4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/export.h>
9 #include <linux/utsname.h>
10 #include <linux/mman.h>
11 #include <linux/reboot.h>
12 #include <linux/prctl.h>
13 #include <linux/highuid.h>
15 #include <linux/kmod.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/file.h>
40 #include <linux/mount.h>
41 #include <linux/gfp.h>
42 #include <linux/syscore_ops.h>
43 #include <linux/version.h>
44 #include <linux/ctype.h>
46 #include <linux/compat.h>
47 #include <linux/syscalls.h>
48 #include <linux/kprobes.h>
49 #include <linux/user_namespace.h>
50 #include <linux/binfmts.h>
52 #include <linux/kmsg_dump.h>
53 /* Move somewhere else to avoid recompiling? */
54 #include <generated/utsrelease.h>
56 #include <asm/uaccess.h>
58 #include <asm/unistd.h>
60 #ifndef SET_UNALIGN_CTL
61 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
63 #ifndef GET_UNALIGN_CTL
64 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
67 # define SET_FPEMU_CTL(a,b) (-EINVAL)
70 # define GET_FPEMU_CTL(a,b) (-EINVAL)
73 # define SET_FPEXC_CTL(a,b) (-EINVAL)
76 # define GET_FPEXC_CTL(a,b) (-EINVAL)
79 # define GET_ENDIAN(a,b) (-EINVAL)
82 # define SET_ENDIAN(a,b) (-EINVAL)
85 # define GET_TSC_CTL(a) (-EINVAL)
88 # define SET_TSC_CTL(a) (-EINVAL)
92 * this is where the system-wide overflow UID and GID are defined, for
93 * architectures that now have 32-bit UID/GID but didn't in the past
96 int overflowuid
= DEFAULT_OVERFLOWUID
;
97 int overflowgid
= DEFAULT_OVERFLOWGID
;
99 EXPORT_SYMBOL(overflowuid
);
100 EXPORT_SYMBOL(overflowgid
);
103 * the same as above, but for filesystems which can only store a 16-bit
104 * UID and GID. as such, this is needed on all architectures
107 int fs_overflowuid
= DEFAULT_FS_OVERFLOWUID
;
108 int fs_overflowgid
= DEFAULT_FS_OVERFLOWUID
;
110 EXPORT_SYMBOL(fs_overflowuid
);
111 EXPORT_SYMBOL(fs_overflowgid
);
114 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
119 EXPORT_SYMBOL(cad_pid
);
122 * If set, this is used for preparing the system to power off.
125 void (*pm_power_off_prepare
)(void);
128 * Returns true if current's euid is same as p's uid or euid,
129 * or has CAP_SYS_NICE to p's user_ns.
131 * Called with rcu_read_lock, creds are safe
133 static bool set_one_prio_perm(struct task_struct
*p
)
135 const struct cred
*cred
= current_cred(), *pcred
= __task_cred(p
);
137 if (uid_eq(pcred
->uid
, cred
->euid
) ||
138 uid_eq(pcred
->euid
, cred
->euid
))
140 if (ns_capable(pcred
->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();
182 if (which
> PRIO_USER
|| which
< PRIO_PROCESS
)
185 /* normalize: avoid signed division (rounding problems) */
193 read_lock(&tasklist_lock
);
197 p
= find_task_by_vpid(who
);
201 error
= set_one_prio(p
, niceval
, error
);
205 pgrp
= find_vpid(who
);
207 pgrp
= task_pgrp(current
);
208 do_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
) {
209 error
= set_one_prio(p
, niceval
, error
);
210 } while_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
);
213 uid
= make_kuid(cred
->user_ns
, who
);
217 else if (!uid_eq(uid
, cred
->uid
) &&
218 !(user
= find_user(uid
)))
219 goto out_unlock
; /* No processes for this user */
221 do_each_thread(g
, p
) {
222 if (uid_eq(task_uid(p
), uid
))
223 error
= set_one_prio(p
, niceval
, error
);
224 } while_each_thread(g
, p
);
225 if (!uid_eq(uid
, cred
->uid
))
226 free_uid(user
); /* For find_user() */
230 read_unlock(&tasklist_lock
);
237 * Ugh. To avoid negative return values, "getpriority()" will
238 * not return the normal nice-value, but a negated value that
239 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
240 * to stay compatible.
242 SYSCALL_DEFINE2(getpriority
, int, which
, int, who
)
244 struct task_struct
*g
, *p
;
245 struct user_struct
*user
;
246 const struct cred
*cred
= current_cred();
247 long niceval
, retval
= -ESRCH
;
251 if (which
> PRIO_USER
|| which
< PRIO_PROCESS
)
255 read_lock(&tasklist_lock
);
259 p
= find_task_by_vpid(who
);
263 niceval
= 20 - task_nice(p
);
264 if (niceval
> retval
)
270 pgrp
= find_vpid(who
);
272 pgrp
= task_pgrp(current
);
273 do_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
) {
274 niceval
= 20 - task_nice(p
);
275 if (niceval
> retval
)
277 } while_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
);
280 uid
= make_kuid(cred
->user_ns
, who
);
284 else if (!uid_eq(uid
, cred
->uid
) &&
285 !(user
= find_user(uid
)))
286 goto out_unlock
; /* No processes for this user */
288 do_each_thread(g
, p
) {
289 if (uid_eq(task_uid(p
), uid
)) {
290 niceval
= 20 - task_nice(p
);
291 if (niceval
> retval
)
294 } while_each_thread(g
, p
);
295 if (!uid_eq(uid
, cred
->uid
))
296 free_uid(user
); /* for find_user() */
300 read_unlock(&tasklist_lock
);
307 * emergency_restart - reboot the system
309 * Without shutting down any hardware or taking any locks
310 * reboot the system. This is called when we know we are in
311 * trouble so this is our best effort to reboot. This is
312 * safe to call in interrupt context.
314 void emergency_restart(void)
316 kmsg_dump(KMSG_DUMP_EMERG
);
317 machine_emergency_restart();
319 EXPORT_SYMBOL_GPL(emergency_restart
);
321 void kernel_restart_prepare(char *cmd
)
323 blocking_notifier_call_chain(&reboot_notifier_list
, SYS_RESTART
, cmd
);
324 system_state
= SYSTEM_RESTART
;
325 usermodehelper_disable();
331 * register_reboot_notifier - Register function to be called at reboot time
332 * @nb: Info about notifier function to be called
334 * Registers a function with the list of functions
335 * to be called at reboot time.
337 * Currently always returns zero, as blocking_notifier_chain_register()
338 * always returns zero.
340 int register_reboot_notifier(struct notifier_block
*nb
)
342 return blocking_notifier_chain_register(&reboot_notifier_list
, nb
);
344 EXPORT_SYMBOL(register_reboot_notifier
);
347 * unregister_reboot_notifier - Unregister previously registered reboot notifier
348 * @nb: Hook to be unregistered
350 * Unregisters a previously registered reboot
353 * Returns zero on success, or %-ENOENT on failure.
355 int unregister_reboot_notifier(struct notifier_block
*nb
)
357 return blocking_notifier_chain_unregister(&reboot_notifier_list
, nb
);
359 EXPORT_SYMBOL(unregister_reboot_notifier
);
362 * kernel_restart - reboot the system
363 * @cmd: pointer to buffer containing command to execute for restart
366 * Shutdown everything and perform a clean reboot.
367 * This is not safe to call in interrupt context.
