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 (atomic_read(&new_user
->processes
) >= rlimit(RLIMIT_NPROC
) &&
483 new_user
!= INIT_USER
)
484 current
->flags
|= PF_NPROC_EXCEEDED
;
486 current
->flags
&= ~PF_NPROC_EXCEEDED
;
489 new->user
= new_user
;
494 * Unprivileged users may change the real uid to the effective uid
495 * or vice versa. (BSD-style)
497 * If you set the real uid at all, or set the effective uid to a value not
498 * equal to the real uid, then the saved uid is set to the new effective uid.
500 * This makes it possible for a setuid program to completely drop its
501 * privileges, which is often a useful assertion to make when you are doing
502 * a security audit over a program.
504 * The general idea is that a program which uses just setreuid() will be
505 * 100% compatible with BSD. A program which uses just setuid() will be
506 * 100% compatible with POSIX with saved IDs.
508 long __sys_setreuid(uid_t ruid
, uid_t euid
)
510 struct user_namespace
*ns
= current_user_ns();
511 const struct cred
*old
;
516 kruid
= make_kuid(ns
, ruid
);
517 keuid
= make_kuid(ns
, euid
);
519 if ((ruid
!= (uid_t
) -1) && !uid_valid(kruid
))
521 if ((euid
!= (uid_t
) -1) && !uid_valid(keuid
))
524 new = prepare_creds();
527 old
= current_cred();
530 if (ruid
!= (uid_t
) -1) {
532 if (!uid_eq(old
->uid
, kruid
) &&
533 !uid_eq(old
->euid
, kruid
) &&
534 !ns_capable_setid(old
->user_ns
, CAP_SETUID
))
538 if (euid
!= (uid_t
) -1) {
540 if (!uid_eq(old
->uid
, keuid
) &&
541 !uid_eq(old
->euid
, keuid
) &&
542 !uid_eq(old
->suid
, keuid
) &&
543 !ns_capable_setid(old
->user_ns
, CAP_SETUID
))
547 if (!uid_eq(new->uid
, old
->uid
)) {
548 retval
= set_user(new);
552 if (ruid
!= (uid_t
) -1 ||
553 (euid
!= (uid_t
) -1 && !uid_eq(keuid
, old
->uid
)))
554 new->suid
= new->euid
;
555 new->fsuid
= new->euid
;
557 retval
= security_task_fix_setuid(new, old
, LSM_SETID_RE
);
561 return commit_creds(new);
568 SYSCALL_DEFINE2(setreuid
, uid_t
, ruid
, uid_t
, euid
)
570 return __sys_setreuid(ruid
, euid
);
574 * setuid() is implemented like SysV with SAVED_IDS
576 * Note that SAVED_ID's is deficient in that a setuid root program
577 * like sendmail, for example, cannot set its uid to be a normal
578 * user and then switch back, because if you're root, setuid() sets
579 * the saved uid too. If you don't like this, blame the bright people
580 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
581 * will allow a root program to temporarily drop privileges and be able to
582 * regain them by swapping the real and effective uid.
584 long __sys_setuid(uid_t uid
)
586 struct user_namespace
*ns
= current_user_ns();
587 const struct cred
*old
;
592 kuid
= make_kuid(ns
, uid
);
593 if (!uid_valid(kuid
))
596 new = prepare_creds();
599 old
= current_cred();
602 if (ns_capable_setid(old
->user_ns
, CAP_SETUID
)) {
603 new->suid
= new->uid
= kuid
;
604 if (!uid_eq(kuid
, old
->uid
)) {
605 retval
= set_user(new);
609 } else if (!uid_eq(kuid
, old
->uid
) && !uid_eq(kuid
, new->suid
)) {
613 new->fsuid
= new->euid
= kuid
;
615 retval
= security_task_fix_setuid(new, old
, LSM_SETID_ID
);
619 return commit_creds(new);
626 SYSCALL_DEFINE1(setuid
, uid_t
, uid
)
628 return __sys_setuid(uid
);
633 * This function implements a generic ability to update ruid, euid,
634 * and suid. This allows you to implement the 4.4 compatible seteuid().
636 long __sys_setresuid(uid_t ruid
, uid_t euid
, uid_t suid
)
638 struct user_namespace
*ns
= current_user_ns();
639 const struct cred
*old
;
642 kuid_t kruid
, keuid
, ksuid
;
644 kruid
= make_kuid(ns
, ruid
);
645 keuid
= make_kuid(ns
, euid
);
646 ksuid
= make_kuid(ns
, suid
);
648 if ((ruid
!= (uid_t
) -1) && !uid_valid(kruid
))
651 if ((euid
!= (uid_t
) -1) && !uid_valid(keuid
))
654 if ((suid
!= (uid_t
) -1) && !uid_valid(ksuid
))
657 new = prepare_creds();
661 old
= current_cred();
664 if (!ns_capable_setid(old
->user_ns
, CAP_SETUID
)) {
665 if (ruid
!= (uid_t
) -1 && !uid_eq(kruid
, old
->uid
) &&
666 !uid_eq(kruid
, old
->euid
) && !uid_eq(kruid
, old
->suid
))
668 if (euid
!= (uid_t
) -1 && !uid_eq(keuid
, old
->uid
) &&
669 !uid_eq(keuid
, old
->euid
) && !uid_eq(keuid
, old
->suid
))
671 if (suid
!= (uid_t
) -1 && !uid_eq(ksuid
, old
->uid
) &&
672 !uid_eq(ksuid
, old
->euid
) && !uid_eq(ksuid
, old
->suid
))
676 if (ruid
!= (uid_t
) -1) {
678 if (!uid_eq(kruid
, old
->uid
)) {
679 retval
= set_user(new);
684 if (euid
!= (uid_t
) -1)
686 if (suid
!= (uid_t
) -1)
688 new->fsuid
= new->euid
;
690 retval
= security_task_fix_setuid(new, old
, LSM_SETID_RES
);
694 return commit_creds(new);
701 SYSCALL_DEFINE3(setresuid
, uid_t
, ruid
, uid_t
, euid
, uid_t
, suid
)
703 return __sys_setresuid(ruid
, euid
, suid
);
706 SYSCALL_DEFINE3(getresuid
, uid_t __user
*, ruidp
, uid_t __user
*, euidp
, uid_t __user
*, suidp
)
708 const struct cred
*cred
= current_cred();
710 uid_t ruid
, euid
, suid
;
712 ruid
= from_kuid_munged(cred
->user_ns
, cred
->uid
);
713 euid
= from_kuid_munged(cred
->user_ns
, cred
->euid
);
714 suid
= from_kuid_munged(cred
->user_ns
, cred
->suid
);
716 retval
= put_user(ruid
, ruidp
);
718 retval
= put_user(euid
, euidp
);
720 return put_user(suid
, suidp
);
726 * Same as above, but for rgid, egid, sgid.
728 long __sys_setresgid(gid_t rgid
, gid_t egid
, gid_t sgid
)
730 struct user_namespace
*ns
= current_user_ns();
731 const struct cred
*old
;
734 kgid_t krgid
, kegid
, ksgid
;
736 krgid
= make_kgid(ns
, rgid
);
737 kegid
= make_kgid(ns
, egid
);
738 ksgid
= make_kgid(ns
, sgid
);
740 if ((rgid
!= (gid_t
) -1) && !gid_valid(krgid
))
742 if ((egid
!= (gid_t
) -1) && !gid_valid(kegid
))
744 if ((sgid
!= (gid_t
) -1) && !gid_valid(ksgid
))
747 new = prepare_creds();
750 old
= current_cred();
753 if (!ns_capable_setid(old
->user_ns
, CAP_SETGID
)) {
754 if (rgid
!= (gid_t
) -1 && !gid_eq(krgid
, old
->gid
) &&
755 !gid_eq(krgid
, old
->egid
) && !gid_eq(krgid
, old
->sgid
))
757 if (egid
!= (gid_t
) -1 && !gid_eq(kegid
, old
->gid
) &&
758 !gid_eq(kegid
, old
->egid
) && !gid_eq(kegid
, old
->sgid
))
760 if (sgid
!= (gid_t
) -1 && !gid_eq(ksgid
, old
->gid
) &&
761 !gid_eq(ksgid
, old
->egid
) && !gid_eq(ksgid
, old
->sgid
))
765 if (rgid
!= (gid_t
) -1)
767 if (egid
!= (gid_t
) -1)
769 if (sgid
!= (gid_t
) -1)
771 new->fsgid
= new->egid
;
773 retval
= security_task_fix_setgid(new, old
, LSM_SETID_RES
);
777 return commit_creds(new);
784 SYSCALL_DEFINE3(setresgid
, gid_t
, rgid
, gid_t
, egid
, gid_t
, sgid
)
786 return __sys_setresgid(rgid
, egid
, sgid
);
789 SYSCALL_DEFINE3(getresgid
, gid_t __user
*, rgidp
, gid_t __user
*, egidp
, gid_t __user
*, sgidp
)
791 const struct cred
*cred
= current_cred();
793 gid_t rgid
, egid
, sgid
;
795 rgid
= from_kgid_munged(cred
->user_ns
, cred
->gid
);
796 egid
= from_kgid_munged(cred
->user_ns
, cred
->egid
);
797 sgid
= from_kgid_munged(cred
->user_ns
, cred
->sgid
);
799 retval
= put_user(rgid
, rgidp
);
801 retval
= put_user(egid
, egidp
);
803 retval
= put_user(sgid
, sgidp
);
811 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
812 * is used for "access()" and for the NFS daemon (letting nfsd stay at
813 * whatever uid it wants to). It normally shadows "euid", except when
814 * explicitly set by setfsuid() or for access..
