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/rcupdate.h>
56 #include <linux/uidgid.h>
57 #include <linux/cred.h>
59 #include <linux/kmsg_dump.h>
60 /* Move somewhere else to avoid recompiling? */
61 #include <generated/utsrelease.h>
63 #include <linux/uaccess.h>
65 #include <asm/unistd.h>
67 #ifndef SET_UNALIGN_CTL
68 # define SET_UNALIGN_CTL(a, b) (-EINVAL)
70 #ifndef GET_UNALIGN_CTL
71 # define GET_UNALIGN_CTL(a, b) (-EINVAL)
74 # define SET_FPEMU_CTL(a, b) (-EINVAL)
77 # define GET_FPEMU_CTL(a, b) (-EINVAL)
80 # define SET_FPEXC_CTL(a, b) (-EINVAL)
83 # define GET_FPEXC_CTL(a, b) (-EINVAL)
86 # define GET_ENDIAN(a, b) (-EINVAL)
89 # define SET_ENDIAN(a, b) (-EINVAL)
92 # define GET_TSC_CTL(a) (-EINVAL)
95 # define SET_TSC_CTL(a) (-EINVAL)
97 #ifndef MPX_ENABLE_MANAGEMENT
98 # define MPX_ENABLE_MANAGEMENT() (-EINVAL)
100 #ifndef MPX_DISABLE_MANAGEMENT
101 # define MPX_DISABLE_MANAGEMENT() (-EINVAL)
104 # define GET_FP_MODE(a) (-EINVAL)
107 # define SET_FP_MODE(a,b) (-EINVAL)
111 * this is where the system-wide overflow UID and GID are defined, for
112 * architectures that now have 32-bit UID/GID but didn't in the past
115 int overflowuid
= DEFAULT_OVERFLOWUID
;
116 int overflowgid
= DEFAULT_OVERFLOWGID
;
118 EXPORT_SYMBOL(overflowuid
);
119 EXPORT_SYMBOL(overflowgid
);
122 * the same as above, but for filesystems which can only store a 16-bit
123 * UID and GID. as such, this is needed on all architectures
126 int fs_overflowuid
= DEFAULT_FS_OVERFLOWUID
;
127 int fs_overflowgid
= DEFAULT_FS_OVERFLOWUID
;
129 EXPORT_SYMBOL(fs_overflowuid
);
130 EXPORT_SYMBOL(fs_overflowgid
);
133 * Returns true if current's euid is same as p's uid or euid,
134 * or has CAP_SYS_NICE to p's user_ns.
136 * Called with rcu_read_lock, creds are safe
138 static bool set_one_prio_perm(struct task_struct
*p
)
140 const struct cred
*cred
= current_cred(), *pcred
= __task_cred(p
);
142 if (uid_eq(pcred
->uid
, cred
->euid
) ||
143 uid_eq(pcred
->euid
, cred
->euid
))
145 if (ns_capable(pcred
->user_ns
, CAP_SYS_NICE
))
151 * set the priority of a task
152 * - the caller must hold the RCU read lock
154 static int set_one_prio(struct task_struct
*p
, int niceval
, int error
)
158 if (!set_one_prio_perm(p
)) {
162 if (niceval
< task_nice(p
) && !can_nice(p
, niceval
)) {
166 no_nice
= security_task_setnice(p
, niceval
);
173 set_user_nice(p
, niceval
);
178 SYSCALL_DEFINE3(setpriority
, int, which
, int, who
, int, niceval
)
180 struct task_struct
*g
, *p
;
181 struct user_struct
*user
;
182 const struct cred
*cred
= current_cred();
187 if (which
> PRIO_USER
|| which
< PRIO_PROCESS
)
190 /* normalize: avoid signed division (rounding problems) */
192 if (niceval
< MIN_NICE
)
194 if (niceval
> MAX_NICE
)
198 read_lock(&tasklist_lock
);
202 p
= find_task_by_vpid(who
);
206 error
= set_one_prio(p
, niceval
, error
);
210 pgrp
= find_vpid(who
);
212 pgrp
= task_pgrp(current
);
213 do_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
) {
214 error
= set_one_prio(p
, niceval
, error
);
215 } while_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
);
218 uid
= make_kuid(cred
->user_ns
, who
);
222 else if (!uid_eq(uid
, cred
->uid
)) {
223 user
= find_user(uid
);
225 goto out_unlock
; /* No processes for this user */
227 do_each_thread(g
, p
) {
228 if (uid_eq(task_uid(p
), uid
) && task_pid_vnr(p
))
229 error
= set_one_prio(p
, niceval
, error
);
230 } while_each_thread(g
, p
);
231 if (!uid_eq(uid
, cred
->uid
))
232 free_uid(user
); /* For find_user() */
236 read_unlock(&tasklist_lock
);
243 * Ugh. To avoid negative return values, "getpriority()" will
244 * not return the normal nice-value, but a negated value that
245 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
246 * to stay compatible.
248 SYSCALL_DEFINE2(getpriority
, int, which
, int, who
)
250 struct task_struct
*g
, *p
;
251 struct user_struct
*user
;
252 const struct cred
*cred
= current_cred();
253 long niceval
, retval
= -ESRCH
;
257 if (which
> PRIO_USER
|| which
< PRIO_PROCESS
)
261 read_lock(&tasklist_lock
);
265 p
= find_task_by_vpid(who
);
269 niceval
= nice_to_rlimit(task_nice(p
));
270 if (niceval
> retval
)
276 pgrp
= find_vpid(who
);
278 pgrp
= task_pgrp(current
);
279 do_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
) {
280 niceval
= nice_to_rlimit(task_nice(p
));
281 if (niceval
> retval
)
283 } while_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
);
286 uid
= make_kuid(cred
->user_ns
, who
);
290 else if (!uid_eq(uid
, cred
->uid
)) {
291 user
= find_user(uid
);
293 goto out_unlock
; /* No processes for this user */
295 do_each_thread(g
, p
) {
296 if (uid_eq(task_uid(p
), uid
) && task_pid_vnr(p
)) {
297 niceval
= nice_to_rlimit(task_nice(p
));
298 if (niceval
> retval
)
301 } while_each_thread(g
, p
);
302 if (!uid_eq(uid
, cred
->uid
))
303 free_uid(user
); /* for find_user() */
307 read_unlock(&tasklist_lock
);
314 * Unprivileged users may change the real gid to the effective gid
315 * or vice versa. (BSD-style)
317 * If you set the real gid at all, or set the effective gid to a value not
318 * equal to the real gid, then the saved gid is set to the new effective gid.
320 * This makes it possible for a setgid program to completely drop its
321 * privileges, which is often a useful assertion to make when you are doing
322 * a security audit over a program.
324 * The general idea is that a program which uses just setregid() will be
325 * 100% compatible with BSD. A program which uses just setgid() will be
326 * 100% compatible with POSIX with saved IDs.
328 * SMP: There are not races, the GIDs are checked only by filesystem
329 * operations (as far as semantic preservation is concerned).
331 #ifdef CONFIG_MULTIUSER
332 SYSCALL_DEFINE2(setregid
, gid_t
, rgid
, gid_t
, egid
)
334 struct user_namespace
*ns
= current_user_ns();
335 const struct cred
*old
;
340 krgid
= make_kgid(ns
, rgid
);
341 kegid
= make_kgid(ns
, egid
);
343 if ((rgid
!= (gid_t
) -1) && !gid_valid(krgid
))
345 if ((egid
!= (gid_t
) -1) && !gid_valid(kegid
))
348 new = prepare_creds();
351 old
= current_cred();
354 if (rgid
!= (gid_t
) -1) {
355 if (gid_eq(old
->gid
, krgid
) ||
356 gid_eq(old
->egid
, krgid
) ||
357 ns_capable(old
->user_ns
, CAP_SETGID
))
362 if (egid
!= (gid_t
) -1) {
363 if (gid_eq(old
->gid
, kegid
) ||
364 gid_eq(old
->egid
, kegid
) ||
365 gid_eq(old
->sgid
, kegid
) ||
366 ns_capable(old
->user_ns
, CAP_SETGID
))
372 if (rgid
!= (gid_t
) -1 ||
373 (egid
!= (gid_t
) -1 && !gid_eq(kegid
, old
->gid
)))
374 new->sgid
= new->egid
;
375 new->fsgid
= new->egid
;
377 return commit_creds(new);
385 * setgid() is implemented like SysV w/ SAVED_IDS
387 * SMP: Same implicit races as above.
