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/ptrace.h>
37 #include <linux/fs_struct.h>
39 #include <linux/compat.h>
40 #include <linux/syscalls.h>
41 #include <linux/kprobes.h>
42 #include <linux/user_namespace.h>
44 #include <asm/uaccess.h>
46 #include <asm/unistd.h>
48 #ifndef SET_UNALIGN_CTL
49 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
51 #ifndef GET_UNALIGN_CTL
52 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
55 # define SET_FPEMU_CTL(a,b) (-EINVAL)
58 # define GET_FPEMU_CTL(a,b) (-EINVAL)
61 # define SET_FPEXC_CTL(a,b) (-EINVAL)
64 # define GET_FPEXC_CTL(a,b) (-EINVAL)
67 # define GET_ENDIAN(a,b) (-EINVAL)
70 # define SET_ENDIAN(a,b) (-EINVAL)
73 # define GET_TSC_CTL(a) (-EINVAL)
76 # define SET_TSC_CTL(a) (-EINVAL)
80 * this is where the system-wide overflow UID and GID are defined, for
81 * architectures that now have 32-bit UID/GID but didn't in the past
84 int overflowuid = DEFAULT_OVERFLOWUID;
85 int overflowgid = DEFAULT_OVERFLOWGID;
88 EXPORT_SYMBOL(overflowuid);
89 EXPORT_SYMBOL(overflowgid);
93 * the same as above, but for filesystems which can only store a 16-bit
94 * UID and GID. as such, this is needed on all architectures
97 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
98 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
100 EXPORT_SYMBOL(fs_overflowuid);
101 EXPORT_SYMBOL(fs_overflowgid);
104 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
109 EXPORT_SYMBOL(cad_pid);
112 * If set, this is used for preparing the system to power off.
115 void (*pm_power_off_prepare)(void);
118 * set the priority of a task
119 * - the caller must hold the RCU read lock
121 static int set_one_prio(struct task_struct *p, int niceval, int error)
123 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
126 if (pcred->uid != cred->euid &&
127 pcred->euid != cred->euid && !capable(CAP_SYS_NICE)) {
131 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
135 no_nice = security_task_setnice(p, niceval);
142 set_user_nice(p, niceval);
147 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
149 struct task_struct *g, *p;
150 struct user_struct *user;
151 const struct cred *cred = current_cred();
155 if (which > PRIO_USER || which < PRIO_PROCESS)
158 /* normalize: avoid signed division (rounding problems) */
165 read_lock(&tasklist_lock);
169 p = find_task_by_vpid(who);
173 error = set_one_prio(p, niceval, error);
177 pgrp = find_vpid(who);
179 pgrp = task_pgrp(current);
180 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
181 error = set_one_prio(p, niceval, error);
182 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
185 user = (struct user_struct *) cred->user;
188 else if ((who != cred->uid) &&
189 !(user = find_user(who)))
190 goto out_unlock; /* No processes for this user */
193 if (__task_cred(p)->uid == who)
194 error = set_one_prio(p, niceval, error);
195 while_each_thread(g, p);
196 if (who != cred->uid)
197 free_uid(user); /* For find_user() */
201 read_unlock(&tasklist_lock);
207 * Ugh. To avoid negative return values, "getpriority()" will
208 * not return the normal nice-value, but a negated value that
209 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
210 * to stay compatible.
212 SYSCALL_DEFINE2(getpriority, int, which, int, who)
214 struct task_struct *g, *p;
215 struct user_struct *user;
216 const struct cred *cred = current_cred();
217 long niceval, retval = -ESRCH;
220 if (which > PRIO_USER || which < PRIO_PROCESS)
223 read_lock(&tasklist_lock);
227 p = find_task_by_vpid(who);
231 niceval = 20 - task_nice(p);
232 if (niceval > retval)
238 pgrp = find_vpid(who);
240 pgrp = task_pgrp(current);
241 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
242 niceval = 20 - task_nice(p);
243 if (niceval > retval)
245 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
248 user = (struct user_struct *) cred->user;
251 else if ((who != cred->uid) &&
252 !(user = find_user(who)))
253 goto out_unlock; /* No processes for this user */
256 if (__task_cred(p)->uid == who) {
257 niceval = 20 - task_nice(p);
258 if (niceval > retval)
261 while_each_thread(g, p);
262 if (who != cred->uid)
263 free_uid(user); /* for find_user() */
267 read_unlock(&tasklist_lock);
273 * emergency_restart - reboot the system
275 * Without shutting down any hardware or taking any locks
276 * reboot the system. This is called when we know we are in
277 * trouble so this is our best effort to reboot. This is
278 * safe to call in interrupt context.
