?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge tag 'sound-3.11' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound
[linux-2.6.git] / kernel / sys.c
blob071de900c824540abd7e42ee942e2ed89767643b
1 /*
2 * linux/kernel/sys.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
7 #include <linux/export.h>
8 #include <linux/mm.h>
9 #include <linux/utsname.h>
10 #include <linux/mman.h>
11 #include <linux/reboot.h>
12 #include <linux/prctl.h>
13 #include <linux/highuid.h>
14 #include <linux/fs.h>
15 #include <linux/kmod.h>
16 #include <linux/perf_event.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 #include <linux/personality.h>
37 #include <linux/ptrace.h>
38 #include <linux/fs_struct.h>
39 #include <linux/file.h>
40 #include <linux/mount.h>
41 #include <linux/gfp.h>
42 #include <linux/syscore_ops.h>
43 #include <linux/version.h>
44 #include <linux/ctype.h>
45
46 #include <linux/compat.h>
47 #include <linux/syscalls.h>
48 #include <linux/kprobes.h>
49 #include <linux/user_namespace.h>
50 #include <linux/binfmts.h>
51
52 #include <linux/sched.h>
53 #include <linux/rcupdate.h>
54 #include <linux/uidgid.h>
55 #include <linux/cred.h>
56
57 #include <linux/kmsg_dump.h>
58 /* Move somewhere else to avoid recompiling? */
59 #include <generated/utsrelease.h>
60
61 #include <asm/uaccess.h>
62 #include <asm/io.h>
63 #include <asm/unistd.h>
64
65 #ifndef SET_UNALIGN_CTL
66 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
67 #endif
68 #ifndef GET_UNALIGN_CTL
69 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
70 #endif
71 #ifndef SET_FPEMU_CTL
72 # define SET_FPEMU_CTL(a,b) (-EINVAL)
73 #endif
74 #ifndef GET_FPEMU_CTL
75 # define GET_FPEMU_CTL(a,b) (-EINVAL)
76 #endif
77 #ifndef SET_FPEXC_CTL
78 # define SET_FPEXC_CTL(a,b) (-EINVAL)
79 #endif
80 #ifndef GET_FPEXC_CTL
81 # define GET_FPEXC_CTL(a,b) (-EINVAL)
82 #endif
83 #ifndef GET_ENDIAN
84 # define GET_ENDIAN(a,b) (-EINVAL)
85 #endif
86 #ifndef SET_ENDIAN
87 # define SET_ENDIAN(a,b) (-EINVAL)
88 #endif
89 #ifndef GET_TSC_CTL
90 # define GET_TSC_CTL(a) (-EINVAL)
91 #endif
92 #ifndef SET_TSC_CTL
93 # define SET_TSC_CTL(a) (-EINVAL)
94 #endif
95
96 /*
97 * this is where the system-wide overflow UID and GID are defined, for
98 * architectures that now have 32-bit UID/GID but didn't in the past
99 */
100
101 int overflowuid = DEFAULT_OVERFLOWUID;
102 int overflowgid = DEFAULT_OVERFLOWGID;
103
104 EXPORT_SYMBOL(overflowuid);
105 EXPORT_SYMBOL(overflowgid);
106
107 /*
108 * the same as above, but for filesystems which can only store a 16-bit
109 * UID and GID. as such, this is needed on all architectures
110 */
111
112 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
113 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
114
115 EXPORT_SYMBOL(fs_overflowuid);
116 EXPORT_SYMBOL(fs_overflowgid);
117
118 /*
119 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
120 */
121
122 int C_A_D = 1;
123 struct pid *cad_pid;
124 EXPORT_SYMBOL(cad_pid);
125
126 /*
127 * If set, this is used for preparing the system to power off.
128 */
129
130 void (*pm_power_off_prepare)(void);
131
132 /*
133 * Returns true if current's euid is same as p's uid or euid,
134 * or has CAP_SYS_NICE to p's user_ns.
135 *
136 * Called with rcu_read_lock, creds are safe
137 */
138 static bool set_one_prio_perm(struct task_struct *p)
139 {
140 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
141
142 if (uid_eq(pcred->uid, cred->euid) ||
143 uid_eq(pcred->euid, cred->euid))
144 return true;
145 if (ns_capable(pcred->user_ns, CAP_SYS_NICE))
146 return true;
147 return false;
148 }
149
150 /*
151 * set the priority of a task
152 * - the caller must hold the RCU read lock
153 */
154 static int set_one_prio(struct task_struct *p, int niceval, int error)
155 {
156 int no_nice;
157
158 if (!set_one_prio_perm(p)) {
159 error = -EPERM;
160 goto out;
161 }
162 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
163 error = -EACCES;
164 goto out;
165 }
166 no_nice = security_task_setnice(p, niceval);
167 if (no_nice) {
168 error = no_nice;
169 goto out;
170 }
171 if (error == -ESRCH)
172 error = 0;
173 set_user_nice(p, niceval);
174 out:
175 return error;
176 }
177
178 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
179 {
180 struct task_struct *g, *p;
181 struct user_struct *user;
182 const struct cred *cred = current_cred();
183 int error = -EINVAL;
184 struct pid *pgrp;
185 kuid_t uid;
186
187 if (which > PRIO_USER || which < PRIO_PROCESS)
188 goto out;
189
190 /* normalize: avoid signed division (rounding problems) */
191 error = -ESRCH;
192 if (niceval < -20)
193 niceval = -20;
194 if (niceval > 19)
195 niceval = 19;
196
197 rcu_read_lock();
198 read_lock(&tasklist_lock);
199 switch (which) {
200 case PRIO_PROCESS:
201 if (who)
202 p = find_task_by_vpid(who);
203 else
204 p = current;
205 if (p)
206 error = set_one_prio(p, niceval, error);
207 break;
208 case PRIO_PGRP:
209 if (who)
210 pgrp = find_vpid(who);
211 else
212 pgrp = task_pgrp(current);
213 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
214 error = set_one_prio(p, niceval, error);
215 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
216 break;
217 case PRIO_USER:
218 uid = make_kuid(cred->user_ns, who);
219 user = cred->user;
220 if (!who)
221 uid = cred->uid;
222 else if (!uid_eq(uid, cred->uid) &&
223 !(user = find_user(uid)))
224 goto out_unlock; /* No processes for this user */
225
226 do_each_thread(g, p) {
227 if (uid_eq(task_uid(p), uid))
228 error = set_one_prio(p, niceval, error);
229 } while_each_thread(g, p);
230 if (!uid_eq(uid, cred->uid))
231 free_uid(user); /* For find_user() */
232 break;
233 }
234 out_unlock:
235 read_unlock(&tasklist_lock);
236 rcu_read_unlock();
237 out:
238 return error;
239 }
240
241 /*
242 * Ugh. To avoid negative return values, "getpriority()" will
243 * not return the normal nice-value, but a negated value that
244 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
245 * to stay compatible.
246 */
247 SYSCALL_DEFINE2(getpriority, int, which, int, who)
248 {
249 struct task_struct *g, *p;
250 struct user_struct *user;
251 const struct cred *cred = current_cred();
252 long niceval, retval = -ESRCH;
253 struct pid *pgrp;
254 kuid_t uid;
255
256 if (which > PRIO_USER || which < PRIO_PROCESS)
257 return -EINVAL;
258
259 rcu_read_lock();
260 read_lock(&tasklist_lock);
261 switch (which) {
262 case PRIO_PROCESS:
263 if (who)
264 p = find_task_by_vpid(who);
265 else
266 p = current;
267 if (p) {
268 niceval = 20 - task_nice(p);
269 if (niceval > retval)
270 retval = niceval;
271 }
272 break;
273 case PRIO_PGRP:
274 if (who)
275 pgrp = find_vpid(who);
276 else
277 pgrp = task_pgrp(current);
278 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
279 niceval = 20 - task_nice(p);
280 if (niceval > retval)
281 retval = niceval;
282 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
283 break;
284 case PRIO_USER:
285 uid = make_kuid(cred->user_ns, who);
286 user = cred->user;
287 if (!who)
288 uid = cred->uid;
289 else if (!uid_eq(uid, cred->uid) &&
290 !(user = find_user(uid)))
291 goto out_unlock; /* No processes for this user */
292
293 do_each_thread(g, p) {
294 if (uid_eq(task_uid(p), uid)) {
295 niceval = 20 - task_nice(p);
296 if (niceval > retval)
297 retval = niceval;
298 }
299 } while_each_thread(g, p);
300 if (!uid_eq(uid, cred->uid))
301 free_uid(user); /* for find_user() */
302 break;
303 }
304 out_unlock:
305 read_unlock(&tasklist_lock);
306 rcu_read_unlock();
307
308 return retval;
309 }
310
311 /**
312 * emergency_restart - reboot the system
313 *
314 * Without shutting down any hardware or taking any locks
315 * reboot the system. This is called when we know we are in
316 * trouble so this is our best effort to reboot. This is
317 * safe to call in interrupt context.
318 */
319 void emergency_restart(void)
320 {
321 kmsg_dump(KMSG_DUMP_EMERG);
322 machine_emergency_restart();
323 }
324 EXPORT_SYMBOL_GPL(emergency_restart);
325
326 void kernel_restart_prepare(char *cmd)
327 {
328 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
329 system_state = SYSTEM_RESTART;
330 usermodehelper_disable();
331 device_shutdown();
332 }
333
334 /**
335 * register_reboot_notifier - Register function to be called at reboot time
336 * @nb: Info about notifier function to be called
337 *
338 * Registers a function with the list of functions
339 * to be called at reboot time.
340 *
341 * Currently always returns zero, as blocking_notifier_chain_register()
342 * always returns zero.
