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>
40 #include <linux/compat.h>
41 #include <linux/syscalls.h>
42 #include <linux/kprobes.h>
43 #include <linux/user_namespace.h>
45 #include <asm/uaccess.h>
47 #include <asm/unistd.h>
49 #ifndef SET_UNALIGN_CTL
50 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
52 #ifndef GET_UNALIGN_CTL
53 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
56 # define SET_FPEMU_CTL(a,b) (-EINVAL)
59 # define GET_FPEMU_CTL(a,b) (-EINVAL)
62 # define SET_FPEXC_CTL(a,b) (-EINVAL)
65 # define GET_FPEXC_CTL(a,b) (-EINVAL)
68 # define GET_ENDIAN(a,b) (-EINVAL)
71 # define SET_ENDIAN(a,b) (-EINVAL)
74 # define GET_TSC_CTL(a) (-EINVAL)
77 # define SET_TSC_CTL(a) (-EINVAL)
81 * this is where the system-wide overflow UID and GID are defined, for
82 * architectures that now have 32-bit UID/GID but didn't in the past
85 int overflowuid = DEFAULT_OVERFLOWUID;
86 int overflowgid = DEFAULT_OVERFLOWGID;
89 EXPORT_SYMBOL(overflowuid);
90 EXPORT_SYMBOL(overflowgid);
94 * the same as above, but for filesystems which can only store a 16-bit
95 * UID and GID. as such, this is needed on all architectures
98 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
99 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
101 EXPORT_SYMBOL(fs_overflowuid);
102 EXPORT_SYMBOL(fs_overflowgid);
105 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
110 EXPORT_SYMBOL(cad_pid);
113 * If set, this is used for preparing the system to power off.
116 void (*pm_power_off_prepare)(void);
119 * set the priority of a task
120 * - the caller must hold the RCU read lock
122 static int set_one_prio(struct task_struct *p, int niceval, int error)
124 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
127 if (pcred->uid != cred->euid &&
128 pcred->euid != cred->euid && !capable(CAP_SYS_NICE)) {
132 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
136 no_nice = security_task_setnice(p, niceval);
143 set_user_nice(p, niceval);
148 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
150 struct task_struct *g, *p;
151 struct user_struct *user;
152 const struct cred *cred = current_cred();
156 if (which > PRIO_USER || which < PRIO_PROCESS)
159 /* normalize: avoid signed division (rounding problems) */
167 read_lock(&tasklist_lock);
171 p = find_task_by_vpid(who);
175 error = set_one_prio(p, niceval, error);
179 pgrp = find_vpid(who);
181 pgrp = task_pgrp(current);
182 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
183 error = set_one_prio(p, niceval, error);
184 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
187 user = (struct user_struct *) cred->user;
190 else if ((who != cred->uid) &&
191 !(user = find_user(who)))
192 goto out_unlock; /* No processes for this user */
194 do_each_thread(g, p) {
195 if (__task_cred(p)->uid == who)
196 error = set_one_prio(p, niceval, error);
197 } while_each_thread(g, p);
198 if (who != cred->uid)
199 free_uid(user); /* For find_user() */
203 read_unlock(&tasklist_lock);
210 * Ugh. To avoid negative return values, "getpriority()" will
211 * not return the normal nice-value, but a negated value that
212 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
213 * to stay compatible.
215 SYSCALL_DEFINE2(getpriority, int, which, int, who)
217 struct task_struct *g, *p;
218 struct user_struct *user;
219 const struct cred *cred = current_cred();
220 long niceval, retval = -ESRCH;
223 if (which > PRIO_USER || which < PRIO_PROCESS)
227 read_lock(&tasklist_lock);
231 p = find_task_by_vpid(who);
235 niceval = 20 - task_nice(p);
236 if (niceval > retval)
242 pgrp = find_vpid(who);
244 pgrp = task_pgrp(current);
245 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
246 niceval = 20 - task_nice(p);
247 if (niceval > retval)
249 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
252 user = (struct user_struct *) cred->user;
255 else if ((who != cred->uid) &&
256 !(user = find_user(who)))
257 goto out_unlock; /* No processes for this user */
259 do_each_thread(g, p) {
260 if (__task_cred(p)->uid == who) {
261 niceval = 20 - task_nice(p);
262 if (niceval > retval)
265 } while_each_thread(g, p);
266 if (who != cred->uid)
267 free_uid(user); /* for find_user() */
271 read_unlock(&tasklist_lock);
278 * emergency_restart - reboot the system
280 * Without shutting down any hardware or taking any locks
281 * reboot the system. This is called when we know we are in
282 * trouble so this is our best effort to reboot. This is
283 * safe to call in interrupt context.
