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/smp_lock.h>
12 #include <linux/notifier.h>
13 #include <linux/reboot.h>
14 #include <linux/prctl.h>
15 #include <linux/highuid.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>
37 #include <linux/compat.h>
38 #include <linux/syscalls.h>
39 #include <linux/kprobes.h>
40 #include <linux/user_namespace.h>
42 #include <asm/uaccess.h>
44 #include <asm/unistd.h>
46 #ifndef SET_UNALIGN_CTL
47 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
49 #ifndef GET_UNALIGN_CTL
50 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
53 # define SET_FPEMU_CTL(a,b) (-EINVAL)
56 # define GET_FPEMU_CTL(a,b) (-EINVAL)
59 # define SET_FPEXC_CTL(a,b) (-EINVAL)
62 # define GET_FPEXC_CTL(a,b) (-EINVAL)
65 # define GET_ENDIAN(a,b) (-EINVAL)
68 # define SET_ENDIAN(a,b) (-EINVAL)
71 # define GET_TSC_CTL(a) (-EINVAL)
74 # define SET_TSC_CTL(a) (-EINVAL)
78 * this is where the system-wide overflow UID and GID are defined, for
79 * architectures that now have 32-bit UID/GID but didn't in the past
82 int overflowuid = DEFAULT_OVERFLOWUID;
83 int overflowgid = DEFAULT_OVERFLOWGID;
86 EXPORT_SYMBOL(overflowuid);
87 EXPORT_SYMBOL(overflowgid);
91 * the same as above, but for filesystems which can only store a 16-bit
92 * UID and GID. as such, this is needed on all architectures
95 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
96 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
98 EXPORT_SYMBOL(fs_overflowuid);
99 EXPORT_SYMBOL(fs_overflowgid);
102 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
107 EXPORT_SYMBOL(cad_pid);
110 * If set, this is used for preparing the system to power off.
113 void (*pm_power_off_prepare)(void);
115 static int set_one_prio(struct task_struct *p, int niceval, int error)
119 if (p->uid != current->euid &&
120 p->euid != current->euid && !capable(CAP_SYS_NICE)) {
124 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
128 no_nice = security_task_setnice(p, niceval);
135 set_user_nice(p, niceval);
140 asmlinkage long sys_setpriority(int which, int who, int niceval)
142 struct task_struct *g, *p;
143 struct user_struct *user;
147 if (which > PRIO_USER || which < PRIO_PROCESS)
150 /* normalize: avoid signed division (rounding problems) */
157 read_lock(&tasklist_lock);
161 p = find_task_by_vpid(who);
165 error = set_one_prio(p, niceval, error);
169 pgrp = find_vpid(who);
171 pgrp = task_pgrp(current);
172 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
173 error = set_one_prio(p, niceval, error);
174 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
177 user = current->user;
181 if ((who != current->uid) && !(user = find_user(who)))
182 goto out_unlock; /* No processes for this user */
186 error = set_one_prio(p, niceval, error);
187 while_each_thread(g, p);
188 if (who != current->uid)
189 free_uid(user); /* For find_user() */
193 read_unlock(&tasklist_lock);
199 * Ugh. To avoid negative return values, "getpriority()" will
200 * not return the normal nice-value, but a negated value that
201 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
202 * to stay compatible.
204 asmlinkage long sys_getpriority(int which, int who)
206 struct task_struct *g, *p;
207 struct user_struct *user;
208 long niceval, retval = -ESRCH;
211 if (which > PRIO_USER || which < PRIO_PROCESS)
214 read_lock(&tasklist_lock);
218 p = find_task_by_vpid(who);
222 niceval = 20 - task_nice(p);
223 if (niceval > retval)
229 pgrp = find_vpid(who);
231 pgrp = task_pgrp(current);
232 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
233 niceval = 20 - task_nice(p);
234 if (niceval > retval)
236 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
239 user = current->user;
243 if ((who != current->uid) && !(user = find_user(who)))
244 goto out_unlock; /* No processes for this user */
248 niceval = 20 - task_nice(p);
249 if (niceval > retval)
252 while_each_thread(g, p);
253 if (who != current->uid)
254 free_uid(user); /* for find_user() */
258 read_unlock(&tasklist_lock);
264 * emergency_restart - reboot the system
266 * Without shutting down any hardware or taking any locks
267 * reboot the system. This is called when we know we are in
268 * trouble so this is our best effort to reboot. This is
269 * safe to call in interrupt context.
