File: /usr/src/linux/kernel/sys.c
1 /*
2 * linux/kernel/sys.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
7 #include <linux/module.h>
8 #include <linux/mm.h>
9 #include <linux/utsname.h>
10 #include <linux/mman.h>
11 #include <linux/smp_lock.h>
12 #include <linux/notifier.h>
13 #include <linux/reboot.h>
14 #include <linux/prctl.h>
15 #include <linux/init.h>
16 #include <linux/highuid.h>
17
18 #include <asm/uaccess.h>
19 #include <asm/io.h>
20
21 /*
22 * this is where the system-wide overflow UID and GID are defined, for
23 * architectures that now have 32-bit UID/GID but didn't in the past
24 */
25
26 int overflowuid = DEFAULT_OVERFLOWUID;
27 int overflowgid = DEFAULT_OVERFLOWGID;
28
29 /*
30 * the same as above, but for filesystems which can only store a 16-bit
31 * UID and GID. as such, this is needed on all architectures
32 */
33
34 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
35 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
36
37 /*
38 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
39 */
40
41 int C_A_D = 1;
42 int cad_pid = 1;
43
44
45 /*
46 * Notifier list for kernel code which wants to be called
47 * at shutdown. This is used to stop any idling DMA operations
48 * and the like.
49 */
50
51 static struct notifier_block *reboot_notifier_list;
52 rwlock_t notifier_lock = RW_LOCK_UNLOCKED;
53
54 /**
55 * notifier_chain_register - Add notifier to a notifier chain
56 * @list: Pointer to root list pointer
57 * @n: New entry in notifier chain
58 *
59 * Adds a notifier to a notifier chain.
60 *
61 * Currently always returns zero.
62 */
63
64 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
65 {
66 write_lock(¬ifier_lock);
67 while(*list)
68 {
69 if(n->priority > (*list)->priority)
70 break;
71 list= &((*list)->next);
72 }
73 n->next = *list;
74 *list=n;
75 write_unlock(¬ifier_lock);
76 return 0;
77 }
78
79 /**
80 * notifier_chain_unregister - Remove notifier from a notifier chain
81 * @nl: Pointer to root list pointer
82 * @n: New entry in notifier chain
83 *
84 * Removes a notifier from a notifier chain.
85 *
86 * Returns zero on success, or %-ENOENT on failure.
87 */
88
89 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
90 {
91 write_lock(¬ifier_lock);
92 while((*nl)!=NULL)
93 {
94 if((*nl)==n)
95 {
96 *nl=n->next;
97 write_unlock(¬ifier_lock);
98 return 0;
99 }
100 nl=&((*nl)->next);
101 }
102 write_unlock(¬ifier_lock);
103 return -ENOENT;
104 }
105
106 /**
107 * notifier_call_chain - Call functions in a notifier chain
108 * @n: Pointer to root pointer of notifier chain
109 * @val: Value passed unmodified to notifier function
110 * @v: Pointer passed unmodified to notifier function
111 *
112 * Calls each function in a notifier chain in turn.
113 *
114 * If the return value of the notifier can be and'd
115 * with %NOTIFY_STOP_MASK, then notifier_call_chain
116 * will return immediately, with the return value of
117 * the notifier function which halted execution.
118 * Otherwise, the return value is the return value
119 * of the last notifier function called.
120 */
121
122 int notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
123 {
124 int ret=NOTIFY_DONE;
125 struct notifier_block *nb = *n;
126
127 while(nb)
128 {
129 ret=nb->notifier_call(nb,val,v);
130 if(ret&NOTIFY_STOP_MASK)
131 {
132 return ret;
133 }
134 nb=nb->next;
135 }
136 return ret;
137 }
138
139 /**
140 * register_reboot_notifier - Register function to be called at reboot time
141 * @nb: Info about notifier function to be called
142 *
143 * Registers a function with the list of functions
144 * to be called at reboot time.
145 *
146 * Currently always returns zero, as notifier_chain_register
147 * always returns zero.
148 */
149
150 int register_reboot_notifier(struct notifier_block * nb)
151 {
152 return notifier_chain_register(&reboot_notifier_list, nb);
153 }
154
155 /**
156 * unregister_reboot_notifier - Unregister previously registered reboot notifier
157 * @nb: Hook to be unregistered
158 *
159 * Unregisters a previously registered reboot
160 * notifier function.
161 *
162 * Returns zero on success, or %-ENOENT on failure.
