1 // SPDX-License-Identifier: GPL-2.0
5 * Copyright (C) 1991, 1992 Linus Torvalds
8 #include <linux/export.h>
10 #include <linux/mm_inline.h>
11 #include <linux/utsname.h>
12 #include <linux/mman.h>
13 #include <linux/reboot.h>
14 #include <linux/prctl.h>
15 #include <linux/highuid.h>
17 #include <linux/kmod.h>
18 #include <linux/perf_event.h>
19 #include <linux/resource.h>
20 #include <linux/kernel.h>
21 #include <linux/workqueue.h>
22 #include <linux/capability.h>
23 #include <linux/device.h>
24 #include <linux/key.h>
25 #include <linux/times.h>
26 #include <linux/posix-timers.h>
27 #include <linux/security.h>
28 #include <linux/random.h>
29 #include <linux/suspend.h>
30 #include <linux/tty.h>
31 #include <linux/signal.h>
32 #include <linux/cn_proc.h>
33 #include <linux/getcpu.h>
34 #include <linux/task_io_accounting_ops.h>
35 #include <linux/seccomp.h>
36 #include <linux/cpu.h>
37 #include <linux/personality.h>
38 #include <linux/ptrace.h>
39 #include <linux/fs_struct.h>
40 #include <linux/file.h>
41 #include <linux/mount.h>
42 #include <linux/gfp.h>
43 #include <linux/syscore_ops.h>
44 #include <linux/version.h>
45 #include <linux/ctype.h>
46 #include <linux/syscall_user_dispatch.h>
48 #include <linux/compat.h>
49 #include <linux/syscalls.h>
50 #include <linux/kprobes.h>
51 #include <linux/user_namespace.h>
52 #include <linux/time_namespace.h>
53 #include <linux/binfmts.h>
55 #include <linux/sched.h>
56 #include <linux/sched/autogroup.h>
57 #include <linux/sched/loadavg.h>
58 #include <linux/sched/stat.h>
59 #include <linux/sched/mm.h>
60 #include <linux/sched/coredump.h>
61 #include <linux/sched/task.h>
62 #include <linux/sched/cputime.h>
63 #include <linux/rcupdate.h>
64 #include <linux/uidgid.h>
65 #include <linux/cred.h>
67 #include <linux/nospec.h>
69 #include <linux/kmsg_dump.h>
70 /* Move somewhere else to avoid recompiling? */
71 #include <generated/utsrelease.h>
73 #include <linux/uaccess.h>
75 #include <asm/unistd.h>
79 #ifndef SET_UNALIGN_CTL
80 # define SET_UNALIGN_CTL(a, b) (-EINVAL)
82 #ifndef GET_UNALIGN_CTL
83 # define GET_UNALIGN_CTL(a, b) (-EINVAL)
86 # define SET_FPEMU_CTL(a, b) (-EINVAL)
89 # define GET_FPEMU_CTL(a, b) (-EINVAL)
92 # define SET_FPEXC_CTL(a, b) (-EINVAL)
95 # define GET_FPEXC_CTL(a, b) (-EINVAL)
98 # define GET_ENDIAN(a, b) (-EINVAL)
101 # define SET_ENDIAN(a, b) (-EINVAL)
104 # define GET_TSC_CTL(a) (-EINVAL)
107 # define SET_TSC_CTL(a) (-EINVAL)
110 # define GET_FP_MODE(a) (-EINVAL)
113 # define SET_FP_MODE(a,b) (-EINVAL)
116 # define SVE_SET_VL(a) (-EINVAL)
119 # define SVE_GET_VL() (-EINVAL)
122 # define SME_SET_VL(a) (-EINVAL)
125 # define SME_GET_VL() (-EINVAL)
127 #ifndef PAC_RESET_KEYS
128 # define PAC_RESET_KEYS(a, b) (-EINVAL)
130 #ifndef PAC_SET_ENABLED_KEYS
131 # define PAC_SET_ENABLED_KEYS(a, b, c) (-EINVAL)
133 #ifndef PAC_GET_ENABLED_KEYS
134 # define PAC_GET_ENABLED_KEYS(a) (-EINVAL)
136 #ifndef SET_TAGGED_ADDR_CTRL
137 # define SET_TAGGED_ADDR_CTRL(a) (-EINVAL)
139 #ifndef GET_TAGGED_ADDR_CTRL
140 # define GET_TAGGED_ADDR_CTRL() (-EINVAL)
144 * this is where the system-wide overflow UID and GID are defined, for
145 * architectures that now have 32-bit UID/GID but didn't in the past
148 int overflowuid = DEFAULT_OVERFLOWUID;
149 int overflowgid = DEFAULT_OVERFLOWGID;
151 EXPORT_SYMBOL(overflowuid);
152 EXPORT_SYMBOL(overflowgid);
155 * the same as above, but for filesystems which can only store a 16-bit
156 * UID and GID. as such, this is needed on all architectures
159 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
160 int fs_overflowgid = DEFAULT_FS_OVERFLOWGID;
162 EXPORT_SYMBOL(fs_overflowuid);
163 EXPORT_SYMBOL(fs_overflowgid);
166 * Returns true if current's euid is same as p's uid or euid,
167 * or has CAP_SYS_NICE to p's user_ns.
169 * Called with rcu_read_lock, creds are safe
171 static bool set_one_prio_perm(struct task_struct *p)
173 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
175 if (uid_eq(pcred->uid, cred->euid) ||
176 uid_eq(pcred->euid, cred->euid))
178 if (ns_capable(pcred->user_ns, CAP_SYS_NICE))
184 * set the priority of a task
185 * - the caller must hold the RCU read lock
187 static int set_one_prio(struct task_struct *p, int niceval, int error)
191 if (!set_one_prio_perm(p)) {
195 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
199 no_nice = security_task_setnice(p, niceval);
206 set_user_nice(p, niceval);
211 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
213 struct task_struct *g, *p;
214 struct user_struct *user;
215 const struct cred *cred = current_cred();
220 if (which > PRIO_USER || which < PRIO_PROCESS)
223 /* normalize: avoid signed division (rounding problems) */
225 if (niceval < MIN_NICE)
227 if (niceval > MAX_NICE)
234 p = find_task_by_vpid(who);
238 error = set_one_prio(p, niceval, error);
242 pgrp = find_vpid(who);
244 pgrp = task_pgrp(current);
245 read_lock(&tasklist_lock);
246 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
247 error = set_one_prio(p, niceval, error);
248 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
249 read_unlock(&tasklist_lock);
252 uid = make_kuid(cred->user_ns, who);
256 else if (!uid_eq(uid, cred->uid)) {
257 user = find_user(uid);
259 goto out_unlock; /* No processes for this user */
261 for_each_process_thread(g, p) {
262 if (uid_eq(task_uid(p), uid) && task_pid_vnr(p))
263 error = set_one_prio(p, niceval, error);
265 if (!uid_eq(uid, cred->uid))
266 free_uid(user); /* For find_user() */
276 * Ugh. To avoid negative return values, "getpriority()" will
277 * not return the normal nice-value, but a negated value that
278 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
279 * to stay compatible.
281 SYSCALL_DEFINE2(getpriority, int, which, int, who)
283 struct task_struct *g, *p;
284 struct user_struct *user;
285 const struct cred *cred = current_cred();
286 long niceval, retval = -ESRCH;
290 if (which > PRIO_USER || which < PRIO_PROCESS)
297 p = find_task_by_vpid(who);
301 niceval = nice_to_rlimit(task_nice(p));
302 if (niceval > retval)
308 pgrp = find_vpid(who);
310 pgrp = task_pgrp(current);
311 read_lock(&tasklist_lock);
312 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
313 niceval = nice_to_rlimit(task_nice(p));
314 if (niceval > retval)
316 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
317 read_unlock(&tasklist_lock);
320 uid = make_kuid(cred->user_ns, who);
324 else if (!uid_eq(uid, cred->uid)) {
325 user = find_user(uid);
327 goto out_unlock; /* No processes for this user */
329 for_each_process_thread(g, p) {
330 if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) {
331 niceval = nice_to_rlimit(task_nice(p));
332 if (niceval > retval)
336 if (!uid_eq(uid, cred->uid))
337 free_uid(user); /* for find_user() */
347 * Unprivileged users may change the real gid to the effective gid
348 * or vice versa. (BSD-style)
350 * If you set the real gid at all, or set the effective gid to a value not
351 * equal to the real gid, then the saved gid is set to the new effective gid.
353 * This makes it possible for a setgid program to completely drop its
354 * privileges, which is often a useful assertion to make when you are doing
355 * a security audit over a program.
