1 // SPDX-License-Identifier: GPL-2.0
5 * Copyright (C) 1991, 1992 Linus Torvalds
8 #include <linux/export.h>
10 #include <linux/utsname.h>
11 #include <linux/mman.h>
12 #include <linux/reboot.h>
13 #include <linux/prctl.h>
14 #include <linux/highuid.h>
16 #include <linux/kmod.h>
17 #include <linux/perf_event.h>
18 #include <linux/resource.h>
19 #include <linux/kernel.h>
20 #include <linux/workqueue.h>
21 #include <linux/capability.h>
22 #include <linux/device.h>
23 #include <linux/key.h>
24 #include <linux/times.h>
25 #include <linux/posix-timers.h>
26 #include <linux/security.h>
27 #include <linux/dcookies.h>
28 #include <linux/suspend.h>
29 #include <linux/tty.h>
30 #include <linux/signal.h>
31 #include <linux/cn_proc.h>
32 #include <linux/getcpu.h>
33 #include <linux/task_io_accounting_ops.h>
34 #include <linux/seccomp.h>
35 #include <linux/cpu.h>
36 #include <linux/personality.h>
37 #include <linux/ptrace.h>
38 #include <linux/fs_struct.h>
39 #include <linux/file.h>
40 #include <linux/mount.h>
41 #include <linux/gfp.h>
42 #include <linux/syscore_ops.h>
43 #include <linux/version.h>
44 #include <linux/ctype.h>
45 #include <linux/syscall_user_dispatch.h>
47 #include <linux/compat.h>
48 #include <linux/syscalls.h>
49 #include <linux/kprobes.h>
50 #include <linux/user_namespace.h>
51 #include <linux/time_namespace.h>
52 #include <linux/binfmts.h>
54 #include <linux/sched.h>
55 #include <linux/sched/autogroup.h>
56 #include <linux/sched/loadavg.h>
57 #include <linux/sched/stat.h>
58 #include <linux/sched/mm.h>
59 #include <linux/sched/coredump.h>
60 #include <linux/sched/task.h>
61 #include <linux/sched/cputime.h>
62 #include <linux/rcupdate.h>
63 #include <linux/uidgid.h>
64 #include <linux/cred.h>
66 #include <linux/nospec.h>
68 #include <linux/kmsg_dump.h>
69 /* Move somewhere else to avoid recompiling? */
70 #include <generated/utsrelease.h>
72 #include <linux/uaccess.h>
74 #include <asm/unistd.h>
78 #ifndef SET_UNALIGN_CTL
79 # define SET_UNALIGN_CTL(a, b) (-EINVAL)
81 #ifndef GET_UNALIGN_CTL
82 # define GET_UNALIGN_CTL(a, b) (-EINVAL)
85 # define SET_FPEMU_CTL(a, b) (-EINVAL)
88 # define GET_FPEMU_CTL(a, b) (-EINVAL)
91 # define SET_FPEXC_CTL(a, b) (-EINVAL)
94 # define GET_FPEXC_CTL(a, b) (-EINVAL)
97 # define GET_ENDIAN(a, b) (-EINVAL)
100 # define SET_ENDIAN(a, b) (-EINVAL)
103 # define GET_TSC_CTL(a) (-EINVAL)
106 # define SET_TSC_CTL(a) (-EINVAL)
109 # define GET_FP_MODE(a) (-EINVAL)
112 # define SET_FP_MODE(a,b) (-EINVAL)
115 # define SVE_SET_VL(a) (-EINVAL)
118 # define SVE_GET_VL() (-EINVAL)
120 #ifndef PAC_RESET_KEYS
121 # define PAC_RESET_KEYS(a, b) (-EINVAL)
123 #ifndef SET_TAGGED_ADDR_CTRL
124 # define SET_TAGGED_ADDR_CTRL(a) (-EINVAL)
126 #ifndef GET_TAGGED_ADDR_CTRL
127 # define GET_TAGGED_ADDR_CTRL() (-EINVAL)
131 * this is where the system-wide overflow UID and GID are defined, for
132 * architectures that now have 32-bit UID/GID but didn't in the past
135 int overflowuid
= DEFAULT_OVERFLOWUID
;
136 int overflowgid
= DEFAULT_OVERFLOWGID
;
138 EXPORT_SYMBOL(overflowuid
);
139 EXPORT_SYMBOL(overflowgid
);
142 * the same as above, but for filesystems which can only store a 16-bit
143 * UID and GID. as such, this is needed on all architectures
146 int fs_overflowuid
= DEFAULT_FS_OVERFLOWUID
;
147 int fs_overflowgid
= DEFAULT_FS_OVERFLOWGID
;
149 EXPORT_SYMBOL(fs_overflowuid
);
150 EXPORT_SYMBOL(fs_overflowgid
);
153 * Returns true if current's euid is same as p's uid or euid,
154 * or has CAP_SYS_NICE to p's user_ns.
156 * Called with rcu_read_lock, creds are safe
158 static bool set_one_prio_perm(struct task_struct
*p
)
160 const struct cred
*cred
= current_cred(), *pcred
= __task_cred(p
);
162 if (uid_eq(pcred
->uid
, cred
->euid
) ||
163 uid_eq(pcred
->euid
, cred
->euid
))
165 if (ns_capable(pcred
->user_ns
, CAP_SYS_NICE
))
171 * set the priority of a task
172 * - the caller must hold the RCU read lock
174 static int set_one_prio(struct task_struct
*p
, int niceval
, int error
)
178 if (!set_one_prio_perm(p
)) {
182 if (niceval
< task_nice(p
) && !can_nice(p
, niceval
)) {
186 no_nice
= security_task_setnice(p
, niceval
);
193 set_user_nice(p
, niceval
);
198 SYSCALL_DEFINE3(setpriority
, int, which
, int, who
, int, niceval
)
200 struct task_struct
*g
, *p
;
201 struct user_struct
*user
;
202 const struct cred
*cred
= current_cred();
207 if (which
> PRIO_USER
|| which
< PRIO_PROCESS
)
210 /* normalize: avoid signed division (rounding problems) */
212 if (niceval
< MIN_NICE
)
214 if (niceval
> MAX_NICE
)
218 read_lock(&tasklist_lock
);
222 p
= find_task_by_vpid(who
);
226 error
= set_one_prio(p
, niceval
, error
);
230 pgrp
= find_vpid(who
);
232 pgrp
= task_pgrp(current
);
233 do_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
) {
234 error
= set_one_prio(p
, niceval
, error
);
235 } while_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
);
238 uid
= make_kuid(cred
->user_ns
, who
);
242 else if (!uid_eq(uid
, cred
->uid
)) {
243 user
= find_user(uid
);
245 goto out_unlock
; /* No processes for this user */
247 do_each_thread(g
, p
) {
248 if (uid_eq(task_uid(p
), uid
) && task_pid_vnr(p
))
249 error
= set_one_prio(p
, niceval
, error
);
250 } while_each_thread(g
, p
);
251 if (!uid_eq(uid
, cred
->uid
))
252 free_uid(user
); /* For find_user() */
256 read_unlock(&tasklist_lock
);
263 * Ugh. To avoid negative return values, "getpriority()" will
264 * not return the normal nice-value, but a negated value that
265 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
266 * to stay compatible.
268 SYSCALL_DEFINE2(getpriority
, int, which
, int, who
)
270 struct task_struct
*g
, *p
;
271 struct user_struct
*user
;
272 const struct cred
*cred
= current_cred();
273 long niceval
, retval
= -ESRCH
;
277 if (which
> PRIO_USER
|| which
< PRIO_PROCESS
)
281 read_lock(&tasklist_lock
);
285 p
= find_task_by_vpid(who
);
289 niceval
= nice_to_rlimit(task_nice(p
));
290 if (niceval
> retval
)
296 pgrp
= find_vpid(who
);
298 pgrp
= task_pgrp(current
);
299 do_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
) {
300 niceval
= nice_to_rlimit(task_nice(p
));
301 if (niceval
> retval
)
303 } while_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
);
306 uid
= make_kuid(cred
->user_ns
, who
);
310 else if (!uid_eq(uid
, cred
->uid
)) {
311 user
= find_user(uid
);
313 goto out_unlock
; /* No processes for this user */
315 do_each_thread(g
, p
) {
316 if (uid_eq(task_uid(p
), uid
) && task_pid_vnr(p
)) {
317 niceval
= nice_to_rlimit(task_nice(p
));
318 if (niceval
> retval
)
321 } while_each_thread(g
, p
);
322 if (!uid_eq(uid
, cred
->uid
))
323 free_uid(user
); /* for find_user() */
327 read_unlock(&tasklist_lock
);
334 * Unprivileged users may change the real gid to the effective gid
335 * or vice versa. (BSD-style)
337 * If you set the real gid at all, or set the effective gid to a value not
338 * equal to the real gid, then the saved gid is set to the new effective gid.
340 * This makes it possible for a setgid program to completely drop its
341 * privileges, which is often a useful assertion to make when you are doing
342 * a security audit over a program.
344 * The general idea is that a program which uses just setregid() will be
345 * 100% compatible with BSD. A program which uses just setgid() will be
346 * 100% compatible with POSIX with saved IDs.
348 * SMP: There are not races, the GIDs are checked only by filesystem
349 * operations (as far as semantic preservation is concerned).