163 */
164
165 int unregister_reboot_notifier(struct notifier_block * nb)
166 {
167 return notifier_chain_unregister(&reboot_notifier_list, nb);
168 }
169
170 asmlinkage long sys_ni_syscall(void)
171 {
172 return -ENOSYS;
173 }
174
175 static int proc_sel(struct task_struct *p, int which, int who)
176 {
177 if(p->pid)
178 {
179 switch (which) {
180 case PRIO_PROCESS:
181 if (!who && p == current)
182 return 1;
183 return(p->pid == who);
184 case PRIO_PGRP:
185 if (!who)
186 who = current->pgrp;
187 return(p->pgrp == who);
188 case PRIO_USER:
189 if (!who)
190 who = current->uid;
191 return(p->uid == who);
192 }
193 }
194 return 0;
195 }
196
197 asmlinkage long sys_setpriority(int which, int who, int niceval)
198 {
199 struct task_struct *p;
200 int error;
201
202 if (which > 2 || which < 0)
203 return -EINVAL;
204
205 /* normalize: avoid signed division (rounding problems) */
206 error = -ESRCH;
207 if (niceval < -20)
208 niceval = -20;
209 if (niceval > 19)
210 niceval = 19;
211
212 read_lock(&tasklist_lock);
213 for_each_task(p) {
214 if (!proc_sel(p, which, who))
215 continue;
216 if (p->uid != current->euid &&
217 p->uid != current->uid && !capable(CAP_SYS_NICE)) {
218 error = -EPERM;
219 continue;
220 }
221 if (error == -ESRCH)
222 error = 0;
223 if (niceval < p->nice && !capable(CAP_SYS_NICE))
224 error = -EACCES;
225 else
226 p->nice = niceval;
227 }
228 read_unlock(&tasklist_lock);
229
230 return error;
231 }
232
233 /*
234 * Ugh. To avoid negative return values, "getpriority()" will
235 * not return the normal nice-value, but a negated value that
236 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
237 * to stay compatible.
238 */
239 asmlinkage long sys_getpriority(int which, int who)
240 {
241 struct task_struct *p;
242 long retval = -ESRCH;
243
244 if (which > 2 || which < 0)
245 return -EINVAL;
246
247 read_lock(&tasklist_lock);
248 for_each_task (p) {
249 long niceval;
250 if (!proc_sel(p, which, who))
251 continue;
252 niceval = 20 - p->nice;
253 if (niceval > retval)
254 retval = niceval;
255 }
256 read_unlock(&tasklist_lock);
257
258 return retval;
259 }
260
261
262 /*
263 * Reboot system call: for obvious reasons only root may call it,
264 * and even root needs to set up some magic numbers in the registers
265 * so that some mistake won't make this reboot the whole machine.
266 * You can also set the meaning of the ctrl-alt-del-key here.
267 *
268 * reboot doesn't sync: do that yourself before calling this.
269 */
270 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void * arg)
271 {
272 char buffer[256];
273
274 /* We only trust the superuser with rebooting the system. */
275 if (!capable(CAP_SYS_BOOT))
276 return -EPERM;
277
278 /* For safety, we require "magic" arguments. */
279 if (magic1 != LINUX_REBOOT_MAGIC1 ||
280 (magic2 != LINUX_REBOOT_MAGIC2 && magic2 != LINUX_REBOOT_MAGIC2A &&
281 magic2 != LINUX_REBOOT_MAGIC2B))
282 return -EINVAL;
283
284 lock_kernel();
285 switch (cmd) {
286 case LINUX_REBOOT_CMD_RESTART:
287 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
288 printk(KERN_EMERG "Restarting system.\n");
289 machine_restart(NULL);
290 break;
291
292 case LINUX_REBOOT_CMD_CAD_ON:
293 C_A_D = 1;
294 break;
295
296 case LINUX_REBOOT_CMD_CAD_OFF:
297 C_A_D = 0;
298 break;
299
300 case LINUX_REBOOT_CMD_HALT:
301 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
302 printk(KERN_EMERG "System halted.\n");
303 machine_halt();
304 do_exit(0);
305 break;
306
307 case LINUX_REBOOT_CMD_POWER_OFF:
308 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
309 printk(KERN_EMERG "Power down.\n");
310 machine_power_off();
311 do_exit(0);
312 break;
313
314 case LINUX_REBOOT_CMD_RESTART2:
315 if (strncpy_from_user(&buffer[0], (char *)arg, sizeof(buffer) - 1) < 0) {
316 unlock_kernel();
317 return -EFAULT;
318 }
319 buffer[sizeof(buffer) - 1] = '\0';
320
321 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
322 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
323 machine_restart(buffer);
324 break;
325
326 default:
327 unlock_kernel();
328 return -EINVAL;
329 }
330 unlock_kernel();
331 return 0;
332 }
333
334 static void deferred_cad(void *dummy)
335 {
336 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
337 machine_restart(NULL);
338 }
339
340 /*
341 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
342 * As it's called within an interrupt, it may NOT sync: the only choice
343 * is whether to reboot at once, or just ignore the ctrl-alt-del.