357 * The general idea is that a program which uses just setregid() will be
358 * 100% compatible with BSD. A program which uses just setgid() will be
359 * 100% compatible with POSIX with saved IDs.
361 * SMP: There are not races, the GIDs are checked only by filesystem
362 * operations (as far as semantic preservation is concerned).
364 #ifdef CONFIG_MULTIUSER
365 long __sys_setregid(gid_t rgid, gid_t egid)
367 struct user_namespace *ns = current_user_ns();
368 const struct cred *old;
373 krgid = make_kgid(ns, rgid);
374 kegid = make_kgid(ns, egid);
376 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
378 if ((egid != (gid_t) -1) && !gid_valid(kegid))
381 new = prepare_creds();
384 old = current_cred();
387 if (rgid != (gid_t) -1) {
388 if (gid_eq(old->gid, krgid) ||
389 gid_eq(old->egid, krgid) ||
390 ns_capable_setid(old->user_ns, CAP_SETGID))
395 if (egid != (gid_t) -1) {
396 if (gid_eq(old->gid, kegid) ||
397 gid_eq(old->egid, kegid) ||
398 gid_eq(old->sgid, kegid) ||
399 ns_capable_setid(old->user_ns, CAP_SETGID))
405 if (rgid != (gid_t) -1 ||
406 (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
407 new->sgid = new->egid;
408 new->fsgid = new->egid;
410 retval = security_task_fix_setgid(new, old, LSM_SETID_RE);
414 return commit_creds(new);
421 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
423 return __sys_setregid(rgid, egid);
427 * setgid() is implemented like SysV w/ SAVED_IDS
429 * SMP: Same implicit races as above.
431 long __sys_setgid(gid_t gid)
433 struct user_namespace *ns = current_user_ns();
434 const struct cred *old;
439 kgid = make_kgid(ns, gid);
440 if (!gid_valid(kgid))
443 new = prepare_creds();
446 old = current_cred();
449 if (ns_capable_setid(old->user_ns, CAP_SETGID))
450 new->gid = new->egid = new->sgid = new->fsgid = kgid;
451 else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
452 new->egid = new->fsgid = kgid;
456 retval = security_task_fix_setgid(new, old, LSM_SETID_ID);
460 return commit_creds(new);
467 SYSCALL_DEFINE1(setgid, gid_t, gid)
469 return __sys_setgid(gid);
473 * change the user struct in a credentials set to match the new UID
475 static int set_user(struct cred *new)
477 struct user_struct *new_user;
479 new_user = alloc_uid(new->uid);
484 new->user = new_user;
488 static void flag_nproc_exceeded(struct cred *new)
490 if (new->ucounts == current_ucounts())
494 * We don't fail in case of NPROC limit excess here because too many
495 * poorly written programs don't check set*uid() return code, assuming
496 * it never fails if called by root. We may still enforce NPROC limit
497 * for programs doing set*uid()+execve() by harmlessly deferring the
498 * failure to the execve() stage.
500 if (is_rlimit_overlimit(new->ucounts, UCOUNT_RLIMIT_NPROC, rlimit(RLIMIT_NPROC)) &&
501 new->user != INIT_USER)
502 current->flags |= PF_NPROC_EXCEEDED;
504 current->flags &= ~PF_NPROC_EXCEEDED;
508 * Unprivileged users may change the real uid to the effective uid
509 * or vice versa. (BSD-style)
511 * If you set the real uid at all, or set the effective uid to a value not
512 * equal to the real uid, then the saved uid is set to the new effective uid.
514 * This makes it possible for a setuid program to completely drop its
515 * privileges, which is often a useful assertion to make when you are doing
516 * a security audit over a program.
518 * The general idea is that a program which uses just setreuid() will be
519 * 100% compatible with BSD. A program which uses just setuid() will be
520 * 100% compatible with POSIX with saved IDs.
522 long __sys_setreuid(uid_t ruid, uid_t euid)
524 struct user_namespace *ns = current_user_ns();
525 const struct cred *old;
530 kruid = make_kuid(ns, ruid);
531 keuid = make_kuid(ns, euid);
533 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
535 if ((euid != (uid_t) -1) && !uid_valid(keuid))
538 new = prepare_creds();
541 old = current_cred();
544 if (ruid != (uid_t) -1) {
546 if (!uid_eq(old->uid, kruid) &&
547 !uid_eq(old->euid, kruid) &&
548 !ns_capable_setid(old->user_ns, CAP_SETUID))
552 if (euid != (uid_t) -1) {
554 if (!uid_eq(old->uid, keuid) &&
555 !uid_eq(old->euid, keuid) &&
556 !uid_eq(old->suid, keuid) &&
557 !ns_capable_setid(old->user_ns, CAP_SETUID))
561 if (!uid_eq(new->uid, old->uid)) {
562 retval = set_user(new);
566 if (ruid != (uid_t) -1 ||
567 (euid != (uid_t) -1 && !uid_eq(keuid, old->uid)))
568 new->suid = new->euid;
569 new->fsuid = new->euid;
571 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
575 retval = set_cred_ucounts(new);
579 flag_nproc_exceeded(new);
580 return commit_creds(new);
587 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
589 return __sys_setreuid(ruid, euid);
593 * setuid() is implemented like SysV with SAVED_IDS
595 * Note that SAVED_ID's is deficient in that a setuid root program
596 * like sendmail, for example, cannot set its uid to be a normal
597 * user and then switch back, because if you're root, setuid() sets
598 * the saved uid too. If you don't like this, blame the bright people
599 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
600 * will allow a root program to temporarily drop privileges and be able to
601 * regain them by swapping the real and effective uid.
603 long __sys_setuid(uid_t uid)
605 struct user_namespace *ns = current_user_ns();
606 const struct cred *old;
611 kuid = make_kuid(ns, uid);
612 if (!uid_valid(kuid))
615 new = prepare_creds();
618 old = current_cred();
621 if (ns_capable_setid(old->user_ns, CAP_SETUID)) {
622 new->suid = new->uid = kuid;
623 if (!uid_eq(kuid, old->uid)) {
624 retval = set_user(new);
628 } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) {
632 new->fsuid = new->euid = kuid;
634 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
638 retval = set_cred_ucounts(new);
642 flag_nproc_exceeded(new);
643 return commit_creds(new);
650 SYSCALL_DEFINE1(setuid, uid_t, uid)
652 return __sys_setuid(uid);
657 * This function implements a generic ability to update ruid, euid,
658 * and suid. This allows you to implement the 4.4 compatible seteuid().
660 long __sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
662 struct user_namespace *ns = current_user_ns();
663 const struct cred *old;
666 kuid_t kruid, keuid, ksuid;
667 bool ruid_new, euid_new, suid_new;
669 kruid = make_kuid(ns, ruid);
670 keuid = make_kuid(ns, euid);
671 ksuid = make_kuid(ns, suid);
673 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
676 if ((euid != (uid_t) -1) && !uid_valid(keuid))
679 if ((suid != (uid_t) -1) && !uid_valid(ksuid))
682 old = current_cred();
684 /* check for no-op */
685 if ((ruid == (uid_t) -1 || uid_eq(kruid, old->uid)) &&
686 (euid == (uid_t) -1 || (uid_eq(keuid, old->euid) &&
687 uid_eq(keuid, old->fsuid))) &&
688 (suid == (uid_t) -1 || uid_eq(ksuid, old->suid)))
691 ruid_new = ruid != (uid_t) -1 && !uid_eq(kruid, old->uid) &&
692 !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid);
693 euid_new = euid != (uid_t) -1 && !uid_eq(keuid, old->uid) &&
694 !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid);
695 suid_new = suid != (uid_t) -1 && !uid_eq(ksuid, old->uid) &&
696 !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid);
697 if ((ruid_new || euid_new || suid_new) &&
698 !ns_capable_setid(old->user_ns, CAP_SETUID))
701 new = prepare_creds();
705 if (ruid != (uid_t) -1) {
707 if (!uid_eq(kruid, old->uid)) {
708 retval = set_user(new);
713 if (euid != (uid_t) -1)
715 if (suid != (uid_t) -1)
717 new->fsuid = new->euid;
719 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
723 retval = set_cred_ucounts(new);
727 flag_nproc_exceeded(new);
728 return commit_creds(new);
735 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
737 return __sys_setresuid(ruid, euid, suid);
740 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
742 const struct cred *cred = current_cred();
744 uid_t ruid, euid, suid;
746 ruid = from_kuid_munged(cred->user_ns, cred->uid);
747 euid = from_kuid_munged(cred->user_ns, cred->euid);
748 suid = from_kuid_munged(cred->user_ns, cred->suid);
750 retval = put_user(ruid, ruidp);
752 retval = put_user(euid, euidp);
754 return put_user(suid, suidp);
760 * Same as above, but for rgid, egid, sgid.