369 void kernel_restart(char *cmd
)
371 kernel_restart_prepare(cmd
);
372 disable_nonboot_cpus();
374 printk(KERN_EMERG
"Restarting system.\n");
376 printk(KERN_EMERG
"Restarting system with command '%s'.\n", cmd
);
377 kmsg_dump(KMSG_DUMP_RESTART
);
378 machine_restart(cmd
);
380 EXPORT_SYMBOL_GPL(kernel_restart
);
382 static void kernel_shutdown_prepare(enum system_states state
)
384 blocking_notifier_call_chain(&reboot_notifier_list
,
385 (state
== SYSTEM_HALT
)?SYS_HALT
:SYS_POWER_OFF
, NULL
);
386 system_state
= state
;
387 usermodehelper_disable();
391 * kernel_halt - halt the system
393 * Shutdown everything and perform a clean system halt.
395 void kernel_halt(void)
397 kernel_shutdown_prepare(SYSTEM_HALT
);
399 printk(KERN_EMERG
"System halted.\n");
400 kmsg_dump(KMSG_DUMP_HALT
);
404 EXPORT_SYMBOL_GPL(kernel_halt
);
407 * kernel_power_off - power_off the system
409 * Shutdown everything and perform a clean system power_off.
411 void kernel_power_off(void)
413 kernel_shutdown_prepare(SYSTEM_POWER_OFF
);
414 if (pm_power_off_prepare
)
415 pm_power_off_prepare();
416 disable_nonboot_cpus();
418 printk(KERN_EMERG
"Power down.\n");
419 kmsg_dump(KMSG_DUMP_POWEROFF
);
422 EXPORT_SYMBOL_GPL(kernel_power_off
);
424 static DEFINE_MUTEX(reboot_mutex
);
427 * Reboot system call: for obvious reasons only root may call it,
428 * and even root needs to set up some magic numbers in the registers
429 * so that some mistake won't make this reboot the whole machine.
430 * You can also set the meaning of the ctrl-alt-del-key here.
432 * reboot doesn't sync: do that yourself before calling this.
434 SYSCALL_DEFINE4(reboot
, int, magic1
, int, magic2
, unsigned int, cmd
,
440 /* We only trust the superuser with rebooting the system. */
441 if (!capable(CAP_SYS_BOOT
))
444 /* For safety, we require "magic" arguments. */
445 if (magic1
!= LINUX_REBOOT_MAGIC1
||
446 (magic2
!= LINUX_REBOOT_MAGIC2
&&
447 magic2
!= LINUX_REBOOT_MAGIC2A
&&
448 magic2
!= LINUX_REBOOT_MAGIC2B
&&
449 magic2
!= LINUX_REBOOT_MAGIC2C
))
453 * If pid namespaces are enabled and the current task is in a child
454 * pid_namespace, the command is handled by reboot_pid_ns() which will
457 ret
= reboot_pid_ns(task_active_pid_ns(current
), cmd
);
461 /* Instead of trying to make the power_off code look like
462 * halt when pm_power_off is not set do it the easy way.
464 if ((cmd
== LINUX_REBOOT_CMD_POWER_OFF
) && !pm_power_off
)
465 cmd
= LINUX_REBOOT_CMD_HALT
;
467 mutex_lock(&reboot_mutex
);
469 case LINUX_REBOOT_CMD_RESTART
:
470 kernel_restart(NULL
);
473 case LINUX_REBOOT_CMD_CAD_ON
:
477 case LINUX_REBOOT_CMD_CAD_OFF
:
481 case LINUX_REBOOT_CMD_HALT
:
484 panic("cannot halt");
486 case LINUX_REBOOT_CMD_POWER_OFF
:
491 case LINUX_REBOOT_CMD_RESTART2
:
492 if (strncpy_from_user(&buffer
[0], arg
, sizeof(buffer
) - 1) < 0) {
496 buffer
[sizeof(buffer
) - 1] = '\0';
498 kernel_restart(buffer
);
502 case LINUX_REBOOT_CMD_KEXEC
:
503 ret
= kernel_kexec();
507 #ifdef CONFIG_HIBERNATION
508 case LINUX_REBOOT_CMD_SW_SUSPEND
:
517 mutex_unlock(&reboot_mutex
);
521 static void deferred_cad(struct work_struct
*dummy
)
523 kernel_restart(NULL
);
527 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
528 * As it's called within an interrupt, it may NOT sync: the only choice
529 * is whether to reboot at once, or just ignore the ctrl-alt-del.
531 void ctrl_alt_del(void)
533 static DECLARE_WORK(cad_work
, deferred_cad
);
536 schedule_work(&cad_work
);
538 kill_cad_pid(SIGINT
, 1);
542 * Unprivileged users may change the real gid to the effective gid
543 * or vice versa. (BSD-style)
545 * If you set the real gid at all, or set the effective gid to a value not
546 * equal to the real gid, then the saved gid is set to the new effective gid.
548 * This makes it possible for a setgid program to completely drop its
549 * privileges, which is often a useful assertion to make when you are doing
550 * a security audit over a program.
552 * The general idea is that a program which uses just setregid() will be
553 * 100% compatible with BSD. A program which uses just setgid() will be
554 * 100% compatible with POSIX with saved IDs.
556 * SMP: There are not races, the GIDs are checked only by filesystem
557 * operations (as far as semantic preservation is concerned).
559 SYSCALL_DEFINE2(setregid
, gid_t
, rgid
, gid_t
, egid
)
561 struct user_namespace
*ns
= current_user_ns();
562 const struct cred
*old
;
567 krgid
= make_kgid(ns
, rgid
);
568 kegid
= make_kgid(ns
, egid
);
570 if ((rgid
!= (gid_t
) -1) && !gid_valid(krgid
))
572 if ((egid
!= (gid_t
) -1) && !gid_valid(kegid
))
575 new = prepare_creds();
578 old
= current_cred();
581 if (rgid
!= (gid_t
) -1) {
582 if (gid_eq(old
->gid
, krgid
) ||
583 gid_eq(old
->egid
, krgid
) ||
584 nsown_capable(CAP_SETGID
))
589 if (egid
!= (gid_t
) -1) {
590 if (gid_eq(old
->gid
, kegid
) ||
591 gid_eq(old
->egid
, kegid
) ||
592 gid_eq(old
->sgid
, kegid
) ||
593 nsown_capable(CAP_SETGID
))
599 if (rgid
!= (gid_t
) -1 ||
600 (egid
!= (gid_t
) -1 && !gid_eq(kegid
, old
->gid
)))
601 new->sgid
= new->egid
;
602 new->fsgid
= new->egid
;
604 return commit_creds(new);
612 * setgid() is implemented like SysV w/ SAVED_IDS
614 * SMP: Same implicit races as above.
616 SYSCALL_DEFINE1(setgid
, gid_t
, gid
)
618 struct user_namespace
*ns
= current_user_ns();
619 const struct cred
*old
;
624 kgid
= make_kgid(ns
, gid
);
625 if (!gid_valid(kgid
))
628 new = prepare_creds();
631 old
= current_cred();
634 if (nsown_capable(CAP_SETGID
))
635 new->gid
= new->egid
= new->sgid
= new->fsgid
= kgid
;
636 else if (gid_eq(kgid
, old
->gid
) || gid_eq(kgid
, old
->sgid
))
637 new->egid
= new->fsgid
= kgid
;
641 return commit_creds(new);
649 * change the user struct in a credentials set to match the new UID
651 static int set_user(struct cred
*new)
653 struct user_struct
*new_user
;
655 new_user
= alloc_uid(new->uid
);
660 * We don't fail in case of NPROC limit excess here because too many
661 * poorly written programs don't check set*uid() return code, assuming
662 * it never fails if called by root. We may still enforce NPROC limit
663 * for programs doing set*uid()+execve() by harmlessly deferring the
664 * failure to the execve() stage.
