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/sched.h>
53 #include <linux/rcupdate.h>
54 #include <linux/uidgid.h>
55 #include <linux/cred.h>
57 #include <linux/kmsg_dump.h>
58 /* Move somewhere else to avoid recompiling? */
59 #include <generated/utsrelease.h>
61 #include <asm/uaccess.h>
63 #include <asm/unistd.h>
65 #ifndef SET_UNALIGN_CTL
66 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
68 #ifndef GET_UNALIGN_CTL
69 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
72 # define SET_FPEMU_CTL(a,b) (-EINVAL)
75 # define GET_FPEMU_CTL(a,b) (-EINVAL)
78 # define SET_FPEXC_CTL(a,b) (-EINVAL)
81 # define GET_FPEXC_CTL(a,b) (-EINVAL)
84 # define GET_ENDIAN(a,b) (-EINVAL)
87 # define SET_ENDIAN(a,b) (-EINVAL)
90 # define GET_TSC_CTL(a) (-EINVAL)
93 # define SET_TSC_CTL(a) (-EINVAL)
97 * this is where the system-wide overflow UID and GID are defined, for
98 * architectures that now have 32-bit UID/GID but didn't in the past
101 int overflowuid
= DEFAULT_OVERFLOWUID
;
102 int overflowgid
= DEFAULT_OVERFLOWGID
;
104 EXPORT_SYMBOL(overflowuid
);
105 EXPORT_SYMBOL(overflowgid
);
108 * the same as above, but for filesystems which can only store a 16-bit
109 * UID and GID. as such, this is needed on all architectures
112 int fs_overflowuid
= DEFAULT_FS_OVERFLOWUID
;
113 int fs_overflowgid
= DEFAULT_FS_OVERFLOWUID
;
115 EXPORT_SYMBOL(fs_overflowuid
);
116 EXPORT_SYMBOL(fs_overflowgid
);
119 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
124 EXPORT_SYMBOL(cad_pid
);
127 * If set, this is used for preparing the system to power off.
130 void (*pm_power_off_prepare
)(void);
133 * Returns true if current's euid is same as p's uid or euid,
134 * or has CAP_SYS_NICE to p's user_ns.
136 * Called with rcu_read_lock, creds are safe
138 static bool set_one_prio_perm(struct task_struct
*p
)
140 const struct cred
*cred
= current_cred(), *pcred
= __task_cred(p
);
142 if (uid_eq(pcred
->uid
, cred
->euid
) ||
143 uid_eq(pcred
->euid
, cred
->euid
))
145 if (ns_capable(pcred
->user_ns
, CAP_SYS_NICE
))
151 * set the priority of a task
152 * - the caller must hold the RCU read lock
154 static int set_one_prio(struct task_struct
*p
, int niceval
, int error
)
158 if (!set_one_prio_perm(p
)) {
162 if (niceval
< task_nice(p
) && !can_nice(p
, niceval
)) {
166 no_nice
= security_task_setnice(p
, niceval
);
173 set_user_nice(p
, niceval
);
178 SYSCALL_DEFINE3(setpriority
, int, which
, int, who
, int, niceval
)
180 struct task_struct
*g
, *p
;
181 struct user_struct
*user
;
182 const struct cred
*cred
= current_cred();
187 if (which
> PRIO_USER
|| which
< PRIO_PROCESS
)
190 /* normalize: avoid signed division (rounding problems) */
198 read_lock(&tasklist_lock
);
202 p
= find_task_by_vpid(who
);
206 error
= set_one_prio(p
, niceval
, error
);
210 pgrp
= find_vpid(who
);
212 pgrp
= task_pgrp(current
);
213 do_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
) {
214 error
= set_one_prio(p
, niceval
, error
);
215 } while_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
);
218 uid
= make_kuid(cred
->user_ns
, who
);
222 else if (!uid_eq(uid
, cred
->uid
) &&
223 !(user
= find_user(uid
)))
224 goto out_unlock
; /* No processes for this user */
226 do_each_thread(g
, p
) {
227 if (uid_eq(task_uid(p
), uid
))
228 error
= set_one_prio(p
, niceval
, error
);
229 } while_each_thread(g
, p
);
230 if (!uid_eq(uid
, cred
->uid
))
231 free_uid(user
); /* For find_user() */
235 read_unlock(&tasklist_lock
);
242 * Ugh. To avoid negative return values, "getpriority()" will
243 * not return the normal nice-value, but a negated value that
244 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
245 * to stay compatible.
247 SYSCALL_DEFINE2(getpriority
, int, which
, int, who
)
249 struct task_struct
*g
, *p
;
250 struct user_struct
*user
;
251 const struct cred
*cred
= current_cred();
252 long niceval
, retval
= -ESRCH
;
256 if (which
> PRIO_USER
|| which
< PRIO_PROCESS
)
260 read_lock(&tasklist_lock
);
264 p
= find_task_by_vpid(who
);
268 niceval
= 20 - task_nice(p
);
269 if (niceval
> retval
)
275 pgrp
= find_vpid(who
);
277 pgrp
= task_pgrp(current
);
278 do_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
) {
279 niceval
= 20 - task_nice(p
);
280 if (niceval
> retval
)
282 } while_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
);
285 uid
= make_kuid(cred
->user_ns
, who
);
289 else if (!uid_eq(uid
, cred
->uid
) &&
290 !(user
= find_user(uid
)))
291 goto out_unlock
; /* No processes for this user */
293 do_each_thread(g
, p
) {
294 if (uid_eq(task_uid(p
), uid
)) {
295 niceval
= 20 - task_nice(p
);
296 if (niceval
> retval
)
299 } while_each_thread(g
, p
);
300 if (!uid_eq(uid
, cred
->uid
))
301 free_uid(user
); /* for find_user() */
305 read_unlock(&tasklist_lock
);
312 * emergency_restart - reboot the system
314 * Without shutting down any hardware or taking any locks
315 * reboot the system. This is called when we know we are in
316 * trouble so this is our best effort to reboot. This is
317 * safe to call in interrupt context.
319 void emergency_restart(void)
321 kmsg_dump(KMSG_DUMP_EMERG
);
322 machine_emergency_restart();
324 EXPORT_SYMBOL_GPL(emergency_restart
);
326 void kernel_restart_prepare(char *cmd
)
328 blocking_notifier_call_chain(&reboot_notifier_list
, SYS_RESTART
, cmd
);
329 system_state
= SYSTEM_RESTART
;
330 usermodehelper_disable();
335 * register_reboot_notifier - Register function to be called at reboot time
336 * @nb: Info about notifier function to be called
338 * Registers a function with the list of functions
339 * to be called at reboot time.
341 * Currently always returns zero, as blocking_notifier_chain_register()
342 * always returns zero.
344 int register_reboot_notifier(struct notifier_block
*nb
)
346 return blocking_notifier_chain_register(&reboot_notifier_list
, nb
);
348 EXPORT_SYMBOL(register_reboot_notifier
);
351 * unregister_reboot_notifier - Unregister previously registered reboot notifier
352 * @nb: Hook to be unregistered
354 * Unregisters a previously registered reboot
357 * Returns zero on success, or %-ENOENT on failure.
359 int unregister_reboot_notifier(struct notifier_block
*nb
)
361 return blocking_notifier_chain_unregister(&reboot_notifier_list
, nb
);
363 EXPORT_SYMBOL(unregister_reboot_notifier
);
365 /* Add backwards compatibility for stable trees. */
366 #ifndef PF_NO_SETAFFINITY
367 #define PF_NO_SETAFFINITY PF_THREAD_BOUND
370 static void migrate_to_reboot_cpu(void)
372 /* The boot cpu is always logical cpu 0 */
375 cpu_hotplug_disable();
377 /* Make certain the cpu I'm about to reboot on is online */
378 if (!cpu_online(cpu
))
379 cpu
= cpumask_first(cpu_online_mask
);
381 /* Prevent races with other tasks migrating this task */
382 current
->flags
|= PF_NO_SETAFFINITY
;
384 /* Make certain I only run on the appropriate processor */
385 set_cpus_allowed_ptr(current
, cpumask_of(cpu
));
389 * kernel_restart - reboot the system
390 * @cmd: pointer to buffer containing command to execute for restart
393 * Shutdown everything and perform a clean reboot.
394 * This is not safe to call in interrupt context.
