blob: 5fc10d3e3891a41b65e65e4a1b636db77313cd32 [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
7#include <linux/config.h>
8#include <linux/module.h>
9#include <linux/mm.h>
10#include <linux/utsname.h>
11#include <linux/mman.h>
12#include <linux/smp_lock.h>
13#include <linux/notifier.h>
14#include <linux/reboot.h>
15#include <linux/prctl.h>
16#include <linux/init.h>
17#include <linux/highuid.h>
18#include <linux/fs.h>
Eric W. Biedermandc009d92005-06-25 14:57:52 -070019#include <linux/kernel.h>
20#include <linux/kexec.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021#include <linux/workqueue.h>
22#include <linux/device.h>
23#include <linux/key.h>
24#include <linux/times.h>
25#include <linux/posix-timers.h>
26#include <linux/security.h>
27#include <linux/dcookies.h>
28#include <linux/suspend.h>
29#include <linux/tty.h>
Jesper Juhl7ed20e12005-05-01 08:59:14 -070030#include <linux/signal.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031
32#include <linux/compat.h>
33#include <linux/syscalls.h>
34
35#include <asm/uaccess.h>
36#include <asm/io.h>
37#include <asm/unistd.h>
38
39#ifndef SET_UNALIGN_CTL
40# define SET_UNALIGN_CTL(a,b) (-EINVAL)
41#endif
42#ifndef GET_UNALIGN_CTL
43# define GET_UNALIGN_CTL(a,b) (-EINVAL)
44#endif
45#ifndef SET_FPEMU_CTL
46# define SET_FPEMU_CTL(a,b) (-EINVAL)
47#endif
48#ifndef GET_FPEMU_CTL
49# define GET_FPEMU_CTL(a,b) (-EINVAL)
50#endif
51#ifndef SET_FPEXC_CTL
52# define SET_FPEXC_CTL(a,b) (-EINVAL)
53#endif
54#ifndef GET_FPEXC_CTL
55# define GET_FPEXC_CTL(a,b) (-EINVAL)
56#endif
57
58/*
59 * this is where the system-wide overflow UID and GID are defined, for
60 * architectures that now have 32-bit UID/GID but didn't in the past
61 */
62
63int overflowuid = DEFAULT_OVERFLOWUID;
64int overflowgid = DEFAULT_OVERFLOWGID;
65
66#ifdef CONFIG_UID16
67EXPORT_SYMBOL(overflowuid);
68EXPORT_SYMBOL(overflowgid);
69#endif
70
71/*
72 * the same as above, but for filesystems which can only store a 16-bit
73 * UID and GID. as such, this is needed on all architectures
74 */
75
76int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
77int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
78
79EXPORT_SYMBOL(fs_overflowuid);
80EXPORT_SYMBOL(fs_overflowgid);
81
82/*
83 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
84 */
85
86int C_A_D = 1;
87int cad_pid = 1;
88
89/*
90 * Notifier list for kernel code which wants to be called
91 * at shutdown. This is used to stop any idling DMA operations
92 * and the like.
93 */
94
95static struct notifier_block *reboot_notifier_list;
96static DEFINE_RWLOCK(notifier_lock);
97
98/**
99 * notifier_chain_register - Add notifier to a notifier chain
100 * @list: Pointer to root list pointer
101 * @n: New entry in notifier chain
102 *
103 * Adds a notifier to a notifier chain.
104 *
105 * Currently always returns zero.
106 */
107
108int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
109{
110 write_lock(&notifier_lock);
111 while(*list)
112 {
113 if(n->priority > (*list)->priority)
114 break;
115 list= &((*list)->next);
116 }
117 n->next = *list;
118 *list=n;
119 write_unlock(&notifier_lock);
120 return 0;
121}
122
123EXPORT_SYMBOL(notifier_chain_register);
124
125/**
126 * notifier_chain_unregister - Remove notifier from a notifier chain
127 * @nl: Pointer to root list pointer
128 * @n: New entry in notifier chain
129 *
130 * Removes a notifier from a notifier chain.
131 *
132 * Returns zero on success, or %-ENOENT on failure.
133 */
134
135int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
136{
137 write_lock(&notifier_lock);
138 while((*nl)!=NULL)
139 {
140 if((*nl)==n)
141 {
142 *nl=n->next;
143 write_unlock(&notifier_lock);
144 return 0;
145 }
146 nl=&((*nl)->next);
147 }
148 write_unlock(&notifier_lock);
149 return -ENOENT;
150}
151
152EXPORT_SYMBOL(notifier_chain_unregister);
153
154/**
155 * notifier_call_chain - Call functions in a notifier chain
156 * @n: Pointer to root pointer of notifier chain
157 * @val: Value passed unmodified to notifier function
158 * @v: Pointer passed unmodified to notifier function
159 *
160 * Calls each function in a notifier chain in turn.
161 *
162 * If the return value of the notifier can be and'd
163 * with %NOTIFY_STOP_MASK, then notifier_call_chain
164 * will return immediately, with the return value of
165 * the notifier function which halted execution.
166 * Otherwise, the return value is the return value
167 * of the last notifier function called.
168 */
169
170int notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
171{
172 int ret=NOTIFY_DONE;
173 struct notifier_block *nb = *n;
174
175 while(nb)
176 {
177 ret=nb->notifier_call(nb,val,v);
178 if(ret&NOTIFY_STOP_MASK)
179 {
180 return ret;
181 }
182 nb=nb->next;
183 }
184 return ret;
185}
186
187EXPORT_SYMBOL(notifier_call_chain);
188
189/**
190 * register_reboot_notifier - Register function to be called at reboot time
191 * @nb: Info about notifier function to be called
192 *
193 * Registers a function with the list of functions
194 * to be called at reboot time.
195 *
196 * Currently always returns zero, as notifier_chain_register
197 * always returns zero.
198 */
199
200int register_reboot_notifier(struct notifier_block * nb)
201{
202 return notifier_chain_register(&reboot_notifier_list, nb);
203}
204
205EXPORT_SYMBOL(register_reboot_notifier);
206
207/**
208 * unregister_reboot_notifier - Unregister previously registered reboot notifier
209 * @nb: Hook to be unregistered
210 *
211 * Unregisters a previously registered reboot
212 * notifier function.
213 *
214 * Returns zero on success, or %-ENOENT on failure.
