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>
57 #ifndef SET_UNALIGN_CTL
58 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
60 #ifndef GET_UNALIGN_CTL
61 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
64 # define SET_FPEMU_CTL(a,b) (-EINVAL)
67 # define GET_FPEMU_CTL(a,b) (-EINVAL)
70 # define SET_FPEXC_CTL(a,b) (-EINVAL)
73 # define GET_FPEXC_CTL(a,b) (-EINVAL)
76 # define GET_ENDIAN(a,b) (-EINVAL)
79 # define SET_ENDIAN(a,b) (-EINVAL)
82 # define GET_TSC_CTL(a) (-EINVAL)
85 # define SET_TSC_CTL(a) (-EINVAL)
89 * this is where the system-wide overflow UID and GID are defined, for
90 * architectures that now have 32-bit UID/GID but didn't in the past
93 int overflowuid = DEFAULT_OVERFLOWUID;
94 int overflowgid = DEFAULT_OVERFLOWGID;
97 EXPORT_SYMBOL(overflowuid);
98 EXPORT_SYMBOL(overflowgid);
102 * the same as above, but for filesystems which can only store a 16-bit
103 * UID and GID. as such, this is needed on all architectures
106 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
107 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
109 EXPORT_SYMBOL(fs_overflowuid);
110 EXPORT_SYMBOL(fs_overflowgid);
113 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
118 EXPORT_SYMBOL(cad_pid);
121 * If set, this is used for preparing the system to power off.
124 void (*pm_power_off_prepare)(void);
127 * Returns true if current's euid is same as p's uid or euid,
128 * or has CAP_SYS_NICE to p's user_ns.
130 * Called with rcu_read_lock, creds are safe
132 static bool set_one_prio_perm(struct task_struct *p)
134 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
136 if (pcred->user->user_ns == cred->user->user_ns &&
137 (pcred->uid == cred->euid ||
138 pcred->euid == cred->euid))
140 if (ns_capable(pcred->user->user_ns, CAP_SYS_NICE))
146 * set the priority of a task
147 * - the caller must hold the RCU read lock
149 static int set_one_prio(struct task_struct *p, int niceval, int error)
153 if (!set_one_prio_perm(p)) {
157 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
161 no_nice = security_task_setnice(p, niceval);
168 set_user_nice(p, niceval);
173 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
175 struct task_struct *g, *p;
176 struct user_struct *user;
177 const struct cred *cred = current_cred();
181 if (which > PRIO_USER || which < PRIO_PROCESS)
184 /* normalize: avoid signed division (rounding problems) */
192 read_lock(&tasklist_lock);
196 p = find_task_by_vpid(who);
200 error = set_one_prio(p, niceval, error);
204 pgrp = find_vpid(who);
206 pgrp = task_pgrp(current);
207 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
208 error = set_one_prio(p, niceval, error);
209 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
212 user = (struct user_struct *) cred->user;
215 else if ((who != cred->uid) &&
216 !(user = find_user(who)))
217 goto out_unlock; /* No processes for this user */
219 do_each_thread(g, p) {
220 if (__task_cred(p)->uid == who)
221 error = set_one_prio(p, niceval, error);
222 } while_each_thread(g, p);
223 if (who != cred->uid)
224 free_uid(user); /* For find_user() */
228 read_unlock(&tasklist_lock);
235 * Ugh. To avoid negative return values, "getpriority()" will
236 * not return the normal nice-value, but a negated value that
237 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
238 * to stay compatible.
240 SYSCALL_DEFINE2(getpriority, int, which, int, who)
242 struct task_struct *g, *p;
243 struct user_struct *user;
244 const struct cred *cred = current_cred();
245 long niceval, retval = -ESRCH;
248 if (which > PRIO_USER || which < PRIO_PROCESS)
252 read_lock(&tasklist_lock);
256 p = find_task_by_vpid(who);
260 niceval = 20 - task_nice(p);
261 if (niceval > retval)
267 pgrp = find_vpid(who);
269 pgrp = task_pgrp(current);
270 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
271 niceval = 20 - task_nice(p);
272 if (niceval > retval)
274 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
277 user = (struct user_struct *) cred->user;
280 else if ((who != cred->uid) &&
281 !(user = find_user(who)))
282 goto out_unlock; /* No processes for this user */
284 do_each_thread(g, p) {
285 if (__task_cred(p)->uid == who) {
286 niceval = 20 - task_nice(p);
287 if (niceval > retval)
290 } while_each_thread(g, p);
291 if (who != cred->uid)
292 free_uid(user); /* for find_user() */
296 read_unlock(&tasklist_lock);
303 * emergency_restart - reboot the system
305 * Without shutting down any hardware or taking any locks
306 * reboot the system. This is called when we know we are in
307 * trouble so this is our best effort to reboot. This is
308 * safe to call in interrupt context.