666 if (atomic_read(&new_user
->processes
) >= rlimit(RLIMIT_NPROC
) &&
667 new_user
!= INIT_USER
)
668 current
->flags
|= PF_NPROC_EXCEEDED
;
670 current
->flags
&= ~PF_NPROC_EXCEEDED
;
673 new->user
= new_user
;
678 * Unprivileged users may change the real uid to the effective uid
679 * or vice versa. (BSD-style)
681 * If you set the real uid at all, or set the effective uid to a value not
682 * equal to the real uid, then the saved uid is set to the new effective uid.
684 * This makes it possible for a setuid program to completely drop its
685 * privileges, which is often a useful assertion to make when you are doing
686 * a security audit over a program.
688 * The general idea is that a program which uses just setreuid() will be
689 * 100% compatible with BSD. A program which uses just setuid() will be
690 * 100% compatible with POSIX with saved IDs.
692 SYSCALL_DEFINE2(setreuid
, uid_t
, ruid
, uid_t
, euid
)
694 struct user_namespace
*ns
= current_user_ns();
695 const struct cred
*old
;
700 kruid
= make_kuid(ns
, ruid
);
701 keuid
= make_kuid(ns
, euid
);
703 if ((ruid
!= (uid_t
) -1) && !uid_valid(kruid
))
705 if ((euid
!= (uid_t
) -1) && !uid_valid(keuid
))
708 new = prepare_creds();
711 old
= current_cred();
714 if (ruid
!= (uid_t
) -1) {
716 if (!uid_eq(old
->uid
, kruid
) &&
717 !uid_eq(old
->euid
, kruid
) &&
718 !nsown_capable(CAP_SETUID
))
722 if (euid
!= (uid_t
) -1) {
724 if (!uid_eq(old
->uid
, keuid
) &&
725 !uid_eq(old
->euid
, keuid
) &&
726 !uid_eq(old
->suid
, keuid
) &&
727 !nsown_capable(CAP_SETUID
))
731 if (!uid_eq(new->uid
, old
->uid
)) {
732 retval
= set_user(new);
736 if (ruid
!= (uid_t
) -1 ||
737 (euid
!= (uid_t
) -1 && !uid_eq(keuid
, old
->uid
)))
738 new->suid
= new->euid
;
739 new->fsuid
= new->euid
;
741 retval
= security_task_fix_setuid(new, old
, LSM_SETID_RE
);
745 return commit_creds(new);
753 * setuid() is implemented like SysV with SAVED_IDS
755 * Note that SAVED_ID's is deficient in that a setuid root program
756 * like sendmail, for example, cannot set its uid to be a normal
757 * user and then switch back, because if you're root, setuid() sets
758 * the saved uid too. If you don't like this, blame the bright people
759 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
760 * will allow a root program to temporarily drop privileges and be able to
761 * regain them by swapping the real and effective uid.
763 SYSCALL_DEFINE1(setuid
, uid_t
, uid
)
765 struct user_namespace
*ns
= current_user_ns();
766 const struct cred
*old
;
771 kuid
= make_kuid(ns
, uid
);
772 if (!uid_valid(kuid
))
775 new = prepare_creds();
778 old
= current_cred();
781 if (nsown_capable(CAP_SETUID
)) {
782 new->suid
= new->uid
= kuid
;
783 if (!uid_eq(kuid
, old
->uid
)) {
784 retval
= set_user(new);
788 } else if (!uid_eq(kuid
, old
->uid
) && !uid_eq(kuid
, new->suid
)) {
792 new->fsuid
= new->euid
= kuid
;
794 retval
= security_task_fix_setuid(new, old
, LSM_SETID_ID
);
798 return commit_creds(new);
807 * This function implements a generic ability to update ruid, euid,
808 * and suid. This allows you to implement the 4.4 compatible seteuid().
810 SYSCALL_DEFINE3(setresuid
, uid_t
, ruid
, uid_t
, euid
, uid_t
, suid
)
812 struct user_namespace
*ns
= current_user_ns();
813 const struct cred
*old
;
816 kuid_t kruid
, keuid
, ksuid
;
818 kruid
= make_kuid(ns
, ruid
);
819 keuid
= make_kuid(ns
, euid
);
820 ksuid
= make_kuid(ns
, suid
);
822 if ((ruid
!= (uid_t
) -1) && !uid_valid(kruid
))
825 if ((euid
!= (uid_t
) -1) && !uid_valid(keuid
))
828 if ((suid
!= (uid_t
) -1) && !uid_valid(ksuid
))
831 new = prepare_creds();
835 old
= current_cred();
838 if (!nsown_capable(CAP_SETUID
)) {
839 if (ruid
!= (uid_t
) -1 && !uid_eq(kruid
, old
->uid
) &&
840 !uid_eq(kruid
, old
->euid
) && !uid_eq(kruid
, old
->suid
))
842 if (euid
!= (uid_t
) -1 && !uid_eq(keuid
, old
->uid
) &&
843 !uid_eq(keuid
, old
->euid
) && !uid_eq(keuid
, old
->suid
))
845 if (suid
!= (uid_t
) -1 && !uid_eq(ksuid
, old
->uid
) &&
846 !uid_eq(ksuid
, old
->euid
) && !uid_eq(ksuid
, old
->suid
))
850 if (ruid
!= (uid_t
) -1) {
852 if (!uid_eq(kruid
, old
->uid
)) {
853 retval
= set_user(new);
858 if (euid
!= (uid_t
) -1)
860 if (suid
!= (uid_t
) -1)
862 new->fsuid
= new->euid
;
864 retval
= security_task_fix_setuid(new, old
, LSM_SETID_RES
);
868 return commit_creds(new);
875 SYSCALL_DEFINE3(getresuid
, uid_t __user
*, ruidp
, uid_t __user
*, euidp
, uid_t __user
*, suidp
)
877 const struct cred
*cred
= current_cred();
879 uid_t ruid
, euid
, suid
;
881 ruid
= from_kuid_munged(cred
->user_ns
, cred
->uid
);
882 euid
= from_kuid_munged(cred
->user_ns
, cred
->euid
);
883 suid
= from_kuid_munged(cred
->user_ns
, cred
->suid
);
885 if (!(retval
= put_user(ruid
, ruidp
)) &&
886 !(retval
= put_user(euid
, euidp
)))
887 retval
= put_user(suid
, suidp
);
893 * Same as above, but for rgid, egid, sgid.
