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>
42 #include <linux/compat.h>
43 #include <linux/syscalls.h>
44 #include <linux/kprobes.h>
45 #include <linux/user_namespace.h>
47 #include <linux/kmsg_dump.h>
49 #include <asm/uaccess.h>
51 #include <asm/unistd.h>
53 #ifndef SET_UNALIGN_CTL
54 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
56 #ifndef GET_UNALIGN_CTL
57 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
60 # define SET_FPEMU_CTL(a,b) (-EINVAL)
63 # define GET_FPEMU_CTL(a,b) (-EINVAL)
66 # define SET_FPEXC_CTL(a,b) (-EINVAL)
69 # define GET_FPEXC_CTL(a,b) (-EINVAL)
72 # define GET_ENDIAN(a,b) (-EINVAL)
75 # define SET_ENDIAN(a,b) (-EINVAL)
78 # define GET_TSC_CTL(a) (-EINVAL)
81 # define SET_TSC_CTL(a) (-EINVAL)
85 * this is where the system-wide overflow UID and GID are defined, for
86 * architectures that now have 32-bit UID/GID but didn't in the past
89 int overflowuid = DEFAULT_OVERFLOWUID;
90 int overflowgid = DEFAULT_OVERFLOWGID;
93 EXPORT_SYMBOL(overflowuid);
94 EXPORT_SYMBOL(overflowgid);
98 * the same as above, but for filesystems which can only store a 16-bit
99 * UID and GID. as such, this is needed on all architectures
102 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
103 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
105 EXPORT_SYMBOL(fs_overflowuid);
106 EXPORT_SYMBOL(fs_overflowgid);
109 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
114 EXPORT_SYMBOL(cad_pid);
117 * If set, this is used for preparing the system to power off.
120 void (*pm_power_off_prepare)(void);
123 * Returns true if current's euid is same as p's uid or euid,
124 * or has CAP_SYS_NICE to p's user_ns.
126 * Called with rcu_read_lock, creds are safe
128 static bool set_one_prio_perm(struct task_struct *p)
130 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
132 if (pcred->user->user_ns == cred->user->user_ns &&
133 (pcred->uid == cred->euid ||
134 pcred->euid == cred->euid))
136 if (ns_capable(pcred->user->user_ns, CAP_SYS_NICE))
142 * set the priority of a task
143 * - the caller must hold the RCU read lock
145 static int set_one_prio(struct task_struct *p, int niceval, int error)
149 if (!set_one_prio_perm(p)) {
153 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
157 no_nice = security_task_setnice(p, niceval);
164 set_user_nice(p, niceval);
169 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
171 struct task_struct *g, *p;
172 struct user_struct *user;
173 const struct cred *cred = current_cred();
177 if (which > PRIO_USER || which < PRIO_PROCESS)
180 /* normalize: avoid signed division (rounding problems) */
188 read_lock(&tasklist_lock);
192 p = find_task_by_vpid(who);
196 error = set_one_prio(p, niceval, error);
200 pgrp = find_vpid(who);
202 pgrp = task_pgrp(current);
203 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
204 error = set_one_prio(p, niceval, error);
205 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
208 user = (struct user_struct *) cred->user;
211 else if ((who != cred->uid) &&
212 !(user = find_user(who)))
213 goto out_unlock; /* No processes for this user */
215 do_each_thread(g, p) {
216 if (__task_cred(p)->uid == who)
217 error = set_one_prio(p, niceval, error);
218 } while_each_thread(g, p);
219 if (who != cred->uid)
220 free_uid(user); /* For find_user() */
224 read_unlock(&tasklist_lock);
231 * Ugh. To avoid negative return values, "getpriority()" will
232 * not return the normal nice-value, but a negated value that
233 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
234 * to stay compatible.
236 SYSCALL_DEFINE2(getpriority, int, which, int, who)
238 struct task_struct *g, *p;
239 struct user_struct *user;
240 const struct cred *cred = current_cred();
241 long niceval, retval = -ESRCH;
244 if (which > PRIO_USER || which < PRIO_PROCESS)
248 read_lock(&tasklist_lock);
252 p = find_task_by_vpid(who);
256 niceval = 20 - task_nice(p);
257 if (niceval > retval)
263 pgrp = find_vpid(who);
265 pgrp = task_pgrp(current);
266 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
267 niceval = 20 - task_nice(p);
268 if (niceval > retval)
270 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
273 user = (struct user_struct *) cred->user;
276 else if ((who != cred->uid) &&
277 !(user = find_user(who)))
278 goto out_unlock; /* No processes for this user */
280 do_each_thread(g, p) {
281 if (__task_cred(p)->uid == who) {
282 niceval = 20 - task_nice(p);
283 if (niceval > retval)
286 } while_each_thread(g, p);
287 if (who != cred->uid)
288 free_uid(user); /* for find_user() */
292 read_unlock(&tasklist_lock);
299 * emergency_restart - reboot the system
301 * Without shutting down any hardware or taking any locks
302 * reboot the system. This is called when we know we are in
303 * trouble so this is our best effort to reboot. This is
304 * safe to call in interrupt context.
