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/suspend.h>
28 #include <linux/tty.h>
29 #include <linux/signal.h>
30 #include <linux/cn_proc.h>
31 #include <linux/getcpu.h>
32 #include <linux/task_io_accounting_ops.h>
33 #include <linux/seccomp.h>
34 #include <linux/cpu.h>
35 #include <linux/personality.h>
36 #include <linux/ptrace.h>
37 #include <linux/fs_struct.h>
38 #include <linux/file.h>
39 #include <linux/mount.h>
40 #include <linux/gfp.h>
41 #include <linux/syscore_ops.h>
42 #include <linux/version.h>
43 #include <linux/ctype.h>
44 #include <linux/syscall_user_dispatch.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 PAC_SET_ENABLED_KEYS
123 # define PAC_SET_ENABLED_KEYS(a, b, c) (-EINVAL)
125 #ifndef PAC_GET_ENABLED_KEYS
126 # define PAC_GET_ENABLED_KEYS(a) (-EINVAL)
128 #ifndef SET_TAGGED_ADDR_CTRL
129 # define SET_TAGGED_ADDR_CTRL(a) (-EINVAL)
131 #ifndef GET_TAGGED_ADDR_CTRL
132 # define GET_TAGGED_ADDR_CTRL() (-EINVAL)
136 * this is where the system-wide overflow UID and GID are defined, for
137 * architectures that now have 32-bit UID/GID but didn't in the past
140 int overflowuid
= DEFAULT_OVERFLOWUID
;
141 int overflowgid
= DEFAULT_OVERFLOWGID
;
143 EXPORT_SYMBOL(overflowuid
);
144 EXPORT_SYMBOL(overflowgid
);
147 * the same as above, but for filesystems which can only store a 16-bit
148 * UID and GID. as such, this is needed on all architectures
151 int fs_overflowuid
= DEFAULT_FS_OVERFLOWUID
;
152 int fs_overflowgid
= DEFAULT_FS_OVERFLOWGID
;
154 EXPORT_SYMBOL(fs_overflowuid
);
155 EXPORT_SYMBOL(fs_overflowgid
);
158 * Returns true if current's euid is same as p's uid or euid,
159 * or has CAP_SYS_NICE to p's user_ns.
161 * Called with rcu_read_lock, creds are safe
163 static bool set_one_prio_perm(struct task_struct
*p
)
165 const struct cred
*cred
= current_cred(), *pcred
= __task_cred(p
);
167 if (uid_eq(pcred
->uid
, cred
->euid
) ||
168 uid_eq(pcred
->euid
, cred
->euid
))
170 if (ns_capable(pcred
->user_ns
, CAP_SYS_NICE
))
176 * set the priority of a task
177 * - the caller must hold the RCU read lock
179 static int set_one_prio(struct task_struct
*p
, int niceval
, int error
)
183 if (!set_one_prio_perm(p
)) {
187 if (niceval
< task_nice(p
) && !can_nice(p
, niceval
)) {
191 no_nice
= security_task_setnice(p
, niceval
);
198 set_user_nice(p
, niceval
);
203 SYSCALL_DEFINE3(setpriority
, int, which
, int, who
, int, niceval
)
205 struct task_struct
*g
, *p
;
206 struct user_struct
*user
;
207 const struct cred
*cred
= current_cred();
212 if (which
> PRIO_USER
|| which
< PRIO_PROCESS
)
215 /* normalize: avoid signed division (rounding problems) */
217 if (niceval
< MIN_NICE
)
219 if (niceval
> MAX_NICE
)
223 read_lock(&tasklist_lock
);
227 p
= find_task_by_vpid(who
);
231 error
= set_one_prio(p
, niceval
, error
);
235 pgrp
= find_vpid(who
);
237 pgrp
= task_pgrp(current
);
238 do_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
) {
239 error
= set_one_prio(p
, niceval
, error
);
240 } while_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
);
243 uid
= make_kuid(cred
->user_ns
, who
);
247 else if (!uid_eq(uid
, cred
->uid
)) {
248 user
= find_user(uid
);
250 goto out_unlock
; /* No processes for this user */
252 do_each_thread(g
, p
) {
253 if (uid_eq(task_uid(p
), uid
) && task_pid_vnr(p
))
254 error
= set_one_prio(p
, niceval
, error
);
255 } while_each_thread(g
, p
);
256 if (!uid_eq(uid
, cred
->uid
))
257 free_uid(user
); /* For find_user() */
261 read_unlock(&tasklist_lock
);
268 * Ugh. To avoid negative return values, "getpriority()" will
269 * not return the normal nice-value, but a negated value that
270 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
271 * to stay compatible.
273 SYSCALL_DEFINE2(getpriority
, int, which
, int, who
)
275 struct task_struct
*g
, *p
;
276 struct user_struct
*user
;
277 const struct cred
*cred
= current_cred();
278 long niceval
, retval
= -ESRCH
;
282 if (which
> PRIO_USER
|| which
< PRIO_PROCESS
)
286 read_lock(&tasklist_lock
);
290 p
= find_task_by_vpid(who
);
294 niceval
= nice_to_rlimit(task_nice(p
));
295 if (niceval
> retval
)
301 pgrp
= find_vpid(who
);
303 pgrp
= task_pgrp(current
);
304 do_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
) {
305 niceval
= nice_to_rlimit(task_nice(p
));
306 if (niceval
> retval
)
308 } while_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
);
311 uid
= make_kuid(cred
->user_ns
, who
);
315 else if (!uid_eq(uid
, cred
->uid
)) {
316 user
= find_user(uid
);
318 goto out_unlock
; /* No processes for this user */
320 do_each_thread(g
, p
) {
321 if (uid_eq(task_uid(p
), uid
) && task_pid_vnr(p
)) {
322 niceval
= nice_to_rlimit(task_nice(p
));
323 if (niceval
> retval
)
326 } while_each_thread(g
, p
);
327 if (!uid_eq(uid
, cred
->uid
))
328 free_uid(user
); /* for find_user() */
332 read_unlock(&tasklist_lock
);
339 * Unprivileged users may change the real gid to the effective gid
340 * or vice versa. (BSD-style)
342 * If you set the real gid at all, or set the effective gid to a value not
343 * equal to the real gid, then the saved gid is set to the new effective gid.
345 * This makes it possible for a setgid program to completely drop its
346 * privileges, which is often a useful assertion to make when you are doing
347 * a security audit over a program.
349 * The general idea is that a program which uses just setregid() will be
350 * 100% compatible with BSD. A program which uses just setgid() will be
351 * 100% compatible with POSIX with saved IDs.
353 * SMP: There are not races, the GIDs are checked only by filesystem
354 * operations (as far as semantic preservation is concerned).
356 #ifdef CONFIG_MULTIUSER
357 long __sys_setregid(gid_t rgid
, gid_t egid
)
359 struct user_namespace
*ns
= current_user_ns();
360 const struct cred
*old
;
365 krgid
= make_kgid(ns
, rgid
);
366 kegid
= make_kgid(ns
, egid
);
368 if ((rgid
!= (gid_t
) -1) && !gid_valid(krgid
))
370 if ((egid
!= (gid_t
) -1) && !gid_valid(kegid
))
373 new = prepare_creds();
376 old
= current_cred();
379 if (rgid
!= (gid_t
) -1) {
380 if (gid_eq(old
->gid
, krgid
) ||
381 gid_eq(old
->egid
, krgid
) ||
382 ns_capable_setid(old
->user_ns
, CAP_SETGID
))
387 if (egid
!= (gid_t
) -1) {
388 if (gid_eq(old
->gid
, kegid
) ||
389 gid_eq(old
->egid
, kegid
) ||
390 gid_eq(old
->sgid
, kegid
) ||
391 ns_capable_setid(old
->user_ns
, CAP_SETGID
))
397 if (rgid
!= (gid_t
) -1 ||
398 (egid
!= (gid_t
) -1 && !gid_eq(kegid
, old
->gid
)))
399 new->sgid
= new->egid
;
400 new->fsgid
= new->egid
;
402 retval
= security_task_fix_setgid(new, old
, LSM_SETID_RE
);
406 return commit_creds(new);
413 SYSCALL_DEFINE2(setregid
, gid_t
, rgid
, gid_t
, egid
)
415 return __sys_setregid(rgid
, egid
);
419 * setgid() is implemented like SysV w/ SAVED_IDS
421 * SMP: Same implicit races as above.
423 long __sys_setgid(gid_t gid
)
425 struct user_namespace
*ns
= current_user_ns();
426 const struct cred
*old
;
431 kgid
= make_kgid(ns
, gid
);
432 if (!gid_valid(kgid
))
435 new = prepare_creds();
438 old
= current_cred();
441 if (ns_capable_setid(old
->user_ns
, CAP_SETGID
))
442 new->gid
= new->egid
= new->sgid
= new->fsgid
= kgid
;
443 else if (gid_eq(kgid
, old
->gid
) || gid_eq(kgid
, old
->sgid
))
444 new->egid
= new->fsgid
= kgid
;
448 retval
= security_task_fix_setgid(new, old
, LSM_SETID_ID
);
452 return commit_creds(new);
459 SYSCALL_DEFINE1(setgid
, gid_t
, gid
)
461 return __sys_setgid(gid
);
465 * change the user struct in a credentials set to match the new UID
467 static int set_user(struct cred
*new)
469 struct user_struct
*new_user
;
471 new_user
= alloc_uid(new->uid
);
476 * We don't fail in case of NPROC limit excess here because too many
477 * poorly written programs don't check set*uid() return code, assuming
478 * it never fails if called by root. We may still enforce NPROC limit
479 * for programs doing set*uid()+execve() by harmlessly deferring the
480 * failure to the execve() stage.
