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/workqueue.h>
20 #include <linux/capability.h>
21 #include <linux/device.h>
22 #include <linux/key.h>
23 #include <linux/times.h>
24 #include <linux/posix-timers.h>
25 #include <linux/security.h>
26 #include <linux/dcookies.h>
27 #include <linux/suspend.h>
28 #include <linux/tty.h>
29 #include <linux/signal.h>
30 #include <linux/cn_proc.h>
31 #include <linux/getcpu.h>
32 #include <linux/task_io_accounting_ops.h>
33 #include <linux/seccomp.h>
34 #include <linux/cpu.h>
35 #include <linux/personality.h>
36 #include <linux/ptrace.h>
37 #include <linux/fs_struct.h>
38 #include <linux/file.h>
39 #include <linux/mount.h>
40 #include <linux/gfp.h>
41 #include <linux/syscore_ops.h>
42 #include <linux/version.h>
43 #include <linux/ctype.h>
45 #include <linux/compat.h>
46 #include <linux/syscalls.h>
47 #include <linux/kprobes.h>
48 #include <linux/user_namespace.h>
49 #include <linux/binfmts.h>
51 #include <linux/sched.h>
52 #include <linux/rcupdate.h>
53 #include <linux/uidgid.h>
54 #include <linux/cred.h>
56 #include <linux/kmsg_dump.h>
57 /* Move somewhere else to avoid recompiling? */
58 #include <generated/utsrelease.h>
60 #include <asm/uaccess.h>
62 #include <asm/unistd.h>
64 #ifndef SET_UNALIGN_CTL
65 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
67 #ifndef GET_UNALIGN_CTL
68 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
71 # define SET_FPEMU_CTL(a,b) (-EINVAL)
74 # define GET_FPEMU_CTL(a,b) (-EINVAL)
77 # define SET_FPEXC_CTL(a,b) (-EINVAL)
80 # define GET_FPEXC_CTL(a,b) (-EINVAL)
83 # define GET_ENDIAN(a,b) (-EINVAL)
86 # define SET_ENDIAN(a,b) (-EINVAL)
89 # define GET_TSC_CTL(a) (-EINVAL)
92 # define SET_TSC_CTL(a) (-EINVAL)
96 * this is where the system-wide overflow UID and GID are defined, for
97 * architectures that now have 32-bit UID/GID but didn't in the past
100 int overflowuid
= DEFAULT_OVERFLOWUID
;
101 int overflowgid
= DEFAULT_OVERFLOWGID
;
103 EXPORT_SYMBOL(overflowuid
);
104 EXPORT_SYMBOL(overflowgid
);
107 * the same as above, but for filesystems which can only store a 16-bit
108 * UID and GID. as such, this is needed on all architectures
111 int fs_overflowuid
= DEFAULT_FS_OVERFLOWUID
;
112 int fs_overflowgid
= DEFAULT_FS_OVERFLOWUID
;
114 EXPORT_SYMBOL(fs_overflowuid
);
115 EXPORT_SYMBOL(fs_overflowgid
);
118 * Returns true if current's euid is same as p's uid or euid,
119 * or has CAP_SYS_NICE to p's user_ns.
121 * Called with rcu_read_lock, creds are safe
123 static bool set_one_prio_perm(struct task_struct
*p
)
125 const struct cred
*cred
= current_cred(), *pcred
= __task_cred(p
);
127 if (uid_eq(pcred
->uid
, cred
->euid
) ||
128 uid_eq(pcred
->euid
, cred
->euid
))
130 if (ns_capable(pcred
->user_ns
, CAP_SYS_NICE
))
136 * set the priority of a task
137 * - the caller must hold the RCU read lock
139 static int set_one_prio(struct task_struct
*p
, int niceval
, int error
)
143 if (!set_one_prio_perm(p
)) {
147 if (niceval
< task_nice(p
) && !can_nice(p
, niceval
)) {
151 no_nice
= security_task_setnice(p
, niceval
);
158 set_user_nice(p
, niceval
);
163 SYSCALL_DEFINE3(setpriority
, int, which
, int, who
, int, niceval
)
165 struct task_struct
*g
, *p
;
166 struct user_struct
*user
;
167 const struct cred
*cred
= current_cred();
172 if (which
> PRIO_USER
|| which
< PRIO_PROCESS
)
175 /* normalize: avoid signed division (rounding problems) */
183 read_lock(&tasklist_lock
);
187 p
= find_task_by_vpid(who
);
191 error
= set_one_prio(p
, niceval
, error
);
195 pgrp
= find_vpid(who
);
197 pgrp
= task_pgrp(current
);
198 do_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
) {
199 error
= set_one_prio(p
, niceval
, error
);
200 } while_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
);
203 uid
= make_kuid(cred
->user_ns
, who
);
207 else if (!uid_eq(uid
, cred
->uid
) &&
208 !(user
= find_user(uid
)))
209 goto out_unlock
; /* No processes for this user */
211 do_each_thread(g
, p
) {
212 if (uid_eq(task_uid(p
), uid
))
213 error
= set_one_prio(p
, niceval
, error
);
214 } while_each_thread(g
, p
);
215 if (!uid_eq(uid
, cred
->uid
))
216 free_uid(user
); /* For find_user() */
220 read_unlock(&tasklist_lock
);
227 * Ugh. To avoid negative return values, "getpriority()" will
228 * not return the normal nice-value, but a negated value that
229 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
230 * to stay compatible.
232 SYSCALL_DEFINE2(getpriority
, int, which
, int, who
)
234 struct task_struct
*g
, *p
;
235 struct user_struct
*user
;
236 const struct cred
*cred
= current_cred();
237 long niceval
, retval
= -ESRCH
;
241 if (which
> PRIO_USER
|| which
< PRIO_PROCESS
)
245 read_lock(&tasklist_lock
);
249 p
= find_task_by_vpid(who
);
253 niceval
= 20 - task_nice(p
);
254 if (niceval
> retval
)
260 pgrp
= find_vpid(who
);
262 pgrp
= task_pgrp(current
);
263 do_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
) {
264 niceval
= 20 - task_nice(p
);
265 if (niceval
> retval
)
267 } while_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
);
270 uid
= make_kuid(cred
->user_ns
, who
);
274 else if (!uid_eq(uid
, cred
->uid
) &&
275 !(user
= find_user(uid
)))
276 goto out_unlock
; /* No processes for this user */
278 do_each_thread(g
, p
) {
279 if (uid_eq(task_uid(p
), uid
)) {
280 niceval
= 20 - task_nice(p
);
281 if (niceval
> retval
)
284 } while_each_thread(g
, p
);
285 if (!uid_eq(uid
, cred
->uid
))
286 free_uid(user
); /* for find_user() */
290 read_unlock(&tasklist_lock
);
297 * Unprivileged users may change the real gid to the effective gid
298 * or vice versa. (BSD-style)
300 * If you set the real gid at all, or set the effective gid to a value not
301 * equal to the real gid, then the saved gid is set to the new effective gid.
303 * This makes it possible for a setgid program to completely drop its
304 * privileges, which is often a useful assertion to make when you are doing
305 * a security audit over a program.
307 * The general idea is that a program which uses just setregid() will be
308 * 100% compatible with BSD. A program which uses just setgid() will be
309 * 100% compatible with POSIX with saved IDs.
311 * SMP: There are not races, the GIDs are checked only by filesystem
312 * operations (as far as semantic preservation is concerned).