343 */
344 int register_reboot_notifier(struct notifier_block *nb)
345 {
346 return blocking_notifier_chain_register(&reboot_notifier_list, nb);
347 }
348 EXPORT_SYMBOL(register_reboot_notifier);
349
350 /**
351 * unregister_reboot_notifier - Unregister previously registered reboot notifier
352 * @nb: Hook to be unregistered
353 *
354 * Unregisters a previously registered reboot
355 * notifier function.
356 *
357 * Returns zero on success, or %-ENOENT on failure.
358 */
359 int unregister_reboot_notifier(struct notifier_block *nb)
360 {
361 return blocking_notifier_chain_unregister(&reboot_notifier_list, nb);
362 }
363 EXPORT_SYMBOL(unregister_reboot_notifier);
364
365 /* Add backwards compatibility for stable trees. */
366 #ifndef PF_NO_SETAFFINITY
367 #define PF_NO_SETAFFINITY PF_THREAD_BOUND
368 #endif
369
370 static void migrate_to_reboot_cpu(void)
371 {
372 /* The boot cpu is always logical cpu 0 */
373 int cpu = 0;
374
375 cpu_hotplug_disable();
376
377 /* Make certain the cpu I'm about to reboot on is online */
378 if (!cpu_online(cpu))
379 cpu = cpumask_first(cpu_online_mask);
380
381 /* Prevent races with other tasks migrating this task */
382 current->flags |= PF_NO_SETAFFINITY;
383
384 /* Make certain I only run on the appropriate processor */
385 set_cpus_allowed_ptr(current, cpumask_of(cpu));
386 }
387
388 /**
389 * kernel_restart - reboot the system
390 * @cmd: pointer to buffer containing command to execute for restart
391 * or %NULL
392 *
393 * Shutdown everything and perform a clean reboot.
394 * This is not safe to call in interrupt context.
395 */
396 void kernel_restart(char *cmd)
397 {
398 kernel_restart_prepare(cmd);
399 migrate_to_reboot_cpu();
400 syscore_shutdown();
401 if (!cmd)
402 printk(KERN_EMERG "Restarting system.\n");
403 else
404 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
405 kmsg_dump(KMSG_DUMP_RESTART);
406 machine_restart(cmd);
407 }
408 EXPORT_SYMBOL_GPL(kernel_restart);
409
410 static void kernel_shutdown_prepare(enum system_states state)
411 {
412 blocking_notifier_call_chain(&reboot_notifier_list,
413 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
414 system_state = state;
415 usermodehelper_disable();
416 device_shutdown();
417 }
418 /**
419 * kernel_halt - halt the system
420 *
421 * Shutdown everything and perform a clean system halt.
422 */
423 void kernel_halt(void)
424 {
425 kernel_shutdown_prepare(SYSTEM_HALT);
426 migrate_to_reboot_cpu();
427 syscore_shutdown();
428 printk(KERN_EMERG "System halted.\n");
429 kmsg_dump(KMSG_DUMP_HALT);
430 machine_halt();
431 }
432
433 EXPORT_SYMBOL_GPL(kernel_halt);
434
435 /**
436 * kernel_power_off - power_off the system
437 *
438 * Shutdown everything and perform a clean system power_off.
439 */
440 void kernel_power_off(void)
441 {
442 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
443 if (pm_power_off_prepare)
444 pm_power_off_prepare();
445 migrate_to_reboot_cpu();
446 syscore_shutdown();
447 printk(KERN_EMERG "Power down.\n");
448 kmsg_dump(KMSG_DUMP_POWEROFF);
449 machine_power_off();
450 }
451 EXPORT_SYMBOL_GPL(kernel_power_off);
452
453 static DEFINE_MUTEX(reboot_mutex);
454
455 /*
456 * Reboot system call: for obvious reasons only root may call it,
457 * and even root needs to set up some magic numbers in the registers
458 * so that some mistake won't make this reboot the whole machine.
459 * You can also set the meaning of the ctrl-alt-del-key here.
460 *
461 * reboot doesn't sync: do that yourself before calling this.
462 */
463 SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
464 void __user *, arg)
465 {
466 struct pid_namespace *pid_ns = task_active_pid_ns(current);
467 char buffer[256];
468 int ret = 0;
469
470 /* We only trust the superuser with rebooting the system. */
471 if (!ns_capable(pid_ns->user_ns, CAP_SYS_BOOT))
472 return -EPERM;
473
474 /* For safety, we require "magic" arguments. */
475 if (magic1 != LINUX_REBOOT_MAGIC1 ||
476 (magic2 != LINUX_REBOOT_MAGIC2 &&
477 magic2 != LINUX_REBOOT_MAGIC2A &&
478 magic2 != LINUX_REBOOT_MAGIC2B &&
479 magic2 != LINUX_REBOOT_MAGIC2C))
480 return -EINVAL;
481
482 /*
483 * If pid namespaces are enabled and the current task is in a child
484 * pid_namespace, the command is handled by reboot_pid_ns() which will
485 * call do_exit().
486 */
487 ret = reboot_pid_ns(pid_ns, cmd);
488 if (ret)
489 return ret;
490
491 /* Instead of trying to make the power_off code look like
492 * halt when pm_power_off is not set do it the easy way.
493 */
494 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
495 cmd = LINUX_REBOOT_CMD_HALT;
496
497 mutex_lock(&reboot_mutex);
498 switch (cmd) {
499 case LINUX_REBOOT_CMD_RESTART:
500 kernel_restart(NULL);
501 break;
502
503 case LINUX_REBOOT_CMD_CAD_ON:
504 C_A_D = 1;
505 break;
506
507 case LINUX_REBOOT_CMD_CAD_OFF:
508 C_A_D = 0;
509 break;
510
511 case LINUX_REBOOT_CMD_HALT:
512 kernel_halt();
513 do_exit(0);
514 panic("cannot halt.\n");
515
516 case LINUX_REBOOT_CMD_POWER_OFF:
517 kernel_power_off();
518 do_exit(0);
519 break;
520
521 case LINUX_REBOOT_CMD_RESTART2:
522 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
523 ret = -EFAULT;
524 break;
525 }
526 buffer[sizeof(buffer) - 1] = '\0';
527
528 kernel_restart(buffer);
529 break;
530
531 #ifdef CONFIG_KEXEC
532 case LINUX_REBOOT_CMD_KEXEC:
533 ret = kernel_kexec();
534 break;
535 #endif
536
537 #ifdef CONFIG_HIBERNATION
538 case LINUX_REBOOT_CMD_SW_SUSPEND:
539 ret = hibernate();
540 break;
541 #endif
542
543 default:
544 ret = -EINVAL;
545 break;
546 }
547 mutex_unlock(&reboot_mutex);
548 return ret;
549 }
550
551 static void deferred_cad(struct work_struct *dummy)
552 {
553 kernel_restart(NULL);
554 }
555
556 /*
557 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
558 * As it's called within an interrupt, it may NOT sync: the only choice
559 * is whether to reboot at once, or just ignore the ctrl-alt-del.
560 */
561 void ctrl_alt_del(void)
562 {
563 static DECLARE_WORK(cad_work, deferred_cad);
564
565 if (C_A_D)
566 schedule_work(&cad_work);
567 else
568 kill_cad_pid(SIGINT, 1);
569 }
570
571 /*
572 * Unprivileged users may change the real gid to the effective gid
573 * or vice versa. (BSD-style)
574 *
575 * If you set the real gid at all, or set the effective gid to a value not
576 * equal to the real gid, then the saved gid is set to the new effective gid.
577 *
578 * This makes it possible for a setgid program to completely drop its
579 * privileges, which is often a useful assertion to make when you are doing
580 * a security audit over a program.
581 *
582 * The general idea is that a program which uses just setregid() will be
583 * 100% compatible with BSD. A program which uses just setgid() will be
584 * 100% compatible with POSIX with saved IDs.
585 *
586 * SMP: There are not races, the GIDs are checked only by filesystem
587 * operations (as far as semantic preservation is concerned).
588 */
589 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
590 {
591 struct user_namespace *ns = current_user_ns();
592 const struct cred *old;
593 struct cred *new;
594 int retval;
595 kgid_t krgid, kegid;
596
597 krgid = make_kgid(ns, rgid);
598 kegid = make_kgid(ns, egid);
599
600 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
601 return -EINVAL;
602 if ((egid != (gid_t) -1) && !gid_valid(kegid))
603 return -EINVAL;
604
605 new = prepare_creds();
606 if (!new)
607 return -ENOMEM;
608 old = current_cred();
609
610 retval = -EPERM;
611 if (rgid != (gid_t) -1) {
612 if (gid_eq(old->gid, krgid) ||
613 gid_eq(old->egid, krgid) ||
614 nsown_capable(CAP_SETGID))
615 new->gid = krgid;
616 else
617 goto error;
618 }
619 if (egid != (gid_t) -1) {
620 if (gid_eq(old->gid, kegid) ||
621 gid_eq(old->egid, kegid) ||
622 gid_eq(old->sgid, kegid) ||
623 nsown_capable(CAP_SETGID))
624 new->egid = kegid;
625 else
626 goto error;
627 }
628
629 if (rgid != (gid_t) -1 ||
630 (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
631 new->sgid = new->egid;
632 new->fsgid = new->egid;
633
634 return commit_creds(new);
635
636 error:
637 abort_creds(new);
638 return retval;
639 }
640
641 /*
642 * setgid() is implemented like SysV w/ SAVED_IDS
643 *
644 * SMP: Same implicit races as above.
