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);
116 * set the priority of a task
117 * - the caller must hold the RCU read lock
119 static int set_one_prio(struct task_struct *p, int niceval, int error)
121 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
124 if (pcred->uid != cred->euid &&
125 pcred->euid != cred->euid && !capable(CAP_SYS_NICE)) {
129 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
133 no_nice = security_task_setnice(p, niceval);
140 set_user_nice(p, niceval);
145 asmlinkage long sys_setpriority(int which, int who, int niceval)
147 struct task_struct *g, *p;
148 struct user_struct *user;
149 const struct cred *cred = current_cred();
153 if (which > PRIO_USER || which < PRIO_PROCESS)
156 /* normalize: avoid signed division (rounding problems) */
163 read_lock(&tasklist_lock);
167 p = find_task_by_vpid(who);
171 error = set_one_prio(p, niceval, error);
175 pgrp = find_vpid(who);
177 pgrp = task_pgrp(current);
178 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
179 error = set_one_prio(p, niceval, error);
180 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
186 else if ((who != cred->uid) &&
187 !(user = find_user(who)))
188 goto out_unlock; /* No processes for this user */
191 if (__task_cred(p)->uid == who)
192 error = set_one_prio(p, niceval, error);
193 while_each_thread(g, p);
194 if (who != cred->uid)
195 free_uid(user); /* For find_user() */
199 read_unlock(&tasklist_lock);
205 * Ugh. To avoid negative return values, "getpriority()" will
206 * not return the normal nice-value, but a negated value that
207 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
208 * to stay compatible.
210 asmlinkage long sys_getpriority(int which, int who)
212 struct task_struct *g, *p;
213 struct user_struct *user;
214 const struct cred *cred = current_cred();
215 long niceval, retval = -ESRCH;
218 if (which > PRIO_USER || which < PRIO_PROCESS)
221 read_lock(&tasklist_lock);
225 p = find_task_by_vpid(who);
229 niceval = 20 - task_nice(p);
230 if (niceval > retval)
236 pgrp = find_vpid(who);
238 pgrp = task_pgrp(current);
239 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
240 niceval = 20 - task_nice(p);
241 if (niceval > retval)
243 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
246 user = (struct user_struct *) cred->user;
249 else if ((who != cred->uid) &&
250 !(user = find_user(who)))
251 goto out_unlock; /* No processes for this user */
254 if (__task_cred(p)->uid == who) {
255 niceval = 20 - task_nice(p);
256 if (niceval > retval)
259 while_each_thread(g, p);
260 if (who != cred->uid)
261 free_uid(user); /* for find_user() */
265 read_unlock(&tasklist_lock);
271 * emergency_restart - reboot the system
273 * Without shutting down any hardware or taking any locks
274 * reboot the system. This is called when we know we are in
275 * trouble so this is our best effort to reboot. This is
276 * safe to call in interrupt context.
278 void emergency_restart(void)
280 machine_emergency_restart();
282 EXPORT_SYMBOL_GPL(emergency_restart);
284 void kernel_restart_prepare(char *cmd)
286 blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
287 system_state = SYSTEM_RESTART;
293 * kernel_restart - reboot the system
294 * @cmd: pointer to buffer containing command to execute for restart
297 * Shutdown everything and perform a clean reboot.
298 * This is not safe to call in interrupt context.
300 void kernel_restart(char *cmd)
302 kernel_restart_prepare(cmd);
304 printk(KERN_EMERG "Restarting system.\n");
306 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
307 machine_restart(cmd);
309 EXPORT_SYMBOL_GPL(kernel_restart);
311 static void kernel_shutdown_prepare(enum system_states state)
313 blocking_notifier_call_chain(&reboot_notifier_list,
314 (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL);
315 system_state = state;
319 * kernel_halt - halt the system
321 * Shutdown everything and perform a clean system halt.
323 void kernel_halt(void)
325 kernel_shutdown_prepare(SYSTEM_HALT);
327 printk(KERN_EMERG "System halted.\n");
331 EXPORT_SYMBOL_GPL(kernel_halt);
334 * kernel_power_off - power_off the system
336 * Shutdown everything and perform a clean system power_off.
338 void kernel_power_off(void)
340 kernel_shutdown_prepare(SYSTEM_POWER_OFF);
341 if (pm_power_off_prepare)
342 pm_power_off_prepare();
343 disable_nonboot_cpus();
345 printk(KERN_EMERG "Power down.\n");
348 EXPORT_SYMBOL_GPL(kernel_power_off);
350 * Reboot system call: for obvious reasons only root may call it,
351 * and even root needs to set up some magic numbers in the registers
352 * so that some mistake won't make this reboot the whole machine.
