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/sched/autogroup.h>
53 #include <linux/sched/loadavg.h>
54 #include <linux/sched/mm.h>
55 #include <linux/sched/coredump.h>
56 #include <linux/rcupdate.h>
57 #include <linux/uidgid.h>
58 #include <linux/cred.h>
60 #include <linux/kmsg_dump.h>
61 /* Move somewhere else to avoid recompiling? */
62 #include <generated/utsrelease.h>
64 #include <linux/uaccess.h>
66 #include <asm/unistd.h>
68 #ifndef SET_UNALIGN_CTL
69 # define SET_UNALIGN_CTL(a, b) (-EINVAL)
71 #ifndef GET_UNALIGN_CTL
72 # define GET_UNALIGN_CTL(a, b) (-EINVAL)
75 # define SET_FPEMU_CTL(a, b) (-EINVAL)
78 # define GET_FPEMU_CTL(a, b) (-EINVAL)
81 # define SET_FPEXC_CTL(a, b) (-EINVAL)
84 # define GET_FPEXC_CTL(a, b) (-EINVAL)
87 # define GET_ENDIAN(a, b) (-EINVAL)
90 # define SET_ENDIAN(a, b) (-EINVAL)
93 # define GET_TSC_CTL(a) (-EINVAL)
96 # define SET_TSC_CTL(a) (-EINVAL)
98 #ifndef MPX_ENABLE_MANAGEMENT
99 # define MPX_ENABLE_MANAGEMENT() (-EINVAL)
101 #ifndef MPX_DISABLE_MANAGEMENT
102 # define MPX_DISABLE_MANAGEMENT() (-EINVAL)
105 # define GET_FP_MODE(a) (-EINVAL)
108 # define SET_FP_MODE(a,b) (-EINVAL)
112 * this is where the system-wide overflow UID and GID are defined, for
113 * architectures that now have 32-bit UID/GID but didn't in the past
116 int overflowuid
= DEFAULT_OVERFLOWUID
;
117 int overflowgid
= DEFAULT_OVERFLOWGID
;
119 EXPORT_SYMBOL(overflowuid
);
120 EXPORT_SYMBOL(overflowgid
);
123 * the same as above, but for filesystems which can only store a 16-bit
124 * UID and GID. as such, this is needed on all architectures
127 int fs_overflowuid
= DEFAULT_FS_OVERFLOWUID
;
128 int fs_overflowgid
= DEFAULT_FS_OVERFLOWUID
;
130 EXPORT_SYMBOL(fs_overflowuid
);
131 EXPORT_SYMBOL(fs_overflowgid
);
134 * Returns true if current's euid is same as p's uid or euid,
135 * or has CAP_SYS_NICE to p's user_ns.
137 * Called with rcu_read_lock, creds are safe
139 static bool set_one_prio_perm(struct task_struct
*p
)
141 const struct cred
*cred
= current_cred(), *pcred
= __task_cred(p
);
143 if (uid_eq(pcred
->uid
, cred
->euid
) ||
144 uid_eq(pcred
->euid
, cred
->euid
))
146 if (ns_capable(pcred
->user_ns
, CAP_SYS_NICE
))
152 * set the priority of a task
153 * - the caller must hold the RCU read lock
155 static int set_one_prio(struct task_struct
*p
, int niceval
, int error
)
159 if (!set_one_prio_perm(p
)) {
163 if (niceval
< task_nice(p
) && !can_nice(p
, niceval
)) {
167 no_nice
= security_task_setnice(p
, niceval
);
174 set_user_nice(p
, niceval
);
179 SYSCALL_DEFINE3(setpriority
, int, which
, int, who
, int, niceval
)
181 struct task_struct
*g
, *p
;
182 struct user_struct
*user
;
183 const struct cred
*cred
= current_cred();
188 if (which
> PRIO_USER
|| which
< PRIO_PROCESS
)
191 /* normalize: avoid signed division (rounding problems) */
193 if (niceval
< MIN_NICE
)
195 if (niceval
> MAX_NICE
)
199 read_lock(&tasklist_lock
);
203 p
= find_task_by_vpid(who
);
207 error
= set_one_prio(p
, niceval
, error
);
211 pgrp
= find_vpid(who
);
213 pgrp
= task_pgrp(current
);
214 do_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
) {
215 error
= set_one_prio(p
, niceval
, error
);
216 } while_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
);
219 uid
= make_kuid(cred
->user_ns
, who
);
223 else if (!uid_eq(uid
, cred
->uid
)) {
224 user
= find_user(uid
);
226 goto out_unlock
; /* No processes for this user */
228 do_each_thread(g
, p
) {
229 if (uid_eq(task_uid(p
), uid
) && task_pid_vnr(p
))
230 error
= set_one_prio(p
, niceval
, error
);
231 } while_each_thread(g
, p
);
232 if (!uid_eq(uid
, cred
->uid
))
233 free_uid(user
); /* For find_user() */
237 read_unlock(&tasklist_lock
);
244 * Ugh. To avoid negative return values, "getpriority()" will
245 * not return the normal nice-value, but a negated value that
246 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
247 * to stay compatible.
249 SYSCALL_DEFINE2(getpriority
, int, which
, int, who
)
251 struct task_struct
*g
, *p
;
252 struct user_struct
*user
;
253 const struct cred
*cred
= current_cred();
254 long niceval
, retval
= -ESRCH
;
258 if (which
> PRIO_USER
|| which
< PRIO_PROCESS
)
262 read_lock(&tasklist_lock
);
266 p
= find_task_by_vpid(who
);
270 niceval
= nice_to_rlimit(task_nice(p
));
271 if (niceval
> retval
)
277 pgrp
= find_vpid(who
);
279 pgrp
= task_pgrp(current
);
280 do_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
) {
281 niceval
= nice_to_rlimit(task_nice(p
));
282 if (niceval
> retval
)
284 } while_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
);
287 uid
= make_kuid(cred
->user_ns
, who
);
291 else if (!uid_eq(uid
, cred
->uid
)) {
292 user
= find_user(uid
);
294 goto out_unlock
; /* No processes for this user */
296 do_each_thread(g
, p
) {
297 if (uid_eq(task_uid(p
), uid
) && task_pid_vnr(p
)) {
298 niceval
= nice_to_rlimit(task_nice(p
));
299 if (niceval
> retval
)
302 } while_each_thread(g
, p
);
303 if (!uid_eq(uid
, cred
->uid
))
304 free_uid(user
); /* for find_user() */
308 read_unlock(&tasklist_lock
);
315 * Unprivileged users may change the real gid to the effective gid
316 * or vice versa. (BSD-style)
318 * If you set the real gid at all, or set the effective gid to a value not
319 * equal to the real gid, then the saved gid is set to the new effective gid.
321 * This makes it possible for a setgid program to completely drop its
322 * privileges, which is often a useful assertion to make when you are doing
323 * a security audit over a program.
325 * The general idea is that a program which uses just setregid() will be
326 * 100% compatible with BSD. A program which uses just setgid() will be
327 * 100% compatible with POSIX with saved IDs.
329 * SMP: There are not races, the GIDs are checked only by filesystem
330 * operations (as far as semantic preservation is concerned).
332 #ifdef CONFIG_MULTIUSER
333 SYSCALL_DEFINE2(setregid
, gid_t
, rgid
, gid_t
, egid
)
335 struct user_namespace
*ns
= current_user_ns();
336 const struct cred
*old
;
341 krgid
= make_kgid(ns
, rgid
);
342 kegid
= make_kgid(ns
, egid
);
344 if ((rgid
!= (gid_t
) -1) && !gid_valid(krgid
))
346 if ((egid
!= (gid_t
) -1) && !gid_valid(kegid
))
349 new = prepare_creds();
352 old
= current_cred();
355 if (rgid
!= (gid_t
) -1) {
356 if (gid_eq(old
->gid
, krgid
) ||
357 gid_eq(old
->egid
, krgid
) ||
358 ns_capable(old
->user_ns
, CAP_SETGID
))
363 if (egid
!= (gid_t
) -1) {
364 if (gid_eq(old
->gid
, kegid
) ||
365 gid_eq(old
->egid
, kegid
) ||
366 gid_eq(old
->sgid
, kegid
) ||
367 ns_capable(old
->user_ns
, CAP_SETGID
))
373 if (rgid
!= (gid_t
) -1 ||
374 (egid
!= (gid_t
) -1 && !gid_eq(kegid
, old
->gid
)))
375 new->sgid
= new->egid
;
376 new->fsgid
= new->egid
;
378 return commit_creds(new);
386 * setgid() is implemented like SysV w/ SAVED_IDS
388 * SMP: Same implicit races as above.
