4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/module.h>
9 #include <linux/utsname.h>
10 #include <linux/mman.h>
11 #include <linux/notifier.h>
12 #include <linux/reboot.h>
13 #include <linux/prctl.h>
14 #include <linux/highuid.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/gfp.h>
40 #include <linux/syscore_ops.h>
41 #include <linux/version.h>
42 #include <linux/ctype.h>
44 #include <linux/compat.h>
45 #include <linux/syscalls.h>
46 #include <linux/kprobes.h>
47 #include <linux/user_namespace.h>
49 #include <linux/kmsg_dump.h>
50 /* Move somewhere else to avoid recompiling? */
51 #include <generated/utsrelease.h>
53 #include <asm/uaccess.h>
55 #include <asm/unistd.h>
61 #define restart_dbg(format, arg...) \
62 printk("RESTART_DEBUG : " format "\n" , ## arg)
64 #define restart_dbg(format, arg...) do {} while (0)
69 #ifndef SET_UNALIGN_CTL
70 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
72 #ifndef GET_UNALIGN_CTL
73 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
76 # define SET_FPEMU_CTL(a,b) (-EINVAL)
79 # define GET_FPEMU_CTL(a,b) (-EINVAL)
82 # define SET_FPEXC_CTL(a,b) (-EINVAL)
85 # define GET_FPEXC_CTL(a,b) (-EINVAL)
88 # define GET_ENDIAN(a,b) (-EINVAL)
91 # define SET_ENDIAN(a,b) (-EINVAL)
94 # define GET_TSC_CTL(a) (-EINVAL)
97 # define SET_TSC_CTL(a) (-EINVAL)
101 * this is where the system-wide overflow UID and GID are defined, for
102 * architectures that now have 32-bit UID/GID but didn't in the past
105 int overflowuid = DEFAULT_OVERFLOWUID;
106 int overflowgid = DEFAULT_OVERFLOWGID;
109 EXPORT_SYMBOL(overflowuid);
110 EXPORT_SYMBOL(overflowgid);
114 * the same as above, but for filesystems which can only store a 16-bit
115 * UID and GID. as such, this is needed on all architectures
118 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
119 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
121 EXPORT_SYMBOL(fs_overflowuid);
122 EXPORT_SYMBOL(fs_overflowgid);
125 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
130 EXPORT_SYMBOL(cad_pid);
133 * If set, this is used for preparing the system to power off.
136 void (*pm_power_off_prepare)(void);
139 * Returns true if current's euid is same as p's uid or euid,
140 * or has CAP_SYS_NICE to p's user_ns.
142 * Called with rcu_read_lock, creds are safe
144 static bool set_one_prio_perm(struct task_struct *p)
146 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
148 if (pcred->user->user_ns == cred->user->user_ns &&
149 (pcred->uid == cred->euid ||
150 pcred->euid == cred->euid))
152 if (ns_capable(pcred->user->user_ns, CAP_SYS_NICE))
158 * set the priority of a task
159 * - the caller must hold the RCU read lock
161 static int set_one_prio(struct task_struct *p, int niceval, int error)
165 if (!set_one_prio_perm(p)) {
169 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
173 no_nice = security_task_setnice(p, niceval);
180 set_user_nice(p, niceval);
185 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
187 struct task_struct *g, *p;
188 struct user_struct *user;
189 const struct cred *cred = current_cred();
193 if (which > PRIO_USER || which < PRIO_PROCESS)
196 /* normalize: avoid signed division (rounding problems) */
204 read_lock(&tasklist_lock);
208 p = find_task_by_vpid(who);
212 error = set_one_prio(p, niceval, error);
216 pgrp = find_vpid(who);
218 pgrp = task_pgrp(current);
219 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
220 error = set_one_prio(p, niceval, error);
221 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
224 user = (struct user_struct *) cred->user;
227 else if ((who != cred->uid) &&
228 !(user = find_user(who)))
229 goto out_unlock; /* No processes for this user */
231 do_each_thread(g, p) {
232 if (__task_cred(p)->uid == who)
233 error = set_one_prio(p, niceval, error);
234 } while_each_thread(g, p);
235 if (who != cred->uid)
236 free_uid(user); /* For find_user() */
240 read_unlock(&tasklist_lock);
247 * Ugh. To avoid negative return values, "getpriority()" will
248 * not return the normal nice-value, but a negated value that
249 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
250 * to stay compatible.
