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>
41 #include <linux/compat.h>
42 #include <linux/syscalls.h>
43 #include <linux/kprobes.h>
44 #include <linux/user_namespace.h>
46 #include <asm/uaccess.h>
48 #include <asm/unistd.h>
50 #ifndef SET_UNALIGN_CTL
51 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
53 #ifndef GET_UNALIGN_CTL
54 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
57 # define SET_FPEMU_CTL(a,b) (-EINVAL)
60 # define GET_FPEMU_CTL(a,b) (-EINVAL)
63 # define SET_FPEXC_CTL(a,b) (-EINVAL)
66 # define GET_FPEXC_CTL(a,b) (-EINVAL)
69 # define GET_ENDIAN(a,b) (-EINVAL)
72 # define SET_ENDIAN(a,b) (-EINVAL)
75 # define GET_TSC_CTL(a) (-EINVAL)
78 # define SET_TSC_CTL(a) (-EINVAL)
82 * this is where the system-wide overflow UID and GID are defined, for
83 * architectures that now have 32-bit UID/GID but didn't in the past
86 int overflowuid = DEFAULT_OVERFLOWUID;
87 int overflowgid = DEFAULT_OVERFLOWGID;
90 EXPORT_SYMBOL(overflowuid);
91 EXPORT_SYMBOL(overflowgid);
95 * the same as above, but for filesystems which can only store a 16-bit
96 * UID and GID. as such, this is needed on all architectures
99 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
100 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
102 EXPORT_SYMBOL(fs_overflowuid);
103 EXPORT_SYMBOL(fs_overflowgid);
106 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
111 EXPORT_SYMBOL(cad_pid);
114 * If set, this is used for preparing the system to power off.
117 void (*pm_power_off_prepare)(void);
120 * set the priority of a task
121 * - the caller must hold the RCU read lock
123 static int set_one_prio(struct task_struct *p, int niceval, int error)
125 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
128 if (pcred->uid != cred->euid &&
129 pcred->euid != cred->euid && !capable(CAP_SYS_NICE)) {
133 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
137 no_nice = security_task_setnice(p, niceval);
144 set_user_nice(p, niceval);
149 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
151 struct task_struct *g, *p;
152 struct user_struct *user;
153 const struct cred *cred = current_cred();
157 if (which > PRIO_USER || which < PRIO_PROCESS)
160 /* normalize: avoid signed division (rounding problems) */
168 read_lock(&tasklist_lock);
172 p = find_task_by_vpid(who);
176 error = set_one_prio(p, niceval, error);
180 pgrp = find_vpid(who);
182 pgrp = task_pgrp(current);
183 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
184 error = set_one_prio(p, niceval, error);
185 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
188 user = (struct user_struct *) cred->user;
191 else if ((who != cred->uid) &&
192 !(user = find_user(who)))
193 goto out_unlock; /* No processes for this user */
195 do_each_thread(g, p) {
196 if (__task_cred(p)->uid == who)
197 error = set_one_prio(p, niceval, error);
198 } while_each_thread(g, p);
199 if (who != cred->uid)
200 free_uid(user); /* For find_user() */
204 read_unlock(&tasklist_lock);
211 * Ugh. To avoid negative return values, "getpriority()" will
212 * not return the normal nice-value, but a negated value that
213 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
214 * to stay compatible.
216 SYSCALL_DEFINE2(getpriority, int, which, int, who)
218 struct task_struct *g, *p;
219 struct user_struct *user;
220 const struct cred *cred = current_cred();
221 long niceval, retval = -ESRCH;
224 if (which > PRIO_USER || which < PRIO_PROCESS)
228 read_lock(&tasklist_lock);
232 p = find_task_by_vpid(who);
236 niceval = 20 - task_nice(p);
237 if (niceval > retval)
243 pgrp = find_vpid(who);
245 pgrp = task_pgrp(current);
246 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
247 niceval = 20 - task_nice(p);
248 if (niceval > retval)
250 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
253 user = (struct user_struct *) cred->user;
256 else if ((who != cred->uid) &&
257 !(user = find_user(who)))
258 goto out_unlock; /* No processes for this user */
260 do_each_thread(g, p) {
261 if (__task_cred(p)->uid == who) {
262 niceval = 20 - task_nice(p);
263 if (niceval > retval)
266 } while_each_thread(g, p);
267 if (who != cred->uid)
268 free_uid(user); /* for find_user() */
272 read_unlock(&tasklist_lock);
279 * emergency_restart - reboot the system
281 * Without shutting down any hardware or taking any locks
282 * reboot the system. This is called when we know we are in
283 * trouble so this is our best effort to reboot. This is
284 * safe to call in interrupt context.