285 void emergency_restart(void)
287 machine_emergency_restart();
289 EXPORT_SYMBOL_GPL(emergency_restart);
291 void kernel_restart_prepare(char *cmd)
293 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
294 system_state = SYSTEM_RESTART;
300 * kernel_restart - reboot the system
301 * @cmd: pointer to buffer containing command to execute for restart
304 * Shutdown everything and perform a clean reboot.
305 * This is not safe to call in interrupt context.
307 void kernel_restart(char *cmd)
309 kernel_restart_prepare(cmd);
311 printk(KERN_EMERG "Restarting system.\n");
313 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
314 machine_restart(cmd);
316 EXPORT_SYMBOL_GPL(kernel_restart);
318 static void kernel_shutdown_prepare(enum system_states state)
320 blocking_notifier_call_chain(&reboot_notifier_list,
321 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
322 system_state = state;
326 * kernel_halt - halt the system
328 * Shutdown everything and perform a clean system halt.
330 void kernel_halt(void)
332 kernel_shutdown_prepare(SYSTEM_HALT);
334 printk(KERN_EMERG "System halted.\n");
338 EXPORT_SYMBOL_GPL(kernel_halt);
341 * kernel_power_off - power_off the system
343 * Shutdown everything and perform a clean system power_off.
345 void kernel_power_off(void)
347 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
348 if (pm_power_off_prepare)
349 pm_power_off_prepare();
350 disable_nonboot_cpus();
352 printk(KERN_EMERG "Power down.\n");
355 EXPORT_SYMBOL_GPL(kernel_power_off);
357 static DEFINE_MUTEX(reboot_mutex);
360 * Reboot system call: for obvious reasons only root may call it,
361 * and even root needs to set up some magic numbers in the registers
362 * so that some mistake won't make this reboot the whole machine.
363 * You can also set the meaning of the ctrl-alt-del-key here.
365 * reboot doesn't sync: do that yourself before calling this.
367 SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
373 /* We only trust the superuser with rebooting the system. */
374 if (!capable(CAP_SYS_BOOT))
377 /* For safety, we require "magic" arguments. */
378 if (magic1 != LINUX_REBOOT_MAGIC1 ||
379 (magic2 != LINUX_REBOOT_MAGIC2 &&
380 magic2 != LINUX_REBOOT_MAGIC2A &&
381 magic2 != LINUX_REBOOT_MAGIC2B &&
382 magic2 != LINUX_REBOOT_MAGIC2C))
385 /* Instead of trying to make the power_off code look like
386 * halt when pm_power_off is not set do it the easy way.
388 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
389 cmd = LINUX_REBOOT_CMD_HALT;
391 mutex_lock(&reboot_mutex);
393 case LINUX_REBOOT_CMD_RESTART:
394 kernel_restart(NULL);
397 case LINUX_REBOOT_CMD_CAD_ON:
401 case LINUX_REBOOT_CMD_CAD_OFF:
405 case LINUX_REBOOT_CMD_HALT:
408 panic("cannot halt");
410 case LINUX_REBOOT_CMD_POWER_OFF:
415 case LINUX_REBOOT_CMD_RESTART2:
416 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
420 buffer[sizeof(buffer) - 1] = '\0';
422 kernel_restart(buffer);
426 case LINUX_REBOOT_CMD_KEXEC:
427 ret = kernel_kexec();
431 #ifdef CONFIG_HIBERNATION
432 case LINUX_REBOOT_CMD_SW_SUSPEND:
441 mutex_unlock(&reboot_mutex);
445 static void deferred_cad(struct work_struct *dummy)
447 kernel_restart(NULL);
451 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
452 * As it's called within an interrupt, it may NOT sync: the only choice
453 * is whether to reboot at once, or just ignore the ctrl-alt-del.