271 void emergency_restart(void)
273 machine_emergency_restart();
275 EXPORT_SYMBOL_GPL(emergency_restart);
277 void kernel_restart_prepare(char *cmd)
279 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
280 system_state = SYSTEM_RESTART;
286 * kernel_restart - reboot the system
287 * @cmd: pointer to buffer containing command to execute for restart
290 * Shutdown everything and perform a clean reboot.
291 * This is not safe to call in interrupt context.
293 void kernel_restart(char *cmd)
295 kernel_restart_prepare(cmd);
297 printk(KERN_EMERG "Restarting system.\n");
299 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
300 machine_restart(cmd);
302 EXPORT_SYMBOL_GPL(kernel_restart);
304 static void kernel_shutdown_prepare(enum system_states state)
306 blocking_notifier_call_chain(&reboot_notifier_list,
307 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
308 system_state = state;
312 * kernel_halt - halt the system
314 * Shutdown everything and perform a clean system halt.
316 void kernel_halt(void)
318 kernel_shutdown_prepare(SYSTEM_HALT);
320 printk(KERN_EMERG "System halted.\n");
324 EXPORT_SYMBOL_GPL(kernel_halt);
327 * kernel_power_off - power_off the system
329 * Shutdown everything and perform a clean system power_off.
331 void kernel_power_off(void)
333 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
334 if (pm_power_off_prepare)
335 pm_power_off_prepare();
336 disable_nonboot_cpus();
338 printk(KERN_EMERG "Power down.\n");
341 EXPORT_SYMBOL_GPL(kernel_power_off);
343 * Reboot system call: for obvious reasons only root may call it,
344 * and even root needs to set up some magic numbers in the registers
345 * so that some mistake won't make this reboot the whole machine.
346 * You can also set the meaning of the ctrl-alt-del-key here.
348 * reboot doesn't sync: do that yourself before calling this.
350 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
354 /* We only trust the superuser with rebooting the system. */
355 if (!capable(CAP_SYS_BOOT))
358 /* For safety, we require "magic" arguments. */
359 if (magic1 != LINUX_REBOOT_MAGIC1 ||
360 (magic2 != LINUX_REBOOT_MAGIC2 &&
361 magic2 != LINUX_REBOOT_MAGIC2A &&
362 magic2 != LINUX_REBOOT_MAGIC2B &&
363 magic2 != LINUX_REBOOT_MAGIC2C))
366 /* Instead of trying to make the power_off code look like
367 * halt when pm_power_off is not set do it the easy way.
369 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
370 cmd = LINUX_REBOOT_CMD_HALT;
374 case LINUX_REBOOT_CMD_RESTART:
375 kernel_restart(NULL);
378 case LINUX_REBOOT_CMD_CAD_ON:
382 case LINUX_REBOOT_CMD_CAD_OFF:
386 case LINUX_REBOOT_CMD_HALT:
392 case LINUX_REBOOT_CMD_POWER_OFF:
398 case LINUX_REBOOT_CMD_RESTART2:
399 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
403 buffer[sizeof(buffer) - 1] = '\0';
405 kernel_restart(buffer);
409 case LINUX_REBOOT_CMD_KEXEC:
412 ret = kernel_kexec();
418 #ifdef CONFIG_HIBERNATION
419 case LINUX_REBOOT_CMD_SW_SUSPEND:
421 int ret = hibernate();
435 static void deferred_cad(struct work_struct *dummy)
437 kernel_restart(NULL);
441 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
442 * As it's called within an interrupt, it may NOT sync: the only choice
443 * is whether to reboot at once, or just ignore the ctrl-alt-del.
445 void ctrl_alt_del(void)
447 static DECLARE_WORK(cad_work, deferred_cad);
450 schedule_work(&cad_work);
452 kill_cad_pid(SIGINT, 1);
456 * Unprivileged users may change the real gid to the effective gid
457 * or vice versa. (BSD-style)
459 * If you set the real gid at all, or set the effective gid to a value not
460 * equal to the real gid, then the saved gid is set to the new effective gid.
462 * This makes it possible for a setgid program to completely drop its
463 * privileges, which is often a useful assertion to make when you are doing
464 * a security audit over a program.
466 * The general idea is that a program which uses just setregid() will be
467 * 100% compatible with BSD. A program which uses just setgid() will be
468 * 100% compatible with POSIX with saved IDs.
470 * SMP: There are not races, the GIDs are checked only by filesystem
471 * operations (as far as semantic preservation is concerned).