482 if (is_ucounts_overlimit(new->ucounts
, UCOUNT_RLIMIT_NPROC
, rlimit(RLIMIT_NPROC
)) &&
483 new_user
!= INIT_USER
&&
484 !capable(CAP_SYS_RESOURCE
) && !capable(CAP_SYS_ADMIN
))
485 current
->flags
|= PF_NPROC_EXCEEDED
;
487 current
->flags
&= ~PF_NPROC_EXCEEDED
;
490 new->user
= new_user
;
495 * Unprivileged users may change the real uid to the effective uid
496 * or vice versa. (BSD-style)
498 * If you set the real uid at all, or set the effective uid to a value not
499 * equal to the real uid, then the saved uid is set to the new effective uid.
501 * This makes it possible for a setuid program to completely drop its
502 * privileges, which is often a useful assertion to make when you are doing
503 * a security audit over a program.
505 * The general idea is that a program which uses just setreuid() will be
506 * 100% compatible with BSD. A program which uses just setuid() will be
507 * 100% compatible with POSIX with saved IDs.
509 long __sys_setreuid(uid_t ruid
, uid_t euid
)
511 struct user_namespace
*ns
= current_user_ns();
512 const struct cred
*old
;
517 kruid
= make_kuid(ns
, ruid
);
518 keuid
= make_kuid(ns
, euid
);
520 if ((ruid
!= (uid_t
) -1) && !uid_valid(kruid
))
522 if ((euid
!= (uid_t
) -1) && !uid_valid(keuid
))
525 new = prepare_creds();
528 old
= current_cred();
531 if (ruid
!= (uid_t
) -1) {
533 if (!uid_eq(old
->uid
, kruid
) &&
534 !uid_eq(old
->euid
, kruid
) &&
535 !ns_capable_setid(old
->user_ns
, CAP_SETUID
))
539 if (euid
!= (uid_t
) -1) {
541 if (!uid_eq(old
->uid
, keuid
) &&
542 !uid_eq(old
->euid
, keuid
) &&
543 !uid_eq(old
->suid
, keuid
) &&
544 !ns_capable_setid(old
->user_ns
, CAP_SETUID
))
548 if (!uid_eq(new->uid
, old
->uid
)) {
549 retval
= set_user(new);
553 if (ruid
!= (uid_t
) -1 ||
554 (euid
!= (uid_t
) -1 && !uid_eq(keuid
, old
->uid
)))
555 new->suid
= new->euid
;
556 new->fsuid
= new->euid
;
558 retval
= security_task_fix_setuid(new, old
, LSM_SETID_RE
);
562 retval
= set_cred_ucounts(new);
566 return commit_creds(new);
573 SYSCALL_DEFINE2(setreuid
, uid_t
, ruid
, uid_t
, euid
)
575 return __sys_setreuid(ruid
, euid
);
579 * setuid() is implemented like SysV with SAVED_IDS
581 * Note that SAVED_ID's is deficient in that a setuid root program
582 * like sendmail, for example, cannot set its uid to be a normal
583 * user and then switch back, because if you're root, setuid() sets
584 * the saved uid too. If you don't like this, blame the bright people
585 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
586 * will allow a root program to temporarily drop privileges and be able to
587 * regain them by swapping the real and effective uid.
589 long __sys_setuid(uid_t uid
)
591 struct user_namespace
*ns
= current_user_ns();
592 const struct cred
*old
;
597 kuid
= make_kuid(ns
, uid
);
598 if (!uid_valid(kuid
))
601 new = prepare_creds();
604 old
= current_cred();
607 if (ns_capable_setid(old
->user_ns
, CAP_SETUID
)) {
608 new->suid
= new->uid
= kuid
;
609 if (!uid_eq(kuid
, old
->uid
)) {
610 retval
= set_user(new);
614 } else if (!uid_eq(kuid
, old
->uid
) && !uid_eq(kuid
, new->suid
)) {
618 new->fsuid
= new->euid
= kuid
;
620 retval
= security_task_fix_setuid(new, old
, LSM_SETID_ID
);
624 retval
= set_cred_ucounts(new);
628 return commit_creds(new);
635 SYSCALL_DEFINE1(setuid
, uid_t
, uid
)
637 return __sys_setuid(uid
);
642 * This function implements a generic ability to update ruid, euid,
643 * and suid. This allows you to implement the 4.4 compatible seteuid().
645 long __sys_setresuid(uid_t ruid
, uid_t euid
, uid_t suid
)
647 struct user_namespace
*ns
= current_user_ns();
648 const struct cred
*old
;
651 kuid_t kruid
, keuid
, ksuid
;
653 kruid
= make_kuid(ns
, ruid
);
654 keuid
= make_kuid(ns
, euid
);
655 ksuid
= make_kuid(ns
, suid
);
657 if ((ruid
!= (uid_t
) -1) && !uid_valid(kruid
))
660 if ((euid
!= (uid_t
) -1) && !uid_valid(keuid
))
663 if ((suid
!= (uid_t
) -1) && !uid_valid(ksuid
))
666 new = prepare_creds();
670 old
= current_cred();
673 if (!ns_capable_setid(old
->user_ns
, CAP_SETUID
)) {
674 if (ruid
!= (uid_t
) -1 && !uid_eq(kruid
, old
->uid
) &&
675 !uid_eq(kruid
, old
->euid
) && !uid_eq(kruid
, old
->suid
))
677 if (euid
!= (uid_t
) -1 && !uid_eq(keuid
, old
->uid
) &&
678 !uid_eq(keuid
, old
->euid
) && !uid_eq(keuid
, old
->suid
))
680 if (suid
!= (uid_t
) -1 && !uid_eq(ksuid
, old
->uid
) &&
681 !uid_eq(ksuid
, old
->euid
) && !uid_eq(ksuid
, old
->suid
))
685 if (ruid
!= (uid_t
) -1) {
687 if (!uid_eq(kruid
, old
->uid
)) {
688 retval
= set_user(new);
693 if (euid
!= (uid_t
) -1)
695 if (suid
!= (uid_t
) -1)
697 new->fsuid
= new->euid
;
699 retval
= security_task_fix_setuid(new, old
, LSM_SETID_RES
);
703 retval
= set_cred_ucounts(new);
707 return commit_creds(new);
714 SYSCALL_DEFINE3(setresuid
, uid_t
, ruid
, uid_t
, euid
, uid_t
, suid
)
716 return __sys_setresuid(ruid
, euid
, suid
);
719 SYSCALL_DEFINE3(getresuid
, uid_t __user
*, ruidp
, uid_t __user
*, euidp
, uid_t __user
*, suidp
)
721 const struct cred
*cred
= current_cred();
723 uid_t ruid
, euid
, suid
;
725 ruid
= from_kuid_munged(cred
->user_ns
, cred
->uid
);
726 euid
= from_kuid_munged(cred
->user_ns
, cred
->euid
);
727 suid
= from_kuid_munged(cred
->user_ns
, cred
->suid
);
729 retval
= put_user(ruid
, ruidp
);
731 retval
= put_user(euid
, euidp
);
733 return put_user(suid
, suidp
);
739 * Same as above, but for rgid, egid, sgid.
741 long __sys_setresgid(gid_t rgid
, gid_t egid
, gid_t sgid
)
743 struct user_namespace
*ns
= current_user_ns();
744 const struct cred
*old
;
747 kgid_t krgid
, kegid
, ksgid
;
749 krgid
= make_kgid(ns
, rgid
);
750 kegid
= make_kgid(ns
, egid
);
751 ksgid
= make_kgid(ns
, sgid
);
753 if ((rgid
!= (gid_t
) -1) && !gid_valid(krgid
))
755 if ((egid
!= (gid_t
) -1) && !gid_valid(kegid
))
757 if ((sgid
!= (gid_t
) -1) && !gid_valid(ksgid
))
760 new = prepare_creds();
763 old
= current_cred();
766 if (!ns_capable_setid(old
->user_ns
, CAP_SETGID
)) {
767 if (rgid
!= (gid_t
) -1 && !gid_eq(krgid
, old
->gid
) &&
768 !gid_eq(krgid
, old
->egid
) && !gid_eq(krgid
, old
->sgid
))
770 if (egid
!= (gid_t
) -1 && !gid_eq(kegid
, old
->gid
) &&
771 !gid_eq(kegid
, old
->egid
) && !gid_eq(kegid
, old
->sgid
))
773 if (sgid
!= (gid_t
) -1 && !gid_eq(ksgid
, old
->gid
) &&
774 !gid_eq(ksgid
, old
->egid
) && !gid_eq(ksgid
, old
->sgid
))
778 if (rgid
!= (gid_t
) -1)
780 if (egid
!= (gid_t
) -1)
782 if (sgid
!= (gid_t
) -1)
784 new->fsgid
= new->egid
;
786 retval
= security_task_fix_setgid(new, old
, LSM_SETID_RES
);
790 return commit_creds(new);
797 SYSCALL_DEFINE3(setresgid
, gid_t
, rgid
, gid_t
, egid
, gid_t
, sgid
)
799 return __sys_setresgid(rgid
, egid
, sgid
);
802 SYSCALL_DEFINE3(getresgid
, gid_t __user
*, rgidp
, gid_t __user
*, egidp
, gid_t __user
*, sgidp
)
804 const struct cred
*cred
= current_cred();
806 gid_t rgid
, egid
, sgid
;
808 rgid
= from_kgid_munged(cred
->user_ns
, cred
->gid
);
809 egid
= from_kgid_munged(cred
->user_ns
, cred
->egid
);
810 sgid
= from_kgid_munged(cred
->user_ns
, cred
->sgid
);
812 retval
= put_user(rgid
, rgidp
);
814 retval
= put_user(egid
, egidp
);
816 retval
= put_user(sgid
, sgidp
);
824 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
825 * is used for "access()" and for the NFS daemon (letting nfsd stay at
826 * whatever uid it wants to). It normally shadows "euid", except when
827 * explicitly set by setfsuid() or for access..
