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>
46 #include <linux/compat.h>
47 #include <linux/syscalls.h>
48 #include <linux/kprobes.h>
49 #include <linux/user_namespace.h>
50 #include <linux/time_namespace.h>
51 #include <linux/binfmts.h>
53 #include <linux/sched.h>
54 #include <linux/sched/autogroup.h>
55 #include <linux/sched/loadavg.h>
56 #include <linux/sched/stat.h>
57 #include <linux/sched/mm.h>
58 #include <linux/sched/coredump.h>
59 #include <linux/sched/task.h>
60 #include <linux/sched/cputime.h>
61 #include <linux/rcupdate.h>
62 #include <linux/uidgid.h>
63 #include <linux/cred.h>
65 #include <linux/nospec.h>
67 #include <linux/kmsg_dump.h>
68 /* Move somewhere else to avoid recompiling? */
69 #include <generated/utsrelease.h>
71 #include <linux/uaccess.h>
73 #include <asm/unistd.h>
77 #ifndef SET_UNALIGN_CTL
78 # define SET_UNALIGN_CTL(a, b) (-EINVAL)
80 #ifndef GET_UNALIGN_CTL
81 # define GET_UNALIGN_CTL(a, b) (-EINVAL)
84 # define SET_FPEMU_CTL(a, b) (-EINVAL)
87 # define GET_FPEMU_CTL(a, b) (-EINVAL)
90 # define SET_FPEXC_CTL(a, b) (-EINVAL)
93 # define GET_FPEXC_CTL(a, b) (-EINVAL)
96 # define GET_ENDIAN(a, b) (-EINVAL)
99 # define SET_ENDIAN(a, b) (-EINVAL)
102 # define GET_TSC_CTL(a) (-EINVAL)
105 # define SET_TSC_CTL(a) (-EINVAL)
108 # define GET_FP_MODE(a) (-EINVAL)
111 # define SET_FP_MODE(a,b) (-EINVAL)
114 # define SVE_SET_VL(a) (-EINVAL)
117 # define SVE_GET_VL() (-EINVAL)
119 #ifndef PAC_RESET_KEYS
120 # define PAC_RESET_KEYS(a, b) (-EINVAL)
122 #ifndef SET_TAGGED_ADDR_CTRL
123 # define SET_TAGGED_ADDR_CTRL(a) (-EINVAL)
125 #ifndef GET_TAGGED_ADDR_CTRL
126 # define GET_TAGGED_ADDR_CTRL() (-EINVAL)
130 * this is where the system-wide overflow UID and GID are defined, for
131 * architectures that now have 32-bit UID/GID but didn't in the past
134 int overflowuid
= DEFAULT_OVERFLOWUID
;
135 int overflowgid
= DEFAULT_OVERFLOWGID
;
137 EXPORT_SYMBOL(overflowuid
);
138 EXPORT_SYMBOL(overflowgid
);
141 * the same as above, but for filesystems which can only store a 16-bit
142 * UID and GID. as such, this is needed on all architectures
145 int fs_overflowuid
= DEFAULT_FS_OVERFLOWUID
;
146 int fs_overflowgid
= DEFAULT_FS_OVERFLOWGID
;
148 EXPORT_SYMBOL(fs_overflowuid
);
149 EXPORT_SYMBOL(fs_overflowgid
);
152 * Returns true if current's euid is same as p's uid or euid,
153 * or has CAP_SYS_NICE to p's user_ns.
155 * Called with rcu_read_lock, creds are safe
157 static bool set_one_prio_perm(struct task_struct
*p
)
159 const struct cred
*cred
= current_cred(), *pcred
= __task_cred(p
);
161 if (uid_eq(pcred
->uid
, cred
->euid
) ||
162 uid_eq(pcred
->euid
, cred
->euid
))
164 if (ns_capable(pcred
->user_ns
, CAP_SYS_NICE
))
170 * set the priority of a task
171 * - the caller must hold the RCU read lock
173 static int set_one_prio(struct task_struct
*p
, int niceval
, int error
)
177 if (!set_one_prio_perm(p
)) {
181 if (niceval
< task_nice(p
) && !can_nice(p
, niceval
)) {
185 no_nice
= security_task_setnice(p
, niceval
);
192 set_user_nice(p
, niceval
);
197 SYSCALL_DEFINE3(setpriority
, int, which
, int, who
, int, niceval
)
199 struct task_struct
*g
, *p
;
200 struct user_struct
*user
;
201 const struct cred
*cred
= current_cred();
206 if (which
> PRIO_USER
|| which
< PRIO_PROCESS
)
209 /* normalize: avoid signed division (rounding problems) */
211 if (niceval
< MIN_NICE
)
213 if (niceval
> MAX_NICE
)
217 read_lock(&tasklist_lock
);
221 p
= find_task_by_vpid(who
);
225 error
= set_one_prio(p
, niceval
, error
);
229 pgrp
= find_vpid(who
);
231 pgrp
= task_pgrp(current
);
232 do_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
) {
233 error
= set_one_prio(p
, niceval
, error
);
234 } while_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
);
237 uid
= make_kuid(cred
->user_ns
, who
);
241 else if (!uid_eq(uid
, cred
->uid
)) {
242 user
= find_user(uid
);
244 goto out_unlock
; /* No processes for this user */
246 do_each_thread(g
, p
) {
247 if (uid_eq(task_uid(p
), uid
) && task_pid_vnr(p
))
248 error
= set_one_prio(p
, niceval
, error
);
249 } while_each_thread(g
, p
);
250 if (!uid_eq(uid
, cred
->uid
))
251 free_uid(user
); /* For find_user() */
255 read_unlock(&tasklist_lock
);
262 * Ugh. To avoid negative return values, "getpriority()" will
263 * not return the normal nice-value, but a negated value that
264 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
265 * to stay compatible.
267 SYSCALL_DEFINE2(getpriority
, int, which
, int, who
)
269 struct task_struct
*g
, *p
;
270 struct user_struct
*user
;
271 const struct cred
*cred
= current_cred();
272 long niceval
, retval
= -ESRCH
;
276 if (which
> PRIO_USER
|| which
< PRIO_PROCESS
)
280 read_lock(&tasklist_lock
);
284 p
= find_task_by_vpid(who
);
288 niceval
= nice_to_rlimit(task_nice(p
));
289 if (niceval
> retval
)
295 pgrp
= find_vpid(who
);
297 pgrp
= task_pgrp(current
);
298 do_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
) {
299 niceval
= nice_to_rlimit(task_nice(p
));
300 if (niceval
> retval
)
302 } while_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
);
305 uid
= make_kuid(cred
->user_ns
, who
);
309 else if (!uid_eq(uid
, cred
->uid
)) {
310 user
= find_user(uid
);
312 goto out_unlock
; /* No processes for this user */
314 do_each_thread(g
, p
) {
315 if (uid_eq(task_uid(p
), uid
) && task_pid_vnr(p
)) {
316 niceval
= nice_to_rlimit(task_nice(p
));
317 if (niceval
> retval
)
320 } while_each_thread(g
, p
);
321 if (!uid_eq(uid
, cred
->uid
))
322 free_uid(user
); /* for find_user() */
326 read_unlock(&tasklist_lock
);
333 * Unprivileged users may change the real gid to the effective gid
334 * or vice versa. (BSD-style)
336 * If you set the real gid at all, or set the effective gid to a value not
337 * equal to the real gid, then the saved gid is set to the new effective gid.
339 * This makes it possible for a setgid program to completely drop its
340 * privileges, which is often a useful assertion to make when you are doing
341 * a security audit over a program.
343 * The general idea is that a program which uses just setregid() will be
344 * 100% compatible with BSD. A program which uses just setgid() will be
345 * 100% compatible with POSIX with saved IDs.
347 * SMP: There are not races, the GIDs are checked only by filesystem
348 * operations (as far as semantic preservation is concerned).
350 #ifdef CONFIG_MULTIUSER
351 long __sys_setregid(gid_t rgid
, gid_t egid
)
353 struct user_namespace
*ns
= current_user_ns();
354 const struct cred
*old
;
359 krgid
= make_kgid(ns
, rgid
);
360 kegid
= make_kgid(ns
, egid
);
362 if ((rgid
!= (gid_t
) -1) && !gid_valid(krgid
))
364 if ((egid
!= (gid_t
) -1) && !gid_valid(kegid
))
367 new = prepare_creds();
370 old
= current_cred();
373 if (rgid
!= (gid_t
) -1) {
374 if (gid_eq(old
->gid
, krgid
) ||
375 gid_eq(old
->egid
, krgid
) ||
376 ns_capable_setid(old
->user_ns
, CAP_SETGID
))
381 if (egid
!= (gid_t
) -1) {
382 if (gid_eq(old
->gid
, kegid
) ||
383 gid_eq(old
->egid
, kegid
) ||
384 gid_eq(old
->sgid
, kegid
) ||
385 ns_capable_setid(old
->user_ns
, CAP_SETGID
))
391 if (rgid
!= (gid_t
) -1 ||
392 (egid
!= (gid_t
) -1 && !gid_eq(kegid
, old
->gid
)))
393 new->sgid
= new->egid
;
394 new->fsgid
= new->egid
;
396 retval
= security_task_fix_setgid(new, old
, LSM_SETID_RE
);
400 return commit_creds(new);
407 SYSCALL_DEFINE2(setregid
, gid_t
, rgid
, gid_t
, egid
)
409 return __sys_setregid(rgid
, egid
);
413 * setgid() is implemented like SysV w/ SAVED_IDS
415 * SMP: Same implicit races as above.
