?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb-2.6
[linux-2.6/kmemtrace.git] / kernel / sys.c
bloba626116af5db96b58f47434eb47c6af23da6535e
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/highuid.h>
16 #include <linux/fs.h>
17 #include <linux/resource.h>
18 #include <linux/kernel.h>
19 #include <linux/kexec.h>
20 #include <linux/workqueue.h>
21 #include <linux/capability.h>
22 #include <linux/device.h>
23 #include <linux/key.h>
24 #include <linux/times.h>
25 #include <linux/posix-timers.h>
26 #include <linux/security.h>
27 #include <linux/dcookies.h>
28 #include <linux/suspend.h>
29 #include <linux/tty.h>
30 #include <linux/signal.h>
31 #include <linux/cn_proc.h>
32 #include <linux/getcpu.h>
33 #include <linux/task_io_accounting_ops.h>
34 #include <linux/seccomp.h>
35 #include <linux/cpu.h>
36
37 #include <linux/compat.h>
38 #include <linux/syscalls.h>
39 #include <linux/kprobes.h>
40 #include <linux/user_namespace.h>
41
42 #include <asm/uaccess.h>
43 #include <asm/io.h>
44 #include <asm/unistd.h>
45
46 #ifndef SET_UNALIGN_CTL
47 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
48 #endif
49 #ifndef GET_UNALIGN_CTL
50 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
51 #endif
52 #ifndef SET_FPEMU_CTL
53 # define SET_FPEMU_CTL(a,b) (-EINVAL)
54 #endif
55 #ifndef GET_FPEMU_CTL
56 # define GET_FPEMU_CTL(a,b) (-EINVAL)
57 #endif
58 #ifndef SET_FPEXC_CTL
59 # define SET_FPEXC_CTL(a,b) (-EINVAL)
60 #endif
61 #ifndef GET_FPEXC_CTL
62 # define GET_FPEXC_CTL(a,b) (-EINVAL)
63 #endif
64 #ifndef GET_ENDIAN
65 # define GET_ENDIAN(a,b) (-EINVAL)
66 #endif
67 #ifndef SET_ENDIAN
68 # define SET_ENDIAN(a,b) (-EINVAL)
69 #endif
70
71 /*
72 * this is where the system-wide overflow UID and GID are defined, for
73 * architectures that now have 32-bit UID/GID but didn't in the past
74 */
75
76 int overflowuid = DEFAULT_OVERFLOWUID;
77 int overflowgid = DEFAULT_OVERFLOWGID;
78
79 #ifdef CONFIG_UID16
80 EXPORT_SYMBOL(overflowuid);
81 EXPORT_SYMBOL(overflowgid);
82 #endif
83
84 /*
85 * the same as above, but for filesystems which can only store a 16-bit
86 * UID and GID. as such, this is needed on all architectures
87 */
88
89 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
90 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
91
92 EXPORT_SYMBOL(fs_overflowuid);
93 EXPORT_SYMBOL(fs_overflowgid);
94
95 /*
96 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
97 */
98
99 int C_A_D = 1;
100 struct pid *cad_pid;
101 EXPORT_SYMBOL(cad_pid);
102
103 /*
104 * If set, this is used for preparing the system to power off.
105 */
106
107 void (*pm_power_off_prepare)(void);
108
109 static int set_one_prio(struct task_struct *p, int niceval, int error)
110 {
111 int no_nice;
112
113 if (p->uid != current->euid &&
114 p->euid != current->euid && !capable(CAP_SYS_NICE)) {
115 error = -EPERM;
116 goto out;
117 }
118 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
119 error = -EACCES;
120 goto out;
121 }
122 no_nice = security_task_setnice(p, niceval);
123 if (no_nice) {
124 error = no_nice;
125 goto out;
126 }
127 if (error == -ESRCH)
128 error = 0;
129 set_user_nice(p, niceval);
130 out:
131 return error;
132 }
133
134 asmlinkage long sys_setpriority(int which, int who, int niceval)
135 {
136 struct task_struct *g, *p;
137 struct user_struct *user;
138 int error = -EINVAL;
139 struct pid *pgrp;
140
141 if (which > PRIO_USER || which < PRIO_PROCESS)
142 goto out;
143
144 /* normalize: avoid signed division (rounding problems) */
145 error = -ESRCH;
146 if (niceval < -20)
147 niceval = -20;
148 if (niceval > 19)
149 niceval = 19;
150
151 read_lock(&tasklist_lock);
152 switch (which) {
153 case PRIO_PROCESS:
154 if (who)
155 p = find_task_by_vpid(who);
156 else
157 p = current;
158 if (p)
159 error = set_one_prio(p, niceval, error);
160 break;
161 case PRIO_PGRP:
162 if (who)
163 pgrp = find_vpid(who);
164 else
165 pgrp = task_pgrp(current);
166 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
167 error = set_one_prio(p, niceval, error);
168 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
169 break;
170 case PRIO_USER:
171 user = current->user;
172 if (!who)
173 who = current->uid;
174 else
175 if ((who != current->uid) && !(user = find_user(who)))
176 goto out_unlock; /* No processes for this user */
177
178 do_each_thread(g, p)
179 if (p->uid == who)
180 error = set_one_prio(p, niceval, error);
181 while_each_thread(g, p);
182 if (who != current->uid)
183 free_uid(user); /* For find_user() */
184 break;
185 }
186 out_unlock:
187 read_unlock(&tasklist_lock);
188 out:
189 return error;
190 }
191
192 /*
193 * Ugh. To avoid negative return values, "getpriority()" will
194 * not return the normal nice-value, but a negated value that
195 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
196 * to stay compatible.
197 */
198 asmlinkage long sys_getpriority(int which, int who)
199 {
200 struct task_struct *g, *p;
201 struct user_struct *user;
202 long niceval, retval = -ESRCH;
203 struct pid *pgrp;
204
205 if (which > PRIO_USER || which < PRIO_PROCESS)
206 return -EINVAL;
207
208 read_lock(&tasklist_lock);
209 switch (which) {
210 case PRIO_PROCESS:
211 if (who)
212 p = find_task_by_vpid(who);
213 else
214 p = current;
215 if (p) {
216 niceval = 20 - task_nice(p);
217 if (niceval > retval)
218 retval = niceval;
219 }
220 break;
221 case PRIO_PGRP:
222 if (who)
223 pgrp = find_vpid(who);
224 else
225 pgrp = task_pgrp(current);
226 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
227 niceval = 20 - task_nice(p);
228 if (niceval > retval)
229 retval = niceval;
230 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
231 break;
232 case PRIO_USER:
233 user = current->user;
234 if (!who)
235 who = current->uid;
236 else
237 if ((who != current->uid) && !(user = find_user(who)))
238 goto out_unlock; /* No processes for this user */
239
240 do_each_thread(g, p)
241 if (p->uid == who) {
242 niceval = 20 - task_nice(p);
243 if (niceval > retval)
244 retval = niceval;
245 }
246 while_each_thread(g, p);
247 if (who != current->uid)
248 free_uid(user); /* for find_user() */
249 break;
250 }
251 out_unlock:
252 read_unlock(&tasklist_lock);
253
254 return retval;
255 }
256
257 /**
258 * emergency_restart - reboot the system
259 *
260 * Without shutting down any hardware or taking any locks
261 * reboot the system. This is called when we know we are in
262 * trouble so this is our best effort to reboot. This is
263 * safe to call in interrupt context.