286 void emergency_restart(void)
288 machine_emergency_restart();
290 EXPORT_SYMBOL_GPL(emergency_restart);
292 void kernel_restart_prepare(char *cmd)
294 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
295 system_state = SYSTEM_RESTART;
301 * kernel_restart - reboot the system
302 * @cmd: pointer to buffer containing command to execute for restart
305 * Shutdown everything and perform a clean reboot.
306 * This is not safe to call in interrupt context.
308 void kernel_restart(char *cmd)
310 kernel_restart_prepare(cmd);
312 printk(KERN_EMERG "Restarting system.\n");
314 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
315 machine_restart(cmd);
317 EXPORT_SYMBOL_GPL(kernel_restart);
319 static void kernel_shutdown_prepare(enum system_states state)
321 blocking_notifier_call_chain(&reboot_notifier_list,
322 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
323 system_state = state;
327 * kernel_halt - halt the system
329 * Shutdown everything and perform a clean system halt.
331 void kernel_halt(void)
333 kernel_shutdown_prepare(SYSTEM_HALT);
335 printk(KERN_EMERG "System halted.\n");
339 EXPORT_SYMBOL_GPL(kernel_halt);
342 * kernel_power_off - power_off the system
344 * Shutdown everything and perform a clean system power_off.
346 void kernel_power_off(void)
348 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
349 if (pm_power_off_prepare)
350 pm_power_off_prepare();
351 disable_nonboot_cpus();
353 printk(KERN_EMERG "Power down.\n");
356 EXPORT_SYMBOL_GPL(kernel_power_off);
358 static DEFINE_MUTEX(reboot_mutex);
361 * Reboot system call: for obvious reasons only root may call it,
362 * and even root needs to set up some magic numbers in the registers
363 * so that some mistake won't make this reboot the whole machine.
364 * You can also set the meaning of the ctrl-alt-del-key here.
366 * reboot doesn't sync: do that yourself before calling this.
368 SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
374 /* We only trust the superuser with rebooting the system. */
375 if (!capable(CAP_SYS_BOOT))
378 /* For safety, we require "magic" arguments. */
379 if (magic1 != LINUX_REBOOT_MAGIC1 ||
380 (magic2 != LINUX_REBOOT_MAGIC2 &&
381 magic2 != LINUX_REBOOT_MAGIC2A &&
382 magic2 != LINUX_REBOOT_MAGIC2B &&
383 magic2 != LINUX_REBOOT_MAGIC2C))
386 /* Instead of trying to make the power_off code look like
387 * halt when pm_power_off is not set do it the easy way.
389 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
390 cmd = LINUX_REBOOT_CMD_HALT;
392 mutex_lock(&reboot_mutex);
394 case LINUX_REBOOT_CMD_RESTART:
395 kernel_restart(NULL);
398 case LINUX_REBOOT_CMD_CAD_ON:
402 case LINUX_REBOOT_CMD_CAD_OFF:
406 case LINUX_REBOOT_CMD_HALT:
409 panic("cannot halt");
411 case LINUX_REBOOT_CMD_POWER_OFF:
416 case LINUX_REBOOT_CMD_RESTART2:
417 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
421 buffer[sizeof(buffer) - 1] = '\0';
423 kernel_restart(buffer);
427 case LINUX_REBOOT_CMD_KEXEC:
428 ret = kernel_kexec();
432 #ifdef CONFIG_HIBERNATION
433 case LINUX_REBOOT_CMD_SW_SUSPEND:
442 mutex_unlock(&reboot_mutex);
446 static void deferred_cad(struct work_struct *dummy)
448 kernel_restart(NULL);
452 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
453 * As it's called within an interrupt, it may NOT sync: the only choice
454 * is whether to reboot at once, or just ignore the ctrl-alt-del.
456 void ctrl_alt_del(void)
458 static DECLARE_WORK(cad_work, deferred_cad);
461 schedule_work(&cad_work);
463 kill_cad_pid(SIGINT, 1);
467 * Unprivileged users may change the real gid to the effective gid
468 * or vice versa. (BSD-style)
470 * If you set the real gid at all, or set the effective gid to a value not
471 * equal to the real gid, then the saved gid is set to the new effective gid.