351 #ifdef CONFIG_MULTIUSER
352 long __sys_setregid(gid_t rgid
, gid_t egid
)
354 struct user_namespace
*ns
= current_user_ns();
355 const struct cred
*old
;
360 krgid
= make_kgid(ns
, rgid
);
361 kegid
= make_kgid(ns
, egid
);
363 if ((rgid
!= (gid_t
) -1) && !gid_valid(krgid
))
365 if ((egid
!= (gid_t
) -1) && !gid_valid(kegid
))
368 new = prepare_creds();
371 old
= current_cred();
374 if (rgid
!= (gid_t
) -1) {
375 if (gid_eq(old
->gid
, krgid
) ||
376 gid_eq(old
->egid
, krgid
) ||
377 ns_capable_setid(old
->user_ns
, CAP_SETGID
))
382 if (egid
!= (gid_t
) -1) {
383 if (gid_eq(old
->gid
, kegid
) ||
384 gid_eq(old
->egid
, kegid
) ||
385 gid_eq(old
->sgid
, kegid
) ||
386 ns_capable_setid(old
->user_ns
, CAP_SETGID
))
392 if (rgid
!= (gid_t
) -1 ||
393 (egid
!= (gid_t
) -1 && !gid_eq(kegid
, old
->gid
)))
394 new->sgid
= new->egid
;
395 new->fsgid
= new->egid
;
397 retval
= security_task_fix_setgid(new, old
, LSM_SETID_RE
);
401 return commit_creds(new);
408 SYSCALL_DEFINE2(setregid
, gid_t
, rgid
, gid_t
, egid
)
410 return __sys_setregid(rgid
, egid
);
414 * setgid() is implemented like SysV w/ SAVED_IDS
416 * SMP: Same implicit races as above.
418 long __sys_setgid(gid_t gid
)
420 struct user_namespace
*ns
= current_user_ns();
421 const struct cred
*old
;
426 kgid
= make_kgid(ns
, gid
);
427 if (!gid_valid(kgid
))
430 new = prepare_creds();
433 old
= current_cred();
436 if (ns_capable_setid(old
->user_ns
, CAP_SETGID
))
437 new->gid
= new->egid
= new->sgid
= new->fsgid
= kgid
;
438 else if (gid_eq(kgid
, old
->gid
) || gid_eq(kgid
, old
->sgid
))
439 new->egid
= new->fsgid
= kgid
;
443 retval
= security_task_fix_setgid(new, old
, LSM_SETID_ID
);
447 return commit_creds(new);
454 SYSCALL_DEFINE1(setgid
, gid_t
, gid
)
456 return __sys_setgid(gid
);
460 * change the user struct in a credentials set to match the new UID
462 static int set_user(struct cred
*new)
464 struct user_struct
*new_user
;
466 new_user
= alloc_uid(new->uid
);
471 * We don't fail in case of NPROC limit excess here because too many
472 * poorly written programs don't check set*uid() return code, assuming
473 * it never fails if called by root. We may still enforce NPROC limit
474 * for programs doing set*uid()+execve() by harmlessly deferring the
475 * failure to the execve() stage.
477 if (atomic_read(&new_user
->processes
) >= rlimit(RLIMIT_NPROC
) &&
478 new_user
!= INIT_USER
)
479 current
->flags
|= PF_NPROC_EXCEEDED
;
481 current
->flags
&= ~PF_NPROC_EXCEEDED
;
484 new->user
= new_user
;
489 * Unprivileged users may change the real uid to the effective uid
490 * or vice versa. (BSD-style)
492 * If you set the real uid at all, or set the effective uid to a value not
493 * equal to the real uid, then the saved uid is set to the new effective uid.
495 * This makes it possible for a setuid program to completely drop its
496 * privileges, which is often a useful assertion to make when you are doing
497 * a security audit over a program.
499 * The general idea is that a program which uses just setreuid() will be
500 * 100% compatible with BSD. A program which uses just setuid() will be
501 * 100% compatible with POSIX with saved IDs.
503 long __sys_setreuid(uid_t ruid
, uid_t euid
)
505 struct user_namespace
*ns
= current_user_ns();
506 const struct cred
*old
;
511 kruid
= make_kuid(ns
, ruid
);
512 keuid
= make_kuid(ns
, euid
);
514 if ((ruid
!= (uid_t
) -1) && !uid_valid(kruid
))
516 if ((euid
!= (uid_t
) -1) && !uid_valid(keuid
))
519 new = prepare_creds();
522 old
= current_cred();
525 if (ruid
!= (uid_t
) -1) {
527 if (!uid_eq(old
->uid
, kruid
) &&
528 !uid_eq(old
->euid
, kruid
) &&
529 !ns_capable_setid(old
->user_ns
, CAP_SETUID
))
533 if (euid
!= (uid_t
) -1) {
535 if (!uid_eq(old
->uid
, keuid
) &&
536 !uid_eq(old
->euid
, keuid
) &&
537 !uid_eq(old
->suid
, keuid
) &&
538 !ns_capable_setid(old
->user_ns
, CAP_SETUID
))
542 if (!uid_eq(new->uid
, old
->uid
)) {
543 retval
= set_user(new);
547 if (ruid
!= (uid_t
) -1 ||
548 (euid
!= (uid_t
) -1 && !uid_eq(keuid
, old
->uid
)))
549 new->suid
= new->euid
;
550 new->fsuid
= new->euid
;
552 retval
= security_task_fix_setuid(new, old
, LSM_SETID_RE
);
556 return commit_creds(new);
563 SYSCALL_DEFINE2(setreuid
, uid_t
, ruid
, uid_t
, euid
)
565 return __sys_setreuid(ruid
, euid
);
569 * setuid() is implemented like SysV with SAVED_IDS
571 * Note that SAVED_ID's is deficient in that a setuid root program
572 * like sendmail, for example, cannot set its uid to be a normal
573 * user and then switch back, because if you're root, setuid() sets
574 * the saved uid too. If you don't like this, blame the bright people
575 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
576 * will allow a root program to temporarily drop privileges and be able to
577 * regain them by swapping the real and effective uid.
579 long __sys_setuid(uid_t uid
)
581 struct user_namespace
*ns
= current_user_ns();
582 const struct cred
*old
;
587 kuid
= make_kuid(ns
, uid
);
588 if (!uid_valid(kuid
))
591 new = prepare_creds();
594 old
= current_cred();
597 if (ns_capable_setid(old
->user_ns
, CAP_SETUID
)) {
598 new->suid
= new->uid
= kuid
;
599 if (!uid_eq(kuid
, old
->uid
)) {
600 retval
= set_user(new);
604 } else if (!uid_eq(kuid
, old
->uid
) && !uid_eq(kuid
, new->suid
)) {
608 new->fsuid
= new->euid
= kuid
;
610 retval
= security_task_fix_setuid(new, old
, LSM_SETID_ID
);
614 return commit_creds(new);
621 SYSCALL_DEFINE1(setuid
, uid_t
, uid
)
623 return __sys_setuid(uid
);
628 * This function implements a generic ability to update ruid, euid,
629 * and suid. This allows you to implement the 4.4 compatible seteuid().
631 long __sys_setresuid(uid_t ruid
, uid_t euid
, uid_t suid
)
633 struct user_namespace
*ns
= current_user_ns();
634 const struct cred
*old
;
637 kuid_t kruid
, keuid
, ksuid
;
639 kruid
= make_kuid(ns
, ruid
);
640 keuid
= make_kuid(ns
, euid
);
641 ksuid
= make_kuid(ns
, suid
);
643 if ((ruid
!= (uid_t
) -1) && !uid_valid(kruid
))
646 if ((euid
!= (uid_t
) -1) && !uid_valid(keuid
))
649 if ((suid
!= (uid_t
) -1) && !uid_valid(ksuid
))
652 new = prepare_creds();
656 old
= current_cred();
659 if (!ns_capable_setid(old
->user_ns
, CAP_SETUID
)) {
660 if (ruid
!= (uid_t
) -1 && !uid_eq(kruid
, old
->uid
) &&
661 !uid_eq(kruid
, old
->euid
) && !uid_eq(kruid
, old
->suid
))
663 if (euid
!= (uid_t
) -1 && !uid_eq(keuid
, old
->uid
) &&
664 !uid_eq(keuid
, old
->euid
) && !uid_eq(keuid
, old
->suid
))
666 if (suid
!= (uid_t
) -1 && !uid_eq(ksuid
, old
->uid
) &&
667 !uid_eq(ksuid
, old
->euid
) && !uid_eq(ksuid
, old
->suid
))
671 if (ruid
!= (uid_t
) -1) {
673 if (!uid_eq(kruid
, old
->uid
)) {
674 retval
= set_user(new);
679 if (euid
!= (uid_t
) -1)
681 if (suid
!= (uid_t
) -1)
683 new->fsuid
= new->euid
;
685 retval
= security_task_fix_setuid(new, old
, LSM_SETID_RES
);
689 return commit_creds(new);
696 SYSCALL_DEFINE3(setresuid
, uid_t
, ruid
, uid_t
, euid
, uid_t
, suid
)
698 return __sys_setresuid(ruid
, euid
, suid
);
701 SYSCALL_DEFINE3(getresuid
, uid_t __user
*, ruidp
, uid_t __user
*, euidp
, uid_t __user
*, suidp
)
703 const struct cred
*cred
= current_cred();
705 uid_t ruid
, euid
, suid
;
707 ruid
= from_kuid_munged(cred
->user_ns
, cred
->uid
);
708 euid
= from_kuid_munged(cred
->user_ns
, cred
->euid
);
709 suid
= from_kuid_munged(cred
->user_ns
, cred
->suid
);
711 retval
= put_user(ruid
, ruidp
);
713 retval
= put_user(euid
, euidp
);
715 return put_user(suid
, suidp
);
721 * Same as above, but for rgid, egid, sgid.