390 SYSCALL_DEFINE1(setgid
, gid_t
, gid
)
392 struct user_namespace
*ns
= current_user_ns();
393 const struct cred
*old
;
398 kgid
= make_kgid(ns
, gid
);
399 if (!gid_valid(kgid
))
402 new = prepare_creds();
405 old
= current_cred();
408 if (ns_capable(old
->user_ns
, CAP_SETGID
))
409 new->gid
= new->egid
= new->sgid
= new->fsgid
= kgid
;
410 else if (gid_eq(kgid
, old
->gid
) || gid_eq(kgid
, old
->sgid
))
411 new->egid
= new->fsgid
= kgid
;
415 return commit_creds(new);
423 * change the user struct in a credentials set to match the new UID
425 static int set_user(struct cred
*new)
427 struct user_struct
*new_user
;
429 new_user
= alloc_uid(new->uid
);
434 * We don't fail in case of NPROC limit excess here because too many
435 * poorly written programs don't check set*uid() return code, assuming
436 * it never fails if called by root. We may still enforce NPROC limit
437 * for programs doing set*uid()+execve() by harmlessly deferring the
438 * failure to the execve() stage.
440 if (atomic_read(&new_user
->processes
) >= rlimit(RLIMIT_NPROC
) &&
441 new_user
!= INIT_USER
)
442 current
->flags
|= PF_NPROC_EXCEEDED
;
444 current
->flags
&= ~PF_NPROC_EXCEEDED
;
447 new->user
= new_user
;
452 * Unprivileged users may change the real uid to the effective uid
453 * or vice versa. (BSD-style)
455 * If you set the real uid at all, or set the effective uid to a value not
456 * equal to the real uid, then the saved uid is set to the new effective uid.
458 * This makes it possible for a setuid program to completely drop its
459 * privileges, which is often a useful assertion to make when you are doing
460 * a security audit over a program.
462 * The general idea is that a program which uses just setreuid() will be
463 * 100% compatible with BSD. A program which uses just setuid() will be
464 * 100% compatible with POSIX with saved IDs.
466 SYSCALL_DEFINE2(setreuid
, uid_t
, ruid
, uid_t
, euid
)
468 struct user_namespace
*ns
= current_user_ns();
469 const struct cred
*old
;
474 kruid
= make_kuid(ns
, ruid
);
475 keuid
= make_kuid(ns
, euid
);
477 if ((ruid
!= (uid_t
) -1) && !uid_valid(kruid
))
479 if ((euid
!= (uid_t
) -1) && !uid_valid(keuid
))
482 new = prepare_creds();
485 old
= current_cred();
488 if (ruid
!= (uid_t
) -1) {
490 if (!uid_eq(old
->uid
, kruid
) &&
491 !uid_eq(old
->euid
, kruid
) &&
492 !ns_capable(old
->user_ns
, CAP_SETUID
))
496 if (euid
!= (uid_t
) -1) {
498 if (!uid_eq(old
->uid
, keuid
) &&
499 !uid_eq(old
->euid
, keuid
) &&
500 !uid_eq(old
->suid
, keuid
) &&
501 !ns_capable(old
->user_ns
, CAP_SETUID
))
505 if (!uid_eq(new->uid
, old
->uid
)) {
506 retval
= set_user(new);
510 if (ruid
!= (uid_t
) -1 ||
511 (euid
!= (uid_t
) -1 && !uid_eq(keuid
, old
->uid
)))
512 new->suid
= new->euid
;
513 new->fsuid
= new->euid
;
515 retval
= security_task_fix_setuid(new, old
, LSM_SETID_RE
);
519 return commit_creds(new);
527 * setuid() is implemented like SysV with SAVED_IDS
529 * Note that SAVED_ID's is deficient in that a setuid root program
530 * like sendmail, for example, cannot set its uid to be a normal
531 * user and then switch back, because if you're root, setuid() sets
532 * the saved uid too. If you don't like this, blame the bright people
533 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
534 * will allow a root program to temporarily drop privileges and be able to
535 * regain them by swapping the real and effective uid.
537 SYSCALL_DEFINE1(setuid
, uid_t
, uid
)
539 struct user_namespace
*ns
= current_user_ns();
540 const struct cred
*old
;
545 kuid
= make_kuid(ns
, uid
);
546 if (!uid_valid(kuid
))
549 new = prepare_creds();
552 old
= current_cred();
555 if (ns_capable(old
->user_ns
, CAP_SETUID
)) {
556 new->suid
= new->uid
= kuid
;
557 if (!uid_eq(kuid
, old
->uid
)) {
558 retval
= set_user(new);
562 } else if (!uid_eq(kuid
, old
->uid
) && !uid_eq(kuid
, new->suid
)) {
566 new->fsuid
= new->euid
= kuid
;
568 retval
= security_task_fix_setuid(new, old
, LSM_SETID_ID
);
572 return commit_creds(new);
581 * This function implements a generic ability to update ruid, euid,
582 * and suid. This allows you to implement the 4.4 compatible seteuid().
584 SYSCALL_DEFINE3(setresuid
, uid_t
, ruid
, uid_t
, euid
, uid_t
, suid
)
586 struct user_namespace
*ns
= current_user_ns();
587 const struct cred
*old
;
590 kuid_t kruid
, keuid
, ksuid
;
592 kruid
= make_kuid(ns
, ruid
);
593 keuid
= make_kuid(ns
, euid
);
594 ksuid
= make_kuid(ns
, suid
);
596 if ((ruid
!= (uid_t
) -1) && !uid_valid(kruid
))
599 if ((euid
!= (uid_t
) -1) && !uid_valid(keuid
))
602 if ((suid
!= (uid_t
) -1) && !uid_valid(ksuid
))
605 new = prepare_creds();
609 old
= current_cred();
612 if (!ns_capable(old
->user_ns
, CAP_SETUID
)) {
613 if (ruid
!= (uid_t
) -1 && !uid_eq(kruid
, old
->uid
) &&
614 !uid_eq(kruid
, old
->euid
) && !uid_eq(kruid
, old
->suid
))
616 if (euid
!= (uid_t
) -1 && !uid_eq(keuid
, old
->uid
) &&
617 !uid_eq(keuid
, old
->euid
) && !uid_eq(keuid
, old
->suid
))
619 if (suid
!= (uid_t
) -1 && !uid_eq(ksuid
, old
->uid
) &&
620 !uid_eq(ksuid
, old
->euid
) && !uid_eq(ksuid
, old
->suid
))
624 if (ruid
!= (uid_t
) -1) {
626 if (!uid_eq(kruid
, old
->uid
)) {
627 retval
= set_user(new);
632 if (euid
!= (uid_t
) -1)
634 if (suid
!= (uid_t
) -1)
636 new->fsuid
= new->euid
;
638 retval
= security_task_fix_setuid(new, old
, LSM_SETID_RES
);
642 return commit_creds(new);
649 SYSCALL_DEFINE3(getresuid
, uid_t __user
*, ruidp
, uid_t __user
*, euidp
, uid_t __user
*, suidp
)
651 const struct cred
*cred
= current_cred();
653 uid_t ruid
, euid
, suid
;
655 ruid
= from_kuid_munged(cred
->user_ns
, cred
->uid
);
656 euid
= from_kuid_munged(cred
->user_ns
, cred
->euid
);
657 suid
= from_kuid_munged(cred
->user_ns
, cred
->suid
);
659 retval
= put_user(ruid
, ruidp
);
661 retval
= put_user(euid
, euidp
);
663 return put_user(suid
, suidp
);
669 * Same as above, but for rgid, egid, sgid.