310 void emergency_restart(void)
312 kmsg_dump(KMSG_DUMP_EMERG);
313 machine_emergency_restart();
315 EXPORT_SYMBOL_GPL(emergency_restart);
317 void kernel_restart_prepare(char *cmd)
319 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
320 system_state = SYSTEM_RESTART;
321 usermodehelper_disable();
327 * kernel_restart - reboot the system
328 * @cmd: pointer to buffer containing command to execute for restart
331 * Shutdown everything and perform a clean reboot.
332 * This is not safe to call in interrupt context.
334 void kernel_restart(char *cmd)
336 kernel_restart_prepare(cmd);
338 printk(KERN_EMERG "Restarting system.\n");
340 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
341 kmsg_dump(KMSG_DUMP_RESTART);
342 machine_restart(cmd);
344 EXPORT_SYMBOL_GPL(kernel_restart);
346 static void kernel_shutdown_prepare(enum system_states state)
348 blocking_notifier_call_chain(&reboot_notifier_list,
349 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
350 system_state = state;
351 usermodehelper_disable();
355 * kernel_halt - halt the system
357 * Shutdown everything and perform a clean system halt.
359 void kernel_halt(void)
361 kernel_shutdown_prepare(SYSTEM_HALT);
363 printk(KERN_EMERG "System halted.\n");
364 kmsg_dump(KMSG_DUMP_HALT);
368 EXPORT_SYMBOL_GPL(kernel_halt);
371 * kernel_power_off - power_off the system
373 * Shutdown everything and perform a clean system power_off.
375 void kernel_power_off(void)
377 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
378 if (pm_power_off_prepare)
379 pm_power_off_prepare();
380 disable_nonboot_cpus();
382 printk(KERN_EMERG "Power down.\n");
383 kmsg_dump(KMSG_DUMP_POWEROFF);
386 EXPORT_SYMBOL_GPL(kernel_power_off);
388 static DEFINE_MUTEX(reboot_mutex);
391 * Reboot system call: for obvious reasons only root may call it,
392 * and even root needs to set up some magic numbers in the registers
393 * so that some mistake won't make this reboot the whole machine.
394 * You can also set the meaning of the ctrl-alt-del-key here.
396 * reboot doesn't sync: do that yourself before calling this.
398 SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
404 /* We only trust the superuser with rebooting the system. */
405 if (!capable(CAP_SYS_BOOT))
408 /* For safety, we require "magic" arguments. */
409 if (magic1 != LINUX_REBOOT_MAGIC1 ||
410 (magic2 != LINUX_REBOOT_MAGIC2 &&
411 magic2 != LINUX_REBOOT_MAGIC2A &&
412 magic2 != LINUX_REBOOT_MAGIC2B &&
413 magic2 != LINUX_REBOOT_MAGIC2C))
416 /* Instead of trying to make the power_off code look like
417 * halt when pm_power_off is not set do it the easy way.
419 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
420 cmd = LINUX_REBOOT_CMD_HALT;
422 mutex_lock(&reboot_mutex);
424 case LINUX_REBOOT_CMD_RESTART:
425 kernel_restart(NULL);
428 case LINUX_REBOOT_CMD_CAD_ON:
432 case LINUX_REBOOT_CMD_CAD_OFF:
436 case LINUX_REBOOT_CMD_HALT:
439 panic("cannot halt");
441 case LINUX_REBOOT_CMD_POWER_OFF:
446 case LINUX_REBOOT_CMD_RESTART2:
447 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
451 buffer[sizeof(buffer) - 1] = '\0';
453 kernel_restart(buffer);
457 case LINUX_REBOOT_CMD_KEXEC:
458 ret = kernel_kexec();
462 #ifdef CONFIG_HIBERNATION
463 case LINUX_REBOOT_CMD_SW_SUSPEND:
472 mutex_unlock(&reboot_mutex);
476 static void deferred_cad(struct work_struct *dummy)
478 kernel_restart(NULL);
482 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
483 * As it's called within an interrupt, it may NOT sync: the only choice
484 * is whether to reboot at once, or just ignore the ctrl-alt-del.
486 void ctrl_alt_del(void)
488 static DECLARE_WORK(cad_work, deferred_cad);
491 schedule_work(&cad_work);
493 kill_cad_pid(SIGINT, 1);
497 * Unprivileged users may change the real gid to the effective gid
498 * or vice versa. (BSD-style)
500 * If you set the real gid at all, or set the effective gid to a value not
501 * equal to the real gid, then the saved gid is set to the new effective gid.