417 long __sys_setgid(gid_t gid
)
419 struct user_namespace
*ns
= current_user_ns();
420 const struct cred
*old
;
425 kgid
= make_kgid(ns
, gid
);
426 if (!gid_valid(kgid
))
429 new = prepare_creds();
432 old
= current_cred();
435 if (ns_capable_setid(old
->user_ns
, CAP_SETGID
))
436 new->gid
= new->egid
= new->sgid
= new->fsgid
= kgid
;
437 else if (gid_eq(kgid
, old
->gid
) || gid_eq(kgid
, old
->sgid
))
438 new->egid
= new->fsgid
= kgid
;
442 retval
= security_task_fix_setgid(new, old
, LSM_SETID_ID
);
446 return commit_creds(new);
453 SYSCALL_DEFINE1(setgid
, gid_t
, gid
)
455 return __sys_setgid(gid
);
459 * change the user struct in a credentials set to match the new UID
461 static int set_user(struct cred
*new)
463 struct user_struct
*new_user
;
465 new_user
= alloc_uid(new->uid
);
470 * We don't fail in case of NPROC limit excess here because too many
471 * poorly written programs don't check set*uid() return code, assuming
472 * it never fails if called by root. We may still enforce NPROC limit
473 * for programs doing set*uid()+execve() by harmlessly deferring the
474 * failure to the execve() stage.
476 if (atomic_read(&new_user
->processes
) >= rlimit(RLIMIT_NPROC
) &&
477 new_user
!= INIT_USER
)
478 current
->flags
|= PF_NPROC_EXCEEDED
;
480 current
->flags
&= ~PF_NPROC_EXCEEDED
;
483 new->user
= new_user
;
488 * Unprivileged users may change the real uid to the effective uid
489 * or vice versa. (BSD-style)
491 * If you set the real uid at all, or set the effective uid to a value not
492 * equal to the real uid, then the saved uid is set to the new effective uid.
494 * This makes it possible for a setuid program to completely drop its
495 * privileges, which is often a useful assertion to make when you are doing
496 * a security audit over a program.
498 * The general idea is that a program which uses just setreuid() will be
499 * 100% compatible with BSD. A program which uses just setuid() will be
500 * 100% compatible with POSIX with saved IDs.
502 long __sys_setreuid(uid_t ruid
, uid_t euid
)
504 struct user_namespace
*ns
= current_user_ns();
505 const struct cred
*old
;
510 kruid
= make_kuid(ns
, ruid
);
511 keuid
= make_kuid(ns
, euid
);
513 if ((ruid
!= (uid_t
) -1) && !uid_valid(kruid
))
515 if ((euid
!= (uid_t
) -1) && !uid_valid(keuid
))
518 new = prepare_creds();
521 old
= current_cred();
524 if (ruid
!= (uid_t
) -1) {
526 if (!uid_eq(old
->uid
, kruid
) &&
527 !uid_eq(old
->euid
, kruid
) &&
528 !ns_capable_setid(old
->user_ns
, CAP_SETUID
))
532 if (euid
!= (uid_t
) -1) {
534 if (!uid_eq(old
->uid
, keuid
) &&
535 !uid_eq(old
->euid
, keuid
) &&
536 !uid_eq(old
->suid
, keuid
) &&
537 !ns_capable_setid(old
->user_ns
, CAP_SETUID
))
541 if (!uid_eq(new->uid
, old
->uid
)) {
542 retval
= set_user(new);
546 if (ruid
!= (uid_t
) -1 ||
547 (euid
!= (uid_t
) -1 && !uid_eq(keuid
, old
->uid
)))
548 new->suid
= new->euid
;
549 new->fsuid
= new->euid
;
551 retval
= security_task_fix_setuid(new, old
, LSM_SETID_RE
);
555 return commit_creds(new);
562 SYSCALL_DEFINE2(setreuid
, uid_t
, ruid
, uid_t
, euid
)
564 return __sys_setreuid(ruid
, euid
);
568 * setuid() is implemented like SysV with SAVED_IDS
570 * Note that SAVED_ID's is deficient in that a setuid root program
571 * like sendmail, for example, cannot set its uid to be a normal
572 * user and then switch back, because if you're root, setuid() sets
573 * the saved uid too. If you don't like this, blame the bright people
574 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
575 * will allow a root program to temporarily drop privileges and be able to
576 * regain them by swapping the real and effective uid.
578 long __sys_setuid(uid_t uid
)
580 struct user_namespace
*ns
= current_user_ns();
581 const struct cred
*old
;
586 kuid
= make_kuid(ns
, uid
);
587 if (!uid_valid(kuid
))
590 new = prepare_creds();
593 old
= current_cred();
596 if (ns_capable_setid(old
->user_ns
, CAP_SETUID
)) {
597 new->suid
= new->uid
= kuid
;
598 if (!uid_eq(kuid
, old
->uid
)) {
599 retval
= set_user(new);
603 } else if (!uid_eq(kuid
, old
->uid
) && !uid_eq(kuid
, new->suid
)) {
607 new->fsuid
= new->euid
= kuid
;
609 retval
= security_task_fix_setuid(new, old
, LSM_SETID_ID
);
613 return commit_creds(new);
620 SYSCALL_DEFINE1(setuid
, uid_t
, uid
)
622 return __sys_setuid(uid
);
627 * This function implements a generic ability to update ruid, euid,
628 * and suid. This allows you to implement the 4.4 compatible seteuid().
630 long __sys_setresuid(uid_t ruid
, uid_t euid
, uid_t suid
)
632 struct user_namespace
*ns
= current_user_ns();
633 const struct cred
*old
;
636 kuid_t kruid
, keuid
, ksuid
;
638 kruid
= make_kuid(ns
, ruid
);
639 keuid
= make_kuid(ns
, euid
);
640 ksuid
= make_kuid(ns
, suid
);
642 if ((ruid
!= (uid_t
) -1) && !uid_valid(kruid
))
645 if ((euid
!= (uid_t
) -1) && !uid_valid(keuid
))
648 if ((suid
!= (uid_t
) -1) && !uid_valid(ksuid
))
651 new = prepare_creds();
655 old
= current_cred();
658 if (!ns_capable_setid(old
->user_ns
, CAP_SETUID
)) {
659 if (ruid
!= (uid_t
) -1 && !uid_eq(kruid
, old
->uid
) &&
660 !uid_eq(kruid
, old
->euid
) && !uid_eq(kruid
, old
->suid
))
662 if (euid
!= (uid_t
) -1 && !uid_eq(keuid
, old
->uid
) &&
663 !uid_eq(keuid
, old
->euid
) && !uid_eq(keuid
, old
->suid
))
665 if (suid
!= (uid_t
) -1 && !uid_eq(ksuid
, old
->uid
) &&
666 !uid_eq(ksuid
, old
->euid
) && !uid_eq(ksuid
, old
->suid
))
670 if (ruid
!= (uid_t
) -1) {
672 if (!uid_eq(kruid
, old
->uid
)) {
673 retval
= set_user(new);
678 if (euid
!= (uid_t
) -1)
680 if (suid
!= (uid_t
) -1)
682 new->fsuid
= new->euid
;
684 retval
= security_task_fix_setuid(new, old
, LSM_SETID_RES
);
688 return commit_creds(new);
695 SYSCALL_DEFINE3(setresuid
, uid_t
, ruid
, uid_t
, euid
, uid_t
, suid
)
697 return __sys_setresuid(ruid
, euid
, suid
);
700 SYSCALL_DEFINE3(getresuid
, uid_t __user
*, ruidp
, uid_t __user
*, euidp
, uid_t __user
*, suidp
)
702 const struct cred
*cred
= current_cred();
704 uid_t ruid
, euid
, suid
;
706 ruid
= from_kuid_munged(cred
->user_ns
, cred
->uid
);
707 euid
= from_kuid_munged(cred
->user_ns
, cred
->euid
);
708 suid
= from_kuid_munged(cred
->user_ns
, cred
->suid
);
710 retval
= put_user(ruid
, ruidp
);
712 retval
= put_user(euid
, euidp
);
714 return put_user(suid
, suidp
);
720 * Same as above, but for rgid, egid, sgid.
722 long __sys_setresgid(gid_t rgid
, gid_t egid
, gid_t sgid
)
724 struct user_namespace
*ns
= current_user_ns();
725 const struct cred
*old
;
728 kgid_t krgid
, kegid
, ksgid
;
730 krgid
= make_kgid(ns
, rgid
);
731 kegid
= make_kgid(ns
, egid
);
732 ksgid
= make_kgid(ns
, sgid
);
734 if ((rgid
!= (gid_t
) -1) && !gid_valid(krgid
))
736 if ((egid
!= (gid_t
) -1) && !gid_valid(kegid
))
738 if ((sgid
!= (gid_t
) -1) && !gid_valid(ksgid
))
741 new = prepare_creds();
744 old
= current_cred();
747 if (!ns_capable_setid(old
->user_ns
, CAP_SETGID
)) {
748 if (rgid
!= (gid_t
) -1 && !gid_eq(krgid
, old
->gid
) &&
749 !gid_eq(krgid
, old
->egid
) && !gid_eq(krgid
, old
->sgid
))
751 if (egid
!= (gid_t
) -1 && !gid_eq(kegid
, old
->gid
) &&
752 !gid_eq(kegid
, old
->egid
) && !gid_eq(kegid
, old
->sgid
))
754 if (sgid
!= (gid_t
) -1 && !gid_eq(ksgid
, old
->gid
) &&
755 !gid_eq(ksgid
, old
->egid
) && !gid_eq(ksgid
, old
->sgid
))
759 if (rgid
!= (gid_t
) -1)
761 if (egid
!= (gid_t
) -1)
763 if (sgid
!= (gid_t
) -1)
765 new->fsgid
= new->egid
;
767 retval
= security_task_fix_setgid(new, old
, LSM_SETID_RES
);
771 return commit_creds(new);
778 SYSCALL_DEFINE3(setresgid
, gid_t
, rgid
, gid_t
, egid
, gid_t
, sgid
)
780 return __sys_setresgid(rgid
, egid
, sgid
);
783 SYSCALL_DEFINE3(getresgid
, gid_t __user
*, rgidp
, gid_t __user
*, egidp
, gid_t __user
*, sgidp
)
785 const struct cred
*cred
= current_cred();
787 gid_t rgid
, egid
, sgid
;
789 rgid
= from_kgid_munged(cred
->user_ns
, cred
->gid
);
790 egid
= from_kgid_munged(cred
->user_ns
, cred
->egid
);
791 sgid
= from_kgid_munged(cred
->user_ns
, cred
->sgid
);
793 retval
= put_user(rgid
, rgidp
);
795 retval
= put_user(egid
, egidp
);
797 retval
= put_user(sgid
, sgidp
);
805 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
806 * is used for "access()" and for the NFS daemon (letting nfsd stay at
807 * whatever uid it wants to). It normally shadows "euid", except when
808 * explicitly set by setfsuid() or for access..