280 void emergency_restart(void)
282 machine_emergency_restart();
284 EXPORT_SYMBOL_GPL(emergency_restart);
286 void kernel_restart_prepare(char *cmd)
288 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
289 system_state = SYSTEM_RESTART;
295 * kernel_restart - reboot the system
296 * @cmd: pointer to buffer containing command to execute for restart
299 * Shutdown everything and perform a clean reboot.
300 * This is not safe to call in interrupt context.
302 void kernel_restart(char *cmd)
304 kernel_restart_prepare(cmd);
306 printk(KERN_EMERG "Restarting system.\n");
308 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
309 machine_restart(cmd);
311 EXPORT_SYMBOL_GPL(kernel_restart);
313 static void kernel_shutdown_prepare(enum system_states state)
315 blocking_notifier_call_chain(&reboot_notifier_list,
316 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
317 system_state = state;
321 * kernel_halt - halt the system
323 * Shutdown everything and perform a clean system halt.
325 void kernel_halt(void)
327 kernel_shutdown_prepare(SYSTEM_HALT);
329 printk(KERN_EMERG "System halted.\n");
333 EXPORT_SYMBOL_GPL(kernel_halt);
336 * kernel_power_off - power_off the system
338 * Shutdown everything and perform a clean system power_off.
340 void kernel_power_off(void)
342 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
343 if (pm_power_off_prepare)
344 pm_power_off_prepare();
345 disable_nonboot_cpus();
347 printk(KERN_EMERG "Power down.\n");
350 EXPORT_SYMBOL_GPL(kernel_power_off);
352 static DEFINE_MUTEX(reboot_mutex);
355 * Reboot system call: for obvious reasons only root may call it,
356 * and even root needs to set up some magic numbers in the registers
357 * so that some mistake won't make this reboot the whole machine.
358 * You can also set the meaning of the ctrl-alt-del-key here.
360 * reboot doesn't sync: do that yourself before calling this.
362 SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
368 /* We only trust the superuser with rebooting the system. */
369 if (!capable(CAP_SYS_BOOT))
372 /* For safety, we require "magic" arguments. */
373 if (magic1 != LINUX_REBOOT_MAGIC1 ||
374 (magic2 != LINUX_REBOOT_MAGIC2 &&
375 magic2 != LINUX_REBOOT_MAGIC2A &&
376 magic2 != LINUX_REBOOT_MAGIC2B &&
377 magic2 != LINUX_REBOOT_MAGIC2C))
380 /* Instead of trying to make the power_off code look like
381 * halt when pm_power_off is not set do it the easy way.
383 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
384 cmd = LINUX_REBOOT_CMD_HALT;
386 mutex_lock(&reboot_mutex);
388 case LINUX_REBOOT_CMD_RESTART:
389 kernel_restart(NULL);
392 case LINUX_REBOOT_CMD_CAD_ON:
396 case LINUX_REBOOT_CMD_CAD_OFF:
400 case LINUX_REBOOT_CMD_HALT:
403 panic("cannot halt");
405 case LINUX_REBOOT_CMD_POWER_OFF:
410 case LINUX_REBOOT_CMD_RESTART2:
411 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
415 buffer[sizeof(buffer) - 1] = '\0';
417 kernel_restart(buffer);
421 case LINUX_REBOOT_CMD_KEXEC:
422 ret = kernel_kexec();
426 #ifdef CONFIG_HIBERNATION
427 case LINUX_REBOOT_CMD_SW_SUSPEND:
436 mutex_unlock(&reboot_mutex);
440 static void deferred_cad(struct work_struct *dummy)
442 kernel_restart(NULL);
446 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
447 * As it's called within an interrupt, it may NOT sync: the only choice
448 * is whether to reboot at once, or just ignore the ctrl-alt-del.