503 * This makes it possible for a setgid program to completely drop its
504 * privileges, which is often a useful assertion to make when you are doing
505 * a security audit over a program.
507 * The general idea is that a program which uses just setregid() will be
508 * 100% compatible with BSD. A program which uses just setgid() will be
509 * 100% compatible with POSIX with saved IDs.
511 * SMP: There are not races, the GIDs are checked only by filesystem
512 * operations (as far as semantic preservation is concerned).
514 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
516 const struct cred *old;
520 new = prepare_creds();
523 old = current_cred();
526 if (rgid != (gid_t) -1) {
527 if (old->gid == rgid ||
529 nsown_capable(CAP_SETGID))
534 if (egid != (gid_t) -1) {
535 if (old->gid == egid ||
538 nsown_capable(CAP_SETGID))
544 if (rgid != (gid_t) -1 ||
545 (egid != (gid_t) -1 && egid != old->gid))
546 new->sgid = new->egid;
547 new->fsgid = new->egid;
549 return commit_creds(new);
557 * setgid() is implemented like SysV w/ SAVED_IDS
559 * SMP: Same implicit races as above.
561 SYSCALL_DEFINE1(setgid, gid_t, gid)
563 const struct cred *old;
567 new = prepare_creds();
570 old = current_cred();
573 if (nsown_capable(CAP_SETGID))
574 new->gid = new->egid = new->sgid = new->fsgid = gid;
575 else if (gid == old->gid || gid == old->sgid)
576 new->egid = new->fsgid = gid;
580 return commit_creds(new);
588 * change the user struct in a credentials set to match the new UID
590 static int set_user(struct cred *new)
592 struct user_struct *new_user;
594 new_user = alloc_uid(current_user_ns(), new->uid);
598 if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
599 new_user != INIT_USER) {
605 new->user = new_user;
610 * Unprivileged users may change the real uid to the effective uid
611 * or vice versa. (BSD-style)
613 * If you set the real uid at all, or set the effective uid to a value not
614 * equal to the real uid, then the saved uid is set to the new effective uid.
616 * This makes it possible for a setuid program to completely drop its
617 * privileges, which is often a useful assertion to make when you are doing
618 * a security audit over a program.
620 * The general idea is that a program which uses just setreuid() will be
621 * 100% compatible with BSD. A program which uses just setuid() will be
622 * 100% compatible with POSIX with saved IDs.
624 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
626 const struct cred *old;
630 new = prepare_creds();
633 old = current_cred();
636 if (ruid != (uid_t) -1) {
638 if (old->uid != ruid &&
640 !nsown_capable(CAP_SETUID))
644 if (euid != (uid_t) -1) {
646 if (old->uid != euid &&
649 !nsown_capable(CAP_SETUID))
653 if (new->uid != old->uid) {
654 retval = set_user(new);
658 if (ruid != (uid_t) -1 ||
659 (euid != (uid_t) -1 && euid != old->uid))
660 new->suid = new->euid;
661 new->fsuid = new->euid;
663 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
667 return commit_creds(new);
675 * setuid() is implemented like SysV with SAVED_IDS
677 * Note that SAVED_ID's is deficient in that a setuid root program
678 * like sendmail, for example, cannot set its uid to be a normal
679 * user and then switch back, because if you're root, setuid() sets
680 * the saved uid too. If you don't like this, blame the bright people
681 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
682 * will allow a root program to temporarily drop privileges and be able to
683 * regain them by swapping the real and effective uid.
685 SYSCALL_DEFINE1(setuid, uid_t, uid)
687 const struct cred *old;
691 new = prepare_creds();
694 old = current_cred();
697 if (nsown_capable(CAP_SETUID)) {
698 new->suid = new->uid = uid;
699 if (uid != old->uid) {
700 retval = set_user(new);
704 } else if (uid != old->uid && uid != new->suid) {
708 new->fsuid = new->euid = uid;
710 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
714 return commit_creds(new);
723 * This function implements a generic ability to update ruid, euid,
724 * and suid. This allows you to implement the 4.4 compatible seteuid().