473 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
475 int old_rgid = current->gid;
476 int old_egid = current->egid;
477 int new_rgid = old_rgid;
478 int new_egid = old_egid;
481 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
485 if (rgid != (gid_t) -1) {
486 if ((old_rgid == rgid) ||
487 (current->egid==rgid) ||
493 if (egid != (gid_t) -1) {
494 if ((old_rgid == egid) ||
495 (current->egid == egid) ||
496 (current->sgid == egid) ||
502 if (new_egid != old_egid) {
503 set_dumpable(current->mm, suid_dumpable);
506 if (rgid != (gid_t) -1 ||
507 (egid != (gid_t) -1 && egid != old_rgid))
508 current->sgid = new_egid;
509 current->fsgid = new_egid;
510 current->egid = new_egid;
511 current->gid = new_rgid;
512 key_fsgid_changed(current);
513 proc_id_connector(current, PROC_EVENT_GID);
518 * setgid() is implemented like SysV w/ SAVED_IDS
520 * SMP: Same implicit races as above.
522 asmlinkage long sys_setgid(gid_t gid)
524 int old_egid = current->egid;
527 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
531 if (capable(CAP_SETGID)) {
532 if (old_egid != gid) {
533 set_dumpable(current->mm, suid_dumpable);
536 current->gid = current->egid = current->sgid = current->fsgid = gid;
537 } else if ((gid == current->gid) || (gid == current->sgid)) {
538 if (old_egid != gid) {
539 set_dumpable(current->mm, suid_dumpable);
542 current->egid = current->fsgid = gid;
547 key_fsgid_changed(current);
548 proc_id_connector(current, PROC_EVENT_GID);
552 static int set_user(uid_t new_ruid, int dumpclear)
554 struct user_struct *new_user;
556 new_user = alloc_uid(current->nsproxy->user_ns, new_ruid);
560 if (atomic_read(&new_user->processes) >=
561 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
562 new_user != current->nsproxy->user_ns->root_user) {
567 switch_uid(new_user);
570 set_dumpable(current->mm, suid_dumpable);
573 current->uid = new_ruid;
578 * Unprivileged users may change the real uid to the effective uid
579 * or vice versa. (BSD-style)
581 * If you set the real uid at all, or set the effective uid to a value not
582 * equal to the real uid, then the saved uid is set to the new effective uid.
584 * This makes it possible for a setuid program to completely drop its
585 * privileges, which is often a useful assertion to make when you are doing
586 * a security audit over a program.
588 * The general idea is that a program which uses just setreuid() will be
589 * 100% compatible with BSD. A program which uses just setuid() will be
590 * 100% compatible with POSIX with saved IDs.
592 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
594 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
597 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
601 new_ruid = old_ruid = current->uid;
602 new_euid = old_euid = current->euid;
603 old_suid = current->suid;
605 if (ruid != (uid_t) -1) {
607 if ((old_ruid != ruid) &&
608 (current->euid != ruid) &&
609 !capable(CAP_SETUID))
613 if (euid != (uid_t) -1) {
615 if ((old_ruid != euid) &&
616 (current->euid != euid) &&
617 (current->suid != euid) &&
618 !capable(CAP_SETUID))
622 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
625 if (new_euid != old_euid) {
626 set_dumpable(current->mm, suid_dumpable);
629 current->fsuid = current->euid = new_euid;
630 if (ruid != (uid_t) -1 ||
631 (euid != (uid_t) -1 && euid != old_ruid))
632 current->suid = current->euid;
633 current->fsuid = current->euid;
635 key_fsuid_changed(current);
636 proc_id_connector(current, PROC_EVENT_UID);
638 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
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 asmlinkage long sys_setuid(uid_t uid)
656 int old_euid = current->euid;
657 int old_ruid, old_suid, new_suid;
660 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
664 old_ruid = current->uid;
665 old_suid = current->suid;
668 if (capable(CAP_SETUID)) {
669 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
672 } else if ((uid != current->uid) && (uid != new_suid))
675 if (old_euid != uid) {
676 set_dumpable(current->mm, suid_dumpable);
679 current->fsuid = current->euid = uid;
680 current->suid = new_suid;
682 key_fsuid_changed(current);
683 proc_id_connector(current, PROC_EVENT_UID);
685 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
690 * This function implements a generic ability to update ruid, euid,
691 * and suid. This allows you to implement the 4.4 compatible seteuid().
693 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
695 int old_ruid = current->uid;
696 int old_euid = current->euid;
697 int old_suid = current->suid;
700 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
704 if (!capable(CAP_SETUID)) {
705 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
706 (ruid != current->euid) && (ruid != current->suid))
708 if ((euid != (uid_t) -1) && (euid != current->uid) &&
709 (euid != current->euid) && (euid != current->suid))
711 if ((suid != (uid_t) -1) && (suid != current->uid) &&
712 (suid != current->euid) && (suid != current->suid))
715 if (ruid != (uid_t) -1) {
716 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
719 if (euid != (uid_t) -1) {
720 if (euid != current->euid) {
721 set_dumpable(current->mm, suid_dumpable);
724 current->euid = euid;
726 current->fsuid = current->euid;
727 if (suid != (uid_t) -1)
728 current->suid = suid;
730 key_fsuid_changed(current);
731 proc_id_connector(current, PROC_EVENT_UID);
733 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
736 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
740 if (!(retval = put_user(current->uid, ruid)) &&
741 !(retval = put_user(current->euid, euid)))
742 retval = put_user(current->suid, suid);
748 * Same as above, but for rgid, egid, sgid.