314 SYSCALL_DEFINE2(setregid
, gid_t
, rgid
, gid_t
, egid
)
316 struct user_namespace
*ns
= current_user_ns();
317 const struct cred
*old
;
322 krgid
= make_kgid(ns
, rgid
);
323 kegid
= make_kgid(ns
, egid
);
325 if ((rgid
!= (gid_t
) -1) && !gid_valid(krgid
))
327 if ((egid
!= (gid_t
) -1) && !gid_valid(kegid
))
330 new = prepare_creds();
333 old
= current_cred();
336 if (rgid
!= (gid_t
) -1) {
337 if (gid_eq(old
->gid
, krgid
) ||
338 gid_eq(old
->egid
, krgid
) ||
339 ns_capable(old
->user_ns
, CAP_SETGID
))
344 if (egid
!= (gid_t
) -1) {
345 if (gid_eq(old
->gid
, kegid
) ||
346 gid_eq(old
->egid
, kegid
) ||
347 gid_eq(old
->sgid
, kegid
) ||
348 ns_capable(old
->user_ns
, CAP_SETGID
))
354 if (rgid
!= (gid_t
) -1 ||
355 (egid
!= (gid_t
) -1 && !gid_eq(kegid
, old
->gid
)))
356 new->sgid
= new->egid
;
357 new->fsgid
= new->egid
;
359 return commit_creds(new);
367 * setgid() is implemented like SysV w/ SAVED_IDS
369 * SMP: Same implicit races as above.
371 SYSCALL_DEFINE1(setgid
, gid_t
, gid
)
373 struct user_namespace
*ns
= current_user_ns();
374 const struct cred
*old
;
379 kgid
= make_kgid(ns
, gid
);
380 if (!gid_valid(kgid
))
383 new = prepare_creds();
386 old
= current_cred();
389 if (ns_capable(old
->user_ns
, CAP_SETGID
))
390 new->gid
= new->egid
= new->sgid
= new->fsgid
= kgid
;
391 else if (gid_eq(kgid
, old
->gid
) || gid_eq(kgid
, old
->sgid
))
392 new->egid
= new->fsgid
= kgid
;
396 return commit_creds(new);
404 * change the user struct in a credentials set to match the new UID
406 static int set_user(struct cred
*new)
408 struct user_struct
*new_user
;
410 new_user
= alloc_uid(new->uid
);
415 * We don't fail in case of NPROC limit excess here because too many
416 * poorly written programs don't check set*uid() return code, assuming
417 * it never fails if called by root. We may still enforce NPROC limit
418 * for programs doing set*uid()+execve() by harmlessly deferring the
419 * failure to the execve() stage.
421 if (atomic_read(&new_user
->processes
) >= rlimit(RLIMIT_NPROC
) &&
422 new_user
!= INIT_USER
)
423 current
->flags
|= PF_NPROC_EXCEEDED
;
425 current
->flags
&= ~PF_NPROC_EXCEEDED
;
428 new->user
= new_user
;
433 * Unprivileged users may change the real uid to the effective uid
434 * or vice versa. (BSD-style)
436 * If you set the real uid at all, or set the effective uid to a value not
437 * equal to the real uid, then the saved uid is set to the new effective uid.
439 * This makes it possible for a setuid program to completely drop its
440 * privileges, which is often a useful assertion to make when you are doing
441 * a security audit over a program.
443 * The general idea is that a program which uses just setreuid() will be
444 * 100% compatible with BSD. A program which uses just setuid() will be
445 * 100% compatible with POSIX with saved IDs.
447 SYSCALL_DEFINE2(setreuid
, uid_t
, ruid
, uid_t
, euid
)
449 struct user_namespace
*ns
= current_user_ns();
450 const struct cred
*old
;
455 kruid
= make_kuid(ns
, ruid
);
456 keuid
= make_kuid(ns
, euid
);
458 if ((ruid
!= (uid_t
) -1) && !uid_valid(kruid
))
460 if ((euid
!= (uid_t
) -1) && !uid_valid(keuid
))
463 new = prepare_creds();
466 old
= current_cred();
469 if (ruid
!= (uid_t
) -1) {
471 if (!uid_eq(old
->uid
, kruid
) &&
472 !uid_eq(old
->euid
, kruid
) &&
473 !ns_capable(old
->user_ns
, CAP_SETUID
))
477 if (euid
!= (uid_t
) -1) {
479 if (!uid_eq(old
->uid
, keuid
) &&
480 !uid_eq(old
->euid
, keuid
) &&
481 !uid_eq(old
->suid
, keuid
) &&
482 !ns_capable(old
->user_ns
, CAP_SETUID
))
486 if (!uid_eq(new->uid
, old
->uid
)) {
487 retval
= set_user(new);
491 if (ruid
!= (uid_t
) -1 ||
492 (euid
!= (uid_t
) -1 && !uid_eq(keuid
, old
->uid
)))
493 new->suid
= new->euid
;
494 new->fsuid
= new->euid
;
496 retval
= security_task_fix_setuid(new, old
, LSM_SETID_RE
);
500 return commit_creds(new);
508 * setuid() is implemented like SysV with SAVED_IDS
510 * Note that SAVED_ID's is deficient in that a setuid root program
511 * like sendmail, for example, cannot set its uid to be a normal
512 * user and then switch back, because if you're root, setuid() sets
513 * the saved uid too. If you don't like this, blame the bright people
514 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
515 * will allow a root program to temporarily drop privileges and be able to
516 * regain them by swapping the real and effective uid.
518 SYSCALL_DEFINE1(setuid
, uid_t
, uid
)
520 struct user_namespace
*ns
= current_user_ns();
521 const struct cred
*old
;
526 kuid
= make_kuid(ns
, uid
);
527 if (!uid_valid(kuid
))
530 new = prepare_creds();
533 old
= current_cred();
536 if (ns_capable(old
->user_ns
, CAP_SETUID
)) {
537 new->suid
= new->uid
= kuid
;
538 if (!uid_eq(kuid
, old
->uid
)) {
539 retval
= set_user(new);
543 } else if (!uid_eq(kuid
, old
->uid
) && !uid_eq(kuid
, new->suid
)) {
547 new->fsuid
= new->euid
= kuid
;
549 retval
= security_task_fix_setuid(new, old
, LSM_SETID_ID
);
553 return commit_creds(new);
562 * This function implements a generic ability to update ruid, euid,
563 * and suid. This allows you to implement the 4.4 compatible seteuid().