645 */
646 SYSCALL_DEFINE1(setgid, gid_t, gid)
647 {
648 struct user_namespace *ns = current_user_ns();
649 const struct cred *old;
650 struct cred *new;
651 int retval;
652 kgid_t kgid;
653
654 kgid = make_kgid(ns, gid);
655 if (!gid_valid(kgid))
656 return -EINVAL;
657
658 new = prepare_creds();
659 if (!new)
660 return -ENOMEM;
661 old = current_cred();
662
663 retval = -EPERM;
664 if (nsown_capable(CAP_SETGID))
665 new->gid = new->egid = new->sgid = new->fsgid = kgid;
666 else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
667 new->egid = new->fsgid = kgid;
668 else
669 goto error;
670
671 return commit_creds(new);
672
673 error:
674 abort_creds(new);
675 return retval;
676 }
677
678 /*
679 * change the user struct in a credentials set to match the new UID
680 */
681 static int set_user(struct cred *new)
682 {
683 struct user_struct *new_user;
684
685 new_user = alloc_uid(new->uid);
686 if (!new_user)
687 return -EAGAIN;
688
689 /*
690 * We don't fail in case of NPROC limit excess here because too many
691 * poorly written programs don't check set*uid() return code, assuming
692 * it never fails if called by root. We may still enforce NPROC limit
693 * for programs doing set*uid()+execve() by harmlessly deferring the
694 * failure to the execve() stage.
695 */
696 if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
697 new_user != INIT_USER)
698 current->flags |= PF_NPROC_EXCEEDED;
699 else
700 current->flags &= ~PF_NPROC_EXCEEDED;
701
702 free_uid(new->user);
703 new->user = new_user;
704 return 0;
705 }
706
707 /*
708 * Unprivileged users may change the real uid to the effective uid
709 * or vice versa. (BSD-style)
710 *
711 * If you set the real uid at all, or set the effective uid to a value not
712 * equal to the real uid, then the saved uid is set to the new effective uid.
713 *
714 * This makes it possible for a setuid program to completely drop its
715 * privileges, which is often a useful assertion to make when you are doing
716 * a security audit over a program.
717 *
718 * The general idea is that a program which uses just setreuid() will be
719 * 100% compatible with BSD. A program which uses just setuid() will be
720 * 100% compatible with POSIX with saved IDs.
721 */
722 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
723 {
724 struct user_namespace *ns = current_user_ns();
725 const struct cred *old;
726 struct cred *new;
727 int retval;
728 kuid_t kruid, keuid;
729
730 kruid = make_kuid(ns, ruid);
731 keuid = make_kuid(ns, euid);
732
733 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
734 return -EINVAL;
735 if ((euid != (uid_t) -1) && !uid_valid(keuid))
736 return -EINVAL;
737
738 new = prepare_creds();
739 if (!new)
740 return -ENOMEM;
741 old = current_cred();
742
743 retval = -EPERM;
744 if (ruid != (uid_t) -1) {
745 new->uid = kruid;
746 if (!uid_eq(old->uid, kruid) &&
747 !uid_eq(old->euid, kruid) &&
748 !nsown_capable(CAP_SETUID))
749 goto error;
750 }
751
752 if (euid != (uid_t) -1) {
753 new->euid = keuid;
754 if (!uid_eq(old->uid, keuid) &&
755 !uid_eq(old->euid, keuid) &&
756 !uid_eq(old->suid, keuid) &&
757 !nsown_capable(CAP_SETUID))
758 goto error;
759 }
760
761 if (!uid_eq(new->uid, old->uid)) {
762 retval = set_user(new);
763 if (retval < 0)
764 goto error;
765 }
766 if (ruid != (uid_t) -1 ||
767 (euid != (uid_t) -1 && !uid_eq(keuid, old->uid)))
768 new->suid = new->euid;
769 new->fsuid = new->euid;
770
771 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
772 if (retval < 0)
773 goto error;
774
775 return commit_creds(new);
776
777 error:
778 abort_creds(new);
779 return retval;
780 }
781
782 /*
783 * setuid() is implemented like SysV with SAVED_IDS
784 *
785 * Note that SAVED_ID's is deficient in that a setuid root program
786 * like sendmail, for example, cannot set its uid to be a normal
787 * user and then switch back, because if you're root, setuid() sets
788 * the saved uid too. If you don't like this, blame the bright people
789 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
790 * will allow a root program to temporarily drop privileges and be able to
791 * regain them by swapping the real and effective uid.
792 */
793 SYSCALL_DEFINE1(setuid, uid_t, uid)
794 {
795 struct user_namespace *ns = current_user_ns();
796 const struct cred *old;
797 struct cred *new;
798 int retval;
799 kuid_t kuid;
800
801 kuid = make_kuid(ns, uid);
802 if (!uid_valid(kuid))
803 return -EINVAL;
804
805 new = prepare_creds();
806 if (!new)
807 return -ENOMEM;
808 old = current_cred();
809
810 retval = -EPERM;
811 if (nsown_capable(CAP_SETUID)) {
812 new->suid = new->uid = kuid;
813 if (!uid_eq(kuid, old->uid)) {
814 retval = set_user(new);
815 if (retval < 0)
816 goto error;
817 }
818 } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) {
819 goto error;
820 }
821
822 new->fsuid = new->euid = kuid;
823
824 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
825 if (retval < 0)
826 goto error;
827
828 return commit_creds(new);
829
830 error:
831 abort_creds(new);
832 return retval;
833 }
834
835
836 /*
837 * This function implements a generic ability to update ruid, euid,
838 * and suid. This allows you to implement the 4.4 compatible seteuid().
839 */
840 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
841 {
842 struct user_namespace *ns = current_user_ns();
843 const struct cred *old;
844 struct cred *new;
845 int retval;
846 kuid_t kruid, keuid, ksuid;
847
848 kruid = make_kuid(ns, ruid);
849 keuid = make_kuid(ns, euid);
850 ksuid = make_kuid(ns, suid);
851
852 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
853 return -EINVAL;
854
855 if ((euid != (uid_t) -1) && !uid_valid(keuid))
856 return -EINVAL;
857
858 if ((suid != (uid_t) -1) && !uid_valid(ksuid))
859 return -EINVAL;
860
861 new = prepare_creds();
862 if (!new)
863 return -ENOMEM;
864
865 old = current_cred();
866
867 retval = -EPERM;
868 if (!nsown_capable(CAP_SETUID)) {
869 if (ruid != (uid_t) -1 && !uid_eq(kruid, old->uid) &&
870 !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid))
871 goto error;
872 if (euid != (uid_t) -1 && !uid_eq(keuid, old->uid) &&
873 !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid))
874 goto error;
875 if (suid != (uid_t) -1 && !uid_eq(ksuid, old->uid) &&
876 !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid))
877 goto error;
878 }
879
880 if (ruid != (uid_t) -1) {
881 new->uid = kruid;
882 if (!uid_eq(kruid, old->uid)) {
883 retval = set_user(new);
884 if (retval < 0)
885 goto error;
886 }
887 }
888 if (euid != (uid_t) -1)
889 new->euid = keuid;
890 if (suid != (uid_t) -1)
891 new->suid = ksuid;
892 new->fsuid = new->euid;
893
894 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
895 if (retval < 0)
896 goto error;
897
898 return commit_creds(new);
899
900 error:
901 abort_creds(new);
902 return retval;
903 }
904
905 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
906 {
907 const struct cred *cred = current_cred();
908 int retval;
909 uid_t ruid, euid, suid;
910
911 ruid = from_kuid_munged(cred->user_ns, cred->uid);
912 euid = from_kuid_munged(cred->user_ns, cred->euid);
913 suid = from_kuid_munged(cred->user_ns, cred->suid);
914
915 if (!(retval = put_user(ruid, ruidp)) &&
916 !(retval = put_user(euid, euidp)))
917 retval = put_user(suid, suidp);
918
919 return retval;
920 }
921
922 /*
923 * Same as above, but for rgid, egid, sgid.
924 */
925 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
926 {
927 struct user_namespace *ns = current_user_ns();
928 const struct cred *old;
929 struct cred *new;
930 int retval;
931 kgid_t krgid, kegid, ksgid;
932
933 krgid = make_kgid(ns, rgid);
934 kegid = make_kgid(ns, egid);
935 ksgid = make_kgid(ns, sgid);
936
937 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
938 return -EINVAL;
939 if ((egid != (gid_t) -1) && !gid_valid(kegid))
940 return -EINVAL;
941 if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
942 return -EINVAL;
943
944 new = prepare_creds();
945 if (!new)
946 return -ENOMEM;
947 old = current_cred();
948
949 retval = -EPERM;
950 if (!nsown_capable(CAP_SETGID)) {
951 if (rgid != (gid_t) -1 && !gid_eq(krgid, old->gid) &&
952 !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid))
953 goto error;
954 if (egid != (gid_t) -1 && !gid_eq(kegid, old->gid) &&
955 !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid))
956 goto error;
957 if (sgid != (gid_t) -1 && !gid_eq(ksgid, old->gid) &&
958 !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid))
959 goto error;
960 }
961
962 if (rgid != (gid_t) -1)
963 new->gid = krgid;
964 if (egid != (gid_t) -1)
965 new->egid = kegid;
966 if (sgid != (gid_t) -1)
967 new->sgid = ksgid;
968 new->fsgid = new->egid;
969
970 return commit_creds(new);
971
972 error:
973 abort_creds(new);
974 return retval;
975 }
976
977 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
978 {
979 const struct cred *cred = current_cred();
980 int retval;
981 gid_t rgid, egid, sgid;
982
983 rgid = from_kgid_munged(cred->user_ns, cred->gid);
984 egid = from_kgid_munged(cred->user_ns, cred->egid);
985 sgid = from_kgid_munged(cred->user_ns, cred->sgid);
986
987 if (!(retval = put_user(rgid, rgidp)) &&
988 !(retval = put_user(egid, egidp)))
989 retval = put_user(sgid, sgidp);
990
991 return retval;
992 }
993
994
995 /*
996 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
997 * is used for "access()" and for the NFS daemon (letting nfsd stay at
998 * whatever uid it wants to). It normally shadows "euid", except when
999 * explicitly set by setfsuid() or for access..