750 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
754 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
758 if (!capable(CAP_SETGID)) {
759 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
760 (rgid != current->egid) && (rgid != current->sgid))
762 if ((egid != (gid_t) -1) && (egid != current->gid) &&
763 (egid != current->egid) && (egid != current->sgid))
765 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
766 (sgid != current->egid) && (sgid != current->sgid))
769 if (egid != (gid_t) -1) {
770 if (egid != current->egid) {
771 set_dumpable(current->mm, suid_dumpable);
774 current->egid = egid;
776 current->fsgid = current->egid;
777 if (rgid != (gid_t) -1)
779 if (sgid != (gid_t) -1)
780 current->sgid = sgid;
782 key_fsgid_changed(current);
783 proc_id_connector(current, PROC_EVENT_GID);
787 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
791 if (!(retval = put_user(current->gid, rgid)) &&
792 !(retval = put_user(current->egid, egid)))
793 retval = put_user(current->sgid, sgid);
800 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
801 * is used for "access()" and for the NFS daemon (letting nfsd stay at
802 * whatever uid it wants to). It normally shadows "euid", except when
803 * explicitly set by setfsuid() or for access..
805 asmlinkage long sys_setfsuid(uid_t uid)
809 old_fsuid = current->fsuid;
810 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
813 if (uid == current->uid || uid == current->euid ||
814 uid == current->suid || uid == current->fsuid ||
815 capable(CAP_SETUID)) {
816 if (uid != old_fsuid) {
817 set_dumpable(current->mm, suid_dumpable);
820 current->fsuid = uid;
823 key_fsuid_changed(current);
824 proc_id_connector(current, PROC_EVENT_UID);
826 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
832 * Samma på svenska..
834 asmlinkage long sys_setfsgid(gid_t gid)
838 old_fsgid = current->fsgid;
839 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
842 if (gid == current->gid || gid == current->egid ||
843 gid == current->sgid || gid == current->fsgid ||
844 capable(CAP_SETGID)) {
845 if (gid != old_fsgid) {
846 set_dumpable(current->mm, suid_dumpable);
849 current->fsgid = gid;
850 key_fsgid_changed(current);
851 proc_id_connector(current, PROC_EVENT_GID);
856 asmlinkage long sys_times(struct tms __user * tbuf)
859 * In the SMP world we might just be unlucky and have one of
860 * the times increment as we use it. Since the value is an
861 * atomically safe type this is just fine. Conceptually its
862 * as if the syscall took an instant longer to occur.
866 struct task_struct *tsk = current;
867 struct task_struct *t;
868 cputime_t utime, stime, cutime, cstime;
870 spin_lock_irq(&tsk->sighand->siglock);
871 utime = tsk->signal->utime;
872 stime = tsk->signal->stime;
875 utime = cputime_add(utime, t->utime);
876 stime = cputime_add(stime, t->stime);
880 cutime = tsk->signal->cutime;
881 cstime = tsk->signal->cstime;
882 spin_unlock_irq(&tsk->sighand->siglock);
884 tmp.tms_utime = cputime_to_clock_t(utime);
885 tmp.tms_stime = cputime_to_clock_t(stime);
886 tmp.tms_cutime = cputime_to_clock_t(cutime);
887 tmp.tms_cstime = cputime_to_clock_t(cstime);
888 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
891 return (long) jiffies_64_to_clock_t(get_jiffies_64());
895 * This needs some heavy checking ...
896 * I just haven't the stomach for it. I also don't fully
897 * understand sessions/pgrp etc. Let somebody who does explain it.