762 long __sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
764 struct user_namespace *ns = current_user_ns();
765 const struct cred *old;
768 kgid_t krgid, kegid, ksgid;
769 bool rgid_new, egid_new, sgid_new;
771 krgid = make_kgid(ns, rgid);
772 kegid = make_kgid(ns, egid);
773 ksgid = make_kgid(ns, sgid);
775 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
777 if ((egid != (gid_t) -1) && !gid_valid(kegid))
779 if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
782 old = current_cred();
784 /* check for no-op */
785 if ((rgid == (gid_t) -1 || gid_eq(krgid, old->gid)) &&
786 (egid == (gid_t) -1 || (gid_eq(kegid, old->egid) &&
787 gid_eq(kegid, old->fsgid))) &&
788 (sgid == (gid_t) -1 || gid_eq(ksgid, old->sgid)))
791 rgid_new = rgid != (gid_t) -1 && !gid_eq(krgid, old->gid) &&
792 !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid);
793 egid_new = egid != (gid_t) -1 && !gid_eq(kegid, old->gid) &&
794 !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid);
795 sgid_new = sgid != (gid_t) -1 && !gid_eq(ksgid, old->gid) &&
796 !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid);
797 if ((rgid_new || egid_new || sgid_new) &&
798 !ns_capable_setid(old->user_ns, CAP_SETGID))
801 new = prepare_creds();
805 if (rgid != (gid_t) -1)
807 if (egid != (gid_t) -1)
809 if (sgid != (gid_t) -1)
811 new->fsgid = new->egid;
813 retval = security_task_fix_setgid(new, old, LSM_SETID_RES);
817 return commit_creds(new);
824 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
826 return __sys_setresgid(rgid, egid, sgid);
829 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
831 const struct cred *cred = current_cred();
833 gid_t rgid, egid, sgid;
835 rgid = from_kgid_munged(cred->user_ns, cred->gid);
836 egid = from_kgid_munged(cred->user_ns, cred->egid);
837 sgid = from_kgid_munged(cred->user_ns, cred->sgid);
839 retval = put_user(rgid, rgidp);
841 retval = put_user(egid, egidp);
843 retval = put_user(sgid, sgidp);
851 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
852 * is used for "access()" and for the NFS daemon (letting nfsd stay at
853 * whatever uid it wants to). It normally shadows "euid", except when
854 * explicitly set by setfsuid() or for access..
856 long __sys_setfsuid(uid_t uid)
858 const struct cred *old;
863 old = current_cred();
864 old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
866 kuid = make_kuid(old->user_ns, uid);
867 if (!uid_valid(kuid))
870 new = prepare_creds();
874 if (uid_eq(kuid, old->uid) || uid_eq(kuid, old->euid) ||
875 uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
876 ns_capable_setid(old->user_ns, CAP_SETUID)) {
877 if (!uid_eq(kuid, old->fsuid)) {
879 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
892 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
894 return __sys_setfsuid(uid);
898 * Samma på svenska..
900 long __sys_setfsgid(gid_t gid)
902 const struct cred *old;
907 old = current_cred();
908 old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
910 kgid = make_kgid(old->user_ns, gid);
911 if (!gid_valid(kgid))
914 new = prepare_creds();
918 if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->egid) ||
919 gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
920 ns_capable_setid(old->user_ns, CAP_SETGID)) {
921 if (!gid_eq(kgid, old->fsgid)) {
923 if (security_task_fix_setgid(new,old,LSM_SETID_FS) == 0)
936 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
938 return __sys_setfsgid(gid);
940 #endif /* CONFIG_MULTIUSER */
943 * sys_getpid - return the thread group id of the current process
945 * Note, despite the name, this returns the tgid not the pid. The tgid and
946 * the pid are identical unless CLONE_THREAD was specified on clone() in
947 * which case the tgid is the same in all threads of the same group.
949 * This is SMP safe as current->tgid does not change.
951 SYSCALL_DEFINE0(getpid)
953 return task_tgid_vnr(current);
956 /* Thread ID - the internal kernel "pid" */
957 SYSCALL_DEFINE0(gettid)
959 return task_pid_vnr(current);
963 * Accessing ->real_parent is not SMP-safe, it could
964 * change from under us. However, we can use a stale
965 * value of ->real_parent under rcu_read_lock(), see
966 * release_task()->call_rcu(delayed_put_task_struct).
968 SYSCALL_DEFINE0(getppid)
973 pid = task_tgid_vnr(rcu_dereference(current->real_parent));
979 SYSCALL_DEFINE0(getuid)
981 /* Only we change this so SMP safe */
982 return from_kuid_munged(current_user_ns(), current_uid());
985 SYSCALL_DEFINE0(geteuid)
987 /* Only we change this so SMP safe */
988 return from_kuid_munged(current_user_ns(), current_euid());
991 SYSCALL_DEFINE0(getgid)
993 /* Only we change this so SMP safe */
994 return from_kgid_munged(current_user_ns(), current_gid());
997 SYSCALL_DEFINE0(getegid)
999 /* Only we change this so SMP safe */
1000 return from_kgid_munged(current_user_ns(), current_egid());
1003 static void do_sys_times(struct tms *tms)
1005 u64 tgutime, tgstime, cutime, cstime;
1007 thread_group_cputime_adjusted(current, &tgutime, &tgstime);
1008 cutime = current->signal->cutime;
1009 cstime = current->signal->cstime;
1010 tms->tms_utime = nsec_to_clock_t(tgutime);
1011 tms->tms_stime = nsec_to_clock_t(tgstime);
1012 tms->tms_cutime = nsec_to_clock_t(cutime);
1013 tms->tms_cstime = nsec_to_clock_t(cstime);
1016 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
1022 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1025 force_successful_syscall_return();
1026 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1029 #ifdef CONFIG_COMPAT
1030 static compat_clock_t clock_t_to_compat_clock_t(clock_t x)
1032 return compat_jiffies_to_clock_t(clock_t_to_jiffies(x));
1035 COMPAT_SYSCALL_DEFINE1(times, struct compat_tms __user *, tbuf)
1039 struct compat_tms tmp;
1042 /* Convert our struct tms to the compat version. */
1043 tmp.tms_utime = clock_t_to_compat_clock_t(tms.tms_utime);
1044 tmp.tms_stime = clock_t_to_compat_clock_t(tms.tms_stime);
1045 tmp.tms_cutime = clock_t_to_compat_clock_t(tms.tms_cutime);
1046 tmp.tms_cstime = clock_t_to_compat_clock_t(tms.tms_cstime);
1047 if (copy_to_user(tbuf, &tmp, sizeof(tmp)))
1050 force_successful_syscall_return();
1051 return compat_jiffies_to_clock_t(jiffies);
1056 * This needs some heavy checking ...
1057 * I just haven't the stomach for it. I also don't fully
1058 * understand sessions/pgrp etc. Let somebody who does explain it.
1060 * OK, I think I have the protection semantics right.... this is really
1061 * only important on a multi-user system anyway, to make sure one user
1062 * can't send a signal to a process owned by another. -TYT, 12/12/91
1064 * !PF_FORKNOEXEC check to conform completely to POSIX.
