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;
322 * kernel_restart - reboot the system
323 * @cmd: pointer to buffer containing command to execute for restart
326 * Shutdown everything and perform a clean reboot.
327 * This is not safe to call in interrupt context.
329 void kernel_restart(char *cmd)
331 kernel_restart_prepare(cmd);
333 printk(KERN_EMERG "Restarting system.\n");
335 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
336 kmsg_dump(KMSG_DUMP_RESTART);
337 machine_restart(cmd);
339 EXPORT_SYMBOL_GPL(kernel_restart);
341 static void kernel_shutdown_prepare(enum system_states state)
343 blocking_notifier_call_chain(&reboot_notifier_list,
344 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
345 system_state = state;
349 * kernel_halt - halt the system
351 * Shutdown everything and perform a clean system halt.
353 void kernel_halt(void)
355 kernel_shutdown_prepare(SYSTEM_HALT);
357 printk(KERN_EMERG "System halted.\n");
358 kmsg_dump(KMSG_DUMP_HALT);
362 EXPORT_SYMBOL_GPL(kernel_halt);
365 * kernel_power_off - power_off the system
367 * Shutdown everything and perform a clean system power_off.
369 void kernel_power_off(void)
371 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
372 if (pm_power_off_prepare)
373 pm_power_off_prepare();
374 disable_nonboot_cpus();
376 printk(KERN_EMERG "Power down.\n");
377 kmsg_dump(KMSG_DUMP_POWEROFF);
380 EXPORT_SYMBOL_GPL(kernel_power_off);
382 static DEFINE_MUTEX(reboot_mutex);
385 * Reboot system call: for obvious reasons only root may call it,
386 * and even root needs to set up some magic numbers in the registers
387 * so that some mistake won't make this reboot the whole machine.
388 * You can also set the meaning of the ctrl-alt-del-key here.
390 * reboot doesn't sync: do that yourself before calling this.
392 SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
398 /* We only trust the superuser with rebooting the system. */
399 if (!capable(CAP_SYS_BOOT))
402 /* For safety, we require "magic" arguments. */
403 if (magic1 != LINUX_REBOOT_MAGIC1 ||
404 (magic2 != LINUX_REBOOT_MAGIC2 &&
405 magic2 != LINUX_REBOOT_MAGIC2A &&
406 magic2 != LINUX_REBOOT_MAGIC2B &&
407 magic2 != LINUX_REBOOT_MAGIC2C))
410 /* Instead of trying to make the power_off code look like
411 * halt when pm_power_off is not set do it the easy way.
413 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
414 cmd = LINUX_REBOOT_CMD_HALT;
416 mutex_lock(&reboot_mutex);
418 case LINUX_REBOOT_CMD_RESTART:
419 kernel_restart(NULL);
422 case LINUX_REBOOT_CMD_CAD_ON:
426 case LINUX_REBOOT_CMD_CAD_OFF:
430 case LINUX_REBOOT_CMD_HALT:
433 panic("cannot halt");
435 case LINUX_REBOOT_CMD_POWER_OFF:
440 case LINUX_REBOOT_CMD_RESTART2:
441 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
445 buffer[sizeof(buffer) - 1] = '\0';
447 kernel_restart(buffer);
451 case LINUX_REBOOT_CMD_KEXEC:
452 ret = kernel_kexec();
456 #ifdef CONFIG_HIBERNATION
457 case LINUX_REBOOT_CMD_SW_SUSPEND:
466 mutex_unlock(&reboot_mutex);
470 static void deferred_cad(struct work_struct *dummy)
472 kernel_restart(NULL);
476 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
477 * As it's called within an interrupt, it may NOT sync: the only choice
478 * is whether to reboot at once, or just ignore the ctrl-alt-del.