810 long __sys_setfsuid(uid_t uid
)
812 const struct cred
*old
;
817 old
= current_cred();
818 old_fsuid
= from_kuid_munged(old
->user_ns
, old
->fsuid
);
820 kuid
= make_kuid(old
->user_ns
, uid
);
821 if (!uid_valid(kuid
))
824 new = prepare_creds();
828 if (uid_eq(kuid
, old
->uid
) || uid_eq(kuid
, old
->euid
) ||
829 uid_eq(kuid
, old
->suid
) || uid_eq(kuid
, old
->fsuid
) ||
830 ns_capable_setid(old
->user_ns
, CAP_SETUID
)) {
831 if (!uid_eq(kuid
, old
->fsuid
)) {
833 if (security_task_fix_setuid(new, old
, LSM_SETID_FS
) == 0)
846 SYSCALL_DEFINE1(setfsuid
, uid_t
, uid
)
848 return __sys_setfsuid(uid
);
852 * Samma på svenska..
854 long __sys_setfsgid(gid_t gid
)
856 const struct cred
*old
;
861 old
= current_cred();
862 old_fsgid
= from_kgid_munged(old
->user_ns
, old
->fsgid
);
864 kgid
= make_kgid(old
->user_ns
, gid
);
865 if (!gid_valid(kgid
))
868 new = prepare_creds();
872 if (gid_eq(kgid
, old
->gid
) || gid_eq(kgid
, old
->egid
) ||
873 gid_eq(kgid
, old
->sgid
) || gid_eq(kgid
, old
->fsgid
) ||
874 ns_capable_setid(old
->user_ns
, CAP_SETGID
)) {
875 if (!gid_eq(kgid
, old
->fsgid
)) {
877 if (security_task_fix_setgid(new,old
,LSM_SETID_FS
) == 0)
890 SYSCALL_DEFINE1(setfsgid
, gid_t
, gid
)
892 return __sys_setfsgid(gid
);
894 #endif /* CONFIG_MULTIUSER */
897 * sys_getpid - return the thread group id of the current process
899 * Note, despite the name, this returns the tgid not the pid. The tgid and
900 * the pid are identical unless CLONE_THREAD was specified on clone() in
901 * which case the tgid is the same in all threads of the same group.
903 * This is SMP safe as current->tgid does not change.
905 SYSCALL_DEFINE0(getpid
)
907 return task_tgid_vnr(current
);
910 /* Thread ID - the internal kernel "pid" */
911 SYSCALL_DEFINE0(gettid
)
913 return task_pid_vnr(current
);
917 * Accessing ->real_parent is not SMP-safe, it could
918 * change from under us. However, we can use a stale
919 * value of ->real_parent under rcu_read_lock(), see
920 * release_task()->call_rcu(delayed_put_task_struct).
922 SYSCALL_DEFINE0(getppid
)
927 pid
= task_tgid_vnr(rcu_dereference(current
->real_parent
));
933 SYSCALL_DEFINE0(getuid
)
935 /* Only we change this so SMP safe */
936 return from_kuid_munged(current_user_ns(), current_uid());
939 SYSCALL_DEFINE0(geteuid
)
941 /* Only we change this so SMP safe */
942 return from_kuid_munged(current_user_ns(), current_euid());
945 SYSCALL_DEFINE0(getgid
)
947 /* Only we change this so SMP safe */
948 return from_kgid_munged(current_user_ns(), current_gid());
951 SYSCALL_DEFINE0(getegid
)
953 /* Only we change this so SMP safe */
954 return from_kgid_munged(current_user_ns(), current_egid());
957 static void do_sys_times(struct tms
*tms
)
959 u64 tgutime
, tgstime
, cutime
, cstime
;
961 thread_group_cputime_adjusted(current
, &tgutime
, &tgstime
);
962 cutime
= current
->signal
->cutime
;
963 cstime
= current
->signal
->cstime
;
964 tms
->tms_utime
= nsec_to_clock_t(tgutime
);
965 tms
->tms_stime
= nsec_to_clock_t(tgstime
);
966 tms
->tms_cutime
= nsec_to_clock_t(cutime
);
967 tms
->tms_cstime
= nsec_to_clock_t(cstime
);
970 SYSCALL_DEFINE1(times
, struct tms __user
*, tbuf
)
976 if (copy_to_user(tbuf
, &tmp
, sizeof(struct tms
)))
979 force_successful_syscall_return();
980 return (long) jiffies_64_to_clock_t(get_jiffies_64());
984 static compat_clock_t
clock_t_to_compat_clock_t(clock_t x
)
986 return compat_jiffies_to_clock_t(clock_t_to_jiffies(x
));
989 COMPAT_SYSCALL_DEFINE1(times
, struct compat_tms __user
*, tbuf
)
993 struct compat_tms tmp
;
996 /* Convert our struct tms to the compat version. */
997 tmp
.tms_utime
= clock_t_to_compat_clock_t(tms
.tms_utime
);
998 tmp
.tms_stime
= clock_t_to_compat_clock_t(tms
.tms_stime
);
999 tmp
.tms_cutime
= clock_t_to_compat_clock_t(tms
.tms_cutime
);
1000 tmp
.tms_cstime
= clock_t_to_compat_clock_t(tms
.tms_cstime
);
1001 if (copy_to_user(tbuf
, &tmp
, sizeof(tmp
)))
1004 force_successful_syscall_return();
1005 return compat_jiffies_to_clock_t(jiffies
);
1010 * This needs some heavy checking ...
1011 * I just haven't the stomach for it. I also don't fully
1012 * understand sessions/pgrp etc. Let somebody who does explain it.
1014 * OK, I think I have the protection semantics right.... this is really
1015 * only important on a multi-user system anyway, to make sure one user
1016 * can't send a signal to a process owned by another. -TYT, 12/12/91
1018 * !PF_FORKNOEXEC check to conform completely to POSIX.
1020 SYSCALL_DEFINE2(setpgid
, pid_t
, pid
, pid_t
, pgid
)
1022 struct task_struct
*p
;
1023 struct task_struct
*group_leader
= current
->group_leader
;
1028 pid
= task_pid_vnr(group_leader
);
1035 /* From this point forward we keep holding onto the tasklist lock
1036 * so that our parent does not change from under us. -DaveM
1038 write_lock_irq(&tasklist_lock
);
1041 p
= find_task_by_vpid(pid
);
1046 if (!thread_group_leader(p
))
1049 if (same_thread_group(p
->real_parent
, group_leader
)) {
1051 if (task_session(p
) != task_session(group_leader
))
1054 if (!(p
->flags
& PF_FORKNOEXEC
))
1058 if (p
!= group_leader
)
1063 if (p
->signal
->leader
)
1068 struct task_struct
*g
;
1070 pgrp
= find_vpid(pgid
);
1071 g
= pid_task(pgrp
, PIDTYPE_PGID
);
1072 if (!g
|| task_session(g
) != task_session(group_leader
))
1076 err
= security_task_setpgid(p
, pgid
);
1080 if (task_pgrp(p
) != pgrp
)
1081 change_pid(p
, PIDTYPE_PGID
, pgrp
);
1085 /* All paths lead to here, thus we are safe. -DaveM */
1086 write_unlock_irq(&tasklist_lock
);
1091 static int do_getpgid(pid_t pid
)
1093 struct task_struct
*p
;
1099 grp
= task_pgrp(current
);
1102 p
= find_task_by_vpid(pid
);
1109 retval
= security_task_getpgid(p
);
1113 retval
= pid_vnr(grp
);
1119 SYSCALL_DEFINE1(getpgid
, pid_t
, pid
)
1121 return do_getpgid(pid
);
1124 #ifdef __ARCH_WANT_SYS_GETPGRP
1126 SYSCALL_DEFINE0(getpgrp
)
1128 return do_getpgid(0);
1133 SYSCALL_DEFINE1(getsid
, pid_t
, pid
)
1135 struct task_struct
*p
;
1141 sid
= task_session(current
);
1144 p
= find_task_by_vpid(pid
);
1147 sid
= task_session(p
);
1151 retval
= security_task_getsid(p
);
1155 retval
= pid_vnr(sid
);
1161 static void set_special_pids(struct pid
*pid
)
1163 struct task_struct
*curr
= current
->group_leader
;
1165 if (task_session(curr
) != pid
)
1166 change_pid(curr
, PIDTYPE_SID
, pid
);
1168 if (task_pgrp(curr
) != pid
)
1169 change_pid(curr
, PIDTYPE_PGID
, pid
);
1172 int ksys_setsid(void)
1174 struct task_struct
*group_leader
= current
->group_leader
;
1175 struct pid
*sid
= task_pid(group_leader
);
1176 pid_t session
= pid_vnr(sid
);
1179 write_lock_irq(&tasklist_lock
);
1180 /* Fail if I am already a session leader */
1181 if (group_leader
->signal
->leader
)
1184 /* Fail if a process group id already exists that equals the
1185 * proposed session id.