396 void kernel_restart(char *cmd
)
398 kernel_restart_prepare(cmd
);
399 migrate_to_reboot_cpu();
402 printk(KERN_EMERG
"Restarting system.\n");
404 printk(KERN_EMERG
"Restarting system with command '%s'.\n", cmd
);
405 kmsg_dump(KMSG_DUMP_RESTART
);
406 machine_restart(cmd
);
408 EXPORT_SYMBOL_GPL(kernel_restart
);
410 static void kernel_shutdown_prepare(enum system_states state
)
412 blocking_notifier_call_chain(&reboot_notifier_list
,
413 (state
== SYSTEM_HALT
)?SYS_HALT
:SYS_POWER_OFF
, NULL
);
414 system_state
= state
;
415 usermodehelper_disable();
419 * kernel_halt - halt the system
421 * Shutdown everything and perform a clean system halt.
423 void kernel_halt(void)
425 kernel_shutdown_prepare(SYSTEM_HALT
);
426 migrate_to_reboot_cpu();
428 printk(KERN_EMERG
"System halted.\n");
429 kmsg_dump(KMSG_DUMP_HALT
);
433 EXPORT_SYMBOL_GPL(kernel_halt
);
436 * kernel_power_off - power_off the system
438 * Shutdown everything and perform a clean system power_off.
440 void kernel_power_off(void)
442 kernel_shutdown_prepare(SYSTEM_POWER_OFF
);
443 if (pm_power_off_prepare
)
444 pm_power_off_prepare();
445 migrate_to_reboot_cpu();
447 printk(KERN_EMERG
"Power down.\n");
448 kmsg_dump(KMSG_DUMP_POWEROFF
);
451 EXPORT_SYMBOL_GPL(kernel_power_off
);
453 static DEFINE_MUTEX(reboot_mutex
);
456 * Reboot system call: for obvious reasons only root may call it,
457 * and even root needs to set up some magic numbers in the registers
458 * so that some mistake won't make this reboot the whole machine.
459 * You can also set the meaning of the ctrl-alt-del-key here.
461 * reboot doesn't sync: do that yourself before calling this.
463 SYSCALL_DEFINE4(reboot
, int, magic1
, int, magic2
, unsigned int, cmd
,
466 struct pid_namespace
*pid_ns
= task_active_pid_ns(current
);
470 /* We only trust the superuser with rebooting the system. */
471 if (!ns_capable(pid_ns
->user_ns
, CAP_SYS_BOOT
))
474 /* For safety, we require "magic" arguments. */
475 if (magic1
!= LINUX_REBOOT_MAGIC1
||
476 (magic2
!= LINUX_REBOOT_MAGIC2
&&
477 magic2
!= LINUX_REBOOT_MAGIC2A
&&
478 magic2
!= LINUX_REBOOT_MAGIC2B
&&
479 magic2
!= LINUX_REBOOT_MAGIC2C
))
483 * If pid namespaces are enabled and the current task is in a child
484 * pid_namespace, the command is handled by reboot_pid_ns() which will
487 ret
= reboot_pid_ns(pid_ns
, cmd
);
491 /* Instead of trying to make the power_off code look like
492 * halt when pm_power_off is not set do it the easy way.
494 if ((cmd
== LINUX_REBOOT_CMD_POWER_OFF
) && !pm_power_off
)
495 cmd
= LINUX_REBOOT_CMD_HALT
;
497 mutex_lock(&reboot_mutex
);
499 case LINUX_REBOOT_CMD_RESTART
:
500 kernel_restart(NULL
);
503 case LINUX_REBOOT_CMD_CAD_ON
:
507 case LINUX_REBOOT_CMD_CAD_OFF
:
511 case LINUX_REBOOT_CMD_HALT
:
514 panic("cannot halt");
516 case LINUX_REBOOT_CMD_POWER_OFF
:
521 case LINUX_REBOOT_CMD_RESTART2
:
522 if (strncpy_from_user(&buffer
[0], arg
, sizeof(buffer
) - 1) < 0) {
526 buffer
[sizeof(buffer
) - 1] = '\0';
528 kernel_restart(buffer
);
532 case LINUX_REBOOT_CMD_KEXEC
:
533 ret
= kernel_kexec();
537 #ifdef CONFIG_HIBERNATION
538 case LINUX_REBOOT_CMD_SW_SUSPEND
:
547 mutex_unlock(&reboot_mutex
);
551 static void deferred_cad(struct work_struct
*dummy
)
553 kernel_restart(NULL
);
557 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
558 * As it's called within an interrupt, it may NOT sync: the only choice
559 * is whether to reboot at once, or just ignore the ctrl-alt-del.
561 void ctrl_alt_del(void)
563 static DECLARE_WORK(cad_work
, deferred_cad
);
566 schedule_work(&cad_work
);
568 kill_cad_pid(SIGINT
, 1);
572 * Unprivileged users may change the real gid to the effective gid
573 * or vice versa. (BSD-style)
575 * If you set the real gid at all, or set the effective gid to a value not
576 * equal to the real gid, then the saved gid is set to the new effective gid.
578 * This makes it possible for a setgid program to completely drop its
579 * privileges, which is often a useful assertion to make when you are doing
580 * a security audit over a program.
582 * The general idea is that a program which uses just setregid() will be
583 * 100% compatible with BSD. A program which uses just setgid() will be
584 * 100% compatible with POSIX with saved IDs.
586 * SMP: There are not races, the GIDs are checked only by filesystem
587 * operations (as far as semantic preservation is concerned).
589 SYSCALL_DEFINE2(setregid
, gid_t
, rgid
, gid_t
, egid
)
591 struct user_namespace
*ns
= current_user_ns();
592 const struct cred
*old
;
597 krgid
= make_kgid(ns
, rgid
);
598 kegid
= make_kgid(ns
, egid
);
600 if ((rgid
!= (gid_t
) -1) && !gid_valid(krgid
))
602 if ((egid
!= (gid_t
) -1) && !gid_valid(kegid
))
605 new = prepare_creds();
608 old
= current_cred();
611 if (rgid
!= (gid_t
) -1) {
612 if (gid_eq(old
->gid
, krgid
) ||
613 gid_eq(old
->egid
, krgid
) ||
614 nsown_capable(CAP_SETGID
))
619 if (egid
!= (gid_t
) -1) {
620 if (gid_eq(old
->gid
, kegid
) ||
621 gid_eq(old
->egid
, kegid
) ||
622 gid_eq(old
->sgid
, kegid
) ||
623 nsown_capable(CAP_SETGID
))
629 if (rgid
!= (gid_t
) -1 ||
630 (egid
!= (gid_t
) -1 && !gid_eq(kegid
, old
->gid
)))
631 new->sgid
= new->egid
;
632 new->fsgid
= new->egid
;
634 return commit_creds(new);
642 * setgid() is implemented like SysV w/ SAVED_IDS
644 * SMP: Same implicit races as above.
646 SYSCALL_DEFINE1(setgid
, gid_t
, gid
)
648 struct user_namespace
*ns
= current_user_ns();
649 const struct cred
*old
;
654 kgid
= make_kgid(ns
, gid
);
655 if (!gid_valid(kgid
))
658 new = prepare_creds();
661 old
= current_cred();
664 if (nsown_capable(CAP_SETGID
))
665 new->gid
= new->egid
= new->sgid
= new->fsgid
= kgid
;
666 else if (gid_eq(kgid
, old
->gid
) || gid_eq(kgid
, old
->sgid
))
667 new->egid
= new->fsgid
= kgid
;
671 return commit_creds(new);
679 * change the user struct in a credentials set to match the new UID
681 static int set_user(struct cred
*new)
683 struct user_struct
*new_user
;
685 new_user
= alloc_uid(new->uid
);
690 * We don't fail in case of NPROC limit excess here because too many
691 * poorly written programs don't check set*uid() return code, assuming
692 * it never fails if called by root. We may still enforce NPROC limit
693 * for programs doing set*uid()+execve() by harmlessly deferring the
694 * failure to the execve() stage.
696 if (atomic_read(&new_user
->processes
) >= rlimit(RLIMIT_NPROC
) &&
697 new_user
!= INIT_USER
)
698 current
->flags
|= PF_NPROC_EXCEEDED
;
700 current
->flags
&= ~PF_NPROC_EXCEEDED
;
703 new->user
= new_user
;
708 * Unprivileged users may change the real uid to the effective uid
709 * or vice versa. (BSD-style)
711 * If you set the real uid at all, or set the effective uid to a value not
712 * equal to the real uid, then the saved uid is set to the new effective uid.