726 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
728 const struct cred *old;
732 new = prepare_creds();
736 old = current_cred();
739 if (!nsown_capable(CAP_SETUID)) {
740 if (ruid != (uid_t) -1 && ruid != old->uid &&
741 ruid != old->euid && ruid != old->suid)
743 if (euid != (uid_t) -1 && euid != old->uid &&
744 euid != old->euid && euid != old->suid)
746 if (suid != (uid_t) -1 && suid != old->uid &&
747 suid != old->euid && suid != old->suid)
751 if (ruid != (uid_t) -1) {
753 if (ruid != old->uid) {
754 retval = set_user(new);
759 if (euid != (uid_t) -1)
761 if (suid != (uid_t) -1)
763 new->fsuid = new->euid;
765 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
769 return commit_creds(new);
776 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruid, uid_t __user *, euid, uid_t __user *, suid)
778 const struct cred *cred = current_cred();
781 if (!(retval = put_user(cred->uid, ruid)) &&
782 !(retval = put_user(cred->euid, euid)))
783 retval = put_user(cred->suid, suid);
789 * Same as above, but for rgid, egid, sgid.
791 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
793 const struct cred *old;
797 new = prepare_creds();
800 old = current_cred();
803 if (!nsown_capable(CAP_SETGID)) {
804 if (rgid != (gid_t) -1 && rgid != old->gid &&
805 rgid != old->egid && rgid != old->sgid)
807 if (egid != (gid_t) -1 && egid != old->gid &&
808 egid != old->egid && egid != old->sgid)
810 if (sgid != (gid_t) -1 && sgid != old->gid &&
811 sgid != old->egid && sgid != old->sgid)
815 if (rgid != (gid_t) -1)
817 if (egid != (gid_t) -1)
819 if (sgid != (gid_t) -1)
821 new->fsgid = new->egid;
823 return commit_creds(new);
830 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgid, gid_t __user *, egid, gid_t __user *, sgid)
832 const struct cred *cred = current_cred();
835 if (!(retval = put_user(cred->gid, rgid)) &&
836 !(retval = put_user(cred->egid, egid)))
837 retval = put_user(cred->sgid, sgid);
844 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
845 * is used for "access()" and for the NFS daemon (letting nfsd stay at
846 * whatever uid it wants to). It normally shadows "euid", except when
847 * explicitly set by setfsuid() or for access..
849 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
851 const struct cred *old;
855 new = prepare_creds();
857 return current_fsuid();
858 old = current_cred();
859 old_fsuid = old->fsuid;
861 if (uid == old->uid || uid == old->euid ||
862 uid == old->suid || uid == old->fsuid ||
863 nsown_capable(CAP_SETUID)) {
864 if (uid != old_fsuid) {
866 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
880 * Samma på svenska..
882 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
884 const struct cred *old;
888 new = prepare_creds();
890 return current_fsgid();
891 old = current_cred();
892 old_fsgid = old->fsgid;
894 if (gid == old->gid || gid == old->egid ||
895 gid == old->sgid || gid == old->fsgid ||
896 nsown_capable(CAP_SETGID)) {
897 if (gid != old_fsgid) {
911 void do_sys_times(struct tms *tms)
913 cputime_t tgutime, tgstime, cutime, cstime;
915 spin_lock_irq(¤t->sighand->siglock);
916 thread_group_times(current, &tgutime, &tgstime);
917 cutime = current->signal->cutime;
918 cstime = current->signal->cstime;
919 spin_unlock_irq(¤t->sighand->siglock);
920 tms->tms_utime = cputime_to_clock_t(tgutime);
921 tms->tms_stime = cputime_to_clock_t(tgstime);
922 tms->tms_cutime = cputime_to_clock_t(cutime);
923 tms->tms_cstime = cputime_to_clock_t(cstime);
926 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
932 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
935 force_successful_syscall_return();
936 return (long) jiffies_64_to_clock_t(get_jiffies_64());
940 * This needs some heavy checking ...
941 * I just haven't the stomach for it. I also don't fully
942 * understand sessions/pgrp etc. Let somebody who does explain it.
