1 // SPDX-License-Identifier: GPL-2.0
5 * Copyright (C) 1991, 1992 Linus Torvalds
8 #include <linux/export.h>
10 #include <linux/utsname.h>
11 #include <linux/mman.h>
12 #include <linux/reboot.h>
13 #include <linux/prctl.h>
14 #include <linux/highuid.h>
16 #include <linux/kmod.h>
17 #include <linux/perf_event.h>
18 #include <linux/resource.h>
19 #include <linux/kernel.h>
20 #include <linux/workqueue.h>
21 #include <linux/capability.h>
22 #include <linux/device.h>
23 #include <linux/key.h>
24 #include <linux/times.h>
25 #include <linux/posix-timers.h>
26 #include <linux/security.h>
27 #include <linux/dcookies.h>
28 #include <linux/suspend.h>
29 #include <linux/tty.h>
30 #include <linux/signal.h>
31 #include <linux/cn_proc.h>
32 #include <linux/getcpu.h>
33 #include <linux/task_io_accounting_ops.h>
34 #include <linux/seccomp.h>
35 #include <linux/cpu.h>
36 #include <linux/personality.h>
37 #include <linux/ptrace.h>
38 #include <linux/fs_struct.h>
39 #include <linux/file.h>
40 #include <linux/mount.h>
41 #include <linux/gfp.h>
42 #include <linux/syscore_ops.h>
43 #include <linux/version.h>
44 #include <linux/ctype.h>
46 #include <linux/compat.h>
47 #include <linux/syscalls.h>
48 #include <linux/kprobes.h>
49 #include <linux/user_namespace.h>
50 #include <linux/time_namespace.h>
51 #include <linux/binfmts.h>
53 #include <linux/sched.h>
54 #include <linux/sched/autogroup.h>
55 #include <linux/sched/loadavg.h>
56 #include <linux/sched/stat.h>
57 #include <linux/sched/mm.h>
58 #include <linux/sched/coredump.h>
59 #include <linux/sched/task.h>
60 #include <linux/sched/cputime.h>
61 #include <linux/rcupdate.h>
62 #include <linux/uidgid.h>
63 #include <linux/cred.h>
65 #include <linux/nospec.h>
67 #include <linux/kmsg_dump.h>
68 /* Move somewhere else to avoid recompiling? */
69 #include <generated/utsrelease.h>
71 #include <linux/uaccess.h>
73 #include <asm/unistd.h>
77 #ifndef SET_UNALIGN_CTL
78 # define SET_UNALIGN_CTL(a, b) (-EINVAL)
80 #ifndef GET_UNALIGN_CTL
81 # define GET_UNALIGN_CTL(a, b) (-EINVAL)
84 # define SET_FPEMU_CTL(a, b) (-EINVAL)
87 # define GET_FPEMU_CTL(a, b) (-EINVAL)
90 # define SET_FPEXC_CTL(a, b) (-EINVAL)
93 # define GET_FPEXC_CTL(a, b) (-EINVAL)
96 # define GET_ENDIAN(a, b) (-EINVAL)
99 # define SET_ENDIAN(a, b) (-EINVAL)
102 # define GET_TSC_CTL(a) (-EINVAL)
105 # define SET_TSC_CTL(a) (-EINVAL)
108 # define GET_FP_MODE(a) (-EINVAL)
111 # define SET_FP_MODE(a,b) (-EINVAL)
114 # define SVE_SET_VL(a) (-EINVAL)
117 # define SVE_GET_VL() (-EINVAL)
119 #ifndef PAC_RESET_KEYS
120 # define PAC_RESET_KEYS(a, b) (-EINVAL)
122 #ifndef SET_TAGGED_ADDR_CTRL
123 # define SET_TAGGED_ADDR_CTRL(a) (-EINVAL)
125 #ifndef GET_TAGGED_ADDR_CTRL
126 # define GET_TAGGED_ADDR_CTRL() (-EINVAL)
130 * this is where the system-wide overflow UID and GID are defined, for
131 * architectures that now have 32-bit UID/GID but didn't in the past
134 int overflowuid = DEFAULT_OVERFLOWUID;
135 int overflowgid = DEFAULT_OVERFLOWGID;
137 EXPORT_SYMBOL(overflowuid);
138 EXPORT_SYMBOL(overflowgid);
141 * the same as above, but for filesystems which can only store a 16-bit
142 * UID and GID. as such, this is needed on all architectures
145 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
146 int fs_overflowgid = DEFAULT_FS_OVERFLOWGID;
148 EXPORT_SYMBOL(fs_overflowuid);
149 EXPORT_SYMBOL(fs_overflowgid);
152 * Returns true if current's euid is same as p's uid or euid,
153 * or has CAP_SYS_NICE to p's user_ns.