306 void emergency_restart(void)
308 kmsg_dump(KMSG_DUMP_EMERG);
309 machine_emergency_restart();
311 EXPORT_SYMBOL_GPL(emergency_restart);
313 void kernel_restart_prepare(char *cmd)
315 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
316 system_state = SYSTEM_RESTART;
317 usermodehelper_disable();
323 * kernel_restart - reboot the system
324 * @cmd: pointer to buffer containing command to execute for restart
327 * Shutdown everything and perform a clean reboot.
328 * This is not safe to call in interrupt context.
330 void kernel_restart(char *cmd)
332 kernel_restart_prepare(cmd);
334 printk(KERN_EMERG "Restarting system.\n");
336 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
337 kmsg_dump(KMSG_DUMP_RESTART);
338 machine_restart(cmd);
340 EXPORT_SYMBOL_GPL(kernel_restart);
342 static void kernel_shutdown_prepare(enum system_states state)
344 blocking_notifier_call_chain(&reboot_notifier_list,
345 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
346 system_state = state;
347 usermodehelper_disable();
351 * kernel_halt - halt the system
353 * Shutdown everything and perform a clean system halt.
355 void kernel_halt(void)
357 kernel_shutdown_prepare(SYSTEM_HALT);
359 printk(KERN_EMERG "System halted.\n");
360 kmsg_dump(KMSG_DUMP_HALT);
364 EXPORT_SYMBOL_GPL(kernel_halt);
367 * kernel_power_off - power_off the system
369 * Shutdown everything and perform a clean system power_off.
371 void kernel_power_off(void)
373 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
374 if (pm_power_off_prepare)
375 pm_power_off_prepare();
376 disable_nonboot_cpus();
378 printk(KERN_EMERG "Power down.\n");
379 kmsg_dump(KMSG_DUMP_POWEROFF);
382 EXPORT_SYMBOL_GPL(kernel_power_off);
384 static DEFINE_MUTEX(reboot_mutex);
387 * Reboot system call: for obvious reasons only root may call it,
388 * and even root needs to set up some magic numbers in the registers
389 * so that some mistake won't make this reboot the whole machine.
390 * You can also set the meaning of the ctrl-alt-del-key here.
392 * reboot doesn't sync: do that yourself before calling this.
394 SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
400 /* We only trust the superuser with rebooting the system. */
401 if (!capable(CAP_SYS_BOOT))
404 /* For safety, we require "magic" arguments. */
405 if (magic1 != LINUX_REBOOT_MAGIC1 ||
406 (magic2 != LINUX_REBOOT_MAGIC2 &&
407 magic2 != LINUX_REBOOT_MAGIC2A &&
408 magic2 != LINUX_REBOOT_MAGIC2B &&
409 magic2 != LINUX_REBOOT_MAGIC2C))
412 /* Instead of trying to make the power_off code look like
413 * halt when pm_power_off is not set do it the easy way.
415 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
416 cmd = LINUX_REBOOT_CMD_HALT;
418 mutex_lock(&reboot_mutex);
420 case LINUX_REBOOT_CMD_RESTART:
421 kernel_restart(NULL);
424 case LINUX_REBOOT_CMD_CAD_ON:
428 case LINUX_REBOOT_CMD_CAD_OFF:
432 case LINUX_REBOOT_CMD_HALT:
435 panic("cannot halt");
437 case LINUX_REBOOT_CMD_POWER_OFF:
442 case LINUX_REBOOT_CMD_RESTART2:
443 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
447 buffer[sizeof(buffer) - 1] = '\0';
449 kernel_restart(buffer);
453 case LINUX_REBOOT_CMD_KEXEC:
454 ret = kernel_kexec();
458 #ifdef CONFIG_HIBERNATION
459 case LINUX_REBOOT_CMD_SW_SUSPEND:
468 mutex_unlock(&reboot_mutex);
472 static void deferred_cad(struct work_struct *dummy)
474 kernel_restart(NULL);
478 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
479 * As it's called within an interrupt, it may NOT sync: the only choice
480 * is whether to reboot at once, or just ignore the ctrl-alt-del.