215 */
216
217int unregister_reboot_notifier(struct notifier_block * nb)
218{
219 return notifier_chain_unregister(&reboot_notifier_list, nb);
220}
221
222EXPORT_SYMBOL(unregister_reboot_notifier);
223
224static int set_one_prio(struct task_struct *p, int niceval, int error)
225{
226 int no_nice;
227
228 if (p->uid != current->euid &&
229 p->euid != current->euid && !capable(CAP_SYS_NICE)) {
230 error = -EPERM;
231 goto out;
232 }
Matt Mackalle43379f2005-05-01 08:59:00 -0700233 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234 error = -EACCES;
235 goto out;
236 }
237 no_nice = security_task_setnice(p, niceval);
238 if (no_nice) {
239 error = no_nice;
240 goto out;
241 }
242 if (error == -ESRCH)
243 error = 0;
244 set_user_nice(p, niceval);
245out:
246 return error;
247}
248
249asmlinkage long sys_setpriority(int which, int who, int niceval)
250{
251 struct task_struct *g, *p;
252 struct user_struct *user;
253 int error = -EINVAL;
254
255 if (which > 2 || which < 0)
256 goto out;
257
258 /* normalize: avoid signed division (rounding problems) */
259 error = -ESRCH;
260 if (niceval < -20)
261 niceval = -20;
262 if (niceval > 19)
263 niceval = 19;
264
265 read_lock(&tasklist_lock);
266 switch (which) {
267 case PRIO_PROCESS:
268 if (!who)
269 who = current->pid;
270 p = find_task_by_pid(who);
271 if (p)
272 error = set_one_prio(p, niceval, error);
273 break;
274 case PRIO_PGRP:
275 if (!who)
276 who = process_group(current);
277 do_each_task_pid(who, PIDTYPE_PGID, p) {
278 error = set_one_prio(p, niceval, error);
279 } while_each_task_pid(who, PIDTYPE_PGID, p);
280 break;
281 case PRIO_USER:
282 user = current->user;
283 if (!who)
284 who = current->uid;
285 else
286 if ((who != current->uid) && !(user = find_user(who)))
287 goto out_unlock; /* No processes for this user */
288
289 do_each_thread(g, p)
290 if (p->uid == who)
291 error = set_one_prio(p, niceval, error);
292 while_each_thread(g, p);
293 if (who != current->uid)
294 free_uid(user); /* For find_user() */
295 break;
296 }
297out_unlock:
298 read_unlock(&tasklist_lock);
299out:
300 return error;
301}
302
303/*
304 * Ugh. To avoid negative return values, "getpriority()" will
305 * not return the normal nice-value, but a negated value that
306 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
307 * to stay compatible.
308 */
309asmlinkage long sys_getpriority(int which, int who)
310{
311 struct task_struct *g, *p;
312 struct user_struct *user;
313 long niceval, retval = -ESRCH;
314
315 if (which > 2 || which < 0)
316 return -EINVAL;
317
318 read_lock(&tasklist_lock);
319 switch (which) {
320 case PRIO_PROCESS:
321 if (!who)
322 who = current->pid;
323 p = find_task_by_pid(who);
324 if (p) {
325 niceval = 20 - task_nice(p);
326 if (niceval > retval)
327 retval = niceval;
328 }
329 break;
330 case PRIO_PGRP:
331 if (!who)
332 who = process_group(current);
333 do_each_task_pid(who, PIDTYPE_PGID, p) {
334 niceval = 20 - task_nice(p);
335 if (niceval > retval)
336 retval = niceval;
337 } while_each_task_pid(who, PIDTYPE_PGID, p);
338 break;
339 case PRIO_USER:
340 user = current->user;
341 if (!who)
342 who = current->uid;
343 else
344 if ((who != current->uid) && !(user = find_user(who)))
345 goto out_unlock; /* No processes for this user */
346
347 do_each_thread(g, p)
348 if (p->uid == who) {
349 niceval = 20 - task_nice(p);
350 if (niceval > retval)
351 retval = niceval;
352 }
353 while_each_thread(g, p);
354 if (who != current->uid)
355 free_uid(user); /* for find_user() */
356 break;
357 }
358out_unlock:
359 read_unlock(&tasklist_lock);
360
361 return retval;
362}
363
364
365/*
366 * Reboot system call: for obvious reasons only root may call it,
367 * and even root needs to set up some magic numbers in the registers
368 * so that some mistake won't make this reboot the whole machine.
369 * You can also set the meaning of the ctrl-alt-del-key here.
370 *
371 * reboot doesn't sync: do that yourself before calling this.
372 */
373asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
374{
375 char buffer[256];
376
377 /* We only trust the superuser with rebooting the system. */
378 if (!capable(CAP_SYS_BOOT))
379 return -EPERM;
380
381 /* For safety, we require "magic" arguments. */
382 if (magic1 != LINUX_REBOOT_MAGIC1 ||
383 (magic2 != LINUX_REBOOT_MAGIC2 &&
384 magic2 != LINUX_REBOOT_MAGIC2A &&
385 magic2 != LINUX_REBOOT_MAGIC2B &&
386 magic2 != LINUX_REBOOT_MAGIC2C))
387 return -EINVAL;
388
389 lock_kernel();
390 switch (cmd) {
391 case LINUX_REBOOT_CMD_RESTART:
392 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
393 system_state = SYSTEM_RESTART;
Eric W. Biederman47f61f32005-07-26 11:21:38 -0600394 device_suspend(PMSG_FREEZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 device_shutdown();
396 printk(KERN_EMERG "Restarting system.\n");
397 machine_restart(NULL);
398 break;
399
400 case LINUX_REBOOT_CMD_CAD_ON:
401 C_A_D = 1;
402 break;
403
404 case LINUX_REBOOT_CMD_CAD_OFF:
405 C_A_D = 0;
406 break;
407
408 case LINUX_REBOOT_CMD_HALT:
409 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
410 system_state = SYSTEM_HALT;
Pavel Machek620b0322005-06-25 14:55:11 -0700411 device_suspend(PMSG_SUSPEND);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 device_shutdown();
413 printk(KERN_EMERG "System halted.\n");
414 machine_halt();
415 unlock_kernel();
416 do_exit(0);
417 break;
418
419 case LINUX_REBOOT_CMD_POWER_OFF:
420 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
421 system_state = SYSTEM_POWER_OFF;
Pavel Machek620b0322005-06-25 14:55:11 -0700422 device_suspend(PMSG_SUSPEND);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 device_shutdown();
424 printk(KERN_EMERG "Power down.\n");
425 machine_power_off();
426 unlock_kernel();
427 do_exit(0);
428 break;
429
430 case LINUX_REBOOT_CMD_RESTART2:
431 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
432 unlock_kernel();
433 return -EFAULT;
434 }
435 buffer[sizeof(buffer) - 1] = '\0';
436
437 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
438 system_state = SYSTEM_RESTART;
Pavel Machek620b0322005-06-25 14:55:11 -0700439 device_suspend(PMSG_FREEZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440 device_shutdown();
441 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
442 machine_restart(buffer);
443 break;
444
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700445#ifdef CONFIG_KEXEC
446 case LINUX_REBOOT_CMD_KEXEC:
447 {
448 struct kimage *image;
449 image = xchg(&kexec_image, 0);
450 if (!image) {
451 unlock_kernel();
452 return -EINVAL;
453 }
454 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
455 system_state = SYSTEM_RESTART;
Eric W. Biederman47f61f32005-07-26 11:21:38 -0600456 device_suspend(PMSG_FREEZE);
Eric W. Biedermandc009d92005-06-25 14:57:52 -0700457 device_shutdown();
458 printk(KERN_EMERG "Starting new kernel\n");
459 machine_shutdown();
460 machine_kexec(image);
461 break;
462 }
463#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464#ifdef CONFIG_SOFTWARE_SUSPEND
465 case LINUX_REBOOT_CMD_SW_SUSPEND:
466 {
467 int ret = software_suspend();
468 unlock_kernel();
469 return ret;
470 }
471#endif
472
473 default:
474 unlock_kernel();
475 return -EINVAL;
476 }
477 unlock_kernel();
478 return 0;
479}
480
481static void deferred_cad(void *dummy)
482{
483 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
484 machine_restart(NULL);
485}
486
487/*
488 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
489 * As it's called within an interrupt, it may NOT sync: the only choice
490 * is whether to reboot at once, or just ignore the ctrl-alt-del.
