4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/module.h>
9 #include <linux/utsname.h>
10 #include <linux/mman.h>
11 #include <linux/notifier.h>
12 #include <linux/reboot.h>
13 #include <linux/prctl.h>
14 #include <linux/highuid.h>
16 #include <linux/perf_event.h>
17 #include <linux/resource.h>
18 #include <linux/kernel.h>
19 #include <linux/kexec.h>
20 #include <linux/workqueue.h>
21 #include <linux/capability.h>
22 #include <linux/device.h>
23 #include <linux/key.h>
24 #include <linux/times.h>
25 #include <linux/posix-timers.h>
26 #include <linux/security.h>
27 #include <linux/dcookies.h>
28 #include <linux/suspend.h>
29 #include <linux/tty.h>
30 #include <linux/signal.h>
31 #include <linux/cn_proc.h>
32 #include <linux/getcpu.h>
33 #include <linux/task_io_accounting_ops.h>
34 #include <linux/seccomp.h>
35 #include <linux/cpu.h>
36 #include <linux/personality.h>
37 #include <linux/ptrace.h>
38 #include <linux/fs_struct.h>
39 #include <linux/gfp.h>
40 #include <linux/syscore_ops.h>
41 #include <linux/version.h>
42 #include <linux/ctype.h>
44 #include <linux/compat.h>
45 #include <linux/syscalls.h>
46 #include <linux/kprobes.h>
47 #include <linux/user_namespace.h>
49 #include <linux/kmsg_dump.h>
50 /* Move somewhere else to avoid recompiling? */
51 #include <generated/utsrelease.h>
53 #include <asm/uaccess.h>
55 #include <asm/unistd.h>
57 #ifndef SET_UNALIGN_CTL
58 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
60 #ifndef GET_UNALIGN_CTL
61 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
64 # define SET_FPEMU_CTL(a,b) (-EINVAL)
67 # define GET_FPEMU_CTL(a,b) (-EINVAL)
70 # define SET_FPEXC_CTL(a,b) (-EINVAL)
73 # define GET_FPEXC_CTL(a,b) (-EINVAL)
76 # define GET_ENDIAN(a,b) (-EINVAL)
79 # define SET_ENDIAN(a,b) (-EINVAL)
82 # define GET_TSC_CTL(a) (-EINVAL)
85 # define SET_TSC_CTL(a) (-EINVAL)
89 * this is where the system-wide overflow UID and GID are defined, for
90 * architectures that now have 32-bit UID/GID but didn't in the past
93 int overflowuid = DEFAULT_OVERFLOWUID;
94 int overflowgid = DEFAULT_OVERFLOWGID;
97 EXPORT_SYMBOL(overflowuid);
98 EXPORT_SYMBOL(overflowgid);
102 * the same as above, but for filesystems which can only store a 16-bit
103 * UID and GID. as such, this is needed on all architectures
106 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
107 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
109 EXPORT_SYMBOL(fs_overflowuid);
110 EXPORT_SYMBOL(fs_overflowgid);
113 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
118 EXPORT_SYMBOL(cad_pid);
121 * If set, this is used for preparing the system to power off.
124 void (*pm_power_off_prepare)(void);
127 * Returns true if current's euid is same as p's uid or euid,
128 * or has CAP_SYS_NICE to p's user_ns.
130 * Called with rcu_read_lock, creds are safe
132 static bool set_one_prio_perm(struct task_struct *p)
134 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
136 if (pcred->user->user_ns == cred->user->user_ns &&
137 (pcred->uid == cred->euid ||
138 pcred->euid == cred->euid))
140 if (ns_capable(pcred->user->user_ns, CAP_SYS_NICE))
146 * set the priority of a task
147 * - the caller must hold the RCU read lock
149 static int set_one_prio(struct task_struct *p, int niceval, int error)
153 if (!set_one_prio_perm(p)) {
157 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
161 no_nice = security_task_setnice(p, niceval);
168 set_user_nice(p, niceval);
173 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
175 struct task_struct *g, *p;
176 struct user_struct *user;
177 const struct cred *cred = current_cred();
181 if (which > PRIO_USER || which < PRIO_PROCESS)
184 /* normalize: avoid signed division (rounding problems) */
192 read_lock(&tasklist_lock);
196 p = find_task_by_vpid(who);
200 error = set_one_prio(p, niceval, error);
204 pgrp = find_vpid(who);
206 pgrp = task_pgrp(current);
207 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
208 error = set_one_prio(p, niceval, error);
209 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
212 user = (struct user_struct *) cred->user;
215 else if ((who != cred->uid) &&
216 !(user = find_user(who)))
217 goto out_unlock; /* No processes for this user */
219 do_each_thread(g, p) {
220 if (__task_cred(p)->uid == who)
221 error = set_one_prio(p, niceval, error);
222 } while_each_thread(g, p);
223 if (who != cred->uid)
224 free_uid(user); /* For find_user() */
228 read_unlock(&tasklist_lock);
235 * Ugh. To avoid negative return values, "getpriority()" will
236 * not return the normal nice-value, but a negated value that
237 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
238 * to stay compatible.