723 long __sys_setresgid(gid_t rgid
, gid_t egid
, gid_t sgid
)
725 struct user_namespace
*ns
= current_user_ns();
726 const struct cred
*old
;
729 kgid_t krgid
, kegid
, ksgid
;
731 krgid
= make_kgid(ns
, rgid
);
732 kegid
= make_kgid(ns
, egid
);
733 ksgid
= make_kgid(ns
, sgid
);
735 if ((rgid
!= (gid_t
) -1) && !gid_valid(krgid
))
737 if ((egid
!= (gid_t
) -1) && !gid_valid(kegid
))
739 if ((sgid
!= (gid_t
) -1) && !gid_valid(ksgid
))
742 new = prepare_creds();
745 old
= current_cred();
748 if (!ns_capable_setid(old
->user_ns
, CAP_SETGID
)) {
749 if (rgid
!= (gid_t
) -1 && !gid_eq(krgid
, old
->gid
) &&
750 !gid_eq(krgid
, old
->egid
) && !gid_eq(krgid
, old
->sgid
))
752 if (egid
!= (gid_t
) -1 && !gid_eq(kegid
, old
->gid
) &&
753 !gid_eq(kegid
, old
->egid
) && !gid_eq(kegid
, old
->sgid
))
755 if (sgid
!= (gid_t
) -1 && !gid_eq(ksgid
, old
->gid
) &&
756 !gid_eq(ksgid
, old
->egid
) && !gid_eq(ksgid
, old
->sgid
))
760 if (rgid
!= (gid_t
) -1)
762 if (egid
!= (gid_t
) -1)
764 if (sgid
!= (gid_t
) -1)
766 new->fsgid
= new->egid
;
768 retval
= security_task_fix_setgid(new, old
, LSM_SETID_RES
);
772 return commit_creds(new);
779 SYSCALL_DEFINE3(setresgid
, gid_t
, rgid
, gid_t
, egid
, gid_t
, sgid
)
781 return __sys_setresgid(rgid
, egid
, sgid
);
784 SYSCALL_DEFINE3(getresgid
, gid_t __user
*, rgidp
, gid_t __user
*, egidp
, gid_t __user
*, sgidp
)
786 const struct cred
*cred
= current_cred();
788 gid_t rgid
, egid
, sgid
;
790 rgid
= from_kgid_munged(cred
->user_ns
, cred
->gid
);
791 egid
= from_kgid_munged(cred
->user_ns
, cred
->egid
);
792 sgid
= from_kgid_munged(cred
->user_ns
, cred
->sgid
);
794 retval
= put_user(rgid
, rgidp
);
796 retval
= put_user(egid
, egidp
);
798 retval
= put_user(sgid
, sgidp
);
806 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
807 * is used for "access()" and for the NFS daemon (letting nfsd stay at
808 * whatever uid it wants to). It normally shadows "euid", except when
809 * explicitly set by setfsuid() or for access..
811 long __sys_setfsuid(uid_t uid
)
813 const struct cred
*old
;
818 old
= current_cred();
819 old_fsuid
= from_kuid_munged(old
->user_ns
, old
->fsuid
);
821 kuid
= make_kuid(old
->user_ns
, uid
);
822 if (!uid_valid(kuid
))
825 new = prepare_creds();
829 if (uid_eq(kuid
, old
->uid
) || uid_eq(kuid
, old
->euid
) ||
830 uid_eq(kuid
, old
->suid
) || uid_eq(kuid
, old
->fsuid
) ||
831 ns_capable_setid(old
->user_ns
, CAP_SETUID
)) {
832 if (!uid_eq(kuid
, old
->fsuid
)) {
834 if (security_task_fix_setuid(new, old
, LSM_SETID_FS
) == 0)
847 SYSCALL_DEFINE1(setfsuid
, uid_t
, uid
)
849 return __sys_setfsuid(uid
);
853 * Samma på svenska..
855 long __sys_setfsgid(gid_t gid
)
857 const struct cred
*old
;
862 old
= current_cred();
863 old_fsgid
= from_kgid_munged(old
->user_ns
, old
->fsgid
);
865 kgid
= make_kgid(old
->user_ns
, gid
);
866 if (!gid_valid(kgid
))
869 new = prepare_creds();
873 if (gid_eq(kgid
, old
->gid
) || gid_eq(kgid
, old
->egid
) ||
874 gid_eq(kgid
, old
->sgid
) || gid_eq(kgid
, old
->fsgid
) ||
875 ns_capable_setid(old
->user_ns
, CAP_SETGID
)) {
876 if (!gid_eq(kgid
, old
->fsgid
)) {
878 if (security_task_fix_setgid(new,old
,LSM_SETID_FS
) == 0)
891 SYSCALL_DEFINE1(setfsgid
, gid_t
, gid
)
893 return __sys_setfsgid(gid
);
895 #endif /* CONFIG_MULTIUSER */
898 * sys_getpid - return the thread group id of the current process
900 * Note, despite the name, this returns the tgid not the pid. The tgid and
901 * the pid are identical unless CLONE_THREAD was specified on clone() in
902 * which case the tgid is the same in all threads of the same group.
904 * This is SMP safe as current->tgid does not change.
906 SYSCALL_DEFINE0(getpid
)
908 return task_tgid_vnr(current
);
911 /* Thread ID - the internal kernel "pid" */
912 SYSCALL_DEFINE0(gettid
)
914 return task_pid_vnr(current
);
918 * Accessing ->real_parent is not SMP-safe, it could
919 * change from under us. However, we can use a stale
920 * value of ->real_parent under rcu_read_lock(), see
921 * release_task()->call_rcu(delayed_put_task_struct).
923 SYSCALL_DEFINE0(getppid
)
928 pid
= task_tgid_vnr(rcu_dereference(current
->real_parent
));
934 SYSCALL_DEFINE0(getuid
)
936 /* Only we change this so SMP safe */
937 return from_kuid_munged(current_user_ns(), current_uid());
940 SYSCALL_DEFINE0(geteuid
)
942 /* Only we change this so SMP safe */
943 return from_kuid_munged(current_user_ns(), current_euid());
946 SYSCALL_DEFINE0(getgid
)
948 /* Only we change this so SMP safe */
949 return from_kgid_munged(current_user_ns(), current_gid());
952 SYSCALL_DEFINE0(getegid
)
954 /* Only we change this so SMP safe */
955 return from_kgid_munged(current_user_ns(), current_egid());
958 static void do_sys_times(struct tms
*tms
)
960 u64 tgutime
, tgstime
, cutime
, cstime
;
962 thread_group_cputime_adjusted(current
, &tgutime
, &tgstime
);
963 cutime
= current
->signal
->cutime
;
964 cstime
= current
->signal
->cstime
;
965 tms
->tms_utime
= nsec_to_clock_t(tgutime
);
966 tms
->tms_stime
= nsec_to_clock_t(tgstime
);
967 tms
->tms_cutime
= nsec_to_clock_t(cutime
);
968 tms
->tms_cstime
= nsec_to_clock_t(cstime
);
971 SYSCALL_DEFINE1(times
, struct tms __user
*, tbuf
)
977 if (copy_to_user(tbuf
, &tmp
, sizeof(struct tms
)))
980 force_successful_syscall_return();
981 return (long) jiffies_64_to_clock_t(get_jiffies_64());
985 static compat_clock_t
clock_t_to_compat_clock_t(clock_t x
)
987 return compat_jiffies_to_clock_t(clock_t_to_jiffies(x
));
990 COMPAT_SYSCALL_DEFINE1(times
, struct compat_tms __user
*, tbuf
)
994 struct compat_tms tmp
;
997 /* Convert our struct tms to the compat version. */
998 tmp
.tms_utime
= clock_t_to_compat_clock_t(tms
.tms_utime
);
999 tmp
.tms_stime
= clock_t_to_compat_clock_t(tms
.tms_stime
);
1000 tmp
.tms_cutime
= clock_t_to_compat_clock_t(tms
.tms_cutime
);
1001 tmp
.tms_cstime
= clock_t_to_compat_clock_t(tms
.tms_cstime
);
1002 if (copy_to_user(tbuf
, &tmp
, sizeof(tmp
)))
1005 force_successful_syscall_return();
1006 return compat_jiffies_to_clock_t(jiffies
);
1011 * This needs some heavy checking ...
1012 * I just haven't the stomach for it. I also don't fully
1013 * understand sessions/pgrp etc. Let somebody who does explain it.
1015 * OK, I think I have the protection semantics right.... this is really
1016 * only important on a multi-user system anyway, to make sure one user
1017 * can't send a signal to a process owned by another. -TYT, 12/12/91
1019 * !PF_FORKNOEXEC check to conform completely to POSIX.
1021 SYSCALL_DEFINE2(setpgid
, pid_t
, pid
, pid_t
, pgid
)
1023 struct task_struct
*p
;
1024 struct task_struct
*group_leader
= current
->group_leader
;
1029 pid
= task_pid_vnr(group_leader
);
1036 /* From this point forward we keep holding onto the tasklist lock
1037 * so that our parent does not change from under us. -DaveM
1039 write_lock_irq(&tasklist_lock
);
1042 p
= find_task_by_vpid(pid
);
1047 if (!thread_group_leader(p
))
1050 if (same_thread_group(p
->real_parent
, group_leader
)) {
1052 if (task_session(p
) != task_session(group_leader
))
1055 if (!(p
->flags
& PF_FORKNOEXEC
))
1059 if (p
!= group_leader
)
1064 if (p
->signal
->leader
)
1069 struct task_struct
*g
;
1071 pgrp
= find_vpid(pgid
);
1072 g
= pid_task(pgrp
, PIDTYPE_PGID
);
1073 if (!g
|| task_session(g
) != task_session(group_leader
))
1077 err
= security_task_setpgid(p
, pgid
);
1081 if (task_pgrp(p
) != pgrp
)
1082 change_pid(p
, PIDTYPE_PGID
, pgrp
);
1086 /* All paths lead to here, thus we are safe. -DaveM */
1087 write_unlock_irq(&tasklist_lock
);
1092 static int do_getpgid(pid_t pid
)
1094 struct task_struct
*p
;
1100 grp
= task_pgrp(current
);
1103 p
= find_task_by_vpid(pid
);
1110 retval
= security_task_getpgid(p
);
1114 retval
= pid_vnr(grp
);
1120 SYSCALL_DEFINE1(getpgid
, pid_t
, pid
)
1122 return do_getpgid(pid
);
1125 #ifdef __ARCH_WANT_SYS_GETPGRP
1127 SYSCALL_DEFINE0(getpgrp
)
1129 return do_getpgid(0);
1134 SYSCALL_DEFINE1(getsid
, pid_t
, pid
)
1136 struct task_struct
*p
;
1142 sid
= task_session(current
);
1145 p
= find_task_by_vpid(pid
);
1148 sid
= task_session(p
);
1152 retval
= security_task_getsid(p
);
1156 retval
= pid_vnr(sid
);
1162 static void set_special_pids(struct pid
*pid
)
1164 struct task_struct
*curr
= current
->group_leader
;
1166 if (task_session(curr
) != pid
)
1167 change_pid(curr
, PIDTYPE_SID
, pid
);
1169 if (task_pgrp(curr
) != pid
)
1170 change_pid(curr
, PIDTYPE_PGID
, pid
);
1173 int ksys_setsid(void)
1175 struct task_struct
*group_leader
= current
->group_leader
;
1176 struct pid
*sid
= task_pid(group_leader
);
1177 pid_t session
= pid_vnr(sid
);
1180 write_lock_irq(&tasklist_lock
);
1181 /* Fail if I am already a session leader */
1182 if (group_leader
->signal
->leader
)
1185 /* Fail if a process group id already exists that equals the
1186 * proposed session id.