714 * This makes it possible for a setuid program to completely drop its
715 * privileges, which is often a useful assertion to make when you are doing
716 * a security audit over a program.
718 * The general idea is that a program which uses just setreuid() will be
719 * 100% compatible with BSD. A program which uses just setuid() will be
720 * 100% compatible with POSIX with saved IDs.
722 SYSCALL_DEFINE2(setreuid
, uid_t
, ruid
, uid_t
, euid
)
724 struct user_namespace
*ns
= current_user_ns();
725 const struct cred
*old
;
730 kruid
= make_kuid(ns
, ruid
);
731 keuid
= make_kuid(ns
, euid
);
733 if ((ruid
!= (uid_t
) -1) && !uid_valid(kruid
))
735 if ((euid
!= (uid_t
) -1) && !uid_valid(keuid
))
738 new = prepare_creds();
741 old
= current_cred();
744 if (ruid
!= (uid_t
) -1) {
746 if (!uid_eq(old
->uid
, kruid
) &&
747 !uid_eq(old
->euid
, kruid
) &&
748 !nsown_capable(CAP_SETUID
))
752 if (euid
!= (uid_t
) -1) {
754 if (!uid_eq(old
->uid
, keuid
) &&
755 !uid_eq(old
->euid
, keuid
) &&
756 !uid_eq(old
->suid
, keuid
) &&
757 !nsown_capable(CAP_SETUID
))
761 if (!uid_eq(new->uid
, old
->uid
)) {
762 retval
= set_user(new);
766 if (ruid
!= (uid_t
) -1 ||
767 (euid
!= (uid_t
) -1 && !uid_eq(keuid
, old
->uid
)))
768 new->suid
= new->euid
;
769 new->fsuid
= new->euid
;
771 retval
= security_task_fix_setuid(new, old
, LSM_SETID_RE
);
775 return commit_creds(new);
783 * setuid() is implemented like SysV with SAVED_IDS
785 * Note that SAVED_ID's is deficient in that a setuid root program
786 * like sendmail, for example, cannot set its uid to be a normal
787 * user and then switch back, because if you're root, setuid() sets
788 * the saved uid too. If you don't like this, blame the bright people
789 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
790 * will allow a root program to temporarily drop privileges and be able to
791 * regain them by swapping the real and effective uid.
793 SYSCALL_DEFINE1(setuid
, uid_t
, uid
)
795 struct user_namespace
*ns
= current_user_ns();
796 const struct cred
*old
;
801 kuid
= make_kuid(ns
, uid
);
802 if (!uid_valid(kuid
))
805 new = prepare_creds();
808 old
= current_cred();
811 if (nsown_capable(CAP_SETUID
)) {
812 new->suid
= new->uid
= kuid
;
813 if (!uid_eq(kuid
, old
->uid
)) {
814 retval
= set_user(new);
818 } else if (!uid_eq(kuid
, old
->uid
) && !uid_eq(kuid
, new->suid
)) {
822 new->fsuid
= new->euid
= kuid
;
824 retval
= security_task_fix_setuid(new, old
, LSM_SETID_ID
);
828 return commit_creds(new);
837 * This function implements a generic ability to update ruid, euid,
838 * and suid. This allows you to implement the 4.4 compatible seteuid().
840 SYSCALL_DEFINE3(setresuid
, uid_t
, ruid
, uid_t
, euid
, uid_t
, suid
)
842 struct user_namespace
*ns
= current_user_ns();
843 const struct cred
*old
;
846 kuid_t kruid
, keuid
, ksuid
;
848 kruid
= make_kuid(ns
, ruid
);
849 keuid
= make_kuid(ns
, euid
);
850 ksuid
= make_kuid(ns
, suid
);
852 if ((ruid
!= (uid_t
) -1) && !uid_valid(kruid
))
855 if ((euid
!= (uid_t
) -1) && !uid_valid(keuid
))
858 if ((suid
!= (uid_t
) -1) && !uid_valid(ksuid
))
861 new = prepare_creds();
865 old
= current_cred();
868 if (!nsown_capable(CAP_SETUID
)) {
869 if (ruid
!= (uid_t
) -1 && !uid_eq(kruid
, old
->uid
) &&
870 !uid_eq(kruid
, old
->euid
) && !uid_eq(kruid
, old
->suid
))
872 if (euid
!= (uid_t
) -1 && !uid_eq(keuid
, old
->uid
) &&
873 !uid_eq(keuid
, old
->euid
) && !uid_eq(keuid
, old
->suid
))
875 if (suid
!= (uid_t
) -1 && !uid_eq(ksuid
, old
->uid
) &&
876 !uid_eq(ksuid
, old
->euid
) && !uid_eq(ksuid
, old
->suid
))
880 if (ruid
!= (uid_t
) -1) {
882 if (!uid_eq(kruid
, old
->uid
)) {
883 retval
= set_user(new);
888 if (euid
!= (uid_t
) -1)
890 if (suid
!= (uid_t
) -1)
892 new->fsuid
= new->euid
;
894 retval
= security_task_fix_setuid(new, old
, LSM_SETID_RES
);
898 return commit_creds(new);
905 SYSCALL_DEFINE3(getresuid
, uid_t __user
*, ruidp
, uid_t __user
*, euidp
, uid_t __user
*, suidp
)
907 const struct cred
*cred
= current_cred();
909 uid_t ruid
, euid
, suid
;
911 ruid
= from_kuid_munged(cred
->user_ns
, cred
->uid
);
912 euid
= from_kuid_munged(cred
->user_ns
, cred
->euid
);
913 suid
= from_kuid_munged(cred
->user_ns
, cred
->suid
);
915 if (!(retval
= put_user(ruid
, ruidp
)) &&
916 !(retval
= put_user(euid
, euidp
)))
917 retval
= put_user(suid
, suidp
);
923 * Same as above, but for rgid, egid, sgid.
925 SYSCALL_DEFINE3(setresgid
, gid_t
, rgid
, gid_t
, egid
, gid_t
, sgid
)
927 struct user_namespace
*ns
= current_user_ns();
928 const struct cred
*old
;
931 kgid_t krgid
, kegid
, ksgid
;
933 krgid
= make_kgid(ns
, rgid
);
934 kegid
= make_kgid(ns
, egid
);
935 ksgid
= make_kgid(ns
, sgid
);
937 if ((rgid
!= (gid_t
) -1) && !gid_valid(krgid
))
939 if ((egid
!= (gid_t
) -1) && !gid_valid(kegid
))
941 if ((sgid
!= (gid_t
) -1) && !gid_valid(ksgid
))
944 new = prepare_creds();
947 old
= current_cred();
950 if (!nsown_capable(CAP_SETGID
)) {
951 if (rgid
!= (gid_t
) -1 && !gid_eq(krgid
, old
->gid
) &&
952 !gid_eq(krgid
, old
->egid
) && !gid_eq(krgid
, old
->sgid
))
954 if (egid
!= (gid_t
) -1 && !gid_eq(kegid
, old
->gid
) &&
955 !gid_eq(kegid
, old
->egid
) && !gid_eq(kegid
, old
->sgid
))
957 if (sgid
!= (gid_t
) -1 && !gid_eq(ksgid
, old
->gid
) &&
958 !gid_eq(ksgid
, old
->egid
) && !gid_eq(ksgid
, old
->sgid
))
962 if (rgid
!= (gid_t
) -1)
964 if (egid
!= (gid_t
) -1)
966 if (sgid
!= (gid_t
) -1)
968 new->fsgid
= new->egid
;
970 return commit_creds(new);
977 SYSCALL_DEFINE3(getresgid
, gid_t __user
*, rgidp
, gid_t __user
*, egidp
, gid_t __user
*, sgidp
)
979 const struct cred
*cred
= current_cred();
981 gid_t rgid
, egid
, sgid
;
983 rgid
= from_kgid_munged(cred
->user_ns
, cred
->gid
);
984 egid
= from_kgid_munged(cred
->user_ns
, cred
->egid
);
985 sgid
= from_kgid_munged(cred
->user_ns
, cred
->sgid
);
987 if (!(retval
= put_user(rgid
, rgidp
)) &&
988 !(retval
= put_user(egid
, egidp
)))
989 retval
= put_user(sgid
, sgidp
);
996 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
997 * is used for "access()" and for the NFS daemon (letting nfsd stay at
998 * whatever uid it wants to). It normally shadows "euid", except when
999 * explicitly set by setfsuid() or for access..