450 void ctrl_alt_del(void)
452 static DECLARE_WORK(cad_work, deferred_cad);
455 schedule_work(&cad_work);
457 kill_cad_pid(SIGINT, 1);
461 * Unprivileged users may change the real gid to the effective gid
462 * or vice versa. (BSD-style)
464 * If you set the real gid at all, or set the effective gid to a value not
465 * equal to the real gid, then the saved gid is set to the new effective gid.
467 * This makes it possible for a setgid program to completely drop its
468 * privileges, which is often a useful assertion to make when you are doing
469 * a security audit over a program.
471 * The general idea is that a program which uses just setregid() will be
472 * 100% compatible with BSD. A program which uses just setgid() will be
473 * 100% compatible with POSIX with saved IDs.
475 * SMP: There are not races, the GIDs are checked only by filesystem
476 * operations (as far as semantic preservation is concerned).
478 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
480 const struct cred *old;
484 new = prepare_creds();
487 old = current_cred();
489 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
494 if (rgid != (gid_t) -1) {
495 if (old->gid == rgid ||
502 if (egid != (gid_t) -1) {
503 if (old->gid == egid ||
512 if (rgid != (gid_t) -1 ||
513 (egid != (gid_t) -1 && egid != old->gid))
514 new->sgid = new->egid;
515 new->fsgid = new->egid;
517 return commit_creds(new);
525 * setgid() is implemented like SysV w/ SAVED_IDS
527 * SMP: Same implicit races as above.
529 SYSCALL_DEFINE1(setgid, gid_t, gid)
531 const struct cred *old;
535 new = prepare_creds();
538 old = current_cred();
540 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
545 if (capable(CAP_SETGID))
546 new->gid = new->egid = new->sgid = new->fsgid = gid;
547 else if (gid == old->gid || gid == old->sgid)
548 new->egid = new->fsgid = gid;
552 return commit_creds(new);
560 * change the user struct in a credentials set to match the new UID
562 static int set_user(struct cred *new)
564 struct user_struct *new_user;
566 new_user = alloc_uid(current_user_ns(), new->uid);
570 if (!task_can_switch_user(new_user, current)) {
575 if (atomic_read(&new_user->processes) >=
576 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
577 new_user != INIT_USER) {
583 new->user = new_user;
588 * Unprivileged users may change the real uid to the effective uid
589 * or vice versa. (BSD-style)
591 * If you set the real uid at all, or set the effective uid to a value not
592 * equal to the real uid, then the saved uid is set to the new effective uid.
594 * This makes it possible for a setuid program to completely drop its
595 * privileges, which is often a useful assertion to make when you are doing
596 * a security audit over a program.
598 * The general idea is that a program which uses just setreuid() will be
599 * 100% compatible with BSD. A program which uses just setuid() will be
600 * 100% compatible with POSIX with saved IDs.
602 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
604 const struct cred *old;
608 new = prepare_creds();
611 old = current_cred();
613 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
618 if (ruid != (uid_t) -1) {
620 if (old->uid != ruid &&
622 !capable(CAP_SETUID))
626 if (euid != (uid_t) -1) {
628 if (old->uid != euid &&
631 !capable(CAP_SETUID))
635 if (new->uid != old->uid) {
636 retval = set_user(new);
640 if (ruid != (uid_t) -1 ||
641 (euid != (uid_t) -1 && euid != old->uid))
642 new->suid = new->euid;
643 new->fsuid = new->euid;
645 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
649 return commit_creds(new);
657 * setuid() is implemented like SysV with SAVED_IDS
659 * Note that SAVED_ID's is deficient in that a setuid root program
660 * like sendmail, for example, cannot set its uid to be a normal
661 * user and then switch back, because if you're root, setuid() sets
662 * the saved uid too. If you don't like this, blame the bright people
663 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
664 * will allow a root program to temporarily drop privileges and be able to
665 * regain them by swapping the real and effective uid.
667 SYSCALL_DEFINE1(setuid, uid_t, uid)
669 const struct cred *old;
673 new = prepare_creds();
676 old = current_cred();
678 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
683 if (capable(CAP_SETUID)) {
684 new->suid = new->uid = uid;
685 if (uid != old->uid) {
686 retval = set_user(new);
690 } else if (uid != old->uid && uid != new->suid) {
694 new->fsuid = new->euid = uid;
696 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
700 return commit_creds(new);
709 * This function implements a generic ability to update ruid, euid,
710 * and suid. This allows you to implement the 4.4 compatible seteuid().