829 long __sys_setfsuid(uid_t uid
)
831 const struct cred
*old
;
836 old
= current_cred();
837 old_fsuid
= from_kuid_munged(old
->user_ns
, old
->fsuid
);
839 kuid
= make_kuid(old
->user_ns
, uid
);
840 if (!uid_valid(kuid
))
843 new = prepare_creds();
847 if (uid_eq(kuid
, old
->uid
) || uid_eq(kuid
, old
->euid
) ||
848 uid_eq(kuid
, old
->suid
) || uid_eq(kuid
, old
->fsuid
) ||
849 ns_capable_setid(old
->user_ns
, CAP_SETUID
)) {
850 if (!uid_eq(kuid
, old
->fsuid
)) {
852 if (security_task_fix_setuid(new, old
, LSM_SETID_FS
) == 0)
865 SYSCALL_DEFINE1(setfsuid
, uid_t
, uid
)
867 return __sys_setfsuid(uid
);
871 * Samma på svenska..
873 long __sys_setfsgid(gid_t gid
)
875 const struct cred
*old
;
880 old
= current_cred();
881 old_fsgid
= from_kgid_munged(old
->user_ns
, old
->fsgid
);
883 kgid
= make_kgid(old
->user_ns
, gid
);
884 if (!gid_valid(kgid
))
887 new = prepare_creds();
891 if (gid_eq(kgid
, old
->gid
) || gid_eq(kgid
, old
->egid
) ||
892 gid_eq(kgid
, old
->sgid
) || gid_eq(kgid
, old
->fsgid
) ||
893 ns_capable_setid(old
->user_ns
, CAP_SETGID
)) {
894 if (!gid_eq(kgid
, old
->fsgid
)) {
896 if (security_task_fix_setgid(new,old
,LSM_SETID_FS
) == 0)
909 SYSCALL_DEFINE1(setfsgid
, gid_t
, gid
)
911 return __sys_setfsgid(gid
);
913 #endif /* CONFIG_MULTIUSER */
916 * sys_getpid - return the thread group id of the current process
918 * Note, despite the name, this returns the tgid not the pid. The tgid and
919 * the pid are identical unless CLONE_THREAD was specified on clone() in
920 * which case the tgid is the same in all threads of the same group.
922 * This is SMP safe as current->tgid does not change.
924 SYSCALL_DEFINE0(getpid
)
926 return task_tgid_vnr(current
);
929 /* Thread ID - the internal kernel "pid" */
930 SYSCALL_DEFINE0(gettid
)
932 return task_pid_vnr(current
);
936 * Accessing ->real_parent is not SMP-safe, it could
937 * change from under us. However, we can use a stale
938 * value of ->real_parent under rcu_read_lock(), see
939 * release_task()->call_rcu(delayed_put_task_struct).
941 SYSCALL_DEFINE0(getppid
)
946 pid
= task_tgid_vnr(rcu_dereference(current
->real_parent
));
952 SYSCALL_DEFINE0(getuid
)
954 /* Only we change this so SMP safe */
955 return from_kuid_munged(current_user_ns(), current_uid());
958 SYSCALL_DEFINE0(geteuid
)
960 /* Only we change this so SMP safe */
961 return from_kuid_munged(current_user_ns(), current_euid());
964 SYSCALL_DEFINE0(getgid
)
966 /* Only we change this so SMP safe */
967 return from_kgid_munged(current_user_ns(), current_gid());
970 SYSCALL_DEFINE0(getegid
)
972 /* Only we change this so SMP safe */
973 return from_kgid_munged(current_user_ns(), current_egid());
976 static void do_sys_times(struct tms
*tms
)
978 u64 tgutime
, tgstime
, cutime
, cstime
;
980 thread_group_cputime_adjusted(current
, &tgutime
, &tgstime
);
981 cutime
= current
->signal
->cutime
;
982 cstime
= current
->signal
->cstime
;
983 tms
->tms_utime
= nsec_to_clock_t(tgutime
);
984 tms
->tms_stime
= nsec_to_clock_t(tgstime
);
985 tms
->tms_cutime
= nsec_to_clock_t(cutime
);
986 tms
->tms_cstime
= nsec_to_clock_t(cstime
);
989 SYSCALL_DEFINE1(times
, struct tms __user
*, tbuf
)
995 if (copy_to_user(tbuf
, &tmp
, sizeof(struct tms
)))
998 force_successful_syscall_return();
999 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1002 #ifdef CONFIG_COMPAT
1003 static compat_clock_t
clock_t_to_compat_clock_t(clock_t x
)
1005 return compat_jiffies_to_clock_t(clock_t_to_jiffies(x
));
1008 COMPAT_SYSCALL_DEFINE1(times
, struct compat_tms __user
*, tbuf
)
1012 struct compat_tms tmp
;
1015 /* Convert our struct tms to the compat version. */
1016 tmp
.tms_utime
= clock_t_to_compat_clock_t(tms
.tms_utime
);
1017 tmp
.tms_stime
= clock_t_to_compat_clock_t(tms
.tms_stime
);
1018 tmp
.tms_cutime
= clock_t_to_compat_clock_t(tms
.tms_cutime
);
1019 tmp
.tms_cstime
= clock_t_to_compat_clock_t(tms
.tms_cstime
);
1020 if (copy_to_user(tbuf
, &tmp
, sizeof(tmp
)))
1023 force_successful_syscall_return();
1024 return compat_jiffies_to_clock_t(jiffies
);
1029 * This needs some heavy checking ...
1030 * I just haven't the stomach for it. I also don't fully
1031 * understand sessions/pgrp etc. Let somebody who does explain it.
1033 * OK, I think I have the protection semantics right.... this is really
1034 * only important on a multi-user system anyway, to make sure one user
1035 * can't send a signal to a process owned by another. -TYT, 12/12/91
1037 * !PF_FORKNOEXEC check to conform completely to POSIX.
1039 SYSCALL_DEFINE2(setpgid
, pid_t
, pid
, pid_t
, pgid
)
1041 struct task_struct
*p
;
1042 struct task_struct
*group_leader
= current
->group_leader
;
1047 pid
= task_pid_vnr(group_leader
);
1054 /* From this point forward we keep holding onto the tasklist lock
1055 * so that our parent does not change from under us. -DaveM
1057 write_lock_irq(&tasklist_lock
);
1060 p
= find_task_by_vpid(pid
);
1065 if (!thread_group_leader(p
))
1068 if (same_thread_group(p
->real_parent
, group_leader
)) {
1070 if (task_session(p
) != task_session(group_leader
))
1073 if (!(p
->flags
& PF_FORKNOEXEC
))
1077 if (p
!= group_leader
)
1082 if (p
->signal
->leader
)
1087 struct task_struct
*g
;
1089 pgrp
= find_vpid(pgid
);
1090 g
= pid_task(pgrp
, PIDTYPE_PGID
);
1091 if (!g
|| task_session(g
) != task_session(group_leader
))
1095 err
= security_task_setpgid(p
, pgid
);
1099 if (task_pgrp(p
) != pgrp
)
1100 change_pid(p
, PIDTYPE_PGID
, pgrp
);
1104 /* All paths lead to here, thus we are safe. -DaveM */
1105 write_unlock_irq(&tasklist_lock
);
1110 static int do_getpgid(pid_t pid
)
1112 struct task_struct
*p
;
1118 grp
= task_pgrp(current
);
1121 p
= find_task_by_vpid(pid
);
1128 retval
= security_task_getpgid(p
);
1132 retval
= pid_vnr(grp
);
1138 SYSCALL_DEFINE1(getpgid
, pid_t
, pid
)
1140 return do_getpgid(pid
);
1143 #ifdef __ARCH_WANT_SYS_GETPGRP
1145 SYSCALL_DEFINE0(getpgrp
)
1147 return do_getpgid(0);
1152 SYSCALL_DEFINE1(getsid
, pid_t
, pid
)
1154 struct task_struct
*p
;
1160 sid
= task_session(current
);
1163 p
= find_task_by_vpid(pid
);
1166 sid
= task_session(p
);
1170 retval
= security_task_getsid(p
);
1174 retval
= pid_vnr(sid
);
1180 static void set_special_pids(struct pid
*pid
)
1182 struct task_struct
*curr
= current
->group_leader
;
1184 if (task_session(curr
) != pid
)
1185 change_pid(curr
, PIDTYPE_SID
, pid
);
1187 if (task_pgrp(curr
) != pid
)
1188 change_pid(curr
, PIDTYPE_PGID
, pid
);
1191 int ksys_setsid(void)
1193 struct task_struct
*group_leader
= current
->group_leader
;
1194 struct pid
*sid
= task_pid(group_leader
);
1195 pid_t session
= pid_vnr(sid
);
1198 write_lock_irq(&tasklist_lock
);
1199 /* Fail if I am already a session leader */
1200 if (group_leader
->signal
->leader
)
1203 /* Fail if a process group id already exists that equals the
1204 * proposed session id.