671 SYSCALL_DEFINE3(setresgid
, gid_t
, rgid
, gid_t
, egid
, gid_t
, sgid
)
673 struct user_namespace
*ns
= current_user_ns();
674 const struct cred
*old
;
677 kgid_t krgid
, kegid
, ksgid
;
679 krgid
= make_kgid(ns
, rgid
);
680 kegid
= make_kgid(ns
, egid
);
681 ksgid
= make_kgid(ns
, sgid
);
683 if ((rgid
!= (gid_t
) -1) && !gid_valid(krgid
))
685 if ((egid
!= (gid_t
) -1) && !gid_valid(kegid
))
687 if ((sgid
!= (gid_t
) -1) && !gid_valid(ksgid
))
690 new = prepare_creds();
693 old
= current_cred();
696 if (!ns_capable(old
->user_ns
, CAP_SETGID
)) {
697 if (rgid
!= (gid_t
) -1 && !gid_eq(krgid
, old
->gid
) &&
698 !gid_eq(krgid
, old
->egid
) && !gid_eq(krgid
, old
->sgid
))
700 if (egid
!= (gid_t
) -1 && !gid_eq(kegid
, old
->gid
) &&
701 !gid_eq(kegid
, old
->egid
) && !gid_eq(kegid
, old
->sgid
))
703 if (sgid
!= (gid_t
) -1 && !gid_eq(ksgid
, old
->gid
) &&
704 !gid_eq(ksgid
, old
->egid
) && !gid_eq(ksgid
, old
->sgid
))
708 if (rgid
!= (gid_t
) -1)
710 if (egid
!= (gid_t
) -1)
712 if (sgid
!= (gid_t
) -1)
714 new->fsgid
= new->egid
;
716 return commit_creds(new);
723 SYSCALL_DEFINE3(getresgid
, gid_t __user
*, rgidp
, gid_t __user
*, egidp
, gid_t __user
*, sgidp
)
725 const struct cred
*cred
= current_cred();
727 gid_t rgid
, egid
, sgid
;
729 rgid
= from_kgid_munged(cred
->user_ns
, cred
->gid
);
730 egid
= from_kgid_munged(cred
->user_ns
, cred
->egid
);
731 sgid
= from_kgid_munged(cred
->user_ns
, cred
->sgid
);
733 retval
= put_user(rgid
, rgidp
);
735 retval
= put_user(egid
, egidp
);
737 retval
= put_user(sgid
, sgidp
);
745 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
746 * is used for "access()" and for the NFS daemon (letting nfsd stay at
747 * whatever uid it wants to). It normally shadows "euid", except when
748 * explicitly set by setfsuid() or for access..
750 SYSCALL_DEFINE1(setfsuid
, uid_t
, uid
)
752 const struct cred
*old
;
757 old
= current_cred();
758 old_fsuid
= from_kuid_munged(old
->user_ns
, old
->fsuid
);
760 kuid
= make_kuid(old
->user_ns
, uid
);
761 if (!uid_valid(kuid
))
764 new = prepare_creds();
768 if (uid_eq(kuid
, old
->uid
) || uid_eq(kuid
, old
->euid
) ||
769 uid_eq(kuid
, old
->suid
) || uid_eq(kuid
, old
->fsuid
) ||
770 ns_capable(old
->user_ns
, CAP_SETUID
)) {
771 if (!uid_eq(kuid
, old
->fsuid
)) {
773 if (security_task_fix_setuid(new, old
, LSM_SETID_FS
) == 0)
787 * Samma på svenska..
789 SYSCALL_DEFINE1(setfsgid
, gid_t
, gid
)
791 const struct cred
*old
;
796 old
= current_cred();
797 old_fsgid
= from_kgid_munged(old
->user_ns
, old
->fsgid
);
799 kgid
= make_kgid(old
->user_ns
, gid
);
800 if (!gid_valid(kgid
))
803 new = prepare_creds();
807 if (gid_eq(kgid
, old
->gid
) || gid_eq(kgid
, old
->egid
) ||
808 gid_eq(kgid
, old
->sgid
) || gid_eq(kgid
, old
->fsgid
) ||
809 ns_capable(old
->user_ns
, CAP_SETGID
)) {
810 if (!gid_eq(kgid
, old
->fsgid
)) {
823 #endif /* CONFIG_MULTIUSER */
826 * sys_getpid - return the thread group id of the current process
828 * Note, despite the name, this returns the tgid not the pid. The tgid and
829 * the pid are identical unless CLONE_THREAD was specified on clone() in
830 * which case the tgid is the same in all threads of the same group.
832 * This is SMP safe as current->tgid does not change.
834 SYSCALL_DEFINE0(getpid
)
836 return task_tgid_vnr(current
);
839 /* Thread ID - the internal kernel "pid" */
840 SYSCALL_DEFINE0(gettid
)
842 return task_pid_vnr(current
);
846 * Accessing ->real_parent is not SMP-safe, it could
847 * change from under us. However, we can use a stale
848 * value of ->real_parent under rcu_read_lock(), see
849 * release_task()->call_rcu(delayed_put_task_struct).
851 SYSCALL_DEFINE0(getppid
)
856 pid
= task_tgid_vnr(rcu_dereference(current
->real_parent
));
862 SYSCALL_DEFINE0(getuid
)
864 /* Only we change this so SMP safe */
865 return from_kuid_munged(current_user_ns(), current_uid());
868 SYSCALL_DEFINE0(geteuid
)
870 /* Only we change this so SMP safe */
871 return from_kuid_munged(current_user_ns(), current_euid());
874 SYSCALL_DEFINE0(getgid
)
876 /* Only we change this so SMP safe */
877 return from_kgid_munged(current_user_ns(), current_gid());
880 SYSCALL_DEFINE0(getegid
)
882 /* Only we change this so SMP safe */
883 return from_kgid_munged(current_user_ns(), current_egid());
886 void do_sys_times(struct tms
*tms
)
888 u64 tgutime
, tgstime
, cutime
, cstime
;
890 thread_group_cputime_adjusted(current
, &tgutime
, &tgstime
);
891 cutime
= current
->signal
->cutime
;
892 cstime
= current
->signal
->cstime
;
893 tms
->tms_utime
= nsec_to_clock_t(tgutime
);
894 tms
->tms_stime
= nsec_to_clock_t(tgstime
);
895 tms
->tms_cutime
= nsec_to_clock_t(cutime
);
896 tms
->tms_cstime
= nsec_to_clock_t(cstime
);
899 SYSCALL_DEFINE1(times
, struct tms __user
*, tbuf
)
905 if (copy_to_user(tbuf
, &tmp
, sizeof(struct tms
)))
908 force_successful_syscall_return();
909 return (long) jiffies_64_to_clock_t(get_jiffies_64());
913 * This needs some heavy checking ...
914 * I just haven't the stomach for it. I also don't fully
915 * understand sessions/pgrp etc. Let somebody who does explain it.
917 * OK, I think I have the protection semantics right.... this is really
918 * only important on a multi-user system anyway, to make sure one user
919 * can't send a signal to a process owned by another. -TYT, 12/12/91
921 * !PF_FORKNOEXEC check to conform completely to POSIX.
923 SYSCALL_DEFINE2(setpgid
, pid_t
, pid
, pid_t
, pgid
)
925 struct task_struct
*p
;
926 struct task_struct
*group_leader
= current
->group_leader
;
931 pid
= task_pid_vnr(group_leader
);
938 /* From this point forward we keep holding onto the tasklist lock
939 * so that our parent does not change from under us. -DaveM
941 write_lock_irq(&tasklist_lock
);
944 p
= find_task_by_vpid(pid
);
949 if (!thread_group_leader(p
))
952 if (same_thread_group(p
->real_parent
, group_leader
)) {
954 if (task_session(p
) != task_session(group_leader
))
957 if (!(p
->flags
& PF_FORKNOEXEC
))
961 if (p
!= group_leader
)
966 if (p
->signal
->leader
)
971 struct task_struct
*g
;
973 pgrp
= find_vpid(pgid
);
974 g
= pid_task(pgrp
, PIDTYPE_PGID
);
975 if (!g
|| task_session(g
) != task_session(group_leader
))
979 err
= security_task_setpgid(p
, pgid
);
983 if (task_pgrp(p
) != pgrp
)
984 change_pid(p
, PIDTYPE_PGID
, pgrp
);
988 /* All paths lead to here, thus we are safe. -DaveM */
989 write_unlock_irq(&tasklist_lock
);
994 SYSCALL_DEFINE1(getpgid
, pid_t
, pid
)
996 struct task_struct
*p
;
1002 grp
= task_pgrp(current
);
1005 p
= find_task_by_vpid(pid
);
1012 retval
= security_task_getpgid(p
);
1016 retval
= pid_vnr(grp
);
1022 #ifdef __ARCH_WANT_SYS_GETPGRP
1024 SYSCALL_DEFINE0(getpgrp
)
1026 return sys_getpgid(0);
1031 SYSCALL_DEFINE1(getsid
, pid_t
, pid
)
1033 struct task_struct
*p
;
1039 sid
= task_session(current
);
1042 p
= find_task_by_vpid(pid
);
1045 sid
= task_session(p
);
1049 retval
= security_task_getsid(p
);
1053 retval
= pid_vnr(sid
);
1059 static void set_special_pids(struct pid
*pid
)
1061 struct task_struct
*curr
= current
->group_leader
;
1063 if (task_session(curr
) != pid
)
1064 change_pid(curr
, PIDTYPE_SID
, pid
);
1066 if (task_pgrp(curr
) != pid
)
1067 change_pid(curr
, PIDTYPE_PGID
, pid
);
1070 SYSCALL_DEFINE0(setsid
)
1072 struct task_struct
*group_leader
= current
->group_leader
;
1073 struct pid
*sid
= task_pid(group_leader
);
1074 pid_t session
= pid_vnr(sid
);
1077 write_lock_irq(&tasklist_lock
);
1078 /* Fail if I am already a session leader */
1079 if (group_leader
->signal
->leader
)
1082 /* Fail if a process group id already exists that equals the
1083 * proposed session id.