565 SYSCALL_DEFINE3(setresuid
, uid_t
, ruid
, uid_t
, euid
, uid_t
, suid
)
567 struct user_namespace
*ns
= current_user_ns();
568 const struct cred
*old
;
571 kuid_t kruid
, keuid
, ksuid
;
573 kruid
= make_kuid(ns
, ruid
);
574 keuid
= make_kuid(ns
, euid
);
575 ksuid
= make_kuid(ns
, suid
);
577 if ((ruid
!= (uid_t
) -1) && !uid_valid(kruid
))
580 if ((euid
!= (uid_t
) -1) && !uid_valid(keuid
))
583 if ((suid
!= (uid_t
) -1) && !uid_valid(ksuid
))
586 new = prepare_creds();
590 old
= current_cred();
593 if (!ns_capable(old
->user_ns
, CAP_SETUID
)) {
594 if (ruid
!= (uid_t
) -1 && !uid_eq(kruid
, old
->uid
) &&
595 !uid_eq(kruid
, old
->euid
) && !uid_eq(kruid
, old
->suid
))
597 if (euid
!= (uid_t
) -1 && !uid_eq(keuid
, old
->uid
) &&
598 !uid_eq(keuid
, old
->euid
) && !uid_eq(keuid
, old
->suid
))
600 if (suid
!= (uid_t
) -1 && !uid_eq(ksuid
, old
->uid
) &&
601 !uid_eq(ksuid
, old
->euid
) && !uid_eq(ksuid
, old
->suid
))
605 if (ruid
!= (uid_t
) -1) {
607 if (!uid_eq(kruid
, old
->uid
)) {
608 retval
= set_user(new);
613 if (euid
!= (uid_t
) -1)
615 if (suid
!= (uid_t
) -1)
617 new->fsuid
= new->euid
;
619 retval
= security_task_fix_setuid(new, old
, LSM_SETID_RES
);
623 return commit_creds(new);
630 SYSCALL_DEFINE3(getresuid
, uid_t __user
*, ruidp
, uid_t __user
*, euidp
, uid_t __user
*, suidp
)
632 const struct cred
*cred
= current_cred();
634 uid_t ruid
, euid
, suid
;
636 ruid
= from_kuid_munged(cred
->user_ns
, cred
->uid
);
637 euid
= from_kuid_munged(cred
->user_ns
, cred
->euid
);
638 suid
= from_kuid_munged(cred
->user_ns
, cred
->suid
);
640 if (!(retval
= put_user(ruid
, ruidp
)) &&
641 !(retval
= put_user(euid
, euidp
)))
642 retval
= put_user(suid
, suidp
);
648 * Same as above, but for rgid, egid, sgid.
650 SYSCALL_DEFINE3(setresgid
, gid_t
, rgid
, gid_t
, egid
, gid_t
, sgid
)
652 struct user_namespace
*ns
= current_user_ns();
653 const struct cred
*old
;
656 kgid_t krgid
, kegid
, ksgid
;
658 krgid
= make_kgid(ns
, rgid
);
659 kegid
= make_kgid(ns
, egid
);
660 ksgid
= make_kgid(ns
, sgid
);
662 if ((rgid
!= (gid_t
) -1) && !gid_valid(krgid
))
664 if ((egid
!= (gid_t
) -1) && !gid_valid(kegid
))
666 if ((sgid
!= (gid_t
) -1) && !gid_valid(ksgid
))
669 new = prepare_creds();
672 old
= current_cred();
675 if (!ns_capable(old
->user_ns
, CAP_SETGID
)) {
676 if (rgid
!= (gid_t
) -1 && !gid_eq(krgid
, old
->gid
) &&
677 !gid_eq(krgid
, old
->egid
) && !gid_eq(krgid
, old
->sgid
))
679 if (egid
!= (gid_t
) -1 && !gid_eq(kegid
, old
->gid
) &&
680 !gid_eq(kegid
, old
->egid
) && !gid_eq(kegid
, old
->sgid
))
682 if (sgid
!= (gid_t
) -1 && !gid_eq(ksgid
, old
->gid
) &&
683 !gid_eq(ksgid
, old
->egid
) && !gid_eq(ksgid
, old
->sgid
))
687 if (rgid
!= (gid_t
) -1)
689 if (egid
!= (gid_t
) -1)
691 if (sgid
!= (gid_t
) -1)
693 new->fsgid
= new->egid
;
695 return commit_creds(new);
702 SYSCALL_DEFINE3(getresgid
, gid_t __user
*, rgidp
, gid_t __user
*, egidp
, gid_t __user
*, sgidp
)
704 const struct cred
*cred
= current_cred();
706 gid_t rgid
, egid
, sgid
;
708 rgid
= from_kgid_munged(cred
->user_ns
, cred
->gid
);
709 egid
= from_kgid_munged(cred
->user_ns
, cred
->egid
);
710 sgid
= from_kgid_munged(cred
->user_ns
, cred
->sgid
);
712 if (!(retval
= put_user(rgid
, rgidp
)) &&
713 !(retval
= put_user(egid
, egidp
)))
714 retval
= put_user(sgid
, sgidp
);
721 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
722 * is used for "access()" and for the NFS daemon (letting nfsd stay at
723 * whatever uid it wants to). It normally shadows "euid", except when
724 * explicitly set by setfsuid() or for access..
726 SYSCALL_DEFINE1(setfsuid
, uid_t
, uid
)
728 const struct cred
*old
;
733 old
= current_cred();
734 old_fsuid
= from_kuid_munged(old
->user_ns
, old
->fsuid
);
736 kuid
= make_kuid(old
->user_ns
, uid
);
737 if (!uid_valid(kuid
))
740 new = prepare_creds();
744 if (uid_eq(kuid
, old
->uid
) || uid_eq(kuid
, old
->euid
) ||
745 uid_eq(kuid
, old
->suid
) || uid_eq(kuid
, old
->fsuid
) ||
746 ns_capable(old
->user_ns
, CAP_SETUID
)) {
747 if (!uid_eq(kuid
, old
->fsuid
)) {
749 if (security_task_fix_setuid(new, old
, LSM_SETID_FS
) == 0)
763 * Samma på svenska..
765 SYSCALL_DEFINE1(setfsgid
, gid_t
, gid
)
767 const struct cred
*old
;
772 old
= current_cred();
773 old_fsgid
= from_kgid_munged(old
->user_ns
, old
->fsgid
);
775 kgid
= make_kgid(old
->user_ns
, gid
);
776 if (!gid_valid(kgid
))
779 new = prepare_creds();
783 if (gid_eq(kgid
, old
->gid
) || gid_eq(kgid
, old
->egid
) ||
784 gid_eq(kgid
, old
->sgid
) || gid_eq(kgid
, old
->fsgid
) ||
785 ns_capable(old
->user_ns
, CAP_SETGID
)) {
786 if (!gid_eq(kgid
, old
->fsgid
)) {
801 * sys_getpid - return the thread group id of the current process
803 * Note, despite the name, this returns the tgid not the pid. The tgid and
804 * the pid are identical unless CLONE_THREAD was specified on clone() in
805 * which case the tgid is the same in all threads of the same group.
807 * This is SMP safe as current->tgid does not change.
809 SYSCALL_DEFINE0(getpid
)
811 return task_tgid_vnr(current
);
814 /* Thread ID - the internal kernel "pid" */
815 SYSCALL_DEFINE0(gettid
)
817 return task_pid_vnr(current
);
821 * Accessing ->real_parent is not SMP-safe, it could
822 * change from under us. However, we can use a stale
823 * value of ->real_parent under rcu_read_lock(), see
824 * release_task()->call_rcu(delayed_put_task_struct).