1187 if (pid_task(sid
, PIDTYPE_PGID
))
1190 group_leader
->signal
->leader
= 1;
1191 set_special_pids(sid
);
1193 proc_clear_tty(group_leader
);
1197 write_unlock_irq(&tasklist_lock
);
1199 proc_sid_connector(group_leader
);
1200 sched_autogroup_create_attach(group_leader
);
1205 SYSCALL_DEFINE0(setsid
)
1207 return ksys_setsid();
1210 DECLARE_RWSEM(uts_sem
);
1212 #ifdef COMPAT_UTS_MACHINE
1213 static char compat_uts_machine
[__OLD_UTS_LEN
+1] = COMPAT_UTS_MACHINE
;
1215 static int __init
parse_compat_uts_machine(char *arg
)
1217 strncpy(compat_uts_machine
, arg
, __OLD_UTS_LEN
);
1218 compat_uts_machine
[__OLD_UTS_LEN
] = 0;
1221 early_param("compat_uts_machine", parse_compat_uts_machine
);
1223 #undef COMPAT_UTS_MACHINE
1224 #define COMPAT_UTS_MACHINE compat_uts_machine
1227 #ifdef COMPAT_UTS_MACHINE
1228 #define override_architecture(name) \
1229 (personality(current->personality) == PER_LINUX32 && \
1230 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1231 sizeof(COMPAT_UTS_MACHINE)))
1233 #define override_architecture(name) 0
1237 * Work around broken programs that cannot handle "Linux 3.0".
1238 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1239 * And we map 4.x and later versions to 2.6.60+x, so 4.0/5.0/6.0/... would be
1242 static int override_release(char __user
*release
, size_t len
)
1246 if (current
->personality
& UNAME26
) {
1247 const char *rest
= UTS_RELEASE
;
1248 char buf
[65] = { 0 };
1254 if (*rest
== '.' && ++ndots
>= 3)
1256 if (!isdigit(*rest
) && *rest
!= '.')
1260 v
= ((LINUX_VERSION_CODE
>> 8) & 0xff) + 60;
1261 copy
= clamp_t(size_t, len
, 1, sizeof(buf
));
1262 copy
= scnprintf(buf
, copy
, "2.6.%u%s", v
, rest
);
1263 ret
= copy_to_user(release
, buf
, copy
+ 1);
1268 SYSCALL_DEFINE1(newuname
, struct new_utsname __user
*, name
)
1270 struct new_utsname tmp
;
1272 down_read(&uts_sem
);
1273 memcpy(&tmp
, utsname(), sizeof(tmp
));
1275 if (copy_to_user(name
, &tmp
, sizeof(tmp
)))
1278 if (override_release(name
->release
, sizeof(name
->release
)))
1280 if (override_architecture(name
))
1285 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1289 SYSCALL_DEFINE1(uname
, struct old_utsname __user
*, name
)
1291 struct old_utsname tmp
;
1296 down_read(&uts_sem
);
1297 memcpy(&tmp
, utsname(), sizeof(tmp
));
1299 if (copy_to_user(name
, &tmp
, sizeof(tmp
)))
1302 if (override_release(name
->release
, sizeof(name
->release
)))
1304 if (override_architecture(name
))
1309 SYSCALL_DEFINE1(olduname
, struct oldold_utsname __user
*, name
)
1311 struct oldold_utsname tmp
;
1316 memset(&tmp
, 0, sizeof(tmp
));
1318 down_read(&uts_sem
);
1319 memcpy(&tmp
.sysname
, &utsname()->sysname
, __OLD_UTS_LEN
);
1320 memcpy(&tmp
.nodename
, &utsname()->nodename
, __OLD_UTS_LEN
);
1321 memcpy(&tmp
.release
, &utsname()->release
, __OLD_UTS_LEN
);
1322 memcpy(&tmp
.version
, &utsname()->version
, __OLD_UTS_LEN
);
1323 memcpy(&tmp
.machine
, &utsname()->machine
, __OLD_UTS_LEN
);
1325 if (copy_to_user(name
, &tmp
, sizeof(tmp
)))
1328 if (override_architecture(name
))
1330 if (override_release(name
->release
, sizeof(name
->release
)))
1336 SYSCALL_DEFINE2(sethostname
, char __user
*, name
, int, len
)
1339 char tmp
[__NEW_UTS_LEN
];
1341 if (!ns_capable(current
->nsproxy
->uts_ns
->user_ns
, CAP_SYS_ADMIN
))
1344 if (len
< 0 || len
> __NEW_UTS_LEN
)
1347 if (!copy_from_user(tmp
, name
, len
)) {
1348 struct new_utsname
*u
;
1350 down_write(&uts_sem
);
1352 memcpy(u
->nodename
, tmp
, len
);
1353 memset(u
->nodename
+ len
, 0, sizeof(u
->nodename
) - len
);
1355 uts_proc_notify(UTS_PROC_HOSTNAME
);
1361 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1363 SYSCALL_DEFINE2(gethostname
, char __user
*, name
, int, len
)
1366 struct new_utsname
*u
;
1367 char tmp
[__NEW_UTS_LEN
+ 1];
1371 down_read(&uts_sem
);
1373 i
= 1 + strlen(u
->nodename
);
1376 memcpy(tmp
, u
->nodename
, i
);
1378 if (copy_to_user(name
, tmp
, i
))
1386 * Only setdomainname; getdomainname can be implemented by calling
1389 SYSCALL_DEFINE2(setdomainname
, char __user
*, name
, int, len
)
1392 char tmp
[__NEW_UTS_LEN
];
1394 if (!ns_capable(current
->nsproxy
->uts_ns
->user_ns
, CAP_SYS_ADMIN
))
1396 if (len
< 0 || len
> __NEW_UTS_LEN
)
1400 if (!copy_from_user(tmp
, name
, len
)) {
1401 struct new_utsname
*u
;
1403 down_write(&uts_sem
);
1405 memcpy(u
->domainname
, tmp
, len
);
1406 memset(u
->domainname
+ len
, 0, sizeof(u
->domainname
) - len
);
1408 uts_proc_notify(UTS_PROC_DOMAINNAME
);
1414 SYSCALL_DEFINE2(getrlimit
, unsigned int, resource
, struct rlimit __user
*, rlim
)
1416 struct rlimit value
;
1419 ret
= do_prlimit(current
, resource
, NULL
, &value
);
1421 ret
= copy_to_user(rlim
, &value
, sizeof(*rlim
)) ? -EFAULT
: 0;
1426 #ifdef CONFIG_COMPAT
1428 COMPAT_SYSCALL_DEFINE2(setrlimit
, unsigned int, resource
,
1429 struct compat_rlimit __user
*, rlim
)
1432 struct compat_rlimit r32
;
1434 if (copy_from_user(&r32
, rlim
, sizeof(struct compat_rlimit
)))
1437 if (r32
.rlim_cur
== COMPAT_RLIM_INFINITY
)
1438 r
.rlim_cur
= RLIM_INFINITY
;
1440 r
.rlim_cur
= r32
.rlim_cur
;
1441 if (r32
.rlim_max
== COMPAT_RLIM_INFINITY
)
1442 r
.rlim_max
= RLIM_INFINITY
;
1444 r
.rlim_max
= r32
.rlim_max
;
1445 return do_prlimit(current
, resource
, &r
, NULL
);
1448 COMPAT_SYSCALL_DEFINE2(getrlimit
, unsigned int, resource
,
1449 struct compat_rlimit __user
*, rlim
)
1454 ret
= do_prlimit(current
, resource
, NULL
, &r
);
1456 struct compat_rlimit r32
;
1457 if (r
.rlim_cur
> COMPAT_RLIM_INFINITY
)
1458 r32
.rlim_cur
= COMPAT_RLIM_INFINITY
;
1460 r32
.rlim_cur
= r
.rlim_cur
;
1461 if (r
.rlim_max
> COMPAT_RLIM_INFINITY
)
1462 r32
.rlim_max
= COMPAT_RLIM_INFINITY
;
1464 r32
.rlim_max
= r
.rlim_max
;
1466 if (copy_to_user(rlim
, &r32
, sizeof(struct compat_rlimit
)))
1474 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1477 * Back compatibility for getrlimit. Needed for some apps.