816 long __sys_setfsuid(uid_t uid
)
818 const struct cred
*old
;
823 old
= current_cred();
824 old_fsuid
= from_kuid_munged(old
->user_ns
, old
->fsuid
);
826 kuid
= make_kuid(old
->user_ns
, uid
);
827 if (!uid_valid(kuid
))
830 new = prepare_creds();
834 if (uid_eq(kuid
, old
->uid
) || uid_eq(kuid
, old
->euid
) ||
835 uid_eq(kuid
, old
->suid
) || uid_eq(kuid
, old
->fsuid
) ||
836 ns_capable_setid(old
->user_ns
, CAP_SETUID
)) {
837 if (!uid_eq(kuid
, old
->fsuid
)) {
839 if (security_task_fix_setuid(new, old
, LSM_SETID_FS
) == 0)
852 SYSCALL_DEFINE1(setfsuid
, uid_t
, uid
)
854 return __sys_setfsuid(uid
);
858 * Samma på svenska..
860 long __sys_setfsgid(gid_t gid
)
862 const struct cred
*old
;
867 old
= current_cred();
868 old_fsgid
= from_kgid_munged(old
->user_ns
, old
->fsgid
);
870 kgid
= make_kgid(old
->user_ns
, gid
);
871 if (!gid_valid(kgid
))
874 new = prepare_creds();
878 if (gid_eq(kgid
, old
->gid
) || gid_eq(kgid
, old
->egid
) ||
879 gid_eq(kgid
, old
->sgid
) || gid_eq(kgid
, old
->fsgid
) ||
880 ns_capable_setid(old
->user_ns
, CAP_SETGID
)) {
881 if (!gid_eq(kgid
, old
->fsgid
)) {
883 if (security_task_fix_setgid(new,old
,LSM_SETID_FS
) == 0)
896 SYSCALL_DEFINE1(setfsgid
, gid_t
, gid
)
898 return __sys_setfsgid(gid
);
900 #endif /* CONFIG_MULTIUSER */
903 * sys_getpid - return the thread group id of the current process
905 * Note, despite the name, this returns the tgid not the pid. The tgid and
906 * the pid are identical unless CLONE_THREAD was specified on clone() in
907 * which case the tgid is the same in all threads of the same group.
909 * This is SMP safe as current->tgid does not change.
911 SYSCALL_DEFINE0(getpid
)
913 return task_tgid_vnr(current
);
916 /* Thread ID - the internal kernel "pid" */
917 SYSCALL_DEFINE0(gettid
)
919 return task_pid_vnr(current
);
923 * Accessing ->real_parent is not SMP-safe, it could
924 * change from under us. However, we can use a stale
925 * value of ->real_parent under rcu_read_lock(), see
926 * release_task()->call_rcu(delayed_put_task_struct).
928 SYSCALL_DEFINE0(getppid
)
933 pid
= task_tgid_vnr(rcu_dereference(current
->real_parent
));
939 SYSCALL_DEFINE0(getuid
)
941 /* Only we change this so SMP safe */
942 return from_kuid_munged(current_user_ns(), current_uid());
945 SYSCALL_DEFINE0(geteuid
)
947 /* Only we change this so SMP safe */
948 return from_kuid_munged(current_user_ns(), current_euid());
951 SYSCALL_DEFINE0(getgid
)
953 /* Only we change this so SMP safe */
954 return from_kgid_munged(current_user_ns(), current_gid());
957 SYSCALL_DEFINE0(getegid
)
959 /* Only we change this so SMP safe */
960 return from_kgid_munged(current_user_ns(), current_egid());
963 static void do_sys_times(struct tms
*tms
)
965 u64 tgutime
, tgstime
, cutime
, cstime
;
967 thread_group_cputime_adjusted(current
, &tgutime
, &tgstime
);
968 cutime
= current
->signal
->cutime
;
969 cstime
= current
->signal
->cstime
;
970 tms
->tms_utime
= nsec_to_clock_t(tgutime
);
971 tms
->tms_stime
= nsec_to_clock_t(tgstime
);
972 tms
->tms_cutime
= nsec_to_clock_t(cutime
);
973 tms
->tms_cstime
= nsec_to_clock_t(cstime
);
976 SYSCALL_DEFINE1(times
, struct tms __user
*, tbuf
)
982 if (copy_to_user(tbuf
, &tmp
, sizeof(struct tms
)))
985 force_successful_syscall_return();
986 return (long) jiffies_64_to_clock_t(get_jiffies_64());
990 static compat_clock_t
clock_t_to_compat_clock_t(clock_t x
)
992 return compat_jiffies_to_clock_t(clock_t_to_jiffies(x
));
995 COMPAT_SYSCALL_DEFINE1(times
, struct compat_tms __user
*, tbuf
)
999 struct compat_tms tmp
;
1002 /* Convert our struct tms to the compat version. */
1003 tmp
.tms_utime
= clock_t_to_compat_clock_t(tms
.tms_utime
);
1004 tmp
.tms_stime
= clock_t_to_compat_clock_t(tms
.tms_stime
);
1005 tmp
.tms_cutime
= clock_t_to_compat_clock_t(tms
.tms_cutime
);
1006 tmp
.tms_cstime
= clock_t_to_compat_clock_t(tms
.tms_cstime
);
1007 if (copy_to_user(tbuf
, &tmp
, sizeof(tmp
)))
1010 force_successful_syscall_return();
1011 return compat_jiffies_to_clock_t(jiffies
);
1016 * This needs some heavy checking ...
1017 * I just haven't the stomach for it. I also don't fully
1018 * understand sessions/pgrp etc. Let somebody who does explain it.
1020 * OK, I think I have the protection semantics right.... this is really
1021 * only important on a multi-user system anyway, to make sure one user
1022 * can't send a signal to a process owned by another. -TYT, 12/12/91
1024 * !PF_FORKNOEXEC check to conform completely to POSIX.
