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) */
166 read_lock(&tasklist_lock);
170 p = find_task_by_vpid(who);
174 error = set_one_prio(p, niceval, error);
178 pgrp = find_vpid(who);
180 pgrp = task_pgrp(current);
181 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
182 error = set_one_prio(p, niceval, error);
183 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
186 user = (struct user_struct *) cred->user;
189 else if ((who != cred->uid) &&
190 !(user = find_user(who)))
191 goto out_unlock; /* No processes for this user */
193 do_each_thread(g, p) {
194 if (__task_cred(p)->uid == who)
195 error = set_one_prio(p, niceval, error);
196 } while_each_thread(g, p);
197 if (who != cred->uid)
198 free_uid(user); /* For find_user() */
202 read_unlock(&tasklist_lock);
209 * Ugh. To avoid negative return values, "getpriority()" will
210 * not return the normal nice-value, but a negated value that
211 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
212 * to stay compatible.
214 SYSCALL_DEFINE2(getpriority, int, which, int, who)
216 struct task_struct *g, *p;
217 struct user_struct *user;
218 const struct cred *cred = current_cred();
219 long niceval, retval = -ESRCH;
222 if (which > PRIO_USER || which < PRIO_PROCESS)
225 read_lock(&tasklist_lock);
229 p = find_task_by_vpid(who);
233 niceval = 20 - task_nice(p);
234 if (niceval > retval)
240 pgrp = find_vpid(who);
242 pgrp = task_pgrp(current);
243 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
244 niceval = 20 - task_nice(p);
245 if (niceval > retval)
247 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
250 user = (struct user_struct *) cred->user;
253 else if ((who != cred->uid) &&
254 !(user = find_user(who)))
255 goto out_unlock; /* No processes for this user */
257 do_each_thread(g, p) {
258 if (__task_cred(p)->uid == who) {
259 niceval = 20 - task_nice(p);
260 if (niceval > retval)
263 } while_each_thread(g, p);
264 if (who != cred->uid)
265 free_uid(user); /* for find_user() */
269 read_unlock(&tasklist_lock);
275 * emergency_restart - reboot the system
277 * Without shutting down any hardware or taking any locks
278 * reboot the system. This is called when we know we are in
279 * trouble so this is our best effort to reboot. This is
280 * safe to call in interrupt context.
282 void emergency_restart(void)
284 machine_emergency_restart();
286 EXPORT_SYMBOL_GPL(emergency_restart);
288 void kernel_restart_prepare(char *cmd)
290 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
291 system_state = SYSTEM_RESTART;
297 * kernel_restart - reboot the system
298 * @cmd: pointer to buffer containing command to execute for restart
301 * Shutdown everything and perform a clean reboot.
302 * This is not safe to call in interrupt context.
304 void kernel_restart(char *cmd)
306 kernel_restart_prepare(cmd);
308 printk(KERN_EMERG "Restarting system.\n");
310 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
311 machine_restart(cmd);
313 EXPORT_SYMBOL_GPL(kernel_restart);
315 static void kernel_shutdown_prepare(enum system_states state)
317 blocking_notifier_call_chain(&reboot_notifier_list,
318 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
319 system_state = state;
323 * kernel_halt - halt the system
325 * Shutdown everything and perform a clean system halt.
327 void kernel_halt(void)
329 kernel_shutdown_prepare(SYSTEM_HALT);
331 printk(KERN_EMERG "System halted.\n");
335 EXPORT_SYMBOL_GPL(kernel_halt);
338 * kernel_power_off - power_off the system
340 * Shutdown everything and perform a clean system power_off.
342 void kernel_power_off(void)
344 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
345 if (pm_power_off_prepare)
346 pm_power_off_prepare();
347 disable_nonboot_cpus();
349 printk(KERN_EMERG "Power down.\n");
352 EXPORT_SYMBOL_GPL(kernel_power_off);
354 static DEFINE_MUTEX(reboot_mutex);
357 * Reboot system call: for obvious reasons only root may call it,
358 * and even root needs to set up some magic numbers in the registers
359 * so that some mistake won't make this reboot the whole machine.
360 * You can also set the meaning of the ctrl-alt-del-key here.