344 */
345 void ctrl_alt_del(void)
346 {
347 static struct tq_struct cad_tq = {
348 routine: deferred_cad,
349 };
350
351 if (C_A_D)
352 schedule_task(&cad_tq);
353 else
354 kill_proc(cad_pid, SIGINT, 1);
355 }
356
357
358 /*
359 * Unprivileged users may change the real gid to the effective gid
360 * or vice versa. (BSD-style)
361 *
362 * If you set the real gid at all, or set the effective gid to a value not
363 * equal to the real gid, then the saved gid is set to the new effective gid.
364 *
365 * This makes it possible for a setgid program to completely drop its
366 * privileges, which is often a useful assertion to make when you are doing
367 * a security audit over a program.
368 *
369 * The general idea is that a program which uses just setregid() will be
370 * 100% compatible with BSD. A program which uses just setgid() will be
371 * 100% compatible with POSIX with saved IDs.
372 *
373 * SMP: There are not races, the GIDs are checked only by filesystem
374 * operations (as far as semantic preservation is concerned).
375 */
376 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
377 {
378 int old_rgid = current->gid;
379 int old_egid = current->egid;
380 int new_rgid = old_rgid;
381 int new_egid = old_egid;
382
383 if (rgid != (gid_t) -1) {
384 if ((old_rgid == rgid) ||
385 (current->egid==rgid) ||
386 capable(CAP_SETGID))
387 new_rgid = rgid;
388 else
389 return -EPERM;
390 }
391 if (egid != (gid_t) -1) {
392 if ((old_rgid == egid) ||
393 (current->egid == egid) ||
394 (current->sgid == egid) ||
395 capable(CAP_SETGID))
396 new_egid = egid;
397 else {
398 return -EPERM;
399 }
400 }
401 if (new_egid != old_egid)
402 {
403 current->mm->dumpable = 0;
404 wmb();
405 }
406 if (rgid != (gid_t) -1 ||
407 (egid != (gid_t) -1 && egid != old_rgid))
408 current->sgid = new_egid;
409 current->fsgid = new_egid;
410 current->egid = new_egid;
411 current->gid = new_rgid;
412 return 0;
413 }
414
415 /*
416 * setgid() is implemented like SysV w/ SAVED_IDS
417 *
418 * SMP: Same implicit races as above.
419 */
420 asmlinkage long sys_setgid(gid_t gid)
421 {
422 int old_egid = current->egid;
423
424 if (capable(CAP_SETGID))
425 {
426 if(old_egid != gid)
427 {
428 current->mm->dumpable=0;
429 wmb();
430 }
431 current->gid = current->egid = current->sgid = current->fsgid = gid;
432 }
433 else if ((gid == current->gid) || (gid == current->sgid))
434 {
435 if(old_egid != gid)
436 {
437 current->mm->dumpable=0;
438 wmb();
439 }
440 current->egid = current->fsgid = gid;
441 }
442 else
443 return -EPERM;
444 return 0;
445 }
446
447 /*
448 * cap_emulate_setxuid() fixes the effective / permitted capabilities of
449 * a process after a call to setuid, setreuid, or setresuid.
450 *
451 * 1) When set*uiding _from_ one of {r,e,s}uid == 0 _to_ all of
452 * {r,e,s}uid != 0, the permitted and effective capabilities are
453 * cleared.
454 *
455 * 2) When set*uiding _from_ euid == 0 _to_ euid != 0, the effective
456 * capabilities of the process are cleared.
457 *
458 * 3) When set*uiding _from_ euid != 0 _to_ euid == 0, the effective
459 * capabilities are set to the permitted capabilities.
460 *
461 * fsuid is handled elsewhere. fsuid == 0 and {r,e,s}uid!= 0 should
462 * never happen.
463 *
464 * -astor
465 *
466 * cevans - New behaviour, Oct '99
467 * A process may, via prctl(), elect to keep its capabilities when it
468 * calls setuid() and switches away from uid==0. Both permitted and
469 * effective sets will be retained.
470 * Without this change, it was impossible for a daemon to drop only some
471 * of its privilege. The call to setuid(!=0) would drop all privileges!
472 * Keeping uid 0 is not an option because uid 0 owns too many vital
473 * files..
474 * Thanks to Olaf Kirch and Peter Benie for spotting this.