1188 if (pid_task(sid
, PIDTYPE_PGID
))
1191 group_leader
->signal
->leader
= 1;
1192 set_special_pids(sid
);
1194 proc_clear_tty(group_leader
);
1198 write_unlock_irq(&tasklist_lock
);
1200 proc_sid_connector(group_leader
);
1201 sched_autogroup_create_attach(group_leader
);
1206 SYSCALL_DEFINE0(setsid
)
1208 return ksys_setsid();
1211 DECLARE_RWSEM(uts_sem
);
1213 #ifdef COMPAT_UTS_MACHINE
1214 static char compat_uts_machine
[__OLD_UTS_LEN
+1] = COMPAT_UTS_MACHINE
;
1216 static int __init
parse_compat_uts_machine(char *arg
)
1218 strncpy(compat_uts_machine
, arg
, __OLD_UTS_LEN
);
1219 compat_uts_machine
[__OLD_UTS_LEN
] = 0;
1222 early_param("compat_uts_machine", parse_compat_uts_machine
);
1224 #undef COMPAT_UTS_MACHINE
1225 #define COMPAT_UTS_MACHINE compat_uts_machine
1228 #ifdef COMPAT_UTS_MACHINE
1229 #define override_architecture(name) \
1230 (personality(current->personality) == PER_LINUX32 && \
1231 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1232 sizeof(COMPAT_UTS_MACHINE)))
1234 #define override_architecture(name) 0
1238 * Work around broken programs that cannot handle "Linux 3.0".
1239 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1240 * And we map 4.x and later versions to 2.6.60+x, so 4.0/5.0/6.0/... would be
1243 static int override_release(char __user
*release
, size_t len
)
1247 if (current
->personality
& UNAME26
) {
1248 const char *rest
= UTS_RELEASE
;
1249 char buf
[65] = { 0 };
1255 if (*rest
== '.' && ++ndots
>= 3)
1257 if (!isdigit(*rest
) && *rest
!= '.')
1261 v
= ((LINUX_VERSION_CODE
>> 8) & 0xff) + 60;
1262 copy
= clamp_t(size_t, len
, 1, sizeof(buf
));
1263 copy
= scnprintf(buf
, copy
, "2.6.%u%s", v
, rest
);
1264 ret
= copy_to_user(release
, buf
, copy
+ 1);
1269 SYSCALL_DEFINE1(newuname
, struct new_utsname __user
*, name
)
1271 struct new_utsname tmp
;
1273 down_read(&uts_sem
);
1274 memcpy(&tmp
, utsname(), sizeof(tmp
));
1276 if (copy_to_user(name
, &tmp
, sizeof(tmp
)))
1279 if (override_release(name
->release
, sizeof(name
->release
)))
1281 if (override_architecture(name
))
1286 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1290 SYSCALL_DEFINE1(uname
, struct old_utsname __user
*, name
)
1292 struct old_utsname tmp
;
1297 down_read(&uts_sem
);
1298 memcpy(&tmp
, utsname(), sizeof(tmp
));
1300 if (copy_to_user(name
, &tmp
, sizeof(tmp
)))
1303 if (override_release(name
->release
, sizeof(name
->release
)))
1305 if (override_architecture(name
))
1310 SYSCALL_DEFINE1(olduname
, struct oldold_utsname __user
*, name
)
1312 struct oldold_utsname tmp
;
1317 memset(&tmp
, 0, sizeof(tmp
));
1319 down_read(&uts_sem
);
1320 memcpy(&tmp
.sysname
, &utsname()->sysname
, __OLD_UTS_LEN
);
1321 memcpy(&tmp
.nodename
, &utsname()->nodename
, __OLD_UTS_LEN
);
1322 memcpy(&tmp
.release
, &utsname()->release
, __OLD_UTS_LEN
);
1323 memcpy(&tmp
.version
, &utsname()->version
, __OLD_UTS_LEN
);
1324 memcpy(&tmp
.machine
, &utsname()->machine
, __OLD_UTS_LEN
);
1326 if (copy_to_user(name
, &tmp
, sizeof(tmp
)))
1329 if (override_architecture(name
))
1331 if (override_release(name
->release
, sizeof(name
->release
)))
1337 SYSCALL_DEFINE2(sethostname
, char __user
*, name
, int, len
)
1340 char tmp
[__NEW_UTS_LEN
];
1342 if (!ns_capable(current
->nsproxy
->uts_ns
->user_ns
, CAP_SYS_ADMIN
))
1345 if (len
< 0 || len
> __NEW_UTS_LEN
)
1348 if (!copy_from_user(tmp
, name
, len
)) {
1349 struct new_utsname
*u
;
1351 down_write(&uts_sem
);
1353 memcpy(u
->nodename
, tmp
, len
);
1354 memset(u
->nodename
+ len
, 0, sizeof(u
->nodename
) - len
);
1356 uts_proc_notify(UTS_PROC_HOSTNAME
);
1362 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1364 SYSCALL_DEFINE2(gethostname
, char __user
*, name
, int, len
)
1367 struct new_utsname
*u
;
1368 char tmp
[__NEW_UTS_LEN
+ 1];
1372 down_read(&uts_sem
);
1374 i
= 1 + strlen(u
->nodename
);
1377 memcpy(tmp
, u
->nodename
, i
);
1379 if (copy_to_user(name
, tmp
, i
))
1387 * Only setdomainname; getdomainname can be implemented by calling
1390 SYSCALL_DEFINE2(setdomainname
, char __user
*, name
, int, len
)
1393 char tmp
[__NEW_UTS_LEN
];
1395 if (!ns_capable(current
->nsproxy
->uts_ns
->user_ns
, CAP_SYS_ADMIN
))
1397 if (len
< 0 || len
> __NEW_UTS_LEN
)
1401 if (!copy_from_user(tmp
, name
, len
)) {
1402 struct new_utsname
*u
;
1404 down_write(&uts_sem
);
1406 memcpy(u
->domainname
, tmp
, len
);
1407 memset(u
->domainname
+ len
, 0, sizeof(u
->domainname
) - len
);
1409 uts_proc_notify(UTS_PROC_DOMAINNAME
);
1415 SYSCALL_DEFINE2(getrlimit
, unsigned int, resource
, struct rlimit __user
*, rlim
)
1417 struct rlimit value
;
1420 ret
= do_prlimit(current
, resource
, NULL
, &value
);
1422 ret
= copy_to_user(rlim
, &value
, sizeof(*rlim
)) ? -EFAULT
: 0;
1427 #ifdef CONFIG_COMPAT
1429 COMPAT_SYSCALL_DEFINE2(setrlimit
, unsigned int, resource
,
1430 struct compat_rlimit __user
*, rlim
)
1433 struct compat_rlimit r32
;
1435 if (copy_from_user(&r32
, rlim
, sizeof(struct compat_rlimit
)))
1438 if (r32
.rlim_cur
== COMPAT_RLIM_INFINITY
)
1439 r
.rlim_cur
= RLIM_INFINITY
;
1441 r
.rlim_cur
= r32
.rlim_cur
;
1442 if (r32
.rlim_max
== COMPAT_RLIM_INFINITY
)
1443 r
.rlim_max
= RLIM_INFINITY
;
1445 r
.rlim_max
= r32
.rlim_max
;
1446 return do_prlimit(current
, resource
, &r
, NULL
);
1449 COMPAT_SYSCALL_DEFINE2(getrlimit
, unsigned int, resource
,
1450 struct compat_rlimit __user
*, rlim
)
1455 ret
= do_prlimit(current
, resource
, NULL
, &r
);
1457 struct compat_rlimit r32
;
1458 if (r
.rlim_cur
> COMPAT_RLIM_INFINITY
)
1459 r32
.rlim_cur
= COMPAT_RLIM_INFINITY
;
1461 r32
.rlim_cur
= r
.rlim_cur
;
1462 if (r
.rlim_max
> COMPAT_RLIM_INFINITY
)
1463 r32
.rlim_max
= COMPAT_RLIM_INFINITY
;
1465 r32
.rlim_max
= r
.rlim_max
;
1467 if (copy_to_user(rlim
, &r32
, sizeof(struct compat_rlimit
)))
1475 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1478 * Back compatibility for getrlimit. Needed for some apps.