944 * OK, I think I have the protection semantics right.... this is really
945 * only important on a multi-user system anyway, to make sure one user
946 * can't send a signal to a process owned by another. -TYT, 12/12/91
948 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
951 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
953 struct task_struct *p;
954 struct task_struct *group_leader = current->group_leader;
959 pid = task_pid_vnr(group_leader);
966 /* From this point forward we keep holding onto the tasklist lock
967 * so that our parent does not change from under us. -DaveM
969 write_lock_irq(&tasklist_lock);
972 p = find_task_by_vpid(pid);
977 if (!thread_group_leader(p))
980 if (same_thread_group(p->real_parent, group_leader)) {
982 if (task_session(p) != task_session(group_leader))
989 if (p != group_leader)
994 if (p->signal->leader)
999 struct task_struct *g;
1001 pgrp = find_vpid(pgid);
1002 g = pid_task(pgrp, PIDTYPE_PGID);
1003 if (!g || task_session(g) != task_session(group_leader))
1007 err = security_task_setpgid(p, pgid);
1011 if (task_pgrp(p) != pgrp)
1012 change_pid(p, PIDTYPE_PGID, pgrp);
1016 /* All paths lead to here, thus we are safe. -DaveM */
1017 write_unlock_irq(&tasklist_lock);
1022 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1024 struct task_struct *p;
1030 grp = task_pgrp(current);
1033 p = find_task_by_vpid(pid);
1040 retval = security_task_getpgid(p);
1044 retval = pid_vnr(grp);
1050 #ifdef __ARCH_WANT_SYS_GETPGRP
1052 SYSCALL_DEFINE0(getpgrp)
1054 return sys_getpgid(0);
1059 SYSCALL_DEFINE1(getsid, pid_t, pid)
1061 struct task_struct *p;
1067 sid = task_session(current);
1070 p = find_task_by_vpid(pid);
1073 sid = task_session(p);
1077 retval = security_task_getsid(p);
1081 retval = pid_vnr(sid);
1087 SYSCALL_DEFINE0(setsid)
1089 struct task_struct *group_leader = current->group_leader;
1090 struct pid *sid = task_pid(group_leader);
1091 pid_t session = pid_vnr(sid);
1094 write_lock_irq(&tasklist_lock);
1095 /* Fail if I am already a session leader */
1096 if (group_leader->signal->leader)
1099 /* Fail if a process group id already exists that equals the
1100 * proposed session id.
1102 if (pid_task(sid, PIDTYPE_PGID))
1105 group_leader->signal->leader = 1;
1106 __set_special_pids(sid);
1108 proc_clear_tty(group_leader);
1112 write_unlock_irq(&tasklist_lock);
1114 proc_sid_connector(group_leader);
1115 sched_autogroup_create_attach(group_leader);
1120 DECLARE_RWSEM(uts_sem);
1122 #ifdef COMPAT_UTS_MACHINE
1123 #define override_architecture(name) \
1124 (personality(current->personality) == PER_LINUX32 && \
1125 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1126 sizeof(COMPAT_UTS_MACHINE)))
1128 #define override_architecture(name) 0
1132 * Work around broken programs that cannot handle "Linux 3.0".
1133 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1135 static int override_release(char __user *release, int len)
1140 if (current->personality & UNAME26) {
1141 char *rest = UTS_RELEASE;
1146 if (*rest == '.' && ++ndots >= 3)
1148 if (!isdigit(*rest) && *rest != '.')
1152 v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 40;
1153 snprintf(buf, len, "2.6.%u%s", v, rest);
1154 ret = copy_to_user(release, buf, len);
1159 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1163 down_read(&uts_sem);
1164 if (copy_to_user(name, utsname(), sizeof *name))
1168 if (!errno && override_release(name->release, sizeof(name->release)))
1170 if (!errno && override_architecture(name))
1175 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1179 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1186 down_read(&uts_sem);
1187 if (copy_to_user(name, utsname(), sizeof(*name)))
1191 if (!error && override_release(name->release, sizeof(name->release)))
1193 if (!error && override_architecture(name))
1198 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1204 if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
1207 down_read(&uts_sem);
1208 error = __copy_to_user(&name->sysname, &utsname()->sysname,
1210 error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
1211 error |= __copy_to_user(&name->nodename, &utsname()->nodename,
1213 error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
1214 error |= __copy_to_user(&name->release, &utsname()->release,
1216 error |= __put_user(0, name->release + __OLD_UTS_LEN);
1217 error |= __copy_to_user(&name->version, &utsname()->version,
1219 error |= __put_user(0, name->version + __OLD_UTS_LEN);
1220 error |= __copy_to_user(&name->machine, &utsname()->machine,
1222 error |= __put_user(0, name->machine + __OLD_UTS_LEN);
1225 if (!error && override_architecture(name))
1227 if (!error && override_release(name->release, sizeof(name->release)))
1229 return error ? -EFAULT : 0;
1233 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1236 char tmp[__NEW_UTS_LEN];
1238 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1241 if (len < 0 || len > __NEW_UTS_LEN)
1243 down_write(&uts_sem);
1245 if (!copy_from_user(tmp, name, len)) {
1246 struct new_utsname *u = utsname();
1248 memcpy(u->nodename, tmp, len);
1249 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1256 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1258 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1261 struct new_utsname *u;
1265 down_read(&uts_sem);
1267 i = 1 + strlen(u->nodename);
1271 if (copy_to_user(name, u->nodename, i))
1280 * Only setdomainname; getdomainname can be implemented by calling
1283 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1286 char tmp[__NEW_UTS_LEN];
1288 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1290 if (len < 0 || len > __NEW_UTS_LEN)
1293 down_write(&uts_sem);
1295 if (!copy_from_user(tmp, name, len)) {
1296 struct new_utsname *u = utsname();
1298 memcpy(u->domainname, tmp, len);
1299 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1306 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1308 struct rlimit value;
1311 ret = do_prlimit(current, resource, NULL, &value);
1313 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1318 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1321 * Back compatibility for getrlimit. Needed for some apps.