712 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
714 const struct cred *old;
718 new = prepare_creds();
722 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
725 old = current_cred();
728 if (!capable(CAP_SETUID)) {
729 if (ruid != (uid_t) -1 && ruid != old->uid &&
730 ruid != old->euid && ruid != old->suid)
732 if (euid != (uid_t) -1 && euid != old->uid &&
733 euid != old->euid && euid != old->suid)
735 if (suid != (uid_t) -1 && suid != old->uid &&
736 suid != old->euid && suid != old->suid)
740 if (ruid != (uid_t) -1) {
742 if (ruid != old->uid) {
743 retval = set_user(new);
748 if (euid != (uid_t) -1)
750 if (suid != (uid_t) -1)
752 new->fsuid = new->euid;
754 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
758 return commit_creds(new);
765 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruid, uid_t __user *, euid, uid_t __user *, suid)
767 const struct cred *cred = current_cred();
770 if (!(retval = put_user(cred->uid, ruid)) &&
771 !(retval = put_user(cred->euid, euid)))
772 retval = put_user(cred->suid, suid);
778 * Same as above, but for rgid, egid, sgid.
780 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
782 const struct cred *old;
786 new = prepare_creds();
789 old = current_cred();
791 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
796 if (!capable(CAP_SETGID)) {
797 if (rgid != (gid_t) -1 && rgid != old->gid &&
798 rgid != old->egid && rgid != old->sgid)
800 if (egid != (gid_t) -1 && egid != old->gid &&
801 egid != old->egid && egid != old->sgid)
803 if (sgid != (gid_t) -1 && sgid != old->gid &&
804 sgid != old->egid && sgid != old->sgid)
808 if (rgid != (gid_t) -1)
810 if (egid != (gid_t) -1)
812 if (sgid != (gid_t) -1)
814 new->fsgid = new->egid;
816 return commit_creds(new);
823 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgid, gid_t __user *, egid, gid_t __user *, sgid)
825 const struct cred *cred = current_cred();
828 if (!(retval = put_user(cred->gid, rgid)) &&
829 !(retval = put_user(cred->egid, egid)))
830 retval = put_user(cred->sgid, sgid);
837 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
838 * is used for "access()" and for the NFS daemon (letting nfsd stay at
839 * whatever uid it wants to). It normally shadows "euid", except when
840 * explicitly set by setfsuid() or for access..
842 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
844 const struct cred *old;
848 new = prepare_creds();
850 return current_fsuid();
851 old = current_cred();
852 old_fsuid = old->fsuid;
854 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS) < 0)
857 if (uid == old->uid || uid == old->euid ||
858 uid == old->suid || uid == old->fsuid ||
859 capable(CAP_SETUID)) {
860 if (uid != old_fsuid) {
862 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
877 * Samma på svenska..
879 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
881 const struct cred *old;
885 new = prepare_creds();
887 return current_fsgid();
888 old = current_cred();
889 old_fsgid = old->fsgid;
891 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
894 if (gid == old->gid || gid == old->egid ||
895 gid == old->sgid || gid == old->fsgid ||
896 capable(CAP_SETGID)) {
897 if (gid != old_fsgid) {
912 void do_sys_times(struct tms *tms)
914 struct task_cputime cputime;
915 cputime_t cutime, cstime;
917 thread_group_cputime(current, &cputime);
918 spin_lock_irq(¤t->sighand->siglock);
919 cutime = current->signal->cutime;
920 cstime = current->signal->cstime;
921 spin_unlock_irq(¤t->sighand->siglock);
922 tms->tms_utime = cputime_to_clock_t(cputime.utime);
923 tms->tms_stime = cputime_to_clock_t(cputime.stime);
924 tms->tms_cutime = cputime_to_clock_t(cutime);
925 tms->tms_cstime = cputime_to_clock_t(cstime);
928 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
934 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
937 force_successful_syscall_return();
938 return (long) jiffies_64_to_clock_t(get_jiffies_64());
942 * This needs some heavy checking ...
943 * I just haven't the stomach for it. I also don't fully
944 * understand sessions/pgrp etc. Let somebody who does explain it.