389 SYSCALL_DEFINE1(setgid
, gid_t
, gid
)
391 struct user_namespace
*ns
= current_user_ns();
392 const struct cred
*old
;
397 kgid
= make_kgid(ns
, gid
);
398 if (!gid_valid(kgid
))
401 new = prepare_creds();
404 old
= current_cred();
407 if (ns_capable(old
->user_ns
, CAP_SETGID
))
408 new->gid
= new->egid
= new->sgid
= new->fsgid
= kgid
;
409 else if (gid_eq(kgid
, old
->gid
) || gid_eq(kgid
, old
->sgid
))
410 new->egid
= new->fsgid
= kgid
;
414 return commit_creds(new);
422 * change the user struct in a credentials set to match the new UID
424 static int set_user(struct cred
*new)
426 struct user_struct
*new_user
;
428 new_user
= alloc_uid(new->uid
);
433 * We don't fail in case of NPROC limit excess here because too many
434 * poorly written programs don't check set*uid() return code, assuming
435 * it never fails if called by root. We may still enforce NPROC limit
436 * for programs doing set*uid()+execve() by harmlessly deferring the
437 * failure to the execve() stage.
439 if (atomic_read(&new_user
->processes
) >= rlimit(RLIMIT_NPROC
) &&
440 new_user
!= INIT_USER
)
441 current
->flags
|= PF_NPROC_EXCEEDED
;
443 current
->flags
&= ~PF_NPROC_EXCEEDED
;
446 new->user
= new_user
;
451 * Unprivileged users may change the real uid to the effective uid
452 * or vice versa. (BSD-style)
454 * If you set the real uid at all, or set the effective uid to a value not
455 * equal to the real uid, then the saved uid is set to the new effective uid.
457 * This makes it possible for a setuid program to completely drop its
458 * privileges, which is often a useful assertion to make when you are doing
459 * a security audit over a program.
461 * The general idea is that a program which uses just setreuid() will be
462 * 100% compatible with BSD. A program which uses just setuid() will be
463 * 100% compatible with POSIX with saved IDs.
465 SYSCALL_DEFINE2(setreuid
, uid_t
, ruid
, uid_t
, euid
)
467 struct user_namespace
*ns
= current_user_ns();
468 const struct cred
*old
;
473 kruid
= make_kuid(ns
, ruid
);
474 keuid
= make_kuid(ns
, euid
);
476 if ((ruid
!= (uid_t
) -1) && !uid_valid(kruid
))
478 if ((euid
!= (uid_t
) -1) && !uid_valid(keuid
))
481 new = prepare_creds();
484 old
= current_cred();
487 if (ruid
!= (uid_t
) -1) {
489 if (!uid_eq(old
->uid
, kruid
) &&
490 !uid_eq(old
->euid
, kruid
) &&
491 !ns_capable(old
->user_ns
, CAP_SETUID
))
495 if (euid
!= (uid_t
) -1) {
497 if (!uid_eq(old
->uid
, keuid
) &&
498 !uid_eq(old
->euid
, keuid
) &&
499 !uid_eq(old
->suid
, keuid
) &&
500 !ns_capable(old
->user_ns
, CAP_SETUID
))
504 if (!uid_eq(new->uid
, old
->uid
)) {
505 retval
= set_user(new);
509 if (ruid
!= (uid_t
) -1 ||
510 (euid
!= (uid_t
) -1 && !uid_eq(keuid
, old
->uid
)))
511 new->suid
= new->euid
;
512 new->fsuid
= new->euid
;
514 retval
= security_task_fix_setuid(new, old
, LSM_SETID_RE
);
518 return commit_creds(new);
526 * setuid() is implemented like SysV with SAVED_IDS
528 * Note that SAVED_ID's is deficient in that a setuid root program
529 * like sendmail, for example, cannot set its uid to be a normal
530 * user and then switch back, because if you're root, setuid() sets
531 * the saved uid too. If you don't like this, blame the bright people
532 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
533 * will allow a root program to temporarily drop privileges and be able to
534 * regain them by swapping the real and effective uid.
536 SYSCALL_DEFINE1(setuid
, uid_t
, uid
)
538 struct user_namespace
*ns
= current_user_ns();
539 const struct cred
*old
;
544 kuid
= make_kuid(ns
, uid
);
545 if (!uid_valid(kuid
))
548 new = prepare_creds();
551 old
= current_cred();
554 if (ns_capable(old
->user_ns
, CAP_SETUID
)) {
555 new->suid
= new->uid
= kuid
;
556 if (!uid_eq(kuid
, old
->uid
)) {
557 retval
= set_user(new);
561 } else if (!uid_eq(kuid
, old
->uid
) && !uid_eq(kuid
, new->suid
)) {
565 new->fsuid
= new->euid
= kuid
;
567 retval
= security_task_fix_setuid(new, old
, LSM_SETID_ID
);
571 return commit_creds(new);
580 * This function implements a generic ability to update ruid, euid,
581 * and suid. This allows you to implement the 4.4 compatible seteuid().
583 SYSCALL_DEFINE3(setresuid
, uid_t
, ruid
, uid_t
, euid
, uid_t
, suid
)
585 struct user_namespace
*ns
= current_user_ns();
586 const struct cred
*old
;
589 kuid_t kruid
, keuid
, ksuid
;
591 kruid
= make_kuid(ns
, ruid
);
592 keuid
= make_kuid(ns
, euid
);
593 ksuid
= make_kuid(ns
, suid
);
595 if ((ruid
!= (uid_t
) -1) && !uid_valid(kruid
))
598 if ((euid
!= (uid_t
) -1) && !uid_valid(keuid
))
601 if ((suid
!= (uid_t
) -1) && !uid_valid(ksuid
))
604 new = prepare_creds();
608 old
= current_cred();
611 if (!ns_capable(old
->user_ns
, CAP_SETUID
)) {
612 if (ruid
!= (uid_t
) -1 && !uid_eq(kruid
, old
->uid
) &&
613 !uid_eq(kruid
, old
->euid
) && !uid_eq(kruid
, old
->suid
))
615 if (euid
!= (uid_t
) -1 && !uid_eq(keuid
, old
->uid
) &&
616 !uid_eq(keuid
, old
->euid
) && !uid_eq(keuid
, old
->suid
))
618 if (suid
!= (uid_t
) -1 && !uid_eq(ksuid
, old
->uid
) &&
619 !uid_eq(ksuid
, old
->euid
) && !uid_eq(ksuid
, old
->suid
))
623 if (ruid
!= (uid_t
) -1) {
625 if (!uid_eq(kruid
, old
->uid
)) {
626 retval
= set_user(new);
631 if (euid
!= (uid_t
) -1)
633 if (suid
!= (uid_t
) -1)
635 new->fsuid
= new->euid
;
637 retval
= security_task_fix_setuid(new, old
, LSM_SETID_RES
);
641 return commit_creds(new);
648 SYSCALL_DEFINE3(getresuid
, uid_t __user
*, ruidp
, uid_t __user
*, euidp
, uid_t __user
*, suidp
)
650 const struct cred
*cred
= current_cred();
652 uid_t ruid
, euid
, suid
;
654 ruid
= from_kuid_munged(cred
->user_ns
, cred
->uid
);
655 euid
= from_kuid_munged(cred
->user_ns
, cred
->euid
);
656 suid
= from_kuid_munged(cred
->user_ns
, cred
->suid
);
658 retval
= put_user(ruid
, ruidp
);
660 retval
= put_user(euid
, euidp
);
662 return put_user(suid
, suidp
);
668 * Same as above, but for rgid, egid, sgid.