1480 SYSCALL_DEFINE2(old_getrlimit
, unsigned int, resource
,
1481 struct rlimit __user
*, rlim
)
1484 if (resource
>= RLIM_NLIMITS
)
1487 resource
= array_index_nospec(resource
, RLIM_NLIMITS
);
1488 task_lock(current
->group_leader
);
1489 x
= current
->signal
->rlim
[resource
];
1490 task_unlock(current
->group_leader
);
1491 if (x
.rlim_cur
> 0x7FFFFFFF)
1492 x
.rlim_cur
= 0x7FFFFFFF;
1493 if (x
.rlim_max
> 0x7FFFFFFF)
1494 x
.rlim_max
= 0x7FFFFFFF;
1495 return copy_to_user(rlim
, &x
, sizeof(x
)) ? -EFAULT
: 0;
1498 #ifdef CONFIG_COMPAT
1499 COMPAT_SYSCALL_DEFINE2(old_getrlimit
, unsigned int, resource
,
1500 struct compat_rlimit __user
*, rlim
)
1504 if (resource
>= RLIM_NLIMITS
)
1507 resource
= array_index_nospec(resource
, RLIM_NLIMITS
);
1508 task_lock(current
->group_leader
);
1509 r
= current
->signal
->rlim
[resource
];
1510 task_unlock(current
->group_leader
);
1511 if (r
.rlim_cur
> 0x7FFFFFFF)
1512 r
.rlim_cur
= 0x7FFFFFFF;
1513 if (r
.rlim_max
> 0x7FFFFFFF)
1514 r
.rlim_max
= 0x7FFFFFFF;
1516 if (put_user(r
.rlim_cur
, &rlim
->rlim_cur
) ||
1517 put_user(r
.rlim_max
, &rlim
->rlim_max
))
1525 static inline bool rlim64_is_infinity(__u64 rlim64
)
1527 #if BITS_PER_LONG < 64
1528 return rlim64
>= ULONG_MAX
;
1530 return rlim64
== RLIM64_INFINITY
;
1534 static void rlim_to_rlim64(const struct rlimit
*rlim
, struct rlimit64
*rlim64
)
1536 if (rlim
->rlim_cur
== RLIM_INFINITY
)
1537 rlim64
->rlim_cur
= RLIM64_INFINITY
;
1539 rlim64
->rlim_cur
= rlim
->rlim_cur
;
1540 if (rlim
->rlim_max
== RLIM_INFINITY
)
1541 rlim64
->rlim_max
= RLIM64_INFINITY
;
1543 rlim64
->rlim_max
= rlim
->rlim_max
;
1546 static void rlim64_to_rlim(const struct rlimit64
*rlim64
, struct rlimit
*rlim
)
1548 if (rlim64_is_infinity(rlim64
->rlim_cur
))
1549 rlim
->rlim_cur
= RLIM_INFINITY
;
1551 rlim
->rlim_cur
= (unsigned long)rlim64
->rlim_cur
;
1552 if (rlim64_is_infinity(rlim64
->rlim_max
))
1553 rlim
->rlim_max
= RLIM_INFINITY
;
1555 rlim
->rlim_max
= (unsigned long)rlim64
->rlim_max
;
1558 /* make sure you are allowed to change @tsk limits before calling this */
1559 int do_prlimit(struct task_struct
*tsk
, unsigned int resource
,
1560 struct rlimit
*new_rlim
, struct rlimit
*old_rlim
)
1562 struct rlimit
*rlim
;
1565 if (resource
>= RLIM_NLIMITS
)
1568 if (new_rlim
->rlim_cur
> new_rlim
->rlim_max
)
1570 if (resource
== RLIMIT_NOFILE
&&
1571 new_rlim
->rlim_max
> sysctl_nr_open
)
1575 /* protect tsk->signal and tsk->sighand from disappearing */
1576 read_lock(&tasklist_lock
);
1577 if (!tsk
->sighand
) {
1582 rlim
= tsk
->signal
->rlim
+ resource
;
1583 task_lock(tsk
->group_leader
);
1585 /* Keep the capable check against init_user_ns until
1586 cgroups can contain all limits */
1587 if (new_rlim
->rlim_max
> rlim
->rlim_max
&&
1588 !capable(CAP_SYS_RESOURCE
))
1591 retval
= security_task_setrlimit(tsk
, resource
, new_rlim
);
1599 task_unlock(tsk
->group_leader
);
1602 * RLIMIT_CPU handling. Arm the posix CPU timer if the limit is not
1603 * infite. In case of RLIM_INFINITY the posix CPU timer code
1604 * ignores the rlimit.
1606 if (!retval
&& new_rlim
&& resource
== RLIMIT_CPU
&&
1607 new_rlim
->rlim_cur
!= RLIM_INFINITY
&&
1608 IS_ENABLED(CONFIG_POSIX_TIMERS
))
1609 update_rlimit_cpu(tsk
, new_rlim
->rlim_cur
);
1611 read_unlock(&tasklist_lock
);
1615 /* rcu lock must be held */
1616 static int check_prlimit_permission(struct task_struct
*task
,
1619 const struct cred
*cred
= current_cred(), *tcred
;
1622 if (current
== task
)
1625 tcred
= __task_cred(task
);
1626 id_match
= (uid_eq(cred
->uid
, tcred
->euid
) &&
1627 uid_eq(cred
->uid
, tcred
->suid
) &&
1628 uid_eq(cred
->uid
, tcred
->uid
) &&
1629 gid_eq(cred
->gid
, tcred
->egid
) &&
1630 gid_eq(cred
->gid
, tcred
->sgid
) &&
1631 gid_eq(cred
->gid
, tcred
->gid
));
1632 if (!id_match
&& !ns_capable(tcred
->user_ns
, CAP_SYS_RESOURCE
))
1635 return security_task_prlimit(cred
, tcred
, flags
);
1638 SYSCALL_DEFINE4(prlimit64
, pid_t
, pid
, unsigned int, resource
,
1639 const struct rlimit64 __user
*, new_rlim
,
1640 struct rlimit64 __user
*, old_rlim
)
1642 struct rlimit64 old64
, new64
;
1643 struct rlimit old
, new;
1644 struct task_struct
*tsk
;
1645 unsigned int checkflags
= 0;
1649 checkflags
|= LSM_PRLIMIT_READ
;
1652 if (copy_from_user(&new64
, new_rlim
, sizeof(new64
)))
1654 rlim64_to_rlim(&new64
, &new);
1655 checkflags
|= LSM_PRLIMIT_WRITE
;
1659 tsk
= pid
? find_task_by_vpid(pid
) : current
;
1664 ret
= check_prlimit_permission(tsk
, checkflags
);
1669 get_task_struct(tsk
);
1672 ret
= do_prlimit(tsk
, resource
, new_rlim
? &new : NULL
,
1673 old_rlim
? &old
: NULL
);
1675 if (!ret
&& old_rlim
) {
1676 rlim_to_rlim64(&old
, &old64
);
1677 if (copy_to_user(old_rlim
, &old64
, sizeof(old64
)))
1681 put_task_struct(tsk
);
1685 SYSCALL_DEFINE2(setrlimit
, unsigned int, resource
, struct rlimit __user
*, rlim
)
1687 struct rlimit new_rlim
;
1689 if (copy_from_user(&new_rlim
, rlim
, sizeof(*rlim
)))
1691 return do_prlimit(current
, resource
, &new_rlim
, NULL
);
1695 * It would make sense to put struct rusage in the task_struct,
1696 * except that would make the task_struct be *really big*. After
1697 * task_struct gets moved into malloc'ed memory, it would
1698 * make sense to do this. It will make moving the rest of the information
1699 * a lot simpler! (Which we're not doing right now because we're not
1700 * measuring them yet).
1702 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1703 * races with threads incrementing their own counters. But since word
1704 * reads are atomic, we either get new values or old values and we don't
1705 * care which for the sums. We always take the siglock to protect reading
1706 * the c* fields from p->signal from races with exit.c updating those
1707 * fields when reaping, so a sample either gets all the additions of a
1708 * given child after it's reaped, or none so this sample is before reaping.
1711 * We need to take the siglock for CHILDEREN, SELF and BOTH
1712 * for the cases current multithreaded, non-current single threaded
1713 * non-current multithreaded. Thread traversal is now safe with
1715 * Strictly speaking, we donot need to take the siglock if we are current and
1716 * single threaded, as no one else can take our signal_struct away, no one
1717 * else can reap the children to update signal->c* counters, and no one else
1718 * can race with the signal-> fields. If we do not take any lock, the
1719 * signal-> fields could be read out of order while another thread was just
1720 * exiting. So we should place a read memory barrier when we avoid the lock.
1721 * On the writer side, write memory barrier is implied in __exit_signal
1722 * as __exit_signal releases the siglock spinlock after updating the signal->
1723 * fields. But we don't do this yet to keep things simple.
