -EINVAL;
2136 break;
2137
2138 case PR_GET_KEEPCAPS:
2139 if (current->keep_capabilities)
2140 error = 1;
2141 break;
2142 case PR_SET_KEEPCAPS:
2143 if (arg2 != 0 && arg2 != 1) {
2144 error = -EINVAL;
2145 break;
2146 }
2147 current->keep_capabilities = arg2;
2148 break;
2149 case PR_SET_NAME: {
2150 struct task_struct *me = current;
2151 unsigned char ncomm[sizeof(me->comm)];
2152
2153 ncomm[sizeof(me->comm)-1] = 0;
2154 if (strncpy_from_user(ncomm, (char __user *)arg2,
2155 sizeof(me->comm)-1) < 0)
2156 return -EFAULT;
2157 set_task_comm(me, ncomm);
2158 return 0;
2159 }
2160 case PR_GET_NAME: {
2161 struct task_struct *me = current;
2162 unsigned char tcomm[sizeof(me->comm)];
2163
2164 get_task_comm(tcomm, me);
2165 if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))
2166 return -EFAULT;
2167 return 0;
2168 }
Anton Blanchard651d7652006-06-07 16:10:19 +10002169 case PR_GET_ENDIAN:
2170 error = GET_ENDIAN(current, arg2);
2171 break;
2172 case PR_SET_ENDIAN:
2173 error = SET_ENDIAN(current, arg2);
2174 break;
2175
Linus Torvalds1da177e2005-04-16 15:20:36 -07002176 default:
2177 error = -EINVAL;
2178 break;
2179 }
2180 return error;
2181}
Andi Kleen3cfc3482006-09-26 10:52:28 +02002182
2183asmlinkage long sys_getcpu(unsigned __user *cpup, unsigned __user *nodep,
2184 struct getcpu_cache __user *cache)
2185{
2186 int err = 0;
2187 int cpu = raw_smp_processor_id();
2188 if (cpup)
2189 err |= put_user(cpu, cpup);
2190 if (nodep)
2191 err |= put_user(cpu_to_node(cpu), nodep);
2192 if (cache) {
2193 /*
2194 * The cache is not needed for this implementation,
2195 * but make sure user programs pass something
2196 * valid. vsyscall implementations can instead make
2197 * good use of the cache. Only use t0 and t1 because
2198 * these are available in both 32bit and 64bit ABI (no
2199 * need for a compat_getcpu). 32bit has enough
2200 * padding
2201 */
2202 unsigned long t0, t1;
Andi Kleen34596dc2006-09-30 01:47:55 +02002203 get_user(t0, &cache->blob[0]);
2204 get_user(t1, &cache->blob[1]);
Andi Kleen3cfc3482006-09-26 10:52:28 +02002205 t0++;
2206 t1++;
Andi Kleen34596dc2006-09-30 01:47:55 +02002207 put_user(t0, &cache->blob[0]);
2208 put_user(t1, &cache->blob[1]);
Andi Kleen3cfc3482006-09-26 10:52:28 +02002209 }
2210 return err ? -EFAULT : 0;
2211}