1026 SYSCALL_DEFINE2(setpgid
, pid_t
, pid
, pid_t
, pgid
)
1028 struct task_struct
*p
;
1029 struct task_struct
*group_leader
= current
->group_leader
;
1034 pid
= task_pid_vnr(group_leader
);
1041 /* From this point forward we keep holding onto the tasklist lock
1042 * so that our parent does not change from under us. -DaveM
1044 write_lock_irq(&tasklist_lock
);
1047 p
= find_task_by_vpid(pid
);
1052 if (!thread_group_leader(p
))
1055 if (same_thread_group(p
->real_parent
, group_leader
)) {
1057 if (task_session(p
) != task_session(group_leader
))
1060 if (!(p
->flags
& PF_FORKNOEXEC
))
1064 if (p
!= group_leader
)
1069 if (p
->signal
->leader
)
1074 struct task_struct
*g
;
1076 pgrp
= find_vpid(pgid
);
1077 g
= pid_task(pgrp
, PIDTYPE_PGID
);
1078 if (!g
|| task_session(g
) != task_session(group_leader
))
1082 err
= security_task_setpgid(p
, pgid
);
1086 if (task_pgrp(p
) != pgrp
)
1087 change_pid(p
, PIDTYPE_PGID
, pgrp
);
1091 /* All paths lead to here, thus we are safe. -DaveM */
1092 write_unlock_irq(&tasklist_lock
);
1097 static int do_getpgid(pid_t pid
)
1099 struct task_struct
*p
;
1105 grp
= task_pgrp(current
);
1108 p
= find_task_by_vpid(pid
);
1115 retval
= security_task_getpgid(p
);
1119 retval
= pid_vnr(grp
);
1125 SYSCALL_DEFINE1(getpgid
, pid_t
, pid
)
1127 return do_getpgid(pid
);
1130 #ifdef __ARCH_WANT_SYS_GETPGRP
1132 SYSCALL_DEFINE0(getpgrp
)
1134 return do_getpgid(0);
1139 SYSCALL_DEFINE1(getsid
, pid_t
, pid
)
1141 struct task_struct
*p
;
1147 sid
= task_session(current
);
1150 p
= find_task_by_vpid(pid
);
1153 sid
= task_session(p
);
1157 retval
= security_task_getsid(p
);
1161 retval
= pid_vnr(sid
);
1167 static void set_special_pids(struct pid
*pid
)
1169 struct task_struct
*curr
= current
->group_leader
;
1171 if (task_session(curr
) != pid
)
1172 change_pid(curr
, PIDTYPE_SID
, pid
);
1174 if (task_pgrp(curr
) != pid
)
1175 change_pid(curr
, PIDTYPE_PGID
, pid
);
1178 int ksys_setsid(void)
1180 struct task_struct
*group_leader
= current
->group_leader
;
1181 struct pid
*sid
= task_pid(group_leader
);
1182 pid_t session
= pid_vnr(sid
);
1185 write_lock_irq(&tasklist_lock
);
1186 /* Fail if I am already a session leader */
1187 if (group_leader
->signal
->leader
)
1190 /* Fail if a process group id already exists that equals the
1191 * proposed session id.
1193 if (pid_task(sid
, PIDTYPE_PGID
))
1196 group_leader
->signal
->leader
= 1;
1197 set_special_pids(sid
);
1199 proc_clear_tty(group_leader
);
1203 write_unlock_irq(&tasklist_lock
);
1205 proc_sid_connector(group_leader
);
1206 sched_autogroup_create_attach(group_leader
);
1211 SYSCALL_DEFINE0(setsid
)
1213 return ksys_setsid();
1216 DECLARE_RWSEM(uts_sem
);
1218 #ifdef COMPAT_UTS_MACHINE
1219 #define override_architecture(name) \
1220 (personality(current->personality) == PER_LINUX32 && \
1221 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1222 sizeof(COMPAT_UTS_MACHINE)))
1224 #define override_architecture(name) 0
1228 * Work around broken programs that cannot handle "Linux 3.0".
1229 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1230 * And we map 4.x and later versions to 2.6.60+x, so 4.0/5.0/6.0/... would be
1233 static int override_release(char __user
*release
, size_t len
)
1237 if (current
->personality
& UNAME26
) {
1238 const char *rest
= UTS_RELEASE
;
1239 char buf
[65] = { 0 };
1245 if (*rest
== '.' && ++ndots
>= 3)
1247 if (!isdigit(*rest
) && *rest
!= '.')
1251 v
= LINUX_VERSION_PATCHLEVEL
+ 60;
1252 copy
= clamp_t(size_t, len
, 1, sizeof(buf
));
1253 copy
= scnprintf(buf
, copy
, "2.6.%u%s", v
, rest
);
1254 ret
= copy_to_user(release
, buf
, copy
+ 1);
1259 SYSCALL_DEFINE1(newuname
, struct new_utsname __user
*, name
)
1261 struct new_utsname tmp
;
1263 down_read(&uts_sem
);
1264 memcpy(&tmp
, utsname(), sizeof(tmp
));
1266 if (copy_to_user(name
, &tmp
, sizeof(tmp
)))
1269 if (override_release(name
->release
, sizeof(name
->release
)))
1271 if (override_architecture(name
))
1276 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1280 SYSCALL_DEFINE1(uname
, struct old_utsname __user
*, name
)
1282 struct old_utsname tmp
;
1287 down_read(&uts_sem
);
1288 memcpy(&tmp
, utsname(), sizeof(tmp
));
1290 if (copy_to_user(name
, &tmp
, sizeof(tmp
)))
1293 if (override_release(name
->release
, sizeof(name
->release
)))
1295 if (override_architecture(name
))
1300 SYSCALL_DEFINE1(olduname
, struct oldold_utsname __user
*, name
)
1302 struct oldold_utsname tmp
;
1307 memset(&tmp
, 0, sizeof(tmp
));
1309 down_read(&uts_sem
);
1310 memcpy(&tmp
.sysname
, &utsname()->sysname
, __OLD_UTS_LEN
);
1311 memcpy(&tmp
.nodename
, &utsname()->nodename
, __OLD_UTS_LEN
);
1312 memcpy(&tmp
.release
, &utsname()->release
, __OLD_UTS_LEN
);
1313 memcpy(&tmp
.version
, &utsname()->version
, __OLD_UTS_LEN
);
1314 memcpy(&tmp
.machine
, &utsname()->machine
, __OLD_UTS_LEN
);
1316 if (copy_to_user(name
, &tmp
, sizeof(tmp
)))
1319 if (override_architecture(name
))
1321 if (override_release(name
->release
, sizeof(name
->release
)))
1327 SYSCALL_DEFINE2(sethostname
, char __user
*, name
, int, len
)
1330 char tmp
[__NEW_UTS_LEN
];
1332 if (!ns_capable(current
->nsproxy
->uts_ns
->user_ns
, CAP_SYS_ADMIN
))
1335 if (len
< 0 || len
> __NEW_UTS_LEN
)
1338 if (!copy_from_user(tmp
, name
, len
)) {
1339 struct new_utsname
*u
;
1341 down_write(&uts_sem
);
1343 memcpy(u
->nodename
, tmp
, len
);
1344 memset(u
->nodename
+ len
, 0, sizeof(u
->nodename
) - len
);
1346 uts_proc_notify(UTS_PROC_HOSTNAME
);
1352 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1354 SYSCALL_DEFINE2(gethostname
, char __user
*, name
, int, len
)
1357 struct new_utsname
*u
;
1358 char tmp
[__NEW_UTS_LEN
+ 1];
1362 down_read(&uts_sem
);
1364 i
= 1 + strlen(u
->nodename
);
1367 memcpy(tmp
, u
->nodename
, i
);
1369 if (copy_to_user(name
, tmp
, i
))
1377 * Only setdomainname; getdomainname can be implemented by calling
1380 SYSCALL_DEFINE2(setdomainname
, char __user
*, name
, int, len
)
1383 char tmp
[__NEW_UTS_LEN
];
1385 if (!ns_capable(current
->nsproxy
->uts_ns
->user_ns
, CAP_SYS_ADMIN
))
1387 if (len
< 0 || len
> __NEW_UTS_LEN
)
1391 if (!copy_from_user(tmp
, name
, len
)) {
1392 struct new_utsname
*u
;
1394 down_write(&uts_sem
);
1396 memcpy(u
->domainname
, tmp
, len
);
1397 memset(u
->domainname
+ len
, 0, sizeof(u
->domainname
) - len
);
1399 uts_proc_notify(UTS_PROC_DOMAINNAME
);
1405 SYSCALL_DEFINE2(getrlimit
, unsigned int, resource
, struct rlimit __user
*, rlim
)
1407 struct rlimit value
;
1410 ret
= do_prlimit(current
, resource
, NULL
, &value
);
1412 ret
= copy_to_user(rlim
, &value
, sizeof(*rlim
)) ? -EFAULT
: 0;
1417 #ifdef CONFIG_COMPAT
1419 COMPAT_SYSCALL_DEFINE2(setrlimit
, unsigned int, resource
,
1420 struct compat_rlimit __user
*, rlim
)
1423 struct compat_rlimit r32
;
1425 if (copy_from_user(&r32
, rlim
, sizeof(struct compat_rlimit
)))
1428 if (r32
.rlim_cur
== COMPAT_RLIM_INFINITY
)
1429 r
.rlim_cur
= RLIM_INFINITY
;
1431 r
.rlim_cur
= r32
.rlim_cur
;
1432 if (r32
.rlim_max
== COMPAT_RLIM_INFINITY
)
1433 r
.rlim_max
= RLIM_INFINITY
;
1435 r
.rlim_max
= r32
.rlim_max
;
1436 return do_prlimit(current
, resource
, &r
, NULL
);
1439 COMPAT_SYSCALL_DEFINE2(getrlimit
, unsigned int, resource
,
1440 struct compat_rlimit __user
*, rlim
)
1445 ret
= do_prlimit(current
, resource
, NULL
, &r
);
1447 struct compat_rlimit r32
;
1448 if (r
.rlim_cur
> COMPAT_RLIM_INFINITY
)
1449 r32
.rlim_cur
= COMPAT_RLIM_INFINITY
;
1451 r32
.rlim_cur
= r
.rlim_cur
;
1452 if (r
.rlim_max
> COMPAT_RLIM_INFINITY
)
1453 r32
.rlim_max
= COMPAT_RLIM_INFINITY
;
1455 r32
.rlim_max
= r
.rlim_max
;
1457 if (copy_to_user(rlim
, &r32
, sizeof(struct compat_rlimit
)))
1465 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1468 * Back compatibility for getrlimit. Needed for some apps.