480 void ctrl_alt_del(void)
482 static DECLARE_WORK(cad_work, deferred_cad);
485 schedule_work(&cad_work);
487 kill_cad_pid(SIGINT, 1);
491 * Unprivileged users may change the real gid to the effective gid
492 * or vice versa. (BSD-style)
494 * If you set the real gid at all, or set the effective gid to a value not
495 * equal to the real gid, then the saved gid is set to the new effective gid.
497 * This makes it possible for a setgid program to completely drop its
498 * privileges, which is often a useful assertion to make when you are doing
499 * a security audit over a program.
501 * The general idea is that a program which uses just setregid() will be
502 * 100% compatible with BSD. A program which uses just setgid() will be
503 * 100% compatible with POSIX with saved IDs.
505 * SMP: There are not races, the GIDs are checked only by filesystem
506 * operations (as far as semantic preservation is concerned).
508 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
510 const struct cred *old;
514 new = prepare_creds();
517 old = current_cred();
520 if (rgid != (gid_t) -1) {
521 if (old->gid == rgid ||
523 nsown_capable(CAP_SETGID))
528 if (egid != (gid_t) -1) {
529 if (old->gid == egid ||
532 nsown_capable(CAP_SETGID))
538 if (rgid != (gid_t) -1 ||
539 (egid != (gid_t) -1 && egid != old->gid))
540 new->sgid = new->egid;
541 new->fsgid = new->egid;
543 return commit_creds(new);
551 * setgid() is implemented like SysV w/ SAVED_IDS
553 * SMP: Same implicit races as above.
555 SYSCALL_DEFINE1(setgid, gid_t, gid)
557 const struct cred *old;
561 new = prepare_creds();
564 old = current_cred();
567 if (nsown_capable(CAP_SETGID))
568 new->gid = new->egid = new->sgid = new->fsgid = gid;
569 else if (gid == old->gid || gid == old->sgid)
570 new->egid = new->fsgid = gid;
574 return commit_creds(new);
582 * change the user struct in a credentials set to match the new UID
584 static int set_user(struct cred *new)
586 struct user_struct *new_user;
588 new_user = alloc_uid(current_user_ns(), new->uid);
592 if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
593 new_user != INIT_USER) {
599 new->user = new_user;
604 * Unprivileged users may change the real uid to the effective uid
605 * or vice versa. (BSD-style)
607 * If you set the real uid at all, or set the effective uid to a value not
608 * equal to the real uid, then the saved uid is set to the new effective uid.
610 * This makes it possible for a setuid program to completely drop its
611 * privileges, which is often a useful assertion to make when you are doing
612 * a security audit over a program.
614 * The general idea is that a program which uses just setreuid() will be
615 * 100% compatible with BSD. A program which uses just setuid() will be
616 * 100% compatible with POSIX with saved IDs.
618 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
620 const struct cred *old;
624 new = prepare_creds();
627 old = current_cred();
630 if (ruid != (uid_t) -1) {
632 if (old->uid != ruid &&
634 !nsown_capable(CAP_SETUID))
638 if (euid != (uid_t) -1) {
640 if (old->uid != euid &&
643 !nsown_capable(CAP_SETUID))
647 if (new->uid != old->uid) {
648 retval = set_user(new);
652 if (ruid != (uid_t) -1 ||
653 (euid != (uid_t) -1 && euid != old->uid))
654 new->suid = new->euid;
655 new->fsuid = new->euid;
657 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
661 return commit_creds(new);
669 * setuid() is implemented like SysV with SAVED_IDS
671 * Note that SAVED_ID's is deficient in that a setuid root program
672 * like sendmail, for example, cannot set its uid to be a normal
673 * user and then switch back, because if you're root, setuid() sets
674 * the saved uid too. If you don't like this, blame the bright people
675 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
676 * will allow a root program to temporarily drop privileges and be able to
677 * regain them by swapping the real and effective uid.