240 SYSCALL_DEFINE2(getpriority, int, which, int, who)
242 struct task_struct *g, *p;
243 struct user_struct *user;
244 const struct cred *cred = current_cred();
245 long niceval, retval = -ESRCH;
248 if (which > PRIO_USER || which < PRIO_PROCESS)
252 read_lock(&tasklist_lock);
256 p = find_task_by_vpid(who);
260 niceval = 20 - task_nice(p);
261 if (niceval > retval)
267 pgrp = find_vpid(who);
269 pgrp = task_pgrp(current);
270 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
271 niceval = 20 - task_nice(p);
272 if (niceval > retval)
274 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
277 user = (struct user_struct *) cred->user;
280 else if ((who != cred->uid) &&
281 !(user = find_user(who)))
282 goto out_unlock; /* No processes for this user */
284 do_each_thread(g, p) {
285 if (__task_cred(p)->uid == who) {
286 niceval = 20 - task_nice(p);
287 if (niceval > retval)
290 } while_each_thread(g, p);
291 if (who != cred->uid)
292 free_uid(user); /* for find_user() */
296 read_unlock(&tasklist_lock);
303 * emergency_restart - reboot the system
305 * Without shutting down any hardware or taking any locks
306 * reboot the system. This is called when we know we are in
307 * trouble so this is our best effort to reboot. This is
308 * safe to call in interrupt context.
310 void emergency_restart(void)
312 kmsg_dump(KMSG_DUMP_EMERG);
313 machine_emergency_restart();
315 EXPORT_SYMBOL_GPL(emergency_restart);
317 void kernel_restart_prepare(char *cmd)
319 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
320 system_state = SYSTEM_RESTART;
321 usermodehelper_disable();
327 * kernel_restart - reboot the system
328 * @cmd: pointer to buffer containing command to execute for restart
331 * Shutdown everything and perform a clean reboot.
332 * This is not safe to call in interrupt context.
334 void kernel_restart(char *cmd)
336 kernel_restart_prepare(cmd);
337 if (pm_power_off_prepare)
338 pm_power_off_prepare();
339 disable_nonboot_cpus();
342 printk(KERN_EMERG "Restarting system.\n");
344 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
345 printk(KERN_EMERG "pid = %d name:%s\n", task_tgid_vnr(current), current->comm);
347 kmsg_dump(KMSG_DUMP_RESTART);
348 machine_restart(cmd);
350 EXPORT_SYMBOL_GPL(kernel_restart);
352 static void kernel_shutdown_prepare(enum system_states state)
354 blocking_notifier_call_chain(&reboot_notifier_list,
355 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
356 system_state = state;
357 usermodehelper_disable();
361 * kernel_halt - halt the system
363 * Shutdown everything and perform a clean system halt.
365 void kernel_halt(void)
367 kernel_shutdown_prepare(SYSTEM_HALT);
369 printk(KERN_EMERG "System halted.\n");
370 kmsg_dump(KMSG_DUMP_HALT);
374 EXPORT_SYMBOL_GPL(kernel_halt);
377 * kernel_power_off - power_off the system
379 * Shutdown everything and perform a clean system power_off.
381 void kernel_power_off(void)
383 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
384 if (pm_power_off_prepare)
385 pm_power_off_prepare();
386 disable_nonboot_cpus();
388 printk(KERN_EMERG "Power down.\n");
389 kmsg_dump(KMSG_DUMP_POWEROFF);
392 EXPORT_SYMBOL_GPL(kernel_power_off);
394 static DEFINE_MUTEX(reboot_mutex);
397 * Reboot system call: for obvious reasons only root may call it,
398 * and even root needs to set up some magic numbers in the registers
399 * so that some mistake won't make this reboot the whole machine.
400 * You can also set the meaning of the ctrl-alt-del-key here.
402 * reboot doesn't sync: do that yourself before calling this.
404 SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
410 /* We only trust the superuser with rebooting the system. */
411 if (!capable(CAP_SYS_BOOT))
414 /* For safety, we require "magic" arguments. */
415 if (magic1 != LINUX_REBOOT_MAGIC1 ||
416 (magic2 != LINUX_REBOOT_MAGIC2 &&
417 magic2 != LINUX_REBOOT_MAGIC2A &&
418 magic2 != LINUX_REBOOT_MAGIC2B &&
419 magic2 != LINUX_REBOOT_MAGIC2C))
422 /* Instead of trying to make the power_off code look like
423 * halt when pm_power_off is not set do it the easy way.