1727 static void accumulate_thread_rusage(struct task_struct
*t
, struct rusage
*r
)
1729 r
->ru_nvcsw
+= t
->nvcsw
;
1730 r
->ru_nivcsw
+= t
->nivcsw
;
1731 r
->ru_minflt
+= t
->min_flt
;
1732 r
->ru_majflt
+= t
->maj_flt
;
1733 r
->ru_inblock
+= task_io_get_inblock(t
);
1734 r
->ru_oublock
+= task_io_get_oublock(t
);
1737 void getrusage(struct task_struct
*p
, int who
, struct rusage
*r
)
1739 struct task_struct
*t
;
1740 unsigned long flags
;
1741 u64 tgutime
, tgstime
, utime
, stime
;
1742 unsigned long maxrss
= 0;
1744 memset((char *)r
, 0, sizeof (*r
));
1747 if (who
== RUSAGE_THREAD
) {
1748 task_cputime_adjusted(current
, &utime
, &stime
);
1749 accumulate_thread_rusage(p
, r
);
1750 maxrss
= p
->signal
->maxrss
;
1754 if (!lock_task_sighand(p
, &flags
))
1759 case RUSAGE_CHILDREN
:
1760 utime
= p
->signal
->cutime
;
1761 stime
= p
->signal
->cstime
;
1762 r
->ru_nvcsw
= p
->signal
->cnvcsw
;
1763 r
->ru_nivcsw
= p
->signal
->cnivcsw
;
1764 r
->ru_minflt
= p
->signal
->cmin_flt
;
1765 r
->ru_majflt
= p
->signal
->cmaj_flt
;
1766 r
->ru_inblock
= p
->signal
->cinblock
;
1767 r
->ru_oublock
= p
->signal
->coublock
;
1768 maxrss
= p
->signal
->cmaxrss
;
1770 if (who
== RUSAGE_CHILDREN
)
1775 thread_group_cputime_adjusted(p
, &tgutime
, &tgstime
);
1778 r
->ru_nvcsw
+= p
->signal
->nvcsw
;
1779 r
->ru_nivcsw
+= p
->signal
->nivcsw
;
1780 r
->ru_minflt
+= p
->signal
->min_flt
;
1781 r
->ru_majflt
+= p
->signal
->maj_flt
;
1782 r
->ru_inblock
+= p
->signal
->inblock
;
1783 r
->ru_oublock
+= p
->signal
->oublock
;
1784 if (maxrss
< p
->signal
->maxrss
)
1785 maxrss
= p
->signal
->maxrss
;
1788 accumulate_thread_rusage(t
, r
);
1789 } while_each_thread(p
, t
);
1795 unlock_task_sighand(p
, &flags
);
1798 r
->ru_utime
= ns_to_kernel_old_timeval(utime
);
1799 r
->ru_stime
= ns_to_kernel_old_timeval(stime
);
1801 if (who
!= RUSAGE_CHILDREN
) {
1802 struct mm_struct
*mm
= get_task_mm(p
);
1805 setmax_mm_hiwater_rss(&maxrss
, mm
);
1809 r
->ru_maxrss
= maxrss
* (PAGE_SIZE
/ 1024); /* convert pages to KBs */
1812 SYSCALL_DEFINE2(getrusage
, int, who
, struct rusage __user
*, ru
)
1816 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
&&
1817 who
!= RUSAGE_THREAD
)
1820 getrusage(current
, who
, &r
);
1821 return copy_to_user(ru
, &r
, sizeof(r
)) ? -EFAULT
: 0;
1824 #ifdef CONFIG_COMPAT
1825 COMPAT_SYSCALL_DEFINE2(getrusage
, int, who
, struct compat_rusage __user
*, ru
)
1829 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
&&
1830 who
!= RUSAGE_THREAD
)
1833 getrusage(current
, who
, &r
);
1834 return put_compat_rusage(&r
, ru
);
1838 SYSCALL_DEFINE1(umask
, int, mask
)
1840 mask
= xchg(¤t
->fs
->umask
, mask
& S_IRWXUGO
);
1844 static int prctl_set_mm_exe_file(struct mm_struct
*mm
, unsigned int fd
)
1847 struct file
*old_exe
, *exe_file
;
1848 struct inode
*inode
;
1855 inode
= file_inode(exe
.file
);
1858 * Because the original mm->exe_file points to executable file, make
1859 * sure that this one is executable as well, to avoid breaking an
1863 if (!S_ISREG(inode
->i_mode
) || path_noexec(&exe
.file
->f_path
))
1866 err
= inode_permission(inode
, MAY_EXEC
);
1871 * Forbid mm->exe_file change if old file still mapped.
1873 exe_file
= get_mm_exe_file(mm
);
1876 struct vm_area_struct
*vma
;
1879 for (vma
= mm
->mmap
; vma
; vma
= vma
->vm_next
) {
1882 if (path_equal(&vma
->vm_file
->f_path
,
1887 mmap_read_unlock(mm
);
1892 /* set the new file, lockless */
1894 old_exe
= xchg(&mm
->exe_file
, exe
.file
);
1901 mmap_read_unlock(mm
);
1907 * Check arithmetic relations of passed addresses.
1909 * WARNING: we don't require any capability here so be very careful
1910 * in what is allowed for modification from userspace.
1912 static int validate_prctl_map_addr(struct prctl_mm_map
*prctl_map
)
1914 unsigned long mmap_max_addr
= TASK_SIZE
;
1915 int error
= -EINVAL
, i
;
1917 static const unsigned char offsets
[] = {
1918 offsetof(struct prctl_mm_map
, start_code
),
1919 offsetof(struct prctl_mm_map
, end_code
),
1920 offsetof(struct prctl_mm_map
, start_data
),
1921 offsetof(struct prctl_mm_map
, end_data
),
1922 offsetof(struct prctl_mm_map
, start_brk
),
1923 offsetof(struct prctl_mm_map
, brk
),
1924 offsetof(struct prctl_mm_map
, start_stack
),
1925 offsetof(struct prctl_mm_map
, arg_start
),
1926 offsetof(struct prctl_mm_map
, arg_end
),
1927 offsetof(struct prctl_mm_map
, env_start
),
1928 offsetof(struct prctl_mm_map
, env_end
),
1932 * Make sure the members are not somewhere outside
1933 * of allowed address space.
1935 for (i
= 0; i
< ARRAY_SIZE(offsets
); i
++) {
1936 u64 val
= *(u64
*)((char *)prctl_map
+ offsets
[i
]);
1938 if ((unsigned long)val
>= mmap_max_addr
||
1939 (unsigned long)val
< mmap_min_addr
)
1944 * Make sure the pairs are ordered.
1946 #define __prctl_check_order(__m1, __op, __m2) \
1947 ((unsigned long)prctl_map->__m1 __op \
1948 (unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1949 error
= __prctl_check_order(start_code
, <, end_code
);
1950 error
|= __prctl_check_order(start_data
,<=, end_data
);
1951 error
|= __prctl_check_order(start_brk
, <=, brk
);
1952 error
|= __prctl_check_order(arg_start
, <=, arg_end
);
1953 error
|= __prctl_check_order(env_start
, <=, env_end
);
1956 #undef __prctl_check_order
1961 * @brk should be after @end_data in traditional maps.
1963 if (prctl_map
->start_brk
<= prctl_map
->end_data
||
1964 prctl_map
->brk
<= prctl_map
->end_data
)
1968 * Neither we should allow to override limits if they set.
1970 if (check_data_rlimit(rlimit(RLIMIT_DATA
), prctl_map
->brk
,
1971 prctl_map
->start_brk
, prctl_map
->end_data
,
1972 prctl_map
->start_data
))
1980 #ifdef CONFIG_CHECKPOINT_RESTORE
1981 static int prctl_set_mm_map(int opt
, const void __user
*addr
, unsigned long data_size
)
1983 struct prctl_mm_map prctl_map
= { .exe_fd
= (u32
)-1, };
1984 unsigned long user_auxv
[AT_VECTOR_SIZE
];
1985 struct mm_struct
*mm
= current
->mm
;
1988 BUILD_BUG_ON(sizeof(user_auxv
) != sizeof(mm
->saved_auxv
));
1989 BUILD_BUG_ON(sizeof(struct prctl_mm_map
) > 256);
1991 if (opt
== PR_SET_MM_MAP_SIZE
)
1992 return put_user((unsigned int)sizeof(prctl_map
),
1993 (unsigned int __user
*)addr
);
1995 if (data_size
!= sizeof(prctl_map
))
1998 if (copy_from_user(&prctl_map
, addr
, sizeof(prctl_map
)))
2001 error
= validate_prctl_map_addr(&prctl_map
);
2005 if (prctl_map
.auxv_size
) {
2007 * Someone is trying to cheat the auxv vector.
2009 if (!prctl_map
.auxv
||
2010 prctl_map
.auxv_size
> sizeof(mm
->saved_auxv
))
2013 memset(user_auxv
, 0, sizeof(user_auxv
));
2014 if (copy_from_user(user_auxv
,
2015 (const void __user
*)prctl_map
.auxv
,
2016 prctl_map
.auxv_size
))
2019 /* Last entry must be AT_NULL as specification requires */
2020 user_auxv
[AT_VECTOR_SIZE
- 2] = AT_NULL
;
2021 user_auxv
[AT_VECTOR_SIZE
- 1] = AT_NULL
;
2024 if (prctl_map
.exe_fd
!= (u32
)-1) {
2026 * Check if the current user is checkpoint/restore capable.
2027 * At the time of this writing, it checks for CAP_SYS_ADMIN
2028 * or CAP_CHECKPOINT_RESTORE.
2029 * Note that a user with access to ptrace can masquerade an
2030 * arbitrary program as any executable, even setuid ones.
2031 * This may have implications in the tomoyo subsystem.
2033 if (!checkpoint_restore_ns_capable(current_user_ns()))
2036 error
= prctl_set_mm_exe_file(mm
, prctl_map
.exe_fd
);
2042 * arg_lock protects concurent updates but we still need mmap_lock for
2043 * read to exclude races with sys_brk.
2048 * We don't validate if these members are pointing to
2049 * real present VMAs because application may have correspond
2050 * VMAs already unmapped and kernel uses these members for statistics
2051 * output in procfs mostly, except
2053 * - @start_brk/@brk which are used in do_brk_flags but kernel lookups
2054 * for VMAs when updating these memvers so anything wrong written
2055 * here cause kernel to swear at userspace program but won't lead
2056 * to any problem in kernel itself
2059 spin_lock(&mm
->arg_lock
);
2060 mm
->start_code
= prctl_map
.start_code
;
2061 mm
->end_code
= prctl_map
.end_code
;
2062 mm
->start_data
= prctl_map
.start_data
;
2063 mm
->end_data
= prctl_map
.end_data
;
2064 mm
->start_brk
= prctl_map
.start_brk
;
2065 mm
->brk
= prctl_map
.brk
;
2066 mm
->start_stack
= prctl_map
.start_stack
;
2067 mm
->arg_start
= prctl_map
.arg_start
;
2068 mm
->arg_end
= prctl_map
.arg_end
;
2069 mm
->env_start
= prctl_map
.env_start
;
2070 mm
->env_end
= prctl_map
.env_end
;
2071 spin_unlock(&mm
->arg_lock
);
2074 * Note this update of @saved_auxv is lockless thus
2075 * if someone reads this member in procfs while we're
2076 * updating -- it may get partly updated results. It's
2077 * known and acceptable trade off: we leave it as is to
2078 * not introduce additional locks here making the kernel
2081 if (prctl_map
.auxv_size
)
2082 memcpy(mm
->saved_auxv
, user_auxv
, sizeof(user_auxv
));
2084 mmap_read_unlock(mm
);
2087 #endif /* CONFIG_CHECKPOINT_RESTORE */
2089 static int prctl_set_auxv(struct mm_struct
*mm
, unsigned long addr
,
2093 * This doesn't move the auxiliary vector itself since it's pinned to
2094 * mm_struct, but it permits filling the vector with new values. It's
2095 * up to the caller to provide sane values here, otherwise userspace
2096 * tools which use this vector might be unhappy.