1001 SYSCALL_DEFINE1(setfsuid
, uid_t
, uid
)
1003 const struct cred
*old
;
1008 old
= current_cred();
1009 old_fsuid
= from_kuid_munged(old
->user_ns
, old
->fsuid
);
1011 kuid
= make_kuid(old
->user_ns
, uid
);
1012 if (!uid_valid(kuid
))
1015 new = prepare_creds();
1019 if (uid_eq(kuid
, old
->uid
) || uid_eq(kuid
, old
->euid
) ||
1020 uid_eq(kuid
, old
->suid
) || uid_eq(kuid
, old
->fsuid
) ||
1021 nsown_capable(CAP_SETUID
)) {
1022 if (!uid_eq(kuid
, old
->fsuid
)) {
1024 if (security_task_fix_setuid(new, old
, LSM_SETID_FS
) == 0)
1038 * Samma på svenska..
1040 SYSCALL_DEFINE1(setfsgid
, gid_t
, gid
)
1042 const struct cred
*old
;
1047 old
= current_cred();
1048 old_fsgid
= from_kgid_munged(old
->user_ns
, old
->fsgid
);
1050 kgid
= make_kgid(old
->user_ns
, gid
);
1051 if (!gid_valid(kgid
))
1054 new = prepare_creds();
1058 if (gid_eq(kgid
, old
->gid
) || gid_eq(kgid
, old
->egid
) ||
1059 gid_eq(kgid
, old
->sgid
) || gid_eq(kgid
, old
->fsgid
) ||
1060 nsown_capable(CAP_SETGID
)) {
1061 if (!gid_eq(kgid
, old
->fsgid
)) {
1076 * sys_getpid - return the thread group id of the current process
1078 * Note, despite the name, this returns the tgid not the pid. The tgid and
1079 * the pid are identical unless CLONE_THREAD was specified on clone() in
1080 * which case the tgid is the same in all threads of the same group.
1082 * This is SMP safe as current->tgid does not change.
1084 SYSCALL_DEFINE0(getpid
)
1086 return task_tgid_vnr(current
);
1089 /* Thread ID - the internal kernel "pid" */
1090 SYSCALL_DEFINE0(gettid
)
1092 return task_pid_vnr(current
);
1096 * Accessing ->real_parent is not SMP-safe, it could
1097 * change from under us. However, we can use a stale
1098 * value of ->real_parent under rcu_read_lock(), see
1099 * release_task()->call_rcu(delayed_put_task_struct).
1101 SYSCALL_DEFINE0(getppid
)
1106 pid
= task_tgid_vnr(rcu_dereference(current
->real_parent
));
1112 SYSCALL_DEFINE0(getuid
)
1114 /* Only we change this so SMP safe */
1115 return from_kuid_munged(current_user_ns(), current_uid());
1118 SYSCALL_DEFINE0(geteuid
)
1120 /* Only we change this so SMP safe */
1121 return from_kuid_munged(current_user_ns(), current_euid());
1124 SYSCALL_DEFINE0(getgid
)
1126 /* Only we change this so SMP safe */
1127 return from_kgid_munged(current_user_ns(), current_gid());
1130 SYSCALL_DEFINE0(getegid
)
1132 /* Only we change this so SMP safe */
1133 return from_kgid_munged(current_user_ns(), current_egid());
1136 void do_sys_times(struct tms
*tms
)
1138 cputime_t tgutime
, tgstime
, cutime
, cstime
;
1140 spin_lock_irq(¤t
->sighand
->siglock
);
1141 thread_group_cputime_adjusted(current
, &tgutime
, &tgstime
);
1142 cutime
= current
->signal
->cutime
;
1143 cstime
= current
->signal
->cstime
;
1144 spin_unlock_irq(¤t
->sighand
->siglock
);
1145 tms
->tms_utime
= cputime_to_clock_t(tgutime
);
1146 tms
->tms_stime
= cputime_to_clock_t(tgstime
);
1147 tms
->tms_cutime
= cputime_to_clock_t(cutime
);
1148 tms
->tms_cstime
= cputime_to_clock_t(cstime
);
1151 SYSCALL_DEFINE1(times
, struct tms __user
*, tbuf
)
1157 if (copy_to_user(tbuf
, &tmp
, sizeof(struct tms
)))
1160 force_successful_syscall_return();
1161 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1165 * This needs some heavy checking ...
1166 * I just haven't the stomach for it. I also don't fully
1167 * understand sessions/pgrp etc. Let somebody who does explain it.
1169 * OK, I think I have the protection semantics right.... this is really
1170 * only important on a multi-user system anyway, to make sure one user
1171 * can't send a signal to a process owned by another. -TYT, 12/12/91
1173 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1176 SYSCALL_DEFINE2(setpgid
, pid_t
, pid
, pid_t
, pgid
)
1178 struct task_struct
*p
;
1179 struct task_struct
*group_leader
= current
->group_leader
;
1184 pid
= task_pid_vnr(group_leader
);
1191 /* From this point forward we keep holding onto the tasklist lock
1192 * so that our parent does not change from under us. -DaveM
1194 write_lock_irq(&tasklist_lock
);
1197 p
= find_task_by_vpid(pid
);
1202 if (!thread_group_leader(p
))
1205 if (same_thread_group(p
->real_parent
, group_leader
)) {
1207 if (task_session(p
) != task_session(group_leader
))
1214 if (p
!= group_leader
)
1219 if (p
->signal
->leader
)
1224 struct task_struct
*g
;
1226 pgrp
= find_vpid(pgid
);
1227 g
= pid_task(pgrp
, PIDTYPE_PGID
);
1228 if (!g
|| task_session(g
) != task_session(group_leader
))
1232 err
= security_task_setpgid(p
, pgid
);
1236 if (task_pgrp(p
) != pgrp
)
1237 change_pid(p
, PIDTYPE_PGID
, pgrp
);
1241 /* All paths lead to here, thus we are safe. -DaveM */
1242 write_unlock_irq(&tasklist_lock
);
1247 SYSCALL_DEFINE1(getpgid
, pid_t
, pid
)
1249 struct task_struct
*p
;
1255 grp
= task_pgrp(current
);
1258 p
= find_task_by_vpid(pid
);
1265 retval
= security_task_getpgid(p
);
1269 retval
= pid_vnr(grp
);
1275 #ifdef __ARCH_WANT_SYS_GETPGRP
1277 SYSCALL_DEFINE0(getpgrp
)
1279 return sys_getpgid(0);
1284 SYSCALL_DEFINE1(getsid
, pid_t
, pid
)
1286 struct task_struct
*p
;
1292 sid
= task_session(current
);
1295 p
= find_task_by_vpid(pid
);
1298 sid
= task_session(p
);
1302 retval
= security_task_getsid(p
);
1306 retval
= pid_vnr(sid
);
1312 SYSCALL_DEFINE0(setsid
)
1314 struct task_struct
*group_leader
= current
->group_leader
;
1315 struct pid
*sid
= task_pid(group_leader
);
1316 pid_t session
= pid_vnr(sid
);
1319 write_lock_irq(&tasklist_lock
);
1320 /* Fail if I am already a session leader */
1321 if (group_leader
->signal
->leader
)
1324 /* Fail if a process group id already exists that equals the
1325 * proposed session id.
1327 if (pid_task(sid
, PIDTYPE_PGID
))
1330 group_leader
->signal
->leader
= 1;
1331 __set_special_pids(sid
);
1333 proc_clear_tty(group_leader
);
1337 write_unlock_irq(&tasklist_lock
);
1339 proc_sid_connector(group_leader
);
1340 sched_autogroup_create_attach(group_leader
);
1345 DECLARE_RWSEM(uts_sem
);
1347 #ifdef COMPAT_UTS_MACHINE
1348 #define override_architecture(name) \
1349 (personality(current->personality) == PER_LINUX32 && \
1350 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1351 sizeof(COMPAT_UTS_MACHINE)))
1353 #define override_architecture(name) 0
1357 * Work around broken programs that cannot handle "Linux 3.0".