679 SYSCALL_DEFINE1(setuid, uid_t, uid)
681 const struct cred *old;
685 new = prepare_creds();
688 old = current_cred();
691 if (nsown_capable(CAP_SETUID)) {
692 new->suid = new->uid = uid;
693 if (uid != old->uid) {
694 retval = set_user(new);
698 } else if (uid != old->uid && uid != new->suid) {
702 new->fsuid = new->euid = uid;
704 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
708 return commit_creds(new);
717 * This function implements a generic ability to update ruid, euid,
718 * and suid. This allows you to implement the 4.4 compatible seteuid().
720 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
722 const struct cred *old;
726 new = prepare_creds();
730 old = current_cred();
733 if (!nsown_capable(CAP_SETUID)) {
734 if (ruid != (uid_t) -1 && ruid != old->uid &&
735 ruid != old->euid && ruid != old->suid)
737 if (euid != (uid_t) -1 && euid != old->uid &&
738 euid != old->euid && euid != old->suid)
740 if (suid != (uid_t) -1 && suid != old->uid &&
741 suid != old->euid && suid != old->suid)
745 if (ruid != (uid_t) -1) {
747 if (ruid != old->uid) {
748 retval = set_user(new);
753 if (euid != (uid_t) -1)
755 if (suid != (uid_t) -1)
757 new->fsuid = new->euid;
759 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
763 return commit_creds(new);
770 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruid, uid_t __user *, euid, uid_t __user *, suid)
772 const struct cred *cred = current_cred();
775 if (!(retval = put_user(cred->uid, ruid)) &&
776 !(retval = put_user(cred->euid, euid)))
777 retval = put_user(cred->suid, suid);
783 * Same as above, but for rgid, egid, sgid.
785 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
787 const struct cred *old;
791 new = prepare_creds();
794 old = current_cred();
797 if (!nsown_capable(CAP_SETGID)) {
798 if (rgid != (gid_t) -1 && rgid != old->gid &&
799 rgid != old->egid && rgid != old->sgid)
801 if (egid != (gid_t) -1 && egid != old->gid &&
802 egid != old->egid && egid != old->sgid)
804 if (sgid != (gid_t) -1 && sgid != old->gid &&
805 sgid != old->egid && sgid != old->sgid)
809 if (rgid != (gid_t) -1)
811 if (egid != (gid_t) -1)
813 if (sgid != (gid_t) -1)
815 new->fsgid = new->egid;
817 return commit_creds(new);
824 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgid, gid_t __user *, egid, gid_t __user *, sgid)
826 const struct cred *cred = current_cred();
829 if (!(retval = put_user(cred->gid, rgid)) &&
830 !(retval = put_user(cred->egid, egid)))
831 retval = put_user(cred->sgid, sgid);
838 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
839 * is used for "access()" and for the NFS daemon (letting nfsd stay at
840 * whatever uid it wants to). It normally shadows "euid", except when
841 * explicitly set by setfsuid() or for access..
843 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
845 const struct cred *old;
849 new = prepare_creds();
851 return current_fsuid();
852 old = current_cred();
853 old_fsuid = old->fsuid;
855 if (uid == old->uid || uid == old->euid ||
856 uid == old->suid || uid == old->fsuid ||
857 nsown_capable(CAP_SETUID)) {
858 if (uid != old_fsuid) {
860 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
874 * Samma på svenska..
876 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
878 const struct cred *old;
882 new = prepare_creds();
884 return current_fsgid();
885 old = current_cred();
886 old_fsgid = old->fsgid;
888 if (gid == old->gid || gid == old->egid ||
889 gid == old->sgid || gid == old->fsgid ||
890 nsown_capable(CAP_SETGID)) {
891 if (gid != old_fsgid) {
905 void do_sys_times(struct tms *tms)
907 cputime_t tgutime, tgstime, cutime, cstime;
909 spin_lock_irq(¤t->sighand->siglock);
910 thread_group_times(current, &tgutime, &tgstime);
911 cutime = current->signal->cutime;
912 cstime = current->signal->cstime;
913 spin_unlock_irq(¤t->sighand->siglock);
914 tms->tms_utime = cputime_to_clock_t(tgutime);
915 tms->tms_stime = cputime_to_clock_t(tgstime);
916 tms->tms_cutime = cputime_to_clock_t(cutime);
917 tms->tms_cstime = cputime_to_clock_t(cstime);
920 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
926 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
929 force_successful_syscall_return();
930 return (long) jiffies_64_to_clock_t(get_jiffies_64());
934 * This needs some heavy checking ...