670 SYSCALL_DEFINE3(setresgid
, gid_t
, rgid
, gid_t
, egid
, gid_t
, sgid
)
672 struct user_namespace
*ns
= current_user_ns();
673 const struct cred
*old
;
676 kgid_t krgid
, kegid
, ksgid
;
678 krgid
= make_kgid(ns
, rgid
);
679 kegid
= make_kgid(ns
, egid
);
680 ksgid
= make_kgid(ns
, sgid
);
682 if ((rgid
!= (gid_t
) -1) && !gid_valid(krgid
))
684 if ((egid
!= (gid_t
) -1) && !gid_valid(kegid
))
686 if ((sgid
!= (gid_t
) -1) && !gid_valid(ksgid
))
689 new = prepare_creds();
692 old
= current_cred();
695 if (!ns_capable(old
->user_ns
, CAP_SETGID
)) {
696 if (rgid
!= (gid_t
) -1 && !gid_eq(krgid
, old
->gid
) &&
697 !gid_eq(krgid
, old
->egid
) && !gid_eq(krgid
, old
->sgid
))
699 if (egid
!= (gid_t
) -1 && !gid_eq(kegid
, old
->gid
) &&
700 !gid_eq(kegid
, old
->egid
) && !gid_eq(kegid
, old
->sgid
))
702 if (sgid
!= (gid_t
) -1 && !gid_eq(ksgid
, old
->gid
) &&
703 !gid_eq(ksgid
, old
->egid
) && !gid_eq(ksgid
, old
->sgid
))
707 if (rgid
!= (gid_t
) -1)
709 if (egid
!= (gid_t
) -1)
711 if (sgid
!= (gid_t
) -1)
713 new->fsgid
= new->egid
;
715 return commit_creds(new);
722 SYSCALL_DEFINE3(getresgid
, gid_t __user
*, rgidp
, gid_t __user
*, egidp
, gid_t __user
*, sgidp
)
724 const struct cred
*cred
= current_cred();
726 gid_t rgid
, egid
, sgid
;
728 rgid
= from_kgid_munged(cred
->user_ns
, cred
->gid
);
729 egid
= from_kgid_munged(cred
->user_ns
, cred
->egid
);
730 sgid
= from_kgid_munged(cred
->user_ns
, cred
->sgid
);
732 retval
= put_user(rgid
, rgidp
);
734 retval
= put_user(egid
, egidp
);
736 retval
= put_user(sgid
, sgidp
);
744 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
745 * is used for "access()" and for the NFS daemon (letting nfsd stay at
746 * whatever uid it wants to). It normally shadows "euid", except when
747 * explicitly set by setfsuid() or for access..
749 SYSCALL_DEFINE1(setfsuid
, uid_t
, uid
)
751 const struct cred
*old
;
756 old
= current_cred();
757 old_fsuid
= from_kuid_munged(old
->user_ns
, old
->fsuid
);
759 kuid
= make_kuid(old
->user_ns
, uid
);
760 if (!uid_valid(kuid
))
763 new = prepare_creds();
767 if (uid_eq(kuid
, old
->uid
) || uid_eq(kuid
, old
->euid
) ||
768 uid_eq(kuid
, old
->suid
) || uid_eq(kuid
, old
->fsuid
) ||
769 ns_capable(old
->user_ns
, CAP_SETUID
)) {
770 if (!uid_eq(kuid
, old
->fsuid
)) {
772 if (security_task_fix_setuid(new, old
, LSM_SETID_FS
) == 0)
786 * Samma på svenska..
788 SYSCALL_DEFINE1(setfsgid
, gid_t
, gid
)
790 const struct cred
*old
;
795 old
= current_cred();
796 old_fsgid
= from_kgid_munged(old
->user_ns
, old
->fsgid
);
798 kgid
= make_kgid(old
->user_ns
, gid
);
799 if (!gid_valid(kgid
))
802 new = prepare_creds();
806 if (gid_eq(kgid
, old
->gid
) || gid_eq(kgid
, old
->egid
) ||
807 gid_eq(kgid
, old
->sgid
) || gid_eq(kgid
, old
->fsgid
) ||
808 ns_capable(old
->user_ns
, CAP_SETGID
)) {
809 if (!gid_eq(kgid
, old
->fsgid
)) {
822 #endif /* CONFIG_MULTIUSER */
825 * sys_getpid - return the thread group id of the current process
827 * Note, despite the name, this returns the tgid not the pid. The tgid and
828 * the pid are identical unless CLONE_THREAD was specified on clone() in
829 * which case the tgid is the same in all threads of the same group.
831 * This is SMP safe as current->tgid does not change.
833 SYSCALL_DEFINE0(getpid
)
835 return task_tgid_vnr(current
);
838 /* Thread ID - the internal kernel "pid" */
839 SYSCALL_DEFINE0(gettid
)
841 return task_pid_vnr(current
);
845 * Accessing ->real_parent is not SMP-safe, it could
846 * change from under us. However, we can use a stale
847 * value of ->real_parent under rcu_read_lock(), see
848 * release_task()->call_rcu(delayed_put_task_struct).
850 SYSCALL_DEFINE0(getppid
)
855 pid
= task_tgid_vnr(rcu_dereference(current
->real_parent
));
861 SYSCALL_DEFINE0(getuid
)
863 /* Only we change this so SMP safe */
864 return from_kuid_munged(current_user_ns(), current_uid());
867 SYSCALL_DEFINE0(geteuid
)
869 /* Only we change this so SMP safe */
870 return from_kuid_munged(current_user_ns(), current_euid());
873 SYSCALL_DEFINE0(getgid
)
875 /* Only we change this so SMP safe */
876 return from_kgid_munged(current_user_ns(), current_gid());
879 SYSCALL_DEFINE0(getegid
)
881 /* Only we change this so SMP safe */
882 return from_kgid_munged(current_user_ns(), current_egid());
885 void do_sys_times(struct tms
*tms
)
887 u64 tgutime
, tgstime
, cutime
, cstime
;
889 thread_group_cputime_adjusted(current
, &tgutime
, &tgstime
);
890 cutime
= current
->signal
->cutime
;
891 cstime
= current
->signal
->cstime
;
892 tms
->tms_utime
= nsec_to_clock_t(tgutime
);
893 tms
->tms_stime
= nsec_to_clock_t(tgstime
);
894 tms
->tms_cutime
= nsec_to_clock_t(cutime
);
895 tms
->tms_cstime
= nsec_to_clock_t(cstime
);
898 SYSCALL_DEFINE1(times
, struct tms __user
*, tbuf
)
904 if (copy_to_user(tbuf
, &tmp
, sizeof(struct tms
)))
907 force_successful_syscall_return();
908 return (long) jiffies_64_to_clock_t(get_jiffies_64());
912 * This needs some heavy checking ...
913 * I just haven't the stomach for it. I also don't fully
914 * understand sessions/pgrp etc. Let somebody who does explain it.
916 * OK, I think I have the protection semantics right.... this is really
917 * only important on a multi-user system anyway, to make sure one user
918 * can't send a signal to a process owned by another. -TYT, 12/12/91
920 * !PF_FORKNOEXEC check to conform completely to POSIX.
922 SYSCALL_DEFINE2(setpgid
, pid_t
, pid
, pid_t
, pgid
)
924 struct task_struct
*p
;
925 struct task_struct
*group_leader
= current
->group_leader
;
930 pid
= task_pid_vnr(group_leader
);
937 /* From this point forward we keep holding onto the tasklist lock
938 * so that our parent does not change from under us. -DaveM
940 write_lock_irq(&tasklist_lock
);
943 p
= find_task_by_vpid(pid
);
948 if (!thread_group_leader(p
))
951 if (same_thread_group(p
->real_parent
, group_leader
)) {
953 if (task_session(p
) != task_session(group_leader
))
956 if (!(p
->flags
& PF_FORKNOEXEC
))
960 if (p
!= group_leader
)
965 if (p
->signal
->leader
)
970 struct task_struct
*g
;
972 pgrp
= find_vpid(pgid
);
973 g
= pid_task(pgrp
, PIDTYPE_PGID
);
974 if (!g
|| task_session(g
) != task_session(group_leader
))
978 err
= security_task_setpgid(p
, pgid
);
982 if (task_pgrp(p
) != pgrp
)
983 change_pid(p
, PIDTYPE_PGID
, pgrp
);
987 /* All paths lead to here, thus we are safe. -DaveM */
988 write_unlock_irq(&tasklist_lock
);
993 SYSCALL_DEFINE1(getpgid
, pid_t
, pid
)
995 struct task_struct
*p
;
1001 grp
= task_pgrp(current
);
1004 p
= find_task_by_vpid(pid
);
1011 retval
= security_task_getpgid(p
);
1015 retval
= pid_vnr(grp
);
1021 #ifdef __ARCH_WANT_SYS_GETPGRP
1023 SYSCALL_DEFINE0(getpgrp
)
1025 return sys_getpgid(0);
1030 SYSCALL_DEFINE1(getsid
, pid_t
, pid
)
1032 struct task_struct
*p
;
1038 sid
= task_session(current
);
1041 p
= find_task_by_vpid(pid
);
1044 sid
= task_session(p
);
1048 retval
= security_task_getsid(p
);
1052 retval
= pid_vnr(sid
);
1058 static void set_special_pids(struct pid
*pid
)
1060 struct task_struct
*curr
= current
->group_leader
;
1062 if (task_session(curr
) != pid
)
1063 change_pid(curr
, PIDTYPE_SID
, pid
);
1065 if (task_pgrp(curr
) != pid
)
1066 change_pid(curr
, PIDTYPE_PGID
, pid
);
1069 SYSCALL_DEFINE0(setsid
)
1071 struct task_struct
*group_leader
= current
->group_leader
;
1072 struct pid
*sid
= task_pid(group_leader
);
1073 pid_t session
= pid_vnr(sid
);
1076 write_lock_irq(&tasklist_lock
);
1077 /* Fail if I am already a session leader */
1078 if (group_leader
->signal
->leader
)
1081 /* Fail if a process group id already exists that equals the
1082 * proposed session id.