252 SYSCALL_DEFINE2(getpriority, int, which, int, who)
254 struct task_struct *g, *p;
255 struct user_struct *user;
256 const struct cred *cred = current_cred();
257 long niceval, retval = -ESRCH;
260 if (which > PRIO_USER || which < PRIO_PROCESS)
264 read_lock(&tasklist_lock);
268 p = find_task_by_vpid(who);
272 niceval = 20 - task_nice(p);
273 if (niceval > retval)
279 pgrp = find_vpid(who);
281 pgrp = task_pgrp(current);
282 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
283 niceval = 20 - task_nice(p);
284 if (niceval > retval)
286 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
289 user = (struct user_struct *) cred->user;
292 else if ((who != cred->uid) &&
293 !(user = find_user(who)))
294 goto out_unlock; /* No processes for this user */
296 do_each_thread(g, p) {
297 if (__task_cred(p)->uid == who) {
298 niceval = 20 - task_nice(p);
299 if (niceval > retval)
302 } while_each_thread(g, p);
303 if (who != cred->uid)
304 free_uid(user); /* for find_user() */
308 read_unlock(&tasklist_lock);
315 * emergency_restart - reboot the system
317 * Without shutting down any hardware or taking any locks
318 * reboot the system. This is called when we know we are in
319 * trouble so this is our best effort to reboot. This is
320 * safe to call in interrupt context.
322 void emergency_restart(void)
324 kmsg_dump(KMSG_DUMP_EMERG);
325 machine_emergency_restart();
327 EXPORT_SYMBOL_GPL(emergency_restart);
329 void kernel_restart_prepare(char *cmd)
331 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
332 system_state = SYSTEM_RESTART;
333 usermodehelper_disable();
339 * kernel_restart - reboot the system
340 * @cmd: pointer to buffer containing command to execute for restart
343 * Shutdown everything and perform a clean reboot.
344 * This is not safe to call in interrupt context.
346 void kernel_restart(char *cmd)
351 restart_dbg("%s->%d->cmd=%s",__FUNCTION__,__LINE__,cmd);
353 kernel_restart_prepare(cmd);
354 disable_nonboot_cpus();
356 printk(KERN_EMERG "Restarting system.\n");
358 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
359 kmsg_dump(KMSG_DUMP_RESTART);
360 machine_restart(cmd);
362 EXPORT_SYMBOL_GPL(kernel_restart);
364 static void kernel_shutdown_prepare(enum system_states state)
366 blocking_notifier_call_chain(&reboot_notifier_list,
367 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
368 system_state = state;
369 usermodehelper_disable();
373 * kernel_halt - halt the system
375 * Shutdown everything and perform a clean system halt.
377 void kernel_halt(void)
379 kernel_shutdown_prepare(SYSTEM_HALT);
381 printk(KERN_EMERG "System halted.\n");
382 kmsg_dump(KMSG_DUMP_HALT);
386 EXPORT_SYMBOL_GPL(kernel_halt);
389 * kernel_power_off - power_off the system
391 * Shutdown everything and perform a clean system power_off.
393 void kernel_power_off(void)
395 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
396 if (pm_power_off_prepare)
397 pm_power_off_prepare();
398 disable_nonboot_cpus();
400 printk(KERN_EMERG "Power down.\n");
401 kmsg_dump(KMSG_DUMP_POWEROFF);
404 EXPORT_SYMBOL_GPL(kernel_power_off);
406 static DEFINE_MUTEX(reboot_mutex);
409 * Reboot system call: for obvious reasons only root may call it,
410 * and even root needs to set up some magic numbers in the registers
411 * so that some mistake won't make this reboot the whole machine.
412 * You can also set the meaning of the ctrl-alt-del-key here.
414 * reboot doesn't sync: do that yourself before calling this.
416 SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
422 /* We only trust the superuser with rebooting the system. */
423 if (!capable(CAP_SYS_BOOT))
426 /* For safety, we require "magic" arguments. */
427 if (magic1 != LINUX_REBOOT_MAGIC1 ||
428 (magic2 != LINUX_REBOOT_MAGIC2 &&
429 magic2 != LINUX_REBOOT_MAGIC2A &&
430 magic2 != LINUX_REBOOT_MAGIC2B &&
431 magic2 != LINUX_REBOOT_MAGIC2C))
434 /* Instead of trying to make the power_off code look like
435 * halt when pm_power_off is not set do it the easy way.
437 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
438 cmd = LINUX_REBOOT_CMD_HALT;
440 mutex_lock(&reboot_mutex);
442 case LINUX_REBOOT_CMD_RESTART:
446 restart_dbg("%s->%d->cmd=%x",__FUNCTION__,__LINE__,cmd);
448 kernel_restart(NULL);
451 case LINUX_REBOOT_CMD_CAD_ON:
455 case LINUX_REBOOT_CMD_CAD_OFF:
459 case LINUX_REBOOT_CMD_HALT:
462 panic("cannot halt");
464 case LINUX_REBOOT_CMD_POWER_OFF:
468 restart_dbg("%s->%d->cmd=%x",__FUNCTION__,__LINE__,cmd);
474 case LINUX_REBOOT_CMD_RESTART2:
475 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
479 buffer[sizeof(buffer) - 1] = '\0';
483 restart_dbg("%s->%d->cmd=%x args=%s",__FUNCTION__,__LINE__,cmd,buffer);
485 kernel_restart(buffer);
489 case LINUX_REBOOT_CMD_KEXEC:
490 ret = kernel_kexec();
494 #ifdef CONFIG_HIBERNATION
495 case LINUX_REBOOT_CMD_SW_SUSPEND:
504 mutex_unlock(&reboot_mutex);
508 static void deferred_cad(struct work_struct *dummy)
510 kernel_restart(NULL);
514 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
515 * As it's called within an interrupt, it may NOT sync: the only choice
516 * is whether to reboot at once, or just ignore the ctrl-alt-del.