264 */
265 void emergency_restart(void)
266 {
267 machine_emergency_restart();
268 }
269 EXPORT_SYMBOL_GPL(emergency_restart);
270
271 static void kernel_restart_prepare(char *cmd)
272 {
273 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
274 system_state = SYSTEM_RESTART;
275 device_shutdown();
276 sysdev_shutdown();
277 }
278
279 /**
280 * kernel_restart - reboot the system
281 * @cmd: pointer to buffer containing command to execute for restart
282 * or %NULL
283 *
284 * Shutdown everything and perform a clean reboot.
285 * This is not safe to call in interrupt context.
286 */
287 void kernel_restart(char *cmd)
288 {
289 kernel_restart_prepare(cmd);
290 if (!cmd)
291 printk(KERN_EMERG "Restarting system.\n");
292 else
293 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
294 machine_restart(cmd);
295 }
296 EXPORT_SYMBOL_GPL(kernel_restart);
297
298 /**
299 * kernel_kexec - reboot the system
300 *
301 * Move into place and start executing a preloaded standalone
302 * executable. If nothing was preloaded return an error.
303 */
304 static void kernel_kexec(void)
305 {
306 #ifdef CONFIG_KEXEC
307 struct kimage *image;
308 image = xchg(&kexec_image, NULL);
309 if (!image)
310 return;
311 kernel_restart_prepare(NULL);
312 printk(KERN_EMERG "Starting new kernel\n");
313 machine_shutdown();
314 machine_kexec(image);
315 #endif
316 }
317
318 static void kernel_shutdown_prepare(enum system_states state)
319 {
320 blocking_notifier_call_chain(&reboot_notifier_list,
321 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
322 system_state = state;
323 device_shutdown();
324 }
325 /**
326 * kernel_halt - halt the system
327 *
328 * Shutdown everything and perform a clean system halt.
329 */
330 void kernel_halt(void)
331 {
332 kernel_shutdown_prepare(SYSTEM_HALT);
333 sysdev_shutdown();
334 printk(KERN_EMERG "System halted.\n");
335 machine_halt();
336 }
337
338 EXPORT_SYMBOL_GPL(kernel_halt);
339
340 /**
341 * kernel_power_off - power_off the system
342 *
343 * Shutdown everything and perform a clean system power_off.
344 */
345 void kernel_power_off(void)
346 {
347 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
348 if (pm_power_off_prepare)
349 pm_power_off_prepare();
350 disable_nonboot_cpus();
351 sysdev_shutdown();
352 printk(KERN_EMERG "Power down.\n");
353 machine_power_off();
354 }
355 EXPORT_SYMBOL_GPL(kernel_power_off);
356 /*
357 * Reboot system call: for obvious reasons only root may call it,
358 * and even root needs to set up some magic numbers in the registers
359 * so that some mistake won't make this reboot the whole machine.
360 * You can also set the meaning of the ctrl-alt-del-key here.
361 *
362 * reboot doesn't sync: do that yourself before calling this.
363 */
364 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
365 {
366 char buffer[256];
367
368 /* We only trust the superuser with rebooting the system. */
369 if (!capable(CAP_SYS_BOOT))
370 return -EPERM;
371
372 /* For safety, we require "magic" arguments. */
373 if (magic1 != LINUX_REBOOT_MAGIC1 ||
374 (magic2 != LINUX_REBOOT_MAGIC2 &&
375 magic2 != LINUX_REBOOT_MAGIC2A &&
376 magic2 != LINUX_REBOOT_MAGIC2B &&
377 magic2 != LINUX_REBOOT_MAGIC2C))
378 return -EINVAL;
379
380 /* Instead of trying to make the power_off code look like
381 * halt when pm_power_off is not set do it the easy way.
382 */
383 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
384 cmd = LINUX_REBOOT_CMD_HALT;
385
386 lock_kernel();
387 switch (cmd) {
388 case LINUX_REBOOT_CMD_RESTART:
389 kernel_restart(NULL);
390 break;
391
392 case LINUX_REBOOT_CMD_CAD_ON:
393 C_A_D = 1;
394 break;
395
396 case LINUX_REBOOT_CMD_CAD_OFF:
397 C_A_D = 0;
398 break;
399
400 case LINUX_REBOOT_CMD_HALT:
401 kernel_halt();
402 unlock_kernel();
403 do_exit(0);
404 break;
405
406 case LINUX_REBOOT_CMD_POWER_OFF:
407 kernel_power_off();
408 unlock_kernel();
409 do_exit(0);
410 break;
411
412 case LINUX_REBOOT_CMD_RESTART2:
413 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
414 unlock_kernel();
415 return -EFAULT;
416 }
417 buffer[sizeof(buffer) - 1] = '\0';
418
419 kernel_restart(buffer);
420 break;
421
422 case LINUX_REBOOT_CMD_KEXEC:
423 kernel_kexec();
424 unlock_kernel();
425 return -EINVAL;
426
427 #ifdef CONFIG_HIBERNATION
428 case LINUX_REBOOT_CMD_SW_SUSPEND:
429 {
430 int ret = hibernate();
431 unlock_kernel();
432 return ret;
433 }
434 #endif
435
436 default:
437 unlock_kernel();
438 return -EINVAL;
439 }
440 unlock_kernel();
441 return 0;
442 }
443
444 static void deferred_cad(struct work_struct *dummy)
445 {
446 kernel_restart(NULL);
447 }
448
449 /*
450 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
451 * As it's called within an interrupt, it may NOT sync: the only choice
452 * is whether to reboot at once, or just ignore the ctrl-alt-del.
453 */
454 void ctrl_alt_del(void)
455 {
456 static DECLARE_WORK(cad_work, deferred_cad);
457
458 if (C_A_D)
459 schedule_work(&cad_work);
460 else
461 kill_cad_pid(SIGINT, 1);
462 }
463
464 /*
465 * Unprivileged users may change the real gid to the effective gid
466 * or vice versa. (BSD-style)
467 *
468 * If you set the real gid at all, or set the effective gid to a value not
469 * equal to the real gid, then the saved gid is set to the new effective gid.
470 *
471 * This makes it possible for a setgid program to completely drop its
472 * privileges, which is often a useful assertion to make when you are doing
473 * a security audit over a program.