899 * OK, I think I have the protection semantics right.... this is really
900 * only important on a multi-user system anyway, to make sure one user
901 * can't send a signal to a process owned by another. -TYT, 12/12/91
903 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
906 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
908 struct task_struct *p;
909 struct task_struct *group_leader = current->group_leader;
914 pid = task_pid_vnr(group_leader);
920 /* From this point forward we keep holding onto the tasklist lock
921 * so that our parent does not change from under us. -DaveM
923 write_lock_irq(&tasklist_lock);
926 p = find_task_by_vpid(pid);
931 if (!thread_group_leader(p))
934 if (same_thread_group(p->real_parent, group_leader)) {
936 if (task_session(p) != task_session(group_leader))
943 if (p != group_leader)
948 if (p->signal->leader)
953 struct task_struct *g;
955 pgrp = find_vpid(pgid);
956 g = pid_task(pgrp, PIDTYPE_PGID);
957 if (!g || task_session(g) != task_session(group_leader))
961 err = security_task_setpgid(p, pgid);
965 if (task_pgrp(p) != pgrp) {
966 change_pid(p, PIDTYPE_PGID, pgrp);
967 set_task_pgrp(p, pid_nr(pgrp));
972 /* All paths lead to here, thus we are safe. -DaveM */
973 write_unlock_irq(&tasklist_lock);
977 asmlinkage long sys_getpgid(pid_t pid)
979 struct task_struct *p;
985 grp = task_pgrp(current);
988 p = find_task_by_vpid(pid);
995 retval = security_task_getpgid(p);
999 retval = pid_vnr(grp);
1005 #ifdef __ARCH_WANT_SYS_GETPGRP
1007 asmlinkage long sys_getpgrp(void)
1009 return sys_getpgid(0);
1014 asmlinkage long sys_getsid(pid_t pid)
1016 struct task_struct *p;
1022 sid = task_session(current);
1025 p = find_task_by_vpid(pid);
1028 sid = task_session(p);
1032 retval = security_task_getsid(p);
1036 retval = pid_vnr(sid);
1042 asmlinkage long sys_setsid(void)
1044 struct task_struct *group_leader = current->group_leader;
1045 struct pid *sid = task_pid(group_leader);
1046 pid_t session = pid_vnr(sid);
1049 write_lock_irq(&tasklist_lock);
1050 /* Fail if I am already a session leader */
1051 if (group_leader->signal->leader)
1054 /* Fail if a process group id already exists that equals the
1055 * proposed session id.
1057 if (pid_task(sid, PIDTYPE_PGID))
1060 group_leader->signal->leader = 1;
1061 __set_special_pids(sid);
1063 proc_clear_tty(group_leader);
1067 write_unlock_irq(&tasklist_lock);
1072 * Supplementary group IDs
1075 /* init to 2 - one for init_task, one to ensure it is never freed */
1076 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1078 struct group_info *groups_alloc(int gidsetsize)
1080 struct group_info *group_info;
1084 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1085 /* Make sure we always allocate at least one indirect block pointer */
1086 nblocks = nblocks ? : 1;
1087 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1090 group_info->ngroups = gidsetsize;
1091 group_info->nblocks = nblocks;
1092 atomic_set(&group_info->usage, 1);
1094 if (gidsetsize <= NGROUPS_SMALL)
1095 group_info->blocks[0] = group_info->small_block;
1097 for (i = 0; i < nblocks; i++) {
1099 b = (void *)__get_free_page(GFP_USER);
1101 goto out_undo_partial_alloc;
1102 group_info->blocks[i] = b;
1107 out_undo_partial_alloc:
1109 free_page((unsigned long)group_info->blocks[i]);
1115 EXPORT_SYMBOL(groups_alloc);
1117 void groups_free(struct group_info *group_info)
1119 if (group_info->blocks[0] != group_info->small_block) {
1121 for (i = 0; i < group_info->nblocks; i++)
1122 free_page((unsigned long)group_info->blocks[i]);
1127 EXPORT_SYMBOL(groups_free);
1129 /* export the group_info to a user-space array */
1130 static int groups_to_user(gid_t __user *grouplist,
1131 struct group_info *group_info)
1134 unsigned int count = group_info->ngroups;
1136 for (i = 0; i < group_info->nblocks; i++) {
1137 unsigned int cp_count = min(NGROUPS_PER_BLOCK, count);
1138 unsigned int len = cp_count * sizeof(*grouplist);
1140 if (copy_to_user(grouplist, group_info->blocks[i], len))
1143 grouplist += NGROUPS_PER_BLOCK;
1149 /* fill a group_info from a user-space array - it must be allocated already */
1150 static int groups_from_user(struct group_info *group_info,
1151 gid_t __user *grouplist)
1154 unsigned int count = group_info->ngroups;