1085 if (pid_task(sid
, PIDTYPE_PGID
))
1088 group_leader
->signal
->leader
= 1;
1089 set_special_pids(sid
);
1091 proc_clear_tty(group_leader
);
1095 write_unlock_irq(&tasklist_lock
);
1097 proc_sid_connector(group_leader
);
1098 sched_autogroup_create_attach(group_leader
);
1103 DECLARE_RWSEM(uts_sem
);
1105 #ifdef COMPAT_UTS_MACHINE
1106 #define override_architecture(name) \
1107 (personality(current->personality) == PER_LINUX32 && \
1108 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1109 sizeof(COMPAT_UTS_MACHINE)))
1111 #define override_architecture(name) 0
1115 * Work around broken programs that cannot handle "Linux 3.0".
1116 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1117 * And we map 4.x to 2.6.60+x, so 4.0 would be 2.6.60.
1119 static int override_release(char __user
*release
, size_t len
)
1123 if (current
->personality
& UNAME26
) {
1124 const char *rest
= UTS_RELEASE
;
1125 char buf
[65] = { 0 };
1131 if (*rest
== '.' && ++ndots
>= 3)
1133 if (!isdigit(*rest
) && *rest
!= '.')
1137 v
= ((LINUX_VERSION_CODE
>> 8) & 0xff) + 60;
1138 copy
= clamp_t(size_t, len
, 1, sizeof(buf
));
1139 copy
= scnprintf(buf
, copy
, "2.6.%u%s", v
, rest
);
1140 ret
= copy_to_user(release
, buf
, copy
+ 1);
1145 SYSCALL_DEFINE1(newuname
, struct new_utsname __user
*, name
)
1149 down_read(&uts_sem
);
1150 if (copy_to_user(name
, utsname(), sizeof *name
))
1154 if (!errno
&& override_release(name
->release
, sizeof(name
->release
)))
1156 if (!errno
&& override_architecture(name
))
1161 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1165 SYSCALL_DEFINE1(uname
, struct old_utsname __user
*, name
)
1172 down_read(&uts_sem
);
1173 if (copy_to_user(name
, utsname(), sizeof(*name
)))
1177 if (!error
&& override_release(name
->release
, sizeof(name
->release
)))
1179 if (!error
&& override_architecture(name
))
1184 SYSCALL_DEFINE1(olduname
, struct oldold_utsname __user
*, name
)
1190 if (!access_ok(VERIFY_WRITE
, name
, sizeof(struct oldold_utsname
)))
1193 down_read(&uts_sem
);
1194 error
= __copy_to_user(&name
->sysname
, &utsname()->sysname
,
1196 error
|= __put_user(0, name
->sysname
+ __OLD_UTS_LEN
);
1197 error
|= __copy_to_user(&name
->nodename
, &utsname()->nodename
,
1199 error
|= __put_user(0, name
->nodename
+ __OLD_UTS_LEN
);
1200 error
|= __copy_to_user(&name
->release
, &utsname()->release
,
1202 error
|= __put_user(0, name
->release
+ __OLD_UTS_LEN
);
1203 error
|= __copy_to_user(&name
->version
, &utsname()->version
,
1205 error
|= __put_user(0, name
->version
+ __OLD_UTS_LEN
);
1206 error
|= __copy_to_user(&name
->machine
, &utsname()->machine
,
1208 error
|= __put_user(0, name
->machine
+ __OLD_UTS_LEN
);
1211 if (!error
&& override_architecture(name
))
1213 if (!error
&& override_release(name
->release
, sizeof(name
->release
)))
1215 return error
? -EFAULT
: 0;
1219 SYSCALL_DEFINE2(sethostname
, char __user
*, name
, int, len
)
1222 char tmp
[__NEW_UTS_LEN
];
1224 if (!ns_capable(current
->nsproxy
->uts_ns
->user_ns
, CAP_SYS_ADMIN
))
1227 if (len
< 0 || len
> __NEW_UTS_LEN
)
1229 down_write(&uts_sem
);
1231 if (!copy_from_user(tmp
, name
, len
)) {
1232 struct new_utsname
*u
= utsname();
1234 memcpy(u
->nodename
, tmp
, len
);
1235 memset(u
->nodename
+ len
, 0, sizeof(u
->nodename
) - len
);
1237 uts_proc_notify(UTS_PROC_HOSTNAME
);
1243 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1245 SYSCALL_DEFINE2(gethostname
, char __user
*, name
, int, len
)
1248 struct new_utsname
*u
;
1252 down_read(&uts_sem
);
1254 i
= 1 + strlen(u
->nodename
);
1258 if (copy_to_user(name
, u
->nodename
, i
))
1267 * Only setdomainname; getdomainname can be implemented by calling
1270 SYSCALL_DEFINE2(setdomainname
, char __user
*, name
, int, len
)
1273 char tmp
[__NEW_UTS_LEN
];
1275 if (!ns_capable(current
->nsproxy
->uts_ns
->user_ns
, CAP_SYS_ADMIN
))
1277 if (len
< 0 || len
> __NEW_UTS_LEN
)
1280 down_write(&uts_sem
);
1282 if (!copy_from_user(tmp
, name
, len
)) {
1283 struct new_utsname
*u
= utsname();
1285 memcpy(u
->domainname
, tmp
, len
);
1286 memset(u
->domainname
+ len
, 0, sizeof(u
->domainname
) - len
);
1288 uts_proc_notify(UTS_PROC_DOMAINNAME
);
1294 SYSCALL_DEFINE2(getrlimit
, unsigned int, resource
, struct rlimit __user
*, rlim
)
1296 struct rlimit value
;
1299 ret
= do_prlimit(current
, resource
, NULL
, &value
);
1301 ret
= copy_to_user(rlim
, &value
, sizeof(*rlim
)) ? -EFAULT
: 0;
1306 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1309 * Back compatibility for getrlimit. Needed for some apps.