518 void ctrl_alt_del(void)
520 static DECLARE_WORK(cad_work, deferred_cad);
523 schedule_work(&cad_work);
525 kill_cad_pid(SIGINT, 1);
529 * Unprivileged users may change the real gid to the effective gid
530 * or vice versa. (BSD-style)
532 * If you set the real gid at all, or set the effective gid to a value not
533 * equal to the real gid, then the saved gid is set to the new effective gid.
535 * This makes it possible for a setgid program to completely drop its
536 * privileges, which is often a useful assertion to make when you are doing
537 * a security audit over a program.
539 * The general idea is that a program which uses just setregid() will be
540 * 100% compatible with BSD. A program which uses just setgid() will be
541 * 100% compatible with POSIX with saved IDs.
543 * SMP: There are not races, the GIDs are checked only by filesystem
544 * operations (as far as semantic preservation is concerned).
546 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
548 const struct cred *old;
552 new = prepare_creds();
555 old = current_cred();
558 if (rgid != (gid_t) -1) {
559 if (old->gid == rgid ||
561 nsown_capable(CAP_SETGID))
566 if (egid != (gid_t) -1) {
567 if (old->gid == egid ||
570 nsown_capable(CAP_SETGID))
576 if (rgid != (gid_t) -1 ||
577 (egid != (gid_t) -1 && egid != old->gid))
578 new->sgid = new->egid;
579 new->fsgid = new->egid;
581 return commit_creds(new);
589 * setgid() is implemented like SysV w/ SAVED_IDS
591 * SMP: Same implicit races as above.
593 SYSCALL_DEFINE1(setgid, gid_t, gid)
595 const struct cred *old;
599 new = prepare_creds();
602 old = current_cred();
605 if (nsown_capable(CAP_SETGID))
606 new->gid = new->egid = new->sgid = new->fsgid = gid;
607 else if (gid == old->gid || gid == old->sgid)
608 new->egid = new->fsgid = gid;
612 return commit_creds(new);
620 * change the user struct in a credentials set to match the new UID
622 static int set_user(struct cred *new)
624 struct user_struct *new_user;
626 new_user = alloc_uid(current_user_ns(), new->uid);
630 if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
631 new_user != INIT_USER) {
637 new->user = new_user;
642 * Unprivileged users may change the real uid to the effective uid
643 * or vice versa. (BSD-style)
645 * If you set the real uid at all, or set the effective uid to a value not
646 * equal to the real uid, then the saved uid is set to the new effective uid.
648 * This makes it possible for a setuid program to completely drop its
649 * privileges, which is often a useful assertion to make when you are doing
650 * a security audit over a program.
652 * The general idea is that a program which uses just setreuid() will be
653 * 100% compatible with BSD. A program which uses just setuid() will be
654 * 100% compatible with POSIX with saved IDs.
656 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
658 const struct cred *old;
662 new = prepare_creds();
665 old = current_cred();
668 if (ruid != (uid_t) -1) {
670 if (old->uid != ruid &&
672 !nsown_capable(CAP_SETUID))
676 if (euid != (uid_t) -1) {
678 if (old->uid != euid &&
681 !nsown_capable(CAP_SETUID))
685 if (new->uid != old->uid) {
686 retval = set_user(new);
690 if (ruid != (uid_t) -1 ||
691 (euid != (uid_t) -1 && euid != old->uid))
692 new->suid = new->euid;
693 new->fsuid = new->euid;
695 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
699 return commit_creds(new);
707 * setuid() is implemented like SysV with SAVED_IDS
709 * Note that SAVED_ID's is deficient in that a setuid root program
710 * like sendmail, for example, cannot set its uid to be a normal
711 * user and then switch back, because if you're root, setuid() sets
712 * the saved uid too. If you don't like this, blame the bright people
713 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
714 * will allow a root program to temporarily drop privileges and be able to
715 * regain them by swapping the real and effective uid.