1156 for (i = 0; i < group_info->nblocks; i++) {
1157 unsigned int cp_count = min(NGROUPS_PER_BLOCK, count);
1158 unsigned int len = cp_count * sizeof(*grouplist);
1160 if (copy_from_user(group_info->blocks[i], grouplist, len))
1163 grouplist += NGROUPS_PER_BLOCK;
1169 /* a simple Shell sort */
1170 static void groups_sort(struct group_info *group_info)
1172 int base, max, stride;
1173 int gidsetsize = group_info->ngroups;
1175 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1180 max = gidsetsize - stride;
1181 for (base = 0; base < max; base++) {
1183 int right = left + stride;
1184 gid_t tmp = GROUP_AT(group_info, right);
1186 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1187 GROUP_AT(group_info, right) =
1188 GROUP_AT(group_info, left);
1192 GROUP_AT(group_info, right) = tmp;
1198 /* a simple bsearch */
1199 int groups_search(struct group_info *group_info, gid_t grp)
1201 unsigned int left, right;
1207 right = group_info->ngroups;
1208 while (left < right) {
1209 unsigned int mid = (left+right)/2;
1210 int cmp = grp - GROUP_AT(group_info, mid);
1221 /* validate and set current->group_info */
1222 int set_current_groups(struct group_info *group_info)
1225 struct group_info *old_info;
1227 retval = security_task_setgroups(group_info);
1231 groups_sort(group_info);
1232 get_group_info(group_info);
1235 old_info = current->group_info;
1236 current->group_info = group_info;
1237 task_unlock(current);
1239 put_group_info(old_info);
1244 EXPORT_SYMBOL(set_current_groups);
1246 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1251 * SMP: Nobody else can change our grouplist. Thus we are
1258 /* no need to grab task_lock here; it cannot change */
1259 i = current->group_info->ngroups;
1261 if (i > gidsetsize) {
1265 if (groups_to_user(grouplist, current->group_info)) {
1275 * SMP: Our groups are copy-on-write. We can set them safely
1276 * without another task interfering.
1279 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1281 struct group_info *group_info;
1284 if (!capable(CAP_SETGID))
1286 if ((unsigned)gidsetsize > NGROUPS_MAX)
1289 group_info = groups_alloc(gidsetsize);
1292 retval = groups_from_user(group_info, grouplist);
1294 put_group_info(group_info);
1298 retval = set_current_groups(group_info);
1299 put_group_info(group_info);
1305 * Check whether we're fsgid/egid or in the supplemental group..
1307 int in_group_p(gid_t grp)
1310 if (grp != current->fsgid)
1311 retval = groups_search(current->group_info, grp);
1315 EXPORT_SYMBOL(in_group_p);
1317 int in_egroup_p(gid_t grp)
1320 if (grp != current->egid)
1321 retval = groups_search(current->group_info, grp);
1325 EXPORT_SYMBOL(in_egroup_p);
1327 DECLARE_RWSEM(uts_sem);
1329 asmlinkage long sys_newuname(struct new_utsname __user * name)
1333 down_read(&uts_sem);
1334 if (copy_to_user(name, utsname(), sizeof *name))
1340 asmlinkage long sys_sethostname(char __user *name, int len)
1343 char tmp[__NEW_UTS_LEN];
1345 if (!capable(CAP_SYS_ADMIN))
1347 if (len < 0 || len > __NEW_UTS_LEN)
1349 down_write(&uts_sem);
1351 if (!copy_from_user(tmp, name, len)) {
1352 memcpy(utsname()->nodename, tmp, len);
1353 utsname()->nodename[len] = 0;
1360 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1362 asmlinkage long sys_gethostname(char __user *name, int len)
1368 down_read(&uts_sem);
1369 i = 1 + strlen(utsname()->nodename);
1373 if (copy_to_user(name, utsname()->nodename, i))
1382 * Only setdomainname; getdomainname can be implemented by calling
1385 asmlinkage long sys_setdomainname(char __user *name, int len)
1388 char tmp[__NEW_UTS_LEN];
1390 if (!capable(CAP_SYS_ADMIN))
1392 if (len < 0 || len > __NEW_UTS_LEN)
1395 down_write(&uts_sem);
1397 if (!copy_from_user(tmp, name, len)) {
1398 memcpy(utsname()->domainname, tmp, len);
1399 utsname()->domainname[len] = 0;
1406 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1408 if (resource >= RLIM_NLIMITS)
1411 struct rlimit value;
1412 task_lock(current->group_leader);
1413 value = current->signal->rlim[resource];
1414 task_unlock(current->group_leader);
1415 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1419 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1422 * Back compatibility for getrlimit. Needed for some apps.