482 void ctrl_alt_del(void)
484 static DECLARE_WORK(cad_work, deferred_cad);
487 schedule_work(&cad_work);
489 kill_cad_pid(SIGINT, 1);
493 * Unprivileged users may change the real gid to the effective gid
494 * or vice versa. (BSD-style)
496 * If you set the real gid at all, or set the effective gid to a value not
497 * equal to the real gid, then the saved gid is set to the new effective gid.
499 * This makes it possible for a setgid program to completely drop its
500 * privileges, which is often a useful assertion to make when you are doing
501 * a security audit over a program.
503 * The general idea is that a program which uses just setregid() will be
504 * 100% compatible with BSD. A program which uses just setgid() will be
505 * 100% compatible with POSIX with saved IDs.
507 * SMP: There are not races, the GIDs are checked only by filesystem
508 * operations (as far as semantic preservation is concerned).
510 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
512 const struct cred *old;
516 new = prepare_creds();
519 old = current_cred();
522 if (rgid != (gid_t) -1) {
523 if (old->gid == rgid ||
525 nsown_capable(CAP_SETGID))
530 if (egid != (gid_t) -1) {
531 if (old->gid == egid ||
534 nsown_capable(CAP_SETGID))
540 if (rgid != (gid_t) -1 ||
541 (egid != (gid_t) -1 && egid != old->gid))
542 new->sgid = new->egid;
543 new->fsgid = new->egid;
545 return commit_creds(new);
553 * setgid() is implemented like SysV w/ SAVED_IDS
555 * SMP: Same implicit races as above.
557 SYSCALL_DEFINE1(setgid, gid_t, gid)
559 const struct cred *old;
563 new = prepare_creds();
566 old = current_cred();
569 if (nsown_capable(CAP_SETGID))
570 new->gid = new->egid = new->sgid = new->fsgid = gid;
571 else if (gid == old->gid || gid == old->sgid)
572 new->egid = new->fsgid = gid;
576 return commit_creds(new);
584 * change the user struct in a credentials set to match the new UID
586 static int set_user(struct cred *new)
588 struct user_struct *new_user;
590 new_user = alloc_uid(current_user_ns(), new->uid);
594 if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
595 new_user != INIT_USER) {
601 new->user = new_user;
606 * Unprivileged users may change the real uid to the effective uid
607 * or vice versa. (BSD-style)
609 * If you set the real uid at all, or set the effective uid to a value not
610 * equal to the real uid, then the saved uid is set to the new effective uid.
612 * This makes it possible for a setuid program to completely drop its
613 * privileges, which is often a useful assertion to make when you are doing
614 * a security audit over a program.
616 * The general idea is that a program which uses just setreuid() will be
617 * 100% compatible with BSD. A program which uses just setuid() will be
618 * 100% compatible with POSIX with saved IDs.
620 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
622 const struct cred *old;
626 new = prepare_creds();
629 old = current_cred();
632 if (ruid != (uid_t) -1) {
634 if (old->uid != ruid &&
636 !nsown_capable(CAP_SETUID))
640 if (euid != (uid_t) -1) {
642 if (old->uid != euid &&
645 !nsown_capable(CAP_SETUID))
649 if (new->uid != old->uid) {
650 retval = set_user(new);
654 if (ruid != (uid_t) -1 ||
655 (euid != (uid_t) -1 && euid != old->uid))
656 new->suid = new->euid;
657 new->fsuid = new->euid;
659 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
663 return commit_creds(new);
671 * setuid() is implemented like SysV with SAVED_IDS
673 * Note that SAVED_ID's is deficient in that a setuid root program
674 * like sendmail, for example, cannot set its uid to be a normal
675 * user and then switch back, because if you're root, setuid() sets
676 * the saved uid too. If you don't like this, blame the bright people
677 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
678 * will allow a root program to temporarily drop privileges and be able to
679 * regain them by swapping the real and effective uid.
681 SYSCALL_DEFINE1(setuid, uid_t, uid)
683 const struct cred *old;
687 new = prepare_creds();
690 old = current_cred();
693 if (nsown_capable(CAP_SETUID)) {
694 new->suid = new->uid = uid;
695 if (uid != old->uid) {
696 retval = set_user(new);
700 } else if (uid != old->uid && uid != new->suid) {
704 new->fsuid = new->euid = uid;
706 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
710 return commit_creds(new);
719 * This function implements a generic ability to update ruid, euid,
720 * and suid. This allows you to implement the 4.4 compatible seteuid().