475 */
476 static inline void cap_emulate_setxuid(int old_ruid, int old_euid,
477 int old_suid)
478 {
479 if ((old_ruid == 0 || old_euid == 0 || old_suid == 0) &&
480 (current->uid != 0 && current->euid != 0 && current->suid != 0) &&
481 !current->keep_capabilities) {
482 cap_clear(current->cap_permitted);
483 cap_clear(current->cap_effective);
484 }
485 if (old_euid == 0 && current->euid != 0) {
486 cap_clear(current->cap_effective);
487 }
488 if (old_euid != 0 && current->euid == 0) {
489 current->cap_effective = current->cap_permitted;
490 }
491 }
492
493 static int set_user(uid_t new_ruid, int dumpclear)
494 {
495 struct user_struct *new_user, *old_user;
496
497 /* What if a process setreuid()'s and this brings the
498 * new uid over his NPROC rlimit? We can check this now
499 * cheaply with the new uid cache, so if it matters
500 * we should be checking for it. -DaveM
501 */
502 new_user = alloc_uid(new_ruid);
503 if (!new_user)
504 return -EAGAIN;
505 old_user = current->user;
506 atomic_dec(&old_user->processes);
507 atomic_inc(&new_user->processes);
508
509 if(dumpclear)
510 {
511 current->mm->dumpable = 0;
512 wmb();
513 }
514 current->uid = new_ruid;
515 current->user = new_user;
516 free_uid(old_user);
517 return 0;
518 }
519
520 /*
521 * Unprivileged users may change the real uid to the effective uid
522 * or vice versa. (BSD-style)
523 *
524 * If you set the real uid at all, or set the effective uid to a value not
525 * equal to the real uid, then the saved uid is set to the new effective uid.
526 *
527 * This makes it possible for a setuid program to completely drop its
528 * privileges, which is often a useful assertion to make when you are doing
529 * a security audit over a program.
530 *
531 * The general idea is that a program which uses just setreuid() will be
532 * 100% compatible with BSD. A program which uses just setuid() will be
533 * 100% compatible with POSIX with saved IDs.
534 */
535 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
536 {
537 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
538
539 new_ruid = old_ruid = current->uid;
540 new_euid = old_euid = current->euid;
541 old_suid = current->suid;
542
543 if (ruid != (uid_t) -1) {
544 new_ruid = ruid;
545 if ((old_ruid != ruid) &&
546 (current->euid != ruid) &&
547 !capable(CAP_SETUID))
548 return -EPERM;
549 }
550
551 if (euid != (uid_t) -1) {
552 new_euid = euid;
553 if ((old_ruid != euid) &&
554 (current->euid != euid) &&
555 (current->suid != euid) &&
556 !capable(CAP_SETUID))
557 return -EPERM;
558 }
559
560 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
561 return -EAGAIN;
562
563 if (new_euid != old_euid)
564 {
565 current->mm->dumpable=0;
566 wmb();
567 }
568 current->fsuid = current->euid = new_euid;
569 if (ruid != (uid_t) -1 ||
570 (euid != (uid_t) -1 && euid != old_ruid))
571 current->suid = current->euid;
572 current->fsuid = current->euid;
573
574 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
575 cap_emulate_setxuid(old_ruid, old_euid, old_suid);
576 }
577
578 return 0;
579 }
580
581
582
583 /*
584 * setuid() is implemented like SysV with SAVED_IDS
585 *
586 * Note that SAVED_ID's is deficient in that a setuid root program
587 * like sendmail, for example, cannot set its uid to be a normal
588 * user and then switch back, because if you're root, setuid() sets
589 * the saved uid too. If you don't like this, blame the bright people
590 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
591 * will allow a root program to temporarily drop privileges and be able to
592 * regain them by swapping the real and effective uid.
593 */
594 asmlinkage long sys_setuid(uid_t uid)
595 {
596 int old_euid = current->euid;
597 int old_ruid, old_suid, new_ruid, new_suid;
598
599 old_ruid = new_ruid = current->uid;
600 old_suid = current->suid;
601 new_suid = old_suid;
602
603 if (capable(CAP_SETUID)) {
604 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
605 return -EAGAIN;
606 new_suid = uid;
607 } else if ((uid != current->uid) && (uid != new_suid))
608 return -EPERM;
609
610 if (old_euid != uid)
611 {
612 current->mm->dumpable = 0;
613 wmb();
614 }
615 current->fsuid = current->euid = uid;
616 current->suid = new_suid;
617
618 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
619 cap_emulate_setxuid(old_ruid, old_euid, old_suid);
620 }
621
622 return 0;
623 }
624
625
626 /*
627 * This function implements a generic ability to update ruid, euid,
628 * and suid. This allows you to implement the 4.4 compatible seteuid().
629 */
630 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
631 {
632 int old_ruid = current->uid;
633 int old_euid = current->euid;
634 int old_suid = current->suid;
635
636 if (!capable(CAP_SETUID)) {
637 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
638 (ruid != current->euid) && (ruid != current->suid))
639 return -EPERM;
640 if ((euid != (uid_t) -1) && (euid != current->uid) &&
641 (euid != current->euid) && (euid != current->suid))
642 return -EPERM;
643 if ((suid != (uid_t) -1) && (suid != current->uid) &&
644 (suid != current->euid) && (suid != current->suid))
645 return -EPERM;
646 }
647 if (ruid != (uid_t) -1) {
648 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
649 return -EAGAIN;
650 }
651 if (euid != (uid_t) -1) {
652 if (euid != current->euid)
653 {
654 current->mm->dumpable = 0;
655 wmb();
656 }
657 current->euid = euid;
658 current->fsuid = euid;
659 }
660 if (suid != (uid_t) -1)
661 current->suid = suid;
662
663 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
664 cap_emulate_setxuid(old_ruid, old_euid, old_suid);
665 }
666
667 return 0;
668 }
669
670 asmlinkage long sys_getresuid(uid_t *ruid, uid_t *euid, uid_t *suid)
671 {
672 int retval;
673
674 if (!(retval = put_user(current->uid, ruid)) &&
675 !(retval = put_user(current->euid, euid)))
676 retval = put_user(current->suid, suid);
677
678 return retval;
679 }
680
681 /*
682 * Same as above, but for rgid, egid, sgid.