455 void ctrl_alt_del(void)
457 static DECLARE_WORK(cad_work, deferred_cad);
460 schedule_work(&cad_work);
462 kill_cad_pid(SIGINT, 1);
466 * Unprivileged users may change the real gid to the effective gid
467 * or vice versa. (BSD-style)
469 * If you set the real gid at all, or set the effective gid to a value not
470 * equal to the real gid, then the saved gid is set to the new effective gid.
472 * This makes it possible for a setgid program to completely drop its
473 * privileges, which is often a useful assertion to make when you are doing
474 * a security audit over a program.
476 * The general idea is that a program which uses just setregid() will be
477 * 100% compatible with BSD. A program which uses just setgid() will be
478 * 100% compatible with POSIX with saved IDs.
480 * SMP: There are not races, the GIDs are checked only by filesystem
481 * operations (as far as semantic preservation is concerned).
483 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
485 const struct cred *old;
489 new = prepare_creds();
492 old = current_cred();
495 if (rgid != (gid_t) -1) {
496 if (old->gid == rgid ||
503 if (egid != (gid_t) -1) {
504 if (old->gid == egid ||
513 if (rgid != (gid_t) -1 ||
514 (egid != (gid_t) -1 && egid != old->gid))
515 new->sgid = new->egid;
516 new->fsgid = new->egid;
518 return commit_creds(new);
526 * setgid() is implemented like SysV w/ SAVED_IDS
528 * SMP: Same implicit races as above.
530 SYSCALL_DEFINE1(setgid, gid_t, gid)
532 const struct cred *old;
536 new = prepare_creds();
539 old = current_cred();
542 if (capable(CAP_SETGID))
543 new->gid = new->egid = new->sgid = new->fsgid = gid;
544 else if (gid == old->gid || gid == old->sgid)
545 new->egid = new->fsgid = gid;
549 return commit_creds(new);
557 * change the user struct in a credentials set to match the new UID
559 static int set_user(struct cred *new)
561 struct user_struct *new_user;
563 new_user = alloc_uid(current_user_ns(), new->uid);
567 if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
568 new_user != INIT_USER) {
574 new->user = new_user;
579 * Unprivileged users may change the real uid to the effective uid
580 * or vice versa. (BSD-style)
582 * If you set the real uid at all, or set the effective uid to a value not
583 * equal to the real uid, then the saved uid is set to the new effective uid.
585 * This makes it possible for a setuid program to completely drop its
586 * privileges, which is often a useful assertion to make when you are doing
587 * a security audit over a program.
589 * The general idea is that a program which uses just setreuid() will be
590 * 100% compatible with BSD. A program which uses just setuid() will be
591 * 100% compatible with POSIX with saved IDs.
593 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
595 const struct cred *old;
599 new = prepare_creds();
602 old = current_cred();
605 if (ruid != (uid_t) -1) {
607 if (old->uid != ruid &&
609 !capable(CAP_SETUID))
613 if (euid != (uid_t) -1) {
615 if (old->uid != euid &&
618 !capable(CAP_SETUID))
622 if (new->uid != old->uid) {
623 retval = set_user(new);
627 if (ruid != (uid_t) -1 ||
628 (euid != (uid_t) -1 && euid != old->uid))
629 new->suid = new->euid;
630 new->fsuid = new->euid;
632 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
636 return commit_creds(new);
644 * setuid() is implemented like SysV with SAVED_IDS
646 * Note that SAVED_ID's is deficient in that a setuid root program
647 * like sendmail, for example, cannot set its uid to be a normal
648 * user and then switch back, because if you're root, setuid() sets
649 * the saved uid too. If you don't like this, blame the bright people
650 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
651 * will allow a root program to temporarily drop privileges and be able to
652 * regain them by swapping the real and effective uid.