1470 SYSCALL_DEFINE2(old_getrlimit
, unsigned int, resource
,
1471 struct rlimit __user
*, rlim
)
1474 if (resource
>= RLIM_NLIMITS
)
1477 resource
= array_index_nospec(resource
, RLIM_NLIMITS
);
1478 task_lock(current
->group_leader
);
1479 x
= current
->signal
->rlim
[resource
];
1480 task_unlock(current
->group_leader
);
1481 if (x
.rlim_cur
> 0x7FFFFFFF)
1482 x
.rlim_cur
= 0x7FFFFFFF;
1483 if (x
.rlim_max
> 0x7FFFFFFF)
1484 x
.rlim_max
= 0x7FFFFFFF;
1485 return copy_to_user(rlim
, &x
, sizeof(x
)) ? -EFAULT
: 0;
1488 #ifdef CONFIG_COMPAT
1489 COMPAT_SYSCALL_DEFINE2(old_getrlimit
, unsigned int, resource
,
1490 struct compat_rlimit __user
*, rlim
)
1494 if (resource
>= RLIM_NLIMITS
)
1497 resource
= array_index_nospec(resource
, RLIM_NLIMITS
);
1498 task_lock(current
->group_leader
);
1499 r
= current
->signal
->rlim
[resource
];
1500 task_unlock(current
->group_leader
);
1501 if (r
.rlim_cur
> 0x7FFFFFFF)
1502 r
.rlim_cur
= 0x7FFFFFFF;
1503 if (r
.rlim_max
> 0x7FFFFFFF)
1504 r
.rlim_max
= 0x7FFFFFFF;
1506 if (put_user(r
.rlim_cur
, &rlim
->rlim_cur
) ||
1507 put_user(r
.rlim_max
, &rlim
->rlim_max
))
1515 static inline bool rlim64_is_infinity(__u64 rlim64
)
1517 #if BITS_PER_LONG < 64
1518 return rlim64
>= ULONG_MAX
;
1520 return rlim64
== RLIM64_INFINITY
;
1524 static void rlim_to_rlim64(const struct rlimit
*rlim
, struct rlimit64
*rlim64
)
1526 if (rlim
->rlim_cur
== RLIM_INFINITY
)
1527 rlim64
->rlim_cur
= RLIM64_INFINITY
;
1529 rlim64
->rlim_cur
= rlim
->rlim_cur
;
1530 if (rlim
->rlim_max
== RLIM_INFINITY
)
1531 rlim64
->rlim_max
= RLIM64_INFINITY
;
1533 rlim64
->rlim_max
= rlim
->rlim_max
;
1536 static void rlim64_to_rlim(const struct rlimit64
*rlim64
, struct rlimit
*rlim
)
1538 if (rlim64_is_infinity(rlim64
->rlim_cur
))
1539 rlim
->rlim_cur
= RLIM_INFINITY
;
1541 rlim
->rlim_cur
= (unsigned long)rlim64
->rlim_cur
;
1542 if (rlim64_is_infinity(rlim64
->rlim_max
))
1543 rlim
->rlim_max
= RLIM_INFINITY
;
1545 rlim
->rlim_max
= (unsigned long)rlim64
->rlim_max
;
1548 /* make sure you are allowed to change @tsk limits before calling this */
1549 int do_prlimit(struct task_struct
*tsk
, unsigned int resource
,
1550 struct rlimit
*new_rlim
, struct rlimit
*old_rlim
)
1552 struct rlimit
*rlim
;
1555 if (resource
>= RLIM_NLIMITS
)
1558 if (new_rlim
->rlim_cur
> new_rlim
->rlim_max
)
1560 if (resource
== RLIMIT_NOFILE
&&
1561 new_rlim
->rlim_max
> sysctl_nr_open
)
1565 /* protect tsk->signal and tsk->sighand from disappearing */
1566 read_lock(&tasklist_lock
);
1567 if (!tsk
->sighand
) {
1572 rlim
= tsk
->signal
->rlim
+ resource
;
1573 task_lock(tsk
->group_leader
);
1575 /* Keep the capable check against init_user_ns until
1576 cgroups can contain all limits */
1577 if (new_rlim
->rlim_max
> rlim
->rlim_max
&&
1578 !capable(CAP_SYS_RESOURCE
))
1581 retval
= security_task_setrlimit(tsk
, resource
, new_rlim
);
1589 task_unlock(tsk
->group_leader
);
1592 * RLIMIT_CPU handling. Arm the posix CPU timer if the limit is not
1593 * infinite. In case of RLIM_INFINITY the posix CPU timer code
1594 * ignores the rlimit.
1596 if (!retval
&& new_rlim
&& resource
== RLIMIT_CPU
&&
1597 new_rlim
->rlim_cur
!= RLIM_INFINITY
&&
1598 IS_ENABLED(CONFIG_POSIX_TIMERS
))
1599 update_rlimit_cpu(tsk
, new_rlim
->rlim_cur
);
1601 read_unlock(&tasklist_lock
);
1605 /* rcu lock must be held */
1606 static int check_prlimit_permission(struct task_struct
*task
,
1609 const struct cred
*cred
= current_cred(), *tcred
;
1612 if (current
== task
)
1615 tcred
= __task_cred(task
);
1616 id_match
= (uid_eq(cred
->uid
, tcred
->euid
) &&
1617 uid_eq(cred
->uid
, tcred
->suid
) &&
1618 uid_eq(cred
->uid
, tcred
->uid
) &&
1619 gid_eq(cred
->gid
, tcred
->egid
) &&
1620 gid_eq(cred
->gid
, tcred
->sgid
) &&
1621 gid_eq(cred
->gid
, tcred
->gid
));
1622 if (!id_match
&& !ns_capable(tcred
->user_ns
, CAP_SYS_RESOURCE
))
1625 return security_task_prlimit(cred
, tcred
, flags
);
1628 SYSCALL_DEFINE4(prlimit64
, pid_t
, pid
, unsigned int, resource
,
1629 const struct rlimit64 __user
*, new_rlim
,
1630 struct rlimit64 __user
*, old_rlim
)
1632 struct rlimit64 old64
, new64
;
1633 struct rlimit old
, new;
1634 struct task_struct
*tsk
;
1635 unsigned int checkflags
= 0;
1639 checkflags
|= LSM_PRLIMIT_READ
;
1642 if (copy_from_user(&new64
, new_rlim
, sizeof(new64
)))
1644 rlim64_to_rlim(&new64
, &new);
1645 checkflags
|= LSM_PRLIMIT_WRITE
;
1649 tsk
= pid
? find_task_by_vpid(pid
) : current
;
1654 ret
= check_prlimit_permission(tsk
, checkflags
);
1659 get_task_struct(tsk
);
1662 ret
= do_prlimit(tsk
, resource
, new_rlim
? &new : NULL
,
1663 old_rlim
? &old
: NULL
);
1665 if (!ret
&& old_rlim
) {
1666 rlim_to_rlim64(&old
, &old64
);
1667 if (copy_to_user(old_rlim
, &old64
, sizeof(old64
)))
1671 put_task_struct(tsk
);
1675 SYSCALL_DEFINE2(setrlimit
, unsigned int, resource
, struct rlimit __user
*, rlim
)
1677 struct rlimit new_rlim
;
1679 if (copy_from_user(&new_rlim
, rlim
, sizeof(*rlim
)))
1681 return do_prlimit(current
, resource
, &new_rlim
, NULL
);
1685 * It would make sense to put struct rusage in the task_struct,
1686 * except that would make the task_struct be *really big*. After
1687 * task_struct gets moved into malloc'ed memory, it would
1688 * make sense to do this. It will make moving the rest of the information
1689 * a lot simpler! (Which we're not doing right now because we're not
1690 * measuring them yet).