1084 if (pid_task(sid
, PIDTYPE_PGID
))
1087 group_leader
->signal
->leader
= 1;
1088 set_special_pids(sid
);
1090 proc_clear_tty(group_leader
);
1094 write_unlock_irq(&tasklist_lock
);
1096 proc_sid_connector(group_leader
);
1097 sched_autogroup_create_attach(group_leader
);
1102 DECLARE_RWSEM(uts_sem
);
1104 #ifdef COMPAT_UTS_MACHINE
1105 #define override_architecture(name) \
1106 (personality(current->personality) == PER_LINUX32 && \
1107 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1108 sizeof(COMPAT_UTS_MACHINE)))
1110 #define override_architecture(name) 0
1114 * Work around broken programs that cannot handle "Linux 3.0".
1115 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1116 * And we map 4.x to 2.6.60+x, so 4.0 would be 2.6.60.
1118 static int override_release(char __user
*release
, size_t len
)
1122 if (current
->personality
& UNAME26
) {
1123 const char *rest
= UTS_RELEASE
;
1124 char buf
[65] = { 0 };
1130 if (*rest
== '.' && ++ndots
>= 3)
1132 if (!isdigit(*rest
) && *rest
!= '.')
1136 v
= ((LINUX_VERSION_CODE
>> 8) & 0xff) + 60;
1137 copy
= clamp_t(size_t, len
, 1, sizeof(buf
));
1138 copy
= scnprintf(buf
, copy
, "2.6.%u%s", v
, rest
);
1139 ret
= copy_to_user(release
, buf
, copy
+ 1);
1144 SYSCALL_DEFINE1(newuname
, struct new_utsname __user
*, name
)
1148 down_read(&uts_sem
);
1149 if (copy_to_user(name
, utsname(), sizeof *name
))
1153 if (!errno
&& override_release(name
->release
, sizeof(name
->release
)))
1155 if (!errno
&& override_architecture(name
))
1160 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1164 SYSCALL_DEFINE1(uname
, struct old_utsname __user
*, name
)
1171 down_read(&uts_sem
);
1172 if (copy_to_user(name
, utsname(), sizeof(*name
)))
1176 if (!error
&& override_release(name
->release
, sizeof(name
->release
)))
1178 if (!error
&& override_architecture(name
))
1183 SYSCALL_DEFINE1(olduname
, struct oldold_utsname __user
*, name
)
1189 if (!access_ok(VERIFY_WRITE
, name
, sizeof(struct oldold_utsname
)))
1192 down_read(&uts_sem
);
1193 error
= __copy_to_user(&name
->sysname
, &utsname()->sysname
,
1195 error
|= __put_user(0, name
->sysname
+ __OLD_UTS_LEN
);
1196 error
|= __copy_to_user(&name
->nodename
, &utsname()->nodename
,
1198 error
|= __put_user(0, name
->nodename
+ __OLD_UTS_LEN
);
1199 error
|= __copy_to_user(&name
->release
, &utsname()->release
,
1201 error
|= __put_user(0, name
->release
+ __OLD_UTS_LEN
);
1202 error
|= __copy_to_user(&name
->version
, &utsname()->version
,
1204 error
|= __put_user(0, name
->version
+ __OLD_UTS_LEN
);
1205 error
|= __copy_to_user(&name
->machine
, &utsname()->machine
,
1207 error
|= __put_user(0, name
->machine
+ __OLD_UTS_LEN
);
1210 if (!error
&& override_architecture(name
))
1212 if (!error
&& override_release(name
->release
, sizeof(name
->release
)))
1214 return error
? -EFAULT
: 0;
1218 SYSCALL_DEFINE2(sethostname
, char __user
*, name
, int, len
)
1221 char tmp
[__NEW_UTS_LEN
];
1223 if (!ns_capable(current
->nsproxy
->uts_ns
->user_ns
, CAP_SYS_ADMIN
))
1226 if (len
< 0 || len
> __NEW_UTS_LEN
)
1228 down_write(&uts_sem
);
1230 if (!copy_from_user(tmp
, name
, len
)) {
1231 struct new_utsname
*u
= utsname();
1233 memcpy(u
->nodename
, tmp
, len
);
1234 memset(u
->nodename
+ len
, 0, sizeof(u
->nodename
) - len
);
1236 uts_proc_notify(UTS_PROC_HOSTNAME
);
1242 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1244 SYSCALL_DEFINE2(gethostname
, char __user
*, name
, int, len
)
1247 struct new_utsname
*u
;
1251 down_read(&uts_sem
);
1253 i
= 1 + strlen(u
->nodename
);
1257 if (copy_to_user(name
, u
->nodename
, i
))
1266 * Only setdomainname; getdomainname can be implemented by calling
1269 SYSCALL_DEFINE2(setdomainname
, char __user
*, name
, int, len
)
1272 char tmp
[__NEW_UTS_LEN
];
1274 if (!ns_capable(current
->nsproxy
->uts_ns
->user_ns
, CAP_SYS_ADMIN
))
1276 if (len
< 0 || len
> __NEW_UTS_LEN
)
1279 down_write(&uts_sem
);
1281 if (!copy_from_user(tmp
, name
, len
)) {
1282 struct new_utsname
*u
= utsname();
1284 memcpy(u
->domainname
, tmp
, len
);
1285 memset(u
->domainname
+ len
, 0, sizeof(u
->domainname
) - len
);
1287 uts_proc_notify(UTS_PROC_DOMAINNAME
);
1293 SYSCALL_DEFINE2(getrlimit
, unsigned int, resource
, struct rlimit __user
*, rlim
)
1295 struct rlimit value
;
1298 ret
= do_prlimit(current
, resource
, NULL
, &value
);
1300 ret
= copy_to_user(rlim
, &value
, sizeof(*rlim
)) ? -EFAULT
: 0;
1305 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1308 * Back compatibility for getrlimit. Needed for some apps.
1310 SYSCALL_DEFINE2(old_getrlimit
, unsigned int, resource
,
1311 struct rlimit __user
*, rlim
)
1314 if (resource
>= RLIM_NLIMITS
)
1317 task_lock(current
->group_leader
);
1318 x
= current
->signal
->rlim
[resource
];
1319 task_unlock(current
->group_leader
);
1320 if (x
.rlim_cur
> 0x7FFFFFFF)
1321 x
.rlim_cur
= 0x7FFFFFFF;
1322 if (x
.rlim_max
> 0x7FFFFFFF)
1323 x
.rlim_max
= 0x7FFFFFFF;
1324 return copy_to_user(rlim
, &x
, sizeof(x
)) ? -EFAULT
: 0;
1329 static inline bool rlim64_is_infinity(__u64 rlim64
)
1331 #if BITS_PER_LONG < 64
1332 return rlim64
>= ULONG_MAX
;
1334 return rlim64
== RLIM64_INFINITY
;
1338 static void rlim_to_rlim64(const struct rlimit
*rlim
, struct rlimit64
*rlim64
)
1340 if (rlim
->rlim_cur
== RLIM_INFINITY
)
1341 rlim64
->rlim_cur
= RLIM64_INFINITY
;
1343 rlim64
->rlim_cur
= rlim
->rlim_cur
;
1344 if (rlim
->rlim_max
== RLIM_INFINITY
)
1345 rlim64
->rlim_max
= RLIM64_INFINITY
;
1347 rlim64
->rlim_max
= rlim
->rlim_max
;
1350 static void rlim64_to_rlim(const struct rlimit64
*rlim64
, struct rlimit
*rlim
)
1352 if (rlim64_is_infinity(rlim64
->rlim_cur
))
1353 rlim
->rlim_cur
= RLIM_INFINITY
;
1355 rlim
->rlim_cur
= (unsigned long)rlim64
->rlim_cur
;
1356 if (rlim64_is_infinity(rlim64
->rlim_max
))
1357 rlim
->rlim_max
= RLIM_INFINITY
;
1359 rlim
->rlim_max
= (unsigned long)rlim64
->rlim_max
;
1362 /* make sure you are allowed to change @tsk limits before calling this */
1363 int do_prlimit(struct task_struct
*tsk
, unsigned int resource
,
1364 struct rlimit
*new_rlim
, struct rlimit
*old_rlim
)
1366 struct rlimit
*rlim
;
1369 if (resource
>= RLIM_NLIMITS
)
1372 if (new_rlim
->rlim_cur
> new_rlim
->rlim_max
)
1374 if (resource
== RLIMIT_NOFILE
&&
1375 new_rlim
->rlim_max
> sysctl_nr_open
)
1379 /* protect tsk->signal and tsk->sighand from disappearing */
1380 read_lock(&tasklist_lock
);
1381 if (!tsk
->sighand
) {
1386 rlim
= tsk
->signal
->rlim
+ resource
;
1387 task_lock(tsk
->group_leader
);
1389 /* Keep the capable check against init_user_ns until
1390 cgroups can contain all limits */
1391 if (new_rlim
->rlim_max
> rlim
->rlim_max
&&
1392 !capable(CAP_SYS_RESOURCE
))
1395 retval
= security_task_setrlimit(tsk
->group_leader
,
1396 resource
, new_rlim
);
1397 if (resource
== RLIMIT_CPU
&& new_rlim
->rlim_cur
== 0) {
1399 * The caller is asking for an immediate RLIMIT_CPU
1400 * expiry. But we use the zero value to mean "it was
1401 * never set". So let's cheat and make it one second
1404 new_rlim
->rlim_cur
= 1;
1413 task_unlock(tsk
->group_leader
);