654 SYSCALL_DEFINE1(setuid, uid_t, uid)
656 const struct cred *old;
660 new = prepare_creds();
663 old = current_cred();
666 if (capable(CAP_SETUID)) {
667 new->suid = new->uid = uid;
668 if (uid != old->uid) {
669 retval = set_user(new);
673 } else if (uid != old->uid && uid != new->suid) {
677 new->fsuid = new->euid = uid;
679 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
683 return commit_creds(new);
692 * This function implements a generic ability to update ruid, euid,
693 * and suid. This allows you to implement the 4.4 compatible seteuid().
695 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
697 const struct cred *old;
701 new = prepare_creds();
705 old = current_cred();
708 if (!capable(CAP_SETUID)) {
709 if (ruid != (uid_t) -1 && ruid != old->uid &&
710 ruid != old->euid && ruid != old->suid)
712 if (euid != (uid_t) -1 && euid != old->uid &&
713 euid != old->euid && euid != old->suid)
715 if (suid != (uid_t) -1 && suid != old->uid &&
716 suid != old->euid && suid != old->suid)
720 if (ruid != (uid_t) -1) {
722 if (ruid != old->uid) {
723 retval = set_user(new);
728 if (euid != (uid_t) -1)
730 if (suid != (uid_t) -1)
732 new->fsuid = new->euid;
734 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
738 return commit_creds(new);
745 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruid, uid_t __user *, euid, uid_t __user *, suid)
747 const struct cred *cred = current_cred();
750 if (!(retval = put_user(cred->uid, ruid)) &&
751 !(retval = put_user(cred->euid, euid)))
752 retval = put_user(cred->suid, suid);
758 * Same as above, but for rgid, egid, sgid.
760 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
762 const struct cred *old;
766 new = prepare_creds();
769 old = current_cred();
772 if (!capable(CAP_SETGID)) {
773 if (rgid != (gid_t) -1 && rgid != old->gid &&
774 rgid != old->egid && rgid != old->sgid)
776 if (egid != (gid_t) -1 && egid != old->gid &&
777 egid != old->egid && egid != old->sgid)
779 if (sgid != (gid_t) -1 && sgid != old->gid &&
780 sgid != old->egid && sgid != old->sgid)
784 if (rgid != (gid_t) -1)
786 if (egid != (gid_t) -1)
788 if (sgid != (gid_t) -1)
790 new->fsgid = new->egid;
792 return commit_creds(new);
799 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgid, gid_t __user *, egid, gid_t __user *, sgid)
801 const struct cred *cred = current_cred();
804 if (!(retval = put_user(cred->gid, rgid)) &&
805 !(retval = put_user(cred->egid, egid)))
806 retval = put_user(cred->sgid, sgid);
813 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
814 * is used for "access()" and for the NFS daemon (letting nfsd stay at
815 * whatever uid it wants to). It normally shadows "euid", except when
816 * explicitly set by setfsuid() or for access..
818 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
820 const struct cred *old;
824 new = prepare_creds();
826 return current_fsuid();
827 old = current_cred();
828 old_fsuid = old->fsuid;
830 if (uid == old->uid || uid == old->euid ||
831 uid == old->suid || uid == old->fsuid ||
832 capable(CAP_SETUID)) {
833 if (uid != old_fsuid) {
835 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
849 * Samma på svenska..
851 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
853 const struct cred *old;
857 new = prepare_creds();
859 return current_fsgid();
860 old = current_cred();
861 old_fsgid = old->fsgid;
863 if (gid == old->gid || gid == old->egid ||
864 gid == old->sgid || gid == old->fsgid ||
865 capable(CAP_SETGID)) {
866 if (gid != old_fsgid) {
880 void do_sys_times(struct tms *tms)
882 cputime_t tgutime, tgstime, cutime, cstime;
884 spin_lock_irq(¤t->sighand->siglock);
885 thread_group_times(current, &tgutime, &tgstime);
886 cutime = current->signal->cutime;
887 cstime = current->signal->cstime;
888 spin_unlock_irq(¤t->sighand->siglock);
889 tms->tms_utime = cputime_to_clock_t(tgutime);
890 tms->tms_stime = cputime_to_clock_t(tgstime);
891 tms->tms_cutime = cputime_to_clock_t(cutime);
892 tms->tms_cstime = cputime_to_clock_t(cstime);
895 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
901 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
904 force_successful_syscall_return();
905 return (long) jiffies_64_to_clock_t(get_jiffies_64());
909 * This needs some heavy checking ...