425 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
426 cmd = LINUX_REBOOT_CMD_HALT;
428 mutex_lock(&reboot_mutex);
430 case LINUX_REBOOT_CMD_RESTART:
431 kernel_restart(NULL);
434 case LINUX_REBOOT_CMD_CAD_ON:
438 case LINUX_REBOOT_CMD_CAD_OFF:
442 case LINUX_REBOOT_CMD_HALT:
445 panic("cannot halt");
447 case LINUX_REBOOT_CMD_POWER_OFF:
452 case LINUX_REBOOT_CMD_RESTART2:
453 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
457 buffer[sizeof(buffer) - 1] = '\0';
459 kernel_restart(buffer);
463 case LINUX_REBOOT_CMD_KEXEC:
464 ret = kernel_kexec();
468 #ifdef CONFIG_HIBERNATION
469 case LINUX_REBOOT_CMD_SW_SUSPEND:
478 mutex_unlock(&reboot_mutex);
482 static void deferred_cad(struct work_struct *dummy)
484 kernel_restart(NULL);
488 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
489 * As it's called within an interrupt, it may NOT sync: the only choice
490 * is whether to reboot at once, or just ignore the ctrl-alt-del.
492 void ctrl_alt_del(void)
494 static DECLARE_WORK(cad_work, deferred_cad);
497 schedule_work(&cad_work);
499 kill_cad_pid(SIGINT, 1);
503 * Unprivileged users may change the real gid to the effective gid
504 * or vice versa. (BSD-style)
506 * If you set the real gid at all, or set the effective gid to a value not
507 * equal to the real gid, then the saved gid is set to the new effective gid.
509 * This makes it possible for a setgid program to completely drop its
510 * privileges, which is often a useful assertion to make when you are doing
511 * a security audit over a program.
513 * The general idea is that a program which uses just setregid() will be
514 * 100% compatible with BSD. A program which uses just setgid() will be
515 * 100% compatible with POSIX with saved IDs.
517 * SMP: There are not races, the GIDs are checked only by filesystem
518 * operations (as far as semantic preservation is concerned).
520 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
522 const struct cred *old;
526 new = prepare_creds();
529 old = current_cred();
532 if (rgid != (gid_t) -1) {
533 if (old->gid == rgid ||
535 nsown_capable(CAP_SETGID))
540 if (egid != (gid_t) -1) {
541 if (old->gid == egid ||
544 nsown_capable(CAP_SETGID))
550 if (rgid != (gid_t) -1 ||
551 (egid != (gid_t) -1 && egid != old->gid))
552 new->sgid = new->egid;
553 new->fsgid = new->egid;
555 return commit_creds(new);
563 * setgid() is implemented like SysV w/ SAVED_IDS
565 * SMP: Same implicit races as above.
567 SYSCALL_DEFINE1(setgid, gid_t, gid)
569 const struct cred *old;
573 new = prepare_creds();
576 old = current_cred();
579 if (nsown_capable(CAP_SETGID))
580 new->gid = new->egid = new->sgid = new->fsgid = gid;
581 else if (gid == old->gid || gid == old->sgid)
582 new->egid = new->fsgid = gid;
586 return commit_creds(new);
594 * change the user struct in a credentials set to match the new UID
596 static int set_user(struct cred *new)
598 struct user_struct *new_user;
600 new_user = alloc_uid(current_user_ns(), new->uid);
604 if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
605 new_user != INIT_USER) {
611 new->user = new_user;
616 * Unprivileged users may change the real uid to the effective uid
617 * or vice versa. (BSD-style)
619 * If you set the real uid at all, or set the effective uid to a value not
620 * equal to the real uid, then the saved uid is set to the new effective uid.
622 * This makes it possible for a setuid program to completely drop its
623 * privileges, which is often a useful assertion to make when you are doing
624 * a security audit over a program.
626 * The general idea is that a program which uses just setreuid() will be
627 * 100% compatible with BSD. A program which uses just setuid() will be
628 * 100% compatible with POSIX with saved IDs.
630 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
632 const struct cred *old;
636 new = prepare_creds();
639 old = current_cred();
642 if (ruid != (uid_t) -1) {
644 if (old->uid != ruid &&
646 !nsown_capable(CAP_SETUID))
650 if (euid != (uid_t) -1) {
652 if (old->uid != euid &&
655 !nsown_capable(CAP_SETUID))
659 if (new->uid != old->uid) {
660 retval = set_user(new);
664 if (ruid != (uid_t) -1 ||
665 (euid != (uid_t) -1 && euid != old->uid))
666 new->suid = new->euid;
667 new->fsuid = new->euid;
669 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
673 return commit_creds(new);
681 * setuid() is implemented like SysV with SAVED_IDS
683 * Note that SAVED_ID's is deficient in that a setuid root program
684 * like sendmail, for example, cannot set its uid to be a normal
685 * user and then switch back, because if you're root, setuid() sets
686 * the saved uid too. If you don't like this, blame the bright people
687 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
688 * will allow a root program to temporarily drop privileges and be able to
689 * regain them by swapping the real and effective uid.