1000 */
1001 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
1002 {
1003 const struct cred *old;
1004 struct cred *new;
1005 uid_t old_fsuid;
1006 kuid_t kuid;
1007
1008 old = current_cred();
1009 old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
1010
1011 kuid = make_kuid(old->user_ns, uid);
1012 if (!uid_valid(kuid))
1013 return old_fsuid;
1014
1015 new = prepare_creds();
1016 if (!new)
1017 return old_fsuid;
1018
1019 if (uid_eq(kuid, old->uid) || uid_eq(kuid, old->euid) ||
1020 uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
1021 nsown_capable(CAP_SETUID)) {
1022 if (!uid_eq(kuid, old->fsuid)) {
1023 new->fsuid = kuid;
1024 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
1025 goto change_okay;
1026 }
1027 }
1028
1029 abort_creds(new);
1030 return old_fsuid;
1031
1032 change_okay:
1033 commit_creds(new);
1034 return old_fsuid;
1035 }
1036
1037 /*
1038 * Samma på svenska..
1039 */
1040 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
1041 {
1042 const struct cred *old;
1043 struct cred *new;
1044 gid_t old_fsgid;
1045 kgid_t kgid;
1046
1047 old = current_cred();
1048 old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
1049
1050 kgid = make_kgid(old->user_ns, gid);
1051 if (!gid_valid(kgid))
1052 return old_fsgid;
1053
1054 new = prepare_creds();
1055 if (!new)
1056 return old_fsgid;
1057
1058 if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->egid) ||
1059 gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
1060 nsown_capable(CAP_SETGID)) {
1061 if (!gid_eq(kgid, old->fsgid)) {
1062 new->fsgid = kgid;
1063 goto change_okay;
1064 }
1065 }
1066
1067 abort_creds(new);
1068 return old_fsgid;
1069
1070 change_okay:
1071 commit_creds(new);
1072 return old_fsgid;
1073 }
1074
1075 /**
1076 * sys_getpid - return the thread group id of the current process
1077 *
1078 * Note, despite the name, this returns the tgid not the pid. The tgid and
1079 * the pid are identical unless CLONE_THREAD was specified on clone() in
1080 * which case the tgid is the same in all threads of the same group.
1081 *
1082 * This is SMP safe as current->tgid does not change.
1083 */
1084 SYSCALL_DEFINE0(getpid)
1085 {
1086 return task_tgid_vnr(current);
1087 }
1088
1089 /* Thread ID - the internal kernel "pid" */
1090 SYSCALL_DEFINE0(gettid)
1091 {
1092 return task_pid_vnr(current);
1093 }
1094
1095 /*
1096 * Accessing ->real_parent is not SMP-safe, it could
1097 * change from under us. However, we can use a stale
1098 * value of ->real_parent under rcu_read_lock(), see
1099 * release_task()->call_rcu(delayed_put_task_struct).
1100 */
1101 SYSCALL_DEFINE0(getppid)
1102 {
1103 int pid;
1104
1105 rcu_read_lock();
1106 pid = task_tgid_vnr(rcu_dereference(current->real_parent));
1107 rcu_read_unlock();
1108
1109 return pid;
1110 }
1111
1112 SYSCALL_DEFINE0(getuid)
1113 {
1114 /* Only we change this so SMP safe */
1115 return from_kuid_munged(current_user_ns(), current_uid());
1116 }
1117
1118 SYSCALL_DEFINE0(geteuid)
1119 {
1120 /* Only we change this so SMP safe */
1121 return from_kuid_munged(current_user_ns(), current_euid());
1122 }
1123
1124 SYSCALL_DEFINE0(getgid)
1125 {
1126 /* Only we change this so SMP safe */
1127 return from_kgid_munged(current_user_ns(), current_gid());
1128 }
1129
1130 SYSCALL_DEFINE0(getegid)
1131 {
1132 /* Only we change this so SMP safe */
1133 return from_kgid_munged(current_user_ns(), current_egid());
1134 }
1135
1136 void do_sys_times(struct tms *tms)
1137 {
1138 cputime_t tgutime, tgstime, cutime, cstime;
1139
1140 spin_lock_irq(&current->sighand->siglock);
1141 thread_group_cputime_adjusted(current, &tgutime, &tgstime);
1142 cutime = current->signal->cutime;
1143 cstime = current->signal->cstime;
1144 spin_unlock_irq(&current->sighand->siglock);
1145 tms->tms_utime = cputime_to_clock_t(tgutime);
1146 tms->tms_stime = cputime_to_clock_t(tgstime);
1147 tms->tms_cutime = cputime_to_clock_t(cutime);
1148 tms->tms_cstime = cputime_to_clock_t(cstime);
1149 }
1150
1151 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
1152 {
1153 if (tbuf) {
1154 struct tms tmp;
1155
1156 do_sys_times(&tmp);
1157 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1158 return -EFAULT;
1159 }
1160 force_successful_syscall_return();
1161 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1162 }
1163
1164 /*
1165 * This needs some heavy checking ...
1166 * I just haven't the stomach for it. I also don't fully
1167 * understand sessions/pgrp etc. Let somebody who does explain it.
1168 *
1169 * OK, I think I have the protection semantics right.... this is really
1170 * only important on a multi-user system anyway, to make sure one user
1171 * can't send a signal to a process owned by another. -TYT, 12/12/91
1172 *
1173 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1174 * LBT 04.03.94
1175 */
1176 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
1177 {
1178 struct task_struct *p;
1179 struct task_struct *group_leader = current->group_leader;
1180 struct pid *pgrp;
1181 int err;
1182
1183 if (!pid)
1184 pid = task_pid_vnr(group_leader);
1185 if (!pgid)
1186 pgid = pid;
1187 if (pgid < 0)
1188 return -EINVAL;
1189 rcu_read_lock();
1190
1191 /* From this point forward we keep holding onto the tasklist lock
1192 * so that our parent does not change from under us. -DaveM
1193 */
1194 write_lock_irq(&tasklist_lock);
1195
1196 err = -ESRCH;
1197 p = find_task_by_vpid(pid);
1198 if (!p)
1199 goto out;
1200
1201 err = -EINVAL;
1202 if (!thread_group_leader(p))
1203 goto out;
1204
1205 if (same_thread_group(p->real_parent, group_leader)) {
1206 err = -EPERM;
1207 if (task_session(p) != task_session(group_leader))
1208 goto out;
1209 err = -EACCES;
1210 if (p->did_exec)
1211 goto out;
1212 } else {
1213 err = -ESRCH;
1214 if (p != group_leader)
1215 goto out;
1216 }
1217
1218 err = -EPERM;
1219 if (p->signal->leader)
1220 goto out;
1221
1222 pgrp = task_pid(p);
1223 if (pgid != pid) {
1224 struct task_struct *g;
1225
1226 pgrp = find_vpid(pgid);
1227 g = pid_task(pgrp, PIDTYPE_PGID);
1228 if (!g || task_session(g) != task_session(group_leader))
1229 goto out;
1230 }
1231
1232 err = security_task_setpgid(p, pgid);
1233 if (err)
1234 goto out;
1235
1236 if (task_pgrp(p) != pgrp)
1237 change_pid(p, PIDTYPE_PGID, pgrp);
1238
1239 err = 0;
1240 out:
1241 /* All paths lead to here, thus we are safe. -DaveM */
1242 write_unlock_irq(&tasklist_lock);
1243 rcu_read_unlock();
1244 return err;
1245 }
1246
1247 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1248 {
1249 struct task_struct *p;
1250 struct pid *grp;
1251 int retval;
1252
1253 rcu_read_lock();
1254 if (!pid)
1255 grp = task_pgrp(current);
1256 else {
1257 retval = -ESRCH;
1258 p = find_task_by_vpid(pid);
1259 if (!p)
1260 goto out;
1261 grp = task_pgrp(p);
1262 if (!grp)
1263 goto out;
1264
1265 retval = security_task_getpgid(p);
1266 if (retval)
1267 goto out;
1268 }
1269 retval = pid_vnr(grp);
1270 out:
1271 rcu_read_unlock();
1272 return retval;
1273 }
1274
1275 #ifdef __ARCH_WANT_SYS_GETPGRP
1276
1277 SYSCALL_DEFINE0(getpgrp)
1278 {
1279 return sys_getpgid(0);
1280 }
1281
1282 #endif
1283
1284 SYSCALL_DEFINE1(getsid, pid_t, pid)
1285 {
1286 struct task_struct *p;
1287 struct pid *sid;
1288 int retval;
1289
1290 rcu_read_lock();
1291 if (!pid)
1292 sid = task_session(current);
1293 else {
1294 retval = -ESRCH;
1295 p = find_task_by_vpid(pid);
1296 if (!p)
1297 goto out;
1298 sid = task_session(p);
1299 if (!sid)
1300 goto out;
1301
1302 retval = security_task_getsid(p);
1303 if (retval)
1304 goto out;
1305 }
1306 retval = pid_vnr(sid);
1307 out:
1308 rcu_read_unlock();
1309 return retval;
1310 }
1311
1312 static void set_special_pids(struct pid *pid)
1313 {
1314 struct task_struct *curr = current->group_leader;
1315
1316 if (task_session(curr) != pid)
1317 change_pid(curr, PIDTYPE_SID, pid);
1318
1319 if (task_pgrp(curr) != pid)
1320 change_pid(curr, PIDTYPE_PGID, pid);
1321 }
1322
1323 SYSCALL_DEFINE0(setsid)
1324 {
1325 struct task_struct *group_leader = current->group_leader;
1326 struct pid *sid = task_pid(group_leader);
1327 pid_t session = pid_vnr(sid);
1328 int err = -EPERM;
1329
1330 write_lock_irq(&tasklist_lock);
1331 /* Fail if I am already a session leader */
1332 if (group_leader->signal->leader)
1333 goto out;
1334
1335 /* Fail if a process group id already exists that equals the
1336 * proposed session id.