1311 SYSCALL_DEFINE2(old_getrlimit
, unsigned int, resource
,
1312 struct rlimit __user
*, rlim
)
1315 if (resource
>= RLIM_NLIMITS
)
1318 task_lock(current
->group_leader
);
1319 x
= current
->signal
->rlim
[resource
];
1320 task_unlock(current
->group_leader
);
1321 if (x
.rlim_cur
> 0x7FFFFFFF)
1322 x
.rlim_cur
= 0x7FFFFFFF;
1323 if (x
.rlim_max
> 0x7FFFFFFF)
1324 x
.rlim_max
= 0x7FFFFFFF;
1325 return copy_to_user(rlim
, &x
, sizeof(x
)) ? -EFAULT
: 0;
1330 static inline bool rlim64_is_infinity(__u64 rlim64
)
1332 #if BITS_PER_LONG < 64
1333 return rlim64
>= ULONG_MAX
;
1335 return rlim64
== RLIM64_INFINITY
;
1339 static void rlim_to_rlim64(const struct rlimit
*rlim
, struct rlimit64
*rlim64
)
1341 if (rlim
->rlim_cur
== RLIM_INFINITY
)
1342 rlim64
->rlim_cur
= RLIM64_INFINITY
;
1344 rlim64
->rlim_cur
= rlim
->rlim_cur
;
1345 if (rlim
->rlim_max
== RLIM_INFINITY
)
1346 rlim64
->rlim_max
= RLIM64_INFINITY
;
1348 rlim64
->rlim_max
= rlim
->rlim_max
;
1351 static void rlim64_to_rlim(const struct rlimit64
*rlim64
, struct rlimit
*rlim
)
1353 if (rlim64_is_infinity(rlim64
->rlim_cur
))
1354 rlim
->rlim_cur
= RLIM_INFINITY
;
1356 rlim
->rlim_cur
= (unsigned long)rlim64
->rlim_cur
;
1357 if (rlim64_is_infinity(rlim64
->rlim_max
))
1358 rlim
->rlim_max
= RLIM_INFINITY
;
1360 rlim
->rlim_max
= (unsigned long)rlim64
->rlim_max
;
1363 /* make sure you are allowed to change @tsk limits before calling this */
1364 int do_prlimit(struct task_struct
*tsk
, unsigned int resource
,
1365 struct rlimit
*new_rlim
, struct rlimit
*old_rlim
)
1367 struct rlimit
*rlim
;
1370 if (resource
>= RLIM_NLIMITS
)
1373 if (new_rlim
->rlim_cur
> new_rlim
->rlim_max
)
1375 if (resource
== RLIMIT_NOFILE
&&
1376 new_rlim
->rlim_max
> sysctl_nr_open
)
1380 /* protect tsk->signal and tsk->sighand from disappearing */
1381 read_lock(&tasklist_lock
);
1382 if (!tsk
->sighand
) {
1387 rlim
= tsk
->signal
->rlim
+ resource
;
1388 task_lock(tsk
->group_leader
);
1390 /* Keep the capable check against init_user_ns until
1391 cgroups can contain all limits */
1392 if (new_rlim
->rlim_max
> rlim
->rlim_max
&&
1393 !capable(CAP_SYS_RESOURCE
))
1396 retval
= security_task_setrlimit(tsk
->group_leader
,
1397 resource
, new_rlim
);
1398 if (resource
== RLIMIT_CPU
&& new_rlim
->rlim_cur
== 0) {
1400 * The caller is asking for an immediate RLIMIT_CPU
1401 * expiry. But we use the zero value to mean "it was
1402 * never set". So let's cheat and make it one second
1405 new_rlim
->rlim_cur
= 1;
1414 task_unlock(tsk
->group_leader
);
1417 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1418 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1419 * very long-standing error, and fixing it now risks breakage of
1420 * applications, so we live with it
1422 if (!retval
&& new_rlim
&& resource
== RLIMIT_CPU
&&
1423 new_rlim
->rlim_cur
!= RLIM_INFINITY
&&
1424 IS_ENABLED(CONFIG_POSIX_TIMERS
))
1425 update_rlimit_cpu(tsk
, new_rlim
->rlim_cur
);
1427 read_unlock(&tasklist_lock
);
1431 /* rcu lock must be held */
1432 static int check_prlimit_permission(struct task_struct
*task
)
1434 const struct cred
*cred
= current_cred(), *tcred
;
1436 if (current
== task
)
1439 tcred
= __task_cred(task
);
1440 if (uid_eq(cred
->uid
, tcred
->euid
) &&
1441 uid_eq(cred
->uid
, tcred
->suid
) &&
1442 uid_eq(cred
->uid
, tcred
->uid
) &&
1443 gid_eq(cred
->gid
, tcred
->egid
) &&
1444 gid_eq(cred
->gid
, tcred
->sgid
) &&
1445 gid_eq(cred
->gid
, tcred
->gid
))
1447 if (ns_capable(tcred
->user_ns
, CAP_SYS_RESOURCE
))
1453 SYSCALL_DEFINE4(prlimit64
, pid_t
, pid
, unsigned int, resource
,
1454 const struct rlimit64 __user
*, new_rlim
,
1455 struct rlimit64 __user
*, old_rlim
)
1457 struct rlimit64 old64
, new64
;
1458 struct rlimit old
, new;
1459 struct task_struct
*tsk
;
1463 if (copy_from_user(&new64
, new_rlim
, sizeof(new64
)))
1465 rlim64_to_rlim(&new64
, &new);
1469 tsk
= pid
? find_task_by_vpid(pid
) : current
;
1474 ret
= check_prlimit_permission(tsk
);
1479 get_task_struct(tsk
);
1482 ret
= do_prlimit(tsk
, resource
, new_rlim
? &new : NULL
,
1483 old_rlim
? &old
: NULL
);
1485 if (!ret
&& old_rlim
) {
1486 rlim_to_rlim64(&old
, &old64
);
1487 if (copy_to_user(old_rlim
, &old64
, sizeof(old64
)))
1491 put_task_struct(tsk
);
1495 SYSCALL_DEFINE2(setrlimit
, unsigned int, resource
, struct rlimit __user
*, rlim
)
1497 struct rlimit new_rlim
;
1499 if (copy_from_user(&new_rlim
, rlim
, sizeof(*rlim
)))
1501 return do_prlimit(current
, resource
, &new_rlim
, NULL
);
1505 * It would make sense to put struct rusage in the task_struct,
1506 * except that would make the task_struct be *really big*. After
1507 * task_struct gets moved into malloc'ed memory, it would
1508 * make sense to do this. It will make moving the rest of the information
1509 * a lot simpler! (Which we're not doing right now because we're not
1510 * measuring them yet).
1512 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1513 * races with threads incrementing their own counters. But since word
1514 * reads are atomic, we either get new values or old values and we don't
1515 * care which for the sums. We always take the siglock to protect reading
1516 * the c* fields from p->signal from races with exit.c updating those
1517 * fields when reaping, so a sample either gets all the additions of a
1518 * given child after it's reaped, or none so this sample is before reaping.
1521 * We need to take the siglock for CHILDEREN, SELF and BOTH
1522 * for the cases current multithreaded, non-current single threaded
1523 * non-current multithreaded. Thread traversal is now safe with
1525 * Strictly speaking, we donot need to take the siglock if we are current and
1526 * single threaded, as no one else can take our signal_struct away, no one
1527 * else can reap the children to update signal->c* counters, and no one else
1528 * can race with the signal-> fields. If we do not take any lock, the
1529 * signal-> fields could be read out of order while another thread was just
1530 * exiting. So we should place a read memory barrier when we avoid the lock.
1531 * On the writer side, write memory barrier is implied in __exit_signal
1532 * as __exit_signal releases the siglock spinlock after updating the signal->
1533 * fields. But we don't do this yet to keep things simple.
1537 static void accumulate_thread_rusage(struct task_struct
*t
, struct rusage
*r
)
1539 r
->ru_nvcsw
+= t
->nvcsw
;
1540 r
->ru_nivcsw
+= t
->nivcsw
;
1541 r
->ru_minflt
+= t
->min_flt
;
1542 r
->ru_majflt
+= t
->maj_flt
;
1543 r
->ru_inblock
+= task_io_get_inblock(t
);
1544 r
->ru_oublock
+= task_io_get_oublock(t
);
1547 static void k_getrusage(struct task_struct
*p
, int who
, struct rusage
*r
)
1549 struct task_struct
*t
;
1550 unsigned long flags
;
1551 u64 tgutime
, tgstime
, utime
, stime
;
1552 unsigned long maxrss
= 0;
1554 memset((char *)r
, 0, sizeof (*r
));
1557 if (who
== RUSAGE_THREAD
) {
1558 task_cputime_adjusted(current
, &utime
, &stime
);
1559 accumulate_thread_rusage(p
, r
);
1560 maxrss
= p
->signal
->maxrss
;
1564 if (!lock_task_sighand(p
, &flags
))
1569 case RUSAGE_CHILDREN
:
1570 utime
= p
->signal
->cutime
;
1571 stime
= p
->signal
->cstime
;
1572 r
->ru_nvcsw
= p
->signal
->cnvcsw
;
1573 r
->ru_nivcsw
= p
->signal
->cnivcsw
;
1574 r
->ru_minflt
= p
->signal
->cmin_flt
;
1575 r
->ru_majflt
= p
->signal
->cmaj_flt
;
1576 r
->ru_inblock
= p
->signal
->cinblock
;
1577 r
->ru_oublock
= p
->signal
->coublock
;
1578 maxrss
= p
->signal
->cmaxrss
;
1580 if (who
== RUSAGE_CHILDREN
)
1584 thread_group_cputime_adjusted(p
, &tgutime
, &tgstime
);
1587 r
->ru_nvcsw
+= p
->signal
->nvcsw
;
1588 r
->ru_nivcsw
+= p
->signal
->nivcsw
;
1589 r
->ru_minflt
+= p
->signal
->min_flt
;
1590 r
->ru_majflt
+= p
->signal
->maj_flt
;
1591 r
->ru_inblock
+= p
->signal
->inblock
;
1592 r
->ru_oublock
+= p
->signal
->oublock
;
1593 if (maxrss
< p
->signal
->maxrss
)
1594 maxrss
= p
->signal
->maxrss
;
1597 accumulate_thread_rusage(t
, r
);
1598 } while_each_thread(p
, t
);
1604 unlock_task_sighand(p
, &flags
);
1607 r
->ru_utime
= ns_to_timeval(utime
);
1608 r
->ru_stime
= ns_to_timeval(stime
);
1610 if (who
!= RUSAGE_CHILDREN
) {
1611 struct mm_struct
*mm
= get_task_mm(p
);
1614 setmax_mm_hiwater_rss(&maxrss
, mm
);
1618 r
->ru_maxrss
= maxrss
* (PAGE_SIZE
/ 1024); /* convert pages to KBs */
1621 int getrusage(struct task_struct
*p
, int who
, struct rusage __user
*ru
)
1625 k_getrusage(p
, who
, &r
);
1626 return copy_to_user(ru
, &r
, sizeof(r
)) ? -EFAULT
: 0;
1629 SYSCALL_DEFINE2(getrusage
, int, who
, struct rusage __user
*, ru
)
1631 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
&&
1632 who
!= RUSAGE_THREAD
)
1634 return getrusage(current
, who
, ru
);
1637 #ifdef CONFIG_COMPAT
1638 COMPAT_SYSCALL_DEFINE2(getrusage
, int, who
, struct compat_rusage __user
*, ru
)
1642 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
&&
1643 who
!= RUSAGE_THREAD
)
1646 k_getrusage(current
, who
, &r
);
1647 return put_compat_rusage(&r
, ru
);
1651 SYSCALL_DEFINE1(umask
, int, mask
)
1653 mask
= xchg(¤t
->fs
->umask
, mask
& S_IRWXUGO
);
1657 static int prctl_set_mm_exe_file(struct mm_struct
*mm
, unsigned int fd
)
1660 struct file
*old_exe
, *exe_file
;
1661 struct inode
*inode
;
1668 inode
= file_inode(exe
.file
);
1671 * Because the original mm->exe_file points to executable file, make
1672 * sure that this one is executable as well, to avoid breaking an
1676 if (!S_ISREG(inode
->i_mode
) || path_noexec(&exe
.file
->f_path
))
1679 err
= inode_permission(inode
, MAY_EXEC
);
1684 * Forbid mm->exe_file change if old file still mapped.