1358 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1360 static int override_release(char __user
*release
, size_t len
)
1364 if (current
->personality
& UNAME26
) {
1365 const char *rest
= UTS_RELEASE
;
1366 char buf
[65] = { 0 };
1372 if (*rest
== '.' && ++ndots
>= 3)
1374 if (!isdigit(*rest
) && *rest
!= '.')
1378 v
= ((LINUX_VERSION_CODE
>> 8) & 0xff) + 40;
1379 copy
= clamp_t(size_t, len
, 1, sizeof(buf
));
1380 copy
= scnprintf(buf
, copy
, "2.6.%u%s", v
, rest
);
1381 ret
= copy_to_user(release
, buf
, copy
+ 1);
1386 SYSCALL_DEFINE1(newuname
, struct new_utsname __user
*, name
)
1390 down_read(&uts_sem
);
1391 if (copy_to_user(name
, utsname(), sizeof *name
))
1395 if (!errno
&& override_release(name
->release
, sizeof(name
->release
)))
1397 if (!errno
&& override_architecture(name
))
1402 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1406 SYSCALL_DEFINE1(uname
, struct old_utsname __user
*, name
)
1413 down_read(&uts_sem
);
1414 if (copy_to_user(name
, utsname(), sizeof(*name
)))
1418 if (!error
&& override_release(name
->release
, sizeof(name
->release
)))
1420 if (!error
&& override_architecture(name
))
1425 SYSCALL_DEFINE1(olduname
, struct oldold_utsname __user
*, name
)
1431 if (!access_ok(VERIFY_WRITE
, name
, sizeof(struct oldold_utsname
)))
1434 down_read(&uts_sem
);
1435 error
= __copy_to_user(&name
->sysname
, &utsname()->sysname
,
1437 error
|= __put_user(0, name
->sysname
+ __OLD_UTS_LEN
);
1438 error
|= __copy_to_user(&name
->nodename
, &utsname()->nodename
,
1440 error
|= __put_user(0, name
->nodename
+ __OLD_UTS_LEN
);
1441 error
|= __copy_to_user(&name
->release
, &utsname()->release
,
1443 error
|= __put_user(0, name
->release
+ __OLD_UTS_LEN
);
1444 error
|= __copy_to_user(&name
->version
, &utsname()->version
,
1446 error
|= __put_user(0, name
->version
+ __OLD_UTS_LEN
);
1447 error
|= __copy_to_user(&name
->machine
, &utsname()->machine
,
1449 error
|= __put_user(0, name
->machine
+ __OLD_UTS_LEN
);
1452 if (!error
&& override_architecture(name
))
1454 if (!error
&& override_release(name
->release
, sizeof(name
->release
)))
1456 return error
? -EFAULT
: 0;
1460 SYSCALL_DEFINE2(sethostname
, char __user
*, name
, int, len
)
1463 char tmp
[__NEW_UTS_LEN
];
1465 if (!ns_capable(current
->nsproxy
->uts_ns
->user_ns
, CAP_SYS_ADMIN
))
1468 if (len
< 0 || len
> __NEW_UTS_LEN
)
1470 down_write(&uts_sem
);
1472 if (!copy_from_user(tmp
, name
, len
)) {
1473 struct new_utsname
*u
= utsname();
1475 memcpy(u
->nodename
, tmp
, len
);
1476 memset(u
->nodename
+ len
, 0, sizeof(u
->nodename
) - len
);
1478 uts_proc_notify(UTS_PROC_HOSTNAME
);
1484 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1486 SYSCALL_DEFINE2(gethostname
, char __user
*, name
, int, len
)
1489 struct new_utsname
*u
;
1493 down_read(&uts_sem
);
1495 i
= 1 + strlen(u
->nodename
);
1499 if (copy_to_user(name
, u
->nodename
, i
))
1508 * Only setdomainname; getdomainname can be implemented by calling
1511 SYSCALL_DEFINE2(setdomainname
, char __user
*, name
, int, len
)
1514 char tmp
[__NEW_UTS_LEN
];
1516 if (!ns_capable(current
->nsproxy
->uts_ns
->user_ns
, CAP_SYS_ADMIN
))
1518 if (len
< 0 || len
> __NEW_UTS_LEN
)
1521 down_write(&uts_sem
);
1523 if (!copy_from_user(tmp
, name
, len
)) {
1524 struct new_utsname
*u
= utsname();
1526 memcpy(u
->domainname
, tmp
, len
);
1527 memset(u
->domainname
+ len
, 0, sizeof(u
->domainname
) - len
);
1529 uts_proc_notify(UTS_PROC_DOMAINNAME
);
1535 SYSCALL_DEFINE2(getrlimit
, unsigned int, resource
, struct rlimit __user
*, rlim
)
1537 struct rlimit value
;
1540 ret
= do_prlimit(current
, resource
, NULL
, &value
);
1542 ret
= copy_to_user(rlim
, &value
, sizeof(*rlim
)) ? -EFAULT
: 0;
1547 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1550 * Back compatibility for getrlimit. Needed for some apps.
1553 SYSCALL_DEFINE2(old_getrlimit
, unsigned int, resource
,
1554 struct rlimit __user
*, rlim
)
1557 if (resource
>= RLIM_NLIMITS
)
1560 task_lock(current
->group_leader
);
1561 x
= current
->signal
->rlim
[resource
];
1562 task_unlock(current
->group_leader
);
1563 if (x
.rlim_cur
> 0x7FFFFFFF)
1564 x
.rlim_cur
= 0x7FFFFFFF;
1565 if (x
.rlim_max
> 0x7FFFFFFF)
1566 x
.rlim_max
= 0x7FFFFFFF;
1567 return copy_to_user(rlim
, &x
, sizeof(x
))?-EFAULT
:0;
1572 static inline bool rlim64_is_infinity(__u64 rlim64
)
1574 #if BITS_PER_LONG < 64
1575 return rlim64
>= ULONG_MAX
;
1577 return rlim64
== RLIM64_INFINITY
;
1581 static void rlim_to_rlim64(const struct rlimit
*rlim
, struct rlimit64
*rlim64
)
1583 if (rlim
->rlim_cur
== RLIM_INFINITY
)
1584 rlim64
->rlim_cur
= RLIM64_INFINITY
;
1586 rlim64
->rlim_cur
= rlim
->rlim_cur
;
1587 if (rlim
->rlim_max
== RLIM_INFINITY
)
1588 rlim64
->rlim_max
= RLIM64_INFINITY
;
1590 rlim64
->rlim_max
= rlim
->rlim_max
;
1593 static void rlim64_to_rlim(const struct rlimit64
*rlim64
, struct rlimit
*rlim
)
1595 if (rlim64_is_infinity(rlim64
->rlim_cur
))
1596 rlim
->rlim_cur
= RLIM_INFINITY
;
1598 rlim
->rlim_cur
= (unsigned long)rlim64
->rlim_cur
;
1599 if (rlim64_is_infinity(rlim64
->rlim_max
))
1600 rlim
->rlim_max
= RLIM_INFINITY
;
1602 rlim
->rlim_max
= (unsigned long)rlim64
->rlim_max
;
1605 /* make sure you are allowed to change @tsk limits before calling this */
1606 int do_prlimit(struct task_struct
*tsk
, unsigned int resource
,
1607 struct rlimit
*new_rlim
, struct rlimit
*old_rlim
)
1609 struct rlimit
*rlim
;
1612 if (resource
>= RLIM_NLIMITS
)
1615 if (new_rlim
->rlim_cur
> new_rlim
->rlim_max
)
1617 if (resource
== RLIMIT_NOFILE
&&
1618 new_rlim
->rlim_max
> sysctl_nr_open
)
1622 /* protect tsk->signal and tsk->sighand from disappearing */
1623 read_lock(&tasklist_lock
);
1624 if (!tsk
->sighand
) {
1629 rlim
= tsk
->signal
->rlim
+ resource
;
1630 task_lock(tsk
->group_leader
);
1632 /* Keep the capable check against init_user_ns until
1633 cgroups can contain all limits */
1634 if (new_rlim
->rlim_max
> rlim
->rlim_max
&&
1635 !capable(CAP_SYS_RESOURCE
))
1638 retval
= security_task_setrlimit(tsk
->group_leader
,
1639 resource
, new_rlim
);
1640 if (resource
== RLIMIT_CPU
&& new_rlim
->rlim_cur
== 0) {
1642 * The caller is asking for an immediate RLIMIT_CPU
1643 * expiry. But we use the zero value to mean "it was
1644 * never set". So let's cheat and make it one second
1647 new_rlim
->rlim_cur
= 1;
1656 task_unlock(tsk
->group_leader
);
1659 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1660 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1661 * very long-standing error, and fixing it now risks breakage of
1662 * applications, so we live with it
1664 if (!retval
&& new_rlim
&& resource
== RLIMIT_CPU
&&
1665 new_rlim
->rlim_cur
!= RLIM_INFINITY
)