353 * You can also set the meaning of the ctrl-alt-del-key here.
355 * reboot doesn't sync: do that yourself before calling this.
357 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
361 /* We only trust the superuser with rebooting the system. */
362 if (!capable(CAP_SYS_BOOT))
365 /* For safety, we require "magic" arguments. */
366 if (magic1 != LINUX_REBOOT_MAGIC1 ||
367 (magic2 != LINUX_REBOOT_MAGIC2 &&
368 magic2 != LINUX_REBOOT_MAGIC2A &&
369 magic2 != LINUX_REBOOT_MAGIC2B &&
370 magic2 != LINUX_REBOOT_MAGIC2C))
373 /* Instead of trying to make the power_off code look like
374 * halt when pm_power_off is not set do it the easy way.
376 if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
377 cmd = LINUX_REBOOT_CMD_HALT;
381 case LINUX_REBOOT_CMD_RESTART:
382 kernel_restart(NULL);
385 case LINUX_REBOOT_CMD_CAD_ON:
389 case LINUX_REBOOT_CMD_CAD_OFF:
393 case LINUX_REBOOT_CMD_HALT:
399 case LINUX_REBOOT_CMD_POWER_OFF:
405 case LINUX_REBOOT_CMD_RESTART2:
406 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
410 buffer[sizeof(buffer) - 1] = '\0';
412 kernel_restart(buffer);
416 case LINUX_REBOOT_CMD_KEXEC:
419 ret = kernel_kexec();
425 #ifdef CONFIG_HIBERNATION
426 case LINUX_REBOOT_CMD_SW_SUSPEND:
428 int ret = hibernate();
442 static void deferred_cad(struct work_struct *dummy)
444 kernel_restart(NULL);
448 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
449 * As it's called within an interrupt, it may NOT sync: the only choice
450 * is whether to reboot at once, or just ignore the ctrl-alt-del.
452 void ctrl_alt_del(void)
454 static DECLARE_WORK(cad_work, deferred_cad);
457 schedule_work(&cad_work);
459 kill_cad_pid(SIGINT, 1);
463 * Unprivileged users may change the real gid to the effective gid
464 * or vice versa. (BSD-style)
466 * If you set the real gid at all, or set the effective gid to a value not
467 * equal to the real gid, then the saved gid is set to the new effective gid.
469 * This makes it possible for a setgid program to completely drop its
470 * privileges, which is often a useful assertion to make when you are doing
471 * a security audit over a program.
473 * The general idea is that a program which uses just setregid() will be
474 * 100% compatible with BSD. A program which uses just setgid() will be
475 * 100% compatible with POSIX with saved IDs.
477 * SMP: There are not races, the GIDs are checked only by filesystem
478 * operations (as far as semantic preservation is concerned).
480 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
482 struct cred *cred = current->cred;
483 int old_rgid = cred->gid;
484 int old_egid = cred->egid;
485 int new_rgid = old_rgid;
486 int new_egid = old_egid;
489 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
493 if (rgid != (gid_t) -1) {
494 if ((old_rgid == rgid) ||
495 (cred->egid == rgid) ||
501 if (egid != (gid_t) -1) {
502 if ((old_rgid == egid) ||
503 (cred->egid == egid) ||
504 (cred->sgid == egid) ||
510 if (new_egid != old_egid) {
511 set_dumpable(current->mm, suid_dumpable);
514 if (rgid != (gid_t) -1 ||
515 (egid != (gid_t) -1 && egid != old_rgid))
516 cred->sgid = new_egid;
517 cred->fsgid = new_egid;
518 cred->egid = new_egid;
519 cred->gid = new_rgid;
520 key_fsgid_changed(current);
521 proc_id_connector(current, PROC_EVENT_GID);
526 * setgid() is implemented like SysV w/ SAVED_IDS
528 * SMP: Same implicit races as above.
530 asmlinkage long sys_setgid(gid_t gid)
532 struct cred *cred = current->cred;
533 int old_egid = cred->egid;
536 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
540 if (capable(CAP_SETGID)) {
541 if (old_egid != gid) {
542 set_dumpable(current->mm, suid_dumpable);
545 cred->gid = cred->egid = cred->sgid = cred->fsgid = gid;
546 } else if ((gid == cred->gid) || (gid == cred->sgid)) {
547 if (old_egid != gid) {
548 set_dumpable(current->mm, suid_dumpable);
551 cred->egid = cred->fsgid = gid;
556 key_fsgid_changed(current);
557 proc_id_connector(current, PROC_EVENT_GID);
561 static int set_user(uid_t new_ruid, int dumpclear)
563 struct user_struct *new_user;
565 new_user = alloc_uid(current->nsproxy->user_ns, new_ruid);
569 if (atomic_read(&new_user->processes) >=
570 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
571 new_user != current->nsproxy->user_ns->root_user) {
576 switch_uid(new_user);
579 set_dumpable(current->mm, suid_dumpable);
582 current->cred->uid = new_ruid;
587 * Unprivileged users may change the real uid to the effective uid
588 * or vice versa. (BSD-style)
590 * If you set the real uid at all, or set the effective uid to a value not
591 * equal to the real uid, then the saved uid is set to the new effective uid.