1066 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
1068 struct task_struct *p;
1069 struct task_struct *group_leader = current->group_leader;
1074 pid = task_pid_vnr(group_leader);
1081 /* From this point forward we keep holding onto the tasklist lock
1082 * so that our parent does not change from under us. -DaveM
1084 write_lock_irq(&tasklist_lock);
1087 p = find_task_by_vpid(pid);
1092 if (!thread_group_leader(p))
1095 if (same_thread_group(p->real_parent, group_leader)) {
1097 if (task_session(p) != task_session(group_leader))
1100 if (!(p->flags & PF_FORKNOEXEC))
1104 if (p != group_leader)
1109 if (p->signal->leader)
1114 struct task_struct *g;
1116 pgrp = find_vpid(pgid);
1117 g = pid_task(pgrp, PIDTYPE_PGID);
1118 if (!g || task_session(g) != task_session(group_leader))
1122 err = security_task_setpgid(p, pgid);
1126 if (task_pgrp(p) != pgrp)
1127 change_pid(p, PIDTYPE_PGID, pgrp);
1131 /* All paths lead to here, thus we are safe. -DaveM */
1132 write_unlock_irq(&tasklist_lock);
1137 static int do_getpgid(pid_t pid)
1139 struct task_struct *p;
1145 grp = task_pgrp(current);
1148 p = find_task_by_vpid(pid);
1155 retval = security_task_getpgid(p);
1159 retval = pid_vnr(grp);
1165 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1167 return do_getpgid(pid);
1170 #ifdef __ARCH_WANT_SYS_GETPGRP
1172 SYSCALL_DEFINE0(getpgrp)
1174 return do_getpgid(0);
1179 SYSCALL_DEFINE1(getsid, pid_t, pid)
1181 struct task_struct *p;
1187 sid = task_session(current);
1190 p = find_task_by_vpid(pid);
1193 sid = task_session(p);
1197 retval = security_task_getsid(p);
1201 retval = pid_vnr(sid);
1207 static void set_special_pids(struct pid *pid)
1209 struct task_struct *curr = current->group_leader;
1211 if (task_session(curr) != pid)
1212 change_pid(curr, PIDTYPE_SID, pid);
1214 if (task_pgrp(curr) != pid)
1215 change_pid(curr, PIDTYPE_PGID, pid);
1218 int ksys_setsid(void)
1220 struct task_struct *group_leader = current->group_leader;
1221 struct pid *sid = task_pid(group_leader);
1222 pid_t session = pid_vnr(sid);
1225 write_lock_irq(&tasklist_lock);
1226 /* Fail if I am already a session leader */
1227 if (group_leader->signal->leader)
1230 /* Fail if a process group id already exists that equals the
1231 * proposed session id.
1233 if (pid_task(sid, PIDTYPE_PGID))
1236 group_leader->signal->leader = 1;
1237 set_special_pids(sid);
1239 proc_clear_tty(group_leader);
1243 write_unlock_irq(&tasklist_lock);
1245 proc_sid_connector(group_leader);
1246 sched_autogroup_create_attach(group_leader);
1251 SYSCALL_DEFINE0(setsid)
1253 return ksys_setsid();
1256 DECLARE_RWSEM(uts_sem);
1258 #ifdef COMPAT_UTS_MACHINE
1259 #define override_architecture(name) \
1260 (personality(current->personality) == PER_LINUX32 && \
1261 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1262 sizeof(COMPAT_UTS_MACHINE)))
1264 #define override_architecture(name) 0
1268 * Work around broken programs that cannot handle "Linux 3.0".
1269 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1270 * And we map 4.x and later versions to 2.6.60+x, so 4.0/5.0/6.0/... would be
1273 static int override_release(char __user *release, size_t len)
1277 if (current->personality & UNAME26) {
1278 const char *rest = UTS_RELEASE;
1279 char buf[65] = { 0 };
1285 if (*rest == '.' && ++ndots >= 3)
1287 if (!isdigit(*rest) && *rest != '.')
1291 v = LINUX_VERSION_PATCHLEVEL + 60;
1292 copy = clamp_t(size_t, len, 1, sizeof(buf));
1293 copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1294 ret = copy_to_user(release, buf, copy + 1);
1299 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1301 struct new_utsname tmp;
1303 down_read(&uts_sem);
1304 memcpy(&tmp, utsname(), sizeof(tmp));
1306 if (copy_to_user(name, &tmp, sizeof(tmp)))
1309 if (override_release(name->release, sizeof(name->release)))
1311 if (override_architecture(name))
1316 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1320 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1322 struct old_utsname tmp;
1327 down_read(&uts_sem);
1328 memcpy(&tmp, utsname(), sizeof(tmp));
1330 if (copy_to_user(name, &tmp, sizeof(tmp)))
1333 if (override_release(name->release, sizeof(name->release)))
1335 if (override_architecture(name))
1340 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1342 struct oldold_utsname tmp;
1347 memset(&tmp, 0, sizeof(tmp));
1349 down_read(&uts_sem);
1350 memcpy(&tmp.sysname, &utsname()->sysname, __OLD_UTS_LEN);
1351 memcpy(&tmp.nodename, &utsname()->nodename, __OLD_UTS_LEN);
1352 memcpy(&tmp.release, &utsname()->release, __OLD_UTS_LEN);
1353 memcpy(&tmp.version, &utsname()->version, __OLD_UTS_LEN);
1354 memcpy(&tmp.machine, &utsname()->machine, __OLD_UTS_LEN);
1356 if (copy_to_user(name, &tmp, sizeof(tmp)))
1359 if (override_architecture(name))
1361 if (override_release(name->release, sizeof(name->release)))
1367 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1370 char tmp[__NEW_UTS_LEN];
1372 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1375 if (len < 0 || len > __NEW_UTS_LEN)
1378 if (!copy_from_user(tmp, name, len)) {
1379 struct new_utsname *u;
1381 add_device_randomness(tmp, len);
1382 down_write(&uts_sem);
1384 memcpy(u->nodename, tmp, len);
1385 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1387 uts_proc_notify(UTS_PROC_HOSTNAME);
1393 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1395 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1398 struct new_utsname *u;
1399 char tmp[__NEW_UTS_LEN + 1];
1403 down_read(&uts_sem);
1405 i = 1 + strlen(u->nodename);
1408 memcpy(tmp, u->nodename, i);
1410 if (copy_to_user(name, tmp, i))
1418 * Only setdomainname; getdomainname can be implemented by calling
1421 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1424 char tmp[__NEW_UTS_LEN];
1426 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1428 if (len < 0 || len > __NEW_UTS_LEN)
1432 if (!copy_from_user(tmp, name, len)) {
1433 struct new_utsname *u;
1435 add_device_randomness(tmp, len);
1436 down_write(&uts_sem);
1438 memcpy(u->domainname, tmp, len);
1439 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1441 uts_proc_notify(UTS_PROC_DOMAINNAME);
1447 /* make sure you are allowed to change @tsk limits before calling this */
1448 static int do_prlimit(struct task_struct *tsk, unsigned int resource,
1449 struct rlimit *new_rlim, struct rlimit *old_rlim)
1451 struct rlimit *rlim;
1454 if (resource >= RLIM_NLIMITS)
1456 resource = array_index_nospec(resource, RLIM_NLIMITS);
1459 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1461 if (resource == RLIMIT_NOFILE &&
1462 new_rlim->rlim_max > sysctl_nr_open)
1466 /* Holding a refcount on tsk protects tsk->signal from disappearing. */
1467 rlim = tsk->signal->rlim + resource;
1468 task_lock(tsk->group_leader);
1471 * Keep the capable check against init_user_ns until cgroups can
1472 * contain all limits.
1474 if (new_rlim->rlim_max > rlim->rlim_max &&
1475 !capable(CAP_SYS_RESOURCE))
1478 retval = security_task_setrlimit(tsk, resource, new_rlim);
1486 task_unlock(tsk->group_leader);
1489 * RLIMIT_CPU handling. Arm the posix CPU timer if the limit is not
1490 * infinite. In case of RLIM_INFINITY the posix CPU timer code
1491 * ignores the rlimit.
1493 if (!retval && new_rlim && resource == RLIMIT_CPU &&
1494 new_rlim->rlim_cur != RLIM_INFINITY &&
1495 IS_ENABLED(CONFIG_POSIX_TIMERS)) {
1497 * update_rlimit_cpu can fail if the task is exiting, but there
1498 * may be other tasks in the thread group that are not exiting,
1499 * and they need their cpu timers adjusted.
1501 * The group_leader is the last task to be released, so if we
1502 * cannot update_rlimit_cpu on it, then the entire process is
1503 * exiting and we do not need to update at all.
1505 update_rlimit_cpu(tsk->group_leader, new_rlim->rlim_cur);
1511 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1513 struct rlimit value;
1516 ret = do_prlimit(current, resource, NULL, &value);
1518 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1523 #ifdef CONFIG_COMPAT
1525 COMPAT_SYSCALL_DEFINE2(setrlimit, unsigned int, resource,
1526 struct compat_rlimit __user *, rlim)
1529 struct compat_rlimit r32;
1531 if (copy_from_user(&r32, rlim, sizeof(struct compat_rlimit)))
1534 if (r32.rlim_cur == COMPAT_RLIM_INFINITY)
1535 r.rlim_cur = RLIM_INFINITY;
1537 r.rlim_cur = r32.rlim_cur;
1538 if (r32.rlim_max == COMPAT_RLIM_INFINITY)
1539 r.rlim_max = RLIM_INFINITY;
1541 r.rlim_max = r32.rlim_max;
1542 return do_prlimit(current, resource, &r, NULL);
1545 COMPAT_SYSCALL_DEFINE2(getrlimit, unsigned int, resource,
1546 struct compat_rlimit __user *, rlim)
1551 ret = do_prlimit(current, resource, NULL, &r);
1553 struct compat_rlimit r32;
1554 if (r.rlim_cur > COMPAT_RLIM_INFINITY)
1555 r32.rlim_cur = COMPAT_RLIM_INFINITY;
1557 r32.rlim_cur = r.rlim_cur;
1558 if (r.rlim_max > COMPAT_RLIM_INFINITY)
1559 r32.rlim_max = COMPAT_RLIM_INFINITY;
1561 r32.rlim_max = r.rlim_max;
1563 if (copy_to_user(rlim, &r32, sizeof(struct compat_rlimit)))
1571 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1574 * Back compatibility for getrlimit. Needed for some apps.