474 *
475 * The general idea is that a program which uses just setregid() will be
476 * 100% compatible with BSD. A program which uses just setgid() will be
477 * 100% compatible with POSIX with saved IDs.
478 *
479 * SMP: There are not races, the GIDs are checked only by filesystem
480 * operations (as far as semantic preservation is concerned).
481 */
482 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
483 {
484 int old_rgid = current->gid;
485 int old_egid = current->egid;
486 int new_rgid = old_rgid;
487 int new_egid = old_egid;
488 int retval;
489
490 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
491 if (retval)
492 return retval;
493
494 if (rgid != (gid_t) -1) {
495 if ((old_rgid == rgid) ||
496 (current->egid==rgid) ||
497 capable(CAP_SETGID))
498 new_rgid = rgid;
499 else
500 return -EPERM;
501 }
502 if (egid != (gid_t) -1) {
503 if ((old_rgid == egid) ||
504 (current->egid == egid) ||
505 (current->sgid == egid) ||
506 capable(CAP_SETGID))
507 new_egid = egid;
508 else
509 return -EPERM;
510 }
511 if (new_egid != old_egid) {
512 set_dumpable(current->mm, suid_dumpable);
513 smp_wmb();
514 }
515 if (rgid != (gid_t) -1 ||
516 (egid != (gid_t) -1 && egid != old_rgid))
517 current->sgid = new_egid;
518 current->fsgid = new_egid;
519 current->egid = new_egid;
520 current->gid = new_rgid;
521 key_fsgid_changed(current);
522 proc_id_connector(current, PROC_EVENT_GID);
523 return 0;
524 }
525
526 /*
527 * setgid() is implemented like SysV w/ SAVED_IDS
528 *
529 * SMP: Same implicit races as above.
530 */
531 asmlinkage long sys_setgid(gid_t gid)
532 {
533 int old_egid = current->egid;
534 int retval;
535
536 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
537 if (retval)
538 return retval;
539
540 if (capable(CAP_SETGID)) {
541 if (old_egid != gid) {
542 set_dumpable(current->mm, suid_dumpable);
543 smp_wmb();
544 }
545 current->gid = current->egid = current->sgid = current->fsgid = gid;
546 } else if ((gid == current->gid) || (gid == current->sgid)) {
547 if (old_egid != gid) {
548 set_dumpable(current->mm, suid_dumpable);
549 smp_wmb();
550 }
551 current->egid = current->fsgid = gid;
552 }
553 else
554 return -EPERM;
555
556 key_fsgid_changed(current);
557 proc_id_connector(current, PROC_EVENT_GID);
558 return 0;
559 }
560
561 static int set_user(uid_t new_ruid, int dumpclear)
562 {
563 struct user_struct *new_user;
564
565 new_user = alloc_uid(current->nsproxy->user_ns, new_ruid);
566 if (!new_user)
567 return -EAGAIN;
568
569 if (atomic_read(&new_user->processes) >=
570 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
571 new_user != current->nsproxy->user_ns->root_user) {
572 free_uid(new_user);
573 return -EAGAIN;
574 }
575
576 switch_uid(new_user);
577
578 if (dumpclear) {
579 set_dumpable(current->mm, suid_dumpable);
580 smp_wmb();
581 }
582 current->uid = new_ruid;
583 return 0;
584 }
585
586 /*
587 * Unprivileged users may change the real uid to the effective uid
588 * or vice versa. (BSD-style)
589 *
590 * If you set the real uid at all, or set the effective uid to a value not
591 * equal to the real uid, then the saved uid is set to the new effective uid.
592 *
593 * This makes it possible for a setuid program to completely drop its
594 * privileges, which is often a useful assertion to make when you are doing
595 * a security audit over a program.
596 *
597 * The general idea is that a program which uses just setreuid() will be
598 * 100% compatible with BSD. A program which uses just setuid() will be
599 * 100% compatible with POSIX with saved IDs.
600 */
601 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
602 {
603 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
604 int retval;
605
606 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
607 if (retval)
608 return retval;
609
610 new_ruid = old_ruid = current->uid;
611 new_euid = old_euid = current->euid;
612 old_suid = current->suid;
613
614 if (ruid != (uid_t) -1) {
615 new_ruid = ruid;
616 if ((old_ruid != ruid) &&
617 (current->euid != ruid) &&
618 !capable(CAP_SETUID))
619 return -EPERM;
620 }
621
622 if (euid != (uid_t) -1) {
623 new_euid = euid;
624 if ((old_ruid != euid) &&
625 (current->euid != euid) &&
626 (current->suid != euid) &&
627 !capable(CAP_SETUID))
628 return -EPERM;
629 }
630
631 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
632 return -EAGAIN;
633
634 if (new_euid != old_euid) {
635 set_dumpable(current->mm, suid_dumpable);
636 smp_wmb();
637 }
638 current->fsuid = current->euid = new_euid;
639 if (ruid != (uid_t) -1 ||
640 (euid != (uid_t) -1 && euid != old_ruid))
641 current->suid = current->euid;
642 current->fsuid = current->euid;
643
644 key_fsuid_changed(current);
645 proc_id_connector(current, PROC_EVENT_UID);
646
647 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
648 }
649
650
651
652 /*
653 * setuid() is implemented like SysV with SAVED_IDS
654 *
655 * Note that SAVED_ID's is deficient in that a setuid root program
656 * like sendmail, for example, cannot set its uid to be a normal
657 * user and then switch back, because if you're root, setuid() sets
658 * the saved uid too. If you don't like this, blame the bright people
659 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
660 * will allow a root program to temporarily drop privileges and be able to
661 * regain them by swapping the real and effective uid.
662 */
663 asmlinkage long sys_setuid(uid_t uid)
664 {
665 int old_euid = current->euid;
666 int old_ruid, old_suid, new_suid;
667 int retval;
668
669 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
670 if (retval)
671 return retval;
672
673 old_ruid = current->uid;
674 old_suid = current->suid;
675 new_suid = old_suid;
676
677 if (capable(CAP_SETUID)) {
678 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
679 return -EAGAIN;
680 new_suid = uid;
681 } else if ((uid != current->uid) && (uid != new_suid))
682 return -EPERM;
683
684 if (old_euid != uid) {
685 set_dumpable(current->mm, suid_dumpable);
686 smp_wmb();
687 }
688 current->fsuid = current->euid = uid;
689 current->suid = new_suid;
690
691 key_fsuid_changed(current);
692 proc_id_connector(current, PROC_EVENT_UID);
693
694 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
695 }
696
697
698 /*
699 * This function implements a generic ability to update ruid, euid,
700 * and suid. This allows you to implement the 4.4 compatible seteuid().