593 * This makes it possible for a setuid program to completely drop its
594 * privileges, which is often a useful assertion to make when you are doing
595 * a security audit over a program.
597 * The general idea is that a program which uses just setreuid() will be
598 * 100% compatible with BSD. A program which uses just setuid() will be
599 * 100% compatible with POSIX with saved IDs.
601 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
603 struct cred *cred = current->cred;
604 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
607 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
611 new_ruid = old_ruid = cred->uid;
612 new_euid = old_euid = cred->euid;
613 old_suid = cred->suid;
615 if (ruid != (uid_t) -1) {
617 if ((old_ruid != ruid) &&
618 (cred->euid != ruid) &&
619 !capable(CAP_SETUID))
623 if (euid != (uid_t) -1) {
625 if ((old_ruid != euid) &&
626 (cred->euid != euid) &&
627 (cred->suid != euid) &&
628 !capable(CAP_SETUID))
632 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
635 if (new_euid != old_euid) {
636 set_dumpable(current->mm, suid_dumpable);
639 cred->fsuid = cred->euid = new_euid;
640 if (ruid != (uid_t) -1 ||
641 (euid != (uid_t) -1 && euid != old_ruid))
642 cred->suid = cred->euid;
643 cred->fsuid = cred->euid;
645 key_fsuid_changed(current);
646 proc_id_connector(current, PROC_EVENT_UID);
648 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
654 * setuid() is implemented like SysV with SAVED_IDS
656 * Note that SAVED_ID's is deficient in that a setuid root program
657 * like sendmail, for example, cannot set its uid to be a normal
658 * user and then switch back, because if you're root, setuid() sets
659 * the saved uid too. If you don't like this, blame the bright people
660 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
661 * will allow a root program to temporarily drop privileges and be able to
662 * regain them by swapping the real and effective uid.
664 asmlinkage long sys_setuid(uid_t uid)
666 struct cred *cred = current->cred;
667 int old_euid = cred->euid;
668 int old_ruid, old_suid, new_suid;
671 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
675 old_ruid = cred->uid;
676 old_suid = cred->suid;
679 if (capable(CAP_SETUID)) {
680 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
683 } else if ((uid != cred->uid) && (uid != new_suid))
686 if (old_euid != uid) {
687 set_dumpable(current->mm, suid_dumpable);
690 cred->fsuid = cred->euid = uid;
691 cred->suid = new_suid;
693 key_fsuid_changed(current);
694 proc_id_connector(current, PROC_EVENT_UID);
696 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
701 * This function implements a generic ability to update ruid, euid,
702 * and suid. This allows you to implement the 4.4 compatible seteuid().
704 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
706 struct cred *cred = current->cred;
707 int old_ruid = cred->uid;
708 int old_euid = cred->euid;
709 int old_suid = cred->suid;
712 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
716 if (!capable(CAP_SETUID)) {
717 if ((ruid != (uid_t) -1) && (ruid != cred->uid) &&
718 (ruid != cred->euid) && (ruid != cred->suid))
720 if ((euid != (uid_t) -1) && (euid != cred->uid) &&
721 (euid != cred->euid) && (euid != cred->suid))
723 if ((suid != (uid_t) -1) && (suid != cred->uid) &&
724 (suid != cred->euid) && (suid != cred->suid))
727 if (ruid != (uid_t) -1) {
728 if (ruid != cred->uid &&
729 set_user(ruid, euid != cred->euid) < 0)
732 if (euid != (uid_t) -1) {
733 if (euid != cred->euid) {
734 set_dumpable(current->mm, suid_dumpable);
739 cred->fsuid = cred->euid;
740 if (suid != (uid_t) -1)
743 key_fsuid_changed(current);
744 proc_id_connector(current, PROC_EVENT_UID);
746 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
749 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
751 const struct cred *cred = current_cred();
754 if (!(retval = put_user(cred->uid, ruid)) &&
755 !(retval = put_user(cred->euid, euid)))
756 retval = put_user(cred->suid, suid);
762 * Same as above, but for rgid, egid, sgid.