362 * reboot doesn't sync: do that yourself before calling this.
364 SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
370 /* We only trust the superuser with rebooting the system. */
371 if (!capable(CAP_SYS_BOOT))
374 /* For safety, we require "magic" arguments. */
375 if (magic1 != LINUX_REBOOT_MAGIC1 ||
376 (magic2 != LINUX_REBOOT_MAGIC2 &&
377 magic2 != LINUX_REBOOT_MAGIC2A &&
378 magic2 != LINUX_REBOOT_MAGIC2B &&
379 magic2 != LINUX_REBOOT_MAGIC2C))
382 /* Instead of trying to make the power_off code look like
383 * halt when pm_power_off is not set do it the easy way.
385 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
386 cmd = LINUX_REBOOT_CMD_HALT;
388 mutex_lock(&reboot_mutex);
390 case LINUX_REBOOT_CMD_RESTART:
391 kernel_restart(NULL);
394 case LINUX_REBOOT_CMD_CAD_ON:
398 case LINUX_REBOOT_CMD_CAD_OFF:
402 case LINUX_REBOOT_CMD_HALT:
405 panic("cannot halt");
407 case LINUX_REBOOT_CMD_POWER_OFF:
412 case LINUX_REBOOT_CMD_RESTART2:
413 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
417 buffer[sizeof(buffer) - 1] = '\0';
419 kernel_restart(buffer);
423 case LINUX_REBOOT_CMD_KEXEC:
424 ret = kernel_kexec();
428 #ifdef CONFIG_HIBERNATION
429 case LINUX_REBOOT_CMD_SW_SUSPEND:
438 mutex_unlock(&reboot_mutex);
442 static void deferred_cad(struct work_struct *dummy)
444 kernel_restart(NULL);
448 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
449 * As it's called within an interrupt, it may NOT sync: the only choice
450 * is whether to reboot at once, or just ignore the ctrl-alt-del.
452 void ctrl_alt_del(void)
454 static DECLARE_WORK(cad_work, deferred_cad);
457 schedule_work(&cad_work);
459 kill_cad_pid(SIGINT, 1);
463 * Unprivileged users may change the real gid to the effective gid
464 * or vice versa. (BSD-style)
466 * If you set the real gid at all, or set the effective gid to a value not
467 * equal to the real gid, then the saved gid is set to the new effective gid.
469 * This makes it possible for a setgid program to completely drop its
470 * privileges, which is often a useful assertion to make when you are doing
471 * a security audit over a program.
473 * The general idea is that a program which uses just setregid() will be
474 * 100% compatible with BSD. A program which uses just setgid() will be
475 * 100% compatible with POSIX with saved IDs.
477 * SMP: There are not races, the GIDs are checked only by filesystem
478 * operations (as far as semantic preservation is concerned).
480 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
482 const struct cred *old;
486 new = prepare_creds();
489 old = current_cred();
491 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
496 if (rgid != (gid_t) -1) {
497 if (old->gid == rgid ||
504 if (egid != (gid_t) -1) {
505 if (old->gid == egid ||
514 if (rgid != (gid_t) -1 ||
515 (egid != (gid_t) -1 && egid != old->gid))
516 new->sgid = new->egid;
517 new->fsgid = new->egid;
519 return commit_creds(new);
527 * setgid() is implemented like SysV w/ SAVED_IDS
529 * SMP: Same implicit races as above.
531 SYSCALL_DEFINE1(setgid, gid_t, gid)
533 const struct cred *old;
537 new = prepare_creds();
540 old = current_cred();
542 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
547 if (capable(CAP_SETGID))
548 new->gid = new->egid = new->sgid = new->fsgid = gid;
549 else if (gid == old->gid || gid == old->sgid)
550 new->egid = new->fsgid = gid;
554 return commit_creds(new);
562 * change the user struct in a credentials set to match the new UID
564 static int set_user(struct cred *new)
566 struct user_struct *new_user;
568 new_user = alloc_uid(current_user_ns(), new->uid);
572 if (atomic_read(&new_user->processes) >=
573 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
574 new_user != INIT_USER) {
580 new->user = new_user;
585 * Unprivileged users may change the real uid to the effective uid
586 * or vice versa. (BSD-style)
588 * If you set the real uid at all, or set the effective uid to a value not
589 * equal to the real uid, then the saved uid is set to the new effective uid.