1337 */
1338 if (pid_task(sid, PIDTYPE_PGID))
1339 goto out;
1340
1341 group_leader->signal->leader = 1;
1342 set_special_pids(sid);
1343
1344 proc_clear_tty(group_leader);
1345
1346 err = session;
1347 out:
1348 write_unlock_irq(&tasklist_lock);
1349 if (err > 0) {
1350 proc_sid_connector(group_leader);
1351 sched_autogroup_create_attach(group_leader);
1352 }
1353 return err;
1354 }
1355
1356 DECLARE_RWSEM(uts_sem);
1357
1358 #ifdef COMPAT_UTS_MACHINE
1359 #define override_architecture(name) \
1360 (personality(current->personality) == PER_LINUX32 && \
1361 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1362 sizeof(COMPAT_UTS_MACHINE)))
1363 #else
1364 #define override_architecture(name) 0
1365 #endif
1366
1367 /*
1368 * Work around broken programs that cannot handle "Linux 3.0".
1369 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1370 */
1371 static int override_release(char __user *release, size_t len)
1372 {
1373 int ret = 0;
1374
1375 if (current->personality & UNAME26) {
1376 const char *rest = UTS_RELEASE;
1377 char buf[65] = { 0 };
1378 int ndots = 0;
1379 unsigned v;
1380 size_t copy;
1381
1382 while (*rest) {
1383 if (*rest == '.' && ++ndots >= 3)
1384 break;
1385 if (!isdigit(*rest) && *rest != '.')
1386 break;
1387 rest++;
1388 }
1389 v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 40;
1390 copy = clamp_t(size_t, len, 1, sizeof(buf));
1391 copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1392 ret = copy_to_user(release, buf, copy + 1);
1393 }
1394 return ret;
1395 }
1396
1397 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1398 {
1399 int errno = 0;
1400
1401 down_read(&uts_sem);
1402 if (copy_to_user(name, utsname(), sizeof *name))
1403 errno = -EFAULT;
1404 up_read(&uts_sem);
1405
1406 if (!errno && override_release(name->release, sizeof(name->release)))
1407 errno = -EFAULT;
1408 if (!errno && override_architecture(name))
1409 errno = -EFAULT;
1410 return errno;
1411 }
1412
1413 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1414 /*
1415 * Old cruft
1416 */
1417 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1418 {
1419 int error = 0;
1420
1421 if (!name)
1422 return -EFAULT;
1423
1424 down_read(&uts_sem);
1425 if (copy_to_user(name, utsname(), sizeof(*name)))
1426 error = -EFAULT;
1427 up_read(&uts_sem);
1428
1429 if (!error && override_release(name->release, sizeof(name->release)))
1430 error = -EFAULT;
1431 if (!error && override_architecture(name))
1432 error = -EFAULT;
1433 return error;
1434 }
1435
1436 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1437 {
1438 int error;
1439
1440 if (!name)
1441 return -EFAULT;
1442 if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
1443 return -EFAULT;
1444
1445 down_read(&uts_sem);
1446 error = __copy_to_user(&name->sysname, &utsname()->sysname,
1447 __OLD_UTS_LEN);
1448 error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
1449 error |= __copy_to_user(&name->nodename, &utsname()->nodename,
1450 __OLD_UTS_LEN);
1451 error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
1452 error |= __copy_to_user(&name->release, &utsname()->release,
1453 __OLD_UTS_LEN);
1454 error |= __put_user(0, name->release + __OLD_UTS_LEN);
1455 error |= __copy_to_user(&name->version, &utsname()->version,
1456 __OLD_UTS_LEN);
1457 error |= __put_user(0, name->version + __OLD_UTS_LEN);
1458 error |= __copy_to_user(&name->machine, &utsname()->machine,
1459 __OLD_UTS_LEN);
1460 error |= __put_user(0, name->machine + __OLD_UTS_LEN);
1461 up_read(&uts_sem);
1462
1463 if (!error && override_architecture(name))
1464 error = -EFAULT;
1465 if (!error && override_release(name->release, sizeof(name->release)))
1466 error = -EFAULT;
1467 return error ? -EFAULT : 0;
1468 }
1469 #endif
1470
1471 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1472 {
1473 int errno;
1474 char tmp[__NEW_UTS_LEN];
1475
1476 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1477 return -EPERM;
1478
1479 if (len < 0 || len > __NEW_UTS_LEN)
1480 return -EINVAL;
1481 down_write(&uts_sem);
1482 errno = -EFAULT;
1483 if (!copy_from_user(tmp, name, len)) {
1484 struct new_utsname *u = utsname();
1485
1486 memcpy(u->nodename, tmp, len);
1487 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1488 errno = 0;
1489 uts_proc_notify(UTS_PROC_HOSTNAME);
1490 }
1491 up_write(&uts_sem);
1492 return errno;
1493 }
1494
1495 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1496
1497 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1498 {
1499 int i, errno;
1500 struct new_utsname *u;
1501
1502 if (len < 0)
1503 return -EINVAL;
1504 down_read(&uts_sem);
1505 u = utsname();
1506 i = 1 + strlen(u->nodename);
1507 if (i > len)
1508 i = len;
1509 errno = 0;
1510 if (copy_to_user(name, u->nodename, i))
1511 errno = -EFAULT;
1512 up_read(&uts_sem);
1513 return errno;
1514 }
1515
1516 #endif
1517
1518 /*
1519 * Only setdomainname; getdomainname can be implemented by calling
1520 * uname()
1521 */
1522 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1523 {
1524 int errno;
1525 char tmp[__NEW_UTS_LEN];
1526
1527 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1528 return -EPERM;
1529 if (len < 0 || len > __NEW_UTS_LEN)
1530 return -EINVAL;
1531
1532 down_write(&uts_sem);
1533 errno = -EFAULT;
1534 if (!copy_from_user(tmp, name, len)) {
1535 struct new_utsname *u = utsname();
1536
1537 memcpy(u->domainname, tmp, len);
1538 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1539 errno = 0;
1540 uts_proc_notify(UTS_PROC_DOMAINNAME);
1541 }
1542 up_write(&uts_sem);
1543 return errno;
1544 }
1545
1546 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1547 {
1548 struct rlimit value;
1549 int ret;
1550
1551 ret = do_prlimit(current, resource, NULL, &value);
1552 if (!ret)
1553 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1554
1555 return ret;
1556 }
1557
1558 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1559
1560 /*
1561 * Back compatibility for getrlimit. Needed for some apps.