691 SYSCALL_DEFINE1(setuid, uid_t, uid)
693 const struct cred *old;
697 new = prepare_creds();
700 old = current_cred();
703 if (nsown_capable(CAP_SETUID)) {
704 new->suid = new->uid = uid;
705 if (uid != old->uid) {
706 retval = set_user(new);
710 } else if (uid != old->uid && uid != new->suid) {
714 new->fsuid = new->euid = uid;
716 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
720 return commit_creds(new);
729 * This function implements a generic ability to update ruid, euid,
730 * and suid. This allows you to implement the 4.4 compatible seteuid().
732 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
734 const struct cred *old;
738 new = prepare_creds();
742 old = current_cred();
745 if (!nsown_capable(CAP_SETUID)) {
746 if (ruid != (uid_t) -1 && ruid != old->uid &&
747 ruid != old->euid && ruid != old->suid)
749 if (euid != (uid_t) -1 && euid != old->uid &&
750 euid != old->euid && euid != old->suid)
752 if (suid != (uid_t) -1 && suid != old->uid &&
753 suid != old->euid && suid != old->suid)
757 if (ruid != (uid_t) -1) {
759 if (ruid != old->uid) {
760 retval = set_user(new);
765 if (euid != (uid_t) -1)
767 if (suid != (uid_t) -1)
769 new->fsuid = new->euid;
771 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
775 return commit_creds(new);
782 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruid, uid_t __user *, euid, uid_t __user *, suid)
784 const struct cred *cred = current_cred();
787 if (!(retval = put_user(cred->uid, ruid)) &&
788 !(retval = put_user(cred->euid, euid)))
789 retval = put_user(cred->suid, suid);
795 * Same as above, but for rgid, egid, sgid.
797 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
799 const struct cred *old;
803 new = prepare_creds();
806 old = current_cred();
809 if (!nsown_capable(CAP_SETGID)) {
810 if (rgid != (gid_t) -1 && rgid != old->gid &&
811 rgid != old->egid && rgid != old->sgid)
813 if (egid != (gid_t) -1 && egid != old->gid &&
814 egid != old->egid && egid != old->sgid)
816 if (sgid != (gid_t) -1 && sgid != old->gid &&
817 sgid != old->egid && sgid != old->sgid)
821 if (rgid != (gid_t) -1)
823 if (egid != (gid_t) -1)
825 if (sgid != (gid_t) -1)
827 new->fsgid = new->egid;
829 return commit_creds(new);
836 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgid, gid_t __user *, egid, gid_t __user *, sgid)
838 const struct cred *cred = current_cred();
841 if (!(retval = put_user(cred->gid, rgid)) &&
842 !(retval = put_user(cred->egid, egid)))
843 retval = put_user(cred->sgid, sgid);
850 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
851 * is used for "access()" and for the NFS daemon (letting nfsd stay at
852 * whatever uid it wants to). It normally shadows "euid", except when
853 * explicitly set by setfsuid() or for access..
855 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
857 const struct cred *old;
861 new = prepare_creds();
863 return current_fsuid();
864 old = current_cred();
865 old_fsuid = old->fsuid;
867 if (uid == old->uid || uid == old->euid ||
868 uid == old->suid || uid == old->fsuid ||
869 nsown_capable(CAP_SETUID)) {
870 if (uid != old_fsuid) {
872 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
886 * Samma på svenska..
888 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
890 const struct cred *old;
894 new = prepare_creds();
896 return current_fsgid();
897 old = current_cred();
898 old_fsgid = old->fsgid;
900 if (gid == old->gid || gid == old->egid ||
901 gid == old->sgid || gid == old->fsgid ||
902 nsown_capable(CAP_SETGID)) {
903 if (gid != old_fsgid) {
917 void do_sys_times(struct tms *tms)
919 cputime_t tgutime, tgstime, cutime, cstime;
921 spin_lock_irq(¤t->sighand->siglock);
922 thread_group_times(current, &tgutime, &tgstime);
923 cutime = current->signal->cutime;
924 cstime = current->signal->cstime;
925 spin_unlock_irq(¤t->sighand->siglock);
926 tms->tms_utime = cputime_to_clock_t(tgutime);
927 tms->tms_stime = cputime_to_clock_t(tgstime);
928 tms->tms_cutime = cputime_to_clock_t(cutime);
929 tms->tms_cstime = cputime_to_clock_t(cstime);
932 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
938 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
941 force_successful_syscall_return();
942 return (long) jiffies_64_to_clock_t(get_jiffies_64());
946 * This needs some heavy checking ...
947 * I just haven't the stomach for it. I also don't fully
948 * understand sessions/pgrp etc. Let somebody who does explain it.