683 */
684 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
685 {
686 if (!capable(CAP_SETGID)) {
687 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
688 (rgid != current->egid) && (rgid != current->sgid))
689 return -EPERM;
690 if ((egid != (gid_t) -1) && (egid != current->gid) &&
691 (egid != current->egid) && (egid != current->sgid))
692 return -EPERM;
693 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
694 (sgid != current->egid) && (sgid != current->sgid))
695 return -EPERM;
696 }
697 if (egid != (gid_t) -1) {
698 if (egid != current->egid)
699 {
700 current->mm->dumpable = 0;
701 wmb();
702 }
703 current->egid = egid;
704 current->fsgid = egid;
705 }
706 if (rgid != (gid_t) -1)
707 current->gid = rgid;
708 if (sgid != (gid_t) -1)
709 current->sgid = sgid;
710 return 0;
711 }
712
713 asmlinkage long sys_getresgid(gid_t *rgid, gid_t *egid, gid_t *sgid)
714 {
715 int retval;
716
717 if (!(retval = put_user(current->gid, rgid)) &&
718 !(retval = put_user(current->egid, egid)))
719 retval = put_user(current->sgid, sgid);
720
721 return retval;
722 }
723
724
725 /*
726 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
727 * is used for "access()" and for the NFS daemon (letting nfsd stay at
728 * whatever uid it wants to). It normally shadows "euid", except when
729 * explicitly set by setfsuid() or for access..
730 */
731 asmlinkage long sys_setfsuid(uid_t uid)
732 {
733 int old_fsuid;
734
735 old_fsuid = current->fsuid;
736 if (uid == current->uid || uid == current->euid ||
737 uid == current->suid || uid == current->fsuid ||
738 capable(CAP_SETUID))
739 {
740 if (uid != old_fsuid)
741 {
742 current->mm->dumpable = 0;
743 wmb();
744 }
745 current->fsuid = uid;
746 }
747
748 /* We emulate fsuid by essentially doing a scaled-down version
749 * of what we did in setresuid and friends. However, we only
750 * operate on the fs-specific bits of the process' effective
751 * capabilities
752 *
753 * FIXME - is fsuser used for all CAP_FS_MASK capabilities?
754 * if not, we might be a bit too harsh here.
755 */
756
757 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
758 if (old_fsuid == 0 && current->fsuid != 0) {
759 cap_t(current->cap_effective) &= ~CAP_FS_MASK;
760 }
761 if (old_fsuid != 0 && current->fsuid == 0) {
762 cap_t(current->cap_effective) |=
763 (cap_t(current->cap_permitted) & CAP_FS_MASK);
764 }
765 }
766
767 return old_fsuid;
768 }
769
770 /*
771 * Samma på svenska..
772 */
773 asmlinkage long sys_setfsgid(gid_t gid)
774 {
775 int old_fsgid;
776
777 old_fsgid = current->fsgid;
778 if (gid == current->gid || gid == current->egid ||
779 gid == current->sgid || gid == current->fsgid ||
780 capable(CAP_SETGID))
781 {
782 if (gid != old_fsgid)
783 {
784 current->mm->dumpable = 0;
785 wmb();
786 }
787 current->fsgid = gid;
788 }
789 return old_fsgid;
790 }
791
792 asmlinkage long sys_times(struct tms * tbuf)
793 {
794 /*
795 * In the SMP world we might just be unlucky and have one of
796 * the times increment as we use it. Since the value is an
797 * atomically safe type this is just fine. Conceptually its
798 * as if the syscall took an instant longer to occur.
799 */
800 if (tbuf)
801 if (copy_to_user(tbuf, ¤t->times, sizeof(struct tms)))
802 return -EFAULT;
803 return jiffies;
804 }
805
806 /*
807 * This needs some heavy checking ...
808 * I just haven't the stomach for it. I also don't fully
809 * understand sessions/pgrp etc. Let somebody who does explain it.