717 SYSCALL_DEFINE1(setuid, uid_t, uid)
719 const struct cred *old;
723 new = prepare_creds();
726 old = current_cred();
729 if (nsown_capable(CAP_SETUID)) {
730 new->suid = new->uid = uid;
731 if (uid != old->uid) {
732 retval = set_user(new);
736 } else if (uid != old->uid && uid != new->suid) {
740 new->fsuid = new->euid = uid;
742 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
746 return commit_creds(new);
755 * This function implements a generic ability to update ruid, euid,
756 * and suid. This allows you to implement the 4.4 compatible seteuid().
758 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
760 const struct cred *old;
764 new = prepare_creds();
768 old = current_cred();
771 if (!nsown_capable(CAP_SETUID)) {
772 if (ruid != (uid_t) -1 && ruid != old->uid &&
773 ruid != old->euid && ruid != old->suid)
775 if (euid != (uid_t) -1 && euid != old->uid &&
776 euid != old->euid && euid != old->suid)
778 if (suid != (uid_t) -1 && suid != old->uid &&
779 suid != old->euid && suid != old->suid)
783 if (ruid != (uid_t) -1) {
785 if (ruid != old->uid) {
786 retval = set_user(new);
791 if (euid != (uid_t) -1)
793 if (suid != (uid_t) -1)
795 new->fsuid = new->euid;
797 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
801 return commit_creds(new);
808 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruid, uid_t __user *, euid, uid_t __user *, suid)
810 const struct cred *cred = current_cred();
813 if (!(retval = put_user(cred->uid, ruid)) &&
814 !(retval = put_user(cred->euid, euid)))
815 retval = put_user(cred->suid, suid);
821 * Same as above, but for rgid, egid, sgid.
823 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
825 const struct cred *old;
829 new = prepare_creds();
832 old = current_cred();
835 if (!nsown_capable(CAP_SETGID)) {
836 if (rgid != (gid_t) -1 && rgid != old->gid &&
837 rgid != old->egid && rgid != old->sgid)
839 if (egid != (gid_t) -1 && egid != old->gid &&
840 egid != old->egid && egid != old->sgid)
842 if (sgid != (gid_t) -1 && sgid != old->gid &&
843 sgid != old->egid && sgid != old->sgid)
847 if (rgid != (gid_t) -1)
849 if (egid != (gid_t) -1)
851 if (sgid != (gid_t) -1)
853 new->fsgid = new->egid;
855 return commit_creds(new);
862 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgid, gid_t __user *, egid, gid_t __user *, sgid)
864 const struct cred *cred = current_cred();
867 if (!(retval = put_user(cred->gid, rgid)) &&
868 !(retval = put_user(cred->egid, egid)))
869 retval = put_user(cred->sgid, sgid);
876 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
877 * is used for "access()" and for the NFS daemon (letting nfsd stay at
878 * whatever uid it wants to). It normally shadows "euid", except when
879 * explicitly set by setfsuid() or for access..
881 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
883 const struct cred *old;
887 new = prepare_creds();
889 return current_fsuid();
890 old = current_cred();
891 old_fsuid = old->fsuid;
893 if (uid == old->uid || uid == old->euid ||
894 uid == old->suid || uid == old->fsuid ||
895 nsown_capable(CAP_SETUID)) {
896 if (uid != old_fsuid) {
898 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
912 * Samma på svenska..
914 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
916 const struct cred *old;
920 new = prepare_creds();
922 return current_fsgid();
923 old = current_cred();
924 old_fsgid = old->fsgid;
926 if (gid == old->gid || gid == old->egid ||
927 gid == old->sgid || gid == old->fsgid ||
928 nsown_capable(CAP_SETGID)) {
929 if (gid != old_fsgid) {
943 void do_sys_times(struct tms *tms)
945 cputime_t tgutime, tgstime, cutime, cstime;
947 spin_lock_irq(¤t->sighand->siglock);
948 thread_group_times(current, &tgutime, &tgstime);
949 cutime = current->signal->cutime;
950 cstime = current->signal->cstime;
951 spin_unlock_irq(¤t->sighand->siglock);
952 tms->tms_utime = cputime_to_clock_t(tgutime);
953 tms->tms_stime = cputime_to_clock_t(tgstime);
954 tms->tms_cutime = cputime_to_clock_t(cutime);
955 tms->tms_cstime = cputime_to_clock_t(cstime);
958 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
964 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
967 force_successful_syscall_return();
968 return (long) jiffies_64_to_clock_t(get_jiffies_64());
972 * This needs some heavy checking ...
973 * I just haven't the stomach for it. I also don't fully
974 * understand sessions/pgrp etc. Let somebody who does explain it.