1206 if (pid_task(sid
, PIDTYPE_PGID
))
1209 group_leader
->signal
->leader
= 1;
1210 set_special_pids(sid
);
1212 proc_clear_tty(group_leader
);
1216 write_unlock_irq(&tasklist_lock
);
1218 proc_sid_connector(group_leader
);
1219 sched_autogroup_create_attach(group_leader
);
1224 SYSCALL_DEFINE0(setsid
)
1226 return ksys_setsid();
1229 DECLARE_RWSEM(uts_sem
);
1231 #ifdef COMPAT_UTS_MACHINE
1232 #define override_architecture(name) \
1233 (personality(current->personality) == PER_LINUX32 && \
1234 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1235 sizeof(COMPAT_UTS_MACHINE)))
1237 #define override_architecture(name) 0
1241 * Work around broken programs that cannot handle "Linux 3.0".
1242 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1243 * And we map 4.x and later versions to 2.6.60+x, so 4.0/5.0/6.0/... would be
1246 static int override_release(char __user
*release
, size_t len
)
1250 if (current
->personality
& UNAME26
) {
1251 const char *rest
= UTS_RELEASE
;
1252 char buf
[65] = { 0 };
1258 if (*rest
== '.' && ++ndots
>= 3)
1260 if (!isdigit(*rest
) && *rest
!= '.')
1264 v
= LINUX_VERSION_PATCHLEVEL
+ 60;
1265 copy
= clamp_t(size_t, len
, 1, sizeof(buf
));
1266 copy
= scnprintf(buf
, copy
, "2.6.%u%s", v
, rest
);
1267 ret
= copy_to_user(release
, buf
, copy
+ 1);
1272 SYSCALL_DEFINE1(newuname
, struct new_utsname __user
*, name
)
1274 struct new_utsname tmp
;
1276 down_read(&uts_sem
);
1277 memcpy(&tmp
, utsname(), sizeof(tmp
));
1279 if (copy_to_user(name
, &tmp
, sizeof(tmp
)))
1282 if (override_release(name
->release
, sizeof(name
->release
)))
1284 if (override_architecture(name
))
1289 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1293 SYSCALL_DEFINE1(uname
, struct old_utsname __user
*, name
)
1295 struct old_utsname tmp
;
1300 down_read(&uts_sem
);
1301 memcpy(&tmp
, utsname(), sizeof(tmp
));
1303 if (copy_to_user(name
, &tmp
, sizeof(tmp
)))
1306 if (override_release(name
->release
, sizeof(name
->release
)))
1308 if (override_architecture(name
))
1313 SYSCALL_DEFINE1(olduname
, struct oldold_utsname __user
*, name
)
1315 struct oldold_utsname tmp
;
1320 memset(&tmp
, 0, sizeof(tmp
));
1322 down_read(&uts_sem
);
1323 memcpy(&tmp
.sysname
, &utsname()->sysname
, __OLD_UTS_LEN
);
1324 memcpy(&tmp
.nodename
, &utsname()->nodename
, __OLD_UTS_LEN
);
1325 memcpy(&tmp
.release
, &utsname()->release
, __OLD_UTS_LEN
);
1326 memcpy(&tmp
.version
, &utsname()->version
, __OLD_UTS_LEN
);
1327 memcpy(&tmp
.machine
, &utsname()->machine
, __OLD_UTS_LEN
);
1329 if (copy_to_user(name
, &tmp
, sizeof(tmp
)))
1332 if (override_architecture(name
))
1334 if (override_release(name
->release
, sizeof(name
->release
)))
1340 SYSCALL_DEFINE2(sethostname
, char __user
*, name
, int, len
)
1343 char tmp
[__NEW_UTS_LEN
];
1345 if (!ns_capable(current
->nsproxy
->uts_ns
->user_ns
, CAP_SYS_ADMIN
))
1348 if (len
< 0 || len
> __NEW_UTS_LEN
)
1351 if (!copy_from_user(tmp
, name
, len
)) {
1352 struct new_utsname
*u
;
1354 down_write(&uts_sem
);
1356 memcpy(u
->nodename
, tmp
, len
);
1357 memset(u
->nodename
+ len
, 0, sizeof(u
->nodename
) - len
);
1359 uts_proc_notify(UTS_PROC_HOSTNAME
);
1365 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1367 SYSCALL_DEFINE2(gethostname
, char __user
*, name
, int, len
)
1370 struct new_utsname
*u
;
1371 char tmp
[__NEW_UTS_LEN
+ 1];
1375 down_read(&uts_sem
);
1377 i
= 1 + strlen(u
->nodename
);
1380 memcpy(tmp
, u
->nodename
, i
);
1382 if (copy_to_user(name
, tmp
, i
))
1390 * Only setdomainname; getdomainname can be implemented by calling
1393 SYSCALL_DEFINE2(setdomainname
, char __user
*, name
, int, len
)
1396 char tmp
[__NEW_UTS_LEN
];
1398 if (!ns_capable(current
->nsproxy
->uts_ns
->user_ns
, CAP_SYS_ADMIN
))
1400 if (len
< 0 || len
> __NEW_UTS_LEN
)
1404 if (!copy_from_user(tmp
, name
, len
)) {
1405 struct new_utsname
*u
;
1407 down_write(&uts_sem
);
1409 memcpy(u
->domainname
, tmp
, len
);
1410 memset(u
->domainname
+ len
, 0, sizeof(u
->domainname
) - len
);
1412 uts_proc_notify(UTS_PROC_DOMAINNAME
);
1418 SYSCALL_DEFINE2(getrlimit
, unsigned int, resource
, struct rlimit __user
*, rlim
)
1420 struct rlimit value
;
1423 ret
= do_prlimit(current
, resource
, NULL
, &value
);
1425 ret
= copy_to_user(rlim
, &value
, sizeof(*rlim
)) ? -EFAULT
: 0;
1430 #ifdef CONFIG_COMPAT
1432 COMPAT_SYSCALL_DEFINE2(setrlimit
, unsigned int, resource
,
1433 struct compat_rlimit __user
*, rlim
)
1436 struct compat_rlimit r32
;
1438 if (copy_from_user(&r32
, rlim
, sizeof(struct compat_rlimit
)))
1441 if (r32
.rlim_cur
== COMPAT_RLIM_INFINITY
)
1442 r
.rlim_cur
= RLIM_INFINITY
;
1444 r
.rlim_cur
= r32
.rlim_cur
;
1445 if (r32
.rlim_max
== COMPAT_RLIM_INFINITY
)
1446 r
.rlim_max
= RLIM_INFINITY
;
1448 r
.rlim_max
= r32
.rlim_max
;
1449 return do_prlimit(current
, resource
, &r
, NULL
);
1452 COMPAT_SYSCALL_DEFINE2(getrlimit
, unsigned int, resource
,
1453 struct compat_rlimit __user
*, rlim
)
1458 ret
= do_prlimit(current
, resource
, NULL
, &r
);
1460 struct compat_rlimit r32
;
1461 if (r
.rlim_cur
> COMPAT_RLIM_INFINITY
)
1462 r32
.rlim_cur
= COMPAT_RLIM_INFINITY
;
1464 r32
.rlim_cur
= r
.rlim_cur
;
1465 if (r
.rlim_max
> COMPAT_RLIM_INFINITY
)
1466 r32
.rlim_max
= COMPAT_RLIM_INFINITY
;
1468 r32
.rlim_max
= r
.rlim_max
;
1470 if (copy_to_user(rlim
, &r32
, sizeof(struct compat_rlimit
)))
1478 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1481 * Back compatibility for getrlimit. Needed for some apps.