895 SYSCALL_DEFINE3(setresgid
, gid_t
, rgid
, gid_t
, egid
, gid_t
, sgid
)
897 struct user_namespace
*ns
= current_user_ns();
898 const struct cred
*old
;
901 kgid_t krgid
, kegid
, ksgid
;
903 krgid
= make_kgid(ns
, rgid
);
904 kegid
= make_kgid(ns
, egid
);
905 ksgid
= make_kgid(ns
, sgid
);
907 if ((rgid
!= (gid_t
) -1) && !gid_valid(krgid
))
909 if ((egid
!= (gid_t
) -1) && !gid_valid(kegid
))
911 if ((sgid
!= (gid_t
) -1) && !gid_valid(ksgid
))
914 new = prepare_creds();
917 old
= current_cred();
920 if (!nsown_capable(CAP_SETGID
)) {
921 if (rgid
!= (gid_t
) -1 && !gid_eq(krgid
, old
->gid
) &&
922 !gid_eq(krgid
, old
->egid
) && !gid_eq(krgid
, old
->sgid
))
924 if (egid
!= (gid_t
) -1 && !gid_eq(kegid
, old
->gid
) &&
925 !gid_eq(kegid
, old
->egid
) && !gid_eq(kegid
, old
->sgid
))
927 if (sgid
!= (gid_t
) -1 && !gid_eq(ksgid
, old
->gid
) &&
928 !gid_eq(ksgid
, old
->egid
) && !gid_eq(ksgid
, old
->sgid
))
932 if (rgid
!= (gid_t
) -1)
934 if (egid
!= (gid_t
) -1)
936 if (sgid
!= (gid_t
) -1)
938 new->fsgid
= new->egid
;
940 return commit_creds(new);
947 SYSCALL_DEFINE3(getresgid
, gid_t __user
*, rgidp
, gid_t __user
*, egidp
, gid_t __user
*, sgidp
)
949 const struct cred
*cred
= current_cred();
951 gid_t rgid
, egid
, sgid
;
953 rgid
= from_kgid_munged(cred
->user_ns
, cred
->gid
);
954 egid
= from_kgid_munged(cred
->user_ns
, cred
->egid
);
955 sgid
= from_kgid_munged(cred
->user_ns
, cred
->sgid
);
957 if (!(retval
= put_user(rgid
, rgidp
)) &&
958 !(retval
= put_user(egid
, egidp
)))
959 retval
= put_user(sgid
, sgidp
);
966 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
967 * is used for "access()" and for the NFS daemon (letting nfsd stay at
968 * whatever uid it wants to). It normally shadows "euid", except when
969 * explicitly set by setfsuid() or for access..
971 SYSCALL_DEFINE1(setfsuid
, uid_t
, uid
)
973 const struct cred
*old
;
978 old
= current_cred();
979 old_fsuid
= from_kuid_munged(old
->user_ns
, old
->fsuid
);
981 kuid
= make_kuid(old
->user_ns
, uid
);
982 if (!uid_valid(kuid
))
985 new = prepare_creds();
989 if (uid_eq(kuid
, old
->uid
) || uid_eq(kuid
, old
->euid
) ||
990 uid_eq(kuid
, old
->suid
) || uid_eq(kuid
, old
->fsuid
) ||
991 nsown_capable(CAP_SETUID
)) {
992 if (!uid_eq(kuid
, old
->fsuid
)) {
994 if (security_task_fix_setuid(new, old
, LSM_SETID_FS
) == 0)
1008 * Samma på svenska..
1010 SYSCALL_DEFINE1(setfsgid
, gid_t
, gid
)
1012 const struct cred
*old
;
1017 old
= current_cred();
1018 old_fsgid
= from_kgid_munged(old
->user_ns
, old
->fsgid
);
1020 kgid
= make_kgid(old
->user_ns
, gid
);
1021 if (!gid_valid(kgid
))
1024 new = prepare_creds();
1028 if (gid_eq(kgid
, old
->gid
) || gid_eq(kgid
, old
->egid
) ||
1029 gid_eq(kgid
, old
->sgid
) || gid_eq(kgid
, old
->fsgid
) ||
1030 nsown_capable(CAP_SETGID
)) {
1031 if (!gid_eq(kgid
, old
->fsgid
)) {
1045 void do_sys_times(struct tms
*tms
)
1047 cputime_t tgutime
, tgstime
, cutime
, cstime
;
1049 spin_lock_irq(¤t
->sighand
->siglock
);
1050 thread_group_cputime_adjusted(current
, &tgutime
, &tgstime
);
1051 cutime
= current
->signal
->cutime
;
1052 cstime
= current
->signal
->cstime
;
1053 spin_unlock_irq(¤t
->sighand
->siglock
);
1054 tms
->tms_utime
= cputime_to_clock_t(tgutime
);
1055 tms
->tms_stime
= cputime_to_clock_t(tgstime
);
1056 tms
->tms_cutime
= cputime_to_clock_t(cutime
);
1057 tms
->tms_cstime
= cputime_to_clock_t(cstime
);
1060 SYSCALL_DEFINE1(times
, struct tms __user
*, tbuf
)
1066 if (copy_to_user(tbuf
, &tmp
, sizeof(struct tms
)))
1069 force_successful_syscall_return();
1070 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1074 * This needs some heavy checking ...
1075 * I just haven't the stomach for it. I also don't fully
1076 * understand sessions/pgrp etc. Let somebody who does explain it.
1078 * OK, I think I have the protection semantics right.... this is really
1079 * only important on a multi-user system anyway, to make sure one user
1080 * can't send a signal to a process owned by another. -TYT, 12/12/91
1082 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1085 SYSCALL_DEFINE2(setpgid
, pid_t
, pid
, pid_t
, pgid
)
1087 struct task_struct
*p
;
1088 struct task_struct
*group_leader
= current
->group_leader
;
1093 pid
= task_pid_vnr(group_leader
);
1100 /* From this point forward we keep holding onto the tasklist lock
1101 * so that our parent does not change from under us. -DaveM
1103 write_lock_irq(&tasklist_lock
);
1106 p
= find_task_by_vpid(pid
);
1111 if (!thread_group_leader(p
))
1114 if (same_thread_group(p
->real_parent
, group_leader
)) {
1116 if (task_session(p
) != task_session(group_leader
))
1123 if (p
!= group_leader
)
1128 if (p
->signal
->leader
)
1133 struct task_struct
*g
;
1135 pgrp
= find_vpid(pgid
);
1136 g
= pid_task(pgrp
, PIDTYPE_PGID
);
1137 if (!g
|| task_session(g
) != task_session(group_leader
))
1141 err
= security_task_setpgid(p
, pgid
);
1145 if (task_pgrp(p
) != pgrp
)
1146 change_pid(p
, PIDTYPE_PGID
, pgrp
);
1150 /* All paths lead to here, thus we are safe. -DaveM */
1151 write_unlock_irq(&tasklist_lock
);
1156 SYSCALL_DEFINE1(getpgid
, pid_t
, pid
)
1158 struct task_struct
*p
;
1164 grp
= task_pgrp(current
);
1167 p
= find_task_by_vpid(pid
);
1174 retval
= security_task_getpgid(p
);
1178 retval
= pid_vnr(grp
);
1184 #ifdef __ARCH_WANT_SYS_GETPGRP
1186 SYSCALL_DEFINE0(getpgrp
)
1188 return sys_getpgid(0);
1193 SYSCALL_DEFINE1(getsid
, pid_t
, pid
)
1195 struct task_struct
*p
;
1201 sid
= task_session(current
);
1204 p
= find_task_by_vpid(pid
);
1207 sid
= task_session(p
);
1211 retval
= security_task_getsid(p
);
1215 retval
= pid_vnr(sid
);
1221 SYSCALL_DEFINE0(setsid
)
1223 struct task_struct
*group_leader
= current
->group_leader
;
1224 struct pid
*sid
= task_pid(group_leader
);
1225 pid_t session
= pid_vnr(sid
);
1228 write_lock_irq(&tasklist_lock
);
1229 /* Fail if I am already a session leader */
1230 if (group_leader
->signal
->leader
)
1233 /* Fail if a process group id already exists that equals the
1234 * proposed session id.