701 */
702 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
703 {
704 int old_ruid = current->uid;
705 int old_euid = current->euid;
706 int old_suid = current->suid;
707 int retval;
708
709 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
710 if (retval)
711 return retval;
712
713 if (!capable(CAP_SETUID)) {
714 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
715 (ruid != current->euid) && (ruid != current->suid))
716 return -EPERM;
717 if ((euid != (uid_t) -1) && (euid != current->uid) &&
718 (euid != current->euid) && (euid != current->suid))
719 return -EPERM;
720 if ((suid != (uid_t) -1) && (suid != current->uid) &&
721 (suid != current->euid) && (suid != current->suid))
722 return -EPERM;
723 }
724 if (ruid != (uid_t) -1) {
725 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
726 return -EAGAIN;
727 }
728 if (euid != (uid_t) -1) {
729 if (euid != current->euid) {
730 set_dumpable(current->mm, suid_dumpable);
731 smp_wmb();
732 }
733 current->euid = euid;
734 }
735 current->fsuid = current->euid;
736 if (suid != (uid_t) -1)
737 current->suid = suid;
738
739 key_fsuid_changed(current);
740 proc_id_connector(current, PROC_EVENT_UID);
741
742 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
743 }
744
745 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
746 {
747 int retval;
748
749 if (!(retval = put_user(current->uid, ruid)) &&
750 !(retval = put_user(current->euid, euid)))
751 retval = put_user(current->suid, suid);
752
753 return retval;
754 }
755
756 /*
757 * Same as above, but for rgid, egid, sgid.
758 */
759 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
760 {
761 int retval;
762
763 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
764 if (retval)
765 return retval;
766
767 if (!capable(CAP_SETGID)) {
768 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
769 (rgid != current->egid) && (rgid != current->sgid))
770 return -EPERM;
771 if ((egid != (gid_t) -1) && (egid != current->gid) &&
772 (egid != current->egid) && (egid != current->sgid))
773 return -EPERM;
774 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
775 (sgid != current->egid) && (sgid != current->sgid))
776 return -EPERM;
777 }
778 if (egid != (gid_t) -1) {
779 if (egid != current->egid) {
780 set_dumpable(current->mm, suid_dumpable);
781 smp_wmb();
782 }
783 current->egid = egid;
784 }
785 current->fsgid = current->egid;
786 if (rgid != (gid_t) -1)
787 current->gid = rgid;
788 if (sgid != (gid_t) -1)
789 current->sgid = sgid;
790
791 key_fsgid_changed(current);
792 proc_id_connector(current, PROC_EVENT_GID);
793 return 0;
794 }
795
796 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
797 {
798 int retval;
799
800 if (!(retval = put_user(current->gid, rgid)) &&
801 !(retval = put_user(current->egid, egid)))
802 retval = put_user(current->sgid, sgid);
803
804 return retval;
805 }
806
807
808 /*
809 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
810 * is used for "access()" and for the NFS daemon (letting nfsd stay at
811 * whatever uid it wants to). It normally shadows "euid", except when
812 * explicitly set by setfsuid() or for access..
813 */
814 asmlinkage long sys_setfsuid(uid_t uid)
815 {
816 int old_fsuid;
817
818 old_fsuid = current->fsuid;
819 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
820 return old_fsuid;
821
822 if (uid == current->uid || uid == current->euid ||
823 uid == current->suid || uid == current->fsuid ||
824 capable(CAP_SETUID)) {
825 if (uid != old_fsuid) {
826 set_dumpable(current->mm, suid_dumpable);
827 smp_wmb();
828 }
829 current->fsuid = uid;
830 }
831
832 key_fsuid_changed(current);
833 proc_id_connector(current, PROC_EVENT_UID);
834
835 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
836
837 return old_fsuid;
838 }
839
840 /*
841 * Samma på svenska..
842 */
843 asmlinkage long sys_setfsgid(gid_t gid)
844 {
845 int old_fsgid;
846
847 old_fsgid = current->fsgid;
848 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
849 return old_fsgid;
850
851 if (gid == current->gid || gid == current->egid ||
852 gid == current->sgid || gid == current->fsgid ||
853 capable(CAP_SETGID)) {
854 if (gid != old_fsgid) {
855 set_dumpable(current->mm, suid_dumpable);
856 smp_wmb();
857 }
858 current->fsgid = gid;
859 key_fsgid_changed(current);
860 proc_id_connector(current, PROC_EVENT_GID);
861 }
862 return old_fsgid;
863 }
864
865 asmlinkage long sys_times(struct tms __user * tbuf)
866 {
867 /*
868 * In the SMP world we might just be unlucky and have one of
869 * the times increment as we use it. Since the value is an
870 * atomically safe type this is just fine. Conceptually its
871 * as if the syscall took an instant longer to occur.
872 */
873 if (tbuf) {
874 struct tms tmp;
875 struct task_struct *tsk = current;
876 struct task_struct *t;
877 cputime_t utime, stime, cutime, cstime;
878
879 spin_lock_irq(&tsk->sighand->siglock);
880 utime = tsk->signal->utime;
881 stime = tsk->signal->stime;
882 t = tsk;
883 do {
884 utime = cputime_add(utime, t->utime);
885 stime = cputime_add(stime, t->stime);
886 t = next_thread(t);
887 } while (t != tsk);
888
889 cutime = tsk->signal->cutime;
890 cstime = tsk->signal->cstime;
891 spin_unlock_irq(&tsk->sighand->siglock);
892
893 tmp.tms_utime = cputime_to_clock_t(utime);
894 tmp.tms_stime = cputime_to_clock_t(stime);
895 tmp.tms_cutime = cputime_to_clock_t(cutime);
896 tmp.tms_cstime = cputime_to_clock_t(cstime);
897 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
898 return -EFAULT;
899 }
900 return (long) jiffies_64_to_clock_t(get_jiffies_64());
901 }
902
903 /*
904 * This needs some heavy checking ...
905 * I just haven't the stomach for it. I also don't fully
906 * understand sessions/pgrp etc. Let somebody who does explain it.