591 * This makes it possible for a setuid program to completely drop its
592 * privileges, which is often a useful assertion to make when you are doing
593 * a security audit over a program.
595 * The general idea is that a program which uses just setreuid() will be
596 * 100% compatible with BSD. A program which uses just setuid() will be
597 * 100% compatible with POSIX with saved IDs.
599 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
601 const struct cred *old;
605 new = prepare_creds();
608 old = current_cred();
610 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
615 if (ruid != (uid_t) -1) {
617 if (old->uid != ruid &&
619 !capable(CAP_SETUID))
623 if (euid != (uid_t) -1) {
625 if (old->uid != euid &&
628 !capable(CAP_SETUID))
632 if (new->uid != old->uid) {
633 retval = set_user(new);
637 if (ruid != (uid_t) -1 ||
638 (euid != (uid_t) -1 && euid != old->uid))
639 new->suid = new->euid;
640 new->fsuid = new->euid;
642 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
646 return commit_creds(new);
654 * setuid() is implemented like SysV with SAVED_IDS
656 * Note that SAVED_ID's is deficient in that a setuid root program
657 * like sendmail, for example, cannot set its uid to be a normal
658 * user and then switch back, because if you're root, setuid() sets
659 * the saved uid too. If you don't like this, blame the bright people
660 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
661 * will allow a root program to temporarily drop privileges and be able to
662 * regain them by swapping the real and effective uid.
664 SYSCALL_DEFINE1(setuid, uid_t, uid)
666 const struct cred *old;
670 new = prepare_creds();
673 old = current_cred();
675 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
680 if (capable(CAP_SETUID)) {
681 new->suid = new->uid = uid;
682 if (uid != old->uid) {
683 retval = set_user(new);
687 } else if (uid != old->uid && uid != new->suid) {
691 new->fsuid = new->euid = uid;
693 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
697 return commit_creds(new);
706 * This function implements a generic ability to update ruid, euid,
707 * and suid. This allows you to implement the 4.4 compatible seteuid().
709 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
711 const struct cred *old;
715 new = prepare_creds();
719 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
722 old = current_cred();
725 if (!capable(CAP_SETUID)) {
726 if (ruid != (uid_t) -1 && ruid != old->uid &&
727 ruid != old->euid && ruid != old->suid)
729 if (euid != (uid_t) -1 && euid != old->uid &&
730 euid != old->euid && euid != old->suid)
732 if (suid != (uid_t) -1 && suid != old->uid &&
733 suid != old->euid && suid != old->suid)
737 if (ruid != (uid_t) -1) {
739 if (ruid != old->uid) {
740 retval = set_user(new);
745 if (euid != (uid_t) -1)
747 if (suid != (uid_t) -1)
749 new->fsuid = new->euid;
751 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
755 return commit_creds(new);
762 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruid, uid_t __user *, euid, uid_t __user *, suid)
764 const struct cred *cred = current_cred();
767 if (!(retval = put_user(cred->uid, ruid)) &&
768 !(retval = put_user(cred->euid, euid)))
769 retval = put_user(cred->suid, suid);
775 * Same as above, but for rgid, egid, sgid.
777 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
779 const struct cred *old;
783 new = prepare_creds();
786 old = current_cred();
788 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
793 if (!capable(CAP_SETGID)) {
794 if (rgid != (gid_t) -1 && rgid != old->gid &&
795 rgid != old->egid && rgid != old->sgid)
797 if (egid != (gid_t) -1 && egid != old->gid &&
798 egid != old->egid && egid != old->sgid)
800 if (sgid != (gid_t) -1 && sgid != old->gid &&
801 sgid != old->egid && sgid != old->sgid)
805 if (rgid != (gid_t) -1)
807 if (egid != (gid_t) -1)
809 if (sgid != (gid_t) -1)
811 new->fsgid = new->egid;
813 return commit_creds(new);
820 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgid, gid_t __user *, egid, gid_t __user *, sgid)
822 const struct cred *cred = current_cred();
825 if (!(retval = put_user(cred->gid, rgid)) &&
826 !(retval = put_user(cred->egid, egid)))
827 retval = put_user(cred->sgid, sgid);
834 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
835 * is used for "access()" and for the NFS daemon (letting nfsd stay at
836 * whatever uid it wants to). It normally shadows "euid", except when
837 * explicitly set by setfsuid() or for access..