1236 if (pid_task(sid
, PIDTYPE_PGID
))
1239 group_leader
->signal
->leader
= 1;
1240 __set_special_pids(sid
);
1242 proc_clear_tty(group_leader
);
1246 write_unlock_irq(&tasklist_lock
);
1248 proc_sid_connector(group_leader
);
1249 sched_autogroup_create_attach(group_leader
);
1254 DECLARE_RWSEM(uts_sem
);
1256 #ifdef COMPAT_UTS_MACHINE
1257 #define override_architecture(name) \
1258 (personality(current->personality) == PER_LINUX32 && \
1259 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1260 sizeof(COMPAT_UTS_MACHINE)))
1262 #define override_architecture(name) 0
1266 * Work around broken programs that cannot handle "Linux 3.0".
1267 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1269 static int override_release(char __user
*release
, size_t len
)
1273 if (current
->personality
& UNAME26
) {
1274 const char *rest
= UTS_RELEASE
;
1275 char buf
[65] = { 0 };
1281 if (*rest
== '.' && ++ndots
>= 3)
1283 if (!isdigit(*rest
) && *rest
!= '.')
1287 v
= ((LINUX_VERSION_CODE
>> 8) & 0xff) + 40;
1288 copy
= clamp_t(size_t, len
, 1, sizeof(buf
));
1289 copy
= scnprintf(buf
, copy
, "2.6.%u%s", v
, rest
);
1290 ret
= copy_to_user(release
, buf
, copy
+ 1);
1295 SYSCALL_DEFINE1(newuname
, struct new_utsname __user
*, name
)
1299 down_read(&uts_sem
);
1300 if (copy_to_user(name
, utsname(), sizeof *name
))
1304 if (!errno
&& override_release(name
->release
, sizeof(name
->release
)))
1306 if (!errno
&& override_architecture(name
))
1311 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1315 SYSCALL_DEFINE1(uname
, struct old_utsname __user
*, name
)
1322 down_read(&uts_sem
);
1323 if (copy_to_user(name
, utsname(), sizeof(*name
)))
1327 if (!error
&& override_release(name
->release
, sizeof(name
->release
)))
1329 if (!error
&& override_architecture(name
))
1334 SYSCALL_DEFINE1(olduname
, struct oldold_utsname __user
*, name
)
1340 if (!access_ok(VERIFY_WRITE
, name
, sizeof(struct oldold_utsname
)))
1343 down_read(&uts_sem
);
1344 error
= __copy_to_user(&name
->sysname
, &utsname()->sysname
,
1346 error
|= __put_user(0, name
->sysname
+ __OLD_UTS_LEN
);
1347 error
|= __copy_to_user(&name
->nodename
, &utsname()->nodename
,
1349 error
|= __put_user(0, name
->nodename
+ __OLD_UTS_LEN
);
1350 error
|= __copy_to_user(&name
->release
, &utsname()->release
,
1352 error
|= __put_user(0, name
->release
+ __OLD_UTS_LEN
);
1353 error
|= __copy_to_user(&name
->version
, &utsname()->version
,
1355 error
|= __put_user(0, name
->version
+ __OLD_UTS_LEN
);
1356 error
|= __copy_to_user(&name
->machine
, &utsname()->machine
,
1358 error
|= __put_user(0, name
->machine
+ __OLD_UTS_LEN
);
1361 if (!error
&& override_architecture(name
))
1363 if (!error
&& override_release(name
->release
, sizeof(name
->release
)))
1365 return error
? -EFAULT
: 0;
1369 SYSCALL_DEFINE2(sethostname
, char __user
*, name
, int, len
)
1372 char tmp
[__NEW_UTS_LEN
];
1374 if (!ns_capable(current
->nsproxy
->uts_ns
->user_ns
, CAP_SYS_ADMIN
))
1377 if (len
< 0 || len
> __NEW_UTS_LEN
)
1379 down_write(&uts_sem
);
1381 if (!copy_from_user(tmp
, name
, len
)) {
1382 struct new_utsname
*u
= utsname();
1384 memcpy(u
->nodename
, tmp
, len
);
1385 memset(u
->nodename
+ len
, 0, sizeof(u
->nodename
) - len
);
1387 uts_proc_notify(UTS_PROC_HOSTNAME
);
1393 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1395 SYSCALL_DEFINE2(gethostname
, char __user
*, name
, int, len
)
1398 struct new_utsname
*u
;
1402 down_read(&uts_sem
);
1404 i
= 1 + strlen(u
->nodename
);
1408 if (copy_to_user(name
, u
->nodename
, i
))
1417 * Only setdomainname; getdomainname can be implemented by calling
1420 SYSCALL_DEFINE2(setdomainname
, char __user
*, name
, int, len
)
1423 char tmp
[__NEW_UTS_LEN
];
1425 if (!ns_capable(current
->nsproxy
->uts_ns
->user_ns
, CAP_SYS_ADMIN
))
1427 if (len
< 0 || len
> __NEW_UTS_LEN
)
1430 down_write(&uts_sem
);
1432 if (!copy_from_user(tmp
, name
, len
)) {
1433 struct new_utsname
*u
= utsname();
1435 memcpy(u
->domainname
, tmp
, len
);
1436 memset(u
->domainname
+ len
, 0, sizeof(u
->domainname
) - len
);
1438 uts_proc_notify(UTS_PROC_DOMAINNAME
);
1444 SYSCALL_DEFINE2(getrlimit
, unsigned int, resource
, struct rlimit __user
*, rlim
)
1446 struct rlimit value
;
1449 ret
= do_prlimit(current
, resource
, NULL
, &value
);
1451 ret
= copy_to_user(rlim
, &value
, sizeof(*rlim
)) ? -EFAULT
: 0;
1456 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1459 * Back compatibility for getrlimit. Needed for some apps.