2098 unsigned long user_auxv
[AT_VECTOR_SIZE
];
2100 if (len
> sizeof(user_auxv
))
2103 if (copy_from_user(user_auxv
, (const void __user
*)addr
, len
))
2106 /* Make sure the last entry is always AT_NULL */
2107 user_auxv
[AT_VECTOR_SIZE
- 2] = 0;
2108 user_auxv
[AT_VECTOR_SIZE
- 1] = 0;
2110 BUILD_BUG_ON(sizeof(user_auxv
) != sizeof(mm
->saved_auxv
));
2113 memcpy(mm
->saved_auxv
, user_auxv
, len
);
2114 task_unlock(current
);
2119 static int prctl_set_mm(int opt
, unsigned long addr
,
2120 unsigned long arg4
, unsigned long arg5
)
2122 struct mm_struct
*mm
= current
->mm
;
2123 struct prctl_mm_map prctl_map
= {
2128 struct vm_area_struct
*vma
;
2131 if (arg5
|| (arg4
&& (opt
!= PR_SET_MM_AUXV
&&
2132 opt
!= PR_SET_MM_MAP
&&
2133 opt
!= PR_SET_MM_MAP_SIZE
)))
2136 #ifdef CONFIG_CHECKPOINT_RESTORE
2137 if (opt
== PR_SET_MM_MAP
|| opt
== PR_SET_MM_MAP_SIZE
)
2138 return prctl_set_mm_map(opt
, (const void __user
*)addr
, arg4
);
2141 if (!capable(CAP_SYS_RESOURCE
))
2144 if (opt
== PR_SET_MM_EXE_FILE
)
2145 return prctl_set_mm_exe_file(mm
, (unsigned int)addr
);
2147 if (opt
== PR_SET_MM_AUXV
)
2148 return prctl_set_auxv(mm
, addr
, arg4
);
2150 if (addr
>= TASK_SIZE
|| addr
< mmap_min_addr
)
2156 * arg_lock protects concurent updates of arg boundaries, we need
2157 * mmap_lock for a) concurrent sys_brk, b) finding VMA for addr
2161 vma
= find_vma(mm
, addr
);
2163 spin_lock(&mm
->arg_lock
);
2164 prctl_map
.start_code
= mm
->start_code
;
2165 prctl_map
.end_code
= mm
->end_code
;
2166 prctl_map
.start_data
= mm
->start_data
;
2167 prctl_map
.end_data
= mm
->end_data
;
2168 prctl_map
.start_brk
= mm
->start_brk
;
2169 prctl_map
.brk
= mm
->brk
;
2170 prctl_map
.start_stack
= mm
->start_stack
;
2171 prctl_map
.arg_start
= mm
->arg_start
;
2172 prctl_map
.arg_end
= mm
->arg_end
;
2173 prctl_map
.env_start
= mm
->env_start
;
2174 prctl_map
.env_end
= mm
->env_end
;
2177 case PR_SET_MM_START_CODE
:
2178 prctl_map
.start_code
= addr
;
2180 case PR_SET_MM_END_CODE
:
2181 prctl_map
.end_code
= addr
;
2183 case PR_SET_MM_START_DATA
:
2184 prctl_map
.start_data
= addr
;
2186 case PR_SET_MM_END_DATA
:
2187 prctl_map
.end_data
= addr
;
2189 case PR_SET_MM_START_STACK
:
2190 prctl_map
.start_stack
= addr
;
2192 case PR_SET_MM_START_BRK
:
2193 prctl_map
.start_brk
= addr
;
2196 prctl_map
.brk
= addr
;
2198 case PR_SET_MM_ARG_START
:
2199 prctl_map
.arg_start
= addr
;
2201 case PR_SET_MM_ARG_END
:
2202 prctl_map
.arg_end
= addr
;
2204 case PR_SET_MM_ENV_START
:
2205 prctl_map
.env_start
= addr
;
2207 case PR_SET_MM_ENV_END
:
2208 prctl_map
.env_end
= addr
;
2214 error
= validate_prctl_map_addr(&prctl_map
);
2220 * If command line arguments and environment
2221 * are placed somewhere else on stack, we can
2222 * set them up here, ARG_START/END to setup
2223 * command line argumets and ENV_START/END
2226 case PR_SET_MM_START_STACK
:
2227 case PR_SET_MM_ARG_START
:
2228 case PR_SET_MM_ARG_END
:
2229 case PR_SET_MM_ENV_START
:
2230 case PR_SET_MM_ENV_END
:
2237 mm
->start_code
= prctl_map
.start_code
;
2238 mm
->end_code
= prctl_map
.end_code
;
2239 mm
->start_data
= prctl_map
.start_data
;
2240 mm
->end_data
= prctl_map
.end_data
;
2241 mm
->start_brk
= prctl_map
.start_brk
;
2242 mm
->brk
= prctl_map
.brk
;
2243 mm
->start_stack
= prctl_map
.start_stack
;
2244 mm
->arg_start
= prctl_map
.arg_start
;
2245 mm
->arg_end
= prctl_map
.arg_end
;
2246 mm
->env_start
= prctl_map
.env_start
;
2247 mm
->env_end
= prctl_map
.env_end
;
2251 spin_unlock(&mm
->arg_lock
);
2252 mmap_read_unlock(mm
);
2256 #ifdef CONFIG_CHECKPOINT_RESTORE
2257 static int prctl_get_tid_address(struct task_struct
*me
, int __user
* __user
*tid_addr
)
2259 return put_user(me
->clear_child_tid
, tid_addr
);
2262 static int prctl_get_tid_address(struct task_struct
*me
, int __user
* __user
*tid_addr
)
2268 static int propagate_has_child_subreaper(struct task_struct
*p
, void *data
)
2271 * If task has has_child_subreaper - all its decendants
2272 * already have these flag too and new decendants will
2273 * inherit it on fork, skip them.
2275 * If we've found child_reaper - skip descendants in
2276 * it's subtree as they will never get out pidns.
2278 if (p
->signal
->has_child_subreaper
||
2279 is_child_reaper(task_pid(p
)))
2282 p
->signal
->has_child_subreaper
= 1;
2286 int __weak
arch_prctl_spec_ctrl_get(struct task_struct
*t
, unsigned long which
)
2291 int __weak
arch_prctl_spec_ctrl_set(struct task_struct
*t
, unsigned long which
,
2297 #define PR_IO_FLUSHER (PF_MEMALLOC_NOIO | PF_LOCAL_THROTTLE)
2299 SYSCALL_DEFINE5(prctl
, int, option
, unsigned long, arg2
, unsigned long, arg3
,
2300 unsigned long, arg4
, unsigned long, arg5
)
2302 struct task_struct
*me
= current
;
2303 unsigned char comm
[sizeof(me
->comm
)];
2306 error
= security_task_prctl(option
, arg2
, arg3
, arg4
, arg5
);
2307 if (error
!= -ENOSYS
)
2312 case PR_SET_PDEATHSIG
:
2313 if (!valid_signal(arg2
)) {
2317 me
->pdeath_signal
= arg2
;
2319 case PR_GET_PDEATHSIG
:
2320 error
= put_user(me
->pdeath_signal
, (int __user
*)arg2
);
2322 case PR_GET_DUMPABLE
:
2323 error
= get_dumpable(me
->mm
);
2325 case PR_SET_DUMPABLE
:
2326 if (arg2
!= SUID_DUMP_DISABLE
&& arg2
!= SUID_DUMP_USER
) {
2330 set_dumpable(me
->mm
, arg2
);
2333 case PR_SET_UNALIGN
:
2334 error
= SET_UNALIGN_CTL(me
, arg2
);
2336 case PR_GET_UNALIGN
:
2337 error
= GET_UNALIGN_CTL(me
, arg2
);
2340 error
= SET_FPEMU_CTL(me
, arg2
);
2343 error
= GET_FPEMU_CTL(me
, arg2
);
2346 error
= SET_FPEXC_CTL(me
, arg2
);
2349 error
= GET_FPEXC_CTL(me
, arg2
);
2352 error
= PR_TIMING_STATISTICAL
;
2355 if (arg2
!= PR_TIMING_STATISTICAL
)
2359 comm
[sizeof(me
->comm
) - 1] = 0;
2360 if (strncpy_from_user(comm
, (char __user
*)arg2
,
2361 sizeof(me
->comm
) - 1) < 0)
2363 set_task_comm(me
, comm
);
2364 proc_comm_connector(me
);
2367 get_task_comm(comm
, me
);
2368 if (copy_to_user((char __user
*)arg2
, comm
, sizeof(comm
)))
2372 error
= GET_ENDIAN(me
, arg2
);
2375 error
= SET_ENDIAN(me
, arg2
);
2377 case PR_GET_SECCOMP
:
2378 error
= prctl_get_seccomp();
2380 case PR_SET_SECCOMP
:
2381 error
= prctl_set_seccomp(arg2
, (char __user
*)arg3
);
2384 error
= GET_TSC_CTL(arg2
);
2387 error
= SET_TSC_CTL(arg2
);
2389 case PR_TASK_PERF_EVENTS_DISABLE
:
2390 error
= perf_event_task_disable();
2392 case PR_TASK_PERF_EVENTS_ENABLE
:
2393 error
= perf_event_task_enable();
2395 case PR_GET_TIMERSLACK
:
2396 if (current
->timer_slack_ns
> ULONG_MAX
)
2399 error
= current
->timer_slack_ns
;
2401 case PR_SET_TIMERSLACK
:
2403 current
->timer_slack_ns
=
2404 current
->default_timer_slack_ns
;
2406 current
->timer_slack_ns
= arg2
;
2412 case PR_MCE_KILL_CLEAR
:
2415 current
->flags
&= ~PF_MCE_PROCESS
;
2417 case PR_MCE_KILL_SET
:
2418 current
->flags
|= PF_MCE_PROCESS
;
2419 if (arg3
== PR_MCE_KILL_EARLY
)
2420 current
->flags
|= PF_MCE_EARLY
;
2421 else if (arg3
== PR_MCE_KILL_LATE
)
2422 current
->flags
&= ~PF_MCE_EARLY
;
2423 else if (arg3
== PR_MCE_KILL_DEFAULT
)
2425 ~(PF_MCE_EARLY
|PF_MCE_PROCESS
);
2433 case PR_MCE_KILL_GET
:
2434 if (arg2
| arg3
| arg4
| arg5
)
2436 if (current
->flags
& PF_MCE_PROCESS
)
2437 error
= (current
->flags
& PF_MCE_EARLY
) ?