491 */
492void ctrl_alt_del(void)
493{
494 static DECLARE_WORK(cad_work, deferred_cad, NULL);
495
496 if (C_A_D)
497 schedule_work(&cad_work);
498 else
499 kill_proc(cad_pid, SIGINT, 1);
500}
501
502
503/*
504 * Unprivileged users may change the real gid to the effective gid
505 * or vice versa. (BSD-style)
506 *
507 * If you set the real gid at all, or set the effective gid to a value not
508 * equal to the real gid, then the saved gid is set to the new effective gid.
509 *
510 * This makes it possible for a setgid program to completely drop its
511 * privileges, which is often a useful assertion to make when you are doing
512 * a security audit over a program.
513 *
514 * The general idea is that a program which uses just setregid() will be
515 * 100% compatible with BSD. A program which uses just setgid() will be
516 * 100% compatible with POSIX with saved IDs.
517 *
518 * SMP: There are not races, the GIDs are checked only by filesystem
519 * operations (as far as semantic preservation is concerned).
520 */
521asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
522{
523 int old_rgid = current->gid;
524 int old_egid = current->egid;
525 int new_rgid = old_rgid;
526 int new_egid = old_egid;
527 int retval;
528
529 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
530 if (retval)
531 return retval;
532
533 if (rgid != (gid_t) -1) {
534 if ((old_rgid == rgid) ||
535 (current->egid==rgid) ||
536 capable(CAP_SETGID))
537 new_rgid = rgid;
538 else
539 return -EPERM;
540 }
541 if (egid != (gid_t) -1) {
542 if ((old_rgid == egid) ||
543 (current->egid == egid) ||
544 (current->sgid == egid) ||
545 capable(CAP_SETGID))
546 new_egid = egid;
547 else {
548 return -EPERM;
549 }
550 }
551 if (new_egid != old_egid)
552 {
Alan Coxd6e71142005-06-23 00:09:43 -0700553 current->mm->dumpable = suid_dumpable;
akpm@osdl.orgd59dd462005-05-01 08:58:47 -0700554 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 }
556 if (rgid != (gid_t) -1 ||
557 (egid != (gid_t) -1 && egid != old_rgid))
558 current->sgid = new_egid;
559 current->fsgid = new_egid;
560 current->egid = new_egid;
561 current->gid = new_rgid;
562 key_fsgid_changed(current);
563 return 0;
564}
565
566/*
567 * setgid() is implemented like SysV w/ SAVED_IDS
568 *
569 * SMP: Same implicit races as above.
570 */
571asmlinkage long sys_setgid(gid_t gid)
572{
573 int old_egid = current->egid;
574 int retval;
575
576 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
577 if (retval)
578 return retval;
579
580 if (capable(CAP_SETGID))
581 {
582 if(old_egid != gid)
583 {
Alan Coxd6e71142005-06-23 00:09:43 -0700584 current->mm->dumpable = suid_dumpable;
akpm@osdl.orgd59dd462005-05-01 08:58:47 -0700585 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586 }
587 current->gid = current->egid = current->sgid = current->fsgid = gid;
588 }
589 else if ((gid == current->gid) || (gid == current->sgid))
590 {
591 if(old_egid != gid)
592 {
Alan Coxd6e71142005-06-23 00:09:43 -0700593 current->mm->dumpable = suid_dumpable;
akpm@osdl.orgd59dd462005-05-01 08:58:47 -0700594 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 }
596 current->egid = current->fsgid = gid;
597 }
598 else
599 return -EPERM;
600
601 key_fsgid_changed(current);
602 return 0;
603}
604
605static int set_user(uid_t new_ruid, int dumpclear)
606{
607 struct user_struct *new_user;
608
609 new_user = alloc_uid(new_ruid);
610 if (!new_user)
611 return -EAGAIN;
612
613 if (atomic_read(&new_user->processes) >=
614 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
615 new_user != &root_user) {
616 free_uid(new_user);
617 return -EAGAIN;
618 }
619
620 switch_uid(new_user);
621
622 if(dumpclear)
623 {
Alan Coxd6e71142005-06-23 00:09:43 -0700624 current->mm->dumpable = suid_dumpable;
akpm@osdl.orgd59dd462005-05-01 08:58:47 -0700625 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 }
627 current->uid = new_ruid;
628 return 0;
629}
630
631/*
632 * Unprivileged users may change the real uid to the effective uid
633 * or vice versa. (BSD-style)
634 *
635 * If you set the real uid at all, or set the effective uid to a value not
636 * equal to the real uid, then the saved uid is set to the new effective uid.
637 *
638 * This makes it possible for a setuid program to completely drop its
639 * privileges, which is often a useful assertion to make when you are doing
640 * a security audit over a program.
641 *
642 * The general idea is that a program which uses just setreuid() will be
643 * 100% compatible with BSD. A program which uses just setuid() will be
644 * 100% compatible with POSIX with saved IDs.
645 */
646asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
647{
648 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
649 int retval;
650
651 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
652 if (retval)
653 return retval;
654
655 new_ruid = old_ruid = current->uid;
656 new_euid = old_euid = current->euid;
657 old_suid = current->suid;
658
659 if (ruid != (uid_t) -1) {
660 new_ruid = ruid;
661 if ((old_ruid != ruid) &&
662 (current->euid != ruid) &&
663 !capable(CAP_SETUID))
664 return -EPERM;
665 }
666
667 if (euid != (uid_t) -1) {
668 new_euid = euid;
669 if ((old_ruid != euid) &&
670 (current->euid != euid) &&
671 (current->suid != euid) &&
672 !capable(CAP_SETUID))
673 return -EPERM;
674 }
675
676 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
677 return -EAGAIN;
678
679 if (new_euid != old_euid)
680 {
Alan Coxd6e71142005-06-23 00:09:43 -0700681 current->mm->dumpable = suid_dumpable;
akpm@osdl.orgd59dd462005-05-01 08:58:47 -0700682 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683 }
684 current->fsuid = current->euid = new_euid;
685 if (ruid != (uid_t) -1 ||
686 (euid != (uid_t) -1 && euid != old_ruid))
687 current->suid = current->euid;
688 current->fsuid = current->euid;
689
690 key_fsuid_changed(current);
691
692 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
693}
694
695
696
697/*
698 * setuid() is implemented like SysV with SAVED_IDS
699 *
700 * Note that SAVED_ID's is deficient in that a setuid root program
701 * like sendmail, for example, cannot set its uid to be a normal
702 * user and then switch back, because if you're root, setuid() sets
703 * the saved uid too. If you don't like this, blame the bright people
704 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
705 * will allow a root program to temporarily drop privileges and be able to
706 * regain them by swapping the real and effective uid.