810 *
811 * OK, I think I have the protection semantics right.... this is really
812 * only important on a multi-user system anyway, to make sure one user
813 * can't send a signal to a process owned by another. -TYT, 12/12/91
814 *
815 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
816 * LBT 04.03.94
817 */
818
819 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
820 {
821 struct task_struct * p;
822 int err = -EINVAL;
823
824 if (!pid)
825 pid = current->pid;
826 if (!pgid)
827 pgid = pid;
828 if (pgid < 0)
829 return -EINVAL;
830
831 /* From this point forward we keep holding onto the tasklist lock
832 * so that our parent does not change from under us. -DaveM
833 */
834 read_lock(&tasklist_lock);
835
836 err = -ESRCH;
837 p = find_task_by_pid(pid);
838 if (!p)
839 goto out;
840
841 if (p->p_pptr == current || p->p_opptr == current) {
842 err = -EPERM;
843 if (p->session != current->session)
844 goto out;
845 err = -EACCES;
846 if (p->did_exec)
847 goto out;
848 } else if (p != current)
849 goto out;
850 err = -EPERM;
851 if (p->leader)
852 goto out;
853 if (pgid != pid) {
854 struct task_struct * tmp;
855 for_each_task (tmp) {
856 if (tmp->pgrp == pgid &&
857 tmp->session == current->session)
858 goto ok_pgid;
859 }
860 goto out;
861 }
862
863 ok_pgid:
864 p->pgrp = pgid;
865 err = 0;
866 out:
867 /* All paths lead to here, thus we are safe. -DaveM */
868 read_unlock(&tasklist_lock);
869 return err;
870 }
871
872 asmlinkage long sys_getpgid(pid_t pid)
873 {
874 if (!pid) {
875 return current->pgrp;
876 } else {
877 int retval;
878 struct task_struct *p;
879
880 read_lock(&tasklist_lock);
881 p = find_task_by_pid(pid);
882
883 retval = -ESRCH;
884 if (p)
885 retval = p->pgrp;
886 read_unlock(&tasklist_lock);
887 return retval;
888 }
889 }
890
891 asmlinkage long sys_getpgrp(void)
892 {
893 /* SMP - assuming writes are word atomic this is fine */
894 return current->pgrp;
895 }
896
897 asmlinkage long sys_getsid(pid_t pid)
898 {
899 if (!pid) {
900 return current->session;
901 } else {
902 int retval;
903 struct task_struct *p;
904
905 read_lock(&tasklist_lock);
906 p = find_task_by_pid(pid);
907
908 retval = -ESRCH;
909 if(p)
910 retval = p->session;
911 read_unlock(&tasklist_lock);
912 return retval;
913 }
914 }
915
916 asmlinkage long sys_setsid(void)
917 {
918 struct task_struct * p;
919 int err = -EPERM;
920
921 read_lock(&tasklist_lock);
922 for_each_task(p) {
923 if (p->pgrp == current->pid)
924 goto out;
925 }
926
927 current->leader = 1;
928 current->session = current->pgrp = current->pid;
929 current->tty = NULL;
930 current->tty_old_pgrp = 0;
931 err = current->pgrp;
932 out:
933 read_unlock(&tasklist_lock);
934 return err;
935 }
936
937 /*
938 * Supplementary group IDs
939 */
940 asmlinkage long sys_getgroups(int gidsetsize, gid_t *grouplist)
941 {
942 int i;
943
944 /*
945 * SMP: Nobody else can change our grouplist. Thus we are
946 * safe.
947 */
948
949 if (gidsetsize < 0)
950 return -EINVAL;
951 i = current->ngroups;
952 if (gidsetsize) {
953 if (i > gidsetsize)
954 return -EINVAL;
955 if (copy_to_user(grouplist, current->groups, sizeof(gid_t)*i))
956 return -EFAULT;
957 }
958 return i;
959 }
960
961 /*
962 * SMP: Our groups are not shared. We can copy to/from them safely
963 * without another task interfering.
964 */
965
966 asmlinkage long sys_setgroups(int gidsetsize, gid_t *grouplist)
967 {
968 if (!capable(CAP_SETGID))
969 return -EPERM;
970 if ((unsigned) gidsetsize > NGROUPS)
971 return -EINVAL;
972 if(copy_from_user(current->groups, grouplist, gidsetsize * sizeof(gid_t)))
973 return -EFAULT;
974 current->ngroups = gidsetsize;
975 return 0;
976 }
977
978 static int supplemental_group_member(gid_t grp)
979 {
980 int i = current->ngroups;
981
982 if (i) {
983 gid_t *groups = current->groups;
984 do {
985 if (*groups == grp)
986 return 1;
987 groups++;
988 i--;
989 } while (i);
990 }
991 return 0;
992 }
993
994 /*
995 * Check whether we're fsgid/egid or in the supplemental group..