1416 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1417 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1418 * very long-standing error, and fixing it now risks breakage of
1419 * applications, so we live with it
1421 if (!retval
&& new_rlim
&& resource
== RLIMIT_CPU
&&
1422 new_rlim
->rlim_cur
!= RLIM_INFINITY
&&
1423 IS_ENABLED(CONFIG_POSIX_TIMERS
))
1424 update_rlimit_cpu(tsk
, new_rlim
->rlim_cur
);
1426 read_unlock(&tasklist_lock
);
1430 /* rcu lock must be held */
1431 static int check_prlimit_permission(struct task_struct
*task
)
1433 const struct cred
*cred
= current_cred(), *tcred
;
1435 if (current
== task
)
1438 tcred
= __task_cred(task
);
1439 if (uid_eq(cred
->uid
, tcred
->euid
) &&
1440 uid_eq(cred
->uid
, tcred
->suid
) &&
1441 uid_eq(cred
->uid
, tcred
->uid
) &&
1442 gid_eq(cred
->gid
, tcred
->egid
) &&
1443 gid_eq(cred
->gid
, tcred
->sgid
) &&
1444 gid_eq(cred
->gid
, tcred
->gid
))
1446 if (ns_capable(tcred
->user_ns
, CAP_SYS_RESOURCE
))
1452 SYSCALL_DEFINE4(prlimit64
, pid_t
, pid
, unsigned int, resource
,
1453 const struct rlimit64 __user
*, new_rlim
,
1454 struct rlimit64 __user
*, old_rlim
)
1456 struct rlimit64 old64
, new64
;
1457 struct rlimit old
, new;
1458 struct task_struct
*tsk
;
1462 if (copy_from_user(&new64
, new_rlim
, sizeof(new64
)))
1464 rlim64_to_rlim(&new64
, &new);
1468 tsk
= pid
? find_task_by_vpid(pid
) : current
;
1473 ret
= check_prlimit_permission(tsk
);
1478 get_task_struct(tsk
);
1481 ret
= do_prlimit(tsk
, resource
, new_rlim
? &new : NULL
,
1482 old_rlim
? &old
: NULL
);
1484 if (!ret
&& old_rlim
) {
1485 rlim_to_rlim64(&old
, &old64
);
1486 if (copy_to_user(old_rlim
, &old64
, sizeof(old64
)))
1490 put_task_struct(tsk
);
1494 SYSCALL_DEFINE2(setrlimit
, unsigned int, resource
, struct rlimit __user
*, rlim
)
1496 struct rlimit new_rlim
;
1498 if (copy_from_user(&new_rlim
, rlim
, sizeof(*rlim
)))
1500 return do_prlimit(current
, resource
, &new_rlim
, NULL
);
1504 * It would make sense to put struct rusage in the task_struct,
1505 * except that would make the task_struct be *really big*. After
1506 * task_struct gets moved into malloc'ed memory, it would
1507 * make sense to do this. It will make moving the rest of the information
1508 * a lot simpler! (Which we're not doing right now because we're not
1509 * measuring them yet).
1511 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1512 * races with threads incrementing their own counters. But since word
1513 * reads are atomic, we either get new values or old values and we don't
1514 * care which for the sums. We always take the siglock to protect reading
1515 * the c* fields from p->signal from races with exit.c updating those
1516 * fields when reaping, so a sample either gets all the additions of a
1517 * given child after it's reaped, or none so this sample is before reaping.
1520 * We need to take the siglock for CHILDEREN, SELF and BOTH
1521 * for the cases current multithreaded, non-current single threaded
1522 * non-current multithreaded. Thread traversal is now safe with
1524 * Strictly speaking, we donot need to take the siglock if we are current and
1525 * single threaded, as no one else can take our signal_struct away, no one
1526 * else can reap the children to update signal->c* counters, and no one else
1527 * can race with the signal-> fields. If we do not take any lock, the
1528 * signal-> fields could be read out of order while another thread was just
1529 * exiting. So we should place a read memory barrier when we avoid the lock.
1530 * On the writer side, write memory barrier is implied in __exit_signal
1531 * as __exit_signal releases the siglock spinlock after updating the signal->
1532 * fields. But we don't do this yet to keep things simple.
1536 static void accumulate_thread_rusage(struct task_struct
*t
, struct rusage
*r
)
1538 r
->ru_nvcsw
+= t
->nvcsw
;
1539 r
->ru_nivcsw
+= t
->nivcsw
;
1540 r
->ru_minflt
+= t
->min_flt
;
1541 r
->ru_majflt
+= t
->maj_flt
;
1542 r
->ru_inblock
+= task_io_get_inblock(t
);
1543 r
->ru_oublock
+= task_io_get_oublock(t
);
1546 static void k_getrusage(struct task_struct
*p
, int who
, struct rusage
*r
)
1548 struct task_struct
*t
;
1549 unsigned long flags
;
1550 u64 tgutime
, tgstime
, utime
, stime
;
1551 unsigned long maxrss
= 0;
1553 memset((char *)r
, 0, sizeof (*r
));
1556 if (who
== RUSAGE_THREAD
) {
1557 task_cputime_adjusted(current
, &utime
, &stime
);
1558 accumulate_thread_rusage(p
, r
);
1559 maxrss
= p
->signal
->maxrss
;
1563 if (!lock_task_sighand(p
, &flags
))
1568 case RUSAGE_CHILDREN
:
1569 utime
= p
->signal
->cutime
;
1570 stime
= p
->signal
->cstime
;
1571 r
->ru_nvcsw
= p
->signal
->cnvcsw
;
1572 r
->ru_nivcsw
= p
->signal
->cnivcsw
;
1573 r
->ru_minflt
= p
->signal
->cmin_flt
;
1574 r
->ru_majflt
= p
->signal
->cmaj_flt
;
1575 r
->ru_inblock
= p
->signal
->cinblock
;
1576 r
->ru_oublock
= p
->signal
->coublock
;
1577 maxrss
= p
->signal
->cmaxrss
;
1579 if (who
== RUSAGE_CHILDREN
)
1583 thread_group_cputime_adjusted(p
, &tgutime
, &tgstime
);
1586 r
->ru_nvcsw
+= p
->signal
->nvcsw
;
1587 r
->ru_nivcsw
+= p
->signal
->nivcsw
;
1588 r
->ru_minflt
+= p
->signal
->min_flt
;
1589 r
->ru_majflt
+= p
->signal
->maj_flt
;
1590 r
->ru_inblock
+= p
->signal
->inblock
;
1591 r
->ru_oublock
+= p
->signal
->oublock
;
1592 if (maxrss
< p
->signal
->maxrss
)
1593 maxrss
= p
->signal
->maxrss
;
1596 accumulate_thread_rusage(t
, r
);
1597 } while_each_thread(p
, t
);
1603 unlock_task_sighand(p
, &flags
);
1606 r
->ru_utime
= ns_to_timeval(utime
);
1607 r
->ru_stime
= ns_to_timeval(stime
);
1609 if (who
!= RUSAGE_CHILDREN
) {
1610 struct mm_struct
*mm
= get_task_mm(p
);
1613 setmax_mm_hiwater_rss(&maxrss
, mm
);
1617 r
->ru_maxrss
= maxrss
* (PAGE_SIZE
/ 1024); /* convert pages to KBs */
1620 int getrusage(struct task_struct
*p
, int who
, struct rusage __user
*ru
)
1624 k_getrusage(p
, who
, &r
);
1625 return copy_to_user(ru
, &r
, sizeof(r
)) ? -EFAULT
: 0;
1628 SYSCALL_DEFINE2(getrusage
, int, who
, struct rusage __user
*, ru
)
1630 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
&&
1631 who
!= RUSAGE_THREAD
)
1633 return getrusage(current
, who
, ru
);
1636 #ifdef CONFIG_COMPAT
1637 COMPAT_SYSCALL_DEFINE2(getrusage
, int, who
, struct compat_rusage __user
*, ru
)
1641 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
&&
1642 who
!= RUSAGE_THREAD
)
1645 k_getrusage(current
, who
, &r
);
1646 return put_compat_rusage(&r
, ru
);
1650 SYSCALL_DEFINE1(umask
, int, mask
)
1652 mask
= xchg(¤t
->fs
->umask
, mask
& S_IRWXUGO
);
1656 static int prctl_set_mm_exe_file(struct mm_struct
*mm
, unsigned int fd
)
1659 struct file
*old_exe
, *exe_file
;
1660 struct inode
*inode
;
1667 inode
= file_inode(exe
.file
);
1670 * Because the original mm->exe_file points to executable file, make
1671 * sure that this one is executable as well, to avoid breaking an
1675 if (!S_ISREG(inode
->i_mode
) || path_noexec(&exe
.file
->f_path
))
1678 err
= inode_permission(inode
, MAY_EXEC
);
1683 * Forbid mm->exe_file change if old file still mapped.