473 * This makes it possible for a setgid program to completely drop its
474 * privileges, which is often a useful assertion to make when you are doing
475 * a security audit over a program.
477 * The general idea is that a program which uses just setregid() will be
478 * 100% compatible with BSD. A program which uses just setgid() will be
479 * 100% compatible with POSIX with saved IDs.
481 * SMP: There are not races, the GIDs are checked only by filesystem
482 * operations (as far as semantic preservation is concerned).
484 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
486 const struct cred *old;
490 new = prepare_creds();
493 old = current_cred();
495 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
500 if (rgid != (gid_t) -1) {
501 if (old->gid == rgid ||
508 if (egid != (gid_t) -1) {
509 if (old->gid == egid ||
518 if (rgid != (gid_t) -1 ||
519 (egid != (gid_t) -1 && egid != old->gid))
520 new->sgid = new->egid;
521 new->fsgid = new->egid;
523 return commit_creds(new);
531 * setgid() is implemented like SysV w/ SAVED_IDS
533 * SMP: Same implicit races as above.
535 SYSCALL_DEFINE1(setgid, gid_t, gid)
537 const struct cred *old;
541 new = prepare_creds();
544 old = current_cred();
546 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
551 if (capable(CAP_SETGID))
552 new->gid = new->egid = new->sgid = new->fsgid = gid;
553 else if (gid == old->gid || gid == old->sgid)
554 new->egid = new->fsgid = gid;
558 return commit_creds(new);
566 * change the user struct in a credentials set to match the new UID
568 static int set_user(struct cred *new)
570 struct user_struct *new_user;
572 new_user = alloc_uid(current_user_ns(), new->uid);
576 if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
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 cputime_t tgutime, tgstime, cutime, cstime;
916 spin_lock_irq(¤t->sighand->siglock);
917 thread_group_times(current, &tgutime, &tgstime);
918 cutime = current->signal->cutime;
919 cstime = current->signal->cstime;
920 spin_unlock_irq(¤t->sighand->siglock);
921 tms->tms_utime = cputime_to_clock_t(tgutime);
922 tms->tms_stime = cputime_to_clock_t(tgstime);
923 tms->tms_cutime = cputime_to_clock_t(cutime);
924 tms->tms_cstime = cputime_to_clock_t(cstime);
927 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
933 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
936 force_successful_syscall_return();
937 return (long) jiffies_64_to_clock_t(get_jiffies_64());
941 * This needs some heavy checking ...
942 * I just haven't the stomach for it. I also don't fully
943 * understand sessions/pgrp etc. Let somebody who does explain it.
945 * OK, I think I have the protection semantics right.... this is really
946 * only important on a multi-user system anyway, to make sure one user
947 * can't send a signal to a process owned by another. -TYT, 12/12/91
949 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
952 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
954 struct task_struct *p;
955 struct task_struct *group_leader = current->group_leader;
960 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);
1021 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1023 struct task_struct *p;
1029 grp = task_pgrp(current);
1032 p = find_task_by_vpid(pid);
1039 retval = security_task_getpgid(p);
1043 retval = pid_vnr(grp);
1049 #ifdef __ARCH_WANT_SYS_GETPGRP
1051 SYSCALL_DEFINE0(getpgrp)
1053 return sys_getpgid(0);
1058 SYSCALL_DEFINE1(getsid, pid_t, pid)
1060 struct task_struct *p;
1066 sid = task_session(current);
1069 p = find_task_by_vpid(pid);
1072 sid = task_session(p);
1076 retval = security_task_getsid(p);
1080 retval = pid_vnr(sid);
1086 SYSCALL_DEFINE0(setsid)
1088 struct task_struct *group_leader = current->group_leader;
1089 struct pid *sid = task_pid(group_leader);
1090 pid_t session = pid_vnr(sid);
1093 write_lock_irq(&tasklist_lock);
1094 /* Fail if I am already a session leader */
1095 if (group_leader->signal->leader)
1098 /* Fail if a process group id already exists that equals the
1099 * proposed session id.