826 SYSCALL_DEFINE0(getppid
)
831 pid
= task_tgid_vnr(rcu_dereference(current
->real_parent
));
837 SYSCALL_DEFINE0(getuid
)
839 /* Only we change this so SMP safe */
840 return from_kuid_munged(current_user_ns(), current_uid());
843 SYSCALL_DEFINE0(geteuid
)
845 /* Only we change this so SMP safe */
846 return from_kuid_munged(current_user_ns(), current_euid());
849 SYSCALL_DEFINE0(getgid
)
851 /* Only we change this so SMP safe */
852 return from_kgid_munged(current_user_ns(), current_gid());
855 SYSCALL_DEFINE0(getegid
)
857 /* Only we change this so SMP safe */
858 return from_kgid_munged(current_user_ns(), current_egid());
861 void do_sys_times(struct tms
*tms
)
863 cputime_t tgutime
, tgstime
, cutime
, cstime
;
865 spin_lock_irq(¤t
->sighand
->siglock
);
866 thread_group_cputime_adjusted(current
, &tgutime
, &tgstime
);
867 cutime
= current
->signal
->cutime
;
868 cstime
= current
->signal
->cstime
;
869 spin_unlock_irq(¤t
->sighand
->siglock
);
870 tms
->tms_utime
= cputime_to_clock_t(tgutime
);
871 tms
->tms_stime
= cputime_to_clock_t(tgstime
);
872 tms
->tms_cutime
= cputime_to_clock_t(cutime
);
873 tms
->tms_cstime
= cputime_to_clock_t(cstime
);
876 SYSCALL_DEFINE1(times
, struct tms __user
*, tbuf
)
882 if (copy_to_user(tbuf
, &tmp
, sizeof(struct tms
)))
885 force_successful_syscall_return();
886 return (long) jiffies_64_to_clock_t(get_jiffies_64());
890 * This needs some heavy checking ...
891 * I just haven't the stomach for it. I also don't fully
892 * understand sessions/pgrp etc. Let somebody who does explain it.
894 * OK, I think I have the protection semantics right.... this is really
895 * only important on a multi-user system anyway, to make sure one user
896 * can't send a signal to a process owned by another. -TYT, 12/12/91
898 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
901 SYSCALL_DEFINE2(setpgid
, pid_t
, pid
, pid_t
, pgid
)
903 struct task_struct
*p
;
904 struct task_struct
*group_leader
= current
->group_leader
;
909 pid
= task_pid_vnr(group_leader
);
916 /* From this point forward we keep holding onto the tasklist lock
917 * so that our parent does not change from under us. -DaveM
919 write_lock_irq(&tasklist_lock
);
922 p
= find_task_by_vpid(pid
);
927 if (!thread_group_leader(p
))
930 if (same_thread_group(p
->real_parent
, group_leader
)) {
932 if (task_session(p
) != task_session(group_leader
))
939 if (p
!= group_leader
)
944 if (p
->signal
->leader
)
949 struct task_struct
*g
;
951 pgrp
= find_vpid(pgid
);
952 g
= pid_task(pgrp
, PIDTYPE_PGID
);
953 if (!g
|| task_session(g
) != task_session(group_leader
))
957 err
= security_task_setpgid(p
, pgid
);
961 if (task_pgrp(p
) != pgrp
)
962 change_pid(p
, PIDTYPE_PGID
, pgrp
);
966 /* All paths lead to here, thus we are safe. -DaveM */
967 write_unlock_irq(&tasklist_lock
);
972 SYSCALL_DEFINE1(getpgid
, pid_t
, pid
)
974 struct task_struct
*p
;
980 grp
= task_pgrp(current
);
983 p
= find_task_by_vpid(pid
);
990 retval
= security_task_getpgid(p
);
994 retval
= pid_vnr(grp
);
1000 #ifdef __ARCH_WANT_SYS_GETPGRP
1002 SYSCALL_DEFINE0(getpgrp
)
1004 return sys_getpgid(0);
1009 SYSCALL_DEFINE1(getsid
, pid_t
, pid
)
1011 struct task_struct
*p
;
1017 sid
= task_session(current
);
1020 p
= find_task_by_vpid(pid
);
1023 sid
= task_session(p
);
1027 retval
= security_task_getsid(p
);
1031 retval
= pid_vnr(sid
);
1037 static void set_special_pids(struct pid
*pid
)
1039 struct task_struct
*curr
= current
->group_leader
;
1041 if (task_session(curr
) != pid
)
1042 change_pid(curr
, PIDTYPE_SID
, pid
);
1044 if (task_pgrp(curr
) != pid
)
1045 change_pid(curr
, PIDTYPE_PGID
, pid
);
1048 SYSCALL_DEFINE0(setsid
)
1050 struct task_struct
*group_leader
= current
->group_leader
;
1051 struct pid
*sid
= task_pid(group_leader
);
1052 pid_t session
= pid_vnr(sid
);
1055 write_lock_irq(&tasklist_lock
);
1056 /* Fail if I am already a session leader */
1057 if (group_leader
->signal
->leader
)
1060 /* Fail if a process group id already exists that equals the
1061 * proposed session id.
1063 if (pid_task(sid
, PIDTYPE_PGID
))
1066 group_leader
->signal
->leader
= 1;
1067 set_special_pids(sid
);
1069 proc_clear_tty(group_leader
);
1073 write_unlock_irq(&tasklist_lock
);
1075 proc_sid_connector(group_leader
);
1076 sched_autogroup_create_attach(group_leader
);
1081 DECLARE_RWSEM(uts_sem
);
1083 #ifdef COMPAT_UTS_MACHINE
1084 #define override_architecture(name) \
1085 (personality(current->personality) == PER_LINUX32 && \
1086 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1087 sizeof(COMPAT_UTS_MACHINE)))
1089 #define override_architecture(name) 0
1093 * Work around broken programs that cannot handle "Linux 3.0".
1094 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1096 static int override_release(char __user
*release
, size_t len
)
1100 if (current
->personality
& UNAME26
) {
1101 const char *rest
= UTS_RELEASE
;
1102 char buf
[65] = { 0 };
1108 if (*rest
== '.' && ++ndots
>= 3)
1110 if (!isdigit(*rest
) && *rest
!= '.')