764 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
766 struct cred *cred = current->cred;
769 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
773 if (!capable(CAP_SETGID)) {
774 if ((rgid != (gid_t) -1) && (rgid != cred->gid) &&
775 (rgid != cred->egid) && (rgid != cred->sgid))
777 if ((egid != (gid_t) -1) && (egid != cred->gid) &&
778 (egid != cred->egid) && (egid != cred->sgid))
780 if ((sgid != (gid_t) -1) && (sgid != cred->gid) &&
781 (sgid != cred->egid) && (sgid != cred->sgid))
784 if (egid != (gid_t) -1) {
785 if (egid != cred->egid) {
786 set_dumpable(current->mm, suid_dumpable);
791 cred->fsgid = cred->egid;
792 if (rgid != (gid_t) -1)
794 if (sgid != (gid_t) -1)
797 key_fsgid_changed(current);
798 proc_id_connector(current, PROC_EVENT_GID);
802 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
804 const struct cred *cred = current_cred();
807 if (!(retval = put_user(cred->gid, rgid)) &&
808 !(retval = put_user(cred->egid, egid)))
809 retval = put_user(cred->sgid, sgid);
816 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
817 * is used for "access()" and for the NFS daemon (letting nfsd stay at
818 * whatever uid it wants to). It normally shadows "euid", except when
819 * explicitly set by setfsuid() or for access..
821 asmlinkage long sys_setfsuid(uid_t uid)
823 struct cred *cred = current->cred;
826 old_fsuid = cred->fsuid;
827 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
830 if (uid == cred->uid || uid == cred->euid ||
831 uid == cred->suid || uid == cred->fsuid ||
832 capable(CAP_SETUID)) {
833 if (uid != old_fsuid) {
834 set_dumpable(current->mm, suid_dumpable);
840 key_fsuid_changed(current);
841 proc_id_connector(current, PROC_EVENT_UID);
843 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
849 * Samma på svenska..
851 asmlinkage long sys_setfsgid(gid_t gid)
853 struct cred *cred = current->cred;
856 old_fsgid = cred->fsgid;
857 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
860 if (gid == cred->gid || gid == cred->egid ||
861 gid == cred->sgid || gid == cred->fsgid ||
862 capable(CAP_SETGID)) {
863 if (gid != old_fsgid) {
864 set_dumpable(current->mm, suid_dumpable);
868 key_fsgid_changed(current);
869 proc_id_connector(current, PROC_EVENT_GID);
874 void do_sys_times(struct tms *tms)
876 struct task_cputime cputime;
877 cputime_t cutime, cstime;
879 spin_lock_irq(¤t->sighand->siglock);
880 thread_group_cputime(current, &cputime);
881 cutime = current->signal->cutime;
882 cstime = current->signal->cstime;
883 spin_unlock_irq(¤t->sighand->siglock);
884 tms->tms_utime = cputime_to_clock_t(cputime.utime);
885 tms->tms_stime = cputime_to_clock_t(cputime.stime);
886 tms->tms_cutime = cputime_to_clock_t(cutime);
887 tms->tms_cstime = cputime_to_clock_t(cstime);
890 asmlinkage long sys_times(struct tms __user * tbuf)
896 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
899 return (long) jiffies_64_to_clock_t(get_jiffies_64());
903 * This needs some heavy checking ...
904 * I just haven't the stomach for it. I also don't fully
905 * understand sessions/pgrp etc. Let somebody who does explain it.