722 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
724 const struct cred *old;
728 new = prepare_creds();
732 old = current_cred();
735 if (!nsown_capable(CAP_SETUID)) {
736 if (ruid != (uid_t) -1 && ruid != old->uid &&
737 ruid != old->euid && ruid != old->suid)
739 if (euid != (uid_t) -1 && euid != old->uid &&
740 euid != old->euid && euid != old->suid)
742 if (suid != (uid_t) -1 && suid != old->uid &&
743 suid != old->euid && suid != old->suid)
747 if (ruid != (uid_t) -1) {
749 if (ruid != old->uid) {
750 retval = set_user(new);
755 if (euid != (uid_t) -1)
757 if (suid != (uid_t) -1)
759 new->fsuid = new->euid;
761 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
765 return commit_creds(new);
772 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruid, uid_t __user *, euid, uid_t __user *, suid)
774 const struct cred *cred = current_cred();
777 if (!(retval = put_user(cred->uid, ruid)) &&
778 !(retval = put_user(cred->euid, euid)))
779 retval = put_user(cred->suid, suid);
785 * Same as above, but for rgid, egid, sgid.
787 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
789 const struct cred *old;
793 new = prepare_creds();
796 old = current_cred();
799 if (!nsown_capable(CAP_SETGID)) {
800 if (rgid != (gid_t) -1 && rgid != old->gid &&
801 rgid != old->egid && rgid != old->sgid)
803 if (egid != (gid_t) -1 && egid != old->gid &&
804 egid != old->egid && egid != old->sgid)
806 if (sgid != (gid_t) -1 && sgid != old->gid &&
807 sgid != old->egid && sgid != old->sgid)
811 if (rgid != (gid_t) -1)
813 if (egid != (gid_t) -1)
815 if (sgid != (gid_t) -1)
817 new->fsgid = new->egid;
819 return commit_creds(new);
826 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgid, gid_t __user *, egid, gid_t __user *, sgid)
828 const struct cred *cred = current_cred();
831 if (!(retval = put_user(cred->gid, rgid)) &&
832 !(retval = put_user(cred->egid, egid)))
833 retval = put_user(cred->sgid, sgid);
840 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
841 * is used for "access()" and for the NFS daemon (letting nfsd stay at
842 * whatever uid it wants to). It normally shadows "euid", except when
843 * explicitly set by setfsuid() or for access..
845 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
847 const struct cred *old;
851 new = prepare_creds();
853 return current_fsuid();
854 old = current_cred();
855 old_fsuid = old->fsuid;
857 if (uid == old->uid || uid == old->euid ||
858 uid == old->suid || uid == old->fsuid ||
859 nsown_capable(CAP_SETUID)) {
860 if (uid != old_fsuid) {
862 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
876 * Samma på svenska..
878 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
880 const struct cred *old;
884 new = prepare_creds();
886 return current_fsgid();
887 old = current_cred();
888 old_fsgid = old->fsgid;
890 if (gid == old->gid || gid == old->egid ||
891 gid == old->sgid || gid == old->fsgid ||
892 nsown_capable(CAP_SETGID)) {
893 if (gid != old_fsgid) {
907 void do_sys_times(struct tms *tms)
909 cputime_t tgutime, tgstime, cutime, cstime;
911 spin_lock_irq(¤t->sighand->siglock);
912 thread_group_times(current, &tgutime, &tgstime);
913 cutime = current->signal->cutime;
914 cstime = current->signal->cstime;
915 spin_unlock_irq(¤t->sighand->siglock);
916 tms->tms_utime = cputime_to_clock_t(tgutime);
917 tms->tms_stime = cputime_to_clock_t(tgstime);
918 tms->tms_cutime = cputime_to_clock_t(cutime);
919 tms->tms_cstime = cputime_to_clock_t(cstime);
922 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
928 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
931 force_successful_syscall_return();
932 return (long) jiffies_64_to_clock_t(get_jiffies_64());
936 * This needs some heavy checking ...
937 * I just haven't the stomach for it. I also don't fully
938 * understand sessions/pgrp etc. Let somebody who does explain it.