1114 v
= ((LINUX_VERSION_CODE
>> 8) & 0xff) + 40;
1115 copy
= clamp_t(size_t, len
, 1, sizeof(buf
));
1116 copy
= scnprintf(buf
, copy
, "2.6.%u%s", v
, rest
);
1117 ret
= copy_to_user(release
, buf
, copy
+ 1);
1122 SYSCALL_DEFINE1(newuname
, struct new_utsname __user
*, name
)
1126 down_read(&uts_sem
);
1127 if (copy_to_user(name
, utsname(), sizeof *name
))
1131 if (!errno
&& override_release(name
->release
, sizeof(name
->release
)))
1133 if (!errno
&& override_architecture(name
))
1138 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1142 SYSCALL_DEFINE1(uname
, struct old_utsname __user
*, name
)
1149 down_read(&uts_sem
);
1150 if (copy_to_user(name
, utsname(), sizeof(*name
)))
1154 if (!error
&& override_release(name
->release
, sizeof(name
->release
)))
1156 if (!error
&& override_architecture(name
))
1161 SYSCALL_DEFINE1(olduname
, struct oldold_utsname __user
*, name
)
1167 if (!access_ok(VERIFY_WRITE
, name
, sizeof(struct oldold_utsname
)))
1170 down_read(&uts_sem
);
1171 error
= __copy_to_user(&name
->sysname
, &utsname()->sysname
,
1173 error
|= __put_user(0, name
->sysname
+ __OLD_UTS_LEN
);
1174 error
|= __copy_to_user(&name
->nodename
, &utsname()->nodename
,
1176 error
|= __put_user(0, name
->nodename
+ __OLD_UTS_LEN
);
1177 error
|= __copy_to_user(&name
->release
, &utsname()->release
,
1179 error
|= __put_user(0, name
->release
+ __OLD_UTS_LEN
);
1180 error
|= __copy_to_user(&name
->version
, &utsname()->version
,
1182 error
|= __put_user(0, name
->version
+ __OLD_UTS_LEN
);
1183 error
|= __copy_to_user(&name
->machine
, &utsname()->machine
,
1185 error
|= __put_user(0, name
->machine
+ __OLD_UTS_LEN
);
1188 if (!error
&& override_architecture(name
))
1190 if (!error
&& override_release(name
->release
, sizeof(name
->release
)))
1192 return error
? -EFAULT
: 0;
1196 SYSCALL_DEFINE2(sethostname
, char __user
*, name
, int, len
)
1199 char tmp
[__NEW_UTS_LEN
];
1201 if (!ns_capable(current
->nsproxy
->uts_ns
->user_ns
, CAP_SYS_ADMIN
))
1204 if (len
< 0 || len
> __NEW_UTS_LEN
)
1206 down_write(&uts_sem
);
1208 if (!copy_from_user(tmp
, name
, len
)) {
1209 struct new_utsname
*u
= utsname();
1211 memcpy(u
->nodename
, tmp
, len
);
1212 memset(u
->nodename
+ len
, 0, sizeof(u
->nodename
) - len
);
1214 uts_proc_notify(UTS_PROC_HOSTNAME
);
1220 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1222 SYSCALL_DEFINE2(gethostname
, char __user
*, name
, int, len
)
1225 struct new_utsname
*u
;
1229 down_read(&uts_sem
);
1231 i
= 1 + strlen(u
->nodename
);
1235 if (copy_to_user(name
, u
->nodename
, i
))
1244 * Only setdomainname; getdomainname can be implemented by calling
1247 SYSCALL_DEFINE2(setdomainname
, char __user
*, name
, int, len
)
1250 char tmp
[__NEW_UTS_LEN
];
1252 if (!ns_capable(current
->nsproxy
->uts_ns
->user_ns
, CAP_SYS_ADMIN
))
1254 if (len
< 0 || len
> __NEW_UTS_LEN
)
1257 down_write(&uts_sem
);
1259 if (!copy_from_user(tmp
, name
, len
)) {
1260 struct new_utsname
*u
= utsname();
1262 memcpy(u
->domainname
, tmp
, len
);
1263 memset(u
->domainname
+ len
, 0, sizeof(u
->domainname
) - len
);
1265 uts_proc_notify(UTS_PROC_DOMAINNAME
);
1271 SYSCALL_DEFINE2(getrlimit
, unsigned int, resource
, struct rlimit __user
*, rlim
)
1273 struct rlimit value
;
1276 ret
= do_prlimit(current
, resource
, NULL
, &value
);
1278 ret
= copy_to_user(rlim
, &value
, sizeof(*rlim
)) ? -EFAULT
: 0;
1283 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1286 * Back compatibility for getrlimit. Needed for some apps.
1289 SYSCALL_DEFINE2(old_getrlimit
, unsigned int, resource
,
1290 struct rlimit __user
*, rlim
)
1293 if (resource
>= RLIM_NLIMITS
)
1296 task_lock(current
->group_leader
);
1297 x
= current
->signal
->rlim
[resource
];
1298 task_unlock(current
->group_leader
);
1299 if (x
.rlim_cur
> 0x7FFFFFFF)
1300 x
.rlim_cur
= 0x7FFFFFFF;
1301 if (x
.rlim_max
> 0x7FFFFFFF)
1302 x
.rlim_max
= 0x7FFFFFFF;
1303 return copy_to_user(rlim
, &x
, sizeof(x
))?-EFAULT
:0;
1308 static inline bool rlim64_is_infinity(__u64 rlim64
)
1310 #if BITS_PER_LONG < 64
1311 return rlim64
>= ULONG_MAX
;
1313 return rlim64
== RLIM64_INFINITY
;
1317 static void rlim_to_rlim64(const struct rlimit
*rlim
, struct rlimit64
*rlim64
)
1319 if (rlim
->rlim_cur
== RLIM_INFINITY
)
1320 rlim64
->rlim_cur
= RLIM64_INFINITY
;
1322 rlim64
->rlim_cur
= rlim
->rlim_cur
;
1323 if (rlim
->rlim_max
== RLIM_INFINITY
)
1324 rlim64
->rlim_max
= RLIM64_INFINITY
;
1326 rlim64
->rlim_max
= rlim
->rlim_max
;
1329 static void rlim64_to_rlim(const struct rlimit64
*rlim64
, struct rlimit
*rlim
)
1331 if (rlim64_is_infinity(rlim64
->rlim_cur
))
1332 rlim
->rlim_cur
= RLIM_INFINITY
;
1334 rlim
->rlim_cur
= (unsigned long)rlim64
->rlim_cur
;
1335 if (rlim64_is_infinity(rlim64
->rlim_max
))
1336 rlim
->rlim_max
= RLIM_INFINITY
;
1338 rlim
->rlim_max
= (unsigned long)rlim64
->rlim_max
;
1341 /* make sure you are allowed to change @tsk limits before calling this */
1342 int do_prlimit(struct task_struct
*tsk
, unsigned int resource
,
1343 struct rlimit
*new_rlim
, struct rlimit
*old_rlim
)
1345 struct rlimit
*rlim
;
1348 if (resource
>= RLIM_NLIMITS
)
1351 if (new_rlim
->rlim_cur
> new_rlim
->rlim_max
)
1353 if (resource
== RLIMIT_NOFILE
&&
1354 new_rlim
->rlim_max
> sysctl_nr_open
)
1358 /* protect tsk->signal and tsk->sighand from disappearing */
1359 read_lock(&tasklist_lock
);
1360 if (!tsk
->sighand
) {
1365 rlim
= tsk
->signal
->rlim
+ resource
;
1366 task_lock(tsk
->group_leader
);
1368 /* Keep the capable check against init_user_ns until
1369 cgroups can contain all limits */
1370 if (new_rlim
->rlim_max
> rlim
->rlim_max
&&
1371 !capable(CAP_SYS_RESOURCE
))
1374 retval
= security_task_setrlimit(tsk
->group_leader
,
1375 resource
, new_rlim
);
1376 if (resource
== RLIMIT_CPU
&& new_rlim
->rlim_cur
== 0) {
1378 * The caller is asking for an immediate RLIMIT_CPU
1379 * expiry. But we use the zero value to mean "it was
1380 * never set". So let's cheat and make it one second
1383 new_rlim
->rlim_cur
= 1;
1392 task_unlock(tsk
->group_leader
);
1395 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1396 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1397 * very long-standing error, and fixing it now risks breakage of
1398 * applications, so we live with it
1400 if (!retval
&& new_rlim
&& resource
== RLIMIT_CPU
&&
1401 new_rlim
->rlim_cur
!= RLIM_INFINITY
)
1402 update_rlimit_cpu(tsk
, new_rlim
->rlim_cur
);
1404 read_unlock(&tasklist_lock
);
1408 /* rcu lock must be held */
1409 static int check_prlimit_permission(struct task_struct
*task
)
1411 const struct cred
*cred
= current_cred(), *tcred
;
1413 if (current
== task
)
1416 tcred
= __task_cred(task
);
1417 if (uid_eq(cred
->uid
, tcred
->euid
) &&
1418 uid_eq(cred
->uid
, tcred
->suid
) &&
1419 uid_eq(cred
->uid
, tcred
->uid
) &&
1420 gid_eq(cred
->gid
, tcred
->egid
) &&
1421 gid_eq(cred
->gid
, tcred
->sgid
) &&
1422 gid_eq(cred
->gid
, tcred
->gid
))
1424 if (ns_capable(tcred
->user_ns
, CAP_SYS_RESOURCE
))
1430 SYSCALL_DEFINE4(prlimit64
, pid_t
, pid
, unsigned int, resource
,
1431 const struct rlimit64 __user
*, new_rlim
,
1432 struct rlimit64 __user
*, old_rlim
)
1434 struct rlimit64 old64
, new64
;
1435 struct rlimit old
, new;
1436 struct task_struct
*tsk
;
1440 if (copy_from_user(&new64
, new_rlim
, sizeof(new64
)))
1442 rlim64_to_rlim(&new64
, &new);
1446 tsk
= pid
? find_task_by_vpid(pid
) : current
;
1451 ret
= check_prlimit_permission(tsk
);
1456 get_task_struct(tsk
);
1459 ret
= do_prlimit(tsk
, resource
, new_rlim
? &new : NULL
,
1460 old_rlim
? &old
: NULL
);
1462 if (!ret
&& old_rlim
) {
1463 rlim_to_rlim64(&old
, &old64
);
1464 if (copy_to_user(old_rlim
, &old64
, sizeof(old64
)))
1468 put_task_struct(tsk
);
1472 SYSCALL_DEFINE2(setrlimit
, unsigned int, resource
, struct rlimit __user
*, rlim
)
1474 struct rlimit new_rlim
;
1476 if (copy_from_user(&new_rlim
, rlim
, sizeof(*rlim
)))
1478 return do_prlimit(current
, resource
, &new_rlim
, NULL
);
1482 * It would make sense to put struct rusage in the task_struct,
1483 * except that would make the task_struct be *really big*. After
1484 * task_struct gets moved into malloc'ed memory, it would
1485 * make sense to do this. It will make moving the rest of the information
1486 * a lot simpler! (Which we're not doing right now because we're not
1487 * measuring them yet).