1479 SYSCALL_DEFINE2(old_getrlimit
, unsigned int, resource
,
1480 struct rlimit __user
*, rlim
)
1483 if (resource
>= RLIM_NLIMITS
)
1486 resource
= array_index_nospec(resource
, RLIM_NLIMITS
);
1487 task_lock(current
->group_leader
);
1488 x
= current
->signal
->rlim
[resource
];
1489 task_unlock(current
->group_leader
);
1490 if (x
.rlim_cur
> 0x7FFFFFFF)
1491 x
.rlim_cur
= 0x7FFFFFFF;
1492 if (x
.rlim_max
> 0x7FFFFFFF)
1493 x
.rlim_max
= 0x7FFFFFFF;
1494 return copy_to_user(rlim
, &x
, sizeof(x
)) ? -EFAULT
: 0;
1497 #ifdef CONFIG_COMPAT
1498 COMPAT_SYSCALL_DEFINE2(old_getrlimit
, unsigned int, resource
,
1499 struct compat_rlimit __user
*, rlim
)
1503 if (resource
>= RLIM_NLIMITS
)
1506 resource
= array_index_nospec(resource
, RLIM_NLIMITS
);
1507 task_lock(current
->group_leader
);
1508 r
= current
->signal
->rlim
[resource
];
1509 task_unlock(current
->group_leader
);
1510 if (r
.rlim_cur
> 0x7FFFFFFF)
1511 r
.rlim_cur
= 0x7FFFFFFF;
1512 if (r
.rlim_max
> 0x7FFFFFFF)
1513 r
.rlim_max
= 0x7FFFFFFF;
1515 if (put_user(r
.rlim_cur
, &rlim
->rlim_cur
) ||
1516 put_user(r
.rlim_max
, &rlim
->rlim_max
))
1524 static inline bool rlim64_is_infinity(__u64 rlim64
)
1526 #if BITS_PER_LONG < 64
1527 return rlim64
>= ULONG_MAX
;
1529 return rlim64
== RLIM64_INFINITY
;
1533 static void rlim_to_rlim64(const struct rlimit
*rlim
, struct rlimit64
*rlim64
)
1535 if (rlim
->rlim_cur
== RLIM_INFINITY
)
1536 rlim64
->rlim_cur
= RLIM64_INFINITY
;
1538 rlim64
->rlim_cur
= rlim
->rlim_cur
;
1539 if (rlim
->rlim_max
== RLIM_INFINITY
)
1540 rlim64
->rlim_max
= RLIM64_INFINITY
;
1542 rlim64
->rlim_max
= rlim
->rlim_max
;
1545 static void rlim64_to_rlim(const struct rlimit64
*rlim64
, struct rlimit
*rlim
)
1547 if (rlim64_is_infinity(rlim64
->rlim_cur
))
1548 rlim
->rlim_cur
= RLIM_INFINITY
;
1550 rlim
->rlim_cur
= (unsigned long)rlim64
->rlim_cur
;
1551 if (rlim64_is_infinity(rlim64
->rlim_max
))
1552 rlim
->rlim_max
= RLIM_INFINITY
;
1554 rlim
->rlim_max
= (unsigned long)rlim64
->rlim_max
;
1557 /* make sure you are allowed to change @tsk limits before calling this */
1558 int do_prlimit(struct task_struct
*tsk
, unsigned int resource
,
1559 struct rlimit
*new_rlim
, struct rlimit
*old_rlim
)
1561 struct rlimit
*rlim
;
1564 if (resource
>= RLIM_NLIMITS
)
1567 if (new_rlim
->rlim_cur
> new_rlim
->rlim_max
)
1569 if (resource
== RLIMIT_NOFILE
&&
1570 new_rlim
->rlim_max
> sysctl_nr_open
)
1574 /* protect tsk->signal and tsk->sighand from disappearing */
1575 read_lock(&tasklist_lock
);
1576 if (!tsk
->sighand
) {
1581 rlim
= tsk
->signal
->rlim
+ resource
;
1582 task_lock(tsk
->group_leader
);
1584 /* Keep the capable check against init_user_ns until
1585 cgroups can contain all limits */
1586 if (new_rlim
->rlim_max
> rlim
->rlim_max
&&
1587 !capable(CAP_SYS_RESOURCE
))
1590 retval
= security_task_setrlimit(tsk
, resource
, new_rlim
);
1598 task_unlock(tsk
->group_leader
);
1601 * RLIMIT_CPU handling. Arm the posix CPU timer if the limit is not
1602 * infite. In case of RLIM_INFINITY the posix CPU timer code
1603 * ignores the rlimit.
1605 if (!retval
&& new_rlim
&& resource
== RLIMIT_CPU
&&
1606 new_rlim
->rlim_cur
!= RLIM_INFINITY
&&
1607 IS_ENABLED(CONFIG_POSIX_TIMERS
))
1608 update_rlimit_cpu(tsk
, new_rlim
->rlim_cur
);
1610 read_unlock(&tasklist_lock
);
1614 /* rcu lock must be held */
1615 static int check_prlimit_permission(struct task_struct
*task
,
1618 const struct cred
*cred
= current_cred(), *tcred
;
1621 if (current
== task
)
1624 tcred
= __task_cred(task
);
1625 id_match
= (uid_eq(cred
->uid
, tcred
->euid
) &&
1626 uid_eq(cred
->uid
, tcred
->suid
) &&
1627 uid_eq(cred
->uid
, tcred
->uid
) &&
1628 gid_eq(cred
->gid
, tcred
->egid
) &&
1629 gid_eq(cred
->gid
, tcred
->sgid
) &&
1630 gid_eq(cred
->gid
, tcred
->gid
));
1631 if (!id_match
&& !ns_capable(tcred
->user_ns
, CAP_SYS_RESOURCE
))
1634 return security_task_prlimit(cred
, tcred
, flags
);
1637 SYSCALL_DEFINE4(prlimit64
, pid_t
, pid
, unsigned int, resource
,
1638 const struct rlimit64 __user
*, new_rlim
,
1639 struct rlimit64 __user
*, old_rlim
)
1641 struct rlimit64 old64
, new64
;
1642 struct rlimit old
, new;
1643 struct task_struct
*tsk
;
1644 unsigned int checkflags
= 0;
1648 checkflags
|= LSM_PRLIMIT_READ
;
1651 if (copy_from_user(&new64
, new_rlim
, sizeof(new64
)))
1653 rlim64_to_rlim(&new64
, &new);
1654 checkflags
|= LSM_PRLIMIT_WRITE
;
1658 tsk
= pid
? find_task_by_vpid(pid
) : current
;
1663 ret
= check_prlimit_permission(tsk
, checkflags
);
1668 get_task_struct(tsk
);
1671 ret
= do_prlimit(tsk
, resource
, new_rlim
? &new : NULL
,
1672 old_rlim
? &old
: NULL
);
1674 if (!ret
&& old_rlim
) {
1675 rlim_to_rlim64(&old
, &old64
);
1676 if (copy_to_user(old_rlim
, &old64
, sizeof(old64
)))
1680 put_task_struct(tsk
);
1684 SYSCALL_DEFINE2(setrlimit
, unsigned int, resource
, struct rlimit __user
*, rlim
)
1686 struct rlimit new_rlim
;
1688 if (copy_from_user(&new_rlim
, rlim
, sizeof(*rlim
)))
1690 return do_prlimit(current
, resource
, &new_rlim
, NULL
);
1694 * It would make sense to put struct rusage in the task_struct,
1695 * except that would make the task_struct be *really big*. After
1696 * task_struct gets moved into malloc'ed memory, it would
1697 * make sense to do this. It will make moving the rest of the information
1698 * a lot simpler! (Which we're not doing right now because we're not
1699 * measuring them yet).
1701 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1702 * races with threads incrementing their own counters. But since word
1703 * reads are atomic, we either get new values or old values and we don't
1704 * care which for the sums. We always take the siglock to protect reading
1705 * the c* fields from p->signal from races with exit.c updating those
1706 * fields when reaping, so a sample either gets all the additions of a
1707 * given child after it's reaped, or none so this sample is before reaping.
1710 * We need to take the siglock for CHILDEREN, SELF and BOTH
1711 * for the cases current multithreaded, non-current single threaded
1712 * non-current multithreaded. Thread traversal is now safe with
1714 * Strictly speaking, we donot need to take the siglock if we are current and
1715 * single threaded, as no one else can take our signal_struct away, no one
1716 * else can reap the children to update signal->c* counters, and no one else
1717 * can race with the signal-> fields. If we do not take any lock, the
1718 * signal-> fields could be read out of order while another thread was just
1719 * exiting. So we should place a read memory barrier when we avoid the lock.
1720 * On the writer side, write memory barrier is implied in __exit_signal
1721 * as __exit_signal releases the siglock spinlock after updating the signal->
1722 * fields. But we don't do this yet to keep things simple.