707 */
708asmlinkage long sys_setuid(uid_t uid)
709{
710 int old_euid = current->euid;
711 int old_ruid, old_suid, new_ruid, new_suid;
712 int retval;
713
714 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
715 if (retval)
716 return retval;
717
718 old_ruid = new_ruid = current->uid;
719 old_suid = current->suid;
720 new_suid = old_suid;
721
722 if (capable(CAP_SETUID)) {
723 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
724 return -EAGAIN;
725 new_suid = uid;
726 } else if ((uid != current->uid) && (uid != new_suid))
727 return -EPERM;
728
729 if (old_euid != uid)
730 {
Alan Coxd6e71142005-06-23 00:09:43 -0700731 current->mm->dumpable = suid_dumpable;
akpm@osdl.orgd59dd462005-05-01 08:58:47 -0700732 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 }
734 current->fsuid = current->euid = uid;
735 current->suid = new_suid;
736
737 key_fsuid_changed(current);
738
739 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
740}
741
742
743/*
744 * This function implements a generic ability to update ruid, euid,
745 * and suid. This allows you to implement the 4.4 compatible seteuid().
746 */
747asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
748{
749 int old_ruid = current->uid;
750 int old_euid = current->euid;
751 int old_suid = current->suid;
752 int retval;
753
754 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
755 if (retval)
756 return retval;
757
758 if (!capable(CAP_SETUID)) {
759 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
760 (ruid != current->euid) && (ruid != current->suid))
761 return -EPERM;
762 if ((euid != (uid_t) -1) && (euid != current->uid) &&
763 (euid != current->euid) && (euid != current->suid))
764 return -EPERM;
765 if ((suid != (uid_t) -1) && (suid != current->uid) &&
766 (suid != current->euid) && (suid != current->suid))
767 return -EPERM;
768 }
769 if (ruid != (uid_t) -1) {
770 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
771 return -EAGAIN;
772 }
773 if (euid != (uid_t) -1) {
774 if (euid != current->euid)
775 {
Alan Coxd6e71142005-06-23 00:09:43 -0700776 current->mm->dumpable = suid_dumpable;
akpm@osdl.orgd59dd462005-05-01 08:58:47 -0700777 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778 }
779 current->euid = euid;
780 }
781 current->fsuid = current->euid;
782 if (suid != (uid_t) -1)
783 current->suid = suid;
784
785 key_fsuid_changed(current);
786
787 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
788}
789
790asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
791{
792 int retval;
793
794 if (!(retval = put_user(current->uid, ruid)) &&
795 !(retval = put_user(current->euid, euid)))
796 retval = put_user(current->suid, suid);
797
798 return retval;
799}
800
801/*
802 * Same as above, but for rgid, egid, sgid.
803 */
804asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
805{
806 int retval;
807
808 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
809 if (retval)
810 return retval;
811
812 if (!capable(CAP_SETGID)) {
813 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
814 (rgid != current->egid) && (rgid != current->sgid))
815 return -EPERM;
816 if ((egid != (gid_t) -1) && (egid != current->gid) &&
817 (egid != current->egid) && (egid != current->sgid))
818 return -EPERM;
819 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
820 (sgid != current->egid) && (sgid != current->sgid))
821 return -EPERM;
822 }
823 if (egid != (gid_t) -1) {
824 if (egid != current->egid)
825 {
Alan Coxd6e71142005-06-23 00:09:43 -0700826 current->mm->dumpable = suid_dumpable;
akpm@osdl.orgd59dd462005-05-01 08:58:47 -0700827 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828 }
829 current->egid = egid;
830 }
831 current->fsgid = current->egid;
832 if (rgid != (gid_t) -1)
833 current->gid = rgid;
834 if (sgid != (gid_t) -1)
835 current->sgid = sgid;
836
837 key_fsgid_changed(current);
838 return 0;
839}
840
841asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
842{
843 int retval;
844
845 if (!(retval = put_user(current->gid, rgid)) &&
846 !(retval = put_user(current->egid, egid)))
847 retval = put_user(current->sgid, sgid);
848
849 return retval;
850}
851
852
853/*
854 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
855 * is used for "access()" and for the NFS daemon (letting nfsd stay at
856 * whatever uid it wants to). It normally shadows "euid", except when
857 * explicitly set by setfsuid() or for access..
858 */
859asmlinkage long sys_setfsuid(uid_t uid)
860{
861 int old_fsuid;
862
863 old_fsuid = current->fsuid;
864 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
865 return old_fsuid;
866
867 if (uid == current->uid || uid == current->euid ||
868 uid == current->suid || uid == current->fsuid ||
869 capable(CAP_SETUID))
870 {
871 if (uid != old_fsuid)
872 {
Alan Coxd6e71142005-06-23 00:09:43 -0700873 current->mm->dumpable = suid_dumpable;
akpm@osdl.orgd59dd462005-05-01 08:58:47 -0700874 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 }
876 current->fsuid = uid;
877 }
878
879 key_fsuid_changed(current);
880
881 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
882
883 return old_fsuid;
884}
885
886/*
887 * Samma på svenska..
888 */
889asmlinkage long sys_setfsgid(gid_t gid)
890{
891 int old_fsgid;
892
893 old_fsgid = current->fsgid;
894 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
895 return old_fsgid;
896
897 if (gid == current->gid || gid == current->egid ||
898 gid == current->sgid || gid == current->fsgid ||
899 capable(CAP_SETGID))
900 {
901 if (gid != old_fsgid)
902 {
Alan Coxd6e71142005-06-23 00:09:43 -0700903 current->mm->dumpable = suid_dumpable;
akpm@osdl.orgd59dd462005-05-01 08:58:47 -0700904 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905 }
906 current->fsgid = gid;
907 key_fsgid_changed(current);
908 }
909 return old_fsgid;
910}
911
912asmlinkage long sys_times(struct tms __user * tbuf)
913{
914 /*
915 * In the SMP world we might just be unlucky and have one of
916 * the times increment as we use it. Since the value is an
917 * atomically safe type this is just fine. Conceptually its
918 * as if the syscall took an instant longer to occur.