940 * OK, I think I have the protection semantics right.... this is really
941 * only important on a multi-user system anyway, to make sure one user
942 * can't send a signal to a process owned by another. -TYT, 12/12/91
944 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
947 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
949 struct task_struct *p;
950 struct task_struct *group_leader = current->group_leader;
955 pid = task_pid_vnr(group_leader);
962 /* From this point forward we keep holding onto the tasklist lock
963 * so that our parent does not change from under us. -DaveM
965 write_lock_irq(&tasklist_lock);
968 p = find_task_by_vpid(pid);
973 if (!thread_group_leader(p))
976 if (same_thread_group(p->real_parent, group_leader)) {
978 if (task_session(p) != task_session(group_leader))
985 if (p != group_leader)
990 if (p->signal->leader)
995 struct task_struct *g;
997 pgrp = find_vpid(pgid);
998 g = pid_task(pgrp, PIDTYPE_PGID);
999 if (!g || task_session(g) != task_session(group_leader))
1003 err = security_task_setpgid(p, pgid);
1007 if (task_pgrp(p) != pgrp)
1008 change_pid(p, PIDTYPE_PGID, pgrp);
1012 /* All paths lead to here, thus we are safe. -DaveM */
1013 write_unlock_irq(&tasklist_lock);
1018 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1020 struct task_struct *p;
1026 grp = task_pgrp(current);
1029 p = find_task_by_vpid(pid);
1036 retval = security_task_getpgid(p);
1040 retval = pid_vnr(grp);
1046 #ifdef __ARCH_WANT_SYS_GETPGRP
1048 SYSCALL_DEFINE0(getpgrp)
1050 return sys_getpgid(0);
1055 SYSCALL_DEFINE1(getsid, pid_t, pid)
1057 struct task_struct *p;
1063 sid = task_session(current);
1066 p = find_task_by_vpid(pid);
1069 sid = task_session(p);
1073 retval = security_task_getsid(p);
1077 retval = pid_vnr(sid);
1083 SYSCALL_DEFINE0(setsid)
1085 struct task_struct *group_leader = current->group_leader;
1086 struct pid *sid = task_pid(group_leader);
1087 pid_t session = pid_vnr(sid);
1090 write_lock_irq(&tasklist_lock);
1091 /* Fail if I am already a session leader */
1092 if (group_leader->signal->leader)
1095 /* Fail if a process group id already exists that equals the
1096 * proposed session id.
1098 if (pid_task(sid, PIDTYPE_PGID))
1101 group_leader->signal->leader = 1;
1102 __set_special_pids(sid);
1104 proc_clear_tty(group_leader);
1108 write_unlock_irq(&tasklist_lock);
1110 proc_sid_connector(group_leader);
1111 sched_autogroup_create_attach(group_leader);
1116 DECLARE_RWSEM(uts_sem);
1118 #ifdef COMPAT_UTS_MACHINE
1119 #define override_architecture(name) \
1120 (personality(current->personality) == PER_LINUX32 && \
1121 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1122 sizeof(COMPAT_UTS_MACHINE)))
1124 #define override_architecture(name) 0
1127 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1131 down_read(&uts_sem);
1132 if (copy_to_user(name, utsname(), sizeof *name))
1136 if (!errno && override_architecture(name))
1141 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1145 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1152 down_read(&uts_sem);
1153 if (copy_to_user(name, utsname(), sizeof(*name)))
1157 if (!error && override_architecture(name))
1162 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1168 if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
1171 down_read(&uts_sem);
1172 error = __copy_to_user(&name->sysname, &utsname()->sysname,
1174 error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
1175 error |= __copy_to_user(&name->nodename, &utsname()->nodename,
1177 error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
1178 error |= __copy_to_user(&name->release, &utsname()->release,
1180 error |= __put_user(0, name->release + __OLD_UTS_LEN);
1181 error |= __copy_to_user(&name->version, &utsname()->version,
1183 error |= __put_user(0, name->version + __OLD_UTS_LEN);
1184 error |= __copy_to_user(&name->machine, &utsname()->machine,
1186 error |= __put_user(0, name->machine + __OLD_UTS_LEN);
1189 if (!error && override_architecture(name))
1191 return error ? -EFAULT : 0;
1195 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1198 char tmp[__NEW_UTS_LEN];
1200 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1203 if (len < 0 || len > __NEW_UTS_LEN)
1205 down_write(&uts_sem);
1207 if (!copy_from_user(tmp, name, len)) {
1208 struct new_utsname *u = utsname();
1210 memcpy(u->nodename, tmp, len);
1211 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1218 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1220 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1223 struct new_utsname *u;
1227 down_read(&uts_sem);
1229 i = 1 + strlen(u->nodename);
1233 if (copy_to_user(name, u->nodename, i))
1242 * Only setdomainname; getdomainname can be implemented by calling
1245 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1248 char tmp[__NEW_UTS_LEN];
1250 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1252 if (len < 0 || len > __NEW_UTS_LEN)
1255 down_write(&uts_sem);
1257 if (!copy_from_user(tmp, name, len)) {
1258 struct new_utsname *u = utsname();
1260 memcpy(u->domainname, tmp, len);
1261 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1268 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1270 struct rlimit value;
1273 ret = do_prlimit(current, resource, NULL, &value);
1275 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1280 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1283 * Back compatibility for getrlimit. Needed for some apps.