950 * OK, I think I have the protection semantics right.... this is really
951 * only important on a multi-user system anyway, to make sure one user
952 * can't send a signal to a process owned by another. -TYT, 12/12/91
954 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
957 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
959 struct task_struct *p;
960 struct task_struct *group_leader = current->group_leader;
965 pid = task_pid_vnr(group_leader);
972 /* From this point forward we keep holding onto the tasklist lock
973 * so that our parent does not change from under us. -DaveM
975 write_lock_irq(&tasklist_lock);
978 p = find_task_by_vpid(pid);
983 if (!thread_group_leader(p))
986 if (same_thread_group(p->real_parent, group_leader)) {
988 if (task_session(p) != task_session(group_leader))
995 if (p != group_leader)
1000 if (p->signal->leader)
1005 struct task_struct *g;
1007 pgrp = find_vpid(pgid);
1008 g = pid_task(pgrp, PIDTYPE_PGID);
1009 if (!g || task_session(g) != task_session(group_leader))
1013 err = security_task_setpgid(p, pgid);
1017 if (task_pgrp(p) != pgrp)
1018 change_pid(p, PIDTYPE_PGID, pgrp);
1022 /* All paths lead to here, thus we are safe. -DaveM */
1023 write_unlock_irq(&tasklist_lock);
1028 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1030 struct task_struct *p;
1036 grp = task_pgrp(current);
1039 p = find_task_by_vpid(pid);
1046 retval = security_task_getpgid(p);
1050 retval = pid_vnr(grp);
1056 #ifdef __ARCH_WANT_SYS_GETPGRP
1058 SYSCALL_DEFINE0(getpgrp)
1060 return sys_getpgid(0);
1065 SYSCALL_DEFINE1(getsid, pid_t, pid)
1067 struct task_struct *p;
1073 sid = task_session(current);
1076 p = find_task_by_vpid(pid);
1079 sid = task_session(p);
1083 retval = security_task_getsid(p);
1087 retval = pid_vnr(sid);
1093 SYSCALL_DEFINE0(setsid)
1095 struct task_struct *group_leader = current->group_leader;
1096 struct pid *sid = task_pid(group_leader);
1097 pid_t session = pid_vnr(sid);
1100 write_lock_irq(&tasklist_lock);
1101 /* Fail if I am already a session leader */
1102 if (group_leader->signal->leader)
1105 /* Fail if a process group id already exists that equals the
1106 * proposed session id.
1108 if (pid_task(sid, PIDTYPE_PGID))
1111 group_leader->signal->leader = 1;
1112 __set_special_pids(sid);
1114 proc_clear_tty(group_leader);
1118 write_unlock_irq(&tasklist_lock);
1120 proc_sid_connector(group_leader);
1121 sched_autogroup_create_attach(group_leader);
1126 DECLARE_RWSEM(uts_sem);
1128 #ifdef COMPAT_UTS_MACHINE
1129 #define override_architecture(name) \
1130 (personality(current->personality) == PER_LINUX32 && \
1131 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1132 sizeof(COMPAT_UTS_MACHINE)))
1134 #define override_architecture(name) 0
1138 * Work around broken programs that cannot handle "Linux 3.0".
1139 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1141 static int override_release(char __user *release, int len)
1146 if (current->personality & UNAME26) {
1147 char *rest = UTS_RELEASE;
1152 if (*rest == '.' && ++ndots >= 3)
1154 if (!isdigit(*rest) && *rest != '.')
1158 v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 40;
1159 snprintf(buf, len, "2.6.%u%s", v, rest);
1160 ret = copy_to_user(release, buf, len);
1165 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1169 down_read(&uts_sem);
1170 if (copy_to_user(name, utsname(), sizeof *name))
1174 if (!errno && override_release(name->release, sizeof(name->release)))
1176 if (!errno && override_architecture(name))
1181 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1185 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1192 down_read(&uts_sem);
1193 if (copy_to_user(name, utsname(), sizeof(*name)))
1197 if (!error && override_release(name->release, sizeof(name->release)))
1199 if (!error && override_architecture(name))
1204 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1210 if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
1213 down_read(&uts_sem);
1214 error = __copy_to_user(&name->sysname, &utsname()->sysname,
1216 error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
1217 error |= __copy_to_user(&name->nodename, &utsname()->nodename,
1219 error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
1220 error |= __copy_to_user(&name->release, &utsname()->release,
1222 error |= __put_user(0, name->release + __OLD_UTS_LEN);
1223 error |= __copy_to_user(&name->version, &utsname()->version,
1225 error |= __put_user(0, name->version + __OLD_UTS_LEN);
1226 error |= __copy_to_user(&name->machine, &utsname()->machine,
1228 error |= __put_user(0, name->machine + __OLD_UTS_LEN);
1231 if (!error && override_architecture(name))
1233 if (!error && override_release(name->release, sizeof(name->release)))
1235 return error ? -EFAULT : 0;
1239 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1242 char tmp[__NEW_UTS_LEN];
1244 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1247 if (len < 0 || len > __NEW_UTS_LEN)
1249 down_write(&uts_sem);
1251 if (!copy_from_user(tmp, name, len)) {
1252 struct new_utsname *u = utsname();
1254 memcpy(u->nodename, tmp, len);
1255 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1262 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1264 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1267 struct new_utsname *u;
1271 down_read(&uts_sem);
1273 i = 1 + strlen(u->nodename);
1277 if (copy_to_user(name, u->nodename, i))
1286 * Only setdomainname; getdomainname can be implemented by calling
1289 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1292 char tmp[__NEW_UTS_LEN];
1294 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1296 if (len < 0 || len > __NEW_UTS_LEN)
1299 down_write(&uts_sem);
1301 if (!copy_from_user(tmp, name, len)) {
1302 struct new_utsname *u = utsname();
1304 memcpy(u->domainname, tmp, len);
1305 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1312 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1314 struct rlimit value;
1317 ret = do_prlimit(current, resource, NULL, &value);
1319 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1324 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1327 * Back compatibility for getrlimit. Needed for some apps.