1686 exe_file
= get_mm_exe_file(mm
);
1689 struct vm_area_struct
*vma
;
1691 down_read(&mm
->mmap_sem
);
1692 for (vma
= mm
->mmap
; vma
; vma
= vma
->vm_next
) {
1695 if (path_equal(&vma
->vm_file
->f_path
,
1700 up_read(&mm
->mmap_sem
);
1705 /* set the new file, lockless */
1707 old_exe
= xchg(&mm
->exe_file
, exe
.file
);
1714 up_read(&mm
->mmap_sem
);
1720 * WARNING: we don't require any capability here so be very careful
1721 * in what is allowed for modification from userspace.
1723 static int validate_prctl_map(struct prctl_mm_map
*prctl_map
)
1725 unsigned long mmap_max_addr
= TASK_SIZE
;
1726 struct mm_struct
*mm
= current
->mm
;
1727 int error
= -EINVAL
, i
;
1729 static const unsigned char offsets
[] = {
1730 offsetof(struct prctl_mm_map
, start_code
),
1731 offsetof(struct prctl_mm_map
, end_code
),
1732 offsetof(struct prctl_mm_map
, start_data
),
1733 offsetof(struct prctl_mm_map
, end_data
),
1734 offsetof(struct prctl_mm_map
, start_brk
),
1735 offsetof(struct prctl_mm_map
, brk
),
1736 offsetof(struct prctl_mm_map
, start_stack
),
1737 offsetof(struct prctl_mm_map
, arg_start
),
1738 offsetof(struct prctl_mm_map
, arg_end
),
1739 offsetof(struct prctl_mm_map
, env_start
),
1740 offsetof(struct prctl_mm_map
, env_end
),
1744 * Make sure the members are not somewhere outside
1745 * of allowed address space.
1747 for (i
= 0; i
< ARRAY_SIZE(offsets
); i
++) {
1748 u64 val
= *(u64
*)((char *)prctl_map
+ offsets
[i
]);
1750 if ((unsigned long)val
>= mmap_max_addr
||
1751 (unsigned long)val
< mmap_min_addr
)
1756 * Make sure the pairs are ordered.
1758 #define __prctl_check_order(__m1, __op, __m2) \
1759 ((unsigned long)prctl_map->__m1 __op \
1760 (unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1761 error
= __prctl_check_order(start_code
, <, end_code
);
1762 error
|= __prctl_check_order(start_data
, <, end_data
);
1763 error
|= __prctl_check_order(start_brk
, <=, brk
);
1764 error
|= __prctl_check_order(arg_start
, <=, arg_end
);
1765 error
|= __prctl_check_order(env_start
, <=, env_end
);
1768 #undef __prctl_check_order
1773 * @brk should be after @end_data in traditional maps.
1775 if (prctl_map
->start_brk
<= prctl_map
->end_data
||
1776 prctl_map
->brk
<= prctl_map
->end_data
)
1780 * Neither we should allow to override limits if they set.
1782 if (check_data_rlimit(rlimit(RLIMIT_DATA
), prctl_map
->brk
,
1783 prctl_map
->start_brk
, prctl_map
->end_data
,
1784 prctl_map
->start_data
))
1788 * Someone is trying to cheat the auxv vector.
1790 if (prctl_map
->auxv_size
) {
1791 if (!prctl_map
->auxv
|| prctl_map
->auxv_size
> sizeof(mm
->saved_auxv
))
1796 * Finally, make sure the caller has the rights to
1797 * change /proc/pid/exe link: only local root should
1800 if (prctl_map
->exe_fd
!= (u32
)-1) {
1801 struct user_namespace
*ns
= current_user_ns();
1802 const struct cred
*cred
= current_cred();
1804 if (!uid_eq(cred
->uid
, make_kuid(ns
, 0)) ||
1805 !gid_eq(cred
->gid
, make_kgid(ns
, 0)))
1814 #ifdef CONFIG_CHECKPOINT_RESTORE
1815 static int prctl_set_mm_map(int opt
, const void __user
*addr
, unsigned long data_size
)
1817 struct prctl_mm_map prctl_map
= { .exe_fd
= (u32
)-1, };
1818 unsigned long user_auxv
[AT_VECTOR_SIZE
];
1819 struct mm_struct
*mm
= current
->mm
;
1822 BUILD_BUG_ON(sizeof(user_auxv
) != sizeof(mm
->saved_auxv
));
1823 BUILD_BUG_ON(sizeof(struct prctl_mm_map
) > 256);
1825 if (opt
== PR_SET_MM_MAP_SIZE
)
1826 return put_user((unsigned int)sizeof(prctl_map
),
1827 (unsigned int __user
*)addr
);
1829 if (data_size
!= sizeof(prctl_map
))
1832 if (copy_from_user(&prctl_map
, addr
, sizeof(prctl_map
)))
1835 error
= validate_prctl_map(&prctl_map
);
1839 if (prctl_map
.auxv_size
) {
1840 memset(user_auxv
, 0, sizeof(user_auxv
));
1841 if (copy_from_user(user_auxv
,
1842 (const void __user
*)prctl_map
.auxv
,
1843 prctl_map
.auxv_size
))
1846 /* Last entry must be AT_NULL as specification requires */
1847 user_auxv
[AT_VECTOR_SIZE
- 2] = AT_NULL
;
1848 user_auxv
[AT_VECTOR_SIZE
- 1] = AT_NULL
;
1851 if (prctl_map
.exe_fd
!= (u32
)-1) {
1852 error
= prctl_set_mm_exe_file(mm
, prctl_map
.exe_fd
);
1857 down_write(&mm
->mmap_sem
);
1860 * We don't validate if these members are pointing to
1861 * real present VMAs because application may have correspond
1862 * VMAs already unmapped and kernel uses these members for statistics
1863 * output in procfs mostly, except
1865 * - @start_brk/@brk which are used in do_brk but kernel lookups
1866 * for VMAs when updating these memvers so anything wrong written
1867 * here cause kernel to swear at userspace program but won't lead
1868 * to any problem in kernel itself
1871 mm
->start_code
= prctl_map
.start_code
;
1872 mm
->end_code
= prctl_map
.end_code
;
1873 mm
->start_data
= prctl_map
.start_data
;
1874 mm
->end_data
= prctl_map
.end_data
;
1875 mm
->start_brk
= prctl_map
.start_brk
;
1876 mm
->brk
= prctl_map
.brk
;
1877 mm
->start_stack
= prctl_map
.start_stack
;
1878 mm
->arg_start
= prctl_map
.arg_start
;
1879 mm
->arg_end
= prctl_map
.arg_end
;
1880 mm
->env_start
= prctl_map
.env_start
;
1881 mm
->env_end
= prctl_map
.env_end
;
1884 * Note this update of @saved_auxv is lockless thus
1885 * if someone reads this member in procfs while we're
1886 * updating -- it may get partly updated results. It's
1887 * known and acceptable trade off: we leave it as is to
1888 * not introduce additional locks here making the kernel
1891 if (prctl_map
.auxv_size
)
1892 memcpy(mm
->saved_auxv
, user_auxv
, sizeof(user_auxv
));
1894 up_write(&mm
->mmap_sem
);
1897 #endif /* CONFIG_CHECKPOINT_RESTORE */
1899 static int prctl_set_auxv(struct mm_struct
*mm
, unsigned long addr
,
1903 * This doesn't move the auxiliary vector itself since it's pinned to
1904 * mm_struct, but it permits filling the vector with new values. It's
1905 * up to the caller to provide sane values here, otherwise userspace
1906 * tools which use this vector might be unhappy.