839 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
841 const struct cred *old;
845 new = prepare_creds();
847 return current_fsuid();
848 old = current_cred();
849 old_fsuid = old->fsuid;
851 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS) < 0)
854 if (uid == old->uid || uid == old->euid ||
855 uid == old->suid || uid == old->fsuid ||
856 capable(CAP_SETUID)) {
857 if (uid != old_fsuid) {
859 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 (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
891 if (gid == old->gid || gid == old->egid ||
892 gid == old->sgid || gid == old->fsgid ||
893 capable(CAP_SETGID)) {
894 if (gid != old_fsgid) {
909 void do_sys_times(struct tms *tms)
911 cputime_t tgutime, tgstime, cutime, cstime;
913 spin_lock_irq(¤t->sighand->siglock);
914 thread_group_times(current, &tgutime, &tgstime);
915 cutime = current->signal->cutime;
916 cstime = current->signal->cstime;
917 spin_unlock_irq(¤t->sighand->siglock);
918 tms->tms_utime = cputime_to_clock_t(tgutime);
919 tms->tms_stime = cputime_to_clock_t(tgstime);
920 tms->tms_cutime = cputime_to_clock_t(cutime);
921 tms->tms_cstime = cputime_to_clock_t(cstime);
924 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
930 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
933 force_successful_syscall_return();
934 return (long) jiffies_64_to_clock_t(get_jiffies_64());
938 * This needs some heavy checking ...
939 * I just haven't the stomach for it. I also don't fully
940 * understand sessions/pgrp etc. Let somebody who does explain it.
942 * OK, I think I have the protection semantics right.... this is really
943 * only important on a multi-user system anyway, to make sure one user
944 * can't send a signal to a process owned by another. -TYT, 12/12/91
946 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
949 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
951 struct task_struct *p;
952 struct task_struct *group_leader = current->group_leader;
957 pid = task_pid_vnr(group_leader);
963 /* From this point forward we keep holding onto the tasklist lock
964 * so that our parent does not change from under us. -DaveM
966 write_lock_irq(&tasklist_lock);
969 p = find_task_by_vpid(pid);
974 if (!thread_group_leader(p))
977 if (same_thread_group(p->real_parent, group_leader)) {
979 if (task_session(p) != task_session(group_leader))
986 if (p != group_leader)
991 if (p->signal->leader)
996 struct task_struct *g;
998 pgrp = find_vpid(pgid);
999 g = pid_task(pgrp, PIDTYPE_PGID);
1000 if (!g || task_session(g) != task_session(group_leader))
1004 err = security_task_setpgid(p, pgid);
1008 if (task_pgrp(p) != pgrp)
1009 change_pid(p, PIDTYPE_PGID, pgrp);
1013 /* All paths lead to here, thus we are safe. -DaveM */
1014 write_unlock_irq(&tasklist_lock);
1018 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1020 struct task_struct *p;
1026 grp = task_pgrp(current);
1029 p = find_task_by_vpid(pid);
1036 retval = security_task_getpgid(p);
1040 retval = pid_vnr(grp);
1046 #ifdef __ARCH_WANT_SYS_GETPGRP
1048 SYSCALL_DEFINE0(getpgrp)
1050 return sys_getpgid(0);
1055 SYSCALL_DEFINE1(getsid, pid_t, pid)
1057 struct task_struct *p;
1063 sid = task_session(current);
1066 p = find_task_by_vpid(pid);
1069 sid = task_session(p);
1073 retval = security_task_getsid(p);
1077 retval = pid_vnr(sid);
1083 SYSCALL_DEFINE0(setsid)
1085 struct task_struct *group_leader = current->group_leader;
1086 struct pid *sid = task_pid(group_leader);
1087 pid_t session = pid_vnr(sid);
1090 write_lock_irq(&tasklist_lock);
1091 /* Fail if I am already a session leader */
1092 if (group_leader->signal->leader)
1095 /* Fail if a process group id already exists that equals the
1096 * proposed session id.