1462 SYSCALL_DEFINE2(old_getrlimit
, unsigned int, resource
,
1463 struct rlimit __user
*, rlim
)
1466 if (resource
>= RLIM_NLIMITS
)
1469 task_lock(current
->group_leader
);
1470 x
= current
->signal
->rlim
[resource
];
1471 task_unlock(current
->group_leader
);
1472 if (x
.rlim_cur
> 0x7FFFFFFF)
1473 x
.rlim_cur
= 0x7FFFFFFF;
1474 if (x
.rlim_max
> 0x7FFFFFFF)
1475 x
.rlim_max
= 0x7FFFFFFF;
1476 return copy_to_user(rlim
, &x
, sizeof(x
))?-EFAULT
:0;
1481 static inline bool rlim64_is_infinity(__u64 rlim64
)
1483 #if BITS_PER_LONG < 64
1484 return rlim64
>= ULONG_MAX
;
1486 return rlim64
== RLIM64_INFINITY
;
1490 static void rlim_to_rlim64(const struct rlimit
*rlim
, struct rlimit64
*rlim64
)
1492 if (rlim
->rlim_cur
== RLIM_INFINITY
)
1493 rlim64
->rlim_cur
= RLIM64_INFINITY
;
1495 rlim64
->rlim_cur
= rlim
->rlim_cur
;
1496 if (rlim
->rlim_max
== RLIM_INFINITY
)
1497 rlim64
->rlim_max
= RLIM64_INFINITY
;
1499 rlim64
->rlim_max
= rlim
->rlim_max
;
1502 static void rlim64_to_rlim(const struct rlimit64
*rlim64
, struct rlimit
*rlim
)
1504 if (rlim64_is_infinity(rlim64
->rlim_cur
))
1505 rlim
->rlim_cur
= RLIM_INFINITY
;
1507 rlim
->rlim_cur
= (unsigned long)rlim64
->rlim_cur
;
1508 if (rlim64_is_infinity(rlim64
->rlim_max
))
1509 rlim
->rlim_max
= RLIM_INFINITY
;
1511 rlim
->rlim_max
= (unsigned long)rlim64
->rlim_max
;
1514 /* make sure you are allowed to change @tsk limits before calling this */
1515 int do_prlimit(struct task_struct
*tsk
, unsigned int resource
,
1516 struct rlimit
*new_rlim
, struct rlimit
*old_rlim
)
1518 struct rlimit
*rlim
;
1521 if (resource
>= RLIM_NLIMITS
)
1524 if (new_rlim
->rlim_cur
> new_rlim
->rlim_max
)
1526 if (resource
== RLIMIT_NOFILE
&&
1527 new_rlim
->rlim_max
> sysctl_nr_open
)
1531 /* protect tsk->signal and tsk->sighand from disappearing */
1532 read_lock(&tasklist_lock
);
1533 if (!tsk
->sighand
) {
1538 rlim
= tsk
->signal
->rlim
+ resource
;
1539 task_lock(tsk
->group_leader
);
1541 /* Keep the capable check against init_user_ns until
1542 cgroups can contain all limits */
1543 if (new_rlim
->rlim_max
> rlim
->rlim_max
&&
1544 !capable(CAP_SYS_RESOURCE
))
1547 retval
= security_task_setrlimit(tsk
->group_leader
,
1548 resource
, new_rlim
);
1549 if (resource
== RLIMIT_CPU
&& new_rlim
->rlim_cur
== 0) {
1551 * The caller is asking for an immediate RLIMIT_CPU
1552 * expiry. But we use the zero value to mean "it was
1553 * never set". So let's cheat and make it one second
1556 new_rlim
->rlim_cur
= 1;
1565 task_unlock(tsk
->group_leader
);
1568 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1569 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1570 * very long-standing error, and fixing it now risks breakage of
1571 * applications, so we live with it
1573 if (!retval
&& new_rlim
&& resource
== RLIMIT_CPU
&&
1574 new_rlim
->rlim_cur
!= RLIM_INFINITY
)
1575 update_rlimit_cpu(tsk
, new_rlim
->rlim_cur
);
1577 read_unlock(&tasklist_lock
);
1581 /* rcu lock must be held */
1582 static int check_prlimit_permission(struct task_struct
*task
)
1584 const struct cred
*cred
= current_cred(), *tcred
;
1586 if (current
== task
)
1589 tcred
= __task_cred(task
);
1590 if (uid_eq(cred
->uid
, tcred
->euid
) &&
1591 uid_eq(cred
->uid
, tcred
->suid
) &&
1592 uid_eq(cred
->uid
, tcred
->uid
) &&
1593 gid_eq(cred
->gid
, tcred
->egid
) &&
1594 gid_eq(cred
->gid
, tcred
->sgid
) &&
1595 gid_eq(cred
->gid
, tcred
->gid
))
1597 if (ns_capable(tcred
->user_ns
, CAP_SYS_RESOURCE
))
1603 SYSCALL_DEFINE4(prlimit64
, pid_t
, pid
, unsigned int, resource
,
1604 const struct rlimit64 __user
*, new_rlim
,
1605 struct rlimit64 __user
*, old_rlim
)
1607 struct rlimit64 old64
, new64
;
1608 struct rlimit old
, new;
1609 struct task_struct
*tsk
;
1613 if (copy_from_user(&new64
, new_rlim
, sizeof(new64
)))
1615 rlim64_to_rlim(&new64
, &new);
1619 tsk
= pid
? find_task_by_vpid(pid
) : current
;
1624 ret
= check_prlimit_permission(tsk
);
1629 get_task_struct(tsk
);
1632 ret
= do_prlimit(tsk
, resource
, new_rlim
? &new : NULL
,
1633 old_rlim
? &old
: NULL
);
1635 if (!ret
&& old_rlim
) {
1636 rlim_to_rlim64(&old
, &old64
);
1637 if (copy_to_user(old_rlim
, &old64
, sizeof(old64
)))
1641 put_task_struct(tsk
);
1645 SYSCALL_DEFINE2(setrlimit
, unsigned int, resource
, struct rlimit __user
*, rlim
)
1647 struct rlimit new_rlim
;
1649 if (copy_from_user(&new_rlim
, rlim
, sizeof(*rlim
)))
1651 return do_prlimit(current
, resource
, &new_rlim
, NULL
);
1655 * It would make sense to put struct rusage in the task_struct,
1656 * except that would make the task_struct be *really big*. After
1657 * task_struct gets moved into malloc'ed memory, it would
1658 * make sense to do this. It will make moving the rest of the information
1659 * a lot simpler! (Which we're not doing right now because we're not
1660 * measuring them yet).
1662 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1663 * races with threads incrementing their own counters. But since word
1664 * reads are atomic, we either get new values or old values and we don't
1665 * care which for the sums. We always take the siglock to protect reading
1666 * the c* fields from p->signal from races with exit.c updating those
1667 * fields when reaping, so a sample either gets all the additions of a
1668 * given child after it's reaped, or none so this sample is before reaping.
1671 * We need to take the siglock for CHILDEREN, SELF and BOTH
1672 * for the cases current multithreaded, non-current single threaded
1673 * non-current multithreaded. Thread traversal is now safe with
1675 * Strictly speaking, we donot need to take the siglock if we are current and
1676 * single threaded, as no one else can take our signal_struct away, no one
1677 * else can reap the children to update signal->c* counters, and no one else
1678 * can race with the signal-> fields. If we do not take any lock, the
1679 * signal-> fields could be read out of order while another thread was just
1680 * exiting. So we should place a read memory barrier when we avoid the lock.
1681 * On the writer side, write memory barrier is implied in __exit_signal
1682 * as __exit_signal releases the siglock spinlock after updating the signal->
1683 * fields. But we don't do this yet to keep things simple.