1908 unsigned long user_auxv
[AT_VECTOR_SIZE
];
1910 if (len
> sizeof(user_auxv
))
1913 if (copy_from_user(user_auxv
, (const void __user
*)addr
, len
))
1916 /* Make sure the last entry is always AT_NULL */
1917 user_auxv
[AT_VECTOR_SIZE
- 2] = 0;
1918 user_auxv
[AT_VECTOR_SIZE
- 1] = 0;
1920 BUILD_BUG_ON(sizeof(user_auxv
) != sizeof(mm
->saved_auxv
));
1923 memcpy(mm
->saved_auxv
, user_auxv
, len
);
1924 task_unlock(current
);
1929 static int prctl_set_mm(int opt
, unsigned long addr
,
1930 unsigned long arg4
, unsigned long arg5
)
1932 struct mm_struct
*mm
= current
->mm
;
1933 struct prctl_mm_map prctl_map
;
1934 struct vm_area_struct
*vma
;
1937 if (arg5
|| (arg4
&& (opt
!= PR_SET_MM_AUXV
&&
1938 opt
!= PR_SET_MM_MAP
&&
1939 opt
!= PR_SET_MM_MAP_SIZE
)))
1942 #ifdef CONFIG_CHECKPOINT_RESTORE
1943 if (opt
== PR_SET_MM_MAP
|| opt
== PR_SET_MM_MAP_SIZE
)
1944 return prctl_set_mm_map(opt
, (const void __user
*)addr
, arg4
);
1947 if (!capable(CAP_SYS_RESOURCE
))
1950 if (opt
== PR_SET_MM_EXE_FILE
)
1951 return prctl_set_mm_exe_file(mm
, (unsigned int)addr
);
1953 if (opt
== PR_SET_MM_AUXV
)
1954 return prctl_set_auxv(mm
, addr
, arg4
);
1956 if (addr
>= TASK_SIZE
|| addr
< mmap_min_addr
)
1961 down_write(&mm
->mmap_sem
);
1962 vma
= find_vma(mm
, addr
);
1964 prctl_map
.start_code
= mm
->start_code
;
1965 prctl_map
.end_code
= mm
->end_code
;
1966 prctl_map
.start_data
= mm
->start_data
;
1967 prctl_map
.end_data
= mm
->end_data
;
1968 prctl_map
.start_brk
= mm
->start_brk
;
1969 prctl_map
.brk
= mm
->brk
;
1970 prctl_map
.start_stack
= mm
->start_stack
;
1971 prctl_map
.arg_start
= mm
->arg_start
;
1972 prctl_map
.arg_end
= mm
->arg_end
;
1973 prctl_map
.env_start
= mm
->env_start
;
1974 prctl_map
.env_end
= mm
->env_end
;
1975 prctl_map
.auxv
= NULL
;
1976 prctl_map
.auxv_size
= 0;
1977 prctl_map
.exe_fd
= -1;
1980 case PR_SET_MM_START_CODE
:
1981 prctl_map
.start_code
= addr
;
1983 case PR_SET_MM_END_CODE
:
1984 prctl_map
.end_code
= addr
;
1986 case PR_SET_MM_START_DATA
:
1987 prctl_map
.start_data
= addr
;
1989 case PR_SET_MM_END_DATA
:
1990 prctl_map
.end_data
= addr
;
1992 case PR_SET_MM_START_STACK
:
1993 prctl_map
.start_stack
= addr
;
1995 case PR_SET_MM_START_BRK
:
1996 prctl_map
.start_brk
= addr
;
1999 prctl_map
.brk
= addr
;
2001 case PR_SET_MM_ARG_START
:
2002 prctl_map
.arg_start
= addr
;
2004 case PR_SET_MM_ARG_END
:
2005 prctl_map
.arg_end
= addr
;
2007 case PR_SET_MM_ENV_START
:
2008 prctl_map
.env_start
= addr
;
2010 case PR_SET_MM_ENV_END
:
2011 prctl_map
.env_end
= addr
;
2017 error
= validate_prctl_map(&prctl_map
);
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
:
2040 mm
->start_code
= prctl_map
.start_code
;
2041 mm
->end_code
= prctl_map
.end_code
;
2042 mm
->start_data
= prctl_map
.start_data
;
2043 mm
->end_data
= prctl_map
.end_data
;
2044 mm
->start_brk
= prctl_map
.start_brk
;
2045 mm
->brk
= prctl_map
.brk
;
2046 mm
->start_stack
= prctl_map
.start_stack
;
2047 mm
->arg_start
= prctl_map
.arg_start
;
2048 mm
->arg_end
= prctl_map
.arg_end
;
2049 mm
->env_start
= prctl_map
.env_start
;
2050 mm
->env_end
= prctl_map
.env_end
;
2054 up_write(&mm
->mmap_sem
);
2058 #ifdef CONFIG_CHECKPOINT_RESTORE
2059 static int prctl_get_tid_address(struct task_struct
*me
, int __user
**tid_addr
)
2061 return put_user(me
->clear_child_tid
, tid_addr
);
2064 static int prctl_get_tid_address(struct task_struct
*me
, int __user
**tid_addr
)
2070 static int propagate_has_child_subreaper(struct task_struct
*p
, void *data
)
2073 * If task has has_child_subreaper - all its decendants
2074 * already have these flag too and new decendants will
2075 * inherit it on fork, skip them.
2077 * If we've found child_reaper - skip descendants in
2078 * it's subtree as they will never get out pidns.
2080 if (p
->signal
->has_child_subreaper
||
2081 is_child_reaper(task_pid(p
)))
2084 p
->signal
->has_child_subreaper
= 1;
2088 SYSCALL_DEFINE5(prctl
, int, option
, unsigned long, arg2
, unsigned long, arg3
,
2089 unsigned long, arg4
, unsigned long, arg5
)
2091 struct task_struct
*me
= current
;
2092 unsigned char comm
[sizeof(me
->comm
)];
2095 error
= security_task_prctl(option
, arg2
, arg3
, arg4
, arg5
);
2096 if (error
!= -ENOSYS
)
2101 case PR_SET_PDEATHSIG
:
2102 if (!valid_signal(arg2
)) {
2106 me
->pdeath_signal
= arg2
;
2108 case PR_GET_PDEATHSIG
:
2109 error
= put_user(me
->pdeath_signal
, (int __user
*)arg2
);
2111 case PR_GET_DUMPABLE
:
2112 error
= get_dumpable(me
->mm
);
2114 case PR_SET_DUMPABLE
:
2115 if (arg2
!= SUID_DUMP_DISABLE
&& arg2
!= SUID_DUMP_USER
) {
2119 set_dumpable(me
->mm
, arg2
);
2122 case PR_SET_UNALIGN
:
2123 error
= SET_UNALIGN_CTL(me
, arg2
);
2125 case PR_GET_UNALIGN
:
2126 error
= GET_UNALIGN_CTL(me
, arg2
);
2129 error
= SET_FPEMU_CTL(me
, arg2
);
2132 error
= GET_FPEMU_CTL(me
, arg2
);
2135 error
= SET_FPEXC_CTL(me
, arg2
);
2138 error
= GET_FPEXC_CTL(me
, arg2
);
2141 error
= PR_TIMING_STATISTICAL
;
2144 if (arg2
!= PR_TIMING_STATISTICAL
)
2148 comm
[sizeof(me
->comm
) - 1] = 0;
2149 if (strncpy_from_user(comm
, (char __user
*)arg2
,
2150 sizeof(me
->comm
) - 1) < 0)
2152 set_task_comm(me
, comm
);
2153 proc_comm_connector(me
);
2156 get_task_comm(comm
, me
);
2157 if (copy_to_user((char __user
*)arg2
, comm
, sizeof(comm
)))
2161 error
= GET_ENDIAN(me
, arg2
);
2164 error
= SET_ENDIAN(me
, arg2
);
2166 case PR_GET_SECCOMP
:
2167 error
= prctl_get_seccomp();
2169 case PR_SET_SECCOMP
:
2170 error
= prctl_set_seccomp(arg2
, (char __user
*)arg3
);
2173 error
= GET_TSC_CTL(arg2
);
2176 error
= SET_TSC_CTL(arg2
);
2178 case PR_TASK_PERF_EVENTS_DISABLE
:
2179 error
= perf_event_task_disable();
2181 case PR_TASK_PERF_EVENTS_ENABLE
:
2182 error
= perf_event_task_enable();
2184 case PR_GET_TIMERSLACK
:
2185 if (current
->timer_slack_ns
> ULONG_MAX
)
2188 error
= current
->timer_slack_ns
;
2190 case PR_SET_TIMERSLACK
:
2192 current
->timer_slack_ns
=
2193 current
->default_timer_slack_ns
;
2195 current
->timer_slack_ns
= arg2
;
2201 case PR_MCE_KILL_CLEAR
:
2204 current
->flags
&= ~PF_MCE_PROCESS
;
2206 case PR_MCE_KILL_SET
:
2207 current
->flags
|= PF_MCE_PROCESS
;
2208 if (arg3
== PR_MCE_KILL_EARLY
)
2209 current
->flags
|= PF_MCE_EARLY
;
2210 else if (arg3
== PR_MCE_KILL_LATE
)
2211 current
->flags
&= ~PF_MCE_EARLY
;
2212 else if (arg3
== PR_MCE_KILL_DEFAULT
)
2214 ~(PF_MCE_EARLY
|PF_MCE_PROCESS
);
2222 case PR_MCE_KILL_GET
:
2223 if (arg2
| arg3
| arg4
| arg5
)
2225 if (current
->flags
& PF_MCE_PROCESS
)
2226 error
= (current
->flags
& PF_MCE_EARLY
) ?