919 */
920 if (tbuf) {
921 struct tms tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922 cputime_t utime, stime, cutime, cstime;
923
Christoph Lameter71a22242005-06-23 00:10:05 -0700924#ifdef CONFIG_SMP
925 if (thread_group_empty(current)) {
926 /*
927 * Single thread case without the use of any locks.
928 *
929 * We may race with release_task if two threads are
930 * executing. However, release task first adds up the
931 * counters (__exit_signal) before removing the task
932 * from the process tasklist (__unhash_process).
933 * __exit_signal also acquires and releases the
934 * siglock which results in the proper memory ordering
935 * so that the list modifications are always visible
936 * after the counters have been updated.
937 *
938 * If the counters have been updated by the second thread
939 * but the thread has not yet been removed from the list
940 * then the other branch will be executing which will
941 * block on tasklist_lock until the exit handling of the
942 * other task is finished.
943 *
944 * This also implies that the sighand->siglock cannot
945 * be held by another processor. So we can also
946 * skip acquiring that lock.
947 */
948 utime = cputime_add(current->signal->utime, current->utime);
949 stime = cputime_add(current->signal->utime, current->stime);
950 cutime = current->signal->cutime;
951 cstime = current->signal->cstime;
952 } else
953#endif
954 {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955
Christoph Lameter71a22242005-06-23 00:10:05 -0700956 /* Process with multiple threads */
957 struct task_struct *tsk = current;
958 struct task_struct *t;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959
Christoph Lameter71a22242005-06-23 00:10:05 -0700960 read_lock(&tasklist_lock);
961 utime = tsk->signal->utime;
962 stime = tsk->signal->stime;
963 t = tsk;
964 do {
965 utime = cputime_add(utime, t->utime);
966 stime = cputime_add(stime, t->stime);
967 t = next_thread(t);
968 } while (t != tsk);
969
970 /*
971 * While we have tasklist_lock read-locked, no dying thread
972 * can be updating current->signal->[us]time. Instead,
973 * we got their counts included in the live thread loop.
974 * However, another thread can come in right now and
975 * do a wait call that updates current->signal->c[us]time.
976 * To make sure we always see that pair updated atomically,
977 * we take the siglock around fetching them.
978 */
979 spin_lock_irq(&tsk->sighand->siglock);
980 cutime = tsk->signal->cutime;
981 cstime = tsk->signal->cstime;
982 spin_unlock_irq(&tsk->sighand->siglock);
983 read_unlock(&tasklist_lock);
984 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985 tmp.tms_utime = cputime_to_clock_t(utime);
986 tmp.tms_stime = cputime_to_clock_t(stime);
987 tmp.tms_cutime = cputime_to_clock_t(cutime);
988 tmp.tms_cstime = cputime_to_clock_t(cstime);
989 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
990 return -EFAULT;
991 }
992 return (long) jiffies_64_to_clock_t(get_jiffies_64());
993}
994
995/*
996 * This needs some heavy checking ...
997 * I just haven't the stomach for it. I also don't fully
998 * understand sessions/pgrp etc. Let somebody who does explain it.
999 *
1000 * OK, I think I have the protection semantics right.... this is really
1001 * only important on a multi-user system anyway, to make sure one user
1002 * can't send a signal to a process owned by another. -TYT, 12/12/91
1003 *
1004 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1005 * LBT 04.03.94
1006 */
1007
1008asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1009{
1010 struct task_struct *p;
1011 int err = -EINVAL;
1012
1013 if (!pid)
1014 pid = current->pid;
1015 if (!pgid)
1016 pgid = pid;
1017 if (pgid < 0)
1018 return -EINVAL;
1019
1020 /* From this point forward we keep holding onto the tasklist lock
1021 * so that our parent does not change from under us. -DaveM
1022 */
1023 write_lock_irq(&tasklist_lock);
1024
1025 err = -ESRCH;
1026 p = find_task_by_pid(pid);
1027 if (!p)
1028 goto out;
1029
1030 err = -EINVAL;
1031 if (!thread_group_leader(p))
1032 goto out;
1033
1034 if (p->parent == current || p->real_parent == current) {
1035 err = -EPERM;
1036 if (p->signal->session != current->signal->session)
1037 goto out;
1038 err = -EACCES;
1039 if (p->did_exec)
1040 goto out;
1041 } else {
1042 err = -ESRCH;
1043 if (p != current)
1044 goto out;
1045 }
1046
1047 err = -EPERM;
1048 if (p->signal->leader)
1049 goto out;
1050
1051 if (pgid != pid) {
1052 struct task_struct *p;
1053
1054 do_each_task_pid(pgid, PIDTYPE_PGID, p) {
1055 if (p->signal->session == current->signal->session)
1056 goto ok_pgid;
1057 } while_each_task_pid(pgid, PIDTYPE_PGID, p);
1058 goto out;
1059 }
1060
1061ok_pgid:
1062 err = security_task_setpgid(p, pgid);
1063 if (err)
1064 goto out;
1065
1066 if (process_group(p) != pgid) {
1067 detach_pid(p, PIDTYPE_PGID);
1068 p->signal->pgrp = pgid;
1069 attach_pid(p, PIDTYPE_PGID, pgid);
1070 }
1071
1072 err = 0;
1073out:
1074 /* All paths lead to here, thus we are safe. -DaveM */
1075 write_unlock_irq(&tasklist_lock);
1076 return err;
1077}
1078
1079asmlinkage long sys_getpgid(pid_t pid)
1080{
1081 if (!pid) {
1082 return process_group(current);
1083 } else {
1084 int retval;
1085 struct task_struct *p;
1086
1087 read_lock(&tasklist_lock);
1088 p = find_task_by_pid(pid);
1089
1090 retval = -ESRCH;
1091 if (p) {
1092 retval = security_task_getpgid(p);
1093 if (!retval)
1094 retval = process_group(p);
1095 }
1096 read_unlock(&tasklist_lock);
1097 return retval;
1098 }
1099}
1100
1101#ifdef __ARCH_WANT_SYS_GETPGRP
1102
1103asmlinkage long sys_getpgrp(void)
1104{
1105 /* SMP - assuming writes are word atomic this is fine */
1106 return process_group(current);
1107}
1108
1109#endif
1110
1111asmlinkage long sys_getsid(pid_t pid)
1112{
1113 if (!pid) {
1114 return current->signal->session;
1115 } else {
1116 int retval;
1117 struct task_struct *p;
1118
1119 read_lock(&tasklist_lock);
1120 p = find_task_by_pid(pid);
1121
1122 retval = -ESRCH;
1123 if(p) {
1124 retval = security_task_getsid(p);
1125 if (!retval)