1576 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1577 struct rlimit __user *, rlim)
1580 if (resource >= RLIM_NLIMITS)
1583 resource = array_index_nospec(resource, RLIM_NLIMITS);
1584 task_lock(current->group_leader);
1585 x = current->signal->rlim[resource];
1586 task_unlock(current->group_leader);
1587 if (x.rlim_cur > 0x7FFFFFFF)
1588 x.rlim_cur = 0x7FFFFFFF;
1589 if (x.rlim_max > 0x7FFFFFFF)
1590 x.rlim_max = 0x7FFFFFFF;
1591 return copy_to_user(rlim, &x, sizeof(x)) ? -EFAULT : 0;
1594 #ifdef CONFIG_COMPAT
1595 COMPAT_SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1596 struct compat_rlimit __user *, rlim)
1600 if (resource >= RLIM_NLIMITS)
1603 resource = array_index_nospec(resource, RLIM_NLIMITS);
1604 task_lock(current->group_leader);
1605 r = current->signal->rlim[resource];
1606 task_unlock(current->group_leader);
1607 if (r.rlim_cur > 0x7FFFFFFF)
1608 r.rlim_cur = 0x7FFFFFFF;
1609 if (r.rlim_max > 0x7FFFFFFF)
1610 r.rlim_max = 0x7FFFFFFF;
1612 if (put_user(r.rlim_cur, &rlim->rlim_cur) ||
1613 put_user(r.rlim_max, &rlim->rlim_max))
1621 static inline bool rlim64_is_infinity(__u64 rlim64)
1623 #if BITS_PER_LONG < 64
1624 return rlim64 >= ULONG_MAX;
1626 return rlim64 == RLIM64_INFINITY;
1630 static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1632 if (rlim->rlim_cur == RLIM_INFINITY)
1633 rlim64->rlim_cur = RLIM64_INFINITY;
1635 rlim64->rlim_cur = rlim->rlim_cur;
1636 if (rlim->rlim_max == RLIM_INFINITY)
1637 rlim64->rlim_max = RLIM64_INFINITY;
1639 rlim64->rlim_max = rlim->rlim_max;
1642 static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1644 if (rlim64_is_infinity(rlim64->rlim_cur))
1645 rlim->rlim_cur = RLIM_INFINITY;
1647 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1648 if (rlim64_is_infinity(rlim64->rlim_max))
1649 rlim->rlim_max = RLIM_INFINITY;
1651 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1654 /* rcu lock must be held */
1655 static int check_prlimit_permission(struct task_struct *task,
1658 const struct cred *cred = current_cred(), *tcred;
1661 if (current == task)
1664 tcred = __task_cred(task);
1665 id_match = (uid_eq(cred->uid, tcred->euid) &&
1666 uid_eq(cred->uid, tcred->suid) &&
1667 uid_eq(cred->uid, tcred->uid) &&
1668 gid_eq(cred->gid, tcred->egid) &&
1669 gid_eq(cred->gid, tcred->sgid) &&
1670 gid_eq(cred->gid, tcred->gid));
1671 if (!id_match && !ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
1674 return security_task_prlimit(cred, tcred, flags);
1677 SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1678 const struct rlimit64 __user *, new_rlim,
1679 struct rlimit64 __user *, old_rlim)
1681 struct rlimit64 old64, new64;
1682 struct rlimit old, new;
1683 struct task_struct *tsk;
1684 unsigned int checkflags = 0;
1688 checkflags |= LSM_PRLIMIT_READ;
1691 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1693 rlim64_to_rlim(&new64, &new);
1694 checkflags |= LSM_PRLIMIT_WRITE;
1698 tsk = pid ? find_task_by_vpid(pid) : current;
1703 ret = check_prlimit_permission(tsk, checkflags);
1708 get_task_struct(tsk);
1711 ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1712 old_rlim ? &old : NULL);
1714 if (!ret && old_rlim) {
1715 rlim_to_rlim64(&old, &old64);
1716 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1720 put_task_struct(tsk);
1724 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1726 struct rlimit new_rlim;
1728 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1730 return do_prlimit(current, resource, &new_rlim, NULL);
1734 * It would make sense to put struct rusage in the task_struct,
1735 * except that would make the task_struct be *really big*. After
1736 * task_struct gets moved into malloc'ed memory, it would
1737 * make sense to do this. It will make moving the rest of the information
1738 * a lot simpler! (Which we're not doing right now because we're not
1739 * measuring them yet).
1741 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1742 * races with threads incrementing their own counters. But since word
1743 * reads are atomic, we either get new values or old values and we don't
1744 * care which for the sums. We always take the siglock to protect reading
1745 * the c* fields from p->signal from races with exit.c updating those
1746 * fields when reaping, so a sample either gets all the additions of a
1747 * given child after it's reaped, or none so this sample is before reaping.
1750 * We need to take the siglock for CHILDEREN, SELF and BOTH
1751 * for the cases current multithreaded, non-current single threaded
1752 * non-current multithreaded. Thread traversal is now safe with
1754 * Strictly speaking, we donot need to take the siglock if we are current and
1755 * single threaded, as no one else can take our signal_struct away, no one
1756 * else can reap the children to update signal->c* counters, and no one else
1757 * can race with the signal-> fields. If we do not take any lock, the
1758 * signal-> fields could be read out of order while another thread was just
1759 * exiting. So we should place a read memory barrier when we avoid the lock.
1760 * On the writer side, write memory barrier is implied in __exit_signal
1761 * as __exit_signal releases the siglock spinlock after updating the signal->
1762 * fields. But we don't do this yet to keep things simple.
1766 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1768 r->ru_nvcsw += t->nvcsw;
1769 r->ru_nivcsw += t->nivcsw;
1770 r->ru_minflt += t->min_flt;
1771 r->ru_majflt += t->maj_flt;
1772 r->ru_inblock += task_io_get_inblock(t);
1773 r->ru_oublock += task_io_get_oublock(t);
1776 void getrusage(struct task_struct *p, int who, struct rusage *r)
1778 struct task_struct *t;
1779 unsigned long flags;
1780 u64 tgutime, tgstime, utime, stime;
1781 unsigned long maxrss = 0;
1783 memset((char *)r, 0, sizeof (*r));
1786 if (who == RUSAGE_THREAD) {
1787 task_cputime_adjusted(current, &utime, &stime);
1788 accumulate_thread_rusage(p, r);
1789 maxrss = p->signal->maxrss;
1793 if (!lock_task_sighand(p, &flags))
1798 case RUSAGE_CHILDREN:
1799 utime = p->signal->cutime;
1800 stime = p->signal->cstime;
1801 r->ru_nvcsw = p->signal->cnvcsw;
1802 r->ru_nivcsw = p->signal->cnivcsw;
1803 r->ru_minflt = p->signal->cmin_flt;
1804 r->ru_majflt = p->signal->cmaj_flt;
1805 r->ru_inblock = p->signal->cinblock;
1806 r->ru_oublock = p->signal->coublock;
1807 maxrss = p->signal->cmaxrss;
1809 if (who == RUSAGE_CHILDREN)
1814 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1817 r->ru_nvcsw += p->signal->nvcsw;
1818 r->ru_nivcsw += p->signal->nivcsw;
1819 r->ru_minflt += p->signal->min_flt;
1820 r->ru_majflt += p->signal->maj_flt;
1821 r->ru_inblock += p->signal->inblock;
1822 r->ru_oublock += p->signal->oublock;
1823 if (maxrss < p->signal->maxrss)
1824 maxrss = p->signal->maxrss;
1827 accumulate_thread_rusage(t, r);
1828 } while_each_thread(p, t);
1834 unlock_task_sighand(p, &flags);
1837 r->ru_utime = ns_to_kernel_old_timeval(utime);
1838 r->ru_stime = ns_to_kernel_old_timeval(stime);
1840 if (who != RUSAGE_CHILDREN) {
1841 struct mm_struct *mm = get_task_mm(p);
1844 setmax_mm_hiwater_rss(&maxrss, mm);
1848 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1851 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1855 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1856 who != RUSAGE_THREAD)
1859 getrusage(current, who, &r);
1860 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1863 #ifdef CONFIG_COMPAT
1864 COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru)
1868 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1869 who != RUSAGE_THREAD)
1872 getrusage(current, who, &r);
1873 return put_compat_rusage(&r, ru);
1877 SYSCALL_DEFINE1(umask, int, mask)
1879 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1883 static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1886 struct inode *inode;
1893 inode = file_inode(exe.file);
1896 * Because the original mm->exe_file points to executable file, make
1897 * sure that this one is executable as well, to avoid breaking an
1901 if (!S_ISREG(inode->i_mode) || path_noexec(&exe.file->f_path))
1904 err = file_permission(exe.file, MAY_EXEC);
1908 err = replace_mm_exe_file(mm, exe.file);
1915 * Check arithmetic relations of passed addresses.