946 * OK, I think I have the protection semantics right.... this is really
947 * only important on a multi-user system anyway, to make sure one user
948 * can't send a signal to a process owned by another. -TYT, 12/12/91
950 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
953 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
955 struct task_struct *p;
956 struct task_struct *group_leader = current->group_leader;
961 pid = task_pid_vnr(group_leader);
967 /* From this point forward we keep holding onto the tasklist lock
968 * so that our parent does not change from under us. -DaveM
970 write_lock_irq(&tasklist_lock);
973 p = find_task_by_vpid(pid);
978 if (!thread_group_leader(p))
981 if (same_thread_group(p->real_parent, group_leader)) {
983 if (task_session(p) != task_session(group_leader))
990 if (p != group_leader)
995 if (p->signal->leader)
1000 struct task_struct *g;
1002 pgrp = find_vpid(pgid);
1003 g = pid_task(pgrp, PIDTYPE_PGID);
1004 if (!g || task_session(g) != task_session(group_leader))
1008 err = security_task_setpgid(p, pgid);
1012 if (task_pgrp(p) != pgrp)
1013 change_pid(p, PIDTYPE_PGID, pgrp);
1017 /* All paths lead to here, thus we are safe. -DaveM */
1018 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);
1116 DECLARE_RWSEM(uts_sem);
1118 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1122 down_read(&uts_sem);
1123 if (copy_to_user(name, utsname(), sizeof *name))
1129 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1132 char tmp[__NEW_UTS_LEN];
1134 if (!capable(CAP_SYS_ADMIN))
1136 if (len < 0 || len > __NEW_UTS_LEN)
1138 down_write(&uts_sem);
1140 if (!copy_from_user(tmp, name, len)) {
1141 struct new_utsname *u = utsname();
1143 memcpy(u->nodename, tmp, len);
1144 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1151 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1153 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1156 struct new_utsname *u;
1160 down_read(&uts_sem);
1162 i = 1 + strlen(u->nodename);
1166 if (copy_to_user(name, u->nodename, i))
1175 * Only setdomainname; getdomainname can be implemented by calling
1178 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1181 char tmp[__NEW_UTS_LEN];
1183 if (!capable(CAP_SYS_ADMIN))
1185 if (len < 0 || len > __NEW_UTS_LEN)
1188 down_write(&uts_sem);
1190 if (!copy_from_user(tmp, name, len)) {
1191 struct new_utsname *u = utsname();
1193 memcpy(u->domainname, tmp, len);
1194 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1201 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1203 if (resource >= RLIM_NLIMITS)
1206 struct rlimit value;
1207 task_lock(current->group_leader);
1208 value = current->signal->rlim[resource];
1209 task_unlock(current->group_leader);
1210 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1214 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1217 * Back compatibility for getrlimit. Needed for some apps.
1220 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1221 struct rlimit __user *, rlim)
1224 if (resource >= RLIM_NLIMITS)
1227 task_lock(current->group_leader);
1228 x = current->signal->rlim[resource];
1229 task_unlock(current->group_leader);
1230 if (x.rlim_cur > 0x7FFFFFFF)
1231 x.rlim_cur = 0x7FFFFFFF;
1232 if (x.rlim_max > 0x7FFFFFFF)
1233 x.rlim_max = 0x7FFFFFFF;
1234 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1239 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1241 struct rlimit new_rlim, *old_rlim;
1244 if (resource >= RLIM_NLIMITS)
1246 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1248 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1250 old_rlim = current->signal->rlim + resource;
1251 if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1252 !capable(CAP_SYS_RESOURCE))
1254 if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > sysctl_nr_open)
1257 retval = security_task_setrlimit(resource, &new_rlim);
1261 if (resource == RLIMIT_CPU && new_rlim.rlim_cur == 0) {
1263 * The caller is asking for an immediate RLIMIT_CPU
1264 * expiry. But we use the zero value to mean "it was
1265 * never set". So let's cheat and make it one second
1268 new_rlim.rlim_cur = 1;
1271 task_lock(current->group_leader);
1272 *old_rlim = new_rlim;
1273 task_unlock(current->group_leader);
1275 if (resource != RLIMIT_CPU)
1279 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1280 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1281 * very long-standing error, and fixing it now risks breakage of
1282 * applications, so we live with it
1284 if (new_rlim.rlim_cur == RLIM_INFINITY)
1287 update_rlimit_cpu(new_rlim.rlim_cur);
1293 * It would make sense to put struct rusage in the task_struct,
1294 * except that would make the task_struct be *really big*. After
1295 * task_struct gets moved into malloc'ed memory, it would
1296 * make sense to do this. It will make moving the rest of the information
1297 * a lot simpler! (Which we're not doing right now because we're not
1298 * measuring them yet).