1126 retval = p->signal->session;
1127 }
1128 read_unlock(&tasklist_lock);
1129 return retval;
1130 }
1131}
1132
1133asmlinkage long sys_setsid(void)
1134{
1135 struct pid *pid;
1136 int err = -EPERM;
1137
1138 if (!thread_group_leader(current))
1139 return -EINVAL;
1140
1141 down(&tty_sem);
1142 write_lock_irq(&tasklist_lock);
1143
1144 pid = find_pid(PIDTYPE_PGID, current->pid);
1145 if (pid)
1146 goto out;
1147
1148 current->signal->leader = 1;
1149 __set_special_pids(current->pid, current->pid);
1150 current->signal->tty = NULL;
1151 current->signal->tty_old_pgrp = 0;
1152 err = process_group(current);
1153out:
1154 write_unlock_irq(&tasklist_lock);
1155 up(&tty_sem);
1156 return err;
1157}
1158
1159/*
1160 * Supplementary group IDs
1161 */
1162
1163/* init to 2 - one for init_task, one to ensure it is never freed */
1164struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1165
1166struct group_info *groups_alloc(int gidsetsize)
1167{
1168 struct group_info *group_info;
1169 int nblocks;
1170 int i;
1171
1172 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1173 /* Make sure we always allocate at least one indirect block pointer */
1174 nblocks = nblocks ? : 1;
1175 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1176 if (!group_info)
1177 return NULL;
1178 group_info->ngroups = gidsetsize;
1179 group_info->nblocks = nblocks;
1180 atomic_set(&group_info->usage, 1);
1181
1182 if (gidsetsize <= NGROUPS_SMALL) {
1183 group_info->blocks[0] = group_info->small_block;
1184 } else {
1185 for (i = 0; i < nblocks; i++) {
1186 gid_t *b;
1187 b = (void *)__get_free_page(GFP_USER);
1188 if (!b)
1189 goto out_undo_partial_alloc;
1190 group_info->blocks[i] = b;
1191 }
1192 }
1193 return group_info;
1194
1195out_undo_partial_alloc:
1196 while (--i >= 0) {
1197 free_page((unsigned long)group_info->blocks[i]);
1198 }
1199 kfree(group_info);
1200 return NULL;
1201}
1202
1203EXPORT_SYMBOL(groups_alloc);
1204
1205void groups_free(struct group_info *group_info)
1206{
1207 if (group_info->blocks[0] != group_info->small_block) {
1208 int i;
1209 for (i = 0; i < group_info->nblocks; i++)
1210 free_page((unsigned long)group_info->blocks[i]);
1211 }
1212 kfree(group_info);
1213}
1214
1215EXPORT_SYMBOL(groups_free);
1216
1217/* export the group_info to a user-space array */
1218static int groups_to_user(gid_t __user *grouplist,
1219 struct group_info *group_info)
1220{
1221 int i;
1222 int count = group_info->ngroups;
1223
1224 for (i = 0; i < group_info->nblocks; i++) {
1225 int cp_count = min(NGROUPS_PER_BLOCK, count);
1226 int off = i * NGROUPS_PER_BLOCK;
1227 int len = cp_count * sizeof(*grouplist);
1228
1229 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1230 return -EFAULT;
1231
1232 count -= cp_count;
1233 }
1234 return 0;
1235}
1236
1237/* fill a group_info from a user-space array - it must be allocated already */
1238static int groups_from_user(struct group_info *group_info,
1239 gid_t __user *grouplist)
1240 {
1241 int i;
1242 int count = group_info->ngroups;
1243
1244 for (i = 0; i < group_info->nblocks; i++) {
1245 int cp_count = min(NGROUPS_PER_BLOCK, count);
1246 int off = i * NGROUPS_PER_BLOCK;
1247 int len = cp_count * sizeof(*grouplist);
1248
1249 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1250 return -EFAULT;
1251
1252 count -= cp_count;
1253 }
1254 return 0;
1255}
1256
Domen Puncerebe8b542005-05-05 16:16:19 -07001257/* a simple Shell sort */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258static void groups_sort(struct group_info *group_info)
1259{
1260 int base, max, stride;
1261 int gidsetsize = group_info->ngroups;
1262
1263 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1264 ; /* nothing */
1265 stride /= 3;
1266
1267 while (stride) {
1268 max = gidsetsize - stride;
1269 for (base = 0; base < max; base++) {
1270 int left = base;
1271 int right = left + stride;
1272 gid_t tmp = GROUP_AT(group_info, right);
1273
1274 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1275 GROUP_AT(group_info, right) =
1276 GROUP_AT(group_info, left);
1277 right = left;
1278 left -= stride;
1279 }
1280 GROUP_AT(group_info, right) = tmp;
1281 }
1282 stride /= 3;
1283 }
1284}
1285
1286/* a simple bsearch */
David Howells3e301482005-06-23 22:00:56 -07001287int groups_search(struct group_info *group_info, gid_t grp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288{
1289 int left, right;
1290
1291 if (!group_info)
1292 return 0;
1293
1294 left = 0;
1295 right = group_info->ngroups;
1296 while (left < right) {
1297 int mid = (left+right)/2;
1298 int cmp = grp - GROUP_AT(group_info, mid);
1299 if (cmp > 0)
1300 left = mid + 1;
1301 else if (cmp < 0)
1302 right = mid;
1303 else
1304 return 1;
1305 }
1306 return 0;
1307}
1308
1309/* validate and set current->group_info */
1310int set_current_groups(struct group_info *group_info)
1311{
1312 int retval;
1313 struct group_info *old_info;
1314
1315 retval = security_task_setgroups(group_info);
1316 if (retval)
1317 return retval;
1318
1319 groups_sort(group_info);
1320 get_group_info(group_info);
1321
1322 task_lock(current);
1323 old_info = current->group_info;
1324 current->group_info = group_info;
1325 task_unlock(current);
1326
1327 put_group_info(old_info);
1328
1329 return 0;
1330}
1331
1332EXPORT_SYMBOL(set_current_groups);
1333
1334asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1335{
1336 int i = 0;
1337
1338 /*
1339 * SMP: Nobody else can change our grouplist. Thus we are
1340 * safe.
1341 */
1342
1343 if (gidsetsize < 0)
1344 return -EINVAL;
1345
1346 /* no need to grab task_lock here; it cannot change */
1347 get_group_info(current->group_info);
1348 i = current->group_info->ngroups;
1349 if (gidsetsize) {
1350 if (i > gidsetsize) {
1351 i = -EINVAL;
1352 goto out;
1353 }
1354 if (groups_to_user(grouplist, current->group_info)) {
1355 i = -EFAULT;
1356 goto out;
1357 }
1358 }
1359out:
1360 put_group_info(current->group_info);
1361 return i;
1362}
1363
1364/*
1365 * SMP: Our groups are copy-on-write. We can set them safely
1366 * without another task interfering.