996 */
997 int in_group_p(gid_t grp)
998 {
999 int retval = 1;
1000 if (grp != current->fsgid)
1001 retval = supplemental_group_member(grp);
1002 return retval;
1003 }
1004
1005 int in_egroup_p(gid_t grp)
1006 {
1007 int retval = 1;
1008 if (grp != current->egid)
1009 retval = supplemental_group_member(grp);
1010 return retval;
1011 }
1012
1013 DECLARE_RWSEM(uts_sem);
1014
1015 asmlinkage long sys_newuname(struct new_utsname * name)
1016 {
1017 int errno = 0;
1018
1019 down_read(&uts_sem);
1020 if (copy_to_user(name,&system_utsname,sizeof *name))
1021 errno = -EFAULT;
1022 up_read(&uts_sem);
1023 return errno;
1024 }
1025
1026 asmlinkage long sys_sethostname(char *name, int len)
1027 {
1028 int errno;
1029
1030 if (!capable(CAP_SYS_ADMIN))
1031 return -EPERM;
1032 if (len < 0 || len > __NEW_UTS_LEN)
1033 return -EINVAL;
1034 down_write(&uts_sem);
1035 errno = -EFAULT;
1036 if (!copy_from_user(system_utsname.nodename, name, len)) {
1037 system_utsname.nodename[len] = 0;
1038 errno = 0;
1039 }
1040 up_write(&uts_sem);
1041 return errno;
1042 }
1043
1044 asmlinkage long sys_gethostname(char *name, int len)
1045 {
1046 int i, errno;
1047
1048 if (len < 0)
1049 return -EINVAL;
1050 down_read(&uts_sem);
1051 i = 1 + strlen(system_utsname.nodename);
1052 if (i > len)
1053 i = len;
1054 errno = 0;
1055 if (copy_to_user(name, system_utsname.nodename, i))
1056 errno = -EFAULT;
1057 up_read(&uts_sem);
1058 return errno;
1059 }
1060
1061 /*
1062 * Only setdomainname; getdomainname can be implemented by calling
1063 * uname()
1064 */
1065 asmlinkage long sys_setdomainname(char *name, int len)
1066 {
1067 int errno;
1068
1069 if (!capable(CAP_SYS_ADMIN))
1070 return -EPERM;
1071 if (len < 0 || len > __NEW_UTS_LEN)
1072 return -EINVAL;
1073
1074 down_write(&uts_sem);
1075 errno = -EFAULT;
1076 if (!copy_from_user(system_utsname.domainname, name, len)) {
1077 errno = 0;
1078 system_utsname.domainname[len] = 0;
1079 }
1080 up_write(&uts_sem);
1081 return errno;
1082 }
1083
1084 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit *rlim)
1085 {
1086 if (resource >= RLIM_NLIMITS)
1087 return -EINVAL;
1088 else
1089 return copy_to_user(rlim, current->rlim + resource, sizeof(*rlim))
1090 ? -EFAULT : 0;
1091 }
1092
1093 #if !defined(__ia64__)
1094
1095 /*
1096 * Back compatibility for getrlimit. Needed for some apps.
1097 */
1098
1099 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit *rlim)
1100 {
1101 struct rlimit x;
1102 if (resource >= RLIM_NLIMITS)
1103 return -EINVAL;
1104
1105 memcpy(&x, current->rlim + resource, sizeof(*rlim));
1106 if(x.rlim_cur > 0x7FFFFFFF)
1107 x.rlim_cur = 0x7FFFFFFF;
1108 if(x.rlim_max > 0x7FFFFFFF)
1109 x.rlim_max = 0x7FFFFFFF;
1110 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1111 }
1112
1113 #endif
1114
1115 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit *rlim)
1116 {
1117 struct rlimit new_rlim, *old_rlim;
1118
1119 if (resource >= RLIM_NLIMITS)
1120 return -EINVAL;
1121 if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1122 return -EFAULT;
1123 old_rlim = current->rlim + resource;
1124 if (((new_rlim.rlim_cur > old_rlim->rlim_max) ||
1125 (new_rlim.rlim_max > old_rlim->rlim_max)) &&
1126 !capable(CAP_SYS_RESOURCE))
1127 return -EPERM;
1128 if (resource == RLIMIT_NOFILE) {
1129 if (new_rlim.rlim_cur > NR_OPEN || new_rlim.rlim_max > NR_OPEN)
1130 return -EPERM;
1131 }
1132 *old_rlim = new_rlim;
1133 return 0;
1134 }
1135
1136 /*
1137 * It would make sense to put struct rusage in the task_struct,
1138 * except that would make the task_struct be *really big*. After
1139 * task_struct gets moved into malloc'ed memory, it would
1140 * make sense to do this. It will make moving the rest of the information
1141 * a lot simpler! (Which we're not doing right now because we're not
1142 * measuring them yet).
1143 *
1144 * This is SMP safe. Either we are called from sys_getrusage on ourselves
1145 * below (we know we aren't going to exit/disappear and only we change our
1146 * rusage counters), or we are called from wait4() on a process which is
1147 * either stopped or zombied. In the zombied case the task won't get
1148 * reaped till shortly after the call to getrusage(), in both cases the
1149 * task being examined is in a frozen state so the counters won't change.
1150 *
1151 * FIXME! Get the fault counts properly!