1666 update_rlimit_cpu(tsk
, new_rlim
->rlim_cur
);
1668 read_unlock(&tasklist_lock
);
1672 /* rcu lock must be held */
1673 static int check_prlimit_permission(struct task_struct
*task
)
1675 const struct cred
*cred
= current_cred(), *tcred
;
1677 if (current
== task
)
1680 tcred
= __task_cred(task
);
1681 if (uid_eq(cred
->uid
, tcred
->euid
) &&
1682 uid_eq(cred
->uid
, tcred
->suid
) &&
1683 uid_eq(cred
->uid
, tcred
->uid
) &&
1684 gid_eq(cred
->gid
, tcred
->egid
) &&
1685 gid_eq(cred
->gid
, tcred
->sgid
) &&
1686 gid_eq(cred
->gid
, tcred
->gid
))
1688 if (ns_capable(tcred
->user_ns
, CAP_SYS_RESOURCE
))
1694 SYSCALL_DEFINE4(prlimit64
, pid_t
, pid
, unsigned int, resource
,
1695 const struct rlimit64 __user
*, new_rlim
,
1696 struct rlimit64 __user
*, old_rlim
)
1698 struct rlimit64 old64
, new64
;
1699 struct rlimit old
, new;
1700 struct task_struct
*tsk
;
1704 if (copy_from_user(&new64
, new_rlim
, sizeof(new64
)))
1706 rlim64_to_rlim(&new64
, &new);
1710 tsk
= pid
? find_task_by_vpid(pid
) : current
;
1715 ret
= check_prlimit_permission(tsk
);
1720 get_task_struct(tsk
);
1723 ret
= do_prlimit(tsk
, resource
, new_rlim
? &new : NULL
,
1724 old_rlim
? &old
: NULL
);
1726 if (!ret
&& old_rlim
) {
1727 rlim_to_rlim64(&old
, &old64
);
1728 if (copy_to_user(old_rlim
, &old64
, sizeof(old64
)))
1732 put_task_struct(tsk
);
1736 SYSCALL_DEFINE2(setrlimit
, unsigned int, resource
, struct rlimit __user
*, rlim
)
1738 struct rlimit new_rlim
;
1740 if (copy_from_user(&new_rlim
, rlim
, sizeof(*rlim
)))
1742 return do_prlimit(current
, resource
, &new_rlim
, NULL
);
1746 * It would make sense to put struct rusage in the task_struct,
1747 * except that would make the task_struct be *really big*. After
1748 * task_struct gets moved into malloc'ed memory, it would
1749 * make sense to do this. It will make moving the rest of the information
1750 * a lot simpler! (Which we're not doing right now because we're not
1751 * measuring them yet).
1753 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1754 * races with threads incrementing their own counters. But since word
1755 * reads are atomic, we either get new values or old values and we don't
1756 * care which for the sums. We always take the siglock to protect reading
1757 * the c* fields from p->signal from races with exit.c updating those
1758 * fields when reaping, so a sample either gets all the additions of a
1759 * given child after it's reaped, or none so this sample is before reaping.
1762 * We need to take the siglock for CHILDEREN, SELF and BOTH
1763 * for the cases current multithreaded, non-current single threaded
1764 * non-current multithreaded. Thread traversal is now safe with
1766 * Strictly speaking, we donot need to take the siglock if we are current and
1767 * single threaded, as no one else can take our signal_struct away, no one
1768 * else can reap the children to update signal->c* counters, and no one else
1769 * can race with the signal-> fields. If we do not take any lock, the
1770 * signal-> fields could be read out of order while another thread was just
1771 * exiting. So we should place a read memory barrier when we avoid the lock.
1772 * On the writer side, write memory barrier is implied in __exit_signal
1773 * as __exit_signal releases the siglock spinlock after updating the signal->
1774 * fields. But we don't do this yet to keep things simple.
1778 static void accumulate_thread_rusage(struct task_struct
*t
, struct rusage
*r
)
1780 r
->ru_nvcsw
+= t
->nvcsw
;
1781 r
->ru_nivcsw
+= t
->nivcsw
;
1782 r
->ru_minflt
+= t
->min_flt
;
1783 r
->ru_majflt
+= t
->maj_flt
;
1784 r
->ru_inblock
+= task_io_get_inblock(t
);
1785 r
->ru_oublock
+= task_io_get_oublock(t
);
1788 static void k_getrusage(struct task_struct
*p
, int who
, struct rusage
*r
)
1790 struct task_struct
*t
;
1791 unsigned long flags
;
1792 cputime_t tgutime
, tgstime
, utime
, stime
;
1793 unsigned long maxrss
= 0;
1795 memset((char *) r
, 0, sizeof *r
);
1798 if (who
== RUSAGE_THREAD
) {
1799 task_cputime_adjusted(current
, &utime
, &stime
);
1800 accumulate_thread_rusage(p
, r
);
1801 maxrss
= p
->signal
->maxrss
;
1805 if (!lock_task_sighand(p
, &flags
))
1810 case RUSAGE_CHILDREN
:
1811 utime
= p
->signal
->cutime
;
1812 stime
= p
->signal
->cstime
;
1813 r
->ru_nvcsw
= p
->signal
->cnvcsw
;
1814 r
->ru_nivcsw
= p
->signal
->cnivcsw
;
1815 r
->ru_minflt
= p
->signal
->cmin_flt
;
1816 r
->ru_majflt
= p
->signal
->cmaj_flt
;
1817 r
->ru_inblock
= p
->signal
->cinblock
;
1818 r
->ru_oublock
= p
->signal
->coublock
;
1819 maxrss
= p
->signal
->cmaxrss
;
1821 if (who
== RUSAGE_CHILDREN
)
1825 thread_group_cputime_adjusted(p
, &tgutime
, &tgstime
);
1828 r
->ru_nvcsw
+= p
->signal
->nvcsw
;
1829 r
->ru_nivcsw
+= p
->signal
->nivcsw
;
1830 r
->ru_minflt
+= p
->signal
->min_flt
;
1831 r
->ru_majflt
+= p
->signal
->maj_flt
;
1832 r
->ru_inblock
+= p
->signal
->inblock
;
1833 r
->ru_oublock
+= p
->signal
->oublock
;
1834 if (maxrss
< p
->signal
->maxrss
)
1835 maxrss
= p
->signal
->maxrss
;
1838 accumulate_thread_rusage(t
, r
);
1846 unlock_task_sighand(p
, &flags
);
1849 cputime_to_timeval(utime
, &r
->ru_utime
);
1850 cputime_to_timeval(stime
, &r
->ru_stime
);
1852 if (who
!= RUSAGE_CHILDREN
) {
1853 struct mm_struct
*mm
= get_task_mm(p
);
1855 setmax_mm_hiwater_rss(&maxrss
, mm
);
1859 r
->ru_maxrss
= maxrss
* (PAGE_SIZE
/ 1024); /* convert pages to KBs */
1862 int getrusage(struct task_struct
*p
, int who
, struct rusage __user
*ru
)
1865 k_getrusage(p
, who
, &r
);
1866 return copy_to_user(ru
, &r
, sizeof(r
)) ? -EFAULT
: 0;
1869 SYSCALL_DEFINE2(getrusage
, int, who
, struct rusage __user
*, ru
)
1871 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
&&
1872 who
!= RUSAGE_THREAD
)
1874 return getrusage(current
, who
, ru
);
1877 #ifdef CONFIG_COMPAT
1878 COMPAT_SYSCALL_DEFINE2(getrusage
, int, who
, struct compat_rusage __user
*, ru
)
1882 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
&&
1883 who
!= RUSAGE_THREAD
)
1886 k_getrusage(current
, who
, &r
);
1887 return put_compat_rusage(&r
, ru
);
1891 SYSCALL_DEFINE1(umask
, int, mask
)
1893 mask
= xchg(¤t
->fs
->umask
, mask
& S_IRWXUGO
);
1897 static int prctl_set_mm_exe_file(struct mm_struct
*mm
, unsigned int fd
)
1900 struct inode
*inode
;
1907 inode
= file_inode(exe
.file
);
1910 * Because the original mm->exe_file points to executable file, make
1911 * sure that this one is executable as well, to avoid breaking an
1915 if (!S_ISREG(inode
->i_mode
) ||
1916 exe
.file
->f_path
.mnt
->mnt_flags
& MNT_NOEXEC
)
1919 err
= inode_permission(inode
, MAY_EXEC
);
1923 down_write(&mm
->mmap_sem
);
1926 * Forbid mm->exe_file change if old file still mapped.