907 * OK, I think I have the protection semantics right.... this is really
908 * only important on a multi-user system anyway, to make sure one user
909 * can't send a signal to a process owned by another. -TYT, 12/12/91
911 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
914 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
916 struct task_struct *p;
917 struct task_struct *group_leader = current->group_leader;
922 pid = task_pid_vnr(group_leader);
928 /* From this point forward we keep holding onto the tasklist lock
929 * so that our parent does not change from under us. -DaveM
931 write_lock_irq(&tasklist_lock);
934 p = find_task_by_vpid(pid);
939 if (!thread_group_leader(p))
942 if (same_thread_group(p->real_parent, group_leader)) {
944 if (task_session(p) != task_session(group_leader))
951 if (p != group_leader)
956 if (p->signal->leader)
961 struct task_struct *g;
963 pgrp = find_vpid(pgid);
964 g = pid_task(pgrp, PIDTYPE_PGID);
965 if (!g || task_session(g) != task_session(group_leader))
969 err = security_task_setpgid(p, pgid);
973 if (task_pgrp(p) != pgrp) {
974 change_pid(p, PIDTYPE_PGID, pgrp);
975 set_task_pgrp(p, pid_nr(pgrp));
980 /* All paths lead to here, thus we are safe. -DaveM */
981 write_unlock_irq(&tasklist_lock);
985 asmlinkage long sys_getpgid(pid_t pid)
987 struct task_struct *p;
993 grp = task_pgrp(current);
996 p = find_task_by_vpid(pid);
1003 retval = security_task_getpgid(p);
1007 retval = pid_vnr(grp);
1013 #ifdef __ARCH_WANT_SYS_GETPGRP
1015 asmlinkage long sys_getpgrp(void)
1017 return sys_getpgid(0);
1022 asmlinkage long sys_getsid(pid_t pid)
1024 struct task_struct *p;
1030 sid = task_session(current);
1033 p = find_task_by_vpid(pid);
1036 sid = task_session(p);
1040 retval = security_task_getsid(p);
1044 retval = pid_vnr(sid);
1050 asmlinkage long sys_setsid(void)
1052 struct task_struct *group_leader = current->group_leader;
1053 struct pid *sid = task_pid(group_leader);
1054 pid_t session = pid_vnr(sid);
1057 write_lock_irq(&tasklist_lock);
1058 /* Fail if I am already a session leader */
1059 if (group_leader->signal->leader)
1062 /* Fail if a process group id already exists that equals the
1063 * proposed session id.
1065 if (pid_task(sid, PIDTYPE_PGID))
1068 group_leader->signal->leader = 1;
1069 __set_special_pids(sid);
1071 proc_clear_tty(group_leader);
1075 write_unlock_irq(&tasklist_lock);
1080 * Supplementary group IDs
1083 /* init to 2 - one for init_task, one to ensure it is never freed */
1084 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1086 struct group_info *groups_alloc(int gidsetsize)
1088 struct group_info *group_info;
1092 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1093 /* Make sure we always allocate at least one indirect block pointer */
1094 nblocks = nblocks ? : 1;
1095 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1098 group_info->ngroups = gidsetsize;
1099 group_info->nblocks = nblocks;
1100 atomic_set(&group_info->usage, 1);
1102 if (gidsetsize <= NGROUPS_SMALL)
1103 group_info->blocks[0] = group_info->small_block;
1105 for (i = 0; i < nblocks; i++) {
1107 b = (void *)__get_free_page(GFP_USER);
1109 goto out_undo_partial_alloc;
1110 group_info->blocks[i] = b;
1115 out_undo_partial_alloc:
1117 free_page((unsigned long)group_info->blocks[i]);
1123 EXPORT_SYMBOL(groups_alloc);
1125 void groups_free(struct group_info *group_info)
1127 if (group_info->blocks[0] != group_info->small_block) {
1129 for (i = 0; i < group_info->nblocks; i++)
1130 free_page((unsigned long)group_info->blocks[i]);
1135 EXPORT_SYMBOL(groups_free);
1137 /* export the group_info to a user-space array */
1138 static int groups_to_user(gid_t __user *grouplist,
1139 struct group_info *group_info)
1142 unsigned int count = group_info->ngroups;
1144 for (i = 0; i < group_info->nblocks; i++) {
1145 unsigned int cp_count = min(NGROUPS_PER_BLOCK, count);
1146 unsigned int len = cp_count * sizeof(*grouplist);
1148 if (copy_to_user(grouplist, group_info->blocks[i], len))
1151 grouplist += NGROUPS_PER_BLOCK;
1157 /* fill a group_info from a user-space array - it must be allocated already */
1158 static int groups_from_user(struct group_info *group_info,
1159 gid_t __user *grouplist)
1162 unsigned int count = group_info->ngroups;
1164 for (i = 0; i < group_info->nblocks; i++) {
1165 unsigned int cp_count = min(NGROUPS_PER_BLOCK, count);
1166 unsigned int len = cp_count * sizeof(*grouplist);
1168 if (copy_from_user(group_info->blocks[i], grouplist, len))
1171 grouplist += NGROUPS_PER_BLOCK;
1177 /* a simple Shell sort */
1178 static void groups_sort(struct group_info *group_info)