1562 */
1563
1564 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1565 struct rlimit __user *, rlim)
1566 {
1567 struct rlimit x;
1568 if (resource >= RLIM_NLIMITS)
1569 return -EINVAL;
1570
1571 task_lock(current->group_leader);
1572 x = current->signal->rlim[resource];
1573 task_unlock(current->group_leader);
1574 if (x.rlim_cur > 0x7FFFFFFF)
1575 x.rlim_cur = 0x7FFFFFFF;
1576 if (x.rlim_max > 0x7FFFFFFF)
1577 x.rlim_max = 0x7FFFFFFF;
1578 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1579 }
1580
1581 #endif
1582
1583 static inline bool rlim64_is_infinity(__u64 rlim64)
1584 {
1585 #if BITS_PER_LONG < 64
1586 return rlim64 >= ULONG_MAX;
1587 #else
1588 return rlim64 == RLIM64_INFINITY;
1589 #endif
1590 }
1591
1592 static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1593 {
1594 if (rlim->rlim_cur == RLIM_INFINITY)
1595 rlim64->rlim_cur = RLIM64_INFINITY;
1596 else
1597 rlim64->rlim_cur = rlim->rlim_cur;
1598 if (rlim->rlim_max == RLIM_INFINITY)
1599 rlim64->rlim_max = RLIM64_INFINITY;
1600 else
1601 rlim64->rlim_max = rlim->rlim_max;
1602 }
1603
1604 static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1605 {
1606 if (rlim64_is_infinity(rlim64->rlim_cur))
1607 rlim->rlim_cur = RLIM_INFINITY;
1608 else
1609 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1610 if (rlim64_is_infinity(rlim64->rlim_max))
1611 rlim->rlim_max = RLIM_INFINITY;
1612 else
1613 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1614 }
1615
1616 /* make sure you are allowed to change @tsk limits before calling this */
1617 int do_prlimit(struct task_struct *tsk, unsigned int resource,
1618 struct rlimit *new_rlim, struct rlimit *old_rlim)
1619 {
1620 struct rlimit *rlim;
1621 int retval = 0;
1622
1623 if (resource >= RLIM_NLIMITS)
1624 return -EINVAL;
1625 if (new_rlim) {
1626 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1627 return -EINVAL;
1628 if (resource == RLIMIT_NOFILE &&
1629 new_rlim->rlim_max > sysctl_nr_open)
1630 return -EPERM;
1631 }
1632
1633 /* protect tsk->signal and tsk->sighand from disappearing */
1634 read_lock(&tasklist_lock);
1635 if (!tsk->sighand) {
1636 retval = -ESRCH;
1637 goto out;
1638 }
1639
1640 rlim = tsk->signal->rlim + resource;
1641 task_lock(tsk->group_leader);
1642 if (new_rlim) {
1643 /* Keep the capable check against init_user_ns until
1644 cgroups can contain all limits */
1645 if (new_rlim->rlim_max > rlim->rlim_max &&
1646 !capable(CAP_SYS_RESOURCE))
1647 retval = -EPERM;
1648 if (!retval)
1649 retval = security_task_setrlimit(tsk->group_leader,
1650 resource, new_rlim);
1651 if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) {
1652 /*
1653 * The caller is asking for an immediate RLIMIT_CPU
1654 * expiry. But we use the zero value to mean "it was
1655 * never set". So let's cheat and make it one second
1656 * instead
1657 */
1658 new_rlim->rlim_cur = 1;
1659 }
1660 }
1661 if (!retval) {
1662 if (old_rlim)
1663 *old_rlim = *rlim;
1664 if (new_rlim)
1665 *rlim = *new_rlim;
1666 }
1667 task_unlock(tsk->group_leader);
1668
1669 /*
1670 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1671 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1672 * very long-standing error, and fixing it now risks breakage of
1673 * applications, so we live with it
1674 */
1675 if (!retval && new_rlim && resource == RLIMIT_CPU &&
1676 new_rlim->rlim_cur != RLIM_INFINITY)
1677 update_rlimit_cpu(tsk, new_rlim->rlim_cur);
1678 out:
1679 read_unlock(&tasklist_lock);
1680 return retval;
1681 }
1682
1683 /* rcu lock must be held */
1684 static int check_prlimit_permission(struct task_struct *task)
1685 {
1686 const struct cred *cred = current_cred(), *tcred;
1687
1688 if (current == task)
1689 return 0;
1690
1691 tcred = __task_cred(task);
1692 if (uid_eq(cred->uid, tcred->euid) &&
1693 uid_eq(cred->uid, tcred->suid) &&
1694 uid_eq(cred->uid, tcred->uid) &&
1695 gid_eq(cred->gid, tcred->egid) &&
1696 gid_eq(cred->gid, tcred->sgid) &&
1697 gid_eq(cred->gid, tcred->gid))
1698 return 0;
1699 if (ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
1700 return 0;
1701
1702 return -EPERM;
1703 }
1704
1705 SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1706 const struct rlimit64 __user *, new_rlim,
1707 struct rlimit64 __user *, old_rlim)
1708 {
1709 struct rlimit64 old64, new64;
1710 struct rlimit old, new;
1711 struct task_struct *tsk;
1712 int ret;
1713
1714 if (new_rlim) {
1715 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1716 return -EFAULT;
1717 rlim64_to_rlim(&new64, &new);
1718 }
1719
1720 rcu_read_lock();
1721 tsk = pid ? find_task_by_vpid(pid) : current;
1722 if (!tsk) {
1723 rcu_read_unlock();
1724 return -ESRCH;
1725 }
1726 ret = check_prlimit_permission(tsk);
1727 if (ret) {
1728 rcu_read_unlock();
1729 return ret;
1730 }
1731 get_task_struct(tsk);
1732 rcu_read_unlock();
1733
1734 ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1735 old_rlim ? &old : NULL);
1736
1737 if (!ret && old_rlim) {
1738 rlim_to_rlim64(&old, &old64);
1739 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1740 ret = -EFAULT;
1741 }
1742
1743 put_task_struct(tsk);
1744 return ret;
1745 }
1746
1747 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1748 {
1749 struct rlimit new_rlim;
1750
1751 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1752 return -EFAULT;
1753 return do_prlimit(current, resource, &new_rlim, NULL);
1754 }
1755
1756 /*
1757 * It would make sense to put struct rusage in the task_struct,
1758 * except that would make the task_struct be *really big*. After
1759 * task_struct gets moved into malloc'ed memory, it would
1760 * make sense to do this. It will make moving the rest of the information
1761 * a lot simpler! (Which we're not doing right now because we're not
1762 * measuring them yet).
1763 *
1764 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1765 * races with threads incrementing their own counters. But since word
1766 * reads are atomic, we either get new values or old values and we don't
1767 * care which for the sums. We always take the siglock to protect reading
1768 * the c* fields from p->signal from races with exit.c updating those
1769 * fields when reaping, so a sample either gets all the additions of a
1770 * given child after it's reaped, or none so this sample is before reaping.
1771 *
1772 * Locking:
1773 * We need to take the siglock for CHILDEREN, SELF and BOTH
1774 * for the cases current multithreaded, non-current single threaded
1775 * non-current multithreaded. Thread traversal is now safe with
1776 * the siglock held.
1777 * Strictly speaking, we donot need to take the siglock if we are current and
1778 * single threaded, as no one else can take our signal_struct away, no one
1779 * else can reap the children to update signal->c* counters, and no one else
1780 * can race with the signal-> fields. If we do not take any lock, the
1781 * signal-> fields could be read out of order while another thread was just
1782 * exiting. So we should place a read memory barrier when we avoid the lock.
1783 * On the writer side, write memory barrier is implied in __exit_signal
1784 * as __exit_signal releases the siglock spinlock after updating the signal->
1785 * fields. But we don't do this yet to keep things simple.
1786 *
1787 */
1788
1789 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1790 {
1791 r->ru_nvcsw += t->nvcsw;
1792 r->ru_nivcsw += t->nivcsw;
1793 r->ru_minflt += t->min_flt;
1794 r->ru_majflt += t->maj_flt;
1795 r->ru_inblock += task_io_get_inblock(t);
1796 r->ru_oublock += task_io_get_oublock(t);
1797 }
1798
1799 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1800 {
1801 struct task_struct *t;
1802 unsigned long flags;
1803 cputime_t tgutime, tgstime, utime, stime;
1804 unsigned long maxrss = 0;
1805
1806 memset((char *) r, 0, sizeof *r);
1807 utime = stime = 0;
1808
1809 if (who == RUSAGE_THREAD) {
1810 task_cputime_adjusted(current, &utime, &stime);
1811 accumulate_thread_rusage(p, r);
1812 maxrss = p->signal->maxrss;
1813 goto out;
1814 }
1815
1816 if (!lock_task_sighand(p, &flags))
1817 return;
1818
1819 switch (who) {
1820 case RUSAGE_BOTH:
1821 case RUSAGE_CHILDREN:
1822 utime = p->signal->cutime;
1823 stime = p->signal->cstime;
1824 r->ru_nvcsw = p->signal->cnvcsw;
1825 r->ru_nivcsw = p->signal->cnivcsw;
1826 r->ru_minflt = p->signal->cmin_flt;
1827 r->ru_majflt = p->signal->cmaj_flt;
1828 r->ru_inblock = p->signal->cinblock;
1829 r->ru_oublock = p->signal->coublock;
1830 maxrss = p->signal->cmaxrss;
1831
1832 if (who == RUSAGE_CHILDREN)
1833 break;
1834
1835 case RUSAGE_SELF:
1836 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1837 utime += tgutime;
1838 stime += tgstime;
1839 r->ru_nvcsw += p->signal->nvcsw;
1840 r->ru_nivcsw += p->signal->nivcsw;
1841 r->ru_minflt += p->signal->min_flt;
1842 r->ru_majflt += p->signal->maj_flt;
1843 r->ru_inblock += p->signal->inblock;
1844 r->ru_oublock += p->signal->oublock;
1845 if (maxrss < p->signal->maxrss)
1846 maxrss = p->signal->maxrss;
1847 t = p;
1848 do {
1849 accumulate_thread_rusage(t, r);
1850 t = next_thread(t);
1851 } while (t != p);
1852 break;
1853
1854 default:
1855 BUG();
1856 }
1857 unlock_task_sighand(p, &flags);
1858
1859 out:
1860 cputime_to_timeval(utime, &r->ru_utime);
1861 cputime_to_timeval(stime, &r->ru_stime);
1862
1863 if (who != RUSAGE_CHILDREN) {
1864 struct mm_struct *mm = get_task_mm(p);
1865 if (mm) {
1866 setmax_mm_hiwater_rss(&maxrss, mm);
1867 mmput(mm);
1868 }
1869 }
1870 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1871 }
1872
1873 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1874 {
1875 struct rusage r;
1876 k_getrusage(p, who, &r);
1877 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1878 }
1879
1880 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1881 {
1882 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1883 who != RUSAGE_THREAD)
1884 return -EINVAL;
1885 return getrusage(current, who, ru);
1886 }
1887
1888 #ifdef CONFIG_COMPAT
1889 COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru)
1890 {
1891 struct rusage r;
1892
1893 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1894 who != RUSAGE_THREAD)
1895 return -EINVAL;
1896
1897 k_getrusage(current, who, &r);
1898 return put_compat_rusage(&r, ru);
1899 }
1900 #endif
1901
1902 SYSCALL_DEFINE1(umask, int, mask)
1903 {
1904 mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1905 return mask;
1906 }
1907
1908 static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1909 {
1910 struct fd exe;
1911 struct inode *inode;
1912 int err;
1913
1914 exe = fdget(fd);
1915 if (!exe.file)
1916 return -EBADF;
1917
1918 inode = file_inode(exe.file);
1919
1920 /*
1921 * Because the original mm->exe_file points to executable file, make
1922 * sure that this one is executable as well, to avoid breaking an
1923 * overall picture.