935 * I just haven't the stomach for it. I also don't fully
936 * understand sessions/pgrp etc. Let somebody who does explain it.
938 * OK, I think I have the protection semantics right.... this is really
939 * only important on a multi-user system anyway, to make sure one user
940 * can't send a signal to a process owned by another. -TYT, 12/12/91
942 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
945 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
947 struct task_struct *p;
948 struct task_struct *group_leader = current->group_leader;
953 pid = task_pid_vnr(group_leader);
960 /* From this point forward we keep holding onto the tasklist lock
961 * so that our parent does not change from under us. -DaveM
963 write_lock_irq(&tasklist_lock);
966 p = find_task_by_vpid(pid);
971 if (!thread_group_leader(p))
974 if (same_thread_group(p->real_parent, group_leader)) {
976 if (task_session(p) != task_session(group_leader))
983 if (p != group_leader)
988 if (p->signal->leader)
993 struct task_struct *g;
995 pgrp = find_vpid(pgid);
996 g = pid_task(pgrp, PIDTYPE_PGID);
997 if (!g || task_session(g) != task_session(group_leader))
1001 err = security_task_setpgid(p, pgid);
1005 if (task_pgrp(p) != pgrp)
1006 change_pid(p, PIDTYPE_PGID, pgrp);
1010 /* All paths lead to here, thus we are safe. -DaveM */
1011 write_unlock_irq(&tasklist_lock);
1016 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1018 struct task_struct *p;
1024 grp = task_pgrp(current);
1027 p = find_task_by_vpid(pid);
1034 retval = security_task_getpgid(p);
1038 retval = pid_vnr(grp);
1044 #ifdef __ARCH_WANT_SYS_GETPGRP
1046 SYSCALL_DEFINE0(getpgrp)
1048 return sys_getpgid(0);
1053 SYSCALL_DEFINE1(getsid, pid_t, pid)
1055 struct task_struct *p;
1061 sid = task_session(current);
1064 p = find_task_by_vpid(pid);
1067 sid = task_session(p);
1071 retval = security_task_getsid(p);
1075 retval = pid_vnr(sid);
1081 SYSCALL_DEFINE0(setsid)
1083 struct task_struct *group_leader = current->group_leader;
1084 struct pid *sid = task_pid(group_leader);
1085 pid_t session = pid_vnr(sid);
1088 write_lock_irq(&tasklist_lock);
1089 /* Fail if I am already a session leader */
1090 if (group_leader->signal->leader)
1093 /* Fail if a process group id already exists that equals the
1094 * proposed session id.