1098 if (pid_task(sid, PIDTYPE_PGID))
1101 group_leader->signal->leader = 1;
1102 __set_special_pids(sid);
1104 proc_clear_tty(group_leader);
1108 write_unlock_irq(&tasklist_lock);
1110 proc_sid_connector(group_leader);
1114 DECLARE_RWSEM(uts_sem);
1116 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1120 down_read(&uts_sem);
1121 if (copy_to_user(name, utsname(), sizeof *name))
1127 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1130 char tmp[__NEW_UTS_LEN];
1132 if (!capable(CAP_SYS_ADMIN))
1134 if (len < 0 || len > __NEW_UTS_LEN)
1136 down_write(&uts_sem);
1138 if (!copy_from_user(tmp, name, len)) {
1139 struct new_utsname *u = utsname();
1141 memcpy(u->nodename, tmp, len);
1142 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1149 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1151 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1154 struct new_utsname *u;
1158 down_read(&uts_sem);
1160 i = 1 + strlen(u->nodename);
1164 if (copy_to_user(name, u->nodename, i))
1173 * Only setdomainname; getdomainname can be implemented by calling
1176 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1179 char tmp[__NEW_UTS_LEN];
1181 if (!capable(CAP_SYS_ADMIN))
1183 if (len < 0 || len > __NEW_UTS_LEN)
1186 down_write(&uts_sem);
1188 if (!copy_from_user(tmp, name, len)) {
1189 struct new_utsname *u = utsname();
1191 memcpy(u->domainname, tmp, len);
1192 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1199 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1201 if (resource >= RLIM_NLIMITS)
1204 struct rlimit value;
1205 task_lock(current->group_leader);
1206 value = current->signal->rlim[resource];
1207 task_unlock(current->group_leader);
1208 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1212 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1215 * Back compatibility for getrlimit. Needed for some apps.
1218 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1219 struct rlimit __user *, rlim)
1222 if (resource >= RLIM_NLIMITS)
1225 task_lock(current->group_leader);
1226 x = current->signal->rlim[resource];
1227 task_unlock(current->group_leader);
1228 if (x.rlim_cur > 0x7FFFFFFF)
1229 x.rlim_cur = 0x7FFFFFFF;
1230 if (x.rlim_max > 0x7FFFFFFF)
1231 x.rlim_max = 0x7FFFFFFF;
1232 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1237 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1239 struct rlimit new_rlim, *old_rlim;
1242 if (resource >= RLIM_NLIMITS)
1244 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1246 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1248 old_rlim = current->signal->rlim + resource;
1249 if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1250 !capable(CAP_SYS_RESOURCE))
1252 if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > sysctl_nr_open)
1255 retval = security_task_setrlimit(resource, &new_rlim);
1259 if (resource == RLIMIT_CPU && new_rlim.rlim_cur == 0) {
1261 * The caller is asking for an immediate RLIMIT_CPU
1262 * expiry. But we use the zero value to mean "it was
1263 * never set". So let's cheat and make it one second
1266 new_rlim.rlim_cur = 1;
1269 task_lock(current->group_leader);
1270 *old_rlim = new_rlim;
1271 task_unlock(current->group_leader);
1273 if (resource != RLIMIT_CPU)
1277 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1278 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1279 * very long-standing error, and fixing it now risks breakage of
1280 * applications, so we live with it
1282 if (new_rlim.rlim_cur == RLIM_INFINITY)
1285 update_rlimit_cpu(new_rlim.rlim_cur);
1291 * It would make sense to put struct rusage in the task_struct,
1292 * except that would make the task_struct be *really big*. After
1293 * task_struct gets moved into malloc'ed memory, it would
1294 * make sense to do this. It will make moving the rest of the information
1295 * a lot simpler! (Which we're not doing right now because we're not
1296 * measuring them yet).
1298 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1299 * races with threads incrementing their own counters. But since word
1300 * reads are atomic, we either get new values or old values and we don't
1301 * care which for the sums. We always take the siglock to protect reading
1302 * the c* fields from p->signal from races with exit.c updating those
1303 * fields when reaping, so a sample either gets all the additions of a
1304 * given child after it's reaped, or none so this sample is before reaping.