1685 exe_file
= get_mm_exe_file(mm
);
1688 struct vm_area_struct
*vma
;
1690 down_read(&mm
->mmap_sem
);
1691 for (vma
= mm
->mmap
; vma
; vma
= vma
->vm_next
) {
1694 if (path_equal(&vma
->vm_file
->f_path
,
1699 up_read(&mm
->mmap_sem
);
1704 /* set the new file, lockless */
1706 old_exe
= xchg(&mm
->exe_file
, exe
.file
);
1713 up_read(&mm
->mmap_sem
);
1719 * WARNING: we don't require any capability here so be very careful
1720 * in what is allowed for modification from userspace.
1722 static int validate_prctl_map(struct prctl_mm_map
*prctl_map
)
1724 unsigned long mmap_max_addr
= TASK_SIZE
;
1725 struct mm_struct
*mm
= current
->mm
;
1726 int error
= -EINVAL
, i
;
1728 static const unsigned char offsets
[] = {
1729 offsetof(struct prctl_mm_map
, start_code
),
1730 offsetof(struct prctl_mm_map
, end_code
),
1731 offsetof(struct prctl_mm_map
, start_data
),
1732 offsetof(struct prctl_mm_map
, end_data
),
1733 offsetof(struct prctl_mm_map
, start_brk
),
1734 offsetof(struct prctl_mm_map
, brk
),
1735 offsetof(struct prctl_mm_map
, start_stack
),
1736 offsetof(struct prctl_mm_map
, arg_start
),
1737 offsetof(struct prctl_mm_map
, arg_end
),
1738 offsetof(struct prctl_mm_map
, env_start
),
1739 offsetof(struct prctl_mm_map
, env_end
),
1743 * Make sure the members are not somewhere outside
1744 * of allowed address space.
1746 for (i
= 0; i
< ARRAY_SIZE(offsets
); i
++) {
1747 u64 val
= *(u64
*)((char *)prctl_map
+ offsets
[i
]);
1749 if ((unsigned long)val
>= mmap_max_addr
||
1750 (unsigned long)val
< mmap_min_addr
)
1755 * Make sure the pairs are ordered.
1757 #define __prctl_check_order(__m1, __op, __m2) \
1758 ((unsigned long)prctl_map->__m1 __op \
1759 (unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1760 error
= __prctl_check_order(start_code
, <, end_code
);
1761 error
|= __prctl_check_order(start_data
, <, end_data
);
1762 error
|= __prctl_check_order(start_brk
, <=, brk
);
1763 error
|= __prctl_check_order(arg_start
, <=, arg_end
);
1764 error
|= __prctl_check_order(env_start
, <=, env_end
);
1767 #undef __prctl_check_order
1772 * @brk should be after @end_data in traditional maps.
1774 if (prctl_map
->start_brk
<= prctl_map
->end_data
||
1775 prctl_map
->brk
<= prctl_map
->end_data
)
1779 * Neither we should allow to override limits if they set.
1781 if (check_data_rlimit(rlimit(RLIMIT_DATA
), prctl_map
->brk
,
1782 prctl_map
->start_brk
, prctl_map
->end_data
,
1783 prctl_map
->start_data
))
1787 * Someone is trying to cheat the auxv vector.
1789 if (prctl_map
->auxv_size
) {
1790 if (!prctl_map
->auxv
|| prctl_map
->auxv_size
> sizeof(mm
->saved_auxv
))
1795 * Finally, make sure the caller has the rights to
1796 * change /proc/pid/exe link: only local root should
1799 if (prctl_map
->exe_fd
!= (u32
)-1) {
1800 struct user_namespace
*ns
= current_user_ns();
1801 const struct cred
*cred
= current_cred();
1803 if (!uid_eq(cred
->uid
, make_kuid(ns
, 0)) ||
1804 !gid_eq(cred
->gid
, make_kgid(ns
, 0)))
1813 #ifdef CONFIG_CHECKPOINT_RESTORE
1814 static int prctl_set_mm_map(int opt
, const void __user
*addr
, unsigned long data_size
)
1816 struct prctl_mm_map prctl_map
= { .exe_fd
= (u32
)-1, };
1817 unsigned long user_auxv
[AT_VECTOR_SIZE
];
1818 struct mm_struct
*mm
= current
->mm
;
1821 BUILD_BUG_ON(sizeof(user_auxv
) != sizeof(mm
->saved_auxv
));
1822 BUILD_BUG_ON(sizeof(struct prctl_mm_map
) > 256);
1824 if (opt
== PR_SET_MM_MAP_SIZE
)
1825 return put_user((unsigned int)sizeof(prctl_map
),
1826 (unsigned int __user
*)addr
);
1828 if (data_size
!= sizeof(prctl_map
))
1831 if (copy_from_user(&prctl_map
, addr
, sizeof(prctl_map
)))
1834 error
= validate_prctl_map(&prctl_map
);
1838 if (prctl_map
.auxv_size
) {
1839 memset(user_auxv
, 0, sizeof(user_auxv
));
1840 if (copy_from_user(user_auxv
,
1841 (const void __user
*)prctl_map
.auxv
,
1842 prctl_map
.auxv_size
))
1845 /* Last entry must be AT_NULL as specification requires */
1846 user_auxv
[AT_VECTOR_SIZE
- 2] = AT_NULL
;
1847 user_auxv
[AT_VECTOR_SIZE
- 1] = AT_NULL
;
1850 if (prctl_map
.exe_fd
!= (u32
)-1) {
1851 error
= prctl_set_mm_exe_file(mm
, prctl_map
.exe_fd
);
1856 down_write(&mm
->mmap_sem
);
1859 * We don't validate if these members are pointing to
1860 * real present VMAs because application may have correspond
1861 * VMAs already unmapped and kernel uses these members for statistics
1862 * output in procfs mostly, except
1864 * - @start_brk/@brk which are used in do_brk but kernel lookups
1865 * for VMAs when updating these memvers so anything wrong written
1866 * here cause kernel to swear at userspace program but won't lead
1867 * to any problem in kernel itself
1870 mm
->start_code
= prctl_map
.start_code
;
1871 mm
->end_code
= prctl_map
.end_code
;
1872 mm
->start_data
= prctl_map
.start_data
;
1873 mm
->end_data
= prctl_map
.end_data
;
1874 mm
->start_brk
= prctl_map
.start_brk
;
1875 mm
->brk
= prctl_map
.brk
;
1876 mm
->start_stack
= prctl_map
.start_stack
;
1877 mm
->arg_start
= prctl_map
.arg_start
;
1878 mm
->arg_end
= prctl_map
.arg_end
;
1879 mm
->env_start
= prctl_map
.env_start
;
1880 mm
->env_end
= prctl_map
.env_end
;
1883 * Note this update of @saved_auxv is lockless thus
1884 * if someone reads this member in procfs while we're
1885 * updating -- it may get partly updated results. It's
1886 * known and acceptable trade off: we leave it as is to
1887 * not introduce additional locks here making the kernel
1890 if (prctl_map
.auxv_size
)
1891 memcpy(mm
->saved_auxv
, user_auxv
, sizeof(user_auxv
));
1893 up_write(&mm
->mmap_sem
);
1896 #endif /* CONFIG_CHECKPOINT_RESTORE */
1898 static int prctl_set_auxv(struct mm_struct
*mm
, unsigned long addr
,
1902 * This doesn't move the auxiliary vector itself since it's pinned to
1903 * mm_struct, but it permits filling the vector with new values. It's
1904 * up to the caller to provide sane values here, otherwise userspace
1905 * tools which use this vector might be unhappy.