1425 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1428 if (resource >= RLIM_NLIMITS)
1431 task_lock(current->group_leader);
1432 x = current->signal->rlim[resource];
1433 task_unlock(current->group_leader);
1434 if (x.rlim_cur > 0x7FFFFFFF)
1435 x.rlim_cur = 0x7FFFFFFF;
1436 if (x.rlim_max > 0x7FFFFFFF)
1437 x.rlim_max = 0x7FFFFFFF;
1438 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1443 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1445 struct rlimit new_rlim, *old_rlim;
1446 unsigned long it_prof_secs;
1449 if (resource >= RLIM_NLIMITS)
1451 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1453 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1455 old_rlim = current->signal->rlim + resource;
1456 if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1457 !capable(CAP_SYS_RESOURCE))
1459 if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > sysctl_nr_open)
1462 retval = security_task_setrlimit(resource, &new_rlim);
1466 if (resource == RLIMIT_CPU && new_rlim.rlim_cur == 0) {
1468 * The caller is asking for an immediate RLIMIT_CPU
1469 * expiry. But we use the zero value to mean "it was
1470 * never set". So let's cheat and make it one second
1473 new_rlim.rlim_cur = 1;
1476 task_lock(current->group_leader);
1477 *old_rlim = new_rlim;
1478 task_unlock(current->group_leader);
1480 if (resource != RLIMIT_CPU)
1484 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1485 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1486 * very long-standing error, and fixing it now risks breakage of
1487 * applications, so we live with it
1489 if (new_rlim.rlim_cur == RLIM_INFINITY)
1492 it_prof_secs = cputime_to_secs(current->signal->it_prof_expires);
1493 if (it_prof_secs == 0 || new_rlim.rlim_cur <= it_prof_secs) {
1494 unsigned long rlim_cur = new_rlim.rlim_cur;
1497 cputime = secs_to_cputime(rlim_cur);
1498 read_lock(&tasklist_lock);
1499 spin_lock_irq(¤t->sighand->siglock);
1500 set_process_cpu_timer(current, CPUCLOCK_PROF, &cputime, NULL);
1501 spin_unlock_irq(¤t->sighand->siglock);
1502 read_unlock(&tasklist_lock);
1509 * It would make sense to put struct rusage in the task_struct,
1510 * except that would make the task_struct be *really big*. After
1511 * task_struct gets moved into malloc'ed memory, it would
1512 * make sense to do this. It will make moving the rest of the information
1513 * a lot simpler! (Which we're not doing right now because we're not
1514 * measuring them yet).
1516 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1517 * races with threads incrementing their own counters. But since word
1518 * reads are atomic, we either get new values or old values and we don't
1519 * care which for the sums. We always take the siglock to protect reading
1520 * the c* fields from p->signal from races with exit.c updating those
1521 * fields when reaping, so a sample either gets all the additions of a
1522 * given child after it's reaped, or none so this sample is before reaping.
1525 * We need to take the siglock for CHILDEREN, SELF and BOTH
1526 * for the cases current multithreaded, non-current single threaded
1527 * non-current multithreaded. Thread traversal is now safe with
1529 * Strictly speaking, we donot need to take the siglock if we are current and
1530 * single threaded, as no one else can take our signal_struct away, no one
1531 * else can reap the children to update signal->c* counters, and no one else
1532 * can race with the signal-> fields. If we do not take any lock, the
1533 * signal-> fields could be read out of order while another thread was just
1534 * exiting. So we should place a read memory barrier when we avoid the lock.
1535 * On the writer side, write memory barrier is implied in __exit_signal
1536 * as __exit_signal releases the siglock spinlock after updating the signal->
1537 * fields. But we don't do this yet to keep things simple.