1489 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1490 * races with threads incrementing their own counters. But since word
1491 * reads are atomic, we either get new values or old values and we don't
1492 * care which for the sums. We always take the siglock to protect reading
1493 * the c* fields from p->signal from races with exit.c updating those
1494 * fields when reaping, so a sample either gets all the additions of a
1495 * given child after it's reaped, or none so this sample is before reaping.
1498 * We need to take the siglock for CHILDEREN, SELF and BOTH
1499 * for the cases current multithreaded, non-current single threaded
1500 * non-current multithreaded. Thread traversal is now safe with
1502 * Strictly speaking, we donot need to take the siglock if we are current and
1503 * single threaded, as no one else can take our signal_struct away, no one
1504 * else can reap the children to update signal->c* counters, and no one else
1505 * can race with the signal-> fields. If we do not take any lock, the
1506 * signal-> fields could be read out of order while another thread was just
1507 * exiting. So we should place a read memory barrier when we avoid the lock.
1508 * On the writer side, write memory barrier is implied in __exit_signal
1509 * as __exit_signal releases the siglock spinlock after updating the signal->
1510 * fields. But we don't do this yet to keep things simple.
1514 static void accumulate_thread_rusage(struct task_struct
*t
, struct rusage
*r
)
1516 r
->ru_nvcsw
+= t
->nvcsw
;
1517 r
->ru_nivcsw
+= t
->nivcsw
;
1518 r
->ru_minflt
+= t
->min_flt
;
1519 r
->ru_majflt
+= t
->maj_flt
;
1520 r
->ru_inblock
+= task_io_get_inblock(t
);
1521 r
->ru_oublock
+= task_io_get_oublock(t
);
1524 static void k_getrusage(struct task_struct
*p
, int who
, struct rusage
*r
)
1526 struct task_struct
*t
;
1527 unsigned long flags
;
1528 cputime_t tgutime
, tgstime
, utime
, stime
;
1529 unsigned long maxrss
= 0;
1531 memset((char *) r
, 0, sizeof *r
);
1534 if (who
== RUSAGE_THREAD
) {
1535 task_cputime_adjusted(current
, &utime
, &stime
);
1536 accumulate_thread_rusage(p
, r
);
1537 maxrss
= p
->signal
->maxrss
;
1541 if (!lock_task_sighand(p
, &flags
))
1546 case RUSAGE_CHILDREN
:
1547 utime
= p
->signal
->cutime
;
1548 stime
= p
->signal
->cstime
;
1549 r
->ru_nvcsw
= p
->signal
->cnvcsw
;
1550 r
->ru_nivcsw
= p
->signal
->cnivcsw
;
1551 r
->ru_minflt
= p
->signal
->cmin_flt
;
1552 r
->ru_majflt
= p
->signal
->cmaj_flt
;
1553 r
->ru_inblock
= p
->signal
->cinblock
;
1554 r
->ru_oublock
= p
->signal
->coublock
;
1555 maxrss
= p
->signal
->cmaxrss
;
1557 if (who
== RUSAGE_CHILDREN
)
1561 thread_group_cputime_adjusted(p
, &tgutime
, &tgstime
);
1564 r
->ru_nvcsw
+= p
->signal
->nvcsw
;
1565 r
->ru_nivcsw
+= p
->signal
->nivcsw
;
1566 r
->ru_minflt
+= p
->signal
->min_flt
;
1567 r
->ru_majflt
+= p
->signal
->maj_flt
;
1568 r
->ru_inblock
+= p
->signal
->inblock
;
1569 r
->ru_oublock
+= p
->signal
->oublock
;
1570 if (maxrss
< p
->signal
->maxrss
)
1571 maxrss
= p
->signal
->maxrss
;
1574 accumulate_thread_rusage(t
, r
);
1582 unlock_task_sighand(p
, &flags
);
1585 cputime_to_timeval(utime
, &r
->ru_utime
);
1586 cputime_to_timeval(stime
, &r
->ru_stime
);
1588 if (who
!= RUSAGE_CHILDREN
) {
1589 struct mm_struct
*mm
= get_task_mm(p
);
1591 setmax_mm_hiwater_rss(&maxrss
, mm
);
1595 r
->ru_maxrss
= maxrss
* (PAGE_SIZE
/ 1024); /* convert pages to KBs */
1598 int getrusage(struct task_struct
*p
, int who
, struct rusage __user
*ru
)
1601 k_getrusage(p
, who
, &r
);
1602 return copy_to_user(ru
, &r
, sizeof(r
)) ? -EFAULT
: 0;
1605 SYSCALL_DEFINE2(getrusage
, int, who
, struct rusage __user
*, ru
)
1607 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
&&
1608 who
!= RUSAGE_THREAD
)
1610 return getrusage(current
, who
, ru
);
1613 #ifdef CONFIG_COMPAT
1614 COMPAT_SYSCALL_DEFINE2(getrusage
, int, who
, struct compat_rusage __user
*, ru
)
1618 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
&&
1619 who
!= RUSAGE_THREAD
)
1622 k_getrusage(current
, who
, &r
);
1623 return put_compat_rusage(&r
, ru
);
1627 SYSCALL_DEFINE1(umask
, int, mask
)
1629 mask
= xchg(¤t
->fs
->umask
, mask
& S_IRWXUGO
);
1633 static int prctl_set_mm_exe_file(struct mm_struct
*mm
, unsigned int fd
)
1636 struct inode
*inode
;
1643 inode
= file_inode(exe
.file
);
1646 * Because the original mm->exe_file points to executable file, make
1647 * sure that this one is executable as well, to avoid breaking an
1651 if (!S_ISREG(inode
->i_mode
) ||
1652 exe
.file
->f_path
.mnt
->mnt_flags
& MNT_NOEXEC
)
1655 err
= inode_permission(inode
, MAY_EXEC
);
1659 down_write(&mm
->mmap_sem
);
1662 * Forbid mm->exe_file change if old file still mapped.