1687 static void accumulate_thread_rusage(struct task_struct
*t
, struct rusage
*r
)
1689 r
->ru_nvcsw
+= t
->nvcsw
;
1690 r
->ru_nivcsw
+= t
->nivcsw
;
1691 r
->ru_minflt
+= t
->min_flt
;
1692 r
->ru_majflt
+= t
->maj_flt
;
1693 r
->ru_inblock
+= task_io_get_inblock(t
);
1694 r
->ru_oublock
+= task_io_get_oublock(t
);
1697 static void k_getrusage(struct task_struct
*p
, int who
, struct rusage
*r
)
1699 struct task_struct
*t
;
1700 unsigned long flags
;
1701 cputime_t tgutime
, tgstime
, utime
, stime
;
1702 unsigned long maxrss
= 0;
1704 memset((char *) r
, 0, sizeof *r
);
1707 if (who
== RUSAGE_THREAD
) {
1708 task_cputime_adjusted(current
, &utime
, &stime
);
1709 accumulate_thread_rusage(p
, r
);
1710 maxrss
= p
->signal
->maxrss
;
1714 if (!lock_task_sighand(p
, &flags
))
1719 case RUSAGE_CHILDREN
:
1720 utime
= p
->signal
->cutime
;
1721 stime
= p
->signal
->cstime
;
1722 r
->ru_nvcsw
= p
->signal
->cnvcsw
;
1723 r
->ru_nivcsw
= p
->signal
->cnivcsw
;
1724 r
->ru_minflt
= p
->signal
->cmin_flt
;
1725 r
->ru_majflt
= p
->signal
->cmaj_flt
;
1726 r
->ru_inblock
= p
->signal
->cinblock
;
1727 r
->ru_oublock
= p
->signal
->coublock
;
1728 maxrss
= p
->signal
->cmaxrss
;
1730 if (who
== RUSAGE_CHILDREN
)
1734 thread_group_cputime_adjusted(p
, &tgutime
, &tgstime
);
1737 r
->ru_nvcsw
+= p
->signal
->nvcsw
;
1738 r
->ru_nivcsw
+= p
->signal
->nivcsw
;
1739 r
->ru_minflt
+= p
->signal
->min_flt
;
1740 r
->ru_majflt
+= p
->signal
->maj_flt
;
1741 r
->ru_inblock
+= p
->signal
->inblock
;
1742 r
->ru_oublock
+= p
->signal
->oublock
;
1743 if (maxrss
< p
->signal
->maxrss
)
1744 maxrss
= p
->signal
->maxrss
;
1747 accumulate_thread_rusage(t
, r
);
1755 unlock_task_sighand(p
, &flags
);
1758 cputime_to_timeval(utime
, &r
->ru_utime
);
1759 cputime_to_timeval(stime
, &r
->ru_stime
);
1761 if (who
!= RUSAGE_CHILDREN
) {
1762 struct mm_struct
*mm
= get_task_mm(p
);
1764 setmax_mm_hiwater_rss(&maxrss
, mm
);
1768 r
->ru_maxrss
= maxrss
* (PAGE_SIZE
/ 1024); /* convert pages to KBs */
1771 int getrusage(struct task_struct
*p
, int who
, struct rusage __user
*ru
)
1774 k_getrusage(p
, who
, &r
);
1775 return copy_to_user(ru
, &r
, sizeof(r
)) ? -EFAULT
: 0;
1778 SYSCALL_DEFINE2(getrusage
, int, who
, struct rusage __user
*, ru
)
1780 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
&&
1781 who
!= RUSAGE_THREAD
)
1783 return getrusage(current
, who
, ru
);
1786 SYSCALL_DEFINE1(umask
, int, mask
)
1788 mask
= xchg(¤t
->fs
->umask
, mask
& S_IRWXUGO
);
1792 #ifdef CONFIG_CHECKPOINT_RESTORE
1793 static int prctl_set_mm_exe_file(struct mm_struct
*mm
, unsigned int fd
)
1796 struct dentry
*dentry
;
1803 dentry
= exe
.file
->f_path
.dentry
;
1806 * Because the original mm->exe_file points to executable file, make
1807 * sure that this one is executable as well, to avoid breaking an
1811 if (!S_ISREG(dentry
->d_inode
->i_mode
) ||
1812 exe
.file
->f_path
.mnt
->mnt_flags
& MNT_NOEXEC
)
1815 err
= inode_permission(dentry
->d_inode
, MAY_EXEC
);
1819 down_write(&mm
->mmap_sem
);
1822 * Forbid mm->exe_file change if old file still mapped.
1826 struct vm_area_struct
*vma
;
1828 for (vma
= mm
->mmap
; vma
; vma
= vma
->vm_next
)
1830 path_equal(&vma
->vm_file
->f_path
,
1831 &mm
->exe_file
->f_path
))
1836 * The symlink can be changed only once, just to disallow arbitrary
1837 * transitions malicious software might bring in. This means one
1838 * could make a snapshot over all processes running and monitor
1839 * /proc/pid/exe changes to notice unusual activity if needed.
1842 if (test_and_set_bit(MMF_EXE_FILE_CHANGED
, &mm
->flags
))
1846 set_mm_exe_file(mm
, exe
.file
); /* this grabs a reference to exe.file */
1848 up_write(&mm
->mmap_sem
);
1855 static int prctl_set_mm(int opt
, unsigned long addr
,
1856 unsigned long arg4
, unsigned long arg5
)
1858 unsigned long rlim
= rlimit(RLIMIT_DATA
);
1859 struct mm_struct
*mm
= current
->mm
;
1860 struct vm_area_struct
*vma
;
1863 if (arg5
|| (arg4
&& opt
!= PR_SET_MM_AUXV
))
1866 if (!capable(CAP_SYS_RESOURCE
))
1869 if (opt
== PR_SET_MM_EXE_FILE
)
1870 return prctl_set_mm_exe_file(mm
, (unsigned int)addr
);
1872 if (addr
>= TASK_SIZE
|| addr
< mmap_min_addr
)
1877 down_read(&mm
->mmap_sem
);
1878 vma
= find_vma(mm
, addr
);
1881 case PR_SET_MM_START_CODE
:
1882 mm
->start_code
= addr
;
1884 case PR_SET_MM_END_CODE
:
1885 mm
->end_code
= addr
;
1887 case PR_SET_MM_START_DATA
:
1888 mm
->start_data
= addr
;
1890 case PR_SET_MM_END_DATA
:
1891 mm
->end_data
= addr
;
1894 case PR_SET_MM_START_BRK
:
1895 if (addr
<= mm
->end_data
)
1898 if (rlim
< RLIM_INFINITY
&&
1900 (mm
->end_data
- mm
->start_data
) > rlim
)
1903 mm
->start_brk
= addr
;
1907 if (addr
<= mm
->end_data
)
1910 if (rlim
< RLIM_INFINITY
&&
1911 (addr
- mm
->start_brk
) +
1912 (mm
->end_data
- mm
->start_data
) > rlim
)
1919 * If command line arguments and environment
1920 * are placed somewhere else on stack, we can
1921 * set them up here, ARG_START/END to setup
1922 * command line argumets and ENV_START/END
1925 case PR_SET_MM_START_STACK
:
1926 case PR_SET_MM_ARG_START
:
1927 case PR_SET_MM_ARG_END
:
1928 case PR_SET_MM_ENV_START
:
1929 case PR_SET_MM_ENV_END
:
1934 if (opt
== PR_SET_MM_START_STACK
)
1935 mm
->start_stack
= addr
;
1936 else if (opt
== PR_SET_MM_ARG_START
)
1937 mm
->arg_start
= addr
;
1938 else if (opt
== PR_SET_MM_ARG_END
)
1940 else if (opt
== PR_SET_MM_ENV_START
)
1941 mm
->env_start
= addr
;
1942 else if (opt
== PR_SET_MM_ENV_END
)
1947 * This doesn't move auxiliary vector itself
1948 * since it's pinned to mm_struct, but allow
1949 * to fill vector with new values. It's up
1950 * to a caller to provide sane values here
1951 * otherwise user space tools which use this
1952 * vector might be unhappy.