907 *
908 * OK, I think I have the protection semantics right.... this is really
909 * only important on a multi-user system anyway, to make sure one user
910 * can't send a signal to a process owned by another. -TYT, 12/12/91
911 *
912 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
913 * LBT 04.03.94
914 */
915 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
916 {
917 struct task_struct *p;
918 struct task_struct *group_leader = current->group_leader;
919 struct pid *pgrp;
920 int err;
921
922 if (!pid)
923 pid = task_pid_vnr(group_leader);
924 if (!pgid)
925 pgid = pid;
926 if (pgid < 0)
927 return -EINVAL;
928
929 /* From this point forward we keep holding onto the tasklist lock
930 * so that our parent does not change from under us. -DaveM
931 */
932 write_lock_irq(&tasklist_lock);
933
934 err = -ESRCH;
935 p = find_task_by_vpid(pid);
936 if (!p)
937 goto out;
938
939 err = -EINVAL;
940 if (!thread_group_leader(p))
941 goto out;
942
943 if (same_thread_group(p->real_parent, group_leader)) {
944 err = -EPERM;
945 if (task_session(p) != task_session(group_leader))
946 goto out;
947 err = -EACCES;
948 if (p->did_exec)
949 goto out;
950 } else {
951 err = -ESRCH;
952 if (p != group_leader)
953 goto out;
954 }
955
956 err = -EPERM;
957 if (p->signal->leader)
958 goto out;
959
960 pgrp = task_pid(p);
961 if (pgid != pid) {
962 struct task_struct *g;
963
964 pgrp = find_vpid(pgid);
965 g = pid_task(pgrp, PIDTYPE_PGID);
966 if (!g || task_session(g) != task_session(group_leader))
967 goto out;
968 }
969
970 err = security_task_setpgid(p, pgid);
971 if (err)
972 goto out;
973
974 if (task_pgrp(p) != pgrp) {
975 detach_pid(p, PIDTYPE_PGID);
976 attach_pid(p, PIDTYPE_PGID, pgrp);
977 set_task_pgrp(p, pid_nr(pgrp));
978 }
979
980 err = 0;
981 out:
982 /* All paths lead to here, thus we are safe. -DaveM */
983 write_unlock_irq(&tasklist_lock);
984 return err;
985 }
986
987 asmlinkage long sys_getpgid(pid_t pid)
988 {
989 if (!pid)
990 return task_pgrp_vnr(current);
991 else {
992 int retval;
993 struct task_struct *p;
994
995 read_lock(&tasklist_lock);
996 p = find_task_by_vpid(pid);
997 retval = -ESRCH;
998 if (p) {
999 retval = security_task_getpgid(p);
1000 if (!retval)
1001 retval = task_pgrp_vnr(p);
1002 }
1003 read_unlock(&tasklist_lock);
1004 return retval;
1005 }
1006 }
1007
1008 #ifdef __ARCH_WANT_SYS_GETPGRP
1009
1010 asmlinkage long sys_getpgrp(void)
1011 {
1012 /* SMP - assuming writes are word atomic this is fine */
1013 return task_pgrp_vnr(current);
1014 }
1015
1016 #endif
1017
1018 asmlinkage long sys_getsid(pid_t pid)
1019 {
1020 if (!pid)
1021 return task_session_vnr(current);
1022 else {
1023 int retval;
1024 struct task_struct *p;
1025
1026 rcu_read_lock();
1027 p = find_task_by_vpid(pid);
1028 retval = -ESRCH;
1029 if (p) {
1030 retval = security_task_getsid(p);
1031 if (!retval)
1032 retval = task_session_vnr(p);
1033 }
1034 rcu_read_unlock();
1035 return retval;
1036 }
1037 }
1038
1039 asmlinkage long sys_setsid(void)
1040 {
1041 struct task_struct *group_leader = current->group_leader;
1042 struct pid *sid = task_pid(group_leader);
1043 pid_t session = pid_vnr(sid);
1044 int err = -EPERM;
1045
1046 write_lock_irq(&tasklist_lock);
1047 /* Fail if I am already a session leader */
1048 if (group_leader->signal->leader)
1049 goto out;
1050
1051 /* Fail if a process group id already exists that equals the
1052 * proposed session id.
1053 */
1054 if (pid_task(sid, PIDTYPE_PGID))
1055 goto out;
1056
1057 group_leader->signal->leader = 1;
1058 __set_special_pids(sid);
1059
1060 spin_lock(&group_leader->sighand->siglock);
1061 group_leader->signal->tty = NULL;
1062 spin_unlock(&group_leader->sighand->siglock);
1063
1064 err = session;
1065 out:
1066 write_unlock_irq(&tasklist_lock);
1067 return err;
1068 }
1069
1070 /*
1071 * Supplementary group IDs
1072 */
1073
1074 /* init to 2 - one for init_task, one to ensure it is never freed */
1075 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1076
1077 struct group_info *groups_alloc(int gidsetsize)
1078 {
1079 struct group_info *group_info;
1080 int nblocks;
1081 int i;
1082
1083 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1084 /* Make sure we always allocate at least one indirect block pointer */
1085 nblocks = nblocks ? : 1;
1086 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1087 if (!group_info)
1088 return NULL;
1089 group_info->ngroups = gidsetsize;
1090 group_info->nblocks = nblocks;
1091 atomic_set(&group_info->usage, 1);
1092
1093 if (gidsetsize <= NGROUPS_SMALL)
1094 group_info->blocks[0] = group_info->small_block;
1095 else {
1096 for (i = 0; i < nblocks; i++) {
1097 gid_t *b;
1098 b = (void *)__get_free_page(GFP_USER);
1099 if (!b)
1100 goto out_undo_partial_alloc;
1101 group_info->blocks[i] = b;
1102 }
1103 }
1104 return group_info;
1105
1106 out_undo_partial_alloc:
1107 while (--i >= 0) {
1108 free_page((unsigned long)group_info->blocks[i]);
1109 }
1110 kfree(group_info);
1111 return NULL;
1112 }
1113
1114 EXPORT_SYMBOL(groups_alloc);
1115
1116 void groups_free(struct group_info *group_info)
1117 {
1118 if (group_info->blocks[0] != group_info->small_block) {
1119 int i;
1120 for (i = 0; i < group_info->nblocks; i++)
1121 free_page((unsigned long)group_info->blocks[i]);
1122 }
1123 kfree(group_info);
1124 }
1125
1126 EXPORT_SYMBOL(groups_free);
1127
1128 /* export the group_info to a user-space array */
1129 static int groups_to_user(gid_t __user *grouplist,
1130 struct group_info *group_info)
1131 {
1132 int i;
1133 unsigned int count = group_info->ngroups;
1134
1135 for (i = 0; i < group_info->nblocks; i++) {
1136 unsigned int cp_count = min(NGROUPS_PER_BLOCK, count);
1137 unsigned int len = cp_count * sizeof(*grouplist);
1138
1139 if (copy_to_user(grouplist, group_info->blocks[i], len))
1140 return -EFAULT;
1141
1142 grouplist += NGROUPS_PER_BLOCK;
1143 count -= cp_count;
1144 }
1145 return 0;
1146 }
1147
1148 /* fill a group_info from a user-space array - it must be allocated already */
1149 static int groups_from_user(struct group_info *group_info,
1150 gid_t __user *grouplist)
1151 {
1152 int i;
1153 unsigned int count = group_info->ngroups;
1154
1155 for (i = 0; i < group_info->nblocks; i++) {