1286 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1287 struct rlimit __user *, rlim)
1290 if (resource >= RLIM_NLIMITS)
1293 task_lock(current->group_leader);
1294 x = current->signal->rlim[resource];
1295 task_unlock(current->group_leader);
1296 if (x.rlim_cur > 0x7FFFFFFF)
1297 x.rlim_cur = 0x7FFFFFFF;
1298 if (x.rlim_max > 0x7FFFFFFF)
1299 x.rlim_max = 0x7FFFFFFF;
1300 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1305 static inline bool rlim64_is_infinity(__u64 rlim64)
1307 #if BITS_PER_LONG < 64
1308 return rlim64 >= ULONG_MAX;
1310 return rlim64 == RLIM64_INFINITY;
1314 static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1316 if (rlim->rlim_cur == RLIM_INFINITY)
1317 rlim64->rlim_cur = RLIM64_INFINITY;
1319 rlim64->rlim_cur = rlim->rlim_cur;
1320 if (rlim->rlim_max == RLIM_INFINITY)
1321 rlim64->rlim_max = RLIM64_INFINITY;
1323 rlim64->rlim_max = rlim->rlim_max;
1326 static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1328 if (rlim64_is_infinity(rlim64->rlim_cur))
1329 rlim->rlim_cur = RLIM_INFINITY;
1331 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1332 if (rlim64_is_infinity(rlim64->rlim_max))
1333 rlim->rlim_max = RLIM_INFINITY;
1335 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1338 /* make sure you are allowed to change @tsk limits before calling this */
1339 int do_prlimit(struct task_struct *tsk, unsigned int resource,
1340 struct rlimit *new_rlim, struct rlimit *old_rlim)
1342 struct rlimit *rlim;
1345 if (resource >= RLIM_NLIMITS)
1348 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1350 if (resource == RLIMIT_NOFILE &&
1351 new_rlim->rlim_max > sysctl_nr_open)
1355 /* protect tsk->signal and tsk->sighand from disappearing */
1356 read_lock(&tasklist_lock);
1357 if (!tsk->sighand) {
1362 rlim = tsk->signal->rlim + resource;
1363 task_lock(tsk->group_leader);
1365 /* Keep the capable check against init_user_ns until
1366 cgroups can contain all limits */
1367 if (new_rlim->rlim_max > rlim->rlim_max &&
1368 !capable(CAP_SYS_RESOURCE))
1371 retval = security_task_setrlimit(tsk->group_leader,
1372 resource, new_rlim);
1373 if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) {
1375 * The caller is asking for an immediate RLIMIT_CPU
1376 * expiry. But we use the zero value to mean "it was
1377 * never set". So let's cheat and make it one second
1380 new_rlim->rlim_cur = 1;
1389 task_unlock(tsk->group_leader);
1392 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1393 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1394 * very long-standing error, and fixing it now risks breakage of
1395 * applications, so we live with it
1397 if (!retval && new_rlim && resource == RLIMIT_CPU &&
1398 new_rlim->rlim_cur != RLIM_INFINITY)
1399 update_rlimit_cpu(tsk, new_rlim->rlim_cur);
1401 read_unlock(&tasklist_lock);
1405 /* rcu lock must be held */
1406 static int check_prlimit_permission(struct task_struct *task)
1408 const struct cred *cred = current_cred(), *tcred;
1410 if (current == task)
1413 tcred = __task_cred(task);
1414 if (cred->user->user_ns == tcred->user->user_ns &&
1415 (cred->uid == tcred->euid &&
1416 cred->uid == tcred->suid &&
1417 cred->uid == tcred->uid &&
1418 cred->gid == tcred->egid &&
1419 cred->gid == tcred->sgid &&
1420 cred->gid == tcred->gid))
1422 if (ns_capable(tcred->user->user_ns, CAP_SYS_RESOURCE))
1428 SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1429 const struct rlimit64 __user *, new_rlim,
1430 struct rlimit64 __user *, old_rlim)
1432 struct rlimit64 old64, new64;
1433 struct rlimit old, new;
1434 struct task_struct *tsk;
1438 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1440 rlim64_to_rlim(&new64, &new);
1444 tsk = pid ? find_task_by_vpid(pid) : current;
1449 ret = check_prlimit_permission(tsk);
1454 get_task_struct(tsk);
1457 ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1458 old_rlim ? &old : NULL);
1460 if (!ret && old_rlim) {
1461 rlim_to_rlim64(&old, &old64);
1462 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1466 put_task_struct(tsk);
1470 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1472 struct rlimit new_rlim;
1474 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1476 return do_prlimit(current, resource, &new_rlim, NULL);
1480 * It would make sense to put struct rusage in the task_struct,
1481 * except that would make the task_struct be *really big*. After
1482 * task_struct gets moved into malloc'ed memory, it would
1483 * make sense to do this. It will make moving the rest of the information
1484 * a lot simpler! (Which we're not doing right now because we're not
1485 * measuring them yet).