1924 */
1925 err = -EACCES;
1926 if (!S_ISREG(inode->i_mode) ||
1927 exe.file->f_path.mnt->mnt_flags & MNT_NOEXEC)
1928 goto exit;
1929
1930 err = inode_permission(inode, MAY_EXEC);
1931 if (err)
1932 goto exit;
1933
1934 down_write(&mm->mmap_sem);
1935
1936 /*
1937 * Forbid mm->exe_file change if old file still mapped.
1938 */
1939 err = -EBUSY;
1940 if (mm->exe_file) {
1941 struct vm_area_struct *vma;
1942
1943 for (vma = mm->mmap; vma; vma = vma->vm_next)
1944 if (vma->vm_file &&
1945 path_equal(&vma->vm_file->f_path,
1946 &mm->exe_file->f_path))
1947 goto exit_unlock;
1948 }
1949
1950 /*
1951 * The symlink can be changed only once, just to disallow arbitrary
1952 * transitions malicious software might bring in. This means one
1953 * could make a snapshot over all processes running and monitor
1954 * /proc/pid/exe changes to notice unusual activity if needed.
1955 */
1956 err = -EPERM;
1957 if (test_and_set_bit(MMF_EXE_FILE_CHANGED, &mm->flags))
1958 goto exit_unlock;
1959
1960 err = 0;
1961 set_mm_exe_file(mm, exe.file); /* this grabs a reference to exe.file */
1962 exit_unlock:
1963 up_write(&mm->mmap_sem);
1964
1965 exit:
1966 fdput(exe);
1967 return err;
1968 }
1969
1970 static int prctl_set_mm(int opt, unsigned long addr,
1971 unsigned long arg4, unsigned long arg5)
1972 {
1973 unsigned long rlim = rlimit(RLIMIT_DATA);
1974 struct mm_struct *mm = current->mm;
1975 struct vm_area_struct *vma;
1976 int error;
1977
1978 if (arg5 || (arg4 && opt != PR_SET_MM_AUXV))
1979 return -EINVAL;
1980
1981 if (!capable(CAP_SYS_RESOURCE))
1982 return -EPERM;
1983
1984 if (opt == PR_SET_MM_EXE_FILE)
1985 return prctl_set_mm_exe_file(mm, (unsigned int)addr);
1986
1987 if (addr >= TASK_SIZE || addr < mmap_min_addr)
1988 return -EINVAL;
1989
1990 error = -EINVAL;
1991
1992 down_read(&mm->mmap_sem);
1993 vma = find_vma(mm, addr);
1994
1995 switch (opt) {
1996 case PR_SET_MM_START_CODE:
1997 mm->start_code = addr;
1998 break;
1999 case PR_SET_MM_END_CODE:
2000 mm->end_code = addr;
2001 break;
2002 case PR_SET_MM_START_DATA:
2003 mm->start_data = addr;
2004 break;
2005 case PR_SET_MM_END_DATA:
2006 mm->end_data = addr;
2007 break;
2008
2009 case PR_SET_MM_START_BRK:
2010 if (addr <= mm->end_data)
2011 goto out;
2012
2013 if (rlim < RLIM_INFINITY &&
2014 (mm->brk - addr) +
2015 (mm->end_data - mm->start_data) > rlim)
2016 goto out;
2017
2018 mm->start_brk = addr;
2019 break;
2020
2021 case PR_SET_MM_BRK:
2022 if (addr <= mm->end_data)
2023 goto out;
2024
2025 if (rlim < RLIM_INFINITY &&
2026 (addr - mm->start_brk) +
2027 (mm->end_data - mm->start_data) > rlim)
2028 goto out;
2029
2030 mm->brk = addr;
2031 break;
2032
2033 /*
2034 * If command line arguments and environment
2035 * are placed somewhere else on stack, we can
2036 * set them up here, ARG_START/END to setup
2037 * command line argumets and ENV_START/END
2038 * for environment.
2039 */
2040 case PR_SET_MM_START_STACK:
2041 case PR_SET_MM_ARG_START:
2042 case PR_SET_MM_ARG_END:
2043 case PR_SET_MM_ENV_START:
2044 case PR_SET_MM_ENV_END:
2045 if (!vma) {
2046 error = -EFAULT;
2047 goto out;
2048 }
2049 if (opt == PR_SET_MM_START_STACK)
2050 mm->start_stack = addr;
2051 else if (opt == PR_SET_MM_ARG_START)
2052 mm->arg_start = addr;
2053 else if (opt == PR_SET_MM_ARG_END)
2054 mm->arg_end = addr;
2055 else if (opt == PR_SET_MM_ENV_START)
2056 mm->env_start = addr;
2057 else if (opt == PR_SET_MM_ENV_END)
2058 mm->env_end = addr;
2059 break;
2060
2061 /*
2062 * This doesn't move auxiliary vector itself
2063 * since it's pinned to mm_struct, but allow
2064 * to fill vector with new values. It's up
2065 * to a caller to provide sane values here
2066 * otherwise user space tools which use this
2067 * vector might be unhappy.
2068 */
2069 case PR_SET_MM_AUXV: {
2070 unsigned long user_auxv[AT_VECTOR_SIZE];
2071
2072 if (arg4 > sizeof(user_auxv))
2073 goto out;
2074 up_read(&mm->mmap_sem);
2075
2076 if (copy_from_user(user_auxv, (const void __user *)addr, arg4))
2077 return -EFAULT;
2078
2079 /* Make sure the last entry is always AT_NULL */
2080 user_auxv[AT_VECTOR_SIZE - 2] = 0;
2081 user_auxv[AT_VECTOR_SIZE - 1] = 0;
2082
2083 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
2084
2085 task_lock(current);
2086 memcpy(mm->saved_auxv, user_auxv, arg4);
2087 task_unlock(current);
2088
2089 return 0;
2090 }
2091 default:
2092 goto out;
2093 }
2094
2095 error = 0;
2096 out:
2097 up_read(&mm->mmap_sem);
2098 return error;
2099 }
2100
2101 #ifdef CONFIG_CHECKPOINT_RESTORE
2102 static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2103 {
2104 return put_user(me->clear_child_tid, tid_addr);
2105 }
2106 #else
2107 static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2108 {
2109 return -EINVAL;
2110 }
2111 #endif
2112
2113 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2114 unsigned long, arg4, unsigned long, arg5)
2115 {
2116 struct task_struct *me = current;
2117 unsigned char comm[sizeof(me->comm)];
2118 long error;
2119
2120 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2121 if (error != -ENOSYS)
2122 return error;
2123
2124 error = 0;
2125 switch (option) {
2126 case PR_SET_PDEATHSIG:
2127 if (!valid_signal(arg2)) {
2128 error = -EINVAL;
2129 break;
2130 }
2131 me->pdeath_signal = arg2;
2132 break;
2133 case PR_GET_PDEATHSIG:
2134 error = put_user(me->pdeath_signal, (int __user *)arg2);
2135 break;
2136 case PR_GET_DUMPABLE:
2137 error = get_dumpable(me->mm);
2138 break;
2139 case PR_SET_DUMPABLE:
2140 if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) {
2141 error = -EINVAL;
2142 break;
2143 }
2144 set_dumpable(me->mm, arg2);
2145 break;
2146
2147 case PR_SET_UNALIGN:
2148 error = SET_UNALIGN_CTL(me, arg2);
2149 break;
2150 case PR_GET_UNALIGN:
2151 error = GET_UNALIGN_CTL(me, arg2);
2152 break;
2153 case PR_SET_FPEMU:
2154 error = SET_FPEMU_CTL(me, arg2);
2155 break;
2156 case PR_GET_FPEMU:
2157 error = GET_FPEMU_CTL(me, arg2);
2158 break;
2159 case PR_SET_FPEXC:
2160 error = SET_FPEXC_CTL(me, arg2);
2161 break;
2162 case PR_GET_FPEXC:
2163 error = GET_FPEXC_CTL(me, arg2);
2164 break;
2165 case PR_GET_TIMING:
2166 error = PR_TIMING_STATISTICAL;
2167 break;
2168 case PR_SET_TIMING:
2169 if (arg2 != PR_TIMING_STATISTICAL)
2170 error = -EINVAL;
2171 break;
2172 case PR_SET_NAME:
2173 comm[sizeof(me->comm) - 1] = 0;
2174 if (strncpy_from_user(comm, (char __user *)arg2,
2175 sizeof(me->comm) - 1) < 0)
2176 return -EFAULT;
2177 set_task_comm(me, comm);
2178 proc_comm_connector(me);
2179 break;
2180 case PR_GET_NAME:
2181 get_task_comm(comm, me);
2182 if (copy_to_user((char __user *)arg2, comm, sizeof(comm)))
2183 return -EFAULT;
2184 break;
2185 case PR_GET_ENDIAN:
2186 error = GET_ENDIAN(me, arg2);
2187 break;
2188 case PR_SET_ENDIAN:
2189 error = SET_ENDIAN(me, arg2);
2190 break;
2191 case PR_GET_SECCOMP:
2192 error = prctl_get_seccomp();
2193 break;
2194 case PR_SET_SECCOMP:
2195 error = prctl_set_seccomp(arg2, (char __user *)arg3);
2196 break;
2197 case PR_GET_TSC:
2198 error = GET_TSC_CTL(arg2);
2199 break;
2200 case PR_SET_TSC:
2201 error = SET_TSC_CTL(arg2);
2202 break;
2203 case PR_TASK_PERF_EVENTS_DISABLE:
2204 error = perf_event_task_disable();
2205 break;
2206 case PR_TASK_PERF_EVENTS_ENABLE:
2207 error = perf_event_task_enable();
2208 break;
2209 case PR_GET_TIMERSLACK:
2210 error = current->timer_slack_ns;
2211 break;
2212 case PR_SET_TIMERSLACK:
2213 if (arg2 <= 0)
2214 current->timer_slack_ns =
2215 current->default_timer_slack_ns;
2216 else
2217 current->timer_slack_ns = arg2;
2218 break;
2219 case PR_MCE_KILL:
2220 if (arg4 | arg5)
2221 return -EINVAL;
2222 switch (arg2) {
2223 case PR_MCE_KILL_CLEAR:
2224 if (arg3 != 0)
2225 return -EINVAL;
2226 current->flags &= ~PF_MCE_PROCESS;
2227 break;
2228 case PR_MCE_KILL_SET:
2229 current->flags |= PF_MCE_PROCESS;
2230 if (arg3 == PR_MCE_KILL_EARLY)
2231 current->flags |= PF_MCE_EARLY;
2232 else if (arg3 == PR_MCE_KILL_LATE)
2233 current->flags &= ~PF_MCE_EARLY;
2234 else if (arg3 == PR_MCE_KILL_DEFAULT)
2235 current->flags &=
2236 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
2237 else
2238 return -EINVAL;
2239 break;
2240 default:
2241 return -EINVAL;
2242 }
2243 break;
2244 case PR_MCE_KILL_GET:
2245 if (arg2 | arg3 | arg4 | arg5)
2246 return -EINVAL;
2247 if (current->flags & PF_MCE_PROCESS)
2248 error = (current->flags & PF_MCE_EARLY) ?