1300 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1301 * races with threads incrementing their own counters. But since word
1302 * reads are atomic, we either get new values or old values and we don't
1303 * care which for the sums. We always take the siglock to protect reading
1304 * the c* fields from p->signal from races with exit.c updating those
1305 * fields when reaping, so a sample either gets all the additions of a
1306 * given child after it's reaped, or none so this sample is before reaping.
1309 * We need to take the siglock for CHILDEREN, SELF and BOTH
1310 * for the cases current multithreaded, non-current single threaded
1311 * non-current multithreaded. Thread traversal is now safe with
1313 * Strictly speaking, we donot need to take the siglock if we are current and
1314 * single threaded, as no one else can take our signal_struct away, no one
1315 * else can reap the children to update signal->c* counters, and no one else
1316 * can race with the signal-> fields. If we do not take any lock, the
1317 * signal-> fields could be read out of order while another thread was just
1318 * exiting. So we should place a read memory barrier when we avoid the lock.
1319 * On the writer side, write memory barrier is implied in __exit_signal
1320 * as __exit_signal releases the siglock spinlock after updating the signal->
1321 * fields. But we don't do this yet to keep things simple.
1325 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1327 r->ru_nvcsw += t->nvcsw;
1328 r->ru_nivcsw += t->nivcsw;
1329 r->ru_minflt += t->min_flt;
1330 r->ru_majflt += t->maj_flt;
1331 r->ru_inblock += task_io_get_inblock(t);
1332 r->ru_oublock += task_io_get_oublock(t);
1335 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1337 struct task_struct *t;
1338 unsigned long flags;
1339 cputime_t utime, stime;
1340 struct task_cputime cputime;
1341 unsigned long maxrss = 0;
1343 memset((char *) r, 0, sizeof *r);
1344 utime = stime = cputime_zero;
1346 if (who == RUSAGE_THREAD) {
1347 utime = task_utime(current);
1348 stime = task_stime(current);
1349 accumulate_thread_rusage(p, r);
1350 maxrss = p->signal->maxrss;
1354 if (!lock_task_sighand(p, &flags))
1359 case RUSAGE_CHILDREN:
1360 utime = p->signal->cutime;
1361 stime = p->signal->cstime;
1362 r->ru_nvcsw = p->signal->cnvcsw;
1363 r->ru_nivcsw = p->signal->cnivcsw;
1364 r->ru_minflt = p->signal->cmin_flt;
1365 r->ru_majflt = p->signal->cmaj_flt;
1366 r->ru_inblock = p->signal->cinblock;
1367 r->ru_oublock = p->signal->coublock;
1368 maxrss = p->signal->cmaxrss;
1370 if (who == RUSAGE_CHILDREN)
1374 thread_group_cputime(p, &cputime);
1375 utime = cputime_add(utime, cputime.utime);
1376 stime = cputime_add(stime, cputime.stime);
1377 r->ru_nvcsw += p->signal->nvcsw;
1378 r->ru_nivcsw += p->signal->nivcsw;
1379 r->ru_minflt += p->signal->min_flt;
1380 r->ru_majflt += p->signal->maj_flt;
1381 r->ru_inblock += p->signal->inblock;
1382 r->ru_oublock += p->signal->oublock;
1383 if (maxrss < p->signal->maxrss)
1384 maxrss = p->signal->maxrss;
1387 accumulate_thread_rusage(t, r);
1395 unlock_task_sighand(p, &flags);
1398 cputime_to_timeval(utime, &r->ru_utime);
1399 cputime_to_timeval(stime, &r->ru_stime);
1401 if (who != RUSAGE_CHILDREN) {
1402 struct mm_struct *mm = get_task_mm(p);
1404 setmax_mm_hiwater_rss(&maxrss, mm);
1408 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1411 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1414 k_getrusage(p, who, &r);
1415 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1418 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1420 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1421 who != RUSAGE_THREAD)