1487 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1488 * races with threads incrementing their own counters. But since word
1489 * reads are atomic, we either get new values or old values and we don't
1490 * care which for the sums. We always take the siglock to protect reading
1491 * the c* fields from p->signal from races with exit.c updating those
1492 * fields when reaping, so a sample either gets all the additions of a
1493 * given child after it's reaped, or none so this sample is before reaping.
1496 * We need to take the siglock for CHILDEREN, SELF and BOTH
1497 * for the cases current multithreaded, non-current single threaded
1498 * non-current multithreaded. Thread traversal is now safe with
1500 * Strictly speaking, we donot need to take the siglock if we are current and
1501 * single threaded, as no one else can take our signal_struct away, no one
1502 * else can reap the children to update signal->c* counters, and no one else
1503 * can race with the signal-> fields. If we do not take any lock, the
1504 * signal-> fields could be read out of order while another thread was just
1505 * exiting. So we should place a read memory barrier when we avoid the lock.
1506 * On the writer side, write memory barrier is implied in __exit_signal
1507 * as __exit_signal releases the siglock spinlock after updating the signal->
1508 * fields. But we don't do this yet to keep things simple.
1512 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1514 r->ru_nvcsw += t->nvcsw;
1515 r->ru_nivcsw += t->nivcsw;
1516 r->ru_minflt += t->min_flt;
1517 r->ru_majflt += t->maj_flt;
1518 r->ru_inblock += task_io_get_inblock(t);
1519 r->ru_oublock += task_io_get_oublock(t);
1522 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1524 struct task_struct *t;
1525 unsigned long flags;
1526 cputime_t tgutime, tgstime, utime, stime;
1527 unsigned long maxrss = 0;
1529 memset((char *) r, 0, sizeof *r);
1530 utime = stime = cputime_zero;
1532 if (who == RUSAGE_THREAD) {
1533 task_times(current, &utime, &stime);
1534 accumulate_thread_rusage(p, r);
1535 maxrss = p->signal->maxrss;
1539 if (!lock_task_sighand(p, &flags))
1544 case RUSAGE_CHILDREN:
1545 utime = p->signal->cutime;
1546 stime = p->signal->cstime;
1547 r->ru_nvcsw = p->signal->cnvcsw;
1548 r->ru_nivcsw = p->signal->cnivcsw;
1549 r->ru_minflt = p->signal->cmin_flt;
1550 r->ru_majflt = p->signal->cmaj_flt;
1551 r->ru_inblock = p->signal->cinblock;
1552 r->ru_oublock = p->signal->coublock;
1553 maxrss = p->signal->cmaxrss;
1555 if (who == RUSAGE_CHILDREN)
1559 thread_group_times(p, &tgutime, &tgstime);
1560 utime = cputime_add(utime, tgutime);
1561 stime = cputime_add(stime, tgstime);
1562 r->ru_nvcsw += p->signal->nvcsw;
1563 r->ru_nivcsw += p->signal->nivcsw;
1564 r->ru_minflt += p->signal->min_flt;
1565 r->ru_majflt += p->signal->maj_flt;
1566 r->ru_inblock += p->signal->inblock;
1567 r->ru_oublock += p->signal->oublock;
1568 if (maxrss < p->signal->maxrss)
1569 maxrss = p->signal->maxrss;
1572 accumulate_thread_rusage(t, r);
1580 unlock_task_sighand(p, &flags);
1583 cputime_to_timeval(utime, &r->ru_utime);
1584 cputime_to_timeval(stime, &r->ru_stime);
1586 if (who != RUSAGE_CHILDREN) {
1587 struct mm_struct *mm = get_task_mm(p);
1589 setmax_mm_hiwater_rss(&maxrss, mm);
1593 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1596 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1599 k_getrusage(p, who, &r);
1600 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1603 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1605 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1606 who != RUSAGE_THREAD)
1608 return getrusage(current, who, ru);
1611 SYSCALL_DEFINE1(umask, int, mask)
1613 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1617 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