2249 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2250 else
2251 error = PR_MCE_KILL_DEFAULT;
2252 break;
2253 case PR_SET_MM:
2254 error = prctl_set_mm(arg2, arg3, arg4, arg5);
2255 break;
2256 case PR_GET_TID_ADDRESS:
2257 error = prctl_get_tid_address(me, (int __user **)arg2);
2258 break;
2259 case PR_SET_CHILD_SUBREAPER:
2260 me->signal->is_child_subreaper = !!arg2;
2261 break;
2262 case PR_GET_CHILD_SUBREAPER:
2263 error = put_user(me->signal->is_child_subreaper,
2264 (int __user *)arg2);
2265 break;
2266 case PR_SET_NO_NEW_PRIVS:
2267 if (arg2 != 1 || arg3 || arg4 || arg5)
2268 return -EINVAL;
2269
2270 current->no_new_privs = 1;
2271 break;
2272 case PR_GET_NO_NEW_PRIVS:
2273 if (arg2 || arg3 || arg4 || arg5)
2274 return -EINVAL;
2275 return current->no_new_privs ? 1 : 0;
2276 default:
2277 error = -EINVAL;
2278 break;
2279 }
2280 return error;
2281 }
2282
2283 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2284 struct getcpu_cache __user *, unused)
2285 {
2286 int err = 0;
2287 int cpu = raw_smp_processor_id();
2288 if (cpup)
2289 err |= put_user(cpu, cpup);
2290 if (nodep)
2291 err |= put_user(cpu_to_node(cpu), nodep);
2292 return err ? -EFAULT : 0;
2293 }
2294
2295 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
2296
2297 static int __orderly_poweroff(bool force)
2298 {
2299 char **argv;
2300 static char *envp[] = {
2301 "HOME=/",
2302 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
2303 NULL
2304 };
2305 int ret;
2306
2307 argv = argv_split(GFP_KERNEL, poweroff_cmd, NULL);
2308 if (argv) {
2309 ret = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_EXEC);
2310 argv_free(argv);
2311 } else {
2312 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
2313 __func__, poweroff_cmd);
2314 ret = -ENOMEM;
2315 }
2316
2317 if (ret && force) {
2318 printk(KERN_WARNING "Failed to start orderly shutdown: "
2319 "forcing the issue\n");
2320 /*
2321 * I guess this should try to kick off some daemon to sync and
2322 * poweroff asap. Or not even bother syncing if we're doing an
2323 * emergency shutdown?
2324 */
2325 emergency_sync();
2326 kernel_power_off();
2327 }
2328
2329 return ret;
2330 }
2331
2332 static bool poweroff_force;
2333
2334 static void poweroff_work_func(struct work_struct *work)
2335 {
2336 __orderly_poweroff(poweroff_force);
2337 }
2338
2339 static DECLARE_WORK(poweroff_work, poweroff_work_func);
2340
2341 /**
2342 * orderly_poweroff - Trigger an orderly system poweroff
2343 * @force: force poweroff if command execution fails
2344 *
2345 * This may be called from any context to trigger a system shutdown.
2346 * If the orderly shutdown fails, it will force an immediate shutdown.
2347 */
2348 int orderly_poweroff(bool force)
2349 {
2350 if (force) /* do not override the pending "true" */
2351 poweroff_force = true;
2352 schedule_work(&poweroff_work);
2353 return 0;
2354 }
2355 EXPORT_SYMBOL_GPL(orderly_poweroff);
2356
2357 /**
2358 * do_sysinfo - fill in sysinfo struct
2359 * @info: pointer to buffer to fill
2360 */
2361 static int do_sysinfo(struct sysinfo *info)
2362 {
2363 unsigned long mem_total, sav_total;
2364 unsigned int mem_unit, bitcount;
2365 struct timespec tp;
2366
2367 memset(info, 0, sizeof(struct sysinfo));
2368
2369 get_monotonic_boottime(&tp);
2370 info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
2371
2372 get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
2373
2374 info->procs = nr_threads;
2375
2376 si_meminfo(info);
2377 si_swapinfo(info);
2378
2379 /*
2380 * If the sum of all the available memory (i.e. ram + swap)
2381 * is less than can be stored in a 32 bit unsigned long then
2382 * we can be binary compatible with 2.2.x kernels. If not,
2383 * well, in that case 2.2.x was broken anyways...
2384 *
2385 * -Erik Andersen <andersee@debian.org>
2386 */
2387
2388 mem_total = info->totalram + info->totalswap;
2389 if (mem_total < info->totalram || mem_total < info->totalswap)
2390 goto out;
2391 bitcount = 0;
2392 mem_unit = info->mem_unit;
2393 while (mem_unit > 1) {
2394 bitcount++;
2395 mem_unit >>= 1;
2396 sav_total = mem_total;
2397 mem_total <<= 1;
2398 if (mem_total < sav_total)
2399 goto out;
2400 }
2401
2402 /*
2403 * If mem_total did not overflow, multiply all memory values by
2404 * info->mem_unit and set it to 1. This leaves things compatible
2405 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2406 * kernels...
2407 */
2408
2409 info->mem_unit = 1;
2410 info->totalram <<= bitcount;
2411 info->freeram <<= bitcount;
2412 info->sharedram <<= bitcount;
2413 info->bufferram <<= bitcount;
2414 info->totalswap <<= bitcount;
2415 info->freeswap <<= bitcount;
2416 info->totalhigh <<= bitcount;
2417 info->freehigh <<= bitcount;
2418
2419 out:
2420 return 0;
2421 }
2422
2423 SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
2424 {
2425 struct sysinfo val;
2426
2427 do_sysinfo(&val);
2428
2429 if (copy_to_user(info, &val, sizeof(struct sysinfo)))
2430 return -EFAULT;
2431
2432 return 0;
2433 }
2434
2435 #ifdef CONFIG_COMPAT
2436 struct compat_sysinfo {
2437 s32 uptime;
2438 u32 loads[3];
2439 u32 totalram;
2440 u32 freeram;
2441 u32 sharedram;
2442 u32 bufferram;
2443 u32 totalswap;
2444 u32 freeswap;
2445 u16 procs;
2446 u16 pad;
2447 u32 totalhigh;
2448 u32 freehigh;
2449 u32 mem_unit;
2450 char _f[20-2*sizeof(u32)-sizeof(int)];
2451 };
2452
2453 COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2454 {
2455 struct sysinfo s;
2456
2457 do_sysinfo(&s);
2458
2459 /* Check to see if any memory value is too large for 32-bit and scale
2460 * down if needed
2461 */
2462 if ((s.totalram >> 32) || (s.totalswap >> 32)) {
2463 int bitcount = 0;
2464
2465 while (s.mem_unit < PAGE_SIZE) {
2466 s.mem_unit <<= 1;
2467 bitcount++;
2468 }
2469
2470 s.totalram >>= bitcount;
2471 s.freeram >>= bitcount;
2472 s.sharedram >>= bitcount;
2473 s.bufferram >>= bitcount;
2474 s.totalswap >>= bitcount;
2475 s.freeswap >>= bitcount;
2476 s.totalhigh >>= bitcount;
2477 s.freehigh >>= bitcount;
2478 }
2479
2480 if (!access_ok(VERIFY_WRITE, info, sizeof(struct compat_sysinfo)) ||
2481 __put_user(s.uptime, &info->uptime) ||
2482 __put_user(s.loads[0], &info->loads[0]) ||
2483 __put_user(s.loads[1], &info->loads[1]) ||
2484 __put_user(s.loads[2], &info->loads[2]) ||
2485 __put_user(s.totalram, &info->totalram) ||
2486 __put_user(s.freeram, &info->freeram) ||
2487 __put_user(s.sharedram, &info->sharedram) ||
2488 __put_user(s.bufferram, &info->bufferram) ||
2489 __put_user(s.totalswap, &info->totalswap) ||
2490 __put_user(s.freeswap, &info->freeswap) ||
2491 __put_user(s.procs, &info->procs) ||
2492 __put_user(s.totalhigh, &info->totalhigh) ||
2493 __put_user(s.freehigh, &info->freehigh) ||
2494 __put_user(s.mem_unit, &info->mem_unit))
2495 return -EFAULT;
2496
2497 return 0;
2498 }
2499 #endif /* CONFIG_COMPAT */
Linus' kernel tree