1917 * WARNING: we don't require any capability here so be very careful
1918 * in what is allowed for modification from userspace.
1920 static int validate_prctl_map_addr(struct prctl_mm_map *prctl_map)
1922 unsigned long mmap_max_addr = TASK_SIZE;
1923 int error = -EINVAL, i;
1925 static const unsigned char offsets[] = {
1926 offsetof(struct prctl_mm_map, start_code),
1927 offsetof(struct prctl_mm_map, end_code),
1928 offsetof(struct prctl_mm_map, start_data),
1929 offsetof(struct prctl_mm_map, end_data),
1930 offsetof(struct prctl_mm_map, start_brk),
1931 offsetof(struct prctl_mm_map, brk),
1932 offsetof(struct prctl_mm_map, start_stack),
1933 offsetof(struct prctl_mm_map, arg_start),
1934 offsetof(struct prctl_mm_map, arg_end),
1935 offsetof(struct prctl_mm_map, env_start),
1936 offsetof(struct prctl_mm_map, env_end),
1940 * Make sure the members are not somewhere outside
1941 * of allowed address space.
1943 for (i = 0; i < ARRAY_SIZE(offsets); i++) {
1944 u64 val = *(u64 *)((char *)prctl_map + offsets[i]);
1946 if ((unsigned long)val >= mmap_max_addr ||
1947 (unsigned long)val < mmap_min_addr)
1952 * Make sure the pairs are ordered.
1954 #define __prctl_check_order(__m1, __op, __m2) \
1955 ((unsigned long)prctl_map->__m1 __op \
1956 (unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1957 error = __prctl_check_order(start_code, <, end_code);
1958 error |= __prctl_check_order(start_data,<=, end_data);
1959 error |= __prctl_check_order(start_brk, <=, brk);
1960 error |= __prctl_check_order(arg_start, <=, arg_end);
1961 error |= __prctl_check_order(env_start, <=, env_end);
1964 #undef __prctl_check_order
1969 * Neither we should allow to override limits if they set.
1971 if (check_data_rlimit(rlimit(RLIMIT_DATA), prctl_map->brk,
1972 prctl_map->start_brk, prctl_map->end_data,
1973 prctl_map->start_data))
1981 #ifdef CONFIG_CHECKPOINT_RESTORE
1982 static int prctl_set_mm_map(int opt, const void __user *addr, unsigned long data_size)
1984 struct prctl_mm_map prctl_map = { .exe_fd = (u32)-1, };
1985 unsigned long user_auxv[AT_VECTOR_SIZE];
1986 struct mm_struct *mm = current->mm;
1989 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
1990 BUILD_BUG_ON(sizeof(struct prctl_mm_map) > 256);
1992 if (opt == PR_SET_MM_MAP_SIZE)
1993 return put_user((unsigned int)sizeof(prctl_map),
1994 (unsigned int __user *)addr);
1996 if (data_size != sizeof(prctl_map))
1999 if (copy_from_user(&prctl_map, addr, sizeof(prctl_map)))
2002 error = validate_prctl_map_addr(&prctl_map);
2006 if (prctl_map.auxv_size) {
2008 * Someone is trying to cheat the auxv vector.
2010 if (!prctl_map.auxv ||
2011 prctl_map.auxv_size > sizeof(mm->saved_auxv))
2014 memset(user_auxv, 0, sizeof(user_auxv));
2015 if (copy_from_user(user_auxv,
2016 (const void __user *)prctl_map.auxv,
2017 prctl_map.auxv_size))
2020 /* Last entry must be AT_NULL as specification requires */
2021 user_auxv[AT_VECTOR_SIZE - 2] = AT_NULL;
2022 user_auxv[AT_VECTOR_SIZE - 1] = AT_NULL;
2025 if (prctl_map.exe_fd != (u32)-1) {
2027 * Check if the current user is checkpoint/restore capable.
2028 * At the time of this writing, it checks for CAP_SYS_ADMIN
2029 * or CAP_CHECKPOINT_RESTORE.
2030 * Note that a user with access to ptrace can masquerade an
2031 * arbitrary program as any executable, even setuid ones.
2032 * This may have implications in the tomoyo subsystem.
2034 if (!checkpoint_restore_ns_capable(current_user_ns()))
2037 error = prctl_set_mm_exe_file(mm, prctl_map.exe_fd);
2043 * arg_lock protects concurrent updates but we still need mmap_lock for
2044 * read to exclude races with sys_brk.
2049 * We don't validate if these members are pointing to
2050 * real present VMAs because application may have correspond
2051 * VMAs already unmapped and kernel uses these members for statistics
2052 * output in procfs mostly, except
2054 * - @start_brk/@brk which are used in do_brk_flags but kernel lookups
2055 * for VMAs when updating these members so anything wrong written
2056 * here cause kernel to swear at userspace program but won't lead
2057 * to any problem in kernel itself
2060 spin_lock(&mm->arg_lock);
2061 mm->start_code = prctl_map.start_code;
2062 mm->end_code = prctl_map.end_code;
2063 mm->start_data = prctl_map.start_data;
2064 mm->end_data = prctl_map.end_data;
2065 mm->start_brk = prctl_map.start_brk;
2066 mm->brk = prctl_map.brk;
2067 mm->start_stack = prctl_map.start_stack;
2068 mm->arg_start = prctl_map.arg_start;
2069 mm->arg_end = prctl_map.arg_end;
2070 mm->env_start = prctl_map.env_start;
2071 mm->env_end = prctl_map.env_end;
2072 spin_unlock(&mm->arg_lock);
2075 * Note this update of @saved_auxv is lockless thus
2076 * if someone reads this member in procfs while we're
2077 * updating -- it may get partly updated results. It's
2078 * known and acceptable trade off: we leave it as is to
2079 * not introduce additional locks here making the kernel
2082 if (prctl_map.auxv_size)
2083 memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv));
2085 mmap_read_unlock(mm);
2088 #endif /* CONFIG_CHECKPOINT_RESTORE */
2090 static int prctl_set_auxv(struct mm_struct *mm, unsigned long addr,
2094 * This doesn't move the auxiliary vector itself since it's pinned to
2095 * mm_struct, but it permits filling the vector with new values. It's
2096 * up to the caller to provide sane values here, otherwise userspace
2097 * tools which use this vector might be unhappy.