1692 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1693 * races with threads incrementing their own counters. But since word
1694 * reads are atomic, we either get new values or old values and we don't
1695 * care which for the sums. We always take the siglock to protect reading
1696 * the c* fields from p->signal from races with exit.c updating those
1697 * fields when reaping, so a sample either gets all the additions of a
1698 * given child after it's reaped, or none so this sample is before reaping.
1701 * We need to take the siglock for CHILDEREN, SELF and BOTH
1702 * for the cases current multithreaded, non-current single threaded
1703 * non-current multithreaded. Thread traversal is now safe with
1705 * Strictly speaking, we donot need to take the siglock if we are current and
1706 * single threaded, as no one else can take our signal_struct away, no one
1707 * else can reap the children to update signal->c* counters, and no one else
1708 * can race with the signal-> fields. If we do not take any lock, the
1709 * signal-> fields could be read out of order while another thread was just
1710 * exiting. So we should place a read memory barrier when we avoid the lock.
1711 * On the writer side, write memory barrier is implied in __exit_signal
1712 * as __exit_signal releases the siglock spinlock after updating the signal->
1713 * fields. But we don't do this yet to keep things simple.
1717 static void accumulate_thread_rusage(struct task_struct
*t
, struct rusage
*r
)
1719 r
->ru_nvcsw
+= t
->nvcsw
;
1720 r
->ru_nivcsw
+= t
->nivcsw
;
1721 r
->ru_minflt
+= t
->min_flt
;
1722 r
->ru_majflt
+= t
->maj_flt
;
1723 r
->ru_inblock
+= task_io_get_inblock(t
);
1724 r
->ru_oublock
+= task_io_get_oublock(t
);
1727 void getrusage(struct task_struct
*p
, int who
, struct rusage
*r
)
1729 struct task_struct
*t
;
1730 unsigned long flags
;
1731 u64 tgutime
, tgstime
, utime
, stime
;
1732 unsigned long maxrss
= 0;
1734 memset((char *)r
, 0, sizeof (*r
));
1737 if (who
== RUSAGE_THREAD
) {
1738 task_cputime_adjusted(current
, &utime
, &stime
);
1739 accumulate_thread_rusage(p
, r
);
1740 maxrss
= p
->signal
->maxrss
;
1744 if (!lock_task_sighand(p
, &flags
))
1749 case RUSAGE_CHILDREN
:
1750 utime
= p
->signal
->cutime
;
1751 stime
= p
->signal
->cstime
;
1752 r
->ru_nvcsw
= p
->signal
->cnvcsw
;
1753 r
->ru_nivcsw
= p
->signal
->cnivcsw
;
1754 r
->ru_minflt
= p
->signal
->cmin_flt
;
1755 r
->ru_majflt
= p
->signal
->cmaj_flt
;
1756 r
->ru_inblock
= p
->signal
->cinblock
;
1757 r
->ru_oublock
= p
->signal
->coublock
;
1758 maxrss
= p
->signal
->cmaxrss
;
1760 if (who
== RUSAGE_CHILDREN
)
1765 thread_group_cputime_adjusted(p
, &tgutime
, &tgstime
);
1768 r
->ru_nvcsw
+= p
->signal
->nvcsw
;
1769 r
->ru_nivcsw
+= p
->signal
->nivcsw
;
1770 r
->ru_minflt
+= p
->signal
->min_flt
;
1771 r
->ru_majflt
+= p
->signal
->maj_flt
;
1772 r
->ru_inblock
+= p
->signal
->inblock
;
1773 r
->ru_oublock
+= p
->signal
->oublock
;
1774 if (maxrss
< p
->signal
->maxrss
)
1775 maxrss
= p
->signal
->maxrss
;
1778 accumulate_thread_rusage(t
, r
);
1779 } while_each_thread(p
, t
);
1785 unlock_task_sighand(p
, &flags
);
1788 r
->ru_utime
= ns_to_kernel_old_timeval(utime
);
1789 r
->ru_stime
= ns_to_kernel_old_timeval(stime
);
1791 if (who
!= RUSAGE_CHILDREN
) {
1792 struct mm_struct
*mm
= get_task_mm(p
);
1795 setmax_mm_hiwater_rss(&maxrss
, mm
);
1799 r
->ru_maxrss
= maxrss
* (PAGE_SIZE
/ 1024); /* convert pages to KBs */
1802 SYSCALL_DEFINE2(getrusage
, int, who
, struct rusage __user
*, ru
)
1806 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
&&
1807 who
!= RUSAGE_THREAD
)
1810 getrusage(current
, who
, &r
);
1811 return copy_to_user(ru
, &r
, sizeof(r
)) ? -EFAULT
: 0;
1814 #ifdef CONFIG_COMPAT
1815 COMPAT_SYSCALL_DEFINE2(getrusage
, int, who
, struct compat_rusage __user
*, ru
)
1819 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
&&
1820 who
!= RUSAGE_THREAD
)
1823 getrusage(current
, who
, &r
);
1824 return put_compat_rusage(&r
, ru
);
1828 SYSCALL_DEFINE1(umask
, int, mask
)
1830 mask
= xchg(¤t
->fs
->umask
, mask
& S_IRWXUGO
);
1834 static int prctl_set_mm_exe_file(struct mm_struct
*mm
, unsigned int fd
)
1837 struct file
*old_exe
, *exe_file
;
1838 struct inode
*inode
;
1845 inode
= file_inode(exe
.file
);
1848 * Because the original mm->exe_file points to executable file, make
1849 * sure that this one is executable as well, to avoid breaking an
1853 if (!S_ISREG(inode
->i_mode
) || path_noexec(&exe
.file
->f_path
))
1856 err
= file_permission(exe
.file
, MAY_EXEC
);
1861 * Forbid mm->exe_file change if old file still mapped.
1863 exe_file
= get_mm_exe_file(mm
);
1866 struct vm_area_struct
*vma
;
1869 for (vma
= mm
->mmap
; vma
; vma
= vma
->vm_next
) {
1872 if (path_equal(&vma
->vm_file
->f_path
,
1877 mmap_read_unlock(mm
);
1882 /* set the new file, lockless */
1884 old_exe
= xchg(&mm
->exe_file
, exe
.file
);
1891 mmap_read_unlock(mm
);
1897 * Check arithmetic relations of passed addresses.
1899 * WARNING: we don't require any capability here so be very careful
1900 * in what is allowed for modification from userspace.
1902 static int validate_prctl_map_addr(struct prctl_mm_map
*prctl_map
)
1904 unsigned long mmap_max_addr
= TASK_SIZE
;
1905 int error
= -EINVAL
, i
;
1907 static const unsigned char offsets
[] = {
1908 offsetof(struct prctl_mm_map
, start_code
),
1909 offsetof(struct prctl_mm_map
, end_code
),
1910 offsetof(struct prctl_mm_map
, start_data
),
1911 offsetof(struct prctl_mm_map
, end_data
),
1912 offsetof(struct prctl_mm_map
, start_brk
),
1913 offsetof(struct prctl_mm_map
, brk
),
1914 offsetof(struct prctl_mm_map
, start_stack
),
1915 offsetof(struct prctl_mm_map
, arg_start
),
1916 offsetof(struct prctl_mm_map
, arg_end
),
1917 offsetof(struct prctl_mm_map
, env_start
),
1918 offsetof(struct prctl_mm_map
, env_end
),
1922 * Make sure the members are not somewhere outside
1923 * of allowed address space.
1925 for (i
= 0; i
< ARRAY_SIZE(offsets
); i
++) {
1926 u64 val
= *(u64
*)((char *)prctl_map
+ offsets
[i
]);
1928 if ((unsigned long)val
>= mmap_max_addr
||
1929 (unsigned long)val
< mmap_min_addr
)
1934 * Make sure the pairs are ordered.