1930 struct vm_area_struct
*vma
;
1932 for (vma
= mm
->mmap
; vma
; vma
= vma
->vm_next
)
1934 path_equal(&vma
->vm_file
->f_path
,
1935 &mm
->exe_file
->f_path
))
1940 * The symlink can be changed only once, just to disallow arbitrary
1941 * transitions malicious software might bring in. This means one
1942 * could make a snapshot over all processes running and monitor
1943 * /proc/pid/exe changes to notice unusual activity if needed.
1946 if (test_and_set_bit(MMF_EXE_FILE_CHANGED
, &mm
->flags
))
1950 set_mm_exe_file(mm
, exe
.file
); /* this grabs a reference to exe.file */
1952 up_write(&mm
->mmap_sem
);
1959 static int prctl_set_mm(int opt
, unsigned long addr
,
1960 unsigned long arg4
, unsigned long arg5
)
1962 unsigned long rlim
= rlimit(RLIMIT_DATA
);
1963 struct mm_struct
*mm
= current
->mm
;
1964 struct vm_area_struct
*vma
;
1967 if (arg5
|| (arg4
&& opt
!= PR_SET_MM_AUXV
))
1970 if (!capable(CAP_SYS_RESOURCE
))
1973 if (opt
== PR_SET_MM_EXE_FILE
)
1974 return prctl_set_mm_exe_file(mm
, (unsigned int)addr
);
1976 if (addr
>= TASK_SIZE
|| addr
< mmap_min_addr
)
1981 down_read(&mm
->mmap_sem
);
1982 vma
= find_vma(mm
, addr
);
1985 case PR_SET_MM_START_CODE
:
1986 mm
->start_code
= addr
;
1988 case PR_SET_MM_END_CODE
:
1989 mm
->end_code
= addr
;
1991 case PR_SET_MM_START_DATA
:
1992 mm
->start_data
= addr
;
1994 case PR_SET_MM_END_DATA
:
1995 mm
->end_data
= addr
;
1998 case PR_SET_MM_START_BRK
:
1999 if (addr
<= mm
->end_data
)
2002 if (rlim
< RLIM_INFINITY
&&
2004 (mm
->end_data
- mm
->start_data
) > rlim
)
2007 mm
->start_brk
= addr
;
2011 if (addr
<= mm
->end_data
)
2014 if (rlim
< RLIM_INFINITY
&&
2015 (addr
- mm
->start_brk
) +
2016 (mm
->end_data
- mm
->start_data
) > rlim
)
2023 * If command line arguments and environment
2024 * are placed somewhere else on stack, we can
2025 * set them up here, ARG_START/END to setup
2026 * command line argumets and ENV_START/END
2029 case PR_SET_MM_START_STACK
:
2030 case PR_SET_MM_ARG_START
:
2031 case PR_SET_MM_ARG_END
:
2032 case PR_SET_MM_ENV_START
:
2033 case PR_SET_MM_ENV_END
:
2038 if (opt
== PR_SET_MM_START_STACK
)
2039 mm
->start_stack
= addr
;
2040 else if (opt
== PR_SET_MM_ARG_START
)
2041 mm
->arg_start
= addr
;
2042 else if (opt
== PR_SET_MM_ARG_END
)
2044 else if (opt
== PR_SET_MM_ENV_START
)
2045 mm
->env_start
= addr
;
2046 else if (opt
== PR_SET_MM_ENV_END
)
2051 * This doesn't move auxiliary vector itself
2052 * since it's pinned to mm_struct, but allow
2053 * to fill vector with new values. It's up
2054 * to a caller to provide sane values here
2055 * otherwise user space tools which use this
2056 * vector might be unhappy.
2058 case PR_SET_MM_AUXV
: {
2059 unsigned long user_auxv
[AT_VECTOR_SIZE
];
2061 if (arg4
> sizeof(user_auxv
))
2063 up_read(&mm
->mmap_sem
);
2065 if (copy_from_user(user_auxv
, (const void __user
*)addr
, arg4
))
2068 /* Make sure the last entry is always AT_NULL */
2069 user_auxv
[AT_VECTOR_SIZE
- 2] = 0;
2070 user_auxv
[AT_VECTOR_SIZE
- 1] = 0;
2072 BUILD_BUG_ON(sizeof(user_auxv
) != sizeof(mm
->saved_auxv
));
2075 memcpy(mm
->saved_auxv
, user_auxv
, arg4
);
2076 task_unlock(current
);
2086 up_read(&mm
->mmap_sem
);
2090 #ifdef CONFIG_CHECKPOINT_RESTORE
2091 static int prctl_get_tid_address(struct task_struct
*me
, int __user
**tid_addr
)
2093 return put_user(me
->clear_child_tid
, tid_addr
);
2096 static int prctl_get_tid_address(struct task_struct
*me
, int __user
**tid_addr
)
2102 SYSCALL_DEFINE5(prctl
, int, option
, unsigned long, arg2
, unsigned long, arg3
,
2103 unsigned long, arg4
, unsigned long, arg5
)
2105 struct task_struct
*me
= current
;
2106 unsigned char comm
[sizeof(me
->comm
)];
2109 error
= security_task_prctl(option
, arg2
, arg3
, arg4
, arg5
);
2110 if (error
!= -ENOSYS
)
2115 case PR_SET_PDEATHSIG
:
2116 if (!valid_signal(arg2
)) {
2120 me
->pdeath_signal
= arg2
;
2122 case PR_GET_PDEATHSIG
:
2123 error
= put_user(me
->pdeath_signal
, (int __user
*)arg2
);
2125 case PR_GET_DUMPABLE
:
2126 error
= get_dumpable(me
->mm
);
2128 case PR_SET_DUMPABLE
:
2129 if (arg2
!= SUID_DUMP_DISABLE
&& arg2
!= SUID_DUMP_USER
) {
2133 set_dumpable(me
->mm
, arg2
);
2136 case PR_SET_UNALIGN
:
2137 error
= SET_UNALIGN_CTL(me
, arg2
);
2139 case PR_GET_UNALIGN
:
2140 error
= GET_UNALIGN_CTL(me
, arg2
);
2143 error
= SET_FPEMU_CTL(me
, arg2
);
2146 error
= GET_FPEMU_CTL(me
, arg2
);
2149 error
= SET_FPEXC_CTL(me
, arg2
);
2152 error
= GET_FPEXC_CTL(me
, arg2
);
2155 error
= PR_TIMING_STATISTICAL
;
2158 if (arg2
!= PR_TIMING_STATISTICAL
)
2162 comm
[sizeof(me
->comm
) - 1] = 0;
2163 if (strncpy_from_user(comm
, (char __user
*)arg2
,
2164 sizeof(me
->comm
) - 1) < 0)
2166 set_task_comm(me
, comm
);
2167 proc_comm_connector(me
);
2170 get_task_comm(comm
, me
);
2171 if (copy_to_user((char __user
*)arg2
, comm
, sizeof(comm
)))
2175 error
= GET_ENDIAN(me
, arg2
);
2178 error
= SET_ENDIAN(me
, arg2
);
2180 case PR_GET_SECCOMP
:
2181 error
= prctl_get_seccomp();
2183 case PR_SET_SECCOMP
:
2184 error
= prctl_set_seccomp(arg2
, (char __user
*)arg3
);
2187 error
= GET_TSC_CTL(arg2
);
2190 error
= SET_TSC_CTL(arg2
);
2192 case PR_TASK_PERF_EVENTS_DISABLE
:
2193 error
= perf_event_task_disable();
2195 case PR_TASK_PERF_EVENTS_ENABLE
:
2196 error
= perf_event_task_enable();
2198 case PR_GET_TIMERSLACK
:
2199 error
= current
->timer_slack_ns
;
2201 case PR_SET_TIMERSLACK
:
2203 current
->timer_slack_ns
=
2204 current
->default_timer_slack_ns
;
2206 current
->timer_slack_ns
= arg2
;
2212 case PR_MCE_KILL_CLEAR
:
2215 current
->flags
&= ~PF_MCE_PROCESS
;
2217 case PR_MCE_KILL_SET
:
2218 current
->flags
|= PF_MCE_PROCESS
;
2219 if (arg3
== PR_MCE_KILL_EARLY
)
2220 current
->flags
|= PF_MCE_EARLY
;
2221 else if (arg3
== PR_MCE_KILL_LATE
)
2222 current
->flags
&= ~PF_MCE_EARLY
;
2223 else if (arg3
== PR_MCE_KILL_DEFAULT
)
2225 ~(PF_MCE_EARLY
|PF_MCE_PROCESS
);
2233 case PR_MCE_KILL_GET
:
2234 if (arg2
| arg3
| arg4
| arg5
)
2236 if (current
->flags
& PF_MCE_PROCESS
)
2237 error
= (current
->flags
& PF_MCE_EARLY
) ?