1330 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1331 struct rlimit __user *, rlim)
1334 if (resource >= RLIM_NLIMITS)
1337 task_lock(current->group_leader);
1338 x = current->signal->rlim[resource];
1339 task_unlock(current->group_leader);
1340 if (x.rlim_cur > 0x7FFFFFFF)
1341 x.rlim_cur = 0x7FFFFFFF;
1342 if (x.rlim_max > 0x7FFFFFFF)
1343 x.rlim_max = 0x7FFFFFFF;
1344 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1349 static inline bool rlim64_is_infinity(__u64 rlim64)
1351 #if BITS_PER_LONG < 64
1352 return rlim64 >= ULONG_MAX;
1354 return rlim64 == RLIM64_INFINITY;
1358 static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1360 if (rlim->rlim_cur == RLIM_INFINITY)
1361 rlim64->rlim_cur = RLIM64_INFINITY;
1363 rlim64->rlim_cur = rlim->rlim_cur;
1364 if (rlim->rlim_max == RLIM_INFINITY)
1365 rlim64->rlim_max = RLIM64_INFINITY;
1367 rlim64->rlim_max = rlim->rlim_max;
1370 static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1372 if (rlim64_is_infinity(rlim64->rlim_cur))
1373 rlim->rlim_cur = RLIM_INFINITY;
1375 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1376 if (rlim64_is_infinity(rlim64->rlim_max))
1377 rlim->rlim_max = RLIM_INFINITY;
1379 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1382 /* make sure you are allowed to change @tsk limits before calling this */
1383 int do_prlimit(struct task_struct *tsk, unsigned int resource,
1384 struct rlimit *new_rlim, struct rlimit *old_rlim)
1386 struct rlimit *rlim;
1389 if (resource >= RLIM_NLIMITS)
1392 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1394 if (resource == RLIMIT_NOFILE &&
1395 new_rlim->rlim_max > sysctl_nr_open)
1399 /* protect tsk->signal and tsk->sighand from disappearing */
1400 read_lock(&tasklist_lock);
1401 if (!tsk->sighand) {
1406 rlim = tsk->signal->rlim + resource;
1407 task_lock(tsk->group_leader);
1409 /* Keep the capable check against init_user_ns until
1410 cgroups can contain all limits */
1411 if (new_rlim->rlim_max > rlim->rlim_max &&
1412 !capable(CAP_SYS_RESOURCE))
1415 retval = security_task_setrlimit(tsk->group_leader,
1416 resource, new_rlim);
1417 if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) {
1419 * The caller is asking for an immediate RLIMIT_CPU
1420 * expiry. But we use the zero value to mean "it was
1421 * never set". So let's cheat and make it one second
1424 new_rlim->rlim_cur = 1;
1433 task_unlock(tsk->group_leader);
1436 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1437 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1438 * very long-standing error, and fixing it now risks breakage of
1439 * applications, so we live with it
1441 if (!retval && new_rlim && resource == RLIMIT_CPU &&
1442 new_rlim->rlim_cur != RLIM_INFINITY)
1443 update_rlimit_cpu(tsk, new_rlim->rlim_cur);
1445 read_unlock(&tasklist_lock);
1449 /* rcu lock must be held */
1450 static int check_prlimit_permission(struct task_struct *task)
1452 const struct cred *cred = current_cred(), *tcred;
1454 if (current == task)
1457 tcred = __task_cred(task);
1458 if (cred->user->user_ns == tcred->user->user_ns &&
1459 (cred->uid == tcred->euid &&
1460 cred->uid == tcred->suid &&
1461 cred->uid == tcred->uid &&
1462 cred->gid == tcred->egid &&
1463 cred->gid == tcred->sgid &&
1464 cred->gid == tcred->gid))
1466 if (ns_capable(tcred->user->user_ns, CAP_SYS_RESOURCE))
1472 SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1473 const struct rlimit64 __user *, new_rlim,
1474 struct rlimit64 __user *, old_rlim)
1476 struct rlimit64 old64, new64;
1477 struct rlimit old, new;
1478 struct task_struct *tsk;
1482 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1484 rlim64_to_rlim(&new64, &new);
1488 tsk = pid ? find_task_by_vpid(pid) : current;
1493 ret = check_prlimit_permission(tsk);
1498 get_task_struct(tsk);
1501 ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1502 old_rlim ? &old : NULL);
1504 if (!ret && old_rlim) {
1505 rlim_to_rlim64(&old, &old64);
1506 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1510 put_task_struct(tsk);
1514 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1516 struct rlimit new_rlim;
1518 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1520 return do_prlimit(current, resource, &new_rlim, NULL);
1524 * It would make sense to put struct rusage in the task_struct,
1525 * except that would make the task_struct be *really big*. After
1526 * task_struct gets moved into malloc'ed memory, it would
1527 * make sense to do this. It will make moving the rest of the information
1528 * a lot simpler! (Which we're not doing right now because we're not
1529 * measuring them yet).