1666 struct vm_area_struct
*vma
;
1668 for (vma
= mm
->mmap
; vma
; vma
= vma
->vm_next
)
1670 path_equal(&vma
->vm_file
->f_path
,
1671 &mm
->exe_file
->f_path
))
1676 * The symlink can be changed only once, just to disallow arbitrary
1677 * transitions malicious software might bring in. This means one
1678 * could make a snapshot over all processes running and monitor
1679 * /proc/pid/exe changes to notice unusual activity if needed.
1682 if (test_and_set_bit(MMF_EXE_FILE_CHANGED
, &mm
->flags
))
1686 set_mm_exe_file(mm
, exe
.file
); /* this grabs a reference to exe.file */
1688 up_write(&mm
->mmap_sem
);
1695 static int prctl_set_mm(int opt
, unsigned long addr
,
1696 unsigned long arg4
, unsigned long arg5
)
1698 unsigned long rlim
= rlimit(RLIMIT_DATA
);
1699 struct mm_struct
*mm
= current
->mm
;
1700 struct vm_area_struct
*vma
;
1703 if (arg5
|| (arg4
&& opt
!= PR_SET_MM_AUXV
))
1706 if (!capable(CAP_SYS_RESOURCE
))
1709 if (opt
== PR_SET_MM_EXE_FILE
)
1710 return prctl_set_mm_exe_file(mm
, (unsigned int)addr
);
1712 if (addr
>= TASK_SIZE
|| addr
< mmap_min_addr
)
1717 down_read(&mm
->mmap_sem
);
1718 vma
= find_vma(mm
, addr
);
1721 case PR_SET_MM_START_CODE
:
1722 mm
->start_code
= addr
;
1724 case PR_SET_MM_END_CODE
:
1725 mm
->end_code
= addr
;
1727 case PR_SET_MM_START_DATA
:
1728 mm
->start_data
= addr
;
1730 case PR_SET_MM_END_DATA
:
1731 mm
->end_data
= addr
;
1734 case PR_SET_MM_START_BRK
:
1735 if (addr
<= mm
->end_data
)
1738 if (rlim
< RLIM_INFINITY
&&
1740 (mm
->end_data
- mm
->start_data
) > rlim
)
1743 mm
->start_brk
= addr
;
1747 if (addr
<= mm
->end_data
)
1750 if (rlim
< RLIM_INFINITY
&&
1751 (addr
- mm
->start_brk
) +
1752 (mm
->end_data
- mm
->start_data
) > rlim
)
1759 * If command line arguments and environment
1760 * are placed somewhere else on stack, we can
1761 * set them up here, ARG_START/END to setup
1762 * command line argumets and ENV_START/END
1765 case PR_SET_MM_START_STACK
:
1766 case PR_SET_MM_ARG_START
:
1767 case PR_SET_MM_ARG_END
:
1768 case PR_SET_MM_ENV_START
:
1769 case PR_SET_MM_ENV_END
:
1774 if (opt
== PR_SET_MM_START_STACK
)
1775 mm
->start_stack
= addr
;
1776 else if (opt
== PR_SET_MM_ARG_START
)
1777 mm
->arg_start
= addr
;
1778 else if (opt
== PR_SET_MM_ARG_END
)
1780 else if (opt
== PR_SET_MM_ENV_START
)
1781 mm
->env_start
= addr
;
1782 else if (opt
== PR_SET_MM_ENV_END
)
1787 * This doesn't move auxiliary vector itself
1788 * since it's pinned to mm_struct, but allow
1789 * to fill vector with new values. It's up
1790 * to a caller to provide sane values here
1791 * otherwise user space tools which use this
1792 * vector might be unhappy.
1794 case PR_SET_MM_AUXV
: {
1795 unsigned long user_auxv
[AT_VECTOR_SIZE
];
1797 if (arg4
> sizeof(user_auxv
))
1799 up_read(&mm
->mmap_sem
);
1801 if (copy_from_user(user_auxv
, (const void __user
*)addr
, arg4
))
1804 /* Make sure the last entry is always AT_NULL */
1805 user_auxv
[AT_VECTOR_SIZE
- 2] = 0;
1806 user_auxv
[AT_VECTOR_SIZE
- 1] = 0;
1808 BUILD_BUG_ON(sizeof(user_auxv
) != sizeof(mm
->saved_auxv
));
1811 memcpy(mm
->saved_auxv
, user_auxv
, arg4
);
1812 task_unlock(current
);
1822 up_read(&mm
->mmap_sem
);
1826 #ifdef CONFIG_CHECKPOINT_RESTORE
1827 static int prctl_get_tid_address(struct task_struct
*me
, int __user
**tid_addr
)
1829 return put_user(me
->clear_child_tid
, tid_addr
);
1832 static int prctl_get_tid_address(struct task_struct
*me
, int __user
**tid_addr
)
1838 SYSCALL_DEFINE5(prctl
, int, option
, unsigned long, arg2
, unsigned long, arg3
,
1839 unsigned long, arg4
, unsigned long, arg5
)
1841 struct task_struct
*me
= current
;
1842 unsigned char comm
[sizeof(me
->comm
)];
1845 error
= security_task_prctl(option
, arg2
, arg3
, arg4
, arg5
);
1846 if (error
!= -ENOSYS
)
1851 case PR_SET_PDEATHSIG
:
1852 if (!valid_signal(arg2
)) {
1856 me
->pdeath_signal
= arg2
;
1858 case PR_GET_PDEATHSIG
:
1859 error
= put_user(me
->pdeath_signal
, (int __user
*)arg2
);
1861 case PR_GET_DUMPABLE
:
1862 error
= get_dumpable(me
->mm
);
1864 case PR_SET_DUMPABLE
:
1865 if (arg2
!= SUID_DUMP_DISABLE
&& arg2
!= SUID_DUMP_USER
) {
1869 set_dumpable(me
->mm
, arg2
);
1872 case PR_SET_UNALIGN
:
1873 error
= SET_UNALIGN_CTL(me
, arg2
);
1875 case PR_GET_UNALIGN
:
1876 error
= GET_UNALIGN_CTL(me
, arg2
);
1879 error
= SET_FPEMU_CTL(me
, arg2
);
1882 error
= GET_FPEMU_CTL(me
, arg2
);
1885 error
= SET_FPEXC_CTL(me
, arg2
);
1888 error
= GET_FPEXC_CTL(me
, arg2
);
1891 error
= PR_TIMING_STATISTICAL
;
1894 if (arg2
!= PR_TIMING_STATISTICAL
)
1898 comm
[sizeof(me
->comm
) - 1] = 0;
1899 if (strncpy_from_user(comm
, (char __user
*)arg2
,
1900 sizeof(me
->comm
) - 1) < 0)
1902 set_task_comm(me
, comm
);
1903 proc_comm_connector(me
);
1906 get_task_comm(comm
, me
);
1907 if (copy_to_user((char __user
*)arg2
, comm
, sizeof(comm
)))
1911 error
= GET_ENDIAN(me
, arg2
);
1914 error
= SET_ENDIAN(me
, arg2
);
1916 case PR_GET_SECCOMP
:
1917 error
= prctl_get_seccomp();
1919 case PR_SET_SECCOMP
:
1920 error
= prctl_set_seccomp(arg2
, (char __user
*)arg3
);
1923 error
= GET_TSC_CTL(arg2
);
1926 error
= SET_TSC_CTL(arg2
);
1928 case PR_TASK_PERF_EVENTS_DISABLE
:
1929 error
= perf_event_task_disable();
1931 case PR_TASK_PERF_EVENTS_ENABLE
:
1932 error
= perf_event_task_enable();
1934 case PR_GET_TIMERSLACK
:
1935 error
= current
->timer_slack_ns
;
1937 case PR_SET_TIMERSLACK
:
1939 current
->timer_slack_ns
=
1940 current
->default_timer_slack_ns
;
1942 current
->timer_slack_ns
= arg2
;
1948 case PR_MCE_KILL_CLEAR
:
1951 current
->flags
&= ~PF_MCE_PROCESS
;
1953 case PR_MCE_KILL_SET
:
1954 current
->flags
|= PF_MCE_PROCESS
;
1955 if (arg3
== PR_MCE_KILL_EARLY
)
1956 current
->flags
|= PF_MCE_EARLY
;
1957 else if (arg3
== PR_MCE_KILL_LATE
)
1958 current
->flags
&= ~PF_MCE_EARLY
;
1959 else if (arg3
== PR_MCE_KILL_DEFAULT
)
1961 ~(PF_MCE_EARLY
|PF_MCE_PROCESS
);
1969 case PR_MCE_KILL_GET
:
1970 if (arg2
| arg3
| arg4
| arg5
)
1972 if (current
->flags
& PF_MCE_PROCESS
)
1973 error
= (current
->flags
& PF_MCE_EARLY
) ?