1936 #define __prctl_check_order(__m1, __op, __m2) \
1937 ((unsigned long)prctl_map->__m1 __op \
1938 (unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1939 error
= __prctl_check_order(start_code
, <, end_code
);
1940 error
|= __prctl_check_order(start_data
,<=, end_data
);
1941 error
|= __prctl_check_order(start_brk
, <=, brk
);
1942 error
|= __prctl_check_order(arg_start
, <=, arg_end
);
1943 error
|= __prctl_check_order(env_start
, <=, env_end
);
1946 #undef __prctl_check_order
1951 * @brk should be after @end_data in traditional maps.
1953 if (prctl_map
->start_brk
<= prctl_map
->end_data
||
1954 prctl_map
->brk
<= prctl_map
->end_data
)
1958 * Neither we should allow to override limits if they set.
1960 if (check_data_rlimit(rlimit(RLIMIT_DATA
), prctl_map
->brk
,
1961 prctl_map
->start_brk
, prctl_map
->end_data
,
1962 prctl_map
->start_data
))
1970 #ifdef CONFIG_CHECKPOINT_RESTORE
1971 static int prctl_set_mm_map(int opt
, const void __user
*addr
, unsigned long data_size
)
1973 struct prctl_mm_map prctl_map
= { .exe_fd
= (u32
)-1, };
1974 unsigned long user_auxv
[AT_VECTOR_SIZE
];
1975 struct mm_struct
*mm
= current
->mm
;
1978 BUILD_BUG_ON(sizeof(user_auxv
) != sizeof(mm
->saved_auxv
));
1979 BUILD_BUG_ON(sizeof(struct prctl_mm_map
) > 256);
1981 if (opt
== PR_SET_MM_MAP_SIZE
)
1982 return put_user((unsigned int)sizeof(prctl_map
),
1983 (unsigned int __user
*)addr
);
1985 if (data_size
!= sizeof(prctl_map
))
1988 if (copy_from_user(&prctl_map
, addr
, sizeof(prctl_map
)))
1991 error
= validate_prctl_map_addr(&prctl_map
);
1995 if (prctl_map
.auxv_size
) {
1997 * Someone is trying to cheat the auxv vector.
1999 if (!prctl_map
.auxv
||
2000 prctl_map
.auxv_size
> sizeof(mm
->saved_auxv
))
2003 memset(user_auxv
, 0, sizeof(user_auxv
));
2004 if (copy_from_user(user_auxv
,
2005 (const void __user
*)prctl_map
.auxv
,
2006 prctl_map
.auxv_size
))
2009 /* Last entry must be AT_NULL as specification requires */
2010 user_auxv
[AT_VECTOR_SIZE
- 2] = AT_NULL
;
2011 user_auxv
[AT_VECTOR_SIZE
- 1] = AT_NULL
;
2014 if (prctl_map
.exe_fd
!= (u32
)-1) {
2016 * Check if the current user is checkpoint/restore capable.
2017 * At the time of this writing, it checks for CAP_SYS_ADMIN
2018 * or CAP_CHECKPOINT_RESTORE.
2019 * Note that a user with access to ptrace can masquerade an
2020 * arbitrary program as any executable, even setuid ones.
2021 * This may have implications in the tomoyo subsystem.
2023 if (!checkpoint_restore_ns_capable(current_user_ns()))
2026 error
= prctl_set_mm_exe_file(mm
, prctl_map
.exe_fd
);
2032 * arg_lock protects concurrent updates but we still need mmap_lock for
2033 * read to exclude races with sys_brk.
2038 * We don't validate if these members are pointing to
2039 * real present VMAs because application may have correspond
2040 * VMAs already unmapped and kernel uses these members for statistics
2041 * output in procfs mostly, except
2043 * - @start_brk/@brk which are used in do_brk_flags but kernel lookups
2044 * for VMAs when updating these members so anything wrong written
2045 * here cause kernel to swear at userspace program but won't lead
2046 * to any problem in kernel itself
2049 spin_lock(&mm
->arg_lock
);
2050 mm
->start_code
= prctl_map
.start_code
;
2051 mm
->end_code
= prctl_map
.end_code
;
2052 mm
->start_data
= prctl_map
.start_data
;
2053 mm
->end_data
= prctl_map
.end_data
;
2054 mm
->start_brk
= prctl_map
.start_brk
;
2055 mm
->brk
= prctl_map
.brk
;
2056 mm
->start_stack
= prctl_map
.start_stack
;
2057 mm
->arg_start
= prctl_map
.arg_start
;
2058 mm
->arg_end
= prctl_map
.arg_end
;
2059 mm
->env_start
= prctl_map
.env_start
;
2060 mm
->env_end
= prctl_map
.env_end
;
2061 spin_unlock(&mm
->arg_lock
);
2064 * Note this update of @saved_auxv is lockless thus
2065 * if someone reads this member in procfs while we're
2066 * updating -- it may get partly updated results. It's
2067 * known and acceptable trade off: we leave it as is to
2068 * not introduce additional locks here making the kernel
2071 if (prctl_map
.auxv_size
)
2072 memcpy(mm
->saved_auxv
, user_auxv
, sizeof(user_auxv
));
2074 mmap_read_unlock(mm
);
2077 #endif /* CONFIG_CHECKPOINT_RESTORE */
2079 static int prctl_set_auxv(struct mm_struct
*mm
, unsigned long addr
,
2083 * This doesn't move the auxiliary vector itself since it's pinned to
2084 * mm_struct, but it permits filling the vector with new values. It's
2085 * up to the caller to provide sane values here, otherwise userspace
2086 * tools which use this vector might be unhappy.
2088 unsigned long user_auxv
[AT_VECTOR_SIZE
] = {};
2090 if (len
> sizeof(user_auxv
))
2093 if (copy_from_user(user_auxv
, (const void __user
*)addr
, len
))
2096 /* Make sure the last entry is always AT_NULL */
2097 user_auxv
[AT_VECTOR_SIZE
- 2] = 0;
2098 user_auxv
[AT_VECTOR_SIZE
- 1] = 0;
2100 BUILD_BUG_ON(sizeof(user_auxv
) != sizeof(mm
->saved_auxv
));
2103 memcpy(mm
->saved_auxv
, user_auxv
, len
);
2104 task_unlock(current
);
2109 static int prctl_set_mm(int opt
, unsigned long addr
,
2110 unsigned long arg4
, unsigned long arg5
)
2112 struct mm_struct
*mm
= current
->mm
;
2113 struct prctl_mm_map prctl_map
= {
2118 struct vm_area_struct
*vma
;
2121 if (arg5
|| (arg4
&& (opt
!= PR_SET_MM_AUXV
&&
2122 opt
!= PR_SET_MM_MAP
&&
2123 opt
!= PR_SET_MM_MAP_SIZE
)))
2126 #ifdef CONFIG_CHECKPOINT_RESTORE
2127 if (opt
== PR_SET_MM_MAP
|| opt
== PR_SET_MM_MAP_SIZE
)
2128 return prctl_set_mm_map(opt
, (const void __user
*)addr
, arg4
);
2131 if (!capable(CAP_SYS_RESOURCE
))
2134 if (opt
== PR_SET_MM_EXE_FILE
)
2135 return prctl_set_mm_exe_file(mm
, (unsigned int)addr
);
2137 if (opt
== PR_SET_MM_AUXV
)
2138 return prctl_set_auxv(mm
, addr
, arg4
);
2140 if (addr
>= TASK_SIZE
|| addr
< mmap_min_addr
)
2146 * arg_lock protects concurrent updates of arg boundaries, we need