1324 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1325 struct rlimit __user *, rlim)
1328 if (resource >= RLIM_NLIMITS)
1331 task_lock(current->group_leader);
1332 x = current->signal->rlim[resource];
1333 task_unlock(current->group_leader);
1334 if (x.rlim_cur > 0x7FFFFFFF)
1335 x.rlim_cur = 0x7FFFFFFF;
1336 if (x.rlim_max > 0x7FFFFFFF)
1337 x.rlim_max = 0x7FFFFFFF;
1338 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1343 static inline bool rlim64_is_infinity(__u64 rlim64)
1345 #if BITS_PER_LONG < 64
1346 return rlim64 >= ULONG_MAX;
1348 return rlim64 == RLIM64_INFINITY;
1352 static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1354 if (rlim->rlim_cur == RLIM_INFINITY)
1355 rlim64->rlim_cur = RLIM64_INFINITY;
1357 rlim64->rlim_cur = rlim->rlim_cur;
1358 if (rlim->rlim_max == RLIM_INFINITY)
1359 rlim64->rlim_max = RLIM64_INFINITY;
1361 rlim64->rlim_max = rlim->rlim_max;
1364 static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1366 if (rlim64_is_infinity(rlim64->rlim_cur))
1367 rlim->rlim_cur = RLIM_INFINITY;
1369 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1370 if (rlim64_is_infinity(rlim64->rlim_max))
1371 rlim->rlim_max = RLIM_INFINITY;
1373 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1376 /* make sure you are allowed to change @tsk limits before calling this */
1377 int do_prlimit(struct task_struct *tsk, unsigned int resource,
1378 struct rlimit *new_rlim, struct rlimit *old_rlim)
1380 struct rlimit *rlim;
1383 if (resource >= RLIM_NLIMITS)
1386 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1388 if (resource == RLIMIT_NOFILE &&
1389 new_rlim->rlim_max > sysctl_nr_open)
1393 /* protect tsk->signal and tsk->sighand from disappearing */
1394 read_lock(&tasklist_lock);
1395 if (!tsk->sighand) {
1400 rlim = tsk->signal->rlim + resource;
1401 task_lock(tsk->group_leader);
1403 /* Keep the capable check against init_user_ns until
1404 cgroups can contain all limits */
1405 if (new_rlim->rlim_max > rlim->rlim_max &&
1406 !capable(CAP_SYS_RESOURCE))
1409 retval = security_task_setrlimit(tsk->group_leader,
1410 resource, new_rlim);
1411 if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) {
1413 * The caller is asking for an immediate RLIMIT_CPU
1414 * expiry. But we use the zero value to mean "it was
1415 * never set". So let's cheat and make it one second
1418 new_rlim->rlim_cur = 1;
1427 task_unlock(tsk->group_leader);
1430 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1431 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1432 * very long-standing error, and fixing it now risks breakage of
1433 * applications, so we live with it
1435 if (!retval && new_rlim && resource == RLIMIT_CPU &&
1436 new_rlim->rlim_cur != RLIM_INFINITY)
1437 update_rlimit_cpu(tsk, new_rlim->rlim_cur);
1439 read_unlock(&tasklist_lock);
1443 /* rcu lock must be held */
1444 static int check_prlimit_permission(struct task_struct *task)
1446 const struct cred *cred = current_cred(), *tcred;
1448 if (current == task)
1451 tcred = __task_cred(task);
1452 if (cred->user->user_ns == tcred->user->user_ns &&
1453 (cred->uid == tcred->euid &&
1454 cred->uid == tcred->suid &&
1455 cred->uid == tcred->uid &&
1456 cred->gid == tcred->egid &&
1457 cred->gid == tcred->sgid &&
1458 cred->gid == tcred->gid))
1460 if (ns_capable(tcred->user->user_ns, CAP_SYS_RESOURCE))
1466 SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1467 const struct rlimit64 __user *, new_rlim,
1468 struct rlimit64 __user *, old_rlim)
1470 struct rlimit64 old64, new64;
1471 struct rlimit old, new;
1472 struct task_struct *tsk;
1476 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1478 rlim64_to_rlim(&new64, &new);
1482 tsk = pid ? find_task_by_vpid(pid) : current;
1487 ret = check_prlimit_permission(tsk);
1492 get_task_struct(tsk);
1495 ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1496 old_rlim ? &old : NULL);
1498 if (!ret && old_rlim) {
1499 rlim_to_rlim64(&old, &old64);
1500 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1504 put_task_struct(tsk);
1508 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1510 struct rlimit new_rlim;
1512 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1514 return do_prlimit(current, resource, &new_rlim, NULL);
1518 * It would make sense to put struct rusage in the task_struct,
1519 * except that would make the task_struct be *really big*. After
1520 * task_struct gets moved into malloc'ed memory, it would
1521 * make sense to do this. It will make moving the rest of the information
1522 * a lot simpler! (Which we're not doing right now because we're not
1523 * measuring them yet).