1726 static void accumulate_thread_rusage(struct task_struct
*t
, struct rusage
*r
)
1728 r
->ru_nvcsw
+= t
->nvcsw
;
1729 r
->ru_nivcsw
+= t
->nivcsw
;
1730 r
->ru_minflt
+= t
->min_flt
;
1731 r
->ru_majflt
+= t
->maj_flt
;
1732 r
->ru_inblock
+= task_io_get_inblock(t
);
1733 r
->ru_oublock
+= task_io_get_oublock(t
);
1736 void getrusage(struct task_struct
*p
, int who
, struct rusage
*r
)
1738 struct task_struct
*t
;
1739 unsigned long flags
;
1740 u64 tgutime
, tgstime
, utime
, stime
;
1741 unsigned long maxrss
= 0;
1743 memset((char *)r
, 0, sizeof (*r
));
1746 if (who
== RUSAGE_THREAD
) {
1747 task_cputime_adjusted(current
, &utime
, &stime
);
1748 accumulate_thread_rusage(p
, r
);
1749 maxrss
= p
->signal
->maxrss
;
1753 if (!lock_task_sighand(p
, &flags
))
1758 case RUSAGE_CHILDREN
:
1759 utime
= p
->signal
->cutime
;
1760 stime
= p
->signal
->cstime
;
1761 r
->ru_nvcsw
= p
->signal
->cnvcsw
;
1762 r
->ru_nivcsw
= p
->signal
->cnivcsw
;
1763 r
->ru_minflt
= p
->signal
->cmin_flt
;
1764 r
->ru_majflt
= p
->signal
->cmaj_flt
;
1765 r
->ru_inblock
= p
->signal
->cinblock
;
1766 r
->ru_oublock
= p
->signal
->coublock
;
1767 maxrss
= p
->signal
->cmaxrss
;
1769 if (who
== RUSAGE_CHILDREN
)
1774 thread_group_cputime_adjusted(p
, &tgutime
, &tgstime
);
1777 r
->ru_nvcsw
+= p
->signal
->nvcsw
;
1778 r
->ru_nivcsw
+= p
->signal
->nivcsw
;
1779 r
->ru_minflt
+= p
->signal
->min_flt
;
1780 r
->ru_majflt
+= p
->signal
->maj_flt
;
1781 r
->ru_inblock
+= p
->signal
->inblock
;
1782 r
->ru_oublock
+= p
->signal
->oublock
;
1783 if (maxrss
< p
->signal
->maxrss
)
1784 maxrss
= p
->signal
->maxrss
;
1787 accumulate_thread_rusage(t
, r
);
1788 } while_each_thread(p
, t
);
1794 unlock_task_sighand(p
, &flags
);
1797 r
->ru_utime
= ns_to_kernel_old_timeval(utime
);
1798 r
->ru_stime
= ns_to_kernel_old_timeval(stime
);
1800 if (who
!= RUSAGE_CHILDREN
) {
1801 struct mm_struct
*mm
= get_task_mm(p
);
1804 setmax_mm_hiwater_rss(&maxrss
, mm
);
1808 r
->ru_maxrss
= maxrss
* (PAGE_SIZE
/ 1024); /* convert pages to KBs */
1811 SYSCALL_DEFINE2(getrusage
, int, who
, struct rusage __user
*, ru
)
1815 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
&&
1816 who
!= RUSAGE_THREAD
)
1819 getrusage(current
, who
, &r
);
1820 return copy_to_user(ru
, &r
, sizeof(r
)) ? -EFAULT
: 0;
1823 #ifdef CONFIG_COMPAT
1824 COMPAT_SYSCALL_DEFINE2(getrusage
, int, who
, struct compat_rusage __user
*, ru
)
1828 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
&&
1829 who
!= RUSAGE_THREAD
)
1832 getrusage(current
, who
, &r
);
1833 return put_compat_rusage(&r
, ru
);
1837 SYSCALL_DEFINE1(umask
, int, mask
)
1839 mask
= xchg(¤t
->fs
->umask
, mask
& S_IRWXUGO
);
1843 static int prctl_set_mm_exe_file(struct mm_struct
*mm
, unsigned int fd
)
1846 struct file
*old_exe
, *exe_file
;
1847 struct inode
*inode
;
1854 inode
= file_inode(exe
.file
);
1857 * Because the original mm->exe_file points to executable file, make
1858 * sure that this one is executable as well, to avoid breaking an
1862 if (!S_ISREG(inode
->i_mode
) || path_noexec(&exe
.file
->f_path
))
1865 err
= inode_permission(inode
, MAY_EXEC
);
1870 * Forbid mm->exe_file change if old file still mapped.
1872 exe_file
= get_mm_exe_file(mm
);
1875 struct vm_area_struct
*vma
;
1878 for (vma
= mm
->mmap
; vma
; vma
= vma
->vm_next
) {
1881 if (path_equal(&vma
->vm_file
->f_path
,
1886 mmap_read_unlock(mm
);
1891 /* set the new file, lockless */
1893 old_exe
= xchg(&mm
->exe_file
, exe
.file
);
1900 mmap_read_unlock(mm
);
1906 * Check arithmetic relations of passed addresses.
1908 * WARNING: we don't require any capability here so be very careful
1909 * in what is allowed for modification from userspace.
1911 static int validate_prctl_map_addr(struct prctl_mm_map
*prctl_map
)
1913 unsigned long mmap_max_addr
= TASK_SIZE
;
1914 int error
= -EINVAL
, i
;
1916 static const unsigned char offsets
[] = {
1917 offsetof(struct prctl_mm_map
, start_code
),
1918 offsetof(struct prctl_mm_map
, end_code
),
1919 offsetof(struct prctl_mm_map
, start_data
),
1920 offsetof(struct prctl_mm_map
, end_data
),
1921 offsetof(struct prctl_mm_map
, start_brk
),
1922 offsetof(struct prctl_mm_map
, brk
),
1923 offsetof(struct prctl_mm_map
, start_stack
),
1924 offsetof(struct prctl_mm_map
, arg_start
),
1925 offsetof(struct prctl_mm_map
, arg_end
),
1926 offsetof(struct prctl_mm_map
, env_start
),
1927 offsetof(struct prctl_mm_map
, env_end
),
1931 * Make sure the members are not somewhere outside
1932 * of allowed address space.
1934 for (i
= 0; i
< ARRAY_SIZE(offsets
); i
++) {
1935 u64 val
= *(u64
*)((char *)prctl_map
+ offsets
[i
]);
1937 if ((unsigned long)val
>= mmap_max_addr
||
1938 (unsigned long)val
< mmap_min_addr
)
1943 * Make sure the pairs are ordered.
1945 #define __prctl_check_order(__m1, __op, __m2) \
1946 ((unsigned long)prctl_map->__m1 __op \
1947 (unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1948 error
= __prctl_check_order(start_code
, <, end_code
);
1949 error
|= __prctl_check_order(start_data
,<=, end_data
);
1950 error
|= __prctl_check_order(start_brk
, <=, brk
);
1951 error
|= __prctl_check_order(arg_start
, <=, arg_end
);
1952 error
|= __prctl_check_order(env_start
, <=, env_end
);
1955 #undef __prctl_check_order
1960 * @brk should be after @end_data in traditional maps.
1962 if (prctl_map
->start_brk
<= prctl_map
->end_data
||
1963 prctl_map
->brk
<= prctl_map
->end_data
)
1967 * Neither we should allow to override limits if they set.
1969 if (check_data_rlimit(rlimit(RLIMIT_DATA
), prctl_map
->brk
,
1970 prctl_map
->start_brk
, prctl_map
->end_data
,
1971 prctl_map
->start_data
))
1979 #ifdef CONFIG_CHECKPOINT_RESTORE
1980 static int prctl_set_mm_map(int opt
, const void __user
*addr
, unsigned long data_size
)
1982 struct prctl_mm_map prctl_map
= { .exe_fd
= (u32
)-1, };
1983 unsigned long user_auxv
[AT_VECTOR_SIZE
];
1984 struct mm_struct
*mm
= current
->mm
;
1987 BUILD_BUG_ON(sizeof(user_auxv
) != sizeof(mm
->saved_auxv
));
1988 BUILD_BUG_ON(sizeof(struct prctl_mm_map
) > 256);
1990 if (opt
== PR_SET_MM_MAP_SIZE
)
1991 return put_user((unsigned int)sizeof(prctl_map
),
1992 (unsigned int __user
*)addr
);
1994 if (data_size
!= sizeof(prctl_map
))
1997 if (copy_from_user(&prctl_map
, addr
, sizeof(prctl_map
)))
2000 error
= validate_prctl_map_addr(&prctl_map
);
2004 if (prctl_map
.auxv_size
) {
2006 * Someone is trying to cheat the auxv vector.
2008 if (!prctl_map
.auxv
||
2009 prctl_map
.auxv_size
> sizeof(mm
->saved_auxv
))
2012 memset(user_auxv
, 0, sizeof(user_auxv
));
2013 if (copy_from_user(user_auxv
,
2014 (const void __user
*)prctl_map
.auxv
,
2015 prctl_map
.auxv_size
))
2018 /* Last entry must be AT_NULL as specification requires */
2019 user_auxv
[AT_VECTOR_SIZE
- 2] = AT_NULL
;
2020 user_auxv
[AT_VECTOR_SIZE
- 1] = AT_NULL
;
2023 if (prctl_map
.exe_fd
!= (u32
)-1) {
2025 * Check if the current user is checkpoint/restore capable.
2026 * At the time of this writing, it checks for CAP_SYS_ADMIN
2027 * or CAP_CHECKPOINT_RESTORE.
2028 * Note that a user with access to ptrace can masquerade an
2029 * arbitrary program as any executable, even setuid ones.
2030 * This may have implications in the tomoyo subsystem.
2032 if (!checkpoint_restore_ns_capable(current_user_ns()))
2035 error
= prctl_set_mm_exe_file(mm
, prctl_map
.exe_fd
);
2041 * arg_lock protects concurent updates but we still need mmap_lock for
2042 * read to exclude races with sys_brk.
2047 * We don't validate if these members are pointing to
2048 * real present VMAs because application may have correspond
2049 * VMAs already unmapped and kernel uses these members for statistics
2050 * output in procfs mostly, except
2052 * - @start_brk/@brk which are used in do_brk_flags but kernel lookups
2053 * for VMAs when updating these memvers so anything wrong written
2054 * here cause kernel to swear at userspace program but won't lead
2055 * to any problem in kernel itself
2058 spin_lock(&mm
->arg_lock
);
2059 mm
->start_code
= prctl_map
.start_code
;
2060 mm
->end_code
= prctl_map
.end_code
;
2061 mm
->start_data
= prctl_map
.start_data
;
2062 mm
->end_data
= prctl_map
.end_data
;
2063 mm
->start_brk
= prctl_map
.start_brk
;
2064 mm
->brk
= prctl_map
.brk
;
2065 mm
->start_stack
= prctl_map
.start_stack
;
2066 mm
->arg_start
= prctl_map
.arg_start
;
2067 mm
->arg_end
= prctl_map
.arg_end
;
2068 mm
->env_start
= prctl_map
.env_start
;
2069 mm
->env_end
= prctl_map
.env_end
;
2070 spin_unlock(&mm
->arg_lock
);
2073 * Note this update of @saved_auxv is lockless thus
2074 * if someone reads this member in procfs while we're
2075 * updating -- it may get partly updated results. It's
2076 * known and acceptable trade off: we leave it as is to
2077 * not introduce additional locks here making the kernel
2080 if (prctl_map
.auxv_size
)
2081 memcpy(mm
->saved_auxv
, user_auxv
, sizeof(user_auxv
));
2083 mmap_read_unlock(mm
);
2086 #endif /* CONFIG_CHECKPOINT_RESTORE */
2088 static int prctl_set_auxv(struct mm_struct
*mm
, unsigned long addr
,
2092 * This doesn't move the auxiliary vector itself since it's pinned to
2093 * mm_struct, but it permits filling the vector with new values. It's
2094 * up to the caller to provide sane values here, otherwise userspace
2095 * tools which use this vector might be unhappy.