2438 PR_MCE_KILL_EARLY
: PR_MCE_KILL_LATE
;
2440 error
= PR_MCE_KILL_DEFAULT
;
2443 error
= prctl_set_mm(arg2
, arg3
, arg4
, arg5
);
2445 case PR_GET_TID_ADDRESS
:
2446 error
= prctl_get_tid_address(me
, (int __user
* __user
*)arg2
);
2448 case PR_SET_CHILD_SUBREAPER
:
2449 me
->signal
->is_child_subreaper
= !!arg2
;
2453 walk_process_tree(me
, propagate_has_child_subreaper
, NULL
);
2455 case PR_GET_CHILD_SUBREAPER
:
2456 error
= put_user(me
->signal
->is_child_subreaper
,
2457 (int __user
*)arg2
);
2459 case PR_SET_NO_NEW_PRIVS
:
2460 if (arg2
!= 1 || arg3
|| arg4
|| arg5
)
2463 task_set_no_new_privs(current
);
2465 case PR_GET_NO_NEW_PRIVS
:
2466 if (arg2
|| arg3
|| arg4
|| arg5
)
2468 return task_no_new_privs(current
) ? 1 : 0;
2469 case PR_GET_THP_DISABLE
:
2470 if (arg2
|| arg3
|| arg4
|| arg5
)
2472 error
= !!test_bit(MMF_DISABLE_THP
, &me
->mm
->flags
);
2474 case PR_SET_THP_DISABLE
:
2475 if (arg3
|| arg4
|| arg5
)
2477 if (mmap_write_lock_killable(me
->mm
))
2480 set_bit(MMF_DISABLE_THP
, &me
->mm
->flags
);
2482 clear_bit(MMF_DISABLE_THP
, &me
->mm
->flags
);
2483 mmap_write_unlock(me
->mm
);
2485 case PR_MPX_ENABLE_MANAGEMENT
:
2486 case PR_MPX_DISABLE_MANAGEMENT
:
2487 /* No longer implemented: */
2489 case PR_SET_FP_MODE
:
2490 error
= SET_FP_MODE(me
, arg2
);
2492 case PR_GET_FP_MODE
:
2493 error
= GET_FP_MODE(me
);
2496 error
= SVE_SET_VL(arg2
);
2499 error
= SVE_GET_VL();
2501 case PR_GET_SPECULATION_CTRL
:
2502 if (arg3
|| arg4
|| arg5
)
2504 error
= arch_prctl_spec_ctrl_get(me
, arg2
);
2506 case PR_SET_SPECULATION_CTRL
:
2509 error
= arch_prctl_spec_ctrl_set(me
, arg2
, arg3
);
2511 case PR_PAC_RESET_KEYS
:
2512 if (arg3
|| arg4
|| arg5
)
2514 error
= PAC_RESET_KEYS(me
, arg2
);
2516 case PR_SET_TAGGED_ADDR_CTRL
:
2517 if (arg3
|| arg4
|| arg5
)
2519 error
= SET_TAGGED_ADDR_CTRL(arg2
);
2521 case PR_GET_TAGGED_ADDR_CTRL
:
2522 if (arg2
|| arg3
|| arg4
|| arg5
)
2524 error
= GET_TAGGED_ADDR_CTRL();
2526 case PR_SET_IO_FLUSHER
:
2527 if (!capable(CAP_SYS_RESOURCE
))
2530 if (arg3
|| arg4
|| arg5
)
2534 current
->flags
|= PR_IO_FLUSHER
;
2536 current
->flags
&= ~PR_IO_FLUSHER
;
2540 case PR_GET_IO_FLUSHER
:
2541 if (!capable(CAP_SYS_RESOURCE
))
2544 if (arg2
|| arg3
|| arg4
|| arg5
)
2547 error
= (current
->flags
& PR_IO_FLUSHER
) == PR_IO_FLUSHER
;
2549 case PR_SET_SYSCALL_USER_DISPATCH
:
2550 error
= set_syscall_user_dispatch(arg2
, arg3
, arg4
,
2551 (char __user
*) arg5
);
2560 SYSCALL_DEFINE3(getcpu
, unsigned __user
*, cpup
, unsigned __user
*, nodep
,
2561 struct getcpu_cache __user
*, unused
)
2564 int cpu
= raw_smp_processor_id();
2567 err
|= put_user(cpu
, cpup
);
2569 err
|= put_user(cpu_to_node(cpu
), nodep
);
2570 return err
? -EFAULT
: 0;
2574 * do_sysinfo - fill in sysinfo struct
2575 * @info: pointer to buffer to fill
2577 static int do_sysinfo(struct sysinfo
*info
)
2579 unsigned long mem_total
, sav_total
;
2580 unsigned int mem_unit
, bitcount
;
2581 struct timespec64 tp
;
2583 memset(info
, 0, sizeof(struct sysinfo
));
2585 ktime_get_boottime_ts64(&tp
);
2586 timens_add_boottime(&tp
);
2587 info
->uptime
= tp
.tv_sec
+ (tp
.tv_nsec
? 1 : 0);
2589 get_avenrun(info
->loads
, 0, SI_LOAD_SHIFT
- FSHIFT
);
2591 info
->procs
= nr_threads
;
2597 * If the sum of all the available memory (i.e. ram + swap)
2598 * is less than can be stored in a 32 bit unsigned long then
2599 * we can be binary compatible with 2.2.x kernels. If not,
2600 * well, in that case 2.2.x was broken anyways...
2602 * -Erik Andersen <andersee@debian.org>
2605 mem_total
= info
->totalram
+ info
->totalswap
;
2606 if (mem_total
< info
->totalram
|| mem_total
< info
->totalswap
)
2609 mem_unit
= info
->mem_unit
;
2610 while (mem_unit
> 1) {
2613 sav_total
= mem_total
;
2615 if (mem_total
< sav_total
)
2620 * If mem_total did not overflow, multiply all memory values by
2621 * info->mem_unit and set it to 1. This leaves things compatible
2622 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2627 info
->totalram
<<= bitcount
;
2628 info
->freeram
<<= bitcount
;
2629 info
->sharedram
<<= bitcount
;
2630 info
->bufferram
<<= bitcount
;
2631 info
->totalswap
<<= bitcount
;
2632 info
->freeswap
<<= bitcount
;
2633 info
->totalhigh
<<= bitcount
;
2634 info
->freehigh
<<= bitcount
;
2640 SYSCALL_DEFINE1(sysinfo
, struct sysinfo __user
*, info
)
2646 if (copy_to_user(info
, &val
, sizeof(struct sysinfo
)))
2652 #ifdef CONFIG_COMPAT
2653 struct compat_sysinfo
{
2667 char _f
[20-2*sizeof(u32
)-sizeof(int)];
2670 COMPAT_SYSCALL_DEFINE1(sysinfo
, struct compat_sysinfo __user
*, info
)
2673 struct compat_sysinfo s_32
;
2677 /* Check to see if any memory value is too large for 32-bit and scale
2680 if (upper_32_bits(s
.totalram
) || upper_32_bits(s
.totalswap
)) {
2683 while (s
.mem_unit
< PAGE_SIZE
) {
2688 s
.totalram
>>= bitcount
;
2689 s
.freeram
>>= bitcount
;
2690 s
.sharedram
>>= bitcount
;
2691 s
.bufferram
>>= bitcount
;
2692 s
.totalswap
>>= bitcount
;
2693 s
.freeswap
>>= bitcount
;
2694 s
.totalhigh
>>= bitcount
;
2695 s
.freehigh
>>= bitcount
;
2698 memset(&s_32
, 0, sizeof(s_32
));
2699 s_32
.uptime
= s
.uptime
;
2700 s_32
.loads
[0] = s
.loads
[0];
2701 s_32
.loads
[1] = s
.loads
[1];
2702 s_32
.loads
[2] = s
.loads
[2];
2703 s_32
.totalram
= s
.totalram
;
2704 s_32
.freeram
= s
.freeram
;
2705 s_32
.sharedram
= s
.sharedram
;
2706 s_32
.bufferram
= s
.bufferram
;
2707 s_32
.totalswap
= s
.totalswap
;
2708 s_32
.freeswap
= s
.freeswap
;
2709 s_32
.procs
= s
.procs
;
2710 s_32
.totalhigh
= s
.totalhigh
;
2711 s_32
.freehigh
= s
.freehigh
;
2712 s_32
.mem_unit
= s
.mem_unit
;
2713 if (copy_to_user(info
, &s_32
, sizeof(s_32
)))
2717 #endif /* CONFIG_COMPAT */