1101 if (pid_task(sid, PIDTYPE_PGID))
1104 group_leader->signal->leader = 1;
1105 __set_special_pids(sid);
1107 proc_clear_tty(group_leader);
1111 write_unlock_irq(&tasklist_lock);
1113 proc_sid_connector(group_leader);
1117 DECLARE_RWSEM(uts_sem);
1119 #ifdef COMPAT_UTS_MACHINE
1120 #define override_architecture(name) \
1121 (current->personality == PER_LINUX32 && \
1122 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1123 sizeof(COMPAT_UTS_MACHINE)))
1125 #define override_architecture(name) 0
1128 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1132 down_read(&uts_sem);
1133 if (copy_to_user(name, utsname(), sizeof *name))
1137 if (!errno && override_architecture(name))
1142 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1146 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1153 down_read(&uts_sem);
1154 if (copy_to_user(name, utsname(), sizeof(*name)))
1158 if (!error && override_architecture(name))
1163 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1169 if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
1172 down_read(&uts_sem);
1173 error = __copy_to_user(&name->sysname, &utsname()->sysname,
1175 error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
1176 error |= __copy_to_user(&name->nodename, &utsname()->nodename,
1178 error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
1179 error |= __copy_to_user(&name->release, &utsname()->release,
1181 error |= __put_user(0, name->release + __OLD_UTS_LEN);
1182 error |= __copy_to_user(&name->version, &utsname()->version,
1184 error |= __put_user(0, name->version + __OLD_UTS_LEN);
1185 error |= __copy_to_user(&name->machine, &utsname()->machine,
1187 error |= __put_user(0, name->machine + __OLD_UTS_LEN);
1190 if (!error && override_architecture(name))
1192 return error ? -EFAULT : 0;
1196 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1199 char tmp[__NEW_UTS_LEN];
1201 if (!capable(CAP_SYS_ADMIN))
1203 if (len < 0 || len > __NEW_UTS_LEN)
1205 down_write(&uts_sem);
1207 if (!copy_from_user(tmp, name, len)) {
1208 struct new_utsname *u = utsname();
1210 memcpy(u->nodename, tmp, len);
1211 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1218 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1220 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1223 struct new_utsname *u;
1227 down_read(&uts_sem);
1229 i = 1 + strlen(u->nodename);
1233 if (copy_to_user(name, u->nodename, i))
1242 * Only setdomainname; getdomainname can be implemented by calling
1245 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1248 char tmp[__NEW_UTS_LEN];
1250 if (!capable(CAP_SYS_ADMIN))
1252 if (len < 0 || len > __NEW_UTS_LEN)
1255 down_write(&uts_sem);
1257 if (!copy_from_user(tmp, name, len)) {
1258 struct new_utsname *u = utsname();
1260 memcpy(u->domainname, tmp, len);
1261 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1268 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1270 if (resource >= RLIM_NLIMITS)
1273 struct rlimit value;
1274 task_lock(current->group_leader);
1275 value = current->signal->rlim[resource];
1276 task_unlock(current->group_leader);
1277 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1281 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1284 * Back compatibility for getrlimit. Needed for some apps.
1287 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1288 struct rlimit __user *, rlim)
1291 if (resource >= RLIM_NLIMITS)
1294 task_lock(current->group_leader);
1295 x = current->signal->rlim[resource];
1296 task_unlock(current->group_leader);
1297 if (x.rlim_cur > 0x7FFFFFFF)
1298 x.rlim_cur = 0x7FFFFFFF;
1299 if (x.rlim_max > 0x7FFFFFFF)
1300 x.rlim_max = 0x7FFFFFFF;
1301 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1306 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1308 struct rlimit new_rlim, *old_rlim;
1311 if (resource >= RLIM_NLIMITS)
1313 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1315 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1317 old_rlim = current->signal->rlim + resource;
1318 if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1319 !capable(CAP_SYS_RESOURCE))
1321 if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > sysctl_nr_open)
1324 retval = security_task_setrlimit(resource, &new_rlim);
1328 if (resource == RLIMIT_CPU && new_rlim.rlim_cur == 0) {
1330 * The caller is asking for an immediate RLIMIT_CPU
1331 * expiry. But we use the zero value to mean "it was
1332 * never set". So let's cheat and make it one second
1335 new_rlim.rlim_cur = 1;
1338 task_lock(current->group_leader);
1339 *old_rlim = new_rlim;
1340 task_unlock(current->group_leader);
1342 if (resource != RLIMIT_CPU)
1346 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1347 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1348 * very long-standing error, and fixing it now risks breakage of
1349 * applications, so we live with it
1351 if (new_rlim.rlim_cur == RLIM_INFINITY)
1354 update_rlimit_cpu(new_rlim.rlim_cur);
1360 * It would make sense to put struct rusage in the task_struct,
1361 * except that would make the task_struct be *really big*. After
1362 * task_struct gets moved into malloc'ed memory, it would
1363 * make sense to do this. It will make moving the rest of the information
1364 * a lot simpler! (Which we're not doing right now because we're not
1365 * measuring them yet).