976 * OK, I think I have the protection semantics right.... this is really
977 * only important on a multi-user system anyway, to make sure one user
978 * can't send a signal to a process owned by another. -TYT, 12/12/91
980 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
983 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
985 struct task_struct *p;
986 struct task_struct *group_leader = current->group_leader;
991 pid = task_pid_vnr(group_leader);
998 /* From this point forward we keep holding onto the tasklist lock
999 * so that our parent does not change from under us. -DaveM
1001 write_lock_irq(&tasklist_lock);
1004 p = find_task_by_vpid(pid);
1009 if (!thread_group_leader(p))
1012 if (same_thread_group(p->real_parent, group_leader)) {
1014 if (task_session(p) != task_session(group_leader))
1021 if (p != group_leader)
1026 if (p->signal->leader)
1031 struct task_struct *g;
1033 pgrp = find_vpid(pgid);
1034 g = pid_task(pgrp, PIDTYPE_PGID);
1035 if (!g || task_session(g) != task_session(group_leader))
1039 err = security_task_setpgid(p, pgid);
1043 if (task_pgrp(p) != pgrp)
1044 change_pid(p, PIDTYPE_PGID, pgrp);
1048 /* All paths lead to here, thus we are safe. -DaveM */
1049 write_unlock_irq(&tasklist_lock);
1054 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1056 struct task_struct *p;
1062 grp = task_pgrp(current);
1065 p = find_task_by_vpid(pid);
1072 retval = security_task_getpgid(p);
1076 retval = pid_vnr(grp);
1082 #ifdef __ARCH_WANT_SYS_GETPGRP
1084 SYSCALL_DEFINE0(getpgrp)
1086 return sys_getpgid(0);
1091 SYSCALL_DEFINE1(getsid, pid_t, pid)
1093 struct task_struct *p;
1099 sid = task_session(current);
1102 p = find_task_by_vpid(pid);
1105 sid = task_session(p);
1109 retval = security_task_getsid(p);
1113 retval = pid_vnr(sid);
1119 SYSCALL_DEFINE0(setsid)
1121 struct task_struct *group_leader = current->group_leader;
1122 struct pid *sid = task_pid(group_leader);
1123 pid_t session = pid_vnr(sid);
1126 write_lock_irq(&tasklist_lock);
1127 /* Fail if I am already a session leader */
1128 if (group_leader->signal->leader)
1131 /* Fail if a process group id already exists that equals the
1132 * proposed session id.
1134 if (pid_task(sid, PIDTYPE_PGID))
1137 group_leader->signal->leader = 1;
1138 __set_special_pids(sid);
1140 proc_clear_tty(group_leader);
1144 write_unlock_irq(&tasklist_lock);
1146 proc_sid_connector(group_leader);
1147 sched_autogroup_create_attach(group_leader);
1152 DECLARE_RWSEM(uts_sem);
1154 #ifdef COMPAT_UTS_MACHINE
1155 #define override_architecture(name) \
1156 (personality(current->personality) == PER_LINUX32 && \
1157 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1158 sizeof(COMPAT_UTS_MACHINE)))
1160 #define override_architecture(name) 0
1164 * Work around broken programs that cannot handle "Linux 3.0".
1165 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1167 static int override_release(char __user *release, size_t len)
1171 if (current->personality & UNAME26) {
1172 const char *rest = UTS_RELEASE;
1173 char buf[65] = { 0 };
1179 if (*rest == '.' && ++ndots >= 3)
1181 if (!isdigit(*rest) && *rest != '.')
1185 v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 40;
1186 copy = clamp_t(size_t, len, 1, sizeof(buf));
1187 copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1188 ret = copy_to_user(release, buf, copy + 1);
1193 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1197 down_read(&uts_sem);
1198 if (copy_to_user(name, utsname(), sizeof *name))
1202 if (!errno && override_release(name->release, sizeof(name->release)))
1204 if (!errno && override_architecture(name))
1209 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1213 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1220 down_read(&uts_sem);
1221 if (copy_to_user(name, utsname(), sizeof(*name)))
1225 if (!error && override_release(name->release, sizeof(name->release)))
1227 if (!error && override_architecture(name))
1232 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1238 if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
1241 down_read(&uts_sem);
1242 error = __copy_to_user(&name->sysname, &utsname()->sysname,
1244 error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
1245 error |= __copy_to_user(&name->nodename, &utsname()->nodename,
1247 error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
1248 error |= __copy_to_user(&name->release, &utsname()->release,
1250 error |= __put_user(0, name->release + __OLD_UTS_LEN);
1251 error |= __copy_to_user(&name->version, &utsname()->version,
1253 error |= __put_user(0, name->version + __OLD_UTS_LEN);
1254 error |= __copy_to_user(&name->machine, &utsname()->machine,
1256 error |= __put_user(0, name->machine + __OLD_UTS_LEN);
1259 if (!error && override_architecture(name))
1261 if (!error && override_release(name->release, sizeof(name->release)))
1263 return error ? -EFAULT : 0;
1267 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1270 char tmp[__NEW_UTS_LEN];
1272 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1275 if (len < 0 || len > __NEW_UTS_LEN)
1277 down_write(&uts_sem);
1279 if (!copy_from_user(tmp, name, len)) {
1280 struct new_utsname *u = utsname();
1282 memcpy(u->nodename, tmp, len);
1283 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1290 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1292 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1295 struct new_utsname *u;
1299 down_read(&uts_sem);
1301 i = 1 + strlen(u->nodename);
1305 if (copy_to_user(name, u->nodename, i))
1314 * Only setdomainname; getdomainname can be implemented by calling
1317 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1320 char tmp[__NEW_UTS_LEN];
1322 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1324 if (len < 0 || len > __NEW_UTS_LEN)
1327 down_write(&uts_sem);
1329 if (!copy_from_user(tmp, name, len)) {
1330 struct new_utsname *u = utsname();
1332 memcpy(u->domainname, tmp, len);
1333 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1340 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1342 struct rlimit value;
1345 ret = do_prlimit(current, resource, NULL, &value);
1347 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1352 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1355 * Back compatibility for getrlimit. Needed for some apps.