2238 PR_MCE_KILL_EARLY
: PR_MCE_KILL_LATE
;
2240 error
= PR_MCE_KILL_DEFAULT
;
2243 error
= prctl_set_mm(arg2
, arg3
, arg4
, arg5
);
2245 case PR_GET_TID_ADDRESS
:
2246 error
= prctl_get_tid_address(me
, (int __user
**)arg2
);
2248 case PR_SET_CHILD_SUBREAPER
:
2249 me
->signal
->is_child_subreaper
= !!arg2
;
2251 case PR_GET_CHILD_SUBREAPER
:
2252 error
= put_user(me
->signal
->is_child_subreaper
,
2253 (int __user
*)arg2
);
2255 case PR_SET_NO_NEW_PRIVS
:
2256 if (arg2
!= 1 || arg3
|| arg4
|| arg5
)
2259 current
->no_new_privs
= 1;
2261 case PR_GET_NO_NEW_PRIVS
:
2262 if (arg2
|| arg3
|| arg4
|| arg5
)
2264 return current
->no_new_privs
? 1 : 0;
2272 SYSCALL_DEFINE3(getcpu
, unsigned __user
*, cpup
, unsigned __user
*, nodep
,
2273 struct getcpu_cache __user
*, unused
)
2276 int cpu
= raw_smp_processor_id();
2278 err
|= put_user(cpu
, cpup
);
2280 err
|= put_user(cpu_to_node(cpu
), nodep
);
2281 return err
? -EFAULT
: 0;
2284 char poweroff_cmd
[POWEROFF_CMD_PATH_LEN
] = "/sbin/poweroff";
2286 static int __orderly_poweroff(bool force
)
2289 static char *envp
[] = {
2291 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
2296 argv
= argv_split(GFP_KERNEL
, poweroff_cmd
, NULL
);
2298 ret
= call_usermodehelper(argv
[0], argv
, envp
, UMH_WAIT_EXEC
);
2301 printk(KERN_WARNING
"%s failed to allocate memory for \"%s\"\n",
2302 __func__
, poweroff_cmd
);
2307 printk(KERN_WARNING
"Failed to start orderly shutdown: "
2308 "forcing the issue\n");
2310 * I guess this should try to kick off some daemon to sync and
2311 * poweroff asap. Or not even bother syncing if we're doing an
2312 * emergency shutdown?
2321 static bool poweroff_force
;
2323 static void poweroff_work_func(struct work_struct
*work
)
2325 __orderly_poweroff(poweroff_force
);
2328 static DECLARE_WORK(poweroff_work
, poweroff_work_func
);
2331 * orderly_poweroff - Trigger an orderly system poweroff
2332 * @force: force poweroff if command execution fails
2334 * This may be called from any context to trigger a system shutdown.
2335 * If the orderly shutdown fails, it will force an immediate shutdown.
2337 int orderly_poweroff(bool force
)
2339 if (force
) /* do not override the pending "true" */
2340 poweroff_force
= true;
2341 schedule_work(&poweroff_work
);
2344 EXPORT_SYMBOL_GPL(orderly_poweroff
);
2347 * do_sysinfo - fill in sysinfo struct
2348 * @info: pointer to buffer to fill
2350 static int do_sysinfo(struct sysinfo
*info
)
2352 unsigned long mem_total
, sav_total
;
2353 unsigned int mem_unit
, bitcount
;
2356 memset(info
, 0, sizeof(struct sysinfo
));
2359 monotonic_to_bootbased(&tp
);
2360 info
->uptime
= tp
.tv_sec
+ (tp
.tv_nsec
? 1 : 0);
2362 get_avenrun(info
->loads
, 0, SI_LOAD_SHIFT
- FSHIFT
);
2364 info
->procs
= nr_threads
;
2370 * If the sum of all the available memory (i.e. ram + swap)
2371 * is less than can be stored in a 32 bit unsigned long then
2372 * we can be binary compatible with 2.2.x kernels. If not,
2373 * well, in that case 2.2.x was broken anyways...
2375 * -Erik Andersen <andersee@debian.org>
2378 mem_total
= info
->totalram
+ info
->totalswap
;
2379 if (mem_total
< info
->totalram
|| mem_total
< info
->totalswap
)
2382 mem_unit
= info
->mem_unit
;
2383 while (mem_unit
> 1) {
2386 sav_total
= mem_total
;
2388 if (mem_total
< sav_total
)
2393 * If mem_total did not overflow, multiply all memory values by
2394 * info->mem_unit and set it to 1. This leaves things compatible
2395 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2400 info
->totalram
<<= bitcount
;
2401 info
->freeram
<<= bitcount
;
2402 info
->sharedram
<<= bitcount
;
2403 info
->bufferram
<<= bitcount
;
2404 info
->totalswap
<<= bitcount
;
2405 info
->freeswap
<<= bitcount
;
2406 info
->totalhigh
<<= bitcount
;
2407 info
->freehigh
<<= bitcount
;
2413 SYSCALL_DEFINE1(sysinfo
, struct sysinfo __user
*, info
)
2419 if (copy_to_user(info
, &val
, sizeof(struct sysinfo
)))
2425 #ifdef CONFIG_COMPAT
2426 struct compat_sysinfo
{
2440 char _f
[20-2*sizeof(u32
)-sizeof(int)];
2443 COMPAT_SYSCALL_DEFINE1(sysinfo
, struct compat_sysinfo __user
*, info
)
2449 /* Check to see if any memory value is too large for 32-bit and scale
2452 if ((s
.totalram
>> 32) || (s
.totalswap
>> 32)) {
2455 while (s
.mem_unit
< PAGE_SIZE
) {
2460 s
.totalram
>>= bitcount
;
2461 s
.freeram
>>= bitcount
;
2462 s
.sharedram
>>= bitcount
;
2463 s
.bufferram
>>= bitcount
;
2464 s
.totalswap
>>= bitcount
;
2465 s
.freeswap
>>= bitcount
;
2466 s
.totalhigh
>>= bitcount
;
2467 s
.freehigh
>>= bitcount
;
2470 if (!access_ok(VERIFY_WRITE
, info
, sizeof(struct compat_sysinfo
)) ||
2471 __put_user(s
.uptime
, &info
->uptime
) ||
2472 __put_user(s
.loads
[0], &info
->loads
[0]) ||
2473 __put_user(s
.loads
[1], &info
->loads
[1]) ||
2474 __put_user(s
.loads
[2], &info
->loads
[2]) ||
2475 __put_user(s
.totalram
, &info
->totalram
) ||
2476 __put_user(s
.freeram
, &info
->freeram
) ||
2477 __put_user(s
.sharedram
, &info
->sharedram
) ||
2478 __put_user(s
.bufferram
, &info
->bufferram
) ||
2479 __put_user(s
.totalswap
, &info
->totalswap
) ||
2480 __put_user(s
.freeswap
, &info
->freeswap
) ||
2481 __put_user(s
.procs
, &info
->procs
) ||
2482 __put_user(s
.totalhigh
, &info
->totalhigh
) ||
2483 __put_user(s
.freehigh
, &info
->freehigh
) ||
2484 __put_user(s
.mem_unit
, &info
->mem_unit
))
2489 #endif /* CONFIG_COMPAT */