1367 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1368 * races with threads incrementing their own counters. But since word
1369 * reads are atomic, we either get new values or old values and we don't
1370 * care which for the sums. We always take the siglock to protect reading
1371 * the c* fields from p->signal from races with exit.c updating those
1372 * fields when reaping, so a sample either gets all the additions of a
1373 * given child after it's reaped, or none so this sample is before reaping.
1376 * We need to take the siglock for CHILDEREN, SELF and BOTH
1377 * for the cases current multithreaded, non-current single threaded
1378 * non-current multithreaded. Thread traversal is now safe with
1380 * Strictly speaking, we donot need to take the siglock if we are current and
1381 * single threaded, as no one else can take our signal_struct away, no one
1382 * else can reap the children to update signal->c* counters, and no one else
1383 * can race with the signal-> fields. If we do not take any lock, the
1384 * signal-> fields could be read out of order while another thread was just
1385 * exiting. So we should place a read memory barrier when we avoid the lock.
1386 * On the writer side, write memory barrier is implied in __exit_signal
1387 * as __exit_signal releases the siglock spinlock after updating the signal->
1388 * fields. But we don't do this yet to keep things simple.
1392 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1394 r->ru_nvcsw += t->nvcsw;
1395 r->ru_nivcsw += t->nivcsw;
1396 r->ru_minflt += t->min_flt;
1397 r->ru_majflt += t->maj_flt;
1398 r->ru_inblock += task_io_get_inblock(t);
1399 r->ru_oublock += task_io_get_oublock(t);
1402 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1404 struct task_struct *t;
1405 unsigned long flags;
1406 cputime_t tgutime, tgstime, utime, stime;
1407 unsigned long maxrss = 0;
1409 memset((char *) r, 0, sizeof *r);
1410 utime = stime = cputime_zero;
1412 if (who == RUSAGE_THREAD) {
1413 task_times(current, &utime, &stime);
1414 accumulate_thread_rusage(p, r);
1415 maxrss = p->signal->maxrss;
1419 if (!lock_task_sighand(p, &flags))
1424 case RUSAGE_CHILDREN:
1425 utime = p->signal->cutime;
1426 stime = p->signal->cstime;
1427 r->ru_nvcsw = p->signal->cnvcsw;
1428 r->ru_nivcsw = p->signal->cnivcsw;
1429 r->ru_minflt = p->signal->cmin_flt;
1430 r->ru_majflt = p->signal->cmaj_flt;
1431 r->ru_inblock = p->signal->cinblock;
1432 r->ru_oublock = p->signal->coublock;
1433 maxrss = p->signal->cmaxrss;
1435 if (who == RUSAGE_CHILDREN)
1439 thread_group_times(p, &tgutime, &tgstime);
1440 utime = cputime_add(utime, tgutime);
1441 stime = cputime_add(stime, tgstime);
1442 r->ru_nvcsw += p->signal->nvcsw;
1443 r->ru_nivcsw += p->signal->nivcsw;
1444 r->ru_minflt += p->signal->min_flt;
1445 r->ru_majflt += p->signal->maj_flt;
1446 r->ru_inblock += p->signal->inblock;
1447 r->ru_oublock += p->signal->oublock;
1448 if (maxrss < p->signal->maxrss)
1449 maxrss = p->signal->maxrss;
1452 accumulate_thread_rusage(t, r);
1460 unlock_task_sighand(p, &flags);
1463 cputime_to_timeval(utime, &r->ru_utime);
1464 cputime_to_timeval(stime, &r->ru_stime);
1466 if (who != RUSAGE_CHILDREN) {
1467 struct mm_struct *mm = get_task_mm(p);
1469 setmax_mm_hiwater_rss(&maxrss, mm);
1473 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1476 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1479 k_getrusage(p, who, &r);
1480 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1483 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1485 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1486 who != RUSAGE_THREAD)
1488 return getrusage(current, who, ru);
1491 SYSCALL_DEFINE1(umask, int, mask)
1493 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1497 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