1367 */
1368
1369asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1370{
1371 struct group_info *group_info;
1372 int retval;
1373
1374 if (!capable(CAP_SETGID))
1375 return -EPERM;
1376 if ((unsigned)gidsetsize > NGROUPS_MAX)
1377 return -EINVAL;
1378
1379 group_info = groups_alloc(gidsetsize);
1380 if (!group_info)
1381 return -ENOMEM;
1382 retval = groups_from_user(group_info, grouplist);
1383 if (retval) {
1384 put_group_info(group_info);
1385 return retval;
1386 }
1387
1388 retval = set_current_groups(group_info);
1389 put_group_info(group_info);
1390
1391 return retval;
1392}
1393
1394/*
1395 * Check whether we're fsgid/egid or in the supplemental group..
1396 */
1397int in_group_p(gid_t grp)
1398{
1399 int retval = 1;
1400 if (grp != current->fsgid) {
1401 get_group_info(current->group_info);
1402 retval = groups_search(current->group_info, grp);
1403 put_group_info(current->group_info);
1404 }
1405 return retval;
1406}
1407
1408EXPORT_SYMBOL(in_group_p);
1409
1410int in_egroup_p(gid_t grp)
1411{
1412 int retval = 1;
1413 if (grp != current->egid) {
1414 get_group_info(current->group_info);
1415 retval = groups_search(current->group_info, grp);
1416 put_group_info(current->group_info);
1417 }
1418 return retval;
1419}
1420
1421EXPORT_SYMBOL(in_egroup_p);
1422
1423DECLARE_RWSEM(uts_sem);
1424
1425EXPORT_SYMBOL(uts_sem);
1426
1427asmlinkage long sys_newuname(struct new_utsname __user * name)
1428{
1429 int errno = 0;
1430
1431 down_read(&uts_sem);
1432 if (copy_to_user(name,&system_utsname,sizeof *name))
1433 errno = -EFAULT;
1434 up_read(&uts_sem);
1435 return errno;
1436}
1437
1438asmlinkage long sys_sethostname(char __user *name, int len)
1439{
1440 int errno;
1441 char tmp[__NEW_UTS_LEN];
1442
1443 if (!capable(CAP_SYS_ADMIN))
1444 return -EPERM;
1445 if (len < 0 || len > __NEW_UTS_LEN)
1446 return -EINVAL;
1447 down_write(&uts_sem);
1448 errno = -EFAULT;
1449 if (!copy_from_user(tmp, name, len)) {
1450 memcpy(system_utsname.nodename, tmp, len);
1451 system_utsname.nodename[len] = 0;
1452 errno = 0;
1453 }
1454 up_write(&uts_sem);
1455 return errno;
1456}
1457
1458#ifdef __ARCH_WANT_SYS_GETHOSTNAME
1459
1460asmlinkage long sys_gethostname(char __user *name, int len)
1461{
1462 int i, errno;
1463
1464 if (len < 0)
1465 return -EINVAL;
1466 down_read(&uts_sem);
1467 i = 1 + strlen(system_utsname.nodename);
1468 if (i > len)
1469 i = len;
1470 errno = 0;
1471 if (copy_to_user(name, system_utsname.nodename, i))
1472 errno = -EFAULT;
1473 up_read(&uts_sem);
1474 return errno;
1475}
1476
1477#endif
1478
1479/*
1480 * Only setdomainname; getdomainname can be implemented by calling
1481 * uname()
1482 */
1483asmlinkage long sys_setdomainname(char __user *name, int len)
1484{
1485 int errno;
1486 char tmp[__NEW_UTS_LEN];
1487
1488 if (!capable(CAP_SYS_ADMIN))
1489 return -EPERM;
1490 if (len < 0 || len > __NEW_UTS_LEN)
1491 return -EINVAL;
1492
1493 down_write(&uts_sem);
1494 errno = -EFAULT;
1495 if (!copy_from_user(tmp, name, len)) {
1496 memcpy(system_utsname.domainname, tmp, len);
1497 system_utsname.domainname[len] = 0;
1498 errno = 0;
1499 }
1500 up_write(&uts_sem);
1501 return errno;
1502}
1503
1504asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1505{
1506 if (resource >= RLIM_NLIMITS)
1507 return -EINVAL;
1508 else {
1509 struct rlimit value;
1510 task_lock(current->group_leader);
1511 value = current->signal->rlim[resource];
1512 task_unlock(current->group_leader);
1513 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1514 }
1515}
1516
1517#ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1518
1519/*
1520 * Back compatibility for getrlimit. Needed for some apps.
1521 */
1522
1523asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1524{
1525 struct rlimit x;
1526 if (resource >= RLIM_NLIMITS)
1527 return -EINVAL;
1528
1529 task_lock(current->group_leader);
1530 x = current->signal->rlim[resource];
1531 task_unlock(current->group_leader);
1532 if(x.rlim_cur > 0x7FFFFFFF)
1533 x.rlim_cur = 0x7FFFFFFF;
1534 if(x.rlim_max > 0x7FFFFFFF)
1535 x.rlim_max = 0x7FFFFFFF;
1536 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1537}
1538
1539#endif
1540
1541asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1542{
1543 struct rlimit new_rlim, *old_rlim;
1544 int retval;
1545
1546 if (resource >= RLIM_NLIMITS)
1547 return -EINVAL;
1548 if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1549 return -EFAULT;
1550 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1551 return -EINVAL;
1552 old_rlim = current->signal->rlim + resource;
1553 if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1554 !capable(CAP_SYS_RESOURCE))
1555 return -EPERM;
1556 if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1557 return -EPERM;
1558
1559 retval = security_task_setrlimit(resource, &new_rlim);
1560 if (retval)
1561 return retval;
1562
1563 task_lock(current->group_leader);
1564 *old_rlim = new_rlim;
1565 task_unlock(current->group_leader);
1566
1567 if (resource == RLIMIT_CPU && new_rlim.rlim_cur != RLIM_INFINITY &&
1568 (cputime_eq(current->signal->it_prof_expires, cputime_zero) ||
1569 new_rlim.rlim_cur <= cputime_to_secs(
1570 current->signal->it_prof_expires))) {
1571 cputime_t cputime = secs_to_cputime(new_rlim.rlim_cur);
1572 read_lock(&tasklist_lock);
1573 spin_lock_irq(&current->sighand->siglock);
1574 set_process_cpu_timer(current, CPUCLOCK_PROF,
1575 &cputime, NULL);
1576 spin_unlock_irq(&current->sighand->siglock);
1577 read_unlock(&tasklist_lock);
1578 }
1579
1580 return 0;
1581}
1582
1583/*
1584 * It would make sense to put struct rusage in the task_struct,
1585 * except that would make the task_struct be *really big*. After
1586 * task_struct gets moved into malloc'ed memory, it would
1587 * make sense to do this. It will make moving the rest of the information
1588 * a lot simpler! (Which we're not doing right now because we're not
1589 * measuring them yet).
1590 *
1591 * This expects to be called with tasklist_lock read-locked or better,
1592 * and the siglock not locked. It may momentarily take the siglock.