1541 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r,
1542 cputime_t *utimep, cputime_t *stimep)
1544 *utimep = cputime_add(*utimep, t->utime);
1545 *stimep = cputime_add(*stimep, t->stime);
1546 r->ru_nvcsw += t->nvcsw;
1547 r->ru_nivcsw += t->nivcsw;
1548 r->ru_minflt += t->min_flt;
1549 r->ru_majflt += t->maj_flt;
1550 r->ru_inblock += task_io_get_inblock(t);
1551 r->ru_oublock += task_io_get_oublock(t);
1554 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1556 struct task_struct *t;
1557 unsigned long flags;
1558 cputime_t utime, stime;
1560 memset((char *) r, 0, sizeof *r);
1561 utime = stime = cputime_zero;
1563 if (who == RUSAGE_THREAD) {
1564 accumulate_thread_rusage(p, r, &utime, &stime);
1568 if (!lock_task_sighand(p, &flags))
1573 case RUSAGE_CHILDREN:
1574 utime = p->signal->cutime;
1575 stime = p->signal->cstime;
1576 r->ru_nvcsw = p->signal->cnvcsw;
1577 r->ru_nivcsw = p->signal->cnivcsw;
1578 r->ru_minflt = p->signal->cmin_flt;
1579 r->ru_majflt = p->signal->cmaj_flt;
1580 r->ru_inblock = p->signal->cinblock;
1581 r->ru_oublock = p->signal->coublock;
1583 if (who == RUSAGE_CHILDREN)
1587 utime = cputime_add(utime, p->signal->utime);
1588 stime = cputime_add(stime, p->signal->stime);
1589 r->ru_nvcsw += p->signal->nvcsw;
1590 r->ru_nivcsw += p->signal->nivcsw;
1591 r->ru_minflt += p->signal->min_flt;
1592 r->ru_majflt += p->signal->maj_flt;
1593 r->ru_inblock += p->signal->inblock;
1594 r->ru_oublock += p->signal->oublock;
1597 accumulate_thread_rusage(t, r, &utime, &stime);
1605 unlock_task_sighand(p, &flags);
1608 cputime_to_timeval(utime, &r->ru_utime);
1609 cputime_to_timeval(stime, &r->ru_stime);
1612 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1615 k_getrusage(p, who, &r);
1616 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1619 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1621 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1622 who != RUSAGE_THREAD)
1624 return getrusage(current, who, ru);
1627 asmlinkage long sys_umask(int mask)
1629 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1633 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1634 unsigned long arg4, unsigned long arg5)
1638 if (security_task_prctl(option, arg2, arg3, arg4, arg5, &error))
1642 case PR_SET_PDEATHSIG:
1643 if (!valid_signal(arg2)) {
1647 current->pdeath_signal = arg2;
1649 case PR_GET_PDEATHSIG:
1650 error = put_user(current->pdeath_signal, (int __user *)arg2);
1652 case PR_GET_DUMPABLE:
1653 error = get_dumpable(current->mm);
1655 case PR_SET_DUMPABLE:
1656 if (arg2 < 0 || arg2 > 1) {
1660 set_dumpable(current->mm, arg2);
1663 case PR_SET_UNALIGN:
1664 error = SET_UNALIGN_CTL(current, arg2);
1666 case PR_GET_UNALIGN:
1667 error = GET_UNALIGN_CTL(current, arg2);
1670 error = SET_FPEMU_CTL(current, arg2);
1673 error = GET_FPEMU_CTL(current, arg2);
1676 error = SET_FPEXC_CTL(current, arg2);
1679 error = GET_FPEXC_CTL(current, arg2);
1682 error = PR_TIMING_STATISTICAL;
1685 if (arg2 != PR_TIMING_STATISTICAL)
1690 struct task_struct *me = current;
1691 unsigned char ncomm[sizeof(me->comm)];
1693 ncomm[sizeof(me->comm)-1] = 0;
1694 if (strncpy_from_user(ncomm, (char __user *)arg2,
1695 sizeof(me->comm)-1) < 0)
1697 set_task_comm(me, ncomm);
1701 struct task_struct *me = current;
1702 unsigned char tcomm[sizeof(me->comm)];
1704 get_task_comm(tcomm, me);
1705 if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))
1710 error = GET_ENDIAN(current, arg2);
1713 error = SET_ENDIAN(current, arg2);
1716 case PR_GET_SECCOMP:
1717 error = prctl_get_seccomp();
1719 case PR_SET_SECCOMP:
1720 error = prctl_set_seccomp(arg2);
1723 error = GET_TSC_CTL(arg2);
1726 error = SET_TSC_CTL(arg2);
1735 asmlinkage long sys_getcpu(unsigned __user *cpup, unsigned __user *nodep,
1736 struct getcpu_cache __user *unused)
1739 int cpu = raw_smp_processor_id();
1741 err |= put_user(cpu, cpup);
1743 err |= put_user(cpu_to_node(cpu), nodep);
1744 return err ? -EFAULT : 0;
1747 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
1749 static void argv_cleanup(char **argv, char **envp)
1755 * orderly_poweroff - Trigger an orderly system poweroff
1756 * @force: force poweroff if command execution fails
1758 * This may be called from any context to trigger a system shutdown.
1759 * If the orderly shutdown fails, it will force an immediate shutdown.
1761 int orderly_poweroff(bool force)
1764 char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
1765 static char *envp[] = {
1767 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
1771 struct subprocess_info *info;
1774 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
1775 __func__, poweroff_cmd);
1779 info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
1785 call_usermodehelper_setcleanup(info, argv_cleanup);
1787 ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
1791 printk(KERN_WARNING "Failed to start orderly shutdown: "
1792 "forcing the issue\n");
1794 /* I guess this should try to kick off some daemon to
1795 sync and poweroff asap. Or not even bother syncing
1796 if we're doing an emergency shutdown? */
1803 EXPORT_SYMBOL_GPL(orderly_poweroff);