1358 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1359 struct rlimit __user *, rlim)
1362 if (resource >= RLIM_NLIMITS)
1365 task_lock(current->group_leader);
1366 x = current->signal->rlim[resource];
1367 task_unlock(current->group_leader);
1368 if (x.rlim_cur > 0x7FFFFFFF)
1369 x.rlim_cur = 0x7FFFFFFF;
1370 if (x.rlim_max > 0x7FFFFFFF)
1371 x.rlim_max = 0x7FFFFFFF;
1372 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1377 static inline bool rlim64_is_infinity(__u64 rlim64)
1379 #if BITS_PER_LONG < 64
1380 return rlim64 >= ULONG_MAX;
1382 return rlim64 == RLIM64_INFINITY;
1386 static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1388 if (rlim->rlim_cur == RLIM_INFINITY)
1389 rlim64->rlim_cur = RLIM64_INFINITY;
1391 rlim64->rlim_cur = rlim->rlim_cur;
1392 if (rlim->rlim_max == RLIM_INFINITY)
1393 rlim64->rlim_max = RLIM64_INFINITY;
1395 rlim64->rlim_max = rlim->rlim_max;
1398 static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1400 if (rlim64_is_infinity(rlim64->rlim_cur))
1401 rlim->rlim_cur = RLIM_INFINITY;
1403 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1404 if (rlim64_is_infinity(rlim64->rlim_max))
1405 rlim->rlim_max = RLIM_INFINITY;
1407 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1410 /* make sure you are allowed to change @tsk limits before calling this */
1411 int do_prlimit(struct task_struct *tsk, unsigned int resource,
1412 struct rlimit *new_rlim, struct rlimit *old_rlim)
1414 struct rlimit *rlim;
1417 if (resource >= RLIM_NLIMITS)
1420 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1422 if (resource == RLIMIT_NOFILE &&
1423 new_rlim->rlim_max > sysctl_nr_open)
1427 /* protect tsk->signal and tsk->sighand from disappearing */
1428 read_lock(&tasklist_lock);
1429 if (!tsk->sighand) {
1434 rlim = tsk->signal->rlim + resource;
1435 task_lock(tsk->group_leader);
1437 /* Keep the capable check against init_user_ns until
1438 cgroups can contain all limits */
1439 if (new_rlim->rlim_max > rlim->rlim_max &&
1440 !capable(CAP_SYS_RESOURCE))
1443 retval = security_task_setrlimit(tsk->group_leader,
1444 resource, new_rlim);
1445 if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) {
1447 * The caller is asking for an immediate RLIMIT_CPU
1448 * expiry. But we use the zero value to mean "it was
1449 * never set". So let's cheat and make it one second
1452 new_rlim->rlim_cur = 1;
1461 task_unlock(tsk->group_leader);
1464 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1465 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1466 * very long-standing error, and fixing it now risks breakage of
1467 * applications, so we live with it
1469 if (!retval && new_rlim && resource == RLIMIT_CPU &&
1470 new_rlim->rlim_cur != RLIM_INFINITY)
1471 update_rlimit_cpu(tsk, new_rlim->rlim_cur);
1473 read_unlock(&tasklist_lock);
1477 /* rcu lock must be held */
1478 static int check_prlimit_permission(struct task_struct *task)
1480 const struct cred *cred = current_cred(), *tcred;
1482 if (current == task)
1485 tcred = __task_cred(task);
1486 if (cred->user->user_ns == tcred->user->user_ns &&
1487 (cred->uid == tcred->euid &&
1488 cred->uid == tcred->suid &&
1489 cred->uid == tcred->uid &&
1490 cred->gid == tcred->egid &&
1491 cred->gid == tcred->sgid &&
1492 cred->gid == tcred->gid))
1494 if (ns_capable(tcred->user->user_ns, CAP_SYS_RESOURCE))
1500 SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1501 const struct rlimit64 __user *, new_rlim,
1502 struct rlimit64 __user *, old_rlim)
1504 struct rlimit64 old64, new64;
1505 struct rlimit old, new;
1506 struct task_struct *tsk;
1510 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1512 rlim64_to_rlim(&new64, &new);
1516 tsk = pid ? find_task_by_vpid(pid) : current;
1521 ret = check_prlimit_permission(tsk);
1526 get_task_struct(tsk);
1529 ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1530 old_rlim ? &old : NULL);
1532 if (!ret && old_rlim) {
1533 rlim_to_rlim64(&old, &old64);
1534 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1538 put_task_struct(tsk);
1542 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1544 struct rlimit new_rlim;
1546 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1548 return do_prlimit(current, resource, &new_rlim, NULL);
1552 * It would make sense to put struct rusage in the task_struct,
1553 * except that would make the task_struct be *really big*. After
1554 * task_struct gets moved into malloc'ed memory, it would
1555 * make sense to do this. It will make moving the rest of the information
1556 * a lot simpler! (Which we're not doing right now because we're not
1557 * measuring them yet).