2147 * mmap_lock for a) concurrent sys_brk, b) finding VMA for addr
2151 vma
= find_vma(mm
, addr
);
2153 spin_lock(&mm
->arg_lock
);
2154 prctl_map
.start_code
= mm
->start_code
;
2155 prctl_map
.end_code
= mm
->end_code
;
2156 prctl_map
.start_data
= mm
->start_data
;
2157 prctl_map
.end_data
= mm
->end_data
;
2158 prctl_map
.start_brk
= mm
->start_brk
;
2159 prctl_map
.brk
= mm
->brk
;
2160 prctl_map
.start_stack
= mm
->start_stack
;
2161 prctl_map
.arg_start
= mm
->arg_start
;
2162 prctl_map
.arg_end
= mm
->arg_end
;
2163 prctl_map
.env_start
= mm
->env_start
;
2164 prctl_map
.env_end
= mm
->env_end
;
2167 case PR_SET_MM_START_CODE
:
2168 prctl_map
.start_code
= addr
;
2170 case PR_SET_MM_END_CODE
:
2171 prctl_map
.end_code
= addr
;
2173 case PR_SET_MM_START_DATA
:
2174 prctl_map
.start_data
= addr
;
2176 case PR_SET_MM_END_DATA
:
2177 prctl_map
.end_data
= addr
;
2179 case PR_SET_MM_START_STACK
:
2180 prctl_map
.start_stack
= addr
;
2182 case PR_SET_MM_START_BRK
:
2183 prctl_map
.start_brk
= addr
;
2186 prctl_map
.brk
= addr
;
2188 case PR_SET_MM_ARG_START
:
2189 prctl_map
.arg_start
= addr
;
2191 case PR_SET_MM_ARG_END
:
2192 prctl_map
.arg_end
= addr
;
2194 case PR_SET_MM_ENV_START
:
2195 prctl_map
.env_start
= addr
;
2197 case PR_SET_MM_ENV_END
:
2198 prctl_map
.env_end
= addr
;
2204 error
= validate_prctl_map_addr(&prctl_map
);
2210 * If command line arguments and environment
2211 * are placed somewhere else on stack, we can
2212 * set them up here, ARG_START/END to setup
2213 * command line arguments and ENV_START/END
2216 case PR_SET_MM_START_STACK
:
2217 case PR_SET_MM_ARG_START
:
2218 case PR_SET_MM_ARG_END
:
2219 case PR_SET_MM_ENV_START
:
2220 case PR_SET_MM_ENV_END
:
2227 mm
->start_code
= prctl_map
.start_code
;
2228 mm
->end_code
= prctl_map
.end_code
;
2229 mm
->start_data
= prctl_map
.start_data
;
2230 mm
->end_data
= prctl_map
.end_data
;
2231 mm
->start_brk
= prctl_map
.start_brk
;
2232 mm
->brk
= prctl_map
.brk
;
2233 mm
->start_stack
= prctl_map
.start_stack
;
2234 mm
->arg_start
= prctl_map
.arg_start
;
2235 mm
->arg_end
= prctl_map
.arg_end
;
2236 mm
->env_start
= prctl_map
.env_start
;
2237 mm
->env_end
= prctl_map
.env_end
;
2241 spin_unlock(&mm
->arg_lock
);
2242 mmap_read_unlock(mm
);
2246 #ifdef CONFIG_CHECKPOINT_RESTORE
2247 static int prctl_get_tid_address(struct task_struct
*me
, int __user
* __user
*tid_addr
)
2249 return put_user(me
->clear_child_tid
, tid_addr
);
2252 static int prctl_get_tid_address(struct task_struct
*me
, int __user
* __user
*tid_addr
)
2258 static int propagate_has_child_subreaper(struct task_struct
*p
, void *data
)
2261 * If task has has_child_subreaper - all its descendants
2262 * already have these flag too and new descendants will
2263 * inherit it on fork, skip them.
2265 * If we've found child_reaper - skip descendants in
2266 * it's subtree as they will never get out pidns.
2268 if (p
->signal
->has_child_subreaper
||
2269 is_child_reaper(task_pid(p
)))
2272 p
->signal
->has_child_subreaper
= 1;
2276 int __weak
arch_prctl_spec_ctrl_get(struct task_struct
*t
, unsigned long which
)
2281 int __weak
arch_prctl_spec_ctrl_set(struct task_struct
*t
, unsigned long which
,
2287 #define PR_IO_FLUSHER (PF_MEMALLOC_NOIO | PF_LOCAL_THROTTLE)
2289 SYSCALL_DEFINE5(prctl
, int, option
, unsigned long, arg2
, unsigned long, arg3
,
2290 unsigned long, arg4
, unsigned long, arg5
)
2292 struct task_struct
*me
= current
;
2293 unsigned char comm
[sizeof(me
->comm
)];
2296 error
= security_task_prctl(option
, arg2
, arg3
, arg4
, arg5
);
2297 if (error
!= -ENOSYS
)
2302 case PR_SET_PDEATHSIG
:
2303 if (!valid_signal(arg2
)) {
2307 me
->pdeath_signal
= arg2
;
2309 case PR_GET_PDEATHSIG
:
2310 error
= put_user(me
->pdeath_signal
, (int __user
*)arg2
);
2312 case PR_GET_DUMPABLE
:
2313 error
= get_dumpable(me
->mm
);
2315 case PR_SET_DUMPABLE
:
2316 if (arg2
!= SUID_DUMP_DISABLE
&& arg2
!= SUID_DUMP_USER
) {
2320 set_dumpable(me
->mm
, arg2
);
2323 case PR_SET_UNALIGN
:
2324 error
= SET_UNALIGN_CTL(me
, arg2
);
2326 case PR_GET_UNALIGN
:
2327 error
= GET_UNALIGN_CTL(me
, arg2
);
2330 error
= SET_FPEMU_CTL(me
, arg2
);
2333 error
= GET_FPEMU_CTL(me
, arg2
);
2336 error
= SET_FPEXC_CTL(me
, arg2
);
2339 error
= GET_FPEXC_CTL(me
, arg2
);
2342 error
= PR_TIMING_STATISTICAL
;
2345 if (arg2
!= PR_TIMING_STATISTICAL
)
2349 comm
[sizeof(me
->comm
) - 1] = 0;
2350 if (strncpy_from_user(comm
, (char __user
*)arg2
,
2351 sizeof(me
->comm
) - 1) < 0)
2353 set_task_comm(me
, comm
);
2354 proc_comm_connector(me
);
2357 get_task_comm(comm
, me
);
2358 if (copy_to_user((char __user
*)arg2
, comm
, sizeof(comm
)))
2362 error
= GET_ENDIAN(me
, arg2
);
2365 error
= SET_ENDIAN(me
, arg2
);
2367 case PR_GET_SECCOMP
:
2368 error
= prctl_get_seccomp();
2370 case PR_SET_SECCOMP
:
2371 error
= prctl_set_seccomp(arg2
, (char __user
*)arg3
);
2374 error
= GET_TSC_CTL(arg2
);
2377 error
= SET_TSC_CTL(arg2
);
2379 case PR_TASK_PERF_EVENTS_DISABLE
:
2380 error
= perf_event_task_disable();
2382 case PR_TASK_PERF_EVENTS_ENABLE
:
2383 error
= perf_event_task_enable();
2385 case PR_GET_TIMERSLACK
:
2386 if (current
->timer_slack_ns
> ULONG_MAX
)
2389 error
= current
->timer_slack_ns
;
2391 case PR_SET_TIMERSLACK
:
2393 current
->timer_slack_ns
=
2394 current
->default_timer_slack_ns
;
2396 current
->timer_slack_ns
= arg2
;
2402 case PR_MCE_KILL_CLEAR
:
2405 current
->flags
&= ~PF_MCE_PROCESS
;
2407 case PR_MCE_KILL_SET
:
2408 current
->flags
|= PF_MCE_PROCESS
;
2409 if (arg3
== PR_MCE_KILL_EARLY
)
2410 current
->flags
|= PF_MCE_EARLY
;
2411 else if (arg3
== PR_MCE_KILL_LATE
)
2412 current
->flags
&= ~PF_MCE_EARLY
;
2413 else if (arg3
== PR_MCE_KILL_DEFAULT
)
2415 ~(PF_MCE_EARLY
|PF_MCE_PROCESS
);
2423 case PR_MCE_KILL_GET
:
2424 if (arg2
| arg3
| arg4
| arg5
)
2426 if (current
->flags
& PF_MCE_PROCESS
)
2427 error
= (current
->flags
& PF_MCE_EARLY
) ?