1152 */
1153 int getrusage(struct task_struct *p, int who, struct rusage *ru)
1154 {
1155 struct rusage r;
1156
1157 memset((char *) &r, 0, sizeof(r));
1158 switch (who) {
1159 case RUSAGE_SELF:
1160 r.ru_utime.tv_sec = CT_TO_SECS(p->times.tms_utime);
1161 r.ru_utime.tv_usec = CT_TO_USECS(p->times.tms_utime);
1162 r.ru_stime.tv_sec = CT_TO_SECS(p->times.tms_stime);
1163 r.ru_stime.tv_usec = CT_TO_USECS(p->times.tms_stime);
1164 r.ru_minflt = p->min_flt;
1165 r.ru_majflt = p->maj_flt;
1166 r.ru_nswap = p->nswap;
1167 break;
1168 case RUSAGE_CHILDREN:
1169 r.ru_utime.tv_sec = CT_TO_SECS(p->times.tms_cutime);
1170 r.ru_utime.tv_usec = CT_TO_USECS(p->times.tms_cutime);
1171 r.ru_stime.tv_sec = CT_TO_SECS(p->times.tms_cstime);
1172 r.ru_stime.tv_usec = CT_TO_USECS(p->times.tms_cstime);
1173 r.ru_minflt = p->cmin_flt;
1174 r.ru_majflt = p->cmaj_flt;
1175 r.ru_nswap = p->cnswap;
1176 break;
1177 default:
1178 r.ru_utime.tv_sec = CT_TO_SECS(p->times.tms_utime + p->times.tms_cutime);
1179 r.ru_utime.tv_usec = CT_TO_USECS(p->times.tms_utime + p->times.tms_cutime);
1180 r.ru_stime.tv_sec = CT_TO_SECS(p->times.tms_stime + p->times.tms_cstime);
1181 r.ru_stime.tv_usec = CT_TO_USECS(p->times.tms_stime + p->times.tms_cstime);
1182 r.ru_minflt = p->min_flt + p->cmin_flt;
1183 r.ru_majflt = p->maj_flt + p->cmaj_flt;
1184 r.ru_nswap = p->nswap + p->cnswap;
1185 break;
1186 }
1187 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1188 }
1189
1190 asmlinkage long sys_getrusage(int who, struct rusage *ru)
1191 {
1192 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1193 return -EINVAL;
1194 return getrusage(current, who, ru);
1195 }
1196
1197 asmlinkage long sys_umask(int mask)
1198 {
1199 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1200 return mask;
1201 }
1202
1203 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1204 unsigned long arg4, unsigned long arg5)
1205 {
1206 int error = 0;
1207 int sig;
1208
1209 switch (option) {
1210 case PR_SET_PDEATHSIG:
1211 sig = arg2;
1212 if (sig < 0 || sig > _NSIG) {
1213 error = -EINVAL;
1214 break;
1215 }
1216 current->pdeath_signal = sig;
1217 break;
1218 case PR_GET_PDEATHSIG:
1219 error = put_user(current->pdeath_signal, (int *)arg2);
1220 break;
1221 case PR_GET_DUMPABLE:
1222 if (current->mm->dumpable)
1223 error = 1;
1224 break;
1225 case PR_SET_DUMPABLE:
1226 if (arg2 != 0 && arg2 != 1) {
1227 error = -EINVAL;
1228 break;
1229 }
1230 current->mm->dumpable = arg2;
1231 break;
1232 case PR_SET_UNALIGN:
1233 #ifdef SET_UNALIGN_CTL
1234 error = SET_UNALIGN_CTL(current, arg2);
1235 #else
1236 error = -EINVAL;
1237 #endif
1238 break;
1239
1240 case PR_GET_UNALIGN:
1241 #ifdef GET_UNALIGN_CTL
1242 error = GET_UNALIGN_CTL(current, arg2);
1243 #else
1244 error = -EINVAL;
1245 #endif
1246 break;
1247
1248 case PR_GET_KEEPCAPS:
1249 if (current->keep_capabilities)
1250 error = 1;
1251 break;
1252 case PR_SET_KEEPCAPS:
1253 if (arg2 != 0 && arg2 != 1) {
1254 error = -EINVAL;
1255 break;
1256 }
1257 current->keep_capabilities = arg2;
1258 break;
1259 default:
1260 error = -EINVAL;
1261 break;
1262 }
1263 return error;
1264 }
1265
1266 EXPORT_SYMBOL(notifier_chain_register);
1267 EXPORT_SYMBOL(notifier_chain_unregister);
1268 EXPORT_SYMBOL(notifier_call_chain);
1269 EXPORT_SYMBOL(register_reboot_notifier);
1270 EXPORT_SYMBOL(unregister_reboot_notifier);
1271 EXPORT_SYMBOL(in_group_p);
1272 EXPORT_SYMBOL(in_egroup_p);
1273