1180 int base, max, stride;
1181 int gidsetsize = group_info->ngroups;
1183 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1188 max = gidsetsize - stride;
1189 for (base = 0; base < max; base++) {
1191 int right = left + stride;
1192 gid_t tmp = GROUP_AT(group_info, right);
1194 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1195 GROUP_AT(group_info, right) =
1196 GROUP_AT(group_info, left);
1200 GROUP_AT(group_info, right) = tmp;
1206 /* a simple bsearch */
1207 int groups_search(const struct group_info *group_info, gid_t grp)
1209 unsigned int left, right;
1215 right = group_info->ngroups;
1216 while (left < right) {
1217 unsigned int mid = (left+right)/2;
1218 int cmp = grp - GROUP_AT(group_info, mid);
1230 * set_groups - Change a group subscription in a security record
1231 * @sec: The security record to alter
1232 * @group_info: The group list to impose
1234 * Validate a group subscription and, if valid, impose it upon a task security
1237 int set_groups(struct cred *cred, struct group_info *group_info)
1240 struct group_info *old_info;
1242 retval = security_task_setgroups(group_info);
1246 groups_sort(group_info);
1247 get_group_info(group_info);
1249 spin_lock(&cred->lock);
1250 old_info = cred->group_info;
1251 cred->group_info = group_info;
1252 spin_unlock(&cred->lock);
1254 put_group_info(old_info);
1258 EXPORT_SYMBOL(set_groups);
1261 * set_current_groups - Change current's group subscription
1262 * @group_info: The group list to impose
1264 * Validate a group subscription and, if valid, impose it upon current's task
1267 int set_current_groups(struct group_info *group_info)
1269 return set_groups(current->cred, group_info);
1272 EXPORT_SYMBOL(set_current_groups);
1274 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1276 const struct cred *cred = current_cred();
1282 /* no need to grab task_lock here; it cannot change */
1283 i = cred->group_info->ngroups;
1285 if (i > gidsetsize) {
1289 if (groups_to_user(grouplist, cred->group_info)) {
1299 * SMP: Our groups are copy-on-write. We can set them safely
1300 * without another task interfering.
1303 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1305 struct group_info *group_info;
1308 if (!capable(CAP_SETGID))
1310 if ((unsigned)gidsetsize > NGROUPS_MAX)
1313 group_info = groups_alloc(gidsetsize);
1316 retval = groups_from_user(group_info, grouplist);
1318 put_group_info(group_info);
1322 retval = set_current_groups(group_info);
1323 put_group_info(group_info);
1329 * Check whether we're fsgid/egid or in the supplemental group..
1331 int in_group_p(gid_t grp)
1333 const struct cred *cred = current_cred();
1336 if (grp != cred->fsgid)
1337 retval = groups_search(cred->group_info, grp);
1341 EXPORT_SYMBOL(in_group_p);
1343 int in_egroup_p(gid_t grp)
1345 const struct cred *cred = current_cred();
1348 if (grp != cred->egid)
1349 retval = groups_search(cred->group_info, grp);
1353 EXPORT_SYMBOL(in_egroup_p);
1355 DECLARE_RWSEM(uts_sem);
1357 asmlinkage long sys_newuname(struct new_utsname __user * name)
1361 down_read(&uts_sem);
1362 if (copy_to_user(name, utsname(), sizeof *name))
1368 asmlinkage long sys_sethostname(char __user *name, int len)
1371 char tmp[__NEW_UTS_LEN];
1373 if (!capable(CAP_SYS_ADMIN))
1375 if (len < 0 || len > __NEW_UTS_LEN)
1377 down_write(&uts_sem);
1379 if (!copy_from_user(tmp, name, len)) {
1380 struct new_utsname *u = utsname();
1382 memcpy(u->nodename, tmp, len);
1383 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1390 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1392 asmlinkage long sys_gethostname(char __user *name, int len)
1395 struct new_utsname *u;
1399 down_read(&uts_sem);
1401 i = 1 + strlen(u->nodename);
1405 if (copy_to_user(name, u->nodename, i))
1414 * Only setdomainname; getdomainname can be implemented by calling
1417 asmlinkage long sys_setdomainname(char __user *name, int len)
1420 char tmp[__NEW_UTS_LEN];
1422 if (!capable(CAP_SYS_ADMIN))
1424 if (len < 0 || len > __NEW_UTS_LEN)
1427 down_write(&uts_sem);
1429 if (!copy_from_user(tmp, name, len)) {
1430 struct new_utsname *u = utsname();
1432 memcpy(u->domainname, tmp, len);
1433 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1440 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1442 if (resource >= RLIM_NLIMITS)
1445 struct rlimit value;
1446 task_lock(current->group_leader);
1447 value = current->signal->rlim[resource];
1448 task_unlock(current->group_leader);
1449 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1453 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1456 * Back compatibility for getrlimit. Needed for some apps.