2227 PR_MCE_KILL_EARLY
: PR_MCE_KILL_LATE
;
2229 error
= PR_MCE_KILL_DEFAULT
;
2232 error
= prctl_set_mm(arg2
, arg3
, arg4
, arg5
);
2234 case PR_GET_TID_ADDRESS
:
2235 error
= prctl_get_tid_address(me
, (int __user
**)arg2
);
2237 case PR_SET_CHILD_SUBREAPER
:
2238 me
->signal
->is_child_subreaper
= !!arg2
;
2242 walk_process_tree(me
, propagate_has_child_subreaper
, NULL
);
2244 case PR_GET_CHILD_SUBREAPER
:
2245 error
= put_user(me
->signal
->is_child_subreaper
,
2246 (int __user
*)arg2
);
2248 case PR_SET_NO_NEW_PRIVS
:
2249 if (arg2
!= 1 || arg3
|| arg4
|| arg5
)
2252 task_set_no_new_privs(current
);
2254 case PR_GET_NO_NEW_PRIVS
:
2255 if (arg2
|| arg3
|| arg4
|| arg5
)
2257 return task_no_new_privs(current
) ? 1 : 0;
2258 case PR_GET_THP_DISABLE
:
2259 if (arg2
|| arg3
|| arg4
|| arg5
)
2261 error
= !!(me
->mm
->def_flags
& VM_NOHUGEPAGE
);
2263 case PR_SET_THP_DISABLE
:
2264 if (arg3
|| arg4
|| arg5
)
2266 if (down_write_killable(&me
->mm
->mmap_sem
))
2269 me
->mm
->def_flags
|= VM_NOHUGEPAGE
;
2271 me
->mm
->def_flags
&= ~VM_NOHUGEPAGE
;
2272 up_write(&me
->mm
->mmap_sem
);
2274 case PR_MPX_ENABLE_MANAGEMENT
:
2275 if (arg2
|| arg3
|| arg4
|| arg5
)
2277 error
= MPX_ENABLE_MANAGEMENT();
2279 case PR_MPX_DISABLE_MANAGEMENT
:
2280 if (arg2
|| arg3
|| arg4
|| arg5
)
2282 error
= MPX_DISABLE_MANAGEMENT();
2284 case PR_SET_FP_MODE
:
2285 error
= SET_FP_MODE(me
, arg2
);
2287 case PR_GET_FP_MODE
:
2288 error
= GET_FP_MODE(me
);
2297 SYSCALL_DEFINE3(getcpu
, unsigned __user
*, cpup
, unsigned __user
*, nodep
,
2298 struct getcpu_cache __user
*, unused
)
2301 int cpu
= raw_smp_processor_id();
2304 err
|= put_user(cpu
, cpup
);
2306 err
|= put_user(cpu_to_node(cpu
), nodep
);
2307 return err
? -EFAULT
: 0;
2311 * do_sysinfo - fill in sysinfo struct
2312 * @info: pointer to buffer to fill
2314 static int do_sysinfo(struct sysinfo
*info
)
2316 unsigned long mem_total
, sav_total
;
2317 unsigned int mem_unit
, bitcount
;
2320 memset(info
, 0, sizeof(struct sysinfo
));
2322 get_monotonic_boottime(&tp
);
2323 info
->uptime
= tp
.tv_sec
+ (tp
.tv_nsec
? 1 : 0);
2325 get_avenrun(info
->loads
, 0, SI_LOAD_SHIFT
- FSHIFT
);
2327 info
->procs
= nr_threads
;
2333 * If the sum of all the available memory (i.e. ram + swap)
2334 * is less than can be stored in a 32 bit unsigned long then
2335 * we can be binary compatible with 2.2.x kernels. If not,
2336 * well, in that case 2.2.x was broken anyways...
2338 * -Erik Andersen <andersee@debian.org>
2341 mem_total
= info
->totalram
+ info
->totalswap
;
2342 if (mem_total
< info
->totalram
|| mem_total
< info
->totalswap
)
2345 mem_unit
= info
->mem_unit
;
2346 while (mem_unit
> 1) {
2349 sav_total
= mem_total
;
2351 if (mem_total
< sav_total
)
2356 * If mem_total did not overflow, multiply all memory values by
2357 * info->mem_unit and set it to 1. This leaves things compatible
2358 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2363 info
->totalram
<<= bitcount
;
2364 info
->freeram
<<= bitcount
;
2365 info
->sharedram
<<= bitcount
;
2366 info
->bufferram
<<= bitcount
;
2367 info
->totalswap
<<= bitcount
;
2368 info
->freeswap
<<= bitcount
;
2369 info
->totalhigh
<<= bitcount
;
2370 info
->freehigh
<<= bitcount
;
2376 SYSCALL_DEFINE1(sysinfo
, struct sysinfo __user
*, info
)
2382 if (copy_to_user(info
, &val
, sizeof(struct sysinfo
)))
2388 #ifdef CONFIG_COMPAT
2389 struct compat_sysinfo
{
2403 char _f
[20-2*sizeof(u32
)-sizeof(int)];
2406 COMPAT_SYSCALL_DEFINE1(sysinfo
, struct compat_sysinfo __user
*, info
)
2412 /* Check to see if any memory value is too large for 32-bit and scale
2415 if (upper_32_bits(s
.totalram
) || upper_32_bits(s
.totalswap
)) {
2418 while (s
.mem_unit
< PAGE_SIZE
) {
2423 s
.totalram
>>= bitcount
;
2424 s
.freeram
>>= bitcount
;
2425 s
.sharedram
>>= bitcount
;
2426 s
.bufferram
>>= bitcount
;
2427 s
.totalswap
>>= bitcount
;
2428 s
.freeswap
>>= bitcount
;
2429 s
.totalhigh
>>= bitcount
;
2430 s
.freehigh
>>= bitcount
;
2433 if (!access_ok(VERIFY_WRITE
, info
, sizeof(struct compat_sysinfo
)) ||
2434 __put_user(s
.uptime
, &info
->uptime
) ||
2435 __put_user(s
.loads
[0], &info
->loads
[0]) ||
2436 __put_user(s
.loads
[1], &info
->loads
[1]) ||
2437 __put_user(s
.loads
[2], &info
->loads
[2]) ||
2438 __put_user(s
.totalram
, &info
->totalram
) ||
2439 __put_user(s
.freeram
, &info
->freeram
) ||
2440 __put_user(s
.sharedram
, &info
->sharedram
) ||
2441 __put_user(s
.bufferram
, &info
->bufferram
) ||
2442 __put_user(s
.totalswap
, &info
->totalswap
) ||
2443 __put_user(s
.freeswap
, &info
->freeswap
) ||
2444 __put_user(s
.procs
, &info
->procs
) ||
2445 __put_user(s
.totalhigh
, &info
->totalhigh
) ||
2446 __put_user(s
.freehigh
, &info
->freehigh
) ||
2447 __put_user(s
.mem_unit
, &info
->mem_unit
))
2452 #endif /* CONFIG_COMPAT */