1483 SYSCALL_DEFINE2(old_getrlimit
, unsigned int, resource
,
1484 struct rlimit __user
*, rlim
)
1487 if (resource
>= RLIM_NLIMITS
)
1490 resource
= array_index_nospec(resource
, RLIM_NLIMITS
);
1491 task_lock(current
->group_leader
);
1492 x
= current
->signal
->rlim
[resource
];
1493 task_unlock(current
->group_leader
);
1494 if (x
.rlim_cur
> 0x7FFFFFFF)
1495 x
.rlim_cur
= 0x7FFFFFFF;
1496 if (x
.rlim_max
> 0x7FFFFFFF)
1497 x
.rlim_max
= 0x7FFFFFFF;
1498 return copy_to_user(rlim
, &x
, sizeof(x
)) ? -EFAULT
: 0;
1501 #ifdef CONFIG_COMPAT
1502 COMPAT_SYSCALL_DEFINE2(old_getrlimit
, unsigned int, resource
,
1503 struct compat_rlimit __user
*, rlim
)
1507 if (resource
>= RLIM_NLIMITS
)
1510 resource
= array_index_nospec(resource
, RLIM_NLIMITS
);
1511 task_lock(current
->group_leader
);
1512 r
= current
->signal
->rlim
[resource
];
1513 task_unlock(current
->group_leader
);
1514 if (r
.rlim_cur
> 0x7FFFFFFF)
1515 r
.rlim_cur
= 0x7FFFFFFF;
1516 if (r
.rlim_max
> 0x7FFFFFFF)
1517 r
.rlim_max
= 0x7FFFFFFF;
1519 if (put_user(r
.rlim_cur
, &rlim
->rlim_cur
) ||
1520 put_user(r
.rlim_max
, &rlim
->rlim_max
))
1528 static inline bool rlim64_is_infinity(__u64 rlim64
)
1530 #if BITS_PER_LONG < 64
1531 return rlim64
>= ULONG_MAX
;
1533 return rlim64
== RLIM64_INFINITY
;
1537 static void rlim_to_rlim64(const struct rlimit
*rlim
, struct rlimit64
*rlim64
)
1539 if (rlim
->rlim_cur
== RLIM_INFINITY
)
1540 rlim64
->rlim_cur
= RLIM64_INFINITY
;
1542 rlim64
->rlim_cur
= rlim
->rlim_cur
;
1543 if (rlim
->rlim_max
== RLIM_INFINITY
)
1544 rlim64
->rlim_max
= RLIM64_INFINITY
;
1546 rlim64
->rlim_max
= rlim
->rlim_max
;
1549 static void rlim64_to_rlim(const struct rlimit64
*rlim64
, struct rlimit
*rlim
)
1551 if (rlim64_is_infinity(rlim64
->rlim_cur
))
1552 rlim
->rlim_cur
= RLIM_INFINITY
;
1554 rlim
->rlim_cur
= (unsigned long)rlim64
->rlim_cur
;
1555 if (rlim64_is_infinity(rlim64
->rlim_max
))
1556 rlim
->rlim_max
= RLIM_INFINITY
;
1558 rlim
->rlim_max
= (unsigned long)rlim64
->rlim_max
;
1561 /* make sure you are allowed to change @tsk limits before calling this */
1562 int do_prlimit(struct task_struct
*tsk
, unsigned int resource
,
1563 struct rlimit
*new_rlim
, struct rlimit
*old_rlim
)
1565 struct rlimit
*rlim
;
1568 if (resource
>= RLIM_NLIMITS
)
1571 if (new_rlim
->rlim_cur
> new_rlim
->rlim_max
)
1573 if (resource
== RLIMIT_NOFILE
&&
1574 new_rlim
->rlim_max
> sysctl_nr_open
)
1578 /* protect tsk->signal and tsk->sighand from disappearing */
1579 read_lock(&tasklist_lock
);
1580 if (!tsk
->sighand
) {
1585 rlim
= tsk
->signal
->rlim
+ resource
;
1586 task_lock(tsk
->group_leader
);
1588 /* Keep the capable check against init_user_ns until
1589 cgroups can contain all limits */
1590 if (new_rlim
->rlim_max
> rlim
->rlim_max
&&
1591 !capable(CAP_SYS_RESOURCE
))
1594 retval
= security_task_setrlimit(tsk
, resource
, new_rlim
);
1602 task_unlock(tsk
->group_leader
);
1605 * RLIMIT_CPU handling. Arm the posix CPU timer if the limit is not
1606 * infinite. In case of RLIM_INFINITY the posix CPU timer code
1607 * ignores the rlimit.
1609 if (!retval
&& new_rlim
&& resource
== RLIMIT_CPU
&&
1610 new_rlim
->rlim_cur
!= RLIM_INFINITY
&&
1611 IS_ENABLED(CONFIG_POSIX_TIMERS
))
1612 update_rlimit_cpu(tsk
, new_rlim
->rlim_cur
);
1614 read_unlock(&tasklist_lock
);
1618 /* rcu lock must be held */
1619 static int check_prlimit_permission(struct task_struct
*task
,
1622 const struct cred
*cred
= current_cred(), *tcred
;
1625 if (current
== task
)
1628 tcred
= __task_cred(task
);
1629 id_match
= (uid_eq(cred
->uid
, tcred
->euid
) &&
1630 uid_eq(cred
->uid
, tcred
->suid
) &&
1631 uid_eq(cred
->uid
, tcred
->uid
) &&
1632 gid_eq(cred
->gid
, tcred
->egid
) &&
1633 gid_eq(cred
->gid
, tcred
->sgid
) &&
1634 gid_eq(cred
->gid
, tcred
->gid
));
1635 if (!id_match
&& !ns_capable(tcred
->user_ns
, CAP_SYS_RESOURCE
))
1638 return security_task_prlimit(cred
, tcred
, flags
);
1641 SYSCALL_DEFINE4(prlimit64
, pid_t
, pid
, unsigned int, resource
,
1642 const struct rlimit64 __user
*, new_rlim
,
1643 struct rlimit64 __user
*, old_rlim
)
1645 struct rlimit64 old64
, new64
;
1646 struct rlimit old
, new;
1647 struct task_struct
*tsk
;
1648 unsigned int checkflags
= 0;
1652 checkflags
|= LSM_PRLIMIT_READ
;
1655 if (copy_from_user(&new64
, new_rlim
, sizeof(new64
)))
1657 rlim64_to_rlim(&new64
, &new);
1658 checkflags
|= LSM_PRLIMIT_WRITE
;
1662 tsk
= pid
? find_task_by_vpid(pid
) : current
;
1667 ret
= check_prlimit_permission(tsk
, checkflags
);
1672 get_task_struct(tsk
);
1675 ret
= do_prlimit(tsk
, resource
, new_rlim
? &new : NULL
,
1676 old_rlim
? &old
: NULL
);
1678 if (!ret
&& old_rlim
) {
1679 rlim_to_rlim64(&old
, &old64
);
1680 if (copy_to_user(old_rlim
, &old64
, sizeof(old64
)))
1684 put_task_struct(tsk
);
1688 SYSCALL_DEFINE2(setrlimit
, unsigned int, resource
, struct rlimit __user
*, rlim
)
1690 struct rlimit new_rlim
;
1692 if (copy_from_user(&new_rlim
, rlim
, sizeof(*rlim
)))
1694 return do_prlimit(current
, resource
, &new_rlim
, NULL
);
1698 * It would make sense to put struct rusage in the task_struct,
1699 * except that would make the task_struct be *really big*. After
1700 * task_struct gets moved into malloc'ed memory, it would
1701 * make sense to do this. It will make moving the rest of the information
1702 * a lot simpler! (Which we're not doing right now because we're not
1703 * measuring them yet).
1705 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1706 * races with threads incrementing their own counters. But since word
1707 * reads are atomic, we either get new values or old values and we don't
1708 * care which for the sums. We always take the siglock to protect reading
1709 * the c* fields from p->signal from races with exit.c updating those
1710 * fields when reaping, so a sample either gets all the additions of a
1711 * given child after it's reaped, or none so this sample is before reaping.
1714 * We need to take the siglock for CHILDEREN, SELF and BOTH
1715 * for the cases current multithreaded, non-current single threaded
1716 * non-current multithreaded. Thread traversal is now safe with
1718 * Strictly speaking, we donot need to take the siglock if we are current and
1719 * single threaded, as no one else can take our signal_struct away, no one
1720 * else can reap the children to update signal->c* counters, and no one else
1721 * can race with the signal-> fields. If we do not take any lock, the
1722 * signal-> fields could be read out of order while another thread was just
1723 * exiting. So we should place a read memory barrier when we avoid the lock.
1724 * On the writer side, write memory barrier is implied in __exit_signal
1725 * as __exit_signal releases the siglock spinlock after updating the signal->
1726 * fields. But we don't do this yet to keep things simple.
1730 static void accumulate_thread_rusage(struct task_struct
*t
, struct rusage
*r
)
1732 r
->ru_nvcsw
+= t
->nvcsw
;
1733 r
->ru_nivcsw
+= t
->nivcsw
;
1734 r
->ru_minflt
+= t
->min_flt
;
1735 r
->ru_majflt
+= t
->maj_flt
;
1736 r
->ru_inblock
+= task_io_get_inblock(t
);
1737 r
->ru_oublock
+= task_io_get_oublock(t
);
1740 void getrusage(struct task_struct
*p
, int who
, struct rusage
*r
)
1742 struct task_struct
*t
;
1743 unsigned long flags
;
1744 u64 tgutime
, tgstime
, utime
, stime
;
1745 unsigned long maxrss
= 0;
1747 memset((char *)r
, 0, sizeof (*r
));
1750 if (who
== RUSAGE_THREAD
) {
1751 task_cputime_adjusted(current
, &utime
, &stime
);
1752 accumulate_thread_rusage(p
, r
);
1753 maxrss
= p
->signal
->maxrss
;
1757 if (!lock_task_sighand(p
, &flags
))
1762 case RUSAGE_CHILDREN
:
1763 utime
= p
->signal
->cutime
;
1764 stime
= p
->signal
->cstime
;
1765 r
->ru_nvcsw
= p
->signal
->cnvcsw
;
1766 r
->ru_nivcsw
= p
->signal
->cnivcsw
;
1767 r
->ru_minflt
= p
->signal
->cmin_flt
;
1768 r
->ru_majflt
= p
->signal
->cmaj_flt
;
1769 r
->ru_inblock
= p
->signal
->cinblock
;
1770 r
->ru_oublock
= p
->signal
->coublock
;
1771 maxrss
= p
->signal
->cmaxrss
;
1773 if (who
== RUSAGE_CHILDREN
)
1778 thread_group_cputime_adjusted(p
, &tgutime
, &tgstime
);
1781 r
->ru_nvcsw
+= p
->signal
->nvcsw
;
1782 r
->ru_nivcsw
+= p
->signal
->nivcsw
;
1783 r
->ru_minflt
+= p
->signal
->min_flt
;
1784 r
->ru_majflt
+= p
->signal
->maj_flt
;
1785 r
->ru_inblock
+= p
->signal
->inblock
;
1786 r
->ru_oublock
+= p
->signal
->oublock
;
1787 if (maxrss
< p
->signal
->maxrss
)
1788 maxrss
= p
->signal
->maxrss
;
1791 accumulate_thread_rusage(t
, r
);
1792 } while_each_thread(p
, t
);
1798 unlock_task_sighand(p
, &flags
);
1801 r
->ru_utime
= ns_to_kernel_old_timeval(utime
);
1802 r
->ru_stime
= ns_to_kernel_old_timeval(stime
);
1804 if (who
!= RUSAGE_CHILDREN
) {
1805 struct mm_struct
*mm
= get_task_mm(p
);
1808 setmax_mm_hiwater_rss(&maxrss
, mm
);
1812 r
->ru_maxrss
= maxrss
* (PAGE_SIZE
/ 1024); /* convert pages to KBs */
1815 SYSCALL_DEFINE2(getrusage
, int, who
, struct rusage __user
*, ru
)
1819 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
&&
1820 who
!= RUSAGE_THREAD
)
1823 getrusage(current
, who
, &r
);
1824 return copy_to_user(ru
, &r
, sizeof(r
)) ? -EFAULT
: 0;
1827 #ifdef CONFIG_COMPAT
1828 COMPAT_SYSCALL_DEFINE2(getrusage
, int, who
, struct compat_rusage __user
*, ru
)
1832 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
&&
1833 who
!= RUSAGE_THREAD
)
1836 getrusage(current
, who
, &r
);
1837 return put_compat_rusage(&r
, ru
);
1841 SYSCALL_DEFINE1(umask
, int, mask
)
1843 mask
= xchg(¤t
->fs
->umask
, mask
& S_IRWXUGO
);
1847 static int prctl_set_mm_exe_file(struct mm_struct
*mm
, unsigned int fd
)
1850 struct inode
*inode
;
1857 inode
= file_inode(exe
.file
);
1860 * Because the original mm->exe_file points to executable file, make
1861 * sure that this one is executable as well, to avoid breaking an
1865 if (!S_ISREG(inode
->i_mode
) || path_noexec(&exe
.file
->f_path
))
1868 err
= file_permission(exe
.file
, MAY_EXEC
);
1872 err
= replace_mm_exe_file(mm
, exe
.file
);
1879 * Check arithmetic relations of passed addresses.