2099 unsigned long user_auxv[AT_VECTOR_SIZE] = {};
2101 if (len > sizeof(user_auxv))
2104 if (copy_from_user(user_auxv, (const void __user *)addr, len))
2107 /* Make sure the last entry is always AT_NULL */
2108 user_auxv[AT_VECTOR_SIZE - 2] = 0;
2109 user_auxv[AT_VECTOR_SIZE - 1] = 0;
2111 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
2114 memcpy(mm->saved_auxv, user_auxv, len);
2115 task_unlock(current);
2120 static int prctl_set_mm(int opt, unsigned long addr,
2121 unsigned long arg4, unsigned long arg5)
2123 struct mm_struct *mm = current->mm;
2124 struct prctl_mm_map prctl_map = {
2129 struct vm_area_struct *vma;
2132 if (arg5 || (arg4 && (opt != PR_SET_MM_AUXV &&
2133 opt != PR_SET_MM_MAP &&
2134 opt != PR_SET_MM_MAP_SIZE)))
2137 #ifdef CONFIG_CHECKPOINT_RESTORE
2138 if (opt == PR_SET_MM_MAP || opt == PR_SET_MM_MAP_SIZE)
2139 return prctl_set_mm_map(opt, (const void __user *)addr, arg4);
2142 if (!capable(CAP_SYS_RESOURCE))
2145 if (opt == PR_SET_MM_EXE_FILE)
2146 return prctl_set_mm_exe_file(mm, (unsigned int)addr);
2148 if (opt == PR_SET_MM_AUXV)
2149 return prctl_set_auxv(mm, addr, arg4);
2151 if (addr >= TASK_SIZE || addr < mmap_min_addr)
2157 * arg_lock protects concurrent updates of arg boundaries, we need
2158 * mmap_lock for a) concurrent sys_brk, b) finding VMA for addr
2162 vma = find_vma(mm, addr);
2164 spin_lock(&mm->arg_lock);
2165 prctl_map.start_code = mm->start_code;
2166 prctl_map.end_code = mm->end_code;
2167 prctl_map.start_data = mm->start_data;
2168 prctl_map.end_data = mm->end_data;
2169 prctl_map.start_brk = mm->start_brk;
2170 prctl_map.brk = mm->brk;
2171 prctl_map.start_stack = mm->start_stack;
2172 prctl_map.arg_start = mm->arg_start;
2173 prctl_map.arg_end = mm->arg_end;
2174 prctl_map.env_start = mm->env_start;
2175 prctl_map.env_end = mm->env_end;
2178 case PR_SET_MM_START_CODE:
2179 prctl_map.start_code = addr;
2181 case PR_SET_MM_END_CODE:
2182 prctl_map.end_code = addr;
2184 case PR_SET_MM_START_DATA:
2185 prctl_map.start_data = addr;
2187 case PR_SET_MM_END_DATA:
2188 prctl_map.end_data = addr;
2190 case PR_SET_MM_START_STACK:
2191 prctl_map.start_stack = addr;
2193 case PR_SET_MM_START_BRK:
2194 prctl_map.start_brk = addr;
2197 prctl_map.brk = addr;
2199 case PR_SET_MM_ARG_START:
2200 prctl_map.arg_start = addr;
2202 case PR_SET_MM_ARG_END:
2203 prctl_map.arg_end = addr;
2205 case PR_SET_MM_ENV_START:
2206 prctl_map.env_start = addr;
2208 case PR_SET_MM_ENV_END:
2209 prctl_map.env_end = addr;
2215 error = validate_prctl_map_addr(&prctl_map);
2221 * If command line arguments and environment
2222 * are placed somewhere else on stack, we can
2223 * set them up here, ARG_START/END to setup
2224 * command line arguments and ENV_START/END
2227 case PR_SET_MM_START_STACK:
2228 case PR_SET_MM_ARG_START:
2229 case PR_SET_MM_ARG_END:
2230 case PR_SET_MM_ENV_START:
2231 case PR_SET_MM_ENV_END:
2238 mm->start_code = prctl_map.start_code;
2239 mm->end_code = prctl_map.end_code;
2240 mm->start_data = prctl_map.start_data;
2241 mm->end_data = prctl_map.end_data;
2242 mm->start_brk = prctl_map.start_brk;
2243 mm->brk = prctl_map.brk;
2244 mm->start_stack = prctl_map.start_stack;
2245 mm->arg_start = prctl_map.arg_start;
2246 mm->arg_end = prctl_map.arg_end;
2247 mm->env_start = prctl_map.env_start;
2248 mm->env_end = prctl_map.env_end;
2252 spin_unlock(&mm->arg_lock);
2253 mmap_read_unlock(mm);
2257 #ifdef CONFIG_CHECKPOINT_RESTORE
2258 static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr)
2260 return put_user(me->clear_child_tid, tid_addr);
2263 static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr)
2269 static int propagate_has_child_subreaper(struct task_struct *p, void *data)
2272 * If task has has_child_subreaper - all its descendants
2273 * already have these flag too and new descendants will
2274 * inherit it on fork, skip them.
2276 * If we've found child_reaper - skip descendants in
2277 * it's subtree as they will never get out pidns.
2279 if (p->signal->has_child_subreaper ||
2280 is_child_reaper(task_pid(p)))
2283 p->signal->has_child_subreaper = 1;
2287 int __weak arch_prctl_spec_ctrl_get(struct task_struct *t, unsigned long which)
2292 int __weak arch_prctl_spec_ctrl_set(struct task_struct *t, unsigned long which,
2298 #define PR_IO_FLUSHER (PF_MEMALLOC_NOIO | PF_LOCAL_THROTTLE)
2300 #ifdef CONFIG_ANON_VMA_NAME
2302 #define ANON_VMA_NAME_MAX_LEN 80
2303 #define ANON_VMA_NAME_INVALID_CHARS "\\`$[]"
2305 static inline bool is_valid_name_char(char ch)
2307 /* printable ascii characters, excluding ANON_VMA_NAME_INVALID_CHARS */
2308 return ch > 0x1f && ch < 0x7f &&
2309 !strchr(ANON_VMA_NAME_INVALID_CHARS, ch);
2312 static int prctl_set_vma(unsigned long opt, unsigned long addr,
2313 unsigned long size, unsigned long arg)
2315 struct mm_struct *mm = current->mm;
2316 const char __user *uname;
2317 struct anon_vma_name *anon_name = NULL;
2321 case PR_SET_VMA_ANON_NAME:
2322 uname = (const char __user *)arg;
2326 name = strndup_user(uname, ANON_VMA_NAME_MAX_LEN);
2328 return PTR_ERR(name);
2330 for (pch = name; *pch != '\0'; pch++) {
2331 if (!is_valid_name_char(*pch)) {
2336 /* anon_vma has its own copy */
2337 anon_name = anon_vma_name_alloc(name);
2344 mmap_write_lock(mm);
2345 error = madvise_set_anon_name(mm, addr, size, anon_name);
2346 mmap_write_unlock(mm);
2347 anon_vma_name_put(anon_name);
2356 #else /* CONFIG_ANON_VMA_NAME */
2357 static int prctl_set_vma(unsigned long opt, unsigned long start,
2358 unsigned long size, unsigned long arg)
2362 #endif /* CONFIG_ANON_VMA_NAME */
2364 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2365 unsigned long, arg4, unsigned long, arg5)
2367 struct task_struct *me = current;
2368 unsigned char comm[sizeof(me->comm)];
2371 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2372 if (error != -ENOSYS)
2377 case PR_SET_PDEATHSIG:
2378 if (!valid_signal(arg2)) {
2382 me->pdeath_signal = arg2;
2384 case PR_GET_PDEATHSIG:
2385 error = put_user(me->pdeath_signal, (int __user *)arg2);
2387 case PR_GET_DUMPABLE:
2388 error = get_dumpable(me->mm);
2390 case PR_SET_DUMPABLE:
2391 if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) {
2395 set_dumpable(me->mm, arg2);
2398 case PR_SET_UNALIGN:
2399 error = SET_UNALIGN_CTL(me, arg2);
2401 case PR_GET_UNALIGN:
2402 error = GET_UNALIGN_CTL(me, arg2);
2405 error = SET_FPEMU_CTL(me, arg2);
2408 error = GET_FPEMU_CTL(me, arg2);
2411 error = SET_FPEXC_CTL(me, arg2);
2414 error = GET_FPEXC_CTL(me, arg2);
2417 error = PR_TIMING_STATISTICAL;
2420 if (arg2 != PR_TIMING_STATISTICAL)
2424 comm[sizeof(me->comm) - 1] = 0;
2425 if (strncpy_from_user(comm, (char __user *)arg2,
2426 sizeof(me->comm) - 1) < 0)
2428 set_task_comm(me, comm);
2429 proc_comm_connector(me);
2432 get_task_comm(comm, me);
2433 if (copy_to_user((char __user *)arg2, comm, sizeof(comm)))
2437 error = GET_ENDIAN(me, arg2);
2440 error = SET_ENDIAN(me, arg2);
2442 case PR_GET_SECCOMP:
2443 error = prctl_get_seccomp();
2445 case PR_SET_SECCOMP:
2446 error = prctl_set_seccomp(arg2, (char __user *)arg3);
2449 error = GET_TSC_CTL(arg2);
2452 error = SET_TSC_CTL(arg2);
2454 case PR_TASK_PERF_EVENTS_DISABLE:
2455 error = perf_event_task_disable();
2457 case PR_TASK_PERF_EVENTS_ENABLE:
2458 error = perf_event_task_enable();
2460 case PR_GET_TIMERSLACK:
2461 if (current->timer_slack_ns > ULONG_MAX)
2464 error = current->timer_slack_ns;
2466 case PR_SET_TIMERSLACK:
2468 current->timer_slack_ns =
2469 current->default_timer_slack_ns;
2471 current->timer_slack_ns = arg2;
2477 case PR_MCE_KILL_CLEAR:
2480 current->flags &= ~PF_MCE_PROCESS;
2482 case PR_MCE_KILL_SET:
2483 current->flags |= PF_MCE_PROCESS;
2484 if (arg3 == PR_MCE_KILL_EARLY)
2485 current->flags |= PF_MCE_EARLY;
2486 else if (arg3 == PR_MCE_KILL_LATE)
2487 current->flags &= ~PF_MCE_EARLY;
2488 else if (arg3 == PR_MCE_KILL_DEFAULT)
2490 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
2498 case PR_MCE_KILL_GET:
2499 if (arg2 | arg3 | arg4 | arg5)
2501 if (current->flags & PF_MCE_PROCESS)
2502 error = (current->flags & PF_MCE_EARLY) ?