1559 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1560 * races with threads incrementing their own counters. But since word
1561 * reads are atomic, we either get new values or old values and we don't
1562 * care which for the sums. We always take the siglock to protect reading
1563 * the c* fields from p->signal from races with exit.c updating those
1564 * fields when reaping, so a sample either gets all the additions of a
1565 * given child after it's reaped, or none so this sample is before reaping.
1568 * We need to take the siglock for CHILDEREN, SELF and BOTH
1569 * for the cases current multithreaded, non-current single threaded
1570 * non-current multithreaded. Thread traversal is now safe with
1572 * Strictly speaking, we donot need to take the siglock if we are current and
1573 * single threaded, as no one else can take our signal_struct away, no one
1574 * else can reap the children to update signal->c* counters, and no one else
1575 * can race with the signal-> fields. If we do not take any lock, the
1576 * signal-> fields could be read out of order while another thread was just
1577 * exiting. So we should place a read memory barrier when we avoid the lock.
1578 * On the writer side, write memory barrier is implied in __exit_signal
1579 * as __exit_signal releases the siglock spinlock after updating the signal->
1580 * fields. But we don't do this yet to keep things simple.
1584 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1586 r->ru_nvcsw += t->nvcsw;
1587 r->ru_nivcsw += t->nivcsw;
1588 r->ru_minflt += t->min_flt;
1589 r->ru_majflt += t->maj_flt;
1590 r->ru_inblock += task_io_get_inblock(t);
1591 r->ru_oublock += task_io_get_oublock(t);
1594 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1596 struct task_struct *t;
1597 unsigned long flags;
1598 cputime_t tgutime, tgstime, utime, stime;
1599 unsigned long maxrss = 0;
1601 memset((char *) r, 0, sizeof *r);
1602 utime = stime = cputime_zero;
1604 if (who == RUSAGE_THREAD) {
1605 task_times(current, &utime, &stime);
1606 accumulate_thread_rusage(p, r);
1607 maxrss = p->signal->maxrss;
1611 if (!lock_task_sighand(p, &flags))
1616 case RUSAGE_CHILDREN:
1617 utime = p->signal->cutime;
1618 stime = p->signal->cstime;
1619 r->ru_nvcsw = p->signal->cnvcsw;
1620 r->ru_nivcsw = p->signal->cnivcsw;
1621 r->ru_minflt = p->signal->cmin_flt;
1622 r->ru_majflt = p->signal->cmaj_flt;
1623 r->ru_inblock = p->signal->cinblock;
1624 r->ru_oublock = p->signal->coublock;
1625 maxrss = p->signal->cmaxrss;
1627 if (who == RUSAGE_CHILDREN)
1631 thread_group_times(p, &tgutime, &tgstime);
1632 utime = cputime_add(utime, tgutime);
1633 stime = cputime_add(stime, tgstime);
1634 r->ru_nvcsw += p->signal->nvcsw;
1635 r->ru_nivcsw += p->signal->nivcsw;
1636 r->ru_minflt += p->signal->min_flt;
1637 r->ru_majflt += p->signal->maj_flt;
1638 r->ru_inblock += p->signal->inblock;
1639 r->ru_oublock += p->signal->oublock;
1640 if (maxrss < p->signal->maxrss)
1641 maxrss = p->signal->maxrss;
1644 accumulate_thread_rusage(t, r);
1652 unlock_task_sighand(p, &flags);
1655 cputime_to_timeval(utime, &r->ru_utime);
1656 cputime_to_timeval(stime, &r->ru_stime);
1658 if (who != RUSAGE_CHILDREN) {
1659 struct mm_struct *mm = get_task_mm(p);
1661 setmax_mm_hiwater_rss(&maxrss, mm);
1665 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1668 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1671 k_getrusage(p, who, &r);
1672 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1675 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1677 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1678 who != RUSAGE_THREAD)
1680 return getrusage(current, who, ru);
1683 SYSCALL_DEFINE1(umask, int, mask)
1685 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1689 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