1881 * WARNING: we don't require any capability here so be very careful
1882 * in what is allowed for modification from userspace.
1884 static int validate_prctl_map_addr(struct prctl_mm_map
*prctl_map
)
1886 unsigned long mmap_max_addr
= TASK_SIZE
;
1887 int error
= -EINVAL
, i
;
1889 static const unsigned char offsets
[] = {
1890 offsetof(struct prctl_mm_map
, start_code
),
1891 offsetof(struct prctl_mm_map
, end_code
),
1892 offsetof(struct prctl_mm_map
, start_data
),
1893 offsetof(struct prctl_mm_map
, end_data
),
1894 offsetof(struct prctl_mm_map
, start_brk
),
1895 offsetof(struct prctl_mm_map
, brk
),
1896 offsetof(struct prctl_mm_map
, start_stack
),
1897 offsetof(struct prctl_mm_map
, arg_start
),
1898 offsetof(struct prctl_mm_map
, arg_end
),
1899 offsetof(struct prctl_mm_map
, env_start
),
1900 offsetof(struct prctl_mm_map
, env_end
),
1904 * Make sure the members are not somewhere outside
1905 * of allowed address space.
1907 for (i
= 0; i
< ARRAY_SIZE(offsets
); i
++) {
1908 u64 val
= *(u64
*)((char *)prctl_map
+ offsets
[i
]);
1910 if ((unsigned long)val
>= mmap_max_addr
||
1911 (unsigned long)val
< mmap_min_addr
)
1916 * Make sure the pairs are ordered.
1918 #define __prctl_check_order(__m1, __op, __m2) \
1919 ((unsigned long)prctl_map->__m1 __op \
1920 (unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1921 error
= __prctl_check_order(start_code
, <, end_code
);
1922 error
|= __prctl_check_order(start_data
,<=, end_data
);
1923 error
|= __prctl_check_order(start_brk
, <=, brk
);
1924 error
|= __prctl_check_order(arg_start
, <=, arg_end
);
1925 error
|= __prctl_check_order(env_start
, <=, env_end
);
1928 #undef __prctl_check_order
1933 * Neither we should allow to override limits if they set.
1935 if (check_data_rlimit(rlimit(RLIMIT_DATA
), prctl_map
->brk
,
1936 prctl_map
->start_brk
, prctl_map
->end_data
,
1937 prctl_map
->start_data
))
1945 #ifdef CONFIG_CHECKPOINT_RESTORE
1946 static int prctl_set_mm_map(int opt
, const void __user
*addr
, unsigned long data_size
)
1948 struct prctl_mm_map prctl_map
= { .exe_fd
= (u32
)-1, };
1949 unsigned long user_auxv
[AT_VECTOR_SIZE
];
1950 struct mm_struct
*mm
= current
->mm
;
1953 BUILD_BUG_ON(sizeof(user_auxv
) != sizeof(mm
->saved_auxv
));
1954 BUILD_BUG_ON(sizeof(struct prctl_mm_map
) > 256);
1956 if (opt
== PR_SET_MM_MAP_SIZE
)
1957 return put_user((unsigned int)sizeof(prctl_map
),
1958 (unsigned int __user
*)addr
);
1960 if (data_size
!= sizeof(prctl_map
))
1963 if (copy_from_user(&prctl_map
, addr
, sizeof(prctl_map
)))
1966 error
= validate_prctl_map_addr(&prctl_map
);
1970 if (prctl_map
.auxv_size
) {
1972 * Someone is trying to cheat the auxv vector.
1974 if (!prctl_map
.auxv
||
1975 prctl_map
.auxv_size
> sizeof(mm
->saved_auxv
))
1978 memset(user_auxv
, 0, sizeof(user_auxv
));
1979 if (copy_from_user(user_auxv
,
1980 (const void __user
*)prctl_map
.auxv
,
1981 prctl_map
.auxv_size
))
1984 /* Last entry must be AT_NULL as specification requires */
1985 user_auxv
[AT_VECTOR_SIZE
- 2] = AT_NULL
;
1986 user_auxv
[AT_VECTOR_SIZE
- 1] = AT_NULL
;
1989 if (prctl_map
.exe_fd
!= (u32
)-1) {
1991 * Check if the current user is checkpoint/restore capable.
1992 * At the time of this writing, it checks for CAP_SYS_ADMIN
1993 * or CAP_CHECKPOINT_RESTORE.
1994 * Note that a user with access to ptrace can masquerade an
1995 * arbitrary program as any executable, even setuid ones.
1996 * This may have implications in the tomoyo subsystem.
1998 if (!checkpoint_restore_ns_capable(current_user_ns()))
2001 error
= prctl_set_mm_exe_file(mm
, prctl_map
.exe_fd
);
2007 * arg_lock protects concurrent updates but we still need mmap_lock for
2008 * read to exclude races with sys_brk.
2013 * We don't validate if these members are pointing to
2014 * real present VMAs because application may have correspond
2015 * VMAs already unmapped and kernel uses these members for statistics
2016 * output in procfs mostly, except
2018 * - @start_brk/@brk which are used in do_brk_flags but kernel lookups
2019 * for VMAs when updating these members so anything wrong written
2020 * here cause kernel to swear at userspace program but won't lead
2021 * to any problem in kernel itself
2024 spin_lock(&mm
->arg_lock
);
2025 mm
->start_code
= prctl_map
.start_code
;
2026 mm
->end_code
= prctl_map
.end_code
;
2027 mm
->start_data
= prctl_map
.start_data
;
2028 mm
->end_data
= prctl_map
.end_data
;
2029 mm
->start_brk
= prctl_map
.start_brk
;
2030 mm
->brk
= prctl_map
.brk
;
2031 mm
->start_stack
= prctl_map
.start_stack
;
2032 mm
->arg_start
= prctl_map
.arg_start
;
2033 mm
->arg_end
= prctl_map
.arg_end
;
2034 mm
->env_start
= prctl_map
.env_start
;
2035 mm
->env_end
= prctl_map
.env_end
;
2036 spin_unlock(&mm
->arg_lock
);
2039 * Note this update of @saved_auxv is lockless thus
2040 * if someone reads this member in procfs while we're
2041 * updating -- it may get partly updated results. It's
2042 * known and acceptable trade off: we leave it as is to
2043 * not introduce additional locks here making the kernel
2046 if (prctl_map
.auxv_size
)
2047 memcpy(mm
->saved_auxv
, user_auxv
, sizeof(user_auxv
));
2049 mmap_read_unlock(mm
);
2052 #endif /* CONFIG_CHECKPOINT_RESTORE */
2054 static int prctl_set_auxv(struct mm_struct
*mm
, unsigned long addr
,
2058 * This doesn't move the auxiliary vector itself since it's pinned to
2059 * mm_struct, but it permits filling the vector with new values. It's
2060 * up to the caller to provide sane values here, otherwise userspace
2061 * tools which use this vector might be unhappy.