1096 if (pid_task(sid, PIDTYPE_PGID))
1099 group_leader->signal->leader = 1;
1100 __set_special_pids(sid);
1102 proc_clear_tty(group_leader);
1106 write_unlock_irq(&tasklist_lock);
1108 proc_sid_connector(group_leader);
1109 sched_autogroup_create_attach(group_leader);
1114 DECLARE_RWSEM(uts_sem);
1116 #ifdef COMPAT_UTS_MACHINE
1117 #define override_architecture(name) \
1118 (personality(current->personality) == PER_LINUX32 && \
1119 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1120 sizeof(COMPAT_UTS_MACHINE)))
1122 #define override_architecture(name) 0
1125 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1129 down_read(&uts_sem);
1130 if (copy_to_user(name, utsname(), sizeof *name))
1134 if (!errno && override_architecture(name))
1139 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1143 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1150 down_read(&uts_sem);
1151 if (copy_to_user(name, utsname(), sizeof(*name)))
1155 if (!error && override_architecture(name))
1160 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1166 if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
1169 down_read(&uts_sem);
1170 error = __copy_to_user(&name->sysname, &utsname()->sysname,
1172 error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
1173 error |= __copy_to_user(&name->nodename, &utsname()->nodename,
1175 error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
1176 error |= __copy_to_user(&name->release, &utsname()->release,
1178 error |= __put_user(0, name->release + __OLD_UTS_LEN);
1179 error |= __copy_to_user(&name->version, &utsname()->version,
1181 error |= __put_user(0, name->version + __OLD_UTS_LEN);
1182 error |= __copy_to_user(&name->machine, &utsname()->machine,
1184 error |= __put_user(0, name->machine + __OLD_UTS_LEN);
1187 if (!error && override_architecture(name))
1189 return error ? -EFAULT : 0;
1193 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1196 char tmp[__NEW_UTS_LEN];
1198 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1201 if (len < 0 || len > __NEW_UTS_LEN)
1203 down_write(&uts_sem);
1205 if (!copy_from_user(tmp, name, len)) {
1206 struct new_utsname *u = utsname();
1208 memcpy(u->nodename, tmp, len);
1209 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1216 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1218 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1221 struct new_utsname *u;
1225 down_read(&uts_sem);
1227 i = 1 + strlen(u->nodename);
1231 if (copy_to_user(name, u->nodename, i))
1240 * Only setdomainname; getdomainname can be implemented by calling
1243 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1246 char tmp[__NEW_UTS_LEN];
1248 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1250 if (len < 0 || len > __NEW_UTS_LEN)
1253 down_write(&uts_sem);
1255 if (!copy_from_user(tmp, name, len)) {
1256 struct new_utsname *u = utsname();
1258 memcpy(u->domainname, tmp, len);
1259 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1266 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1268 struct rlimit value;
1271 ret = do_prlimit(current, resource, NULL, &value);
1273 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1278 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1281 * Back compatibility for getrlimit. Needed for some apps.
1284 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1285 struct rlimit __user *, rlim)
1288 if (resource >= RLIM_NLIMITS)
1291 task_lock(current->group_leader);
1292 x = current->signal->rlim[resource];
1293 task_unlock(current->group_leader);
1294 if (x.rlim_cur > 0x7FFFFFFF)
1295 x.rlim_cur = 0x7FFFFFFF;
1296 if (x.rlim_max > 0x7FFFFFFF)
1297 x.rlim_max = 0x7FFFFFFF;
1298 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1303 static inline bool rlim64_is_infinity(__u64 rlim64)
1305 #if BITS_PER_LONG < 64
1306 return rlim64 >= ULONG_MAX;
1308 return rlim64 == RLIM64_INFINITY;
1312 static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1314 if (rlim->rlim_cur == RLIM_INFINITY)
1315 rlim64->rlim_cur = RLIM64_INFINITY;
1317 rlim64->rlim_cur = rlim->rlim_cur;
1318 if (rlim->rlim_max == RLIM_INFINITY)
1319 rlim64->rlim_max = RLIM64_INFINITY;
1321 rlim64->rlim_max = rlim->rlim_max;
1324 static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1326 if (rlim64_is_infinity(rlim64->rlim_cur))
1327 rlim->rlim_cur = RLIM_INFINITY;
1329 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1330 if (rlim64_is_infinity(rlim64->rlim_max))
1331 rlim->rlim_max = RLIM_INFINITY;
1333 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1336 /* make sure you are allowed to change @tsk limits before calling this */