1423 return getrusage(current, who, ru);
1426 SYSCALL_DEFINE1(umask, int, mask)
1428 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1432 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
1433 unsigned long, arg4, unsigned long, arg5)
1435 struct task_struct *me = current;
1436 unsigned char comm[sizeof(me->comm)];
1439 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1440 if (error != -ENOSYS)
1445 case PR_SET_PDEATHSIG:
1446 if (!valid_signal(arg2)) {
1450 me->pdeath_signal = arg2;
1453 case PR_GET_PDEATHSIG:
1454 error = put_user(me->pdeath_signal, (int __user *)arg2);
1456 case PR_GET_DUMPABLE:
1457 error = get_dumpable(me->mm);
1459 case PR_SET_DUMPABLE:
1460 if (arg2 < 0 || arg2 > 1) {
1464 set_dumpable(me->mm, arg2);
1468 case PR_SET_UNALIGN:
1469 error = SET_UNALIGN_CTL(me, arg2);
1471 case PR_GET_UNALIGN:
1472 error = GET_UNALIGN_CTL(me, arg2);
1475 error = SET_FPEMU_CTL(me, arg2);
1478 error = GET_FPEMU_CTL(me, arg2);
1481 error = SET_FPEXC_CTL(me, arg2);
1484 error = GET_FPEXC_CTL(me, arg2);
1487 error = PR_TIMING_STATISTICAL;
1490 if (arg2 != PR_TIMING_STATISTICAL)
1497 comm[sizeof(me->comm)-1] = 0;
1498 if (strncpy_from_user(comm, (char __user *)arg2,
1499 sizeof(me->comm) - 1) < 0)
1501 set_task_comm(me, comm);
1504 get_task_comm(comm, me);
1505 if (copy_to_user((char __user *)arg2, comm,
1510 error = GET_ENDIAN(me, arg2);
1513 error = SET_ENDIAN(me, arg2);
1516 case PR_GET_SECCOMP:
1517 error = prctl_get_seccomp();
1519 case PR_SET_SECCOMP:
1520 error = prctl_set_seccomp(arg2);
1523 error = GET_TSC_CTL(arg2);
1526 error = SET_TSC_CTL(arg2);
1528 case PR_TASK_PERF_EVENTS_DISABLE:
1529 error = perf_event_task_disable();
1531 case PR_TASK_PERF_EVENTS_ENABLE:
1532 error = perf_event_task_enable();
1534 case PR_GET_TIMERSLACK:
1535 error = current->timer_slack_ns;
1537 case PR_SET_TIMERSLACK:
1539 current->timer_slack_ns =
1540 current->default_timer_slack_ns;
1542 current->timer_slack_ns = arg2;
1552 current->flags &= ~PF_MCE_PROCESS;
1555 current->flags |= PF_MCE_PROCESS;
1557 current->flags |= PF_MCE_EARLY;
1559 current->flags &= ~PF_MCE_EARLY;
1574 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
1575 struct getcpu_cache __user *, unused)
1578 int cpu = raw_smp_processor_id();
1580 err |= put_user(cpu, cpup);
1582 err |= put_user(cpu_to_node(cpu), nodep);
1583 return err ? -EFAULT : 0;
1586 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
1588 static void argv_cleanup(char **argv, char **envp)
1594 * orderly_poweroff - Trigger an orderly system poweroff
1595 * @force: force poweroff if command execution fails
1597 * This may be called from any context to trigger a system shutdown.
1598 * If the orderly shutdown fails, it will force an immediate shutdown.
1600 int orderly_poweroff(bool force)
1603 char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
1604 static char *envp[] = {
1606 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
1610 struct subprocess_info *info;
1613 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
1614 __func__, poweroff_cmd);
1618 info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
1624 call_usermodehelper_setcleanup(info, argv_cleanup);
1626 ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
1630 printk(KERN_WARNING "Failed to start orderly shutdown: "
1631 "forcing the issue\n");
1633 /* I guess this should try to kick off some daemon to
1634 sync and poweroff asap. Or not even bother syncing
1635 if we're doing an emergency shutdown? */
1642 EXPORT_SYMBOL_GPL(orderly_poweroff);