910 * I just haven't the stomach for it. I also don't fully
911 * understand sessions/pgrp etc. Let somebody who does explain it.
913 * OK, I think I have the protection semantics right.... this is really
914 * only important on a multi-user system anyway, to make sure one user
915 * can't send a signal to a process owned by another. -TYT, 12/12/91
917 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
920 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
922 struct task_struct *p;
923 struct task_struct *group_leader = current->group_leader;
928 pid = task_pid_vnr(group_leader);
934 /* From this point forward we keep holding onto the tasklist lock
935 * so that our parent does not change from under us. -DaveM
937 write_lock_irq(&tasklist_lock);
940 p = find_task_by_vpid(pid);
945 if (!thread_group_leader(p))
948 if (same_thread_group(p->real_parent, group_leader)) {
950 if (task_session(p) != task_session(group_leader))
957 if (p != group_leader)
962 if (p->signal->leader)
967 struct task_struct *g;
969 pgrp = find_vpid(pgid);
970 g = pid_task(pgrp, PIDTYPE_PGID);
971 if (!g || task_session(g) != task_session(group_leader))
975 err = security_task_setpgid(p, pgid);
979 if (task_pgrp(p) != pgrp)
980 change_pid(p, PIDTYPE_PGID, pgrp);
984 /* All paths lead to here, thus we are safe. -DaveM */
985 write_unlock_irq(&tasklist_lock);
989 SYSCALL_DEFINE1(getpgid, pid_t, pid)
991 struct task_struct *p;
997 grp = task_pgrp(current);
1000 p = find_task_by_vpid(pid);
1007 retval = security_task_getpgid(p);
1011 retval = pid_vnr(grp);
1017 #ifdef __ARCH_WANT_SYS_GETPGRP
1019 SYSCALL_DEFINE0(getpgrp)
1021 return sys_getpgid(0);
1026 SYSCALL_DEFINE1(getsid, pid_t, pid)
1028 struct task_struct *p;
1034 sid = task_session(current);
1037 p = find_task_by_vpid(pid);
1040 sid = task_session(p);
1044 retval = security_task_getsid(p);
1048 retval = pid_vnr(sid);
1054 SYSCALL_DEFINE0(setsid)
1056 struct task_struct *group_leader = current->group_leader;
1057 struct pid *sid = task_pid(group_leader);
1058 pid_t session = pid_vnr(sid);
1061 write_lock_irq(&tasklist_lock);
1062 /* Fail if I am already a session leader */
1063 if (group_leader->signal->leader)
1066 /* Fail if a process group id already exists that equals the
1067 * proposed session id.
1069 if (pid_task(sid, PIDTYPE_PGID))
1072 group_leader->signal->leader = 1;
1073 __set_special_pids(sid);
1075 proc_clear_tty(group_leader);
1079 write_unlock_irq(&tasklist_lock);
1081 proc_sid_connector(group_leader);
1085 DECLARE_RWSEM(uts_sem);
1087 #ifdef COMPAT_UTS_MACHINE
1088 #define override_architecture(name) \
1089 (current->personality == PER_LINUX32 && \
1090 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1091 sizeof(COMPAT_UTS_MACHINE)))
1093 #define override_architecture(name) 0
1096 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1100 down_read(&uts_sem);
1101 if (copy_to_user(name, utsname(), sizeof *name))
1105 if (!errno && override_architecture(name))
1110 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1114 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1121 down_read(&uts_sem);
1122 if (copy_to_user(name, utsname(), sizeof(*name)))
1126 if (!error && override_architecture(name))
1131 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1137 if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
1140 down_read(&uts_sem);
1141 error = __copy_to_user(&name->sysname, &utsname()->sysname,
1143 error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
1144 error |= __copy_to_user(&name->nodename, &utsname()->nodename,
1146 error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
1147 error |= __copy_to_user(&name->release, &utsname()->release,
1149 error |= __put_user(0, name->release + __OLD_UTS_LEN);
1150 error |= __copy_to_user(&name->version, &utsname()->version,
1152 error |= __put_user(0, name->version + __OLD_UTS_LEN);
1153 error |= __copy_to_user(&name->machine, &utsname()->machine,
1155 error |= __put_user(0, name->machine + __OLD_UTS_LEN);
1158 if (!error && override_architecture(name))
1160 return error ? -EFAULT : 0;