2428 PR_MCE_KILL_EARLY
: PR_MCE_KILL_LATE
;
2430 error
= PR_MCE_KILL_DEFAULT
;
2433 error
= prctl_set_mm(arg2
, arg3
, arg4
, arg5
);
2435 case PR_GET_TID_ADDRESS
:
2436 error
= prctl_get_tid_address(me
, (int __user
* __user
*)arg2
);
2438 case PR_SET_CHILD_SUBREAPER
:
2439 me
->signal
->is_child_subreaper
= !!arg2
;
2443 walk_process_tree(me
, propagate_has_child_subreaper
, NULL
);
2445 case PR_GET_CHILD_SUBREAPER
:
2446 error
= put_user(me
->signal
->is_child_subreaper
,
2447 (int __user
*)arg2
);
2449 case PR_SET_NO_NEW_PRIVS
:
2450 if (arg2
!= 1 || arg3
|| arg4
|| arg5
)
2453 task_set_no_new_privs(current
);
2455 case PR_GET_NO_NEW_PRIVS
:
2456 if (arg2
|| arg3
|| arg4
|| arg5
)
2458 return task_no_new_privs(current
) ? 1 : 0;
2459 case PR_GET_THP_DISABLE
:
2460 if (arg2
|| arg3
|| arg4
|| arg5
)
2462 error
= !!test_bit(MMF_DISABLE_THP
, &me
->mm
->flags
);
2464 case PR_SET_THP_DISABLE
:
2465 if (arg3
|| arg4
|| arg5
)
2467 if (mmap_write_lock_killable(me
->mm
))
2470 set_bit(MMF_DISABLE_THP
, &me
->mm
->flags
);
2472 clear_bit(MMF_DISABLE_THP
, &me
->mm
->flags
);
2473 mmap_write_unlock(me
->mm
);
2475 case PR_MPX_ENABLE_MANAGEMENT
:
2476 case PR_MPX_DISABLE_MANAGEMENT
:
2477 /* No longer implemented: */
2479 case PR_SET_FP_MODE
:
2480 error
= SET_FP_MODE(me
, arg2
);
2482 case PR_GET_FP_MODE
:
2483 error
= GET_FP_MODE(me
);
2486 error
= SVE_SET_VL(arg2
);
2489 error
= SVE_GET_VL();
2491 case PR_GET_SPECULATION_CTRL
:
2492 if (arg3
|| arg4
|| arg5
)
2494 error
= arch_prctl_spec_ctrl_get(me
, arg2
);
2496 case PR_SET_SPECULATION_CTRL
:
2499 error
= arch_prctl_spec_ctrl_set(me
, arg2
, arg3
);
2501 case PR_PAC_RESET_KEYS
:
2502 if (arg3
|| arg4
|| arg5
)
2504 error
= PAC_RESET_KEYS(me
, arg2
);
2506 case PR_PAC_SET_ENABLED_KEYS
:
2509 error
= PAC_SET_ENABLED_KEYS(me
, arg2
, arg3
);
2511 case PR_PAC_GET_ENABLED_KEYS
:
2512 if (arg2
|| arg3
|| arg4
|| arg5
)
2514 error
= PAC_GET_ENABLED_KEYS(me
);
2516 case PR_SET_TAGGED_ADDR_CTRL
:
2517 if (arg3
|| arg4
|| arg5
)
2519 error
= SET_TAGGED_ADDR_CTRL(arg2
);
2521 case PR_GET_TAGGED_ADDR_CTRL
:
2522 if (arg2
|| arg3
|| arg4
|| arg5
)
2524 error
= GET_TAGGED_ADDR_CTRL();
2526 case PR_SET_IO_FLUSHER
:
2527 if (!capable(CAP_SYS_RESOURCE
))
2530 if (arg3
|| arg4
|| arg5
)
2534 current
->flags
|= PR_IO_FLUSHER
;
2536 current
->flags
&= ~PR_IO_FLUSHER
;
2540 case PR_GET_IO_FLUSHER
:
2541 if (!capable(CAP_SYS_RESOURCE
))
2544 if (arg2
|| arg3
|| arg4
|| arg5
)
2547 error
= (current
->flags
& PR_IO_FLUSHER
) == PR_IO_FLUSHER
;
2549 case PR_SET_SYSCALL_USER_DISPATCH
:
2550 error
= set_syscall_user_dispatch(arg2
, arg3
, arg4
,
2551 (char __user
*) arg5
);
2560 SYSCALL_DEFINE3(getcpu
, unsigned __user
*, cpup
, unsigned __user
*, nodep
,
2561 struct getcpu_cache __user
*, unused
)
2564 int cpu
= raw_smp_processor_id();
2567 err
|= put_user(cpu
, cpup
);
2569 err
|= put_user(cpu_to_node(cpu
), nodep
);
2570 return err
? -EFAULT
: 0;
2574 * do_sysinfo - fill in sysinfo struct
2575 * @info: pointer to buffer to fill
2577 static int do_sysinfo(struct sysinfo
*info
)
2579 unsigned long mem_total
, sav_total
;
2580 unsigned int mem_unit
, bitcount
;
2581 struct timespec64 tp
;
2583 memset(info
, 0, sizeof(struct sysinfo
));
2585 ktime_get_boottime_ts64(&tp
);
2586 timens_add_boottime(&tp
);
2587 info
->uptime
= tp
.tv_sec
+ (tp
.tv_nsec
? 1 : 0);
2589 get_avenrun(info
->loads
, 0, SI_LOAD_SHIFT
- FSHIFT
);
2591 info
->procs
= nr_threads
;
2597 * If the sum of all the available memory (i.e. ram + swap)
2598 * is less than can be stored in a 32 bit unsigned long then
2599 * we can be binary compatible with 2.2.x kernels. If not,
2600 * well, in that case 2.2.x was broken anyways...
2602 * -Erik Andersen <andersee@debian.org>
2605 mem_total
= info
->totalram
+ info
->totalswap
;
2606 if (mem_total
< info
->totalram
|| mem_total
< info
->totalswap
)
2609 mem_unit
= info
->mem_unit
;
2610 while (mem_unit
> 1) {
2613 sav_total
= mem_total
;
2615 if (mem_total
< sav_total
)
2620 * If mem_total did not overflow, multiply all memory values by
2621 * info->mem_unit and set it to 1. This leaves things compatible
2622 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2627 info
->totalram
<<= bitcount
;
2628 info
->freeram
<<= bitcount
;
2629 info
->sharedram
<<= bitcount
;
2630 info
->bufferram
<<= bitcount
;
2631 info
->totalswap
<<= bitcount
;
2632 info
->freeswap
<<= bitcount
;
2633 info
->totalhigh
<<= bitcount
;
2634 info
->freehigh
<<= bitcount
;
2640 SYSCALL_DEFINE1(sysinfo
, struct sysinfo __user
*, info
)
2646 if (copy_to_user(info
, &val
, sizeof(struct sysinfo
)))
2652 #ifdef CONFIG_COMPAT
2653 struct compat_sysinfo
{
2667 char _f
[20-2*sizeof(u32
)-sizeof(int)];
2670 COMPAT_SYSCALL_DEFINE1(sysinfo
, struct compat_sysinfo __user
*, info
)
2673 struct compat_sysinfo s_32
;
2677 /* Check to see if any memory value is too large for 32-bit and scale
2680 if (upper_32_bits(s
.totalram
) || upper_32_bits(s
.totalswap
)) {
2683 while (s
.mem_unit
< PAGE_SIZE
) {
2688 s
.totalram
>>= bitcount
;
2689 s
.freeram
>>= bitcount
;
2690 s
.sharedram
>>= bitcount
;
2691 s
.bufferram
>>= bitcount
;
2692 s
.totalswap
>>= bitcount
;
2693 s
.freeswap
>>= bitcount
;
2694 s
.totalhigh
>>= bitcount
;
2695 s
.freehigh
>>= bitcount
;
2698 memset(&s_32
, 0, sizeof(s_32
));
2699 s_32
.uptime
= s
.uptime
;
2700 s_32
.loads
[0] = s
.loads
[0];
2701 s_32
.loads
[1] = s
.loads
[1];
2702 s_32
.loads
[2] = s
.loads
[2];
2703 s_32
.totalram
= s
.totalram
;
2704 s_32
.freeram
= s
.freeram
;
2705 s_32
.sharedram
= s
.sharedram
;
2706 s_32
.bufferram
= s
.bufferram
;
2707 s_32
.totalswap
= s
.totalswap
;
2708 s_32
.freeswap
= s
.freeswap
;
2709 s_32
.procs
= s
.procs
;
2710 s_32
.totalhigh
= s
.totalhigh
;
2711 s_32
.freehigh
= s
.freehigh
;
2712 s_32
.mem_unit
= s
.mem_unit
;
2713 if (copy_to_user(info
, &s_32
, sizeof(s_32
)))
2717 #endif /* CONFIG_COMPAT */