1337 int do_prlimit(struct task_struct *tsk, unsigned int resource,
1338 struct rlimit *new_rlim, struct rlimit *old_rlim)
1340 struct rlimit *rlim;
1343 if (resource >= RLIM_NLIMITS)
1346 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1348 if (resource == RLIMIT_NOFILE &&
1349 new_rlim->rlim_max > sysctl_nr_open)
1353 /* protect tsk->signal and tsk->sighand from disappearing */
1354 read_lock(&tasklist_lock);
1355 if (!tsk->sighand) {
1360 rlim = tsk->signal->rlim + resource;
1361 task_lock(tsk->group_leader);
1363 /* Keep the capable check against init_user_ns until
1364 cgroups can contain all limits */
1365 if (new_rlim->rlim_max > rlim->rlim_max &&
1366 !capable(CAP_SYS_RESOURCE))
1369 retval = security_task_setrlimit(tsk->group_leader,
1370 resource, new_rlim);
1371 if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) {
1373 * The caller is asking for an immediate RLIMIT_CPU
1374 * expiry. But we use the zero value to mean "it was
1375 * never set". So let's cheat and make it one second
1378 new_rlim->rlim_cur = 1;
1387 task_unlock(tsk->group_leader);
1390 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1391 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1392 * very long-standing error, and fixing it now risks breakage of
1393 * applications, so we live with it
1395 if (!retval && new_rlim && resource == RLIMIT_CPU &&
1396 new_rlim->rlim_cur != RLIM_INFINITY)
1397 update_rlimit_cpu(tsk, new_rlim->rlim_cur);
1399 read_unlock(&tasklist_lock);
1403 /* rcu lock must be held */
1404 static int check_prlimit_permission(struct task_struct *task)
1406 const struct cred *cred = current_cred(), *tcred;
1408 if (current == task)
1411 tcred = __task_cred(task);
1412 if (cred->user->user_ns == tcred->user->user_ns &&
1413 (cred->uid == tcred->euid &&
1414 cred->uid == tcred->suid &&
1415 cred->uid == tcred->uid &&
1416 cred->gid == tcred->egid &&
1417 cred->gid == tcred->sgid &&
1418 cred->gid == tcred->gid))
1420 if (ns_capable(tcred->user->user_ns, CAP_SYS_RESOURCE))
1426 SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1427 const struct rlimit64 __user *, new_rlim,
1428 struct rlimit64 __user *, old_rlim)
1430 struct rlimit64 old64, new64;
1431 struct rlimit old, new;
1432 struct task_struct *tsk;
1436 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1438 rlim64_to_rlim(&new64, &new);
1442 tsk = pid ? find_task_by_vpid(pid) : current;
1447 ret = check_prlimit_permission(tsk);
1452 get_task_struct(tsk);
1455 ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1456 old_rlim ? &old : NULL);
1458 if (!ret && old_rlim) {
1459 rlim_to_rlim64(&old, &old64);
1460 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1464 put_task_struct(tsk);
1468 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1470 struct rlimit new_rlim;
1472 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1474 return do_prlimit(current, resource, &new_rlim, NULL);
1478 * It would make sense to put struct rusage in the task_struct,
1479 * except that would make the task_struct be *really big*. After
1480 * task_struct gets moved into malloc'ed memory, it would
1481 * make sense to do this. It will make moving the rest of the information
1482 * a lot simpler! (Which we're not doing right now because we're not
1483 * measuring them yet).
1485 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1486 * races with threads incrementing their own counters. But since word
1487 * reads are atomic, we either get new values or old values and we don't
1488 * care which for the sums. We always take the siglock to protect reading
1489 * the c* fields from p->signal from races with exit.c updating those
1490 * fields when reaping, so a sample either gets all the additions of a
1491 * given child after it's reaped, or none so this sample is before reaping.
1494 * We need to take the siglock for CHILDEREN, SELF and BOTH
1495 * for the cases current multithreaded, non-current single threaded
1496 * non-current multithreaded. Thread traversal is now safe with
1498 * Strictly speaking, we donot need to take the siglock if we are current and
1499 * single threaded, as no one else can take our signal_struct away, no one
1500 * else can reap the children to update signal->c* counters, and no one else
1501 * can race with the signal-> fields. If we do not take any lock, the
1502 * signal-> fields could be read out of order while another thread was just
1503 * exiting. So we should place a read memory barrier when we avoid the lock.