1307 * We need to take the siglock for CHILDEREN, SELF and BOTH
1308 * for the cases current multithreaded, non-current single threaded
1309 * non-current multithreaded. Thread traversal is now safe with
1311 * Strictly speaking, we donot need to take the siglock if we are current and
1312 * single threaded, as no one else can take our signal_struct away, no one
1313 * else can reap the children to update signal->c* counters, and no one else
1314 * can race with the signal-> fields. If we do not take any lock, the
1315 * signal-> fields could be read out of order while another thread was just
1316 * exiting. So we should place a read memory barrier when we avoid the lock.
1317 * On the writer side, write memory barrier is implied in __exit_signal
1318 * as __exit_signal releases the siglock spinlock after updating the signal->
1319 * fields. But we don't do this yet to keep things simple.
1323 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1325 r->ru_nvcsw += t->nvcsw;
1326 r->ru_nivcsw += t->nivcsw;
1327 r->ru_minflt += t->min_flt;
1328 r->ru_majflt += t->maj_flt;
1329 r->ru_inblock += task_io_get_inblock(t);
1330 r->ru_oublock += task_io_get_oublock(t);
1333 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1335 struct task_struct *t;
1336 unsigned long flags;
1337 cputime_t tgutime, tgstime, utime, stime;
1338 unsigned long maxrss = 0;
1340 memset((char *) r, 0, sizeof *r);
1341 utime = stime = cputime_zero;
1343 if (who == RUSAGE_THREAD) {
1344 task_times(current, &utime, &stime);
1345 accumulate_thread_rusage(p, r);
1346 maxrss = p->signal->maxrss;
1350 if (!lock_task_sighand(p, &flags))
1355 case RUSAGE_CHILDREN:
1356 utime = p->signal->cutime;
1357 stime = p->signal->cstime;
1358 r->ru_nvcsw = p->signal->cnvcsw;
1359 r->ru_nivcsw = p->signal->cnivcsw;
1360 r->ru_minflt = p->signal->cmin_flt;
1361 r->ru_majflt = p->signal->cmaj_flt;
1362 r->ru_inblock = p->signal->cinblock;
1363 r->ru_oublock = p->signal->coublock;
1364 maxrss = p->signal->cmaxrss;
1366 if (who == RUSAGE_CHILDREN)
1370 thread_group_times(p, &tgutime, &tgstime);
1371 utime = cputime_add(utime, tgutime);
1372 stime = cputime_add(stime, tgstime);
1373 r->ru_nvcsw += p->signal->nvcsw;
1374 r->ru_nivcsw += p->signal->nivcsw;
1375 r->ru_minflt += p->signal->min_flt;
1376 r->ru_majflt += p->signal->maj_flt;
1377 r->ru_inblock += p->signal->inblock;
1378 r->ru_oublock += p->signal->oublock;
1379 if (maxrss < p->signal->maxrss)
1380 maxrss = p->signal->maxrss;
1383 accumulate_thread_rusage(t, r);
1391 unlock_task_sighand(p, &flags);
1394 cputime_to_timeval(utime, &r->ru_utime);
1395 cputime_to_timeval(stime, &r->ru_stime);
1397 if (who != RUSAGE_CHILDREN) {
1398 struct mm_struct *mm = get_task_mm(p);
1400 setmax_mm_hiwater_rss(&maxrss, mm);
1404 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1407 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1410 k_getrusage(p, who, &r);
1411 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1414 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1416 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1417 who != RUSAGE_THREAD)
1419 return getrusage(current, who, ru);
1422 SYSCALL_DEFINE1(umask, int, mask)
1424 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1428 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
1429 unsigned long, arg4, unsigned long, arg5)
1431 struct task_struct *me = current;
1432 unsigned char comm[sizeof(me->comm)];
1435 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1436 if (error != -ENOSYS)
1441 case PR_SET_PDEATHSIG:
1442 if (!valid_signal(arg2)) {
1446 me->pdeath_signal = arg2;
1449 case PR_GET_PDEATHSIG:
1450 error = put_user(me->pdeath_signal, (int __user *)arg2);
1452 case PR_GET_DUMPABLE:
1453 error = get_dumpable(me->mm);
1455 case PR_SET_DUMPABLE:
1456 if (arg2 < 0 || arg2 > 1) {
1460 set_dumpable(me->mm, arg2);
1464 case PR_SET_UNALIGN:
1465 error = SET_UNALIGN_CTL(me, arg2);
1467 case PR_GET_UNALIGN:
1468 error = GET_UNALIGN_CTL(me, arg2);
1471 error = SET_FPEMU_CTL(me, arg2);
1474 error = GET_FPEMU_CTL(me, arg2);
1477 error = SET_FPEXC_CTL(me, arg2);
1480 error = GET_FPEXC_CTL(me, arg2);
1483 error = PR_TIMING_STATISTICAL;
1486 if (arg2 != PR_TIMING_STATISTICAL)
1493 comm[sizeof(me->comm)-1] = 0;
1494 if (strncpy_from_user(comm, (char __user *)arg2,
1495 sizeof(me->comm) - 1) < 0)
1497 set_task_comm(me, comm);
1500 get_task_comm(comm, me);
1501 if (copy_to_user((char __user *)arg2, comm,
1506 error = GET_ENDIAN(me, arg2);
1509 error = SET_ENDIAN(me, arg2);
1512 case PR_GET_SECCOMP:
1513 error = prctl_get_seccomp();
1515 case PR_SET_SECCOMP:
1516 error = prctl_set_seccomp(arg2);
1519 error = GET_TSC_CTL(arg2);
1522 error = SET_TSC_CTL(arg2);
1524 case PR_TASK_PERF_EVENTS_DISABLE:
1525 error = perf_event_task_disable();
1527 case PR_TASK_PERF_EVENTS_ENABLE:
1528 error = perf_event_task_enable();
1530 case PR_GET_TIMERSLACK:
1531 error = current->timer_slack_ns;
1533 case PR_SET_TIMERSLACK:
1535 current->timer_slack_ns =
1536 current->default_timer_slack_ns;
1538 current->timer_slack_ns = arg2;
1545 case PR_MCE_KILL_CLEAR:
1548 current->flags &= ~PF_MCE_PROCESS;
1550 case PR_MCE_KILL_SET:
1551 current->flags |= PF_MCE_PROCESS;
1552 if (arg3 == PR_MCE_KILL_EARLY)
1553 current->flags |= PF_MCE_EARLY;
1554 else if (arg3 == PR_MCE_KILL_LATE)
1555 current->flags &= ~PF_MCE_EARLY;
1556 else if (arg3 == PR_MCE_KILL_DEFAULT)
1558 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
1567 case PR_MCE_KILL_GET:
1568 if (arg2 | arg3 | arg4 | arg5)
1570 if (current->flags & PF_MCE_PROCESS)
1571 error = (current->flags & PF_MCE_EARLY) ?
1572 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
1574 error = PR_MCE_KILL_DEFAULT;
1583 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
1584 struct getcpu_cache __user *, unused)
1587 int cpu = raw_smp_processor_id();
1589 err |= put_user(cpu, cpup);
1591 err |= put_user(cpu_to_node(cpu), nodep);
1592 return err ? -EFAULT : 0;
1595 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
1597 static void argv_cleanup(char **argv, char **envp)
1603 * orderly_poweroff - Trigger an orderly system poweroff
1604 * @force: force poweroff if command execution fails
1606 * This may be called from any context to trigger a system shutdown.
1607 * If the orderly shutdown fails, it will force an immediate shutdown.
1609 int orderly_poweroff(bool force)
1612 char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
1613 static char *envp[] = {
1615 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
1619 struct subprocess_info *info;
1622 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
1623 __func__, poweroff_cmd);
1627 info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
1633 call_usermodehelper_setcleanup(info, argv_cleanup);
1635 ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
1639 printk(KERN_WARNING "Failed to start orderly shutdown: "
1640 "forcing the issue\n");
1642 /* I guess this should try to kick off some daemon to
1643 sync and poweroff asap. Or not even bother syncing
1644 if we're doing an emergency shutdown? */
1651 EXPORT_SYMBOL_GPL(orderly_poweroff);