1525 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1526 * races with threads incrementing their own counters. But since word
1527 * reads are atomic, we either get new values or old values and we don't
1528 * care which for the sums. We always take the siglock to protect reading
1529 * the c* fields from p->signal from races with exit.c updating those
1530 * fields when reaping, so a sample either gets all the additions of a
1531 * given child after it's reaped, or none so this sample is before reaping.
1534 * We need to take the siglock for CHILDEREN, SELF and BOTH
1535 * for the cases current multithreaded, non-current single threaded
1536 * non-current multithreaded. Thread traversal is now safe with
1538 * Strictly speaking, we donot need to take the siglock if we are current and
1539 * single threaded, as no one else can take our signal_struct away, no one
1540 * else can reap the children to update signal->c* counters, and no one else
1541 * can race with the signal-> fields. If we do not take any lock, the
1542 * signal-> fields could be read out of order while another thread was just
1543 * exiting. So we should place a read memory barrier when we avoid the lock.
1544 * On the writer side, write memory barrier is implied in __exit_signal
1545 * as __exit_signal releases the siglock spinlock after updating the signal->
1546 * fields. But we don't do this yet to keep things simple.
1550 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1552 r->ru_nvcsw += t->nvcsw;
1553 r->ru_nivcsw += t->nivcsw;
1554 r->ru_minflt += t->min_flt;
1555 r->ru_majflt += t->maj_flt;
1556 r->ru_inblock += task_io_get_inblock(t);
1557 r->ru_oublock += task_io_get_oublock(t);
1560 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1562 struct task_struct *t;
1563 unsigned long flags;
1564 cputime_t tgutime, tgstime, utime, stime;
1565 unsigned long maxrss = 0;
1567 memset((char *) r, 0, sizeof *r);
1568 utime = stime = cputime_zero;
1570 if (who == RUSAGE_THREAD) {
1571 task_times(current, &utime, &stime);
1572 accumulate_thread_rusage(p, r);
1573 maxrss = p->signal->maxrss;
1577 if (!lock_task_sighand(p, &flags))
1582 case RUSAGE_CHILDREN:
1583 utime = p->signal->cutime;
1584 stime = p->signal->cstime;
1585 r->ru_nvcsw = p->signal->cnvcsw;
1586 r->ru_nivcsw = p->signal->cnivcsw;
1587 r->ru_minflt = p->signal->cmin_flt;
1588 r->ru_majflt = p->signal->cmaj_flt;
1589 r->ru_inblock = p->signal->cinblock;
1590 r->ru_oublock = p->signal->coublock;
1591 maxrss = p->signal->cmaxrss;
1593 if (who == RUSAGE_CHILDREN)
1597 thread_group_times(p, &tgutime, &tgstime);
1598 utime = cputime_add(utime, tgutime);
1599 stime = cputime_add(stime, tgstime);
1600 r->ru_nvcsw += p->signal->nvcsw;
1601 r->ru_nivcsw += p->signal->nivcsw;
1602 r->ru_minflt += p->signal->min_flt;
1603 r->ru_majflt += p->signal->maj_flt;
1604 r->ru_inblock += p->signal->inblock;
1605 r->ru_oublock += p->signal->oublock;
1606 if (maxrss < p->signal->maxrss)
1607 maxrss = p->signal->maxrss;
1610 accumulate_thread_rusage(t, r);
1618 unlock_task_sighand(p, &flags);
1621 cputime_to_timeval(utime, &r->ru_utime);
1622 cputime_to_timeval(stime, &r->ru_stime);
1624 if (who != RUSAGE_CHILDREN) {
1625 struct mm_struct *mm = get_task_mm(p);
1627 setmax_mm_hiwater_rss(&maxrss, mm);
1631 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1634 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1637 k_getrusage(p, who, &r);
1638 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1641 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1643 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1644 who != RUSAGE_THREAD)
1646 return getrusage(current, who, ru);
1649 SYSCALL_DEFINE1(umask, int, mask)
1651 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1655 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