1690 unsigned long, arg4, unsigned long, arg5)
1692 struct task_struct *me = current;
1693 unsigned char comm[sizeof(me->comm)];
1696 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1697 if (error != -ENOSYS)
1702 case PR_SET_PDEATHSIG:
1703 if (!valid_signal(arg2)) {
1707 me->pdeath_signal = arg2;
1710 case PR_GET_PDEATHSIG:
1711 error = put_user(me->pdeath_signal, (int __user *)arg2);
1713 case PR_GET_DUMPABLE:
1714 error = get_dumpable(me->mm);
1716 case PR_SET_DUMPABLE:
1717 if (arg2 < 0 || arg2 > 1) {
1721 set_dumpable(me->mm, arg2);
1725 case PR_SET_UNALIGN:
1726 error = SET_UNALIGN_CTL(me, arg2);
1728 case PR_GET_UNALIGN:
1729 error = GET_UNALIGN_CTL(me, arg2);
1732 error = SET_FPEMU_CTL(me, arg2);
1735 error = GET_FPEMU_CTL(me, arg2);
1738 error = SET_FPEXC_CTL(me, arg2);
1741 error = GET_FPEXC_CTL(me, arg2);
1744 error = PR_TIMING_STATISTICAL;
1747 if (arg2 != PR_TIMING_STATISTICAL)
1754 comm[sizeof(me->comm)-1] = 0;
1755 if (strncpy_from_user(comm, (char __user *)arg2,
1756 sizeof(me->comm) - 1) < 0)
1758 set_task_comm(me, comm);
1761 get_task_comm(comm, me);
1762 if (copy_to_user((char __user *)arg2, comm,
1767 error = GET_ENDIAN(me, arg2);
1770 error = SET_ENDIAN(me, arg2);
1773 case PR_GET_SECCOMP:
1774 error = prctl_get_seccomp();
1776 case PR_SET_SECCOMP:
1777 error = prctl_set_seccomp(arg2);
1780 error = GET_TSC_CTL(arg2);
1783 error = SET_TSC_CTL(arg2);
1785 case PR_TASK_PERF_EVENTS_DISABLE:
1786 error = perf_event_task_disable();
1788 case PR_TASK_PERF_EVENTS_ENABLE:
1789 error = perf_event_task_enable();
1791 case PR_GET_TIMERSLACK:
1792 error = current->timer_slack_ns;
1794 case PR_SET_TIMERSLACK:
1796 current->timer_slack_ns =
1797 current->default_timer_slack_ns;
1799 current->timer_slack_ns = arg2;
1806 case PR_MCE_KILL_CLEAR:
1809 current->flags &= ~PF_MCE_PROCESS;
1811 case PR_MCE_KILL_SET:
1812 current->flags |= PF_MCE_PROCESS;
1813 if (arg3 == PR_MCE_KILL_EARLY)
1814 current->flags |= PF_MCE_EARLY;
1815 else if (arg3 == PR_MCE_KILL_LATE)
1816 current->flags &= ~PF_MCE_EARLY;
1817 else if (arg3 == PR_MCE_KILL_DEFAULT)
1819 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
1828 case PR_MCE_KILL_GET:
1829 if (arg2 | arg3 | arg4 | arg5)
1831 if (current->flags & PF_MCE_PROCESS)
1832 error = (current->flags & PF_MCE_EARLY) ?
1833 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
1835 error = PR_MCE_KILL_DEFAULT;
1844 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
1845 struct getcpu_cache __user *, unused)
1848 int cpu = raw_smp_processor_id();
1850 err |= put_user(cpu, cpup);
1852 err |= put_user(cpu_to_node(cpu), nodep);
1853 return err ? -EFAULT : 0;
1856 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
1858 static void argv_cleanup(struct subprocess_info *info)
1860 argv_free(info->argv);
1864 * orderly_poweroff - Trigger an orderly system poweroff
1865 * @force: force poweroff if command execution fails
1867 * This may be called from any context to trigger a system shutdown.
1868 * If the orderly shutdown fails, it will force an immediate shutdown.
1870 int orderly_poweroff(bool force)
1873 char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
1874 static char *envp[] = {
1876 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
1880 struct subprocess_info *info;
1883 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
1884 __func__, poweroff_cmd);
1888 info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
1894 call_usermodehelper_setfns(info, NULL, argv_cleanup, NULL);
1896 ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
1900 printk(KERN_WARNING "Failed to start orderly shutdown: "
1901 "forcing the issue\n");
1903 /* I guess this should try to kick off some daemon to
1904 sync and poweroff asap. Or not even bother syncing
1905 if we're doing an emergency shutdown? */
1912 EXPORT_SYMBOL_GPL(orderly_poweroff);