1531 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1532 * races with threads incrementing their own counters. But since word
1533 * reads are atomic, we either get new values or old values and we don't
1534 * care which for the sums. We always take the siglock to protect reading
1535 * the c* fields from p->signal from races with exit.c updating those
1536 * fields when reaping, so a sample either gets all the additions of a
1537 * given child after it's reaped, or none so this sample is before reaping.
1540 * We need to take the siglock for CHILDEREN, SELF and BOTH
1541 * for the cases current multithreaded, non-current single threaded
1542 * non-current multithreaded. Thread traversal is now safe with
1544 * Strictly speaking, we donot need to take the siglock if we are current and
1545 * single threaded, as no one else can take our signal_struct away, no one
1546 * else can reap the children to update signal->c* counters, and no one else
1547 * can race with the signal-> fields. If we do not take any lock, the
1548 * signal-> fields could be read out of order while another thread was just
1549 * exiting. So we should place a read memory barrier when we avoid the lock.
1550 * On the writer side, write memory barrier is implied in __exit_signal
1551 * as __exit_signal releases the siglock spinlock after updating the signal->
1552 * fields. But we don't do this yet to keep things simple.
1556 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1558 r->ru_nvcsw += t->nvcsw;
1559 r->ru_nivcsw += t->nivcsw;
1560 r->ru_minflt += t->min_flt;
1561 r->ru_majflt += t->maj_flt;
1562 r->ru_inblock += task_io_get_inblock(t);
1563 r->ru_oublock += task_io_get_oublock(t);
1566 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1568 struct task_struct *t;
1569 unsigned long flags;
1570 cputime_t tgutime, tgstime, utime, stime;
1571 unsigned long maxrss = 0;
1573 memset((char *) r, 0, sizeof *r);
1574 utime = stime = cputime_zero;
1576 if (who == RUSAGE_THREAD) {
1577 task_times(current, &utime, &stime);
1578 accumulate_thread_rusage(p, r);
1579 maxrss = p->signal->maxrss;
1583 if (!lock_task_sighand(p, &flags))
1588 case RUSAGE_CHILDREN:
1589 utime = p->signal->cutime;
1590 stime = p->signal->cstime;
1591 r->ru_nvcsw = p->signal->cnvcsw;
1592 r->ru_nivcsw = p->signal->cnivcsw;
1593 r->ru_minflt = p->signal->cmin_flt;
1594 r->ru_majflt = p->signal->cmaj_flt;
1595 r->ru_inblock = p->signal->cinblock;
1596 r->ru_oublock = p->signal->coublock;
1597 maxrss = p->signal->cmaxrss;
1599 if (who == RUSAGE_CHILDREN)
1603 thread_group_times(p, &tgutime, &tgstime);
1604 utime = cputime_add(utime, tgutime);
1605 stime = cputime_add(stime, tgstime);
1606 r->ru_nvcsw += p->signal->nvcsw;
1607 r->ru_nivcsw += p->signal->nivcsw;
1608 r->ru_minflt += p->signal->min_flt;
1609 r->ru_majflt += p->signal->maj_flt;
1610 r->ru_inblock += p->signal->inblock;
1611 r->ru_oublock += p->signal->oublock;
1612 if (maxrss < p->signal->maxrss)
1613 maxrss = p->signal->maxrss;
1616 accumulate_thread_rusage(t, r);
1624 unlock_task_sighand(p, &flags);
1627 cputime_to_timeval(utime, &r->ru_utime);
1628 cputime_to_timeval(stime, &r->ru_stime);
1630 if (who != RUSAGE_CHILDREN) {
1631 struct mm_struct *mm = get_task_mm(p);
1633 setmax_mm_hiwater_rss(&maxrss, mm);
1637 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1640 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1643 k_getrusage(p, who, &r);
1644 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1647 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1649 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1650 who != RUSAGE_THREAD)
1652 return getrusage(current, who, ru);
1655 SYSCALL_DEFINE1(umask, int, mask)
1657 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1661 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