2063 unsigned long user_auxv
[AT_VECTOR_SIZE
] = {};
2065 if (len
> sizeof(user_auxv
))
2068 if (copy_from_user(user_auxv
, (const void __user
*)addr
, len
))
2071 /* Make sure the last entry is always AT_NULL */
2072 user_auxv
[AT_VECTOR_SIZE
- 2] = 0;
2073 user_auxv
[AT_VECTOR_SIZE
- 1] = 0;
2075 BUILD_BUG_ON(sizeof(user_auxv
) != sizeof(mm
->saved_auxv
));
2078 memcpy(mm
->saved_auxv
, user_auxv
, len
);
2079 task_unlock(current
);
2084 static int prctl_set_mm(int opt
, unsigned long addr
,
2085 unsigned long arg4
, unsigned long arg5
)
2087 struct mm_struct
*mm
= current
->mm
;
2088 struct prctl_mm_map prctl_map
= {
2093 struct vm_area_struct
*vma
;
2096 if (arg5
|| (arg4
&& (opt
!= PR_SET_MM_AUXV
&&
2097 opt
!= PR_SET_MM_MAP
&&
2098 opt
!= PR_SET_MM_MAP_SIZE
)))
2101 #ifdef CONFIG_CHECKPOINT_RESTORE
2102 if (opt
== PR_SET_MM_MAP
|| opt
== PR_SET_MM_MAP_SIZE
)
2103 return prctl_set_mm_map(opt
, (const void __user
*)addr
, arg4
);
2106 if (!capable(CAP_SYS_RESOURCE
))
2109 if (opt
== PR_SET_MM_EXE_FILE
)
2110 return prctl_set_mm_exe_file(mm
, (unsigned int)addr
);
2112 if (opt
== PR_SET_MM_AUXV
)
2113 return prctl_set_auxv(mm
, addr
, arg4
);
2115 if (addr
>= TASK_SIZE
|| addr
< mmap_min_addr
)
2121 * arg_lock protects concurrent updates of arg boundaries, we need
2122 * mmap_lock for a) concurrent sys_brk, b) finding VMA for addr
2126 vma
= find_vma(mm
, addr
);
2128 spin_lock(&mm
->arg_lock
);
2129 prctl_map
.start_code
= mm
->start_code
;
2130 prctl_map
.end_code
= mm
->end_code
;
2131 prctl_map
.start_data
= mm
->start_data
;
2132 prctl_map
.end_data
= mm
->end_data
;
2133 prctl_map
.start_brk
= mm
->start_brk
;
2134 prctl_map
.brk
= mm
->brk
;
2135 prctl_map
.start_stack
= mm
->start_stack
;
2136 prctl_map
.arg_start
= mm
->arg_start
;
2137 prctl_map
.arg_end
= mm
->arg_end
;
2138 prctl_map
.env_start
= mm
->env_start
;
2139 prctl_map
.env_end
= mm
->env_end
;
2142 case PR_SET_MM_START_CODE
:
2143 prctl_map
.start_code
= addr
;
2145 case PR_SET_MM_END_CODE
:
2146 prctl_map
.end_code
= addr
;
2148 case PR_SET_MM_START_DATA
:
2149 prctl_map
.start_data
= addr
;
2151 case PR_SET_MM_END_DATA
:
2152 prctl_map
.end_data
= addr
;
2154 case PR_SET_MM_START_STACK
:
2155 prctl_map
.start_stack
= addr
;
2157 case PR_SET_MM_START_BRK
:
2158 prctl_map
.start_brk
= addr
;
2161 prctl_map
.brk
= addr
;
2163 case PR_SET_MM_ARG_START
:
2164 prctl_map
.arg_start
= addr
;
2166 case PR_SET_MM_ARG_END
:
2167 prctl_map
.arg_end
= addr
;
2169 case PR_SET_MM_ENV_START
:
2170 prctl_map
.env_start
= addr
;
2172 case PR_SET_MM_ENV_END
:
2173 prctl_map
.env_end
= addr
;
2179 error
= validate_prctl_map_addr(&prctl_map
);
2185 * If command line arguments and environment
2186 * are placed somewhere else on stack, we can
2187 * set them up here, ARG_START/END to setup
2188 * command line arguments and ENV_START/END
2191 case PR_SET_MM_START_STACK
:
2192 case PR_SET_MM_ARG_START
:
2193 case PR_SET_MM_ARG_END
:
2194 case PR_SET_MM_ENV_START
:
2195 case PR_SET_MM_ENV_END
:
2202 mm
->start_code
= prctl_map
.start_code
;
2203 mm
->end_code
= prctl_map
.end_code
;
2204 mm
->start_data
= prctl_map
.start_data
;
2205 mm
->end_data
= prctl_map
.end_data
;
2206 mm
->start_brk
= prctl_map
.start_brk
;
2207 mm
->brk
= prctl_map
.brk
;
2208 mm
->start_stack
= prctl_map
.start_stack
;
2209 mm
->arg_start
= prctl_map
.arg_start
;
2210 mm
->arg_end
= prctl_map
.arg_end
;
2211 mm
->env_start
= prctl_map
.env_start
;
2212 mm
->env_end
= prctl_map
.env_end
;
2216 spin_unlock(&mm
->arg_lock
);
2217 mmap_read_unlock(mm
);
2221 #ifdef CONFIG_CHECKPOINT_RESTORE
2222 static int prctl_get_tid_address(struct task_struct
*me
, int __user
* __user
*tid_addr
)
2224 return put_user(me
->clear_child_tid
, tid_addr
);
2227 static int prctl_get_tid_address(struct task_struct
*me
, int __user
* __user
*tid_addr
)
2233 static int propagate_has_child_subreaper(struct task_struct
*p
, void *data
)
2236 * If task has has_child_subreaper - all its descendants
2237 * already have these flag too and new descendants will
2238 * inherit it on fork, skip them.
2240 * If we've found child_reaper - skip descendants in
2241 * it's subtree as they will never get out pidns.
2243 if (p
->signal
->has_child_subreaper
||
2244 is_child_reaper(task_pid(p
)))
2247 p
->signal
->has_child_subreaper
= 1;
2251 int __weak
arch_prctl_spec_ctrl_get(struct task_struct
*t
, unsigned long which
)
2256 int __weak
arch_prctl_spec_ctrl_set(struct task_struct
*t
, unsigned long which
,
2262 #define PR_IO_FLUSHER (PF_MEMALLOC_NOIO | PF_LOCAL_THROTTLE)
2264 SYSCALL_DEFINE5(prctl
, int, option
, unsigned long, arg2
, unsigned long, arg3
,
2265 unsigned long, arg4
, unsigned long, arg5
)
2267 struct task_struct
*me
= current
;
2268 unsigned char comm
[sizeof(me
->comm
)];
2271 error
= security_task_prctl(option
, arg2
, arg3
, arg4
, arg5
);
2272 if (error
!= -ENOSYS
)
2277 case PR_SET_PDEATHSIG
:
2278 if (!valid_signal(arg2
)) {
2282 me
->pdeath_signal
= arg2
;
2284 case PR_GET_PDEATHSIG
:
2285 error
= put_user(me
->pdeath_signal
, (int __user
*)arg2
);
2287 case PR_GET_DUMPABLE
:
2288 error
= get_dumpable(me
->mm
);
2290 case PR_SET_DUMPABLE
:
2291 if (arg2
!= SUID_DUMP_DISABLE
&& arg2
!= SUID_DUMP_USER
) {
2295 set_dumpable(me
->mm
, arg2
);
2298 case PR_SET_UNALIGN
:
2299 error
= SET_UNALIGN_CTL(me
, arg2
);
2301 case PR_GET_UNALIGN
:
2302 error
= GET_UNALIGN_CTL(me
, arg2
);
2305 error
= SET_FPEMU_CTL(me
, arg2
);
2308 error
= GET_FPEMU_CTL(me
, arg2
);
2311 error
= SET_FPEXC_CTL(me
, arg2
);
2314 error
= GET_FPEXC_CTL(me
, arg2
);
2317 error
= PR_TIMING_STATISTICAL
;
2320 if (arg2
!= PR_TIMING_STATISTICAL
)
2324 comm
[sizeof(me
->comm
) - 1] = 0;
2325 if (strncpy_from_user(comm
, (char __user
*)arg2
,
2326 sizeof(me
->comm
) - 1) < 0)
2328 set_task_comm(me
, comm
);
2329 proc_comm_connector(me
);
2332 get_task_comm(comm
, me
);
2333 if (copy_to_user((char __user
*)arg2
, comm
, sizeof(comm
)))
2337 error
= GET_ENDIAN(me
, arg2
);
2340 error
= SET_ENDIAN(me
, arg2
);
2342 case PR_GET_SECCOMP
:
2343 error
= prctl_get_seccomp();
2345 case PR_SET_SECCOMP
:
2346 error
= prctl_set_seccomp(arg2
, (char __user
*)arg3
);
2349 error
= GET_TSC_CTL(arg2
);
2352 error
= SET_TSC_CTL(arg2
);
2354 case PR_TASK_PERF_EVENTS_DISABLE
:
2355 error
= perf_event_task_disable();
2357 case PR_TASK_PERF_EVENTS_ENABLE
:
2358 error
= perf_event_task_enable();
2360 case PR_GET_TIMERSLACK
:
2361 if (current
->timer_slack_ns
> ULONG_MAX
)
2364 error
= current
->timer_slack_ns
;
2366 case PR_SET_TIMERSLACK
:
2368 current
->timer_slack_ns
=
2369 current
->default_timer_slack_ns
;
2371 current
->timer_slack_ns
= arg2
;
2377 case PR_MCE_KILL_CLEAR
:
2380 current
->flags
&= ~PF_MCE_PROCESS
;
2382 case PR_MCE_KILL_SET
:
2383 current
->flags
|= PF_MCE_PROCESS
;
2384 if (arg3
== PR_MCE_KILL_EARLY
)
2385 current
->flags
|= PF_MCE_EARLY
;
2386 else if (arg3
== PR_MCE_KILL_LATE
)
2387 current
->flags
&= ~PF_MCE_EARLY
;
2388 else if (arg3
== PR_MCE_KILL_DEFAULT
)
2390 ~(PF_MCE_EARLY
|PF_MCE_PROCESS
);
2398 case PR_MCE_KILL_GET
:
2399 if (arg2
| arg3
| arg4
| arg5
)
2401 if (current
->flags
& PF_MCE_PROCESS
)
2402 error
= (current
->flags
& PF_MCE_EARLY
) ?