1618 unsigned long, arg4, unsigned long, arg5)
1620 struct task_struct *me = current;
1621 unsigned char comm[sizeof(me->comm)];
1624 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1625 if (error != -ENOSYS)
1630 case PR_SET_PDEATHSIG:
1631 if (!valid_signal(arg2)) {
1635 me->pdeath_signal = arg2;
1638 case PR_GET_PDEATHSIG:
1639 error = put_user(me->pdeath_signal, (int __user *)arg2);
1641 case PR_GET_DUMPABLE:
1642 error = get_dumpable(me->mm);
1644 case PR_SET_DUMPABLE:
1645 if (arg2 < 0 || arg2 > 1) {
1649 set_dumpable(me->mm, arg2);
1653 case PR_SET_UNALIGN:
1654 error = SET_UNALIGN_CTL(me, arg2);
1656 case PR_GET_UNALIGN:
1657 error = GET_UNALIGN_CTL(me, arg2);
1660 error = SET_FPEMU_CTL(me, arg2);
1663 error = GET_FPEMU_CTL(me, arg2);
1666 error = SET_FPEXC_CTL(me, arg2);
1669 error = GET_FPEXC_CTL(me, arg2);
1672 error = PR_TIMING_STATISTICAL;
1675 if (arg2 != PR_TIMING_STATISTICAL)
1682 comm[sizeof(me->comm)-1] = 0;
1683 if (strncpy_from_user(comm, (char __user *)arg2,
1684 sizeof(me->comm) - 1) < 0)
1686 set_task_comm(me, comm);
1689 get_task_comm(comm, me);
1690 if (copy_to_user((char __user *)arg2, comm,
1695 error = GET_ENDIAN(me, arg2);
1698 error = SET_ENDIAN(me, arg2);
1701 case PR_GET_SECCOMP:
1702 error = prctl_get_seccomp();
1704 case PR_SET_SECCOMP:
1705 error = prctl_set_seccomp(arg2);
1708 error = GET_TSC_CTL(arg2);
1711 error = SET_TSC_CTL(arg2);
1713 case PR_TASK_PERF_EVENTS_DISABLE:
1714 error = perf_event_task_disable();
1716 case PR_TASK_PERF_EVENTS_ENABLE:
1717 error = perf_event_task_enable();
1719 case PR_GET_TIMERSLACK:
1720 error = current->timer_slack_ns;
1722 case PR_SET_TIMERSLACK:
1724 current->timer_slack_ns =
1725 current->default_timer_slack_ns;
1727 current->timer_slack_ns = arg2;
1734 case PR_MCE_KILL_CLEAR:
1737 current->flags &= ~PF_MCE_PROCESS;
1739 case PR_MCE_KILL_SET:
1740 current->flags |= PF_MCE_PROCESS;
1741 if (arg3 == PR_MCE_KILL_EARLY)
1742 current->flags |= PF_MCE_EARLY;
1743 else if (arg3 == PR_MCE_KILL_LATE)
1744 current->flags &= ~PF_MCE_EARLY;
1745 else if (arg3 == PR_MCE_KILL_DEFAULT)
1747 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
1756 case PR_MCE_KILL_GET:
1757 if (arg2 | arg3 | arg4 | arg5)
1759 if (current->flags & PF_MCE_PROCESS)
1760 error = (current->flags & PF_MCE_EARLY) ?
1761 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
1763 error = PR_MCE_KILL_DEFAULT;
1772 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
1773 struct getcpu_cache __user *, unused)
1776 int cpu = raw_smp_processor_id();
1778 err |= put_user(cpu, cpup);
1780 err |= put_user(cpu_to_node(cpu), nodep);
1781 return err ? -EFAULT : 0;
1784 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
1786 static void argv_cleanup(struct subprocess_info *info)
1788 argv_free(info->argv);
1792 * orderly_poweroff - Trigger an orderly system poweroff
1793 * @force: force poweroff if command execution fails
1795 * This may be called from any context to trigger a system shutdown.
1796 * If the orderly shutdown fails, it will force an immediate shutdown.
1798 int orderly_poweroff(bool force)
1801 char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
1802 static char *envp[] = {
1804 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
1808 struct subprocess_info *info;
1811 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
1812 __func__, poweroff_cmd);
1816 info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
1822 call_usermodehelper_setfns(info, NULL, argv_cleanup, NULL);
1824 ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
1828 printk(KERN_WARNING "Failed to start orderly shutdown: "
1829 "forcing the issue\n");
1831 /* I guess this should try to kick off some daemon to
1832 sync and poweroff asap. Or not even bother syncing
1833 if we're doing an emergency shutdown? */
1840 EXPORT_SYMBOL_GPL(orderly_poweroff);