1662 unsigned long, arg4, unsigned long, arg5)
1664 struct task_struct *me = current;
1665 unsigned char comm[sizeof(me->comm)];
1668 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1669 if (error != -ENOSYS)
1674 case PR_SET_PDEATHSIG:
1675 if (!valid_signal(arg2)) {
1679 me->pdeath_signal = arg2;
1682 case PR_GET_PDEATHSIG:
1683 error = put_user(me->pdeath_signal, (int __user *)arg2);
1685 case PR_GET_DUMPABLE:
1686 error = get_dumpable(me->mm);
1688 case PR_SET_DUMPABLE:
1689 if (arg2 < 0 || arg2 > 1) {
1693 set_dumpable(me->mm, arg2);
1697 case PR_SET_UNALIGN:
1698 error = SET_UNALIGN_CTL(me, arg2);
1700 case PR_GET_UNALIGN:
1701 error = GET_UNALIGN_CTL(me, arg2);
1704 error = SET_FPEMU_CTL(me, arg2);
1707 error = GET_FPEMU_CTL(me, arg2);
1710 error = SET_FPEXC_CTL(me, arg2);
1713 error = GET_FPEXC_CTL(me, arg2);
1716 error = PR_TIMING_STATISTICAL;
1719 if (arg2 != PR_TIMING_STATISTICAL)
1726 comm[sizeof(me->comm)-1] = 0;
1727 if (strncpy_from_user(comm, (char __user *)arg2,
1728 sizeof(me->comm) - 1) < 0)
1730 set_task_comm(me, comm);
1733 get_task_comm(comm, me);
1734 if (copy_to_user((char __user *)arg2, comm,
1739 error = GET_ENDIAN(me, arg2);
1742 error = SET_ENDIAN(me, arg2);
1745 case PR_GET_SECCOMP:
1746 error = prctl_get_seccomp();
1748 case PR_SET_SECCOMP:
1749 error = prctl_set_seccomp(arg2);
1752 error = GET_TSC_CTL(arg2);
1755 error = SET_TSC_CTL(arg2);
1757 case PR_TASK_PERF_EVENTS_DISABLE:
1758 error = perf_event_task_disable();
1760 case PR_TASK_PERF_EVENTS_ENABLE:
1761 error = perf_event_task_enable();
1763 case PR_GET_TIMERSLACK:
1764 error = current->timer_slack_ns;
1766 case PR_SET_TIMERSLACK:
1768 current->timer_slack_ns =
1769 current->default_timer_slack_ns;
1771 current->timer_slack_ns = arg2;
1778 case PR_MCE_KILL_CLEAR:
1781 current->flags &= ~PF_MCE_PROCESS;
1783 case PR_MCE_KILL_SET:
1784 current->flags |= PF_MCE_PROCESS;
1785 if (arg3 == PR_MCE_KILL_EARLY)
1786 current->flags |= PF_MCE_EARLY;
1787 else if (arg3 == PR_MCE_KILL_LATE)
1788 current->flags &= ~PF_MCE_EARLY;
1789 else if (arg3 == PR_MCE_KILL_DEFAULT)
1791 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
1800 case PR_MCE_KILL_GET:
1801 if (arg2 | arg3 | arg4 | arg5)
1803 if (current->flags & PF_MCE_PROCESS)
1804 error = (current->flags & PF_MCE_EARLY) ?
1805 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
1807 error = PR_MCE_KILL_DEFAULT;
1816 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
1817 struct getcpu_cache __user *, unused)
1820 int cpu = raw_smp_processor_id();
1822 err |= put_user(cpu, cpup);
1824 err |= put_user(cpu_to_node(cpu), nodep);
1825 return err ? -EFAULT : 0;
1828 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
1830 static void argv_cleanup(struct subprocess_info *info)
1832 argv_free(info->argv);
1836 * orderly_poweroff - Trigger an orderly system poweroff
1837 * @force: force poweroff if command execution fails
1839 * This may be called from any context to trigger a system shutdown.
1840 * If the orderly shutdown fails, it will force an immediate shutdown.
1842 int orderly_poweroff(bool force)
1845 char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
1846 static char *envp[] = {
1848 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
1852 struct subprocess_info *info;
1855 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
1856 __func__, poweroff_cmd);
1860 info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
1866 call_usermodehelper_setfns(info, NULL, argv_cleanup, NULL);
1868 ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
1872 printk(KERN_WARNING "Failed to start orderly shutdown: "
1873 "forcing the issue\n");
1875 /* I guess this should try to kick off some daemon to
1876 sync and poweroff asap. Or not even bother syncing
1877 if we're doing an emergency shutdown? */
1884 EXPORT_SYMBOL_GPL(orderly_poweroff);