2097 unsigned long user_auxv
[AT_VECTOR_SIZE
];
2099 if (len
> sizeof(user_auxv
))
2102 if (copy_from_user(user_auxv
, (const void __user
*)addr
, len
))
2105 /* Make sure the last entry is always AT_NULL */
2106 user_auxv
[AT_VECTOR_SIZE
- 2] = 0;
2107 user_auxv
[AT_VECTOR_SIZE
- 1] = 0;
2109 BUILD_BUG_ON(sizeof(user_auxv
) != sizeof(mm
->saved_auxv
));
2112 memcpy(mm
->saved_auxv
, user_auxv
, len
);
2113 task_unlock(current
);
2118 static int prctl_set_mm(int opt
, unsigned long addr
,
2119 unsigned long arg4
, unsigned long arg5
)
2121 struct mm_struct
*mm
= current
->mm
;
2122 struct prctl_mm_map prctl_map
= {
2127 struct vm_area_struct
*vma
;
2130 if (arg5
|| (arg4
&& (opt
!= PR_SET_MM_AUXV
&&
2131 opt
!= PR_SET_MM_MAP
&&
2132 opt
!= PR_SET_MM_MAP_SIZE
)))
2135 #ifdef CONFIG_CHECKPOINT_RESTORE
2136 if (opt
== PR_SET_MM_MAP
|| opt
== PR_SET_MM_MAP_SIZE
)
2137 return prctl_set_mm_map(opt
, (const void __user
*)addr
, arg4
);
2140 if (!capable(CAP_SYS_RESOURCE
))
2143 if (opt
== PR_SET_MM_EXE_FILE
)
2144 return prctl_set_mm_exe_file(mm
, (unsigned int)addr
);
2146 if (opt
== PR_SET_MM_AUXV
)
2147 return prctl_set_auxv(mm
, addr
, arg4
);
2149 if (addr
>= TASK_SIZE
|| addr
< mmap_min_addr
)
2155 * arg_lock protects concurent updates of arg boundaries, we need
2156 * mmap_lock for a) concurrent sys_brk, b) finding VMA for addr
2160 vma
= find_vma(mm
, addr
);
2162 spin_lock(&mm
->arg_lock
);
2163 prctl_map
.start_code
= mm
->start_code
;
2164 prctl_map
.end_code
= mm
->end_code
;
2165 prctl_map
.start_data
= mm
->start_data
;
2166 prctl_map
.end_data
= mm
->end_data
;
2167 prctl_map
.start_brk
= mm
->start_brk
;
2168 prctl_map
.brk
= mm
->brk
;
2169 prctl_map
.start_stack
= mm
->start_stack
;
2170 prctl_map
.arg_start
= mm
->arg_start
;
2171 prctl_map
.arg_end
= mm
->arg_end
;
2172 prctl_map
.env_start
= mm
->env_start
;
2173 prctl_map
.env_end
= mm
->env_end
;
2176 case PR_SET_MM_START_CODE
:
2177 prctl_map
.start_code
= addr
;
2179 case PR_SET_MM_END_CODE
:
2180 prctl_map
.end_code
= addr
;
2182 case PR_SET_MM_START_DATA
:
2183 prctl_map
.start_data
= addr
;
2185 case PR_SET_MM_END_DATA
:
2186 prctl_map
.end_data
= addr
;
2188 case PR_SET_MM_START_STACK
:
2189 prctl_map
.start_stack
= addr
;
2191 case PR_SET_MM_START_BRK
:
2192 prctl_map
.start_brk
= addr
;
2195 prctl_map
.brk
= addr
;
2197 case PR_SET_MM_ARG_START
:
2198 prctl_map
.arg_start
= addr
;
2200 case PR_SET_MM_ARG_END
:
2201 prctl_map
.arg_end
= addr
;
2203 case PR_SET_MM_ENV_START
:
2204 prctl_map
.env_start
= addr
;
2206 case PR_SET_MM_ENV_END
:
2207 prctl_map
.env_end
= addr
;
2213 error
= validate_prctl_map_addr(&prctl_map
);
2219 * If command line arguments and environment
2220 * are placed somewhere else on stack, we can
2221 * set them up here, ARG_START/END to setup
2222 * command line argumets and ENV_START/END
2225 case PR_SET_MM_START_STACK
:
2226 case PR_SET_MM_ARG_START
:
2227 case PR_SET_MM_ARG_END
:
2228 case PR_SET_MM_ENV_START
:
2229 case PR_SET_MM_ENV_END
:
2236 mm
->start_code
= prctl_map
.start_code
;
2237 mm
->end_code
= prctl_map
.end_code
;
2238 mm
->start_data
= prctl_map
.start_data
;
2239 mm
->end_data
= prctl_map
.end_data
;
2240 mm
->start_brk
= prctl_map
.start_brk
;
2241 mm
->brk
= prctl_map
.brk
;
2242 mm
->start_stack
= prctl_map
.start_stack
;
2243 mm
->arg_start
= prctl_map
.arg_start
;
2244 mm
->arg_end
= prctl_map
.arg_end
;
2245 mm
->env_start
= prctl_map
.env_start
;
2246 mm
->env_end
= prctl_map
.env_end
;
2250 spin_unlock(&mm
->arg_lock
);
2251 mmap_read_unlock(mm
);
2255 #ifdef CONFIG_CHECKPOINT_RESTORE
2256 static int prctl_get_tid_address(struct task_struct
*me
, int __user
* __user
*tid_addr
)
2258 return put_user(me
->clear_child_tid
, tid_addr
);
2261 static int prctl_get_tid_address(struct task_struct
*me
, int __user
* __user
*tid_addr
)
2267 static int propagate_has_child_subreaper(struct task_struct
*p
, void *data
)
2270 * If task has has_child_subreaper - all its decendants
2271 * already have these flag too and new decendants will
2272 * inherit it on fork, skip them.
2274 * If we've found child_reaper - skip descendants in
2275 * it's subtree as they will never get out pidns.
2277 if (p
->signal
->has_child_subreaper
||
2278 is_child_reaper(task_pid(p
)))
2281 p
->signal
->has_child_subreaper
= 1;
2285 int __weak
arch_prctl_spec_ctrl_get(struct task_struct
*t
, unsigned long which
)
2290 int __weak
arch_prctl_spec_ctrl_set(struct task_struct
*t
, unsigned long which
,
2296 #define PR_IO_FLUSHER (PF_MEMALLOC_NOIO | PF_LOCAL_THROTTLE)
2298 SYSCALL_DEFINE5(prctl
, int, option
, unsigned long, arg2
, unsigned long, arg3
,
2299 unsigned long, arg4
, unsigned long, arg5
)
2301 struct task_struct
*me
= current
;
2302 unsigned char comm
[sizeof(me
->comm
)];
2305 error
= security_task_prctl(option
, arg2
, arg3
, arg4
, arg5
);
2306 if (error
!= -ENOSYS
)
2311 case PR_SET_PDEATHSIG
:
2312 if (!valid_signal(arg2
)) {
2316 me
->pdeath_signal
= arg2
;
2318 case PR_GET_PDEATHSIG
:
2319 error
= put_user(me
->pdeath_signal
, (int __user
*)arg2
);
2321 case PR_GET_DUMPABLE
:
2322 error
= get_dumpable(me
->mm
);
2324 case PR_SET_DUMPABLE
:
2325 if (arg2
!= SUID_DUMP_DISABLE
&& arg2
!= SUID_DUMP_USER
) {
2329 set_dumpable(me
->mm
, arg2
);
2332 case PR_SET_UNALIGN
:
2333 error
= SET_UNALIGN_CTL(me
, arg2
);
2335 case PR_GET_UNALIGN
:
2336 error
= GET_UNALIGN_CTL(me
, arg2
);
2339 error
= SET_FPEMU_CTL(me
, arg2
);
2342 error
= GET_FPEMU_CTL(me
, arg2
);
2345 error
= SET_FPEXC_CTL(me
, arg2
);
2348 error
= GET_FPEXC_CTL(me
, arg2
);
2351 error
= PR_TIMING_STATISTICAL
;
2354 if (arg2
!= PR_TIMING_STATISTICAL
)
2358 comm
[sizeof(me
->comm
) - 1] = 0;
2359 if (strncpy_from_user(comm
, (char __user
*)arg2
,
2360 sizeof(me
->comm
) - 1) < 0)
2362 set_task_comm(me
, comm
);
2363 proc_comm_connector(me
);
2366 get_task_comm(comm
, me
);
2367 if (copy_to_user((char __user
*)arg2
, comm
, sizeof(comm
)))
2371 error
= GET_ENDIAN(me
, arg2
);
2374 error
= SET_ENDIAN(me
, arg2
);
2376 case PR_GET_SECCOMP
:
2377 error
= prctl_get_seccomp();
2379 case PR_SET_SECCOMP
:
2380 error
= prctl_set_seccomp(arg2
, (char __user
*)arg3
);
2383 error
= GET_TSC_CTL(arg2
);
2386 error
= SET_TSC_CTL(arg2
);
2388 case PR_TASK_PERF_EVENTS_DISABLE
:
2389 error
= perf_event_task_disable();
2391 case PR_TASK_PERF_EVENTS_ENABLE
:
2392 error
= perf_event_task_enable();
2394 case PR_GET_TIMERSLACK
:
2395 if (current
->timer_slack_ns
> ULONG_MAX
)
2398 error
= current
->timer_slack_ns
;
2400 case PR_SET_TIMERSLACK
:
2402 current
->timer_slack_ns
=
2403 current
->default_timer_slack_ns
;
2405 current
->timer_slack_ns
= arg2
;
2411 case PR_MCE_KILL_CLEAR
:
2414 current
->flags
&= ~PF_MCE_PROCESS
;
2416 case PR_MCE_KILL_SET
:
2417 current
->flags
|= PF_MCE_PROCESS
;
2418 if (arg3
== PR_MCE_KILL_EARLY
)
2419 current
->flags
|= PF_MCE_EARLY
;
2420 else if (arg3
== PR_MCE_KILL_LATE
)
2421 current
->flags
&= ~PF_MCE_EARLY
;
2422 else if (arg3
== PR_MCE_KILL_DEFAULT
)
2424 ~(PF_MCE_EARLY
|PF_MCE_PROCESS
);
2432 case PR_MCE_KILL_GET
:
2433 if (arg2
| arg3
| arg4
| arg5
)
2435 if (current
->flags
& PF_MCE_PROCESS
)
2436 error
= (current
->flags
& PF_MCE_EARLY
) ?