1974 PR_MCE_KILL_EARLY
: PR_MCE_KILL_LATE
;
1976 error
= PR_MCE_KILL_DEFAULT
;
1979 error
= prctl_set_mm(arg2
, arg3
, arg4
, arg5
);
1981 case PR_GET_TID_ADDRESS
:
1982 error
= prctl_get_tid_address(me
, (int __user
**)arg2
);
1984 case PR_SET_CHILD_SUBREAPER
:
1985 me
->signal
->is_child_subreaper
= !!arg2
;
1987 case PR_GET_CHILD_SUBREAPER
:
1988 error
= put_user(me
->signal
->is_child_subreaper
,
1989 (int __user
*)arg2
);
1991 case PR_SET_NO_NEW_PRIVS
:
1992 if (arg2
!= 1 || arg3
|| arg4
|| arg5
)
1995 current
->no_new_privs
= 1;
1997 case PR_GET_NO_NEW_PRIVS
:
1998 if (arg2
|| arg3
|| arg4
|| arg5
)
2000 return current
->no_new_privs
? 1 : 0;
2008 SYSCALL_DEFINE3(getcpu
, unsigned __user
*, cpup
, unsigned __user
*, nodep
,
2009 struct getcpu_cache __user
*, unused
)
2012 int cpu
= raw_smp_processor_id();
2014 err
|= put_user(cpu
, cpup
);
2016 err
|= put_user(cpu_to_node(cpu
), nodep
);
2017 return err
? -EFAULT
: 0;
2021 * do_sysinfo - fill in sysinfo struct
2022 * @info: pointer to buffer to fill
2024 static int do_sysinfo(struct sysinfo
*info
)
2026 unsigned long mem_total
, sav_total
;
2027 unsigned int mem_unit
, bitcount
;
2030 memset(info
, 0, sizeof(struct sysinfo
));
2032 get_monotonic_boottime(&tp
);
2033 info
->uptime
= tp
.tv_sec
+ (tp
.tv_nsec
? 1 : 0);
2035 get_avenrun(info
->loads
, 0, SI_LOAD_SHIFT
- FSHIFT
);
2037 info
->procs
= nr_threads
;
2043 * If the sum of all the available memory (i.e. ram + swap)
2044 * is less than can be stored in a 32 bit unsigned long then
2045 * we can be binary compatible with 2.2.x kernels. If not,
2046 * well, in that case 2.2.x was broken anyways...
2048 * -Erik Andersen <andersee@debian.org>
2051 mem_total
= info
->totalram
+ info
->totalswap
;
2052 if (mem_total
< info
->totalram
|| mem_total
< info
->totalswap
)
2055 mem_unit
= info
->mem_unit
;
2056 while (mem_unit
> 1) {
2059 sav_total
= mem_total
;
2061 if (mem_total
< sav_total
)
2066 * If mem_total did not overflow, multiply all memory values by
2067 * info->mem_unit and set it to 1. This leaves things compatible
2068 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2073 info
->totalram
<<= bitcount
;
2074 info
->freeram
<<= bitcount
;
2075 info
->sharedram
<<= bitcount
;
2076 info
->bufferram
<<= bitcount
;
2077 info
->totalswap
<<= bitcount
;
2078 info
->freeswap
<<= bitcount
;
2079 info
->totalhigh
<<= bitcount
;
2080 info
->freehigh
<<= bitcount
;
2086 SYSCALL_DEFINE1(sysinfo
, struct sysinfo __user
*, info
)
2092 if (copy_to_user(info
, &val
, sizeof(struct sysinfo
)))
2098 #ifdef CONFIG_COMPAT
2099 struct compat_sysinfo
{
2113 char _f
[20-2*sizeof(u32
)-sizeof(int)];
2116 COMPAT_SYSCALL_DEFINE1(sysinfo
, struct compat_sysinfo __user
*, info
)
2122 /* Check to see if any memory value is too large for 32-bit and scale
2125 if ((s
.totalram
>> 32) || (s
.totalswap
>> 32)) {
2128 while (s
.mem_unit
< PAGE_SIZE
) {
2133 s
.totalram
>>= bitcount
;
2134 s
.freeram
>>= bitcount
;
2135 s
.sharedram
>>= bitcount
;
2136 s
.bufferram
>>= bitcount
;
2137 s
.totalswap
>>= bitcount
;
2138 s
.freeswap
>>= bitcount
;
2139 s
.totalhigh
>>= bitcount
;
2140 s
.freehigh
>>= bitcount
;
2143 if (!access_ok(VERIFY_WRITE
, info
, sizeof(struct compat_sysinfo
)) ||
2144 __put_user(s
.uptime
, &info
->uptime
) ||
2145 __put_user(s
.loads
[0], &info
->loads
[0]) ||
2146 __put_user(s
.loads
[1], &info
->loads
[1]) ||
2147 __put_user(s
.loads
[2], &info
->loads
[2]) ||
2148 __put_user(s
.totalram
, &info
->totalram
) ||
2149 __put_user(s
.freeram
, &info
->freeram
) ||
2150 __put_user(s
.sharedram
, &info
->sharedram
) ||
2151 __put_user(s
.bufferram
, &info
->bufferram
) ||
2152 __put_user(s
.totalswap
, &info
->totalswap
) ||
2153 __put_user(s
.freeswap
, &info
->freeswap
) ||
2154 __put_user(s
.procs
, &info
->procs
) ||
2155 __put_user(s
.totalhigh
, &info
->totalhigh
) ||
2156 __put_user(s
.freehigh
, &info
->freehigh
) ||
2157 __put_user(s
.mem_unit
, &info
->mem_unit
))
2162 #endif /* CONFIG_COMPAT */