1164 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1167 char tmp[__NEW_UTS_LEN];
1169 if (!capable(CAP_SYS_ADMIN))
1171 if (len < 0 || len > __NEW_UTS_LEN)
1173 down_write(&uts_sem);
1175 if (!copy_from_user(tmp, name, len)) {
1176 struct new_utsname *u = utsname();
1178 memcpy(u->nodename, tmp, len);
1179 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1186 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1188 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1191 struct new_utsname *u;
1195 down_read(&uts_sem);
1197 i = 1 + strlen(u->nodename);
1201 if (copy_to_user(name, u->nodename, i))
1210 * Only setdomainname; getdomainname can be implemented by calling
1213 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1216 char tmp[__NEW_UTS_LEN];
1218 if (!capable(CAP_SYS_ADMIN))
1220 if (len < 0 || len > __NEW_UTS_LEN)
1223 down_write(&uts_sem);
1225 if (!copy_from_user(tmp, name, len)) {
1226 struct new_utsname *u = utsname();
1228 memcpy(u->domainname, tmp, len);
1229 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1236 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1238 if (resource >= RLIM_NLIMITS)
1241 struct rlimit value;
1242 task_lock(current->group_leader);
1243 value = current->signal->rlim[resource];
1244 task_unlock(current->group_leader);
1245 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1249 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1252 * Back compatibility for getrlimit. Needed for some apps.
1255 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1256 struct rlimit __user *, rlim)
1259 if (resource >= RLIM_NLIMITS)
1262 task_lock(current->group_leader);
1263 x = current->signal->rlim[resource];
1264 task_unlock(current->group_leader);
1265 if (x.rlim_cur > 0x7FFFFFFF)
1266 x.rlim_cur = 0x7FFFFFFF;
1267 if (x.rlim_max > 0x7FFFFFFF)
1268 x.rlim_max = 0x7FFFFFFF;
1269 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1274 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1276 struct rlimit new_rlim, *old_rlim;
1279 if (resource >= RLIM_NLIMITS)
1281 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1283 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1285 old_rlim = current->signal->rlim + resource;
1286 if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1287 !capable(CAP_SYS_RESOURCE))
1289 if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > sysctl_nr_open)
1292 retval = security_task_setrlimit(resource, &new_rlim);
1296 if (resource == RLIMIT_CPU && new_rlim.rlim_cur == 0) {
1298 * The caller is asking for an immediate RLIMIT_CPU
1299 * expiry. But we use the zero value to mean "it was
1300 * never set". So let's cheat and make it one second
1303 new_rlim.rlim_cur = 1;
1306 task_lock(current->group_leader);
1307 *old_rlim = new_rlim;
1308 task_unlock(current->group_leader);
1310 if (resource != RLIMIT_CPU)
1314 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1315 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1316 * very long-standing error, and fixing it now risks breakage of
1317 * applications, so we live with it
1319 if (new_rlim.rlim_cur == RLIM_INFINITY)
1322 update_rlimit_cpu(new_rlim.rlim_cur);
1328 * It would make sense to put struct rusage in the task_struct,
1329 * except that would make the task_struct be *really big*. After
1330 * task_struct gets moved into malloc'ed memory, it would
1331 * make sense to do this. It will make moving the rest of the information
1332 * a lot simpler! (Which we're not doing right now because we're not
1333 * measuring them yet).
1335 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1336 * races with threads incrementing their own counters. But since word
1337 * reads are atomic, we either get new values or old values and we don't
1338 * care which for the sums. We always take the siglock to protect reading
1339 * the c* fields from p->signal from races with exit.c updating those
1340 * fields when reaping, so a sample either gets all the additions of a
1341 * given child after it's reaped, or none so this sample is before reaping.