1156 unsigned int cp_count = min(NGROUPS_PER_BLOCK, count);
1157 unsigned int len = cp_count * sizeof(*grouplist);
1158
1159 if (copy_from_user(group_info->blocks[i], grouplist, len))
1160 return -EFAULT;
1161
1162 grouplist += NGROUPS_PER_BLOCK;
1163 count -= cp_count;
1164 }
1165 return 0;
1166 }
1167
1168 /* a simple Shell sort */
1169 static void groups_sort(struct group_info *group_info)
1170 {
1171 int base, max, stride;
1172 int gidsetsize = group_info->ngroups;
1173
1174 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1175 ; /* nothing */
1176 stride /= 3;
1177
1178 while (stride) {
1179 max = gidsetsize - stride;
1180 for (base = 0; base < max; base++) {
1181 int left = base;
1182 int right = left + stride;
1183 gid_t tmp = GROUP_AT(group_info, right);
1184
1185 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1186 GROUP_AT(group_info, right) =
1187 GROUP_AT(group_info, left);
1188 right = left;
1189 left -= stride;
1190 }
1191 GROUP_AT(group_info, right) = tmp;
1192 }
1193 stride /= 3;
1194 }
1195 }
1196
1197 /* a simple bsearch */
1198 int groups_search(struct group_info *group_info, gid_t grp)
1199 {
1200 unsigned int left, right;
1201
1202 if (!group_info)
1203 return 0;
1204
1205 left = 0;
1206 right = group_info->ngroups;
1207 while (left < right) {
1208 unsigned int mid = (left+right)/2;
1209 int cmp = grp - GROUP_AT(group_info, mid);
1210 if (cmp > 0)
1211 left = mid + 1;
1212 else if (cmp < 0)
1213 right = mid;
1214 else
1215 return 1;
1216 }
1217 return 0;
1218 }
1219
1220 /* validate and set current->group_info */
1221 int set_current_groups(struct group_info *group_info)
1222 {
1223 int retval;
1224 struct group_info *old_info;
1225
1226 retval = security_task_setgroups(group_info);
1227 if (retval)
1228 return retval;
1229
1230 groups_sort(group_info);
1231 get_group_info(group_info);
1232
1233 task_lock(current);
1234 old_info = current->group_info;
1235 current->group_info = group_info;
1236 task_unlock(current);
1237
1238 put_group_info(old_info);
1239
1240 return 0;
1241 }
1242
1243 EXPORT_SYMBOL(set_current_groups);
1244
1245 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1246 {
1247 int i = 0;
1248
1249 /*
1250 * SMP: Nobody else can change our grouplist. Thus we are
1251 * safe.
1252 */
1253
1254 if (gidsetsize < 0)
1255 return -EINVAL;
1256
1257 /* no need to grab task_lock here; it cannot change */
1258 i = current->group_info->ngroups;
1259 if (gidsetsize) {
1260 if (i > gidsetsize) {
1261 i = -EINVAL;
1262 goto out;
1263 }
1264 if (groups_to_user(grouplist, current->group_info)) {
1265 i = -EFAULT;
1266 goto out;
1267 }
1268 }
1269 out:
1270 return i;
1271 }
1272
1273 /*
1274 * SMP: Our groups are copy-on-write. We can set them safely
1275 * without another task interfering.
1276 */
1277
1278 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1279 {
1280 struct group_info *group_info;
1281 int retval;
1282
1283 if (!capable(CAP_SETGID))
1284 return -EPERM;
1285 if ((unsigned)gidsetsize > NGROUPS_MAX)
1286 return -EINVAL;
1287
1288 group_info = groups_alloc(gidsetsize);
1289 if (!group_info)
1290 return -ENOMEM;
1291 retval = groups_from_user(group_info, grouplist);
1292 if (retval) {
1293 put_group_info(group_info);
1294 return retval;
1295 }
1296
1297 retval = set_current_groups(group_info);
1298 put_group_info(group_info);
1299
1300 return retval;
1301 }
1302
1303 /*
1304 * Check whether we're fsgid/egid or in the supplemental group..
1305 */
1306 int in_group_p(gid_t grp)
1307 {
1308 int retval = 1;
1309 if (grp != current->fsgid)
1310 retval = groups_search(current->group_info, grp);
1311 return retval;
1312 }
1313
1314 EXPORT_SYMBOL(in_group_p);
1315
1316 int in_egroup_p(gid_t grp)
1317 {
1318 int retval = 1;
1319 if (grp != current->egid)
1320 retval = groups_search(current->group_info, grp);
1321 return retval;
1322 }
1323
1324 EXPORT_SYMBOL(in_egroup_p);
1325
1326 DECLARE_RWSEM(uts_sem);
1327
1328 EXPORT_SYMBOL(uts_sem);
1329
1330 asmlinkage long sys_newuname(struct new_utsname __user * name)
1331 {
1332 int errno = 0;
1333
1334 down_read(&uts_sem);
1335 if (copy_to_user(name, utsname(), sizeof *name))
1336 errno = -EFAULT;
1337 up_read(&uts_sem);
1338 return errno;
1339 }
1340
1341 asmlinkage long sys_sethostname(char __user *name, int len)
1342 {
1343 int errno;
1344 char tmp[__NEW_UTS_LEN];
1345
1346 if (!capable(CAP_SYS_ADMIN))
1347 return -EPERM;
1348 if (len < 0 || len > __NEW_UTS_LEN)
1349 return -EINVAL;
1350 down_write(&uts_sem);
1351 errno = -EFAULT;
1352 if (!copy_from_user(tmp, name, len)) {
1353 memcpy(utsname()->nodename, tmp, len);
1354 utsname()->nodename[len] = 0;
1355 errno = 0;
1356 }
1357 up_write(&uts_sem);
1358 return errno;
1359 }
1360
1361 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1362
1363 asmlinkage long sys_gethostname(char __user *name, int len)
1364 {
1365 int i, errno;
1366
1367 if (len < 0)
1368 return -EINVAL;
1369 down_read(&uts_sem);
1370 i = 1 + strlen(utsname()->nodename);
1371 if (i > len)
1372 i = len;
1373 errno = 0;
1374 if (copy_to_user(name, utsname()->nodename, i))
1375 errno = -EFAULT;
1376 up_read(&uts_sem);
1377 return errno;
1378 }
1379
1380 #endif
1381
1382 /*
1383 * Only setdomainname; getdomainname can be implemented by calling
1384 * uname()
1385 */
1386 asmlinkage long sys_setdomainname(char __user *name, int len)
1387 {
1388 int errno;
1389 char tmp[__NEW_UTS_LEN];
1390
1391 if (!capable(CAP_SYS_ADMIN))
1392 return -EPERM;
1393 if (len < 0 || len > __NEW_UTS_LEN)
1394 return -EINVAL;
1395
1396 down_write(&uts_sem);
1397 errno = -EFAULT;
1398 if (!copy_from_user(tmp, name, len)) {
1399 memcpy(utsname()->domainname, tmp, len);
1400 utsname()->domainname[len] = 0;
1401 errno = 0;
1402 }
1403 up_write(&uts_sem);
1404 return errno;
1405 }
1406
1407 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1408 {
1409 if (resource >= RLIM_NLIMITS)
1410 return -EINVAL;
1411 else {
1412 struct rlimit value;
1413 task_lock(current->group_leader);
1414 value = current->signal->rlim[resource];
1415 task_unlock(current->group_leader);
1416 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1417 }
1418 }
1419
1420 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1421
1422 /*
1423 * Back compatibility for getrlimit. Needed for some apps.