2503 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2505 error = PR_MCE_KILL_DEFAULT;
2508 error = prctl_set_mm(arg2, arg3, arg4, arg5);
2510 case PR_GET_TID_ADDRESS:
2511 error = prctl_get_tid_address(me, (int __user * __user *)arg2);
2513 case PR_SET_CHILD_SUBREAPER:
2514 me->signal->is_child_subreaper = !!arg2;
2518 walk_process_tree(me, propagate_has_child_subreaper, NULL);
2520 case PR_GET_CHILD_SUBREAPER:
2521 error = put_user(me->signal->is_child_subreaper,
2522 (int __user *)arg2);
2524 case PR_SET_NO_NEW_PRIVS:
2525 if (arg2 != 1 || arg3 || arg4 || arg5)
2528 task_set_no_new_privs(current);
2530 case PR_GET_NO_NEW_PRIVS:
2531 if (arg2 || arg3 || arg4 || arg5)
2533 return task_no_new_privs(current) ? 1 : 0;
2534 case PR_GET_THP_DISABLE:
2535 if (arg2 || arg3 || arg4 || arg5)
2537 error = !!test_bit(MMF_DISABLE_THP, &me->mm->flags);
2539 case PR_SET_THP_DISABLE:
2540 if (arg3 || arg4 || arg5)
2542 if (mmap_write_lock_killable(me->mm))
2545 set_bit(MMF_DISABLE_THP, &me->mm->flags);
2547 clear_bit(MMF_DISABLE_THP, &me->mm->flags);
2548 mmap_write_unlock(me->mm);
2550 case PR_MPX_ENABLE_MANAGEMENT:
2551 case PR_MPX_DISABLE_MANAGEMENT:
2552 /* No longer implemented: */
2554 case PR_SET_FP_MODE:
2555 error = SET_FP_MODE(me, arg2);
2557 case PR_GET_FP_MODE:
2558 error = GET_FP_MODE(me);
2561 error = SVE_SET_VL(arg2);
2564 error = SVE_GET_VL();
2567 error = SME_SET_VL(arg2);
2570 error = SME_GET_VL();
2572 case PR_GET_SPECULATION_CTRL:
2573 if (arg3 || arg4 || arg5)
2575 error = arch_prctl_spec_ctrl_get(me, arg2);
2577 case PR_SET_SPECULATION_CTRL:
2580 error = arch_prctl_spec_ctrl_set(me, arg2, arg3);
2582 case PR_PAC_RESET_KEYS:
2583 if (arg3 || arg4 || arg5)
2585 error = PAC_RESET_KEYS(me, arg2);
2587 case PR_PAC_SET_ENABLED_KEYS:
2590 error = PAC_SET_ENABLED_KEYS(me, arg2, arg3);
2592 case PR_PAC_GET_ENABLED_KEYS:
2593 if (arg2 || arg3 || arg4 || arg5)
2595 error = PAC_GET_ENABLED_KEYS(me);
2597 case PR_SET_TAGGED_ADDR_CTRL:
2598 if (arg3 || arg4 || arg5)
2600 error = SET_TAGGED_ADDR_CTRL(arg2);
2602 case PR_GET_TAGGED_ADDR_CTRL:
2603 if (arg2 || arg3 || arg4 || arg5)
2605 error = GET_TAGGED_ADDR_CTRL();
2607 case PR_SET_IO_FLUSHER:
2608 if (!capable(CAP_SYS_RESOURCE))
2611 if (arg3 || arg4 || arg5)
2615 current->flags |= PR_IO_FLUSHER;
2617 current->flags &= ~PR_IO_FLUSHER;
2621 case PR_GET_IO_FLUSHER:
2622 if (!capable(CAP_SYS_RESOURCE))
2625 if (arg2 || arg3 || arg4 || arg5)
2628 error = (current->flags & PR_IO_FLUSHER) == PR_IO_FLUSHER;
2630 case PR_SET_SYSCALL_USER_DISPATCH:
2631 error = set_syscall_user_dispatch(arg2, arg3, arg4,
2632 (char __user *) arg5);
2634 #ifdef CONFIG_SCHED_CORE
2636 error = sched_core_share_pid(arg2, arg3, arg4, arg5);
2640 error = prctl_set_vma(arg2, arg3, arg4, arg5);
2649 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2650 struct getcpu_cache __user *, unused)
2653 int cpu = raw_smp_processor_id();
2656 err |= put_user(cpu, cpup);
2658 err |= put_user(cpu_to_node(cpu), nodep);
2659 return err ? -EFAULT : 0;
2663 * do_sysinfo - fill in sysinfo struct
2664 * @info: pointer to buffer to fill
2666 static int do_sysinfo(struct sysinfo *info)
2668 unsigned long mem_total, sav_total;
2669 unsigned int mem_unit, bitcount;
2670 struct timespec64 tp;
2672 memset(info, 0, sizeof(struct sysinfo));
2674 ktime_get_boottime_ts64(&tp);
2675 timens_add_boottime(&tp);
2676 info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
2678 get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
2680 info->procs = nr_threads;
2686 * If the sum of all the available memory (i.e. ram + swap)
2687 * is less than can be stored in a 32 bit unsigned long then
2688 * we can be binary compatible with 2.2.x kernels. If not,
2689 * well, in that case 2.2.x was broken anyways...
2691 * -Erik Andersen <andersee@debian.org>
2694 mem_total = info->totalram + info->totalswap;
2695 if (mem_total < info->totalram || mem_total < info->totalswap)
2698 mem_unit = info->mem_unit;
2699 while (mem_unit > 1) {
2702 sav_total = mem_total;
2704 if (mem_total < sav_total)
2709 * If mem_total did not overflow, multiply all memory values by
2710 * info->mem_unit and set it to 1. This leaves things compatible
2711 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2716 info->totalram <<= bitcount;
2717 info->freeram <<= bitcount;
2718 info->sharedram <<= bitcount;
2719 info->bufferram <<= bitcount;
2720 info->totalswap <<= bitcount;
2721 info->freeswap <<= bitcount;
2722 info->totalhigh <<= bitcount;
2723 info->freehigh <<= bitcount;
2729 SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
2735 if (copy_to_user(info, &val, sizeof(struct sysinfo)))
2741 #ifdef CONFIG_COMPAT
2742 struct compat_sysinfo {
2756 char _f[20-2*sizeof(u32)-sizeof(int)];
2759 COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2762 struct compat_sysinfo s_32;
2766 /* Check to see if any memory value is too large for 32-bit and scale
2769 if (upper_32_bits(s.totalram) || upper_32_bits(s.totalswap)) {
2772 while (s.mem_unit < PAGE_SIZE) {
2777 s.totalram >>= bitcount;
2778 s.freeram >>= bitcount;
2779 s.sharedram >>= bitcount;
2780 s.bufferram >>= bitcount;
2781 s.totalswap >>= bitcount;
2782 s.freeswap >>= bitcount;
2783 s.totalhigh >>= bitcount;
2784 s.freehigh >>= bitcount;
2787 memset(&s_32, 0, sizeof(s_32));
2788 s_32.uptime = s.uptime;
2789 s_32.loads[0] = s.loads[0];
2790 s_32.loads[1] = s.loads[1];
2791 s_32.loads[2] = s.loads[2];
2792 s_32.totalram = s.totalram;
2793 s_32.freeram = s.freeram;
2794 s_32.sharedram = s.sharedram;
2795 s_32.bufferram = s.bufferram;
2796 s_32.totalswap = s.totalswap;
2797 s_32.freeswap = s.freeswap;
2798 s_32.procs = s.procs;
2799 s_32.totalhigh = s.totalhigh;
2800 s_32.freehigh = s.freehigh;
2801 s_32.mem_unit = s.mem_unit;
2802 if (copy_to_user(info, &s_32, sizeof(s_32)))
2806 #endif /* CONFIG_COMPAT */