1954 case PR_SET_MM_AUXV
: {
1955 unsigned long user_auxv
[AT_VECTOR_SIZE
];
1957 if (arg4
> sizeof(user_auxv
))
1959 up_read(&mm
->mmap_sem
);
1961 if (copy_from_user(user_auxv
, (const void __user
*)addr
, arg4
))
1964 /* Make sure the last entry is always AT_NULL */
1965 user_auxv
[AT_VECTOR_SIZE
- 2] = 0;
1966 user_auxv
[AT_VECTOR_SIZE
- 1] = 0;
1968 BUILD_BUG_ON(sizeof(user_auxv
) != sizeof(mm
->saved_auxv
));
1971 memcpy(mm
->saved_auxv
, user_auxv
, arg4
);
1972 task_unlock(current
);
1982 up_read(&mm
->mmap_sem
);
1986 static int prctl_get_tid_address(struct task_struct
*me
, int __user
**tid_addr
)
1988 return put_user(me
->clear_child_tid
, tid_addr
);
1991 #else /* CONFIG_CHECKPOINT_RESTORE */
1992 static int prctl_set_mm(int opt
, unsigned long addr
,
1993 unsigned long arg4
, unsigned long arg5
)
1997 static int prctl_get_tid_address(struct task_struct
*me
, int __user
**tid_addr
)
2003 SYSCALL_DEFINE5(prctl
, int, option
, unsigned long, arg2
, unsigned long, arg3
,
2004 unsigned long, arg4
, unsigned long, arg5
)
2006 struct task_struct
*me
= current
;
2007 unsigned char comm
[sizeof(me
->comm
)];
2010 error
= security_task_prctl(option
, arg2
, arg3
, arg4
, arg5
);
2011 if (error
!= -ENOSYS
)
2016 case PR_SET_PDEATHSIG
:
2017 if (!valid_signal(arg2
)) {
2021 me
->pdeath_signal
= arg2
;
2023 case PR_GET_PDEATHSIG
:
2024 error
= put_user(me
->pdeath_signal
, (int __user
*)arg2
);
2026 case PR_GET_DUMPABLE
:
2027 error
= get_dumpable(me
->mm
);
2029 case PR_SET_DUMPABLE
:
2030 if (arg2
!= SUID_DUMP_DISABLE
&& arg2
!= SUID_DUMP_USER
) {
2034 set_dumpable(me
->mm
, arg2
);
2037 case PR_SET_UNALIGN
:
2038 error
= SET_UNALIGN_CTL(me
, arg2
);
2040 case PR_GET_UNALIGN
:
2041 error
= GET_UNALIGN_CTL(me
, arg2
);
2044 error
= SET_FPEMU_CTL(me
, arg2
);
2047 error
= GET_FPEMU_CTL(me
, arg2
);
2050 error
= SET_FPEXC_CTL(me
, arg2
);
2053 error
= GET_FPEXC_CTL(me
, arg2
);
2056 error
= PR_TIMING_STATISTICAL
;
2059 if (arg2
!= PR_TIMING_STATISTICAL
)
2063 comm
[sizeof(me
->comm
) - 1] = 0;
2064 if (strncpy_from_user(comm
, (char __user
*)arg2
,
2065 sizeof(me
->comm
) - 1) < 0)
2067 set_task_comm(me
, comm
);
2068 proc_comm_connector(me
);
2071 get_task_comm(comm
, me
);
2072 if (copy_to_user((char __user
*)arg2
, comm
, sizeof(comm
)))
2076 error
= GET_ENDIAN(me
, arg2
);
2079 error
= SET_ENDIAN(me
, arg2
);
2081 case PR_GET_SECCOMP
:
2082 error
= prctl_get_seccomp();
2084 case PR_SET_SECCOMP
:
2085 error
= prctl_set_seccomp(arg2
, (char __user
*)arg3
);
2088 error
= GET_TSC_CTL(arg2
);
2091 error
= SET_TSC_CTL(arg2
);
2093 case PR_TASK_PERF_EVENTS_DISABLE
:
2094 error
= perf_event_task_disable();
2096 case PR_TASK_PERF_EVENTS_ENABLE
:
2097 error
= perf_event_task_enable();
2099 case PR_GET_TIMERSLACK
:
2100 error
= current
->timer_slack_ns
;
2102 case PR_SET_TIMERSLACK
:
2104 current
->timer_slack_ns
=
2105 current
->default_timer_slack_ns
;
2107 current
->timer_slack_ns
= arg2
;
2113 case PR_MCE_KILL_CLEAR
:
2116 current
->flags
&= ~PF_MCE_PROCESS
;
2118 case PR_MCE_KILL_SET
:
2119 current
->flags
|= PF_MCE_PROCESS
;
2120 if (arg3
== PR_MCE_KILL_EARLY
)
2121 current
->flags
|= PF_MCE_EARLY
;
2122 else if (arg3
== PR_MCE_KILL_LATE
)
2123 current
->flags
&= ~PF_MCE_EARLY
;
2124 else if (arg3
== PR_MCE_KILL_DEFAULT
)
2126 ~(PF_MCE_EARLY
|PF_MCE_PROCESS
);
2134 case PR_MCE_KILL_GET
:
2135 if (arg2
| arg3
| arg4
| arg5
)
2137 if (current
->flags
& PF_MCE_PROCESS
)
2138 error
= (current
->flags
& PF_MCE_EARLY
) ?
2139 PR_MCE_KILL_EARLY
: PR_MCE_KILL_LATE
;
2141 error
= PR_MCE_KILL_DEFAULT
;
2144 error
= prctl_set_mm(arg2
, arg3
, arg4
, arg5
);
2146 case PR_GET_TID_ADDRESS
:
2147 error
= prctl_get_tid_address(me
, (int __user
**)arg2
);
2149 case PR_SET_CHILD_SUBREAPER
:
2150 me
->signal
->is_child_subreaper
= !!arg2
;
2152 case PR_GET_CHILD_SUBREAPER
:
2153 error
= put_user(me
->signal
->is_child_subreaper
,
2154 (int __user
*)arg2
);
2156 case PR_SET_NO_NEW_PRIVS
:
2157 if (arg2
!= 1 || arg3
|| arg4
|| arg5
)
2160 current
->no_new_privs
= 1;
2162 case PR_GET_NO_NEW_PRIVS
:
2163 if (arg2
|| arg3
|| arg4
|| arg5
)
2165 return current
->no_new_privs
? 1 : 0;
2173 SYSCALL_DEFINE3(getcpu
, unsigned __user
*, cpup
, unsigned __user
*, nodep
,
2174 struct getcpu_cache __user
*, unused
)
2177 int cpu
= raw_smp_processor_id();
2179 err
|= put_user(cpu
, cpup
);
2181 err
|= put_user(cpu_to_node(cpu
), nodep
);
2182 return err
? -EFAULT
: 0;
2185 char poweroff_cmd
[POWEROFF_CMD_PATH_LEN
] = "/sbin/poweroff";
2187 static void argv_cleanup(struct subprocess_info
*info
)
2189 argv_free(info
->argv
);
2192 static int __orderly_poweroff(void)
2196 static char *envp
[] = {
2198 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
2203 argv
= argv_split(GFP_ATOMIC
, poweroff_cmd
, &argc
);
2205 printk(KERN_WARNING
"%s failed to allocate memory for \"%s\"\n",
2206 __func__
, poweroff_cmd
);
2210 ret
= call_usermodehelper_fns(argv
[0], argv
, envp
, UMH_WAIT_EXEC
,
2211 NULL
, argv_cleanup
, NULL
);
2219 * orderly_poweroff - Trigger an orderly system poweroff
2220 * @force: force poweroff if command execution fails
2222 * This may be called from any context to trigger a system shutdown.
2223 * If the orderly shutdown fails, it will force an immediate shutdown.
2225 int orderly_poweroff(bool force
)
2227 int ret
= __orderly_poweroff();
2230 printk(KERN_WARNING
"Failed to start orderly shutdown: "
2231 "forcing the issue\n");
2234 * I guess this should try to kick off some daemon to sync and
2235 * poweroff asap. Or not even bother syncing if we're doing an
2236 * emergency shutdown?
2244 EXPORT_SYMBOL_GPL(orderly_poweroff
);