1424 */
1425
1426 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1427 {
1428 struct rlimit x;
1429 if (resource >= RLIM_NLIMITS)
1430 return -EINVAL;
1431
1432 task_lock(current->group_leader);
1433 x = current->signal->rlim[resource];
1434 task_unlock(current->group_leader);
1435 if (x.rlim_cur > 0x7FFFFFFF)
1436 x.rlim_cur = 0x7FFFFFFF;
1437 if (x.rlim_max > 0x7FFFFFFF)
1438 x.rlim_max = 0x7FFFFFFF;
1439 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1440 }
1441
1442 #endif
1443
1444 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1445 {
1446 struct rlimit new_rlim, *old_rlim;
1447 unsigned long it_prof_secs;
1448 int retval;
1449
1450 if (resource >= RLIM_NLIMITS)
1451 return -EINVAL;
1452 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1453 return -EFAULT;
1454 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1455 return -EINVAL;
1456 old_rlim = current->signal->rlim + resource;
1457 if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1458 !capable(CAP_SYS_RESOURCE))
1459 return -EPERM;
1460 if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > sysctl_nr_open)
1461 return -EPERM;
1462
1463 retval = security_task_setrlimit(resource, &new_rlim);
1464 if (retval)
1465 return retval;
1466
1467 if (resource == RLIMIT_CPU && new_rlim.rlim_cur == 0) {
1468 /*
1469 * The caller is asking for an immediate RLIMIT_CPU
1470 * expiry. But we use the zero value to mean "it was
1471 * never set". So let's cheat and make it one second
1472 * instead
1473 */
1474 new_rlim.rlim_cur = 1;
1475 }
1476
1477 task_lock(current->group_leader);
1478 *old_rlim = new_rlim;
1479 task_unlock(current->group_leader);
1480
1481 if (resource != RLIMIT_CPU)
1482 goto out;
1483
1484 /*
1485 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1486 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1487 * very long-standing error, and fixing it now risks breakage of
1488 * applications, so we live with it
1489 */
1490 if (new_rlim.rlim_cur == RLIM_INFINITY)
1491 goto out;
1492
1493 it_prof_secs = cputime_to_secs(current->signal->it_prof_expires);
1494 if (it_prof_secs == 0 || new_rlim.rlim_cur <= it_prof_secs) {
1495 unsigned long rlim_cur = new_rlim.rlim_cur;
1496 cputime_t cputime;
1497
1498 cputime = secs_to_cputime(rlim_cur);
1499 read_lock(&tasklist_lock);
1500 spin_lock_irq(&current->sighand->siglock);
1501 set_process_cpu_timer(current, CPUCLOCK_PROF, &cputime, NULL);
1502 spin_unlock_irq(&current->sighand->siglock);
1503 read_unlock(&tasklist_lock);
1504 }
1505 out:
1506 return 0;
1507 }
1508
1509 /*
1510 * It would make sense to put struct rusage in the task_struct,
1511 * except that would make the task_struct be *really big*. After
1512 * task_struct gets moved into malloc'ed memory, it would
1513 * make sense to do this. It will make moving the rest of the information
1514 * a lot simpler! (Which we're not doing right now because we're not
1515 * measuring them yet).
1516 *
1517 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1518 * races with threads incrementing their own counters. But since word
1519 * reads are atomic, we either get new values or old values and we don't
1520 * care which for the sums. We always take the siglock to protect reading
1521 * the c* fields from p->signal from races with exit.c updating those
1522 * fields when reaping, so a sample either gets all the additions of a
1523 * given child after it's reaped, or none so this sample is before reaping.
1524 *
1525 * Locking:
1526 * We need to take the siglock for CHILDEREN, SELF and BOTH
1527 * for the cases current multithreaded, non-current single threaded
1528 * non-current multithreaded. Thread traversal is now safe with
1529 * the siglock held.
1530 * Strictly speaking, we donot need to take the siglock if we are current and
1531 * single threaded, as no one else can take our signal_struct away, no one
1532 * else can reap the children to update signal->c* counters, and no one else
1533 * can race with the signal-> fields. If we do not take any lock, the
1534 * signal-> fields could be read out of order while another thread was just
1535 * exiting. So we should place a read memory barrier when we avoid the lock.
1536 * On the writer side, write memory barrier is implied in __exit_signal
1537 * as __exit_signal releases the siglock spinlock after updating the signal->
1538 * fields. But we don't do this yet to keep things simple.
1539 *
1540 */
1541
1542 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1543 {
1544 struct task_struct *t;
1545 unsigned long flags;
1546 cputime_t utime, stime;
1547
1548 memset((char *) r, 0, sizeof *r);
1549 utime = stime = cputime_zero;
1550
1551 rcu_read_lock();
1552 if (!lock_task_sighand(p, &flags)) {
1553 rcu_read_unlock();
1554 return;
1555 }
1556
1557 switch (who) {
1558 case RUSAGE_BOTH:
1559 case RUSAGE_CHILDREN:
1560 utime = p->signal->cutime;
1561 stime = p->signal->cstime;
1562 r->ru_nvcsw = p->signal->cnvcsw;
1563 r->ru_nivcsw = p->signal->cnivcsw;
1564 r->ru_minflt = p->signal->cmin_flt;
1565 r->ru_majflt = p->signal->cmaj_flt;
1566 r->ru_inblock = p->signal->cinblock;
1567 r->ru_oublock = p->signal->coublock;
1568
1569 if (who == RUSAGE_CHILDREN)
1570 break;
1571
1572 case RUSAGE_SELF:
1573 utime = cputime_add(utime, p->signal->utime);
1574 stime = cputime_add(stime, p->signal->stime);
1575 r->ru_nvcsw += p->signal->nvcsw;
1576 r->ru_nivcsw += p->signal->nivcsw;
1577 r->ru_minflt += p->signal->min_flt;
1578 r->ru_majflt += p->signal->maj_flt;
1579 r->ru_inblock += p->signal->inblock;
1580 r->ru_oublock += p->signal->oublock;
1581 t = p;
1582 do {
1583 utime = cputime_add(utime, t->utime);
1584 stime = cputime_add(stime, t->stime);
1585 r->ru_nvcsw += t->nvcsw;
1586 r->ru_nivcsw += t->nivcsw;
1587 r->ru_minflt += t->min_flt;