2437 PR_MCE_KILL_EARLY
: PR_MCE_KILL_LATE
;
2439 error
= PR_MCE_KILL_DEFAULT
;
2442 error
= prctl_set_mm(arg2
, arg3
, arg4
, arg5
);
2444 case PR_GET_TID_ADDRESS
:
2445 error
= prctl_get_tid_address(me
, (int __user
* __user
*)arg2
);
2447 case PR_SET_CHILD_SUBREAPER
:
2448 me
->signal
->is_child_subreaper
= !!arg2
;
2452 walk_process_tree(me
, propagate_has_child_subreaper
, NULL
);
2454 case PR_GET_CHILD_SUBREAPER
:
2455 error
= put_user(me
->signal
->is_child_subreaper
,
2456 (int __user
*)arg2
);
2458 case PR_SET_NO_NEW_PRIVS
:
2459 if (arg2
!= 1 || arg3
|| arg4
|| arg5
)
2462 task_set_no_new_privs(current
);
2464 case PR_GET_NO_NEW_PRIVS
:
2465 if (arg2
|| arg3
|| arg4
|| arg5
)
2467 return task_no_new_privs(current
) ? 1 : 0;
2468 case PR_GET_THP_DISABLE
:
2469 if (arg2
|| arg3
|| arg4
|| arg5
)
2471 error
= !!test_bit(MMF_DISABLE_THP
, &me
->mm
->flags
);
2473 case PR_SET_THP_DISABLE
:
2474 if (arg3
|| arg4
|| arg5
)
2476 if (mmap_write_lock_killable(me
->mm
))
2479 set_bit(MMF_DISABLE_THP
, &me
->mm
->flags
);
2481 clear_bit(MMF_DISABLE_THP
, &me
->mm
->flags
);
2482 mmap_write_unlock(me
->mm
);
2484 case PR_MPX_ENABLE_MANAGEMENT
:
2485 case PR_MPX_DISABLE_MANAGEMENT
:
2486 /* No longer implemented: */
2488 case PR_SET_FP_MODE
:
2489 error
= SET_FP_MODE(me
, arg2
);
2491 case PR_GET_FP_MODE
:
2492 error
= GET_FP_MODE(me
);
2495 error
= SVE_SET_VL(arg2
);
2498 error
= SVE_GET_VL();
2500 case PR_GET_SPECULATION_CTRL
:
2501 if (arg3
|| arg4
|| arg5
)
2503 error
= arch_prctl_spec_ctrl_get(me
, arg2
);
2505 case PR_SET_SPECULATION_CTRL
:
2508 error
= arch_prctl_spec_ctrl_set(me
, arg2
, arg3
);
2510 case PR_PAC_RESET_KEYS
:
2511 if (arg3
|| arg4
|| arg5
)
2513 error
= PAC_RESET_KEYS(me
, arg2
);
2515 case PR_SET_TAGGED_ADDR_CTRL
:
2516 if (arg3
|| arg4
|| arg5
)
2518 error
= SET_TAGGED_ADDR_CTRL(arg2
);
2520 case PR_GET_TAGGED_ADDR_CTRL
:
2521 if (arg2
|| arg3
|| arg4
|| arg5
)
2523 error
= GET_TAGGED_ADDR_CTRL();
2525 case PR_SET_IO_FLUSHER
:
2526 if (!capable(CAP_SYS_RESOURCE
))
2529 if (arg3
|| arg4
|| arg5
)
2533 current
->flags
|= PR_IO_FLUSHER
;
2535 current
->flags
&= ~PR_IO_FLUSHER
;
2539 case PR_GET_IO_FLUSHER
:
2540 if (!capable(CAP_SYS_RESOURCE
))
2543 if (arg2
|| arg3
|| arg4
|| arg5
)
2546 error
= (current
->flags
& PR_IO_FLUSHER
) == PR_IO_FLUSHER
;
2555 SYSCALL_DEFINE3(getcpu
, unsigned __user
*, cpup
, unsigned __user
*, nodep
,
2556 struct getcpu_cache __user
*, unused
)
2559 int cpu
= raw_smp_processor_id();
2562 err
|= put_user(cpu
, cpup
);
2564 err
|= put_user(cpu_to_node(cpu
), nodep
);
2565 return err
? -EFAULT
: 0;
2569 * do_sysinfo - fill in sysinfo struct
2570 * @info: pointer to buffer to fill
2572 static int do_sysinfo(struct sysinfo
*info
)
2574 unsigned long mem_total
, sav_total
;
2575 unsigned int mem_unit
, bitcount
;
2576 struct timespec64 tp
;
2578 memset(info
, 0, sizeof(struct sysinfo
));
2580 ktime_get_boottime_ts64(&tp
);
2581 timens_add_boottime(&tp
);
2582 info
->uptime
= tp
.tv_sec
+ (tp
.tv_nsec
? 1 : 0);
2584 get_avenrun(info
->loads
, 0, SI_LOAD_SHIFT
- FSHIFT
);
2586 info
->procs
= nr_threads
;
2592 * If the sum of all the available memory (i.e. ram + swap)
2593 * is less than can be stored in a 32 bit unsigned long then
2594 * we can be binary compatible with 2.2.x kernels. If not,
2595 * well, in that case 2.2.x was broken anyways...
2597 * -Erik Andersen <andersee@debian.org>
2600 mem_total
= info
->totalram
+ info
->totalswap
;
2601 if (mem_total
< info
->totalram
|| mem_total
< info
->totalswap
)
2604 mem_unit
= info
->mem_unit
;
2605 while (mem_unit
> 1) {
2608 sav_total
= mem_total
;
2610 if (mem_total
< sav_total
)
2615 * If mem_total did not overflow, multiply all memory values by
2616 * info->mem_unit and set it to 1. This leaves things compatible
2617 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2622 info
->totalram
<<= bitcount
;
2623 info
->freeram
<<= bitcount
;
2624 info
->sharedram
<<= bitcount
;
2625 info
->bufferram
<<= bitcount
;
2626 info
->totalswap
<<= bitcount
;
2627 info
->freeswap
<<= bitcount
;
2628 info
->totalhigh
<<= bitcount
;
2629 info
->freehigh
<<= bitcount
;
2635 SYSCALL_DEFINE1(sysinfo
, struct sysinfo __user
*, info
)
2641 if (copy_to_user(info
, &val
, sizeof(struct sysinfo
)))
2647 #ifdef CONFIG_COMPAT
2648 struct compat_sysinfo
{
2662 char _f
[20-2*sizeof(u32
)-sizeof(int)];
2665 COMPAT_SYSCALL_DEFINE1(sysinfo
, struct compat_sysinfo __user
*, info
)
2668 struct compat_sysinfo s_32
;
2672 /* Check to see if any memory value is too large for 32-bit and scale
2675 if (upper_32_bits(s
.totalram
) || upper_32_bits(s
.totalswap
)) {
2678 while (s
.mem_unit
< PAGE_SIZE
) {
2683 s
.totalram
>>= bitcount
;
2684 s
.freeram
>>= bitcount
;
2685 s
.sharedram
>>= bitcount
;
2686 s
.bufferram
>>= bitcount
;
2687 s
.totalswap
>>= bitcount
;
2688 s
.freeswap
>>= bitcount
;
2689 s
.totalhigh
>>= bitcount
;
2690 s
.freehigh
>>= bitcount
;
2693 memset(&s_32
, 0, sizeof(s_32
));
2694 s_32
.uptime
= s
.uptime
;
2695 s_32
.loads
[0] = s
.loads
[0];
2696 s_32
.loads
[1] = s
.loads
[1];
2697 s_32
.loads
[2] = s
.loads
[2];
2698 s_32
.totalram
= s
.totalram
;
2699 s_32
.freeram
= s
.freeram
;
2700 s_32
.sharedram
= s
.sharedram
;
2701 s_32
.bufferram
= s
.bufferram
;
2702 s_32
.totalswap
= s
.totalswap
;
2703 s_32
.freeswap
= s
.freeswap
;
2704 s_32
.procs
= s
.procs
;
2705 s_32
.totalhigh
= s
.totalhigh
;
2706 s_32
.freehigh
= s
.freehigh
;
2707 s_32
.mem_unit
= s
.mem_unit
;
2708 if (copy_to_user(info
, &s_32
, sizeof(s_32
)))
2712 #endif /* CONFIG_COMPAT */