1504 * On the writer side, write memory barrier is implied in __exit_signal
1505 * as __exit_signal releases the siglock spinlock after updating the signal->
1506 * fields. But we don't do this yet to keep things simple.
1510 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1512 r->ru_nvcsw += t->nvcsw;
1513 r->ru_nivcsw += t->nivcsw;
1514 r->ru_minflt += t->min_flt;
1515 r->ru_majflt += t->maj_flt;
1516 r->ru_inblock += task_io_get_inblock(t);
1517 r->ru_oublock += task_io_get_oublock(t);
1520 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1522 struct task_struct *t;
1523 unsigned long flags;
1524 cputime_t tgutime, tgstime, utime, stime;
1525 unsigned long maxrss = 0;
1527 memset((char *) r, 0, sizeof *r);
1528 utime = stime = cputime_zero;
1530 if (who == RUSAGE_THREAD) {
1531 task_times(current, &utime, &stime);
1532 accumulate_thread_rusage(p, r);
1533 maxrss = p->signal->maxrss;
1537 if (!lock_task_sighand(p, &flags))
1542 case RUSAGE_CHILDREN:
1543 utime = p->signal->cutime;
1544 stime = p->signal->cstime;
1545 r->ru_nvcsw = p->signal->cnvcsw;
1546 r->ru_nivcsw = p->signal->cnivcsw;
1547 r->ru_minflt = p->signal->cmin_flt;
1548 r->ru_majflt = p->signal->cmaj_flt;
1549 r->ru_inblock = p->signal->cinblock;
1550 r->ru_oublock = p->signal->coublock;
1551 maxrss = p->signal->cmaxrss;
1553 if (who == RUSAGE_CHILDREN)
1557 thread_group_times(p, &tgutime, &tgstime);
1558 utime = cputime_add(utime, tgutime);
1559 stime = cputime_add(stime, tgstime);
1560 r->ru_nvcsw += p->signal->nvcsw;
1561 r->ru_nivcsw += p->signal->nivcsw;
1562 r->ru_minflt += p->signal->min_flt;
1563 r->ru_majflt += p->signal->maj_flt;
1564 r->ru_inblock += p->signal->inblock;
1565 r->ru_oublock += p->signal->oublock;
1566 if (maxrss < p->signal->maxrss)
1567 maxrss = p->signal->maxrss;
1570 accumulate_thread_rusage(t, r);
1578 unlock_task_sighand(p, &flags);
1581 cputime_to_timeval(utime, &r->ru_utime);
1582 cputime_to_timeval(stime, &r->ru_stime);
1584 if (who != RUSAGE_CHILDREN) {
1585 struct mm_struct *mm = get_task_mm(p);
1587 setmax_mm_hiwater_rss(&maxrss, mm);
1591 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1594 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1597 k_getrusage(p, who, &r);
1598 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1601 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1603 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1604 who != RUSAGE_THREAD)
1606 return getrusage(current, who, ru);
1609 SYSCALL_DEFINE1(umask, int, mask)
1611 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1615 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
1616 unsigned long, arg4, unsigned long, arg5)
1618 struct task_struct *me = current;
1619 unsigned char comm[sizeof(me->comm)];
1622 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1623 if (error != -ENOSYS)
1628 case PR_SET_PDEATHSIG:
1629 if (!valid_signal(arg2)) {
1633 me->pdeath_signal = arg2;
1636 case PR_GET_PDEATHSIG:
1637 error = put_user(me->pdeath_signal, (int __user *)arg2);
1639 case PR_GET_DUMPABLE:
1640 error = get_dumpable(me->mm);
1642 case PR_SET_DUMPABLE:
1643 if (arg2 < 0 || arg2 > 1) {
1647 set_dumpable(me->mm, arg2);
1651 case PR_SET_UNALIGN:
1652 error = SET_UNALIGN_CTL(me, arg2);