1344 * We need to take the siglock for CHILDEREN, SELF and BOTH
1345 * for the cases current multithreaded, non-current single threaded
1346 * non-current multithreaded. Thread traversal is now safe with
1348 * Strictly speaking, we donot need to take the siglock if we are current and
1349 * single threaded, as no one else can take our signal_struct away, no one
1350 * else can reap the children to update signal->c* counters, and no one else
1351 * can race with the signal-> fields. If we do not take any lock, the
1352 * signal-> fields could be read out of order while another thread was just
1353 * exiting. So we should place a read memory barrier when we avoid the lock.
1354 * On the writer side, write memory barrier is implied in __exit_signal
1355 * as __exit_signal releases the siglock spinlock after updating the signal->
1356 * fields. But we don't do this yet to keep things simple.
1360 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1362 r->ru_nvcsw += t->nvcsw;
1363 r->ru_nivcsw += t->nivcsw;
1364 r->ru_minflt += t->min_flt;
1365 r->ru_majflt += t->maj_flt;
1366 r->ru_inblock += task_io_get_inblock(t);
1367 r->ru_oublock += task_io_get_oublock(t);
1370 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1372 struct task_struct *t;
1373 unsigned long flags;
1374 cputime_t tgutime, tgstime, utime, stime;
1375 unsigned long maxrss = 0;
1377 memset((char *) r, 0, sizeof *r);
1378 utime = stime = cputime_zero;
1380 if (who == RUSAGE_THREAD) {
1381 task_times(current, &utime, &stime);
1382 accumulate_thread_rusage(p, r);
1383 maxrss = p->signal->maxrss;
1387 if (!lock_task_sighand(p, &flags))
1392 case RUSAGE_CHILDREN:
1393 utime = p->signal->cutime;
1394 stime = p->signal->cstime;
1395 r->ru_nvcsw = p->signal->cnvcsw;
1396 r->ru_nivcsw = p->signal->cnivcsw;
1397 r->ru_minflt = p->signal->cmin_flt;
1398 r->ru_majflt = p->signal->cmaj_flt;
1399 r->ru_inblock = p->signal->cinblock;
1400 r->ru_oublock = p->signal->coublock;
1401 maxrss = p->signal->cmaxrss;
1403 if (who == RUSAGE_CHILDREN)
1407 thread_group_times(p, &tgutime, &tgstime);
1408 utime = cputime_add(utime, tgutime);
1409 stime = cputime_add(stime, tgstime);
1410 r->ru_nvcsw += p->signal->nvcsw;
1411 r->ru_nivcsw += p->signal->nivcsw;
1412 r->ru_minflt += p->signal->min_flt;
1413 r->ru_majflt += p->signal->maj_flt;
1414 r->ru_inblock += p->signal->inblock;
1415 r->ru_oublock += p->signal->oublock;
1416 if (maxrss < p->signal->maxrss)
1417 maxrss = p->signal->maxrss;
1420 accumulate_thread_rusage(t, r);
1428 unlock_task_sighand(p, &flags);
1431 cputime_to_timeval(utime, &r->ru_utime);
1432 cputime_to_timeval(stime, &r->ru_stime);
1434 if (who != RUSAGE_CHILDREN) {
1435 struct mm_struct *mm = get_task_mm(p);
1437 setmax_mm_hiwater_rss(&maxrss, mm);
1441 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1444 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1447 k_getrusage(p, who, &r);
1448 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1451 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1453 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1454 who != RUSAGE_THREAD)
1456 return getrusage(current, who, ru);
1459 SYSCALL_DEFINE1(umask, int, mask)
1461 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1465 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
1466 unsigned long, arg4, unsigned long, arg5)
1468 struct task_struct *me = current;
1469 unsigned char comm[sizeof(me->comm)];
1472 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1473 if (error != -ENOSYS)
1478 case PR_SET_PDEATHSIG:
1479 if (!valid_signal(arg2)) {
1483 me->pdeath_signal = arg2;
1486 case PR_GET_PDEATHSIG:
1487 error = put_user(me->pdeath_signal, (int __user *)arg2);
1489 case PR_GET_DUMPABLE:
1490 error = get_dumpable(me->mm);
1492 case PR_SET_DUMPABLE:
1493 if (arg2 < 0 || arg2 > 1) {
1497 set_dumpable(me->mm, arg2);
1501 case PR_SET_UNALIGN:
1502 error = SET_UNALIGN_CTL(me, arg2);
1504 case PR_GET_UNALIGN:
1505 error = GET_UNALIGN_CTL(me, arg2);
1508 error = SET_FPEMU_CTL(me, arg2);
1511 error = GET_FPEMU_CTL(me, arg2);
1514 error = SET_FPEXC_CTL(me, arg2);
1517 error = GET_FPEXC_CTL(me, arg2);
1520 error = PR_TIMING_STATISTICAL;
1523 if (arg2 != PR_TIMING_STATISTICAL)
1530 comm[sizeof(me->comm)-1] = 0;
1531 if (strncpy_from_user(comm, (char __user *)arg2,
1532 sizeof(me->comm) - 1) < 0)
1534 set_task_comm(me, comm);
1537 get_task_comm(comm, me);
1538 if (copy_to_user((char __user *)arg2, comm,
1543 error = GET_ENDIAN(me, arg2);
1546 error = SET_ENDIAN(me, arg2);
1549 case PR_GET_SECCOMP:
1550 error = prctl_get_seccomp();
1552 case PR_SET_SECCOMP:
1553 error = prctl_set_seccomp(arg2);
1556 error = GET_TSC_CTL(arg2);
1559 error = SET_TSC_CTL(arg2);
1561 case PR_TASK_PERF_EVENTS_DISABLE:
1562 error = perf_event_task_disable();
1564 case PR_TASK_PERF_EVENTS_ENABLE:
1565 error = perf_event_task_enable();
1567 case PR_GET_TIMERSLACK:
1568 error = current->timer_slack_ns;
1570 case PR_SET_TIMERSLACK:
1572 current->timer_slack_ns =
1573 current->default_timer_slack_ns;
1575 current->timer_slack_ns = arg2;
1582 case PR_MCE_KILL_CLEAR:
1585 current->flags &= ~PF_MCE_PROCESS;
1587 case PR_MCE_KILL_SET:
1588 current->flags |= PF_MCE_PROCESS;
1589 if (arg3 == PR_MCE_KILL_EARLY)
1590 current->flags |= PF_MCE_EARLY;
1591 else if (arg3 == PR_MCE_KILL_LATE)
1592 current->flags &= ~PF_MCE_EARLY;
1593 else if (arg3 == PR_MCE_KILL_DEFAULT)
1595 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
1604 case PR_MCE_KILL_GET:
1605 if (arg2 | arg3 | arg4 | arg5)
1607 if (current->flags & PF_MCE_PROCESS)
1608 error = (current->flags & PF_MCE_EARLY) ?
1609 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
1611 error = PR_MCE_KILL_DEFAULT;
1620 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
1621 struct getcpu_cache __user *, unused)
1624 int cpu = raw_smp_processor_id();
1626 err |= put_user(cpu, cpup);
1628 err |= put_user(cpu_to_node(cpu), nodep);
1629 return err ? -EFAULT : 0;
1632 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
1634 static void argv_cleanup(char **argv, char **envp)
1640 * orderly_poweroff - Trigger an orderly system poweroff
1641 * @force: force poweroff if command execution fails
1643 * This may be called from any context to trigger a system shutdown.
1644 * If the orderly shutdown fails, it will force an immediate shutdown.
1646 int orderly_poweroff(bool force)
1649 char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
1650 static char *envp[] = {
1652 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
1656 struct subprocess_info *info;
1659 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
1660 __func__, poweroff_cmd);
1664 info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
1670 call_usermodehelper_setcleanup(info, argv_cleanup);
1672 ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
1676 printk(KERN_WARNING "Failed to start orderly shutdown: "
1677 "forcing the issue\n");
1679 /* I guess this should try to kick off some daemon to
1680 sync and poweroff asap. Or not even bother syncing
1681 if we're doing an emergency shutdown? */
1688 EXPORT_SYMBOL_GPL(orderly_poweroff);