1459 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1462 if (resource >= RLIM_NLIMITS)
1465 task_lock(current->group_leader);
1466 x = current->signal->rlim[resource];
1467 task_unlock(current->group_leader);
1468 if (x.rlim_cur > 0x7FFFFFFF)
1469 x.rlim_cur = 0x7FFFFFFF;
1470 if (x.rlim_max > 0x7FFFFFFF)
1471 x.rlim_max = 0x7FFFFFFF;
1472 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1477 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1479 struct rlimit new_rlim, *old_rlim;
1482 if (resource >= RLIM_NLIMITS)
1484 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1486 old_rlim = current->signal->rlim + resource;
1487 if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1488 !capable(CAP_SYS_RESOURCE))
1491 if (resource == RLIMIT_NOFILE) {
1492 if (new_rlim.rlim_max == RLIM_INFINITY)
1493 new_rlim.rlim_max = sysctl_nr_open;
1494 if (new_rlim.rlim_cur == RLIM_INFINITY)
1495 new_rlim.rlim_cur = sysctl_nr_open;
1496 if (new_rlim.rlim_max > sysctl_nr_open)
1500 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1503 retval = security_task_setrlimit(resource, &new_rlim);
1507 if (resource == RLIMIT_CPU && new_rlim.rlim_cur == 0) {
1509 * The caller is asking for an immediate RLIMIT_CPU
1510 * expiry. But we use the zero value to mean "it was
1511 * never set". So let's cheat and make it one second
1514 new_rlim.rlim_cur = 1;
1517 task_lock(current->group_leader);
1518 *old_rlim = new_rlim;
1519 task_unlock(current->group_leader);
1521 if (resource != RLIMIT_CPU)
1525 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1526 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1527 * very long-standing error, and fixing it now risks breakage of
1528 * applications, so we live with it
1530 if (new_rlim.rlim_cur == RLIM_INFINITY)
1533 update_rlimit_cpu(new_rlim.rlim_cur);
1539 * It would make sense to put struct rusage in the task_struct,
1540 * except that would make the task_struct be *really big*. After
1541 * task_struct gets moved into malloc'ed memory, it would
1542 * make sense to do this. It will make moving the rest of the information
1543 * a lot simpler! (Which we're not doing right now because we're not
1544 * measuring them yet).
1546 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1547 * races with threads incrementing their own counters. But since word
1548 * reads are atomic, we either get new values or old values and we don't
1549 * care which for the sums. We always take the siglock to protect reading
1550 * the c* fields from p->signal from races with exit.c updating those
1551 * fields when reaping, so a sample either gets all the additions of a
1552 * given child after it's reaped, or none so this sample is before reaping.
1555 * We need to take the siglock for CHILDEREN, SELF and BOTH
1556 * for the cases current multithreaded, non-current single threaded
1557 * non-current multithreaded. Thread traversal is now safe with
1559 * Strictly speaking, we donot need to take the siglock if we are current and
1560 * single threaded, as no one else can take our signal_struct away, no one
1561 * else can reap the children to update signal->c* counters, and no one else
1562 * can race with the signal-> fields. If we do not take any lock, the
1563 * signal-> fields could be read out of order while another thread was just
1564 * exiting. So we should place a read memory barrier when we avoid the lock.
1565 * On the writer side, write memory barrier is implied in __exit_signal
1566 * as __exit_signal releases the siglock spinlock after updating the signal->
1567 * fields. But we don't do this yet to keep things simple.
1571 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1573 r->ru_nvcsw += t->nvcsw;
1574 r->ru_nivcsw += t->nivcsw;
1575 r->ru_minflt += t->min_flt;
1576 r->ru_majflt += t->maj_flt;
1577 r->ru_inblock += task_io_get_inblock(t);
1578 r->ru_oublock += task_io_get_oublock(t);
1581 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1583 struct task_struct *t;
1584 unsigned long flags;
1585 cputime_t utime, stime;
1586 struct task_cputime cputime;
1588 memset((char *) r, 0, sizeof *r);
1589 utime = stime = cputime_zero;
1591 if (who == RUSAGE_THREAD) {
1592 accumulate_thread_rusage(p, r);
1596 if (!lock_task_sighand(p, &flags))
1601 case RUSAGE_CHILDREN:
1602 utime = p->signal->cutime;
1603 stime = p->signal->cstime;