1907 unsigned long user_auxv
[AT_VECTOR_SIZE
];
1909 if (len
> sizeof(user_auxv
))
1912 if (copy_from_user(user_auxv
, (const void __user
*)addr
, len
))
1915 /* Make sure the last entry is always AT_NULL */
1916 user_auxv
[AT_VECTOR_SIZE
- 2] = 0;
1917 user_auxv
[AT_VECTOR_SIZE
- 1] = 0;
1919 BUILD_BUG_ON(sizeof(user_auxv
) != sizeof(mm
->saved_auxv
));
1922 memcpy(mm
->saved_auxv
, user_auxv
, len
);
1923 task_unlock(current
);
1928 static int prctl_set_mm(int opt
, unsigned long addr
,
1929 unsigned long arg4
, unsigned long arg5
)
1931 struct mm_struct
*mm
= current
->mm
;
1932 struct prctl_mm_map prctl_map
;
1933 struct vm_area_struct
*vma
;
1936 if (arg5
|| (arg4
&& (opt
!= PR_SET_MM_AUXV
&&
1937 opt
!= PR_SET_MM_MAP
&&
1938 opt
!= PR_SET_MM_MAP_SIZE
)))
1941 #ifdef CONFIG_CHECKPOINT_RESTORE
1942 if (opt
== PR_SET_MM_MAP
|| opt
== PR_SET_MM_MAP_SIZE
)
1943 return prctl_set_mm_map(opt
, (const void __user
*)addr
, arg4
);
1946 if (!capable(CAP_SYS_RESOURCE
))
1949 if (opt
== PR_SET_MM_EXE_FILE
)
1950 return prctl_set_mm_exe_file(mm
, (unsigned int)addr
);
1952 if (opt
== PR_SET_MM_AUXV
)
1953 return prctl_set_auxv(mm
, addr
, arg4
);
1955 if (addr
>= TASK_SIZE
|| addr
< mmap_min_addr
)
1960 down_write(&mm
->mmap_sem
);
1961 vma
= find_vma(mm
, addr
);
1963 prctl_map
.start_code
= mm
->start_code
;
1964 prctl_map
.end_code
= mm
->end_code
;
1965 prctl_map
.start_data
= mm
->start_data
;
1966 prctl_map
.end_data
= mm
->end_data
;
1967 prctl_map
.start_brk
= mm
->start_brk
;
1968 prctl_map
.brk
= mm
->brk
;
1969 prctl_map
.start_stack
= mm
->start_stack
;
1970 prctl_map
.arg_start
= mm
->arg_start
;
1971 prctl_map
.arg_end
= mm
->arg_end
;
1972 prctl_map
.env_start
= mm
->env_start
;
1973 prctl_map
.env_end
= mm
->env_end
;
1974 prctl_map
.auxv
= NULL
;
1975 prctl_map
.auxv_size
= 0;
1976 prctl_map
.exe_fd
= -1;
1979 case PR_SET_MM_START_CODE
:
1980 prctl_map
.start_code
= addr
;
1982 case PR_SET_MM_END_CODE
:
1983 prctl_map
.end_code
= addr
;
1985 case PR_SET_MM_START_DATA
:
1986 prctl_map
.start_data
= addr
;
1988 case PR_SET_MM_END_DATA
:
1989 prctl_map
.end_data
= addr
;
1991 case PR_SET_MM_START_STACK
:
1992 prctl_map
.start_stack
= addr
;
1994 case PR_SET_MM_START_BRK
:
1995 prctl_map
.start_brk
= addr
;
1998 prctl_map
.brk
= addr
;
2000 case PR_SET_MM_ARG_START
:
2001 prctl_map
.arg_start
= addr
;
2003 case PR_SET_MM_ARG_END
:
2004 prctl_map
.arg_end
= addr
;
2006 case PR_SET_MM_ENV_START
:
2007 prctl_map
.env_start
= addr
;
2009 case PR_SET_MM_ENV_END
:
2010 prctl_map
.env_end
= addr
;
2016 error
= validate_prctl_map(&prctl_map
);
2022 * If command line arguments and environment
2023 * are placed somewhere else on stack, we can
2024 * set them up here, ARG_START/END to setup
2025 * command line argumets and ENV_START/END
2028 case PR_SET_MM_START_STACK
:
2029 case PR_SET_MM_ARG_START
:
2030 case PR_SET_MM_ARG_END
:
2031 case PR_SET_MM_ENV_START
:
2032 case PR_SET_MM_ENV_END
:
2039 mm
->start_code
= prctl_map
.start_code
;
2040 mm
->end_code
= prctl_map
.end_code
;
2041 mm
->start_data
= prctl_map
.start_data
;
2042 mm
->end_data
= prctl_map
.end_data
;
2043 mm
->start_brk
= prctl_map
.start_brk
;
2044 mm
->brk
= prctl_map
.brk
;
2045 mm
->start_stack
= prctl_map
.start_stack
;
2046 mm
->arg_start
= prctl_map
.arg_start
;
2047 mm
->arg_end
= prctl_map
.arg_end
;
2048 mm
->env_start
= prctl_map
.env_start
;
2049 mm
->env_end
= prctl_map
.env_end
;
2053 up_write(&mm
->mmap_sem
);
2057 #ifdef CONFIG_CHECKPOINT_RESTORE
2058 static int prctl_get_tid_address(struct task_struct
*me
, int __user
**tid_addr
)
2060 return put_user(me
->clear_child_tid
, tid_addr
);
2063 static int prctl_get_tid_address(struct task_struct
*me
, int __user
**tid_addr
)
2069 static int propagate_has_child_subreaper(struct task_struct
*p
, void *data
)
2072 * If task has has_child_subreaper - all its decendants
2073 * already have these flag too and new decendants will
2074 * inherit it on fork, skip them.
2076 * If we've found child_reaper - skip descendants in
2077 * it's subtree as they will never get out pidns.
2079 if (p
->signal
->has_child_subreaper
||
2080 is_child_reaper(task_pid(p
)))
2083 p
->signal
->has_child_subreaper
= 1;
2087 SYSCALL_DEFINE5(prctl
, int, option
, unsigned long, arg2
, unsigned long, arg3
,
2088 unsigned long, arg4
, unsigned long, arg5
)
2090 struct task_struct
*me
= current
;
2091 unsigned char comm
[sizeof(me
->comm
)];
2094 error
= security_task_prctl(option
, arg2
, arg3
, arg4
, arg5
);
2095 if (error
!= -ENOSYS
)
2100 case PR_SET_PDEATHSIG
:
2101 if (!valid_signal(arg2
)) {
2105 me
->pdeath_signal
= arg2
;
2107 case PR_GET_PDEATHSIG
:
2108 error
= put_user(me
->pdeath_signal
, (int __user
*)arg2
);
2110 case PR_GET_DUMPABLE
:
2111 error
= get_dumpable(me
->mm
);
2113 case PR_SET_DUMPABLE
:
2114 if (arg2
!= SUID_DUMP_DISABLE
&& arg2
!= SUID_DUMP_USER
) {
2118 set_dumpable(me
->mm
, arg2
);
2121 case PR_SET_UNALIGN
:
2122 error
= SET_UNALIGN_CTL(me
, arg2
);
2124 case PR_GET_UNALIGN
:
2125 error
= GET_UNALIGN_CTL(me
, arg2
);
2128 error
= SET_FPEMU_CTL(me
, arg2
);
2131 error
= GET_FPEMU_CTL(me
, arg2
);
2134 error
= SET_FPEXC_CTL(me
, arg2
);
2137 error
= GET_FPEXC_CTL(me
, arg2
);
2140 error
= PR_TIMING_STATISTICAL
;
2143 if (arg2
!= PR_TIMING_STATISTICAL
)
2147 comm
[sizeof(me
->comm
) - 1] = 0;
2148 if (strncpy_from_user(comm
, (char __user
*)arg2
,
2149 sizeof(me
->comm
) - 1) < 0)
2151 set_task_comm(me
, comm
);
2152 proc_comm_connector(me
);
2155 get_task_comm(comm
, me
);
2156 if (copy_to_user((char __user
*)arg2
, comm
, sizeof(comm
)))
2160 error
= GET_ENDIAN(me
, arg2
);
2163 error
= SET_ENDIAN(me
, arg2
);
2165 case PR_GET_SECCOMP
:
2166 error
= prctl_get_seccomp();
2168 case PR_SET_SECCOMP
:
2169 error
= prctl_set_seccomp(arg2
, (char __user
*)arg3
);
2172 error
= GET_TSC_CTL(arg2
);
2175 error
= SET_TSC_CTL(arg2
);
2177 case PR_TASK_PERF_EVENTS_DISABLE
:
2178 error
= perf_event_task_disable();
2180 case PR_TASK_PERF_EVENTS_ENABLE
:
2181 error
= perf_event_task_enable();
2183 case PR_GET_TIMERSLACK
:
2184 if (current
->timer_slack_ns
> ULONG_MAX
)
2187 error
= current
->timer_slack_ns
;
2189 case PR_SET_TIMERSLACK
:
2191 current
->timer_slack_ns
=
2192 current
->default_timer_slack_ns
;
2194 current
->timer_slack_ns
= arg2
;
2200 case PR_MCE_KILL_CLEAR
:
2203 current
->flags
&= ~PF_MCE_PROCESS
;
2205 case PR_MCE_KILL_SET
:
2206 current
->flags
|= PF_MCE_PROCESS
;
2207 if (arg3
== PR_MCE_KILL_EARLY
)
2208 current
->flags
|= PF_MCE_EARLY
;
2209 else if (arg3
== PR_MCE_KILL_LATE
)
2210 current
->flags
&= ~PF_MCE_EARLY
;
2211 else if (arg3
== PR_MCE_KILL_DEFAULT
)
2213 ~(PF_MCE_EARLY
|PF_MCE_PROCESS
);
2221 case PR_MCE_KILL_GET
:
2222 if (arg2
| arg3
| arg4
| arg5
)
2224 if (current
->flags
& PF_MCE_PROCESS
)
2225 error
= (current
->flags
& PF_MCE_EARLY
) ?