1654 case PR_GET_UNALIGN:
1655 error = GET_UNALIGN_CTL(me, arg2);
1658 error = SET_FPEMU_CTL(me, arg2);
1661 error = GET_FPEMU_CTL(me, arg2);
1664 error = SET_FPEXC_CTL(me, arg2);
1667 error = GET_FPEXC_CTL(me, arg2);
1670 error = PR_TIMING_STATISTICAL;
1673 if (arg2 != PR_TIMING_STATISTICAL)
1680 comm[sizeof(me->comm)-1] = 0;
1681 if (strncpy_from_user(comm, (char __user *)arg2,
1682 sizeof(me->comm) - 1) < 0)
1684 set_task_comm(me, comm);
1687 get_task_comm(comm, me);
1688 if (copy_to_user((char __user *)arg2, comm,
1693 error = GET_ENDIAN(me, arg2);
1696 error = SET_ENDIAN(me, arg2);
1699 case PR_GET_SECCOMP:
1700 error = prctl_get_seccomp();
1702 case PR_SET_SECCOMP:
1703 error = prctl_set_seccomp(arg2);
1706 error = GET_TSC_CTL(arg2);
1709 error = SET_TSC_CTL(arg2);
1711 case PR_TASK_PERF_EVENTS_DISABLE:
1712 error = perf_event_task_disable();
1714 case PR_TASK_PERF_EVENTS_ENABLE:
1715 error = perf_event_task_enable();
1717 case PR_GET_TIMERSLACK:
1718 error = current->timer_slack_ns;
1720 case PR_SET_TIMERSLACK:
1722 current->timer_slack_ns =
1723 current->default_timer_slack_ns;
1725 current->timer_slack_ns = arg2;
1732 case PR_MCE_KILL_CLEAR:
1735 current->flags &= ~PF_MCE_PROCESS;
1737 case PR_MCE_KILL_SET:
1738 current->flags |= PF_MCE_PROCESS;
1739 if (arg3 == PR_MCE_KILL_EARLY)
1740 current->flags |= PF_MCE_EARLY;
1741 else if (arg3 == PR_MCE_KILL_LATE)
1742 current->flags &= ~PF_MCE_EARLY;
1743 else if (arg3 == PR_MCE_KILL_DEFAULT)
1745 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
1754 case PR_MCE_KILL_GET:
1755 if (arg2 | arg3 | arg4 | arg5)
1757 if (current->flags & PF_MCE_PROCESS)
1758 error = (current->flags & PF_MCE_EARLY) ?
1759 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
1761 error = PR_MCE_KILL_DEFAULT;
1770 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
1771 struct getcpu_cache __user *, unused)
1774 int cpu = raw_smp_processor_id();
1776 err |= put_user(cpu, cpup);
1778 err |= put_user(cpu_to_node(cpu), nodep);
1779 return err ? -EFAULT : 0;
1782 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
1784 static void argv_cleanup(struct subprocess_info *info)
1786 argv_free(info->argv);
1790 * orderly_poweroff - Trigger an orderly system poweroff
1791 * @force: force poweroff if command execution fails
1793 * This may be called from any context to trigger a system shutdown.
1794 * If the orderly shutdown fails, it will force an immediate shutdown.
1796 int orderly_poweroff(bool force)
1799 char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
1800 static char *envp[] = {
1802 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
1806 struct subprocess_info *info;
1809 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
1810 __func__, poweroff_cmd);
1814 info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
1820 call_usermodehelper_setfns(info, NULL, argv_cleanup, NULL);
1822 ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
1826 printk(KERN_WARNING "Failed to start orderly shutdown: "
1827 "forcing the issue\n");
1829 /* I guess this should try to kick off some daemon to
1830 sync and poweroff asap. Or not even bother syncing
1831 if we're doing an emergency shutdown? */
1838 EXPORT_SYMBOL_GPL(orderly_poweroff);