2403 PR_MCE_KILL_EARLY
: PR_MCE_KILL_LATE
;
2405 error
= PR_MCE_KILL_DEFAULT
;
2408 error
= prctl_set_mm(arg2
, arg3
, arg4
, arg5
);
2410 case PR_GET_TID_ADDRESS
:
2411 error
= prctl_get_tid_address(me
, (int __user
* __user
*)arg2
);
2413 case PR_SET_CHILD_SUBREAPER
:
2414 me
->signal
->is_child_subreaper
= !!arg2
;
2418 walk_process_tree(me
, propagate_has_child_subreaper
, NULL
);
2420 case PR_GET_CHILD_SUBREAPER
:
2421 error
= put_user(me
->signal
->is_child_subreaper
,
2422 (int __user
*)arg2
);
2424 case PR_SET_NO_NEW_PRIVS
:
2425 if (arg2
!= 1 || arg3
|| arg4
|| arg5
)
2428 task_set_no_new_privs(current
);
2430 case PR_GET_NO_NEW_PRIVS
:
2431 if (arg2
|| arg3
|| arg4
|| arg5
)
2433 return task_no_new_privs(current
) ? 1 : 0;
2434 case PR_GET_THP_DISABLE
:
2435 if (arg2
|| arg3
|| arg4
|| arg5
)
2437 error
= !!test_bit(MMF_DISABLE_THP
, &me
->mm
->flags
);
2439 case PR_SET_THP_DISABLE
:
2440 if (arg3
|| arg4
|| arg5
)
2442 if (mmap_write_lock_killable(me
->mm
))
2445 set_bit(MMF_DISABLE_THP
, &me
->mm
->flags
);
2447 clear_bit(MMF_DISABLE_THP
, &me
->mm
->flags
);
2448 mmap_write_unlock(me
->mm
);
2450 case PR_MPX_ENABLE_MANAGEMENT
:
2451 case PR_MPX_DISABLE_MANAGEMENT
:
2452 /* No longer implemented: */
2454 case PR_SET_FP_MODE
:
2455 error
= SET_FP_MODE(me
, arg2
);
2457 case PR_GET_FP_MODE
:
2458 error
= GET_FP_MODE(me
);
2461 error
= SVE_SET_VL(arg2
);
2464 error
= SVE_GET_VL();
2466 case PR_GET_SPECULATION_CTRL
:
2467 if (arg3
|| arg4
|| arg5
)
2469 error
= arch_prctl_spec_ctrl_get(me
, arg2
);
2471 case PR_SET_SPECULATION_CTRL
:
2474 error
= arch_prctl_spec_ctrl_set(me
, arg2
, arg3
);
2476 case PR_PAC_RESET_KEYS
:
2477 if (arg3
|| arg4
|| arg5
)
2479 error
= PAC_RESET_KEYS(me
, arg2
);
2481 case PR_PAC_SET_ENABLED_KEYS
:
2484 error
= PAC_SET_ENABLED_KEYS(me
, arg2
, arg3
);
2486 case PR_PAC_GET_ENABLED_KEYS
:
2487 if (arg2
|| arg3
|| arg4
|| arg5
)
2489 error
= PAC_GET_ENABLED_KEYS(me
);
2491 case PR_SET_TAGGED_ADDR_CTRL
:
2492 if (arg3
|| arg4
|| arg5
)
2494 error
= SET_TAGGED_ADDR_CTRL(arg2
);
2496 case PR_GET_TAGGED_ADDR_CTRL
:
2497 if (arg2
|| arg3
|| arg4
|| arg5
)
2499 error
= GET_TAGGED_ADDR_CTRL();
2501 case PR_SET_IO_FLUSHER
:
2502 if (!capable(CAP_SYS_RESOURCE
))
2505 if (arg3
|| arg4
|| arg5
)
2509 current
->flags
|= PR_IO_FLUSHER
;
2511 current
->flags
&= ~PR_IO_FLUSHER
;
2515 case PR_GET_IO_FLUSHER
:
2516 if (!capable(CAP_SYS_RESOURCE
))
2519 if (arg2
|| arg3
|| arg4
|| arg5
)
2522 error
= (current
->flags
& PR_IO_FLUSHER
) == PR_IO_FLUSHER
;
2524 case PR_SET_SYSCALL_USER_DISPATCH
:
2525 error
= set_syscall_user_dispatch(arg2
, arg3
, arg4
,
2526 (char __user
*) arg5
);
2528 #ifdef CONFIG_SCHED_CORE
2530 error
= sched_core_share_pid(arg2
, arg3
, arg4
, arg5
);
2540 SYSCALL_DEFINE3(getcpu
, unsigned __user
*, cpup
, unsigned __user
*, nodep
,
2541 struct getcpu_cache __user
*, unused
)
2544 int cpu
= raw_smp_processor_id();
2547 err
|= put_user(cpu
, cpup
);
2549 err
|= put_user(cpu_to_node(cpu
), nodep
);
2550 return err
? -EFAULT
: 0;
2554 * do_sysinfo - fill in sysinfo struct
2555 * @info: pointer to buffer to fill
2557 static int do_sysinfo(struct sysinfo
*info
)
2559 unsigned long mem_total
, sav_total
;
2560 unsigned int mem_unit
, bitcount
;
2561 struct timespec64 tp
;
2563 memset(info
, 0, sizeof(struct sysinfo
));
2565 ktime_get_boottime_ts64(&tp
);
2566 timens_add_boottime(&tp
);
2567 info
->uptime
= tp
.tv_sec
+ (tp
.tv_nsec
? 1 : 0);
2569 get_avenrun(info
->loads
, 0, SI_LOAD_SHIFT
- FSHIFT
);
2571 info
->procs
= nr_threads
;
2577 * If the sum of all the available memory (i.e. ram + swap)
2578 * is less than can be stored in a 32 bit unsigned long then
2579 * we can be binary compatible with 2.2.x kernels. If not,
2580 * well, in that case 2.2.x was broken anyways...
2582 * -Erik Andersen <andersee@debian.org>
2585 mem_total
= info
->totalram
+ info
->totalswap
;
2586 if (mem_total
< info
->totalram
|| mem_total
< info
->totalswap
)
2589 mem_unit
= info
->mem_unit
;
2590 while (mem_unit
> 1) {
2593 sav_total
= mem_total
;
2595 if (mem_total
< sav_total
)
2600 * If mem_total did not overflow, multiply all memory values by
2601 * info->mem_unit and set it to 1. This leaves things compatible
2602 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2607 info
->totalram
<<= bitcount
;
2608 info
->freeram
<<= bitcount
;
2609 info
->sharedram
<<= bitcount
;
2610 info
->bufferram
<<= bitcount
;
2611 info
->totalswap
<<= bitcount
;
2612 info
->freeswap
<<= bitcount
;
2613 info
->totalhigh
<<= bitcount
;
2614 info
->freehigh
<<= bitcount
;
2620 SYSCALL_DEFINE1(sysinfo
, struct sysinfo __user
*, info
)
2626 if (copy_to_user(info
, &val
, sizeof(struct sysinfo
)))
2632 #ifdef CONFIG_COMPAT
2633 struct compat_sysinfo
{
2647 char _f
[20-2*sizeof(u32
)-sizeof(int)];
2650 COMPAT_SYSCALL_DEFINE1(sysinfo
, struct compat_sysinfo __user
*, info
)
2653 struct compat_sysinfo s_32
;
2657 /* Check to see if any memory value is too large for 32-bit and scale
2660 if (upper_32_bits(s
.totalram
) || upper_32_bits(s
.totalswap
)) {
2663 while (s
.mem_unit
< PAGE_SIZE
) {
2668 s
.totalram
>>= bitcount
;
2669 s
.freeram
>>= bitcount
;
2670 s
.sharedram
>>= bitcount
;
2671 s
.bufferram
>>= bitcount
;
2672 s
.totalswap
>>= bitcount
;
2673 s
.freeswap
>>= bitcount
;
2674 s
.totalhigh
>>= bitcount
;
2675 s
.freehigh
>>= bitcount
;
2678 memset(&s_32
, 0, sizeof(s_32
));
2679 s_32
.uptime
= s
.uptime
;
2680 s_32
.loads
[0] = s
.loads
[0];
2681 s_32
.loads
[1] = s
.loads
[1];
2682 s_32
.loads
[2] = s
.loads
[2];
2683 s_32
.totalram
= s
.totalram
;
2684 s_32
.freeram
= s
.freeram
;
2685 s_32
.sharedram
= s
.sharedram
;
2686 s_32
.bufferram
= s
.bufferram
;
2687 s_32
.totalswap
= s
.totalswap
;
2688 s_32
.freeswap
= s
.freeswap
;
2689 s_32
.procs
= s
.procs
;
2690 s_32
.totalhigh
= s
.totalhigh
;
2691 s_32
.freehigh
= s
.freehigh
;
2692 s_32
.mem_unit
= s
.mem_unit
;
2693 if (copy_to_user(info
, &s_32
, sizeof(s_32
)))
2697 #endif /* CONFIG_COMPAT */