2226 PR_MCE_KILL_EARLY
: PR_MCE_KILL_LATE
;
2228 error
= PR_MCE_KILL_DEFAULT
;
2231 error
= prctl_set_mm(arg2
, arg3
, arg4
, arg5
);
2233 case PR_GET_TID_ADDRESS
:
2234 error
= prctl_get_tid_address(me
, (int __user
**)arg2
);
2236 case PR_SET_CHILD_SUBREAPER
:
2237 me
->signal
->is_child_subreaper
= !!arg2
;
2241 walk_process_tree(me
, propagate_has_child_subreaper
, NULL
);
2243 case PR_GET_CHILD_SUBREAPER
:
2244 error
= put_user(me
->signal
->is_child_subreaper
,
2245 (int __user
*)arg2
);
2247 case PR_SET_NO_NEW_PRIVS
:
2248 if (arg2
!= 1 || arg3
|| arg4
|| arg5
)
2251 task_set_no_new_privs(current
);
2253 case PR_GET_NO_NEW_PRIVS
:
2254 if (arg2
|| arg3
|| arg4
|| arg5
)
2256 return task_no_new_privs(current
) ? 1 : 0;
2257 case PR_GET_THP_DISABLE
:
2258 if (arg2
|| arg3
|| arg4
|| arg5
)
2260 error
= !!(me
->mm
->def_flags
& VM_NOHUGEPAGE
);
2262 case PR_SET_THP_DISABLE
:
2263 if (arg3
|| arg4
|| arg5
)
2265 if (down_write_killable(&me
->mm
->mmap_sem
))
2268 me
->mm
->def_flags
|= VM_NOHUGEPAGE
;
2270 me
->mm
->def_flags
&= ~VM_NOHUGEPAGE
;
2271 up_write(&me
->mm
->mmap_sem
);
2273 case PR_MPX_ENABLE_MANAGEMENT
:
2274 if (arg2
|| arg3
|| arg4
|| arg5
)
2276 error
= MPX_ENABLE_MANAGEMENT();
2278 case PR_MPX_DISABLE_MANAGEMENT
:
2279 if (arg2
|| arg3
|| arg4
|| arg5
)
2281 error
= MPX_DISABLE_MANAGEMENT();
2283 case PR_SET_FP_MODE
:
2284 error
= SET_FP_MODE(me
, arg2
);
2286 case PR_GET_FP_MODE
:
2287 error
= GET_FP_MODE(me
);
2296 SYSCALL_DEFINE3(getcpu
, unsigned __user
*, cpup
, unsigned __user
*, nodep
,
2297 struct getcpu_cache __user
*, unused
)
2300 int cpu
= raw_smp_processor_id();
2303 err
|= put_user(cpu
, cpup
);
2305 err
|= put_user(cpu_to_node(cpu
), nodep
);
2306 return err
? -EFAULT
: 0;
2310 * do_sysinfo - fill in sysinfo struct
2311 * @info: pointer to buffer to fill
2313 static int do_sysinfo(struct sysinfo
*info
)
2315 unsigned long mem_total
, sav_total
;
2316 unsigned int mem_unit
, bitcount
;
2319 memset(info
, 0, sizeof(struct sysinfo
));
2321 get_monotonic_boottime(&tp
);
2322 info
->uptime
= tp
.tv_sec
+ (tp
.tv_nsec
? 1 : 0);
2324 get_avenrun(info
->loads
, 0, SI_LOAD_SHIFT
- FSHIFT
);
2326 info
->procs
= nr_threads
;
2332 * If the sum of all the available memory (i.e. ram + swap)
2333 * is less than can be stored in a 32 bit unsigned long then
2334 * we can be binary compatible with 2.2.x kernels. If not,
2335 * well, in that case 2.2.x was broken anyways...
2337 * -Erik Andersen <andersee@debian.org>
2340 mem_total
= info
->totalram
+ info
->totalswap
;
2341 if (mem_total
< info
->totalram
|| mem_total
< info
->totalswap
)
2344 mem_unit
= info
->mem_unit
;
2345 while (mem_unit
> 1) {
2348 sav_total
= mem_total
;
2350 if (mem_total
< sav_total
)
2355 * If mem_total did not overflow, multiply all memory values by
2356 * info->mem_unit and set it to 1. This leaves things compatible
2357 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2362 info
->totalram
<<= bitcount
;
2363 info
->freeram
<<= bitcount
;
2364 info
->sharedram
<<= bitcount
;
2365 info
->bufferram
<<= bitcount
;
2366 info
->totalswap
<<= bitcount
;
2367 info
->freeswap
<<= bitcount
;
2368 info
->totalhigh
<<= bitcount
;
2369 info
->freehigh
<<= bitcount
;
2375 SYSCALL_DEFINE1(sysinfo
, struct sysinfo __user
*, info
)
2381 if (copy_to_user(info
, &val
, sizeof(struct sysinfo
)))
2387 #ifdef CONFIG_COMPAT
2388 struct compat_sysinfo
{
2402 char _f
[20-2*sizeof(u32
)-sizeof(int)];
2405 COMPAT_SYSCALL_DEFINE1(sysinfo
, struct compat_sysinfo __user
*, info
)
2411 /* Check to see if any memory value is too large for 32-bit and scale
2414 if (upper_32_bits(s
.totalram
) || upper_32_bits(s
.totalswap
)) {
2417 while (s
.mem_unit
< PAGE_SIZE
) {
2422 s
.totalram
>>= bitcount
;
2423 s
.freeram
>>= bitcount
;
2424 s
.sharedram
>>= bitcount
;
2425 s
.bufferram
>>= bitcount
;
2426 s
.totalswap
>>= bitcount
;
2427 s
.freeswap
>>= bitcount
;
2428 s
.totalhigh
>>= bitcount
;
2429 s
.freehigh
>>= bitcount
;
2432 if (!access_ok(VERIFY_WRITE
, info
, sizeof(struct compat_sysinfo
)) ||
2433 __put_user(s
.uptime
, &info
->uptime
) ||
2434 __put_user(s
.loads
[0], &info
->loads
[0]) ||
2435 __put_user(s
.loads
[1], &info
->loads
[1]) ||
2436 __put_user(s
.loads
[2], &info
->loads
[2]) ||
2437 __put_user(s
.totalram
, &info
->totalram
) ||
2438 __put_user(s
.freeram
, &info
->freeram
) ||
2439 __put_user(s
.sharedram
, &info
->sharedram
) ||
2440 __put_user(s
.bufferram
, &info
->bufferram
) ||
2441 __put_user(s
.totalswap
, &info
->totalswap
) ||
2442 __put_user(s
.freeswap
, &info
->freeswap
) ||
2443 __put_user(s
.procs
, &info
->procs
) ||
2444 __put_user(s
.totalhigh
, &info
->totalhigh
) ||
2445 __put_user(s
.freehigh
, &info
->freehigh
) ||
2446 __put_user(s
.mem_unit
, &info
->mem_unit
))
2451 #endif /* CONFIG_COMPAT */