1593 *
1594 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1595 * races with threads incrementing their own counters. But since word
1596 * reads are atomic, we either get new values or old values and we don't
1597 * care which for the sums. We always take the siglock to protect reading
1598 * the c* fields from p->signal from races with exit.c updating those
1599 * fields when reaping, so a sample either gets all the additions of a
1600 * given child after it's reaped, or none so this sample is before reaping.
1601 */
1602
1603static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1604{
1605 struct task_struct *t;
1606 unsigned long flags;
1607 cputime_t utime, stime;
1608
1609 memset((char *) r, 0, sizeof *r);
1610
1611 if (unlikely(!p->signal))
1612 return;
1613
1614 switch (who) {
1615 case RUSAGE_CHILDREN:
1616 spin_lock_irqsave(&p->sighand->siglock, flags);
1617 utime = p->signal->cutime;
1618 stime = p->signal->cstime;
1619 r->ru_nvcsw = p->signal->cnvcsw;
1620 r->ru_nivcsw = p->signal->cnivcsw;
1621 r->ru_minflt = p->signal->cmin_flt;
1622 r->ru_majflt = p->signal->cmaj_flt;
1623 spin_unlock_irqrestore(&p->sighand->siglock, flags);
1624 cputime_to_timeval(utime, &r->ru_utime);
1625 cputime_to_timeval(stime, &r->ru_stime);
1626 break;
1627 case RUSAGE_SELF:
1628 spin_lock_irqsave(&p->sighand->siglock, flags);
1629 utime = stime = cputime_zero;
1630 goto sum_group;
1631 case RUSAGE_BOTH:
1632 spin_lock_irqsave(&p->sighand->siglock, flags);
1633 utime = p->signal->cutime;
1634 stime = p->signal->cstime;
1635 r->ru_nvcsw = p->signal->cnvcsw;
1636 r->ru_nivcsw = p->signal->cnivcsw;
1637 r->ru_minflt = p->signal->cmin_flt;
1638 r->ru_majflt = p->signal->cmaj_flt;
1639 sum_group:
1640 utime = cputime_add(utime, p->signal->utime);
1641 stime = cputime_add(stime, p->signal->stime);
1642 r->ru_nvcsw += p->signal->nvcsw;
1643 r->ru_nivcsw += p->signal->nivcsw;
1644 r->ru_minflt += p->signal->min_flt;
1645 r->ru_majflt += p->signal->maj_flt;
1646 t = p;
1647 do {
1648 utime = cputime_add(utime, t->utime);
1649 stime = cputime_add(stime, t->stime);
1650 r->ru_nvcsw += t->nvcsw;
1651 r->ru_nivcsw += t->nivcsw;
1652 r->ru_minflt += t->min_flt;
1653 r->ru_majflt += t->maj_flt;
1654 t = next_thread(t);
1655 } while (t != p);
1656 spin_unlock_irqrestore(&p->sighand->siglock, flags);
1657 cputime_to_timeval(utime, &r->ru_utime);
1658 cputime_to_timeval(stime, &r->ru_stime);
1659 break;
1660 default:
1661 BUG();
1662 }
1663}
1664
1665int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1666{
1667 struct rusage r;
1668 read_lock(&tasklist_lock);
1669 k_getrusage(p, who, &r);
1670 read_unlock(&tasklist_lock);
1671 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1672}
1673
1674asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1675{
1676 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1677 return -EINVAL;
1678 return getrusage(current, who, ru);
1679}
1680
1681asmlinkage long sys_umask(int mask)
1682{
1683 mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1684 return mask;
1685}
1686
1687asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1688 unsigned long arg4, unsigned long arg5)
1689{
1690 long error;
1691 int sig;
1692
1693 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1694 if (error)
1695 return error;
1696
1697 switch (option) {
1698 case PR_SET_PDEATHSIG:
1699 sig = arg2;
Jesper Juhl7ed20e12005-05-01 08:59:14 -07001700 if (!valid_signal(sig)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001701 error = -EINVAL;
1702 break;
1703 }
1704 current->pdeath_signal = sig;
1705 break;
1706 case PR_GET_PDEATHSIG:
1707 error = put_user(current->pdeath_signal, (int __user *)arg2);
1708 break;
1709 case PR_GET_DUMPABLE:
1710 if (current->mm->dumpable)
1711 error = 1;
1712 break;
1713 case PR_SET_DUMPABLE:
Alan Coxd6e71142005-06-23 00:09:43 -07001714 if (arg2 < 0 || arg2 > 2) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715 error = -EINVAL;
1716 break;
1717 }
1718 current->mm->dumpable = arg2;
1719 break;
1720
1721 case PR_SET_UNALIGN:
1722 error = SET_UNALIGN_CTL(current, arg2);
1723 break;
1724 case PR_GET_UNALIGN:
1725 error = GET_UNALIGN_CTL(current, arg2);
1726 break;
1727 case PR_SET_FPEMU:
1728 error = SET_FPEMU_CTL(current, arg2);
1729 break;
1730 case PR_GET_FPEMU:
1731 error = GET_FPEMU_CTL(current, arg2);
1732 break;
1733 case PR_SET_FPEXC:
1734 error = SET_FPEXC_CTL(current, arg2);
1735 break;
1736 case PR_GET_FPEXC:
1737 error = GET_FPEXC_CTL(current, arg2);
1738 break;
1739 case PR_GET_TIMING:
1740 error = PR_TIMING_STATISTICAL;
1741 break;
1742 case PR_SET_TIMING:
1743 if (arg2 == PR_TIMING_STATISTICAL)
1744 error = 0;
1745 else
1746 error = -EINVAL;
1747 break;
1748
1749 case PR_GET_KEEPCAPS:
1750 if (current->keep_capabilities)
1751 error = 1;
1752 break;
1753 case PR_SET_KEEPCAPS:
1754 if (arg2 != 0 && arg2 != 1) {
1755 error = -EINVAL;
1756 break;
1757 }
1758 current->keep_capabilities = arg2;
1759 break;
1760 case PR_SET_NAME: {
1761 struct task_struct *me = current;
1762 unsigned char ncomm[sizeof(me->comm)];
1763
1764 ncomm[sizeof(me->comm)-1] = 0;
1765 if (strncpy_from_user(ncomm, (char __user *)arg2,
1766 sizeof(me->comm)-1) < 0)
1767 return -EFAULT;
1768 set_task_comm(me, ncomm);
1769 return 0;
1770 }
1771 case PR_GET_NAME: {
1772 struct task_struct *me = current;
1773 unsigned char tcomm[sizeof(me->comm)];
1774
1775 get_task_comm(tcomm, me);
1776 if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))
1777 return -EFAULT;
1778 return 0;
1779 }
1780 default:
1781 error = -EINVAL;
1782 break;
1783 }
1784 return error;
1785}