1588 r->ru_majflt += t->maj_flt;
1589 r->ru_inblock += task_io_get_inblock(t);
1590 r->ru_oublock += task_io_get_oublock(t);
1591 t = next_thread(t);
1592 } while (t != p);
1593 break;
1594
1595 default:
1596 BUG();
1597 }
1598
1599 unlock_task_sighand(p, &flags);
1600 rcu_read_unlock();
1601
1602 cputime_to_timeval(utime, &r->ru_utime);
1603 cputime_to_timeval(stime, &r->ru_stime);
1604 }
1605
1606 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1607 {
1608 struct rusage r;
1609 k_getrusage(p, who, &r);
1610 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1611 }
1612
1613 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1614 {
1615 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1616 return -EINVAL;
1617 return getrusage(current, who, ru);
1618 }
1619
1620 asmlinkage long sys_umask(int mask)
1621 {
1622 mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1623 return mask;
1624 }
1625
1626 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1627 unsigned long arg4, unsigned long arg5)
1628 {
1629 long error;
1630
1631 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1632 if (error)
1633 return error;
1634
1635 switch (option) {
1636 case PR_SET_PDEATHSIG:
1637 if (!valid_signal(arg2)) {
1638 error = -EINVAL;
1639 break;
1640 }
1641 current->pdeath_signal = arg2;
1642 break;
1643 case PR_GET_PDEATHSIG:
1644 error = put_user(current->pdeath_signal, (int __user *)arg2);
1645 break;
1646 case PR_GET_DUMPABLE:
1647 error = get_dumpable(current->mm);
1648 break;
1649 case PR_SET_DUMPABLE:
1650 if (arg2 < 0 || arg2 > 1) {
1651 error = -EINVAL;
1652 break;
1653 }
1654 set_dumpable(current->mm, arg2);
1655 break;
1656
1657 case PR_SET_UNALIGN:
1658 error = SET_UNALIGN_CTL(current, arg2);
1659 break;
1660 case PR_GET_UNALIGN:
1661 error = GET_UNALIGN_CTL(current, arg2);
1662 break;
1663 case PR_SET_FPEMU:
1664 error = SET_FPEMU_CTL(current, arg2);
1665 break;
1666 case PR_GET_FPEMU:
1667 error = GET_FPEMU_CTL(current, arg2);
1668 break;
1669 case PR_SET_FPEXC:
1670 error = SET_FPEXC_CTL(current, arg2);
1671 break;
1672 case PR_GET_FPEXC:
1673 error = GET_FPEXC_CTL(current, arg2);
1674 break;
1675 case PR_GET_TIMING:
1676 error = PR_TIMING_STATISTICAL;
1677 break;
1678 case PR_SET_TIMING:
1679 if (arg2 == PR_TIMING_STATISTICAL)
1680 error = 0;
1681 else
1682 error = -EINVAL;
1683 break;
1684
1685 case PR_GET_KEEPCAPS:
1686 if (current->keep_capabilities)
1687 error = 1;
1688 break;
1689 case PR_SET_KEEPCAPS:
1690 if (arg2 != 0 && arg2 != 1) {
1691 error = -EINVAL;
1692 break;
1693 }
1694 current->keep_capabilities = arg2;
1695 break;
1696 case PR_SET_NAME: {
1697 struct task_struct *me = current;
1698 unsigned char ncomm[sizeof(me->comm)];
1699
1700 ncomm[sizeof(me->comm)-1] = 0;
1701 if (strncpy_from_user(ncomm, (char __user *)arg2,
1702 sizeof(me->comm)-1) < 0)
1703 return -EFAULT;
1704 set_task_comm(me, ncomm);
1705 return 0;
1706 }
1707 case PR_GET_NAME: {
1708 struct task_struct *me = current;
1709 unsigned char tcomm[sizeof(me->comm)];
1710
1711 get_task_comm(tcomm, me);
1712 if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))
1713 return -EFAULT;
1714 return 0;
1715 }
1716 case PR_GET_ENDIAN:
1717 error = GET_ENDIAN(current, arg2);
1718 break;
1719 case PR_SET_ENDIAN:
1720 error = SET_ENDIAN(current, arg2);
1721 break;
1722
1723 case PR_GET_SECCOMP:
1724 error = prctl_get_seccomp();
1725 break;
1726 case PR_SET_SECCOMP:
1727 error = prctl_set_seccomp(arg2);
1728 break;
1729
1730 case PR_CAPBSET_READ:
1731 if (!cap_valid(arg2))
1732 return -EINVAL;
1733 return !!cap_raised(current->cap_bset, arg2);
1734 case PR_CAPBSET_DROP:
1735 #ifdef CONFIG_SECURITY_FILE_CAPABILITIES
1736 return cap_prctl_drop(arg2);
1737 #else
1738 return -EINVAL;
1739 #endif
1740
1741 default:
1742 error = -EINVAL;
1743 break;
1744 }
1745 return error;
1746 }
1747
1748 asmlinkage long sys_getcpu(unsigned __user *cpup, unsigned __user *nodep,
1749 struct getcpu_cache __user *unused)
1750 {
1751 int err = 0;
1752 int cpu = raw_smp_processor_id();
1753 if (cpup)
1754 err |= put_user(cpu, cpup);
1755 if (nodep)
1756 err |= put_user(cpu_to_node(cpu), nodep);
1757 return err ? -EFAULT : 0;
1758 }
1759
1760 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
1761
1762 static void argv_cleanup(char **argv, char **envp)
1763 {
1764 argv_free(argv);
1765 }
1766
1767 /**
1768 * orderly_poweroff - Trigger an orderly system poweroff
1769 * @force: force poweroff if command execution fails
1770 *
1771 * This may be called from any context to trigger a system shutdown.
1772 * If the orderly shutdown fails, it will force an immediate shutdown.
1773 */
1774 int orderly_poweroff(bool force)
1775 {
1776 int argc;
1777 char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
1778 static char *envp[] = {
1779 "HOME=/",
1780 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
1781 NULL
1782 };
1783 int ret = -ENOMEM;
1784 struct subprocess_info *info;
1785
1786 if (argv == NULL) {
1787 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
1788 __func__, poweroff_cmd);
1789 goto out;
1790 }
1791
1792 info = call_usermodehelper_setup(argv[0], argv, envp);
1793 if (info == NULL) {
1794 argv_free(argv);
1795 goto out;
1796 }
1797
1798 call_usermodehelper_setcleanup(info, argv_cleanup);
1799
1800 ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
1801
1802 out:
1803 if (ret && force) {
1804 printk(KERN_WARNING "Failed to start orderly shutdown: "
1805 "forcing the issue\n");
1806
1807 /* I guess this should try to kick off some daemon to
1808 sync and poweroff asap. Or not even bother syncing
1809 if we're doing an emergency shutdown? */
1810 emergency_sync();
1811 kernel_power_off();
1812 }
1813
1814 return ret;
1815 }
1816 EXPORT_SYMBOL_GPL(orderly_poweroff);
kmemtrace - Kernel memory tracer