1656 unsigned long, arg4, unsigned long, arg5)
1658 struct task_struct *me = current;
1659 unsigned char comm[sizeof(me->comm)];
1662 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1663 if (error != -ENOSYS)
1668 case PR_SET_PDEATHSIG:
1669 if (!valid_signal(arg2)) {
1673 me->pdeath_signal = arg2;
1676 case PR_GET_PDEATHSIG:
1677 error = put_user(me->pdeath_signal, (int __user *)arg2);
1679 case PR_GET_DUMPABLE:
1680 error = get_dumpable(me->mm);
1682 case PR_SET_DUMPABLE:
1683 if (arg2 < 0 || arg2 > 1) {
1687 set_dumpable(me->mm, arg2);
1691 case PR_SET_UNALIGN:
1692 error = SET_UNALIGN_CTL(me, arg2);
1694 case PR_GET_UNALIGN:
1695 error = GET_UNALIGN_CTL(me, arg2);
1698 error = SET_FPEMU_CTL(me, arg2);
1701 error = GET_FPEMU_CTL(me, arg2);
1704 error = SET_FPEXC_CTL(me, arg2);
1707 error = GET_FPEXC_CTL(me, arg2);
1710 error = PR_TIMING_STATISTICAL;
1713 if (arg2 != PR_TIMING_STATISTICAL)
1720 comm[sizeof(me->comm)-1] = 0;
1721 if (strncpy_from_user(comm, (char __user *)arg2,
1722 sizeof(me->comm) - 1) < 0)
1724 set_task_comm(me, comm);
1727 get_task_comm(comm, me);
1728 if (copy_to_user((char __user *)arg2, comm,
1733 error = GET_ENDIAN(me, arg2);
1736 error = SET_ENDIAN(me, arg2);
1739 case PR_GET_SECCOMP:
1740 error = prctl_get_seccomp();
1742 case PR_SET_SECCOMP:
1743 error = prctl_set_seccomp(arg2);
1746 error = GET_TSC_CTL(arg2);
1749 error = SET_TSC_CTL(arg2);
1751 case PR_TASK_PERF_EVENTS_DISABLE:
1752 error = perf_event_task_disable();
1754 case PR_TASK_PERF_EVENTS_ENABLE:
1755 error = perf_event_task_enable();
1757 case PR_GET_TIMERSLACK:
1758 error = current->timer_slack_ns;
1760 case PR_SET_TIMERSLACK:
1762 current->timer_slack_ns =
1763 current->default_timer_slack_ns;
1765 current->timer_slack_ns = arg2;
1772 case PR_MCE_KILL_CLEAR:
1775 current->flags &= ~PF_MCE_PROCESS;
1777 case PR_MCE_KILL_SET:
1778 current->flags |= PF_MCE_PROCESS;
1779 if (arg3 == PR_MCE_KILL_EARLY)
1780 current->flags |= PF_MCE_EARLY;
1781 else if (arg3 == PR_MCE_KILL_LATE)
1782 current->flags &= ~PF_MCE_EARLY;
1783 else if (arg3 == PR_MCE_KILL_DEFAULT)
1785 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
1794 case PR_MCE_KILL_GET:
1795 if (arg2 | arg3 | arg4 | arg5)
1797 if (current->flags & PF_MCE_PROCESS)
1798 error = (current->flags & PF_MCE_EARLY) ?
1799 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
1801 error = PR_MCE_KILL_DEFAULT;
1810 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
1811 struct getcpu_cache __user *, unused)
1814 int cpu = raw_smp_processor_id();
1816 err |= put_user(cpu, cpup);
1818 err |= put_user(cpu_to_node(cpu), nodep);
1819 return err ? -EFAULT : 0;
1822 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
1824 static void argv_cleanup(struct subprocess_info *info)
1826 argv_free(info->argv);
1830 * orderly_poweroff - Trigger an orderly system poweroff
1831 * @force: force poweroff if command execution fails
1833 * This may be called from any context to trigger a system shutdown.
1834 * If the orderly shutdown fails, it will force an immediate shutdown.
1836 int orderly_poweroff(bool force)
1839 char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
1840 static char *envp[] = {
1842 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
1846 struct subprocess_info *info;
1849 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
1850 __func__, poweroff_cmd);
1854 info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
1860 call_usermodehelper_setfns(info, NULL, argv_cleanup, NULL);
1862 ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
1866 printk(KERN_WARNING "Failed to start orderly shutdown: "
1867 "forcing the issue\n");
1869 /* I guess this should try to kick off some daemon to
1870 sync and poweroff asap. Or not even bother syncing
1871 if we're doing an emergency shutdown? */
1878 EXPORT_SYMBOL_GPL(orderly_poweroff);