1604 r->ru_nvcsw = p->signal->cnvcsw;
1605 r->ru_nivcsw = p->signal->cnivcsw;
1606 r->ru_minflt = p->signal->cmin_flt;
1607 r->ru_majflt = p->signal->cmaj_flt;
1608 r->ru_inblock = p->signal->cinblock;
1609 r->ru_oublock = p->signal->coublock;
1611 if (who == RUSAGE_CHILDREN)
1615 thread_group_cputime(p, &cputime);
1616 utime = cputime_add(utime, cputime.utime);
1617 stime = cputime_add(stime, cputime.stime);
1618 r->ru_nvcsw += p->signal->nvcsw;
1619 r->ru_nivcsw += p->signal->nivcsw;
1620 r->ru_minflt += p->signal->min_flt;
1621 r->ru_majflt += p->signal->maj_flt;
1622 r->ru_inblock += p->signal->inblock;
1623 r->ru_oublock += p->signal->oublock;
1626 accumulate_thread_rusage(t, r);
1634 unlock_task_sighand(p, &flags);
1637 cputime_to_timeval(utime, &r->ru_utime);
1638 cputime_to_timeval(stime, &r->ru_stime);
1641 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1644 k_getrusage(p, who, &r);
1645 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1648 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1650 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1651 who != RUSAGE_THREAD)
1653 return getrusage(current, who, ru);
1656 asmlinkage long sys_umask(int mask)
1658 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1662 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1663 unsigned long arg4, unsigned long arg5)
1665 struct task_struct *me = current;
1666 unsigned char comm[sizeof(me->comm)];
1669 if (security_task_prctl(option, arg2, arg3, arg4, arg5, &error))
1673 case PR_SET_PDEATHSIG:
1674 if (!valid_signal(arg2)) {
1678 me->pdeath_signal = arg2;
1681 case PR_GET_PDEATHSIG:
1682 error = put_user(me->pdeath_signal, (int __user *)arg2);
1684 case PR_GET_DUMPABLE:
1685 error = get_dumpable(me->mm);
1687 case PR_SET_DUMPABLE:
1688 if (arg2 < 0 || arg2 > 1) {
1692 set_dumpable(me->mm, arg2);
1696 case PR_SET_UNALIGN:
1697 error = SET_UNALIGN_CTL(me, arg2);
1699 case PR_GET_UNALIGN:
1700 error = GET_UNALIGN_CTL(me, arg2);
1703 error = SET_FPEMU_CTL(me, arg2);
1706 error = GET_FPEMU_CTL(me, arg2);
1709 error = SET_FPEXC_CTL(me, arg2);
1712 error = GET_FPEXC_CTL(me, arg2);
1715 error = PR_TIMING_STATISTICAL;
1718 if (arg2 != PR_TIMING_STATISTICAL)
1725 comm[sizeof(me->comm)-1] = 0;
1726 if (strncpy_from_user(comm, (char __user *)arg2,
1727 sizeof(me->comm) - 1) < 0)
1729 set_task_comm(me, comm);
1732 get_task_comm(comm, me);
1733 if (copy_to_user((char __user *)arg2, comm,
1738 error = GET_ENDIAN(me, arg2);
1741 error = SET_ENDIAN(me, arg2);
1744 case PR_GET_SECCOMP:
1745 error = prctl_get_seccomp();
1747 case PR_SET_SECCOMP:
1748 error = prctl_set_seccomp(arg2);
1751 error = GET_TSC_CTL(arg2);
1754 error = SET_TSC_CTL(arg2);
1756 case PR_GET_TIMERSLACK:
1757 error = current->timer_slack_ns;
1759 case PR_SET_TIMERSLACK:
1761 current->timer_slack_ns =
1762 current->default_timer_slack_ns;
1764 current->timer_slack_ns = arg2;
1774 asmlinkage long sys_getcpu(unsigned __user *cpup, unsigned __user *nodep,
1775 struct getcpu_cache __user *unused)
1778 int cpu = raw_smp_processor_id();
1780 err |= put_user(cpu, cpup);
1782 err |= put_user(cpu_to_node(cpu), nodep);
1783 return err ? -EFAULT : 0;
1786 char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff";
1788 static void argv_cleanup(char **argv, char **envp)
1794 * orderly_poweroff - Trigger an orderly system poweroff
1795 * @force: force poweroff if command execution fails
1797 * This may be called from any context to trigger a system shutdown.
1798 * If the orderly shutdown fails, it will force an immediate shutdown.
1800 int orderly_poweroff(bool force)
1803 char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc);
1804 static char *envp[] = {
1806 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
1810 struct subprocess_info *info;
1813 printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n",
1814 __func__, poweroff_cmd);
1818 info = call_usermodehelper_setup(argv[0], argv, envp, GFP_ATOMIC);
1824 call_usermodehelper_setcleanup(info, argv_cleanup);
1826 ret = call_usermodehelper_exec(info, UMH_NO_WAIT);
1830 printk(KERN_WARNING "Failed to start orderly shutdown: "
1831 "forcing the issue\n");
1833 /* I guess this should try to kick off some daemon to
1834 sync and poweroff asap. Or not even bother syncing
1835 if we're doing an emergency shutdown? */
1842 EXPORT_SYMBOL_GPL(orderly_poweroff);