1498 unsigned long, arg4, unsigned long, arg5)
1500 struct task_struct *me = current;
1501 unsigned char comm[sizeof(me->comm)];
1504 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1505 if (error != -ENOSYS)
1510 case PR_SET_PDEATHSIG:
1511 if (!valid_signal(arg2)) {
1515 me->pdeath_signal = arg2;
1518 case PR_GET_PDEATHSIG:
1519 error = put_user(me->pdeath_signal, (int __user *)arg2);
1521 case PR_GET_DUMPABLE:
1522 error = get_dumpable(me->mm);
1524 case PR_SET_DUMPABLE:
1525 if (arg2 < 0 || arg2 > 1) {
1529 set_dumpable(me->mm, arg2);
1533 case PR_SET_UNALIGN:
1534 error = SET_UNALIGN_CTL(me, arg2);
1536 case PR_GET_UNALIGN:
1537 error = GET_UNALIGN_CTL(me, arg2);
1540 error = SET_FPEMU_CTL(me, arg2);
1543 error = GET_FPEMU_CTL(me, arg2);
1546 error = SET_FPEXC_CTL(me, arg2);
1549 error = GET_FPEXC_CTL(me, arg2);
1552 error = PR_TIMING_STATISTICAL;
1555 if (arg2 != PR_TIMING_STATISTICAL)
1562 comm[sizeof(me->comm)-1] = 0;
1563 if (strncpy_from_user(comm, (char __user *)arg2,
1564 sizeof(me->comm) - 1) < 0)
1566 set_task_comm(me, comm);
1569 get_task_comm(comm, me);
1570 if (copy_to_user((char __user *)arg2, comm,
1575 error = GET_ENDIAN(me, arg2);
1578 error = SET_ENDIAN(me, arg2);
1581 case PR_GET_SECCOMP:
1582 error = prctl_get_seccomp();
1584 case PR_SET_SECCOMP:
1585 error = prctl_set_seccomp(arg2);
1588 error = GET_TSC_CTL(arg2);
1591 error = SET_TSC_CTL(arg2);
1593 case PR_TASK_PERF_EVENTS_DISABLE:
1594 error = perf_event_task_disable();
1596 case PR_TASK_PERF_EVENTS_ENABLE:
1597 error = perf_event_task_enable();
1599 case PR_GET_TIMERSLACK:
1600 error = current->timer_slack_ns;
1602 case PR_SET_TIMERSLACK:
1604 current->timer_slack_ns =
1605 current->default_timer_slack_ns;
1607 current->timer_slack_ns = arg2;
1614 case PR_MCE_KILL_CLEAR:
1617 current->flags &= ~PF_MCE_PROCESS;
1619 case PR_MCE_KILL_SET:
1620 current->flags |= PF_MCE_PROCESS;
1621 if (arg3 == PR_MCE_KILL_EARLY)
1622 current->flags |= PF_MCE_EARLY;
1623 else if (arg3 == PR_MCE_KILL_LATE)
1624 current->flags &= ~PF_MCE_EARLY;
1625 else if (arg3 == PR_MCE_KILL_DEFAULT)
1627 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
1636 case PR_MCE_KILL_GET:
1637 if (arg2 | arg3 | arg4 | arg5)
1639 if (current->flags & PF_MCE_PROCESS)
1640 error = (current->flags & PF_MCE_EARLY) ?
1641 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
1643 error = PR_MCE_KILL_DEFAULT;
1652 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
1653 struct getcpu_cache __user *, unused)
1656 int cpu = raw_smp_processor_id();
1658 err |= put_user(cpu, cpup);
1660 err |= put_user(cpu_to_node(cpu), nodep);
1661 return err ? -EFAULT : 0;
1664 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
1666 static void argv_cleanup(char **argv, char **envp)
1672 * orderly_poweroff - Trigger an orderly system poweroff
1673 * @force: force poweroff if command execution fails
1675 * This may be called from any context to trigger a system shutdown.
1676 * If the orderly shutdown fails, it will force an immediate shutdown.
1678 int orderly_poweroff(bool force)
1681 char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
1682 static char *envp[] = {
1684 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
1688 struct subprocess_info *info;
1691 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
1692 __func__, poweroff_cmd);
1696 info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
1702 call_usermodehelper_setcleanup(info, argv_cleanup);
1704 ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
1708 printk(KERN_WARNING "Failed to start orderly shutdown: "
1709 "forcing the issue\n");
1711 /* I guess this should try to kick off some daemon to
1712 sync and poweroff asap. Or not even bother syncing
1713 if we're doing an emergency shutdown? */
1720 EXPORT_SYMBOL_GPL(orderly_poweroff);