155 * Called with rcu_read_lock, creds are safe
157 static bool set_one_prio_perm(struct task_struct *p)
159 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
161 if (uid_eq(pcred->uid, cred->euid) ||
162 uid_eq(pcred->euid, cred->euid))
164 if (ns_capable(pcred->user_ns, CAP_SYS_NICE))
170 * set the priority of a task
171 * - the caller must hold the RCU read lock
173 static int set_one_prio(struct task_struct *p, int niceval, int error)
177 if (!set_one_prio_perm(p)) {
181 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
185 no_nice = security_task_setnice(p, niceval);
192 set_user_nice(p, niceval);
197 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
199 struct task_struct *g, *p;
200 struct user_struct *user;
201 const struct cred *cred = current_cred();
206 if (which > PRIO_USER || which < PRIO_PROCESS)
209 /* normalize: avoid signed division (rounding problems) */
211 if (niceval < MIN_NICE)
213 if (niceval > MAX_NICE)
217 read_lock(&tasklist_lock);
221 p = find_task_by_vpid(who);
225 error = set_one_prio(p, niceval, error);
229 pgrp = find_vpid(who);
231 pgrp = task_pgrp(current);
232 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
233 error = set_one_prio(p, niceval, error);
234 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
237 uid = make_kuid(cred->user_ns, who);
241 else if (!uid_eq(uid, cred->uid)) {
242 user = find_user(uid);
244 goto out_unlock; /* No processes for this user */
246 do_each_thread(g, p) {
247 if (uid_eq(task_uid(p), uid) && task_pid_vnr(p))
248 error = set_one_prio(p, niceval, error);
249 } while_each_thread(g, p);
250 if (!uid_eq(uid, cred->uid))
251 free_uid(user); /* For find_user() */
255 read_unlock(&tasklist_lock);
262 * Ugh. To avoid negative return values, "getpriority()" will
263 * not return the normal nice-value, but a negated value that
264 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
265 * to stay compatible.
267 SYSCALL_DEFINE2(getpriority, int, which, int, who)
269 struct task_struct *g, *p;
270 struct user_struct *user;
271 const struct cred *cred = current_cred();
272 long niceval, retval = -ESRCH;
276 if (which > PRIO_USER || which < PRIO_PROCESS)
280 read_lock(&tasklist_lock);
284 p = find_task_by_vpid(who);
288 niceval = nice_to_rlimit(task_nice(p));
289 if (niceval > retval)
295 pgrp = find_vpid(who);
297 pgrp = task_pgrp(current);
298 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
299 niceval = nice_to_rlimit(task_nice(p));
300 if (niceval > retval)
302 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
305 uid = make_kuid(cred->user_ns, who);
309 else if (!uid_eq(uid, cred->uid)) {
310 user = find_user(uid);
312 goto out_unlock; /* No processes for this user */
314 do_each_thread(g, p) {
315 if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) {
316 niceval = nice_to_rlimit(task_nice(p));
317 if (niceval > retval)
320 } while_each_thread(g, p);
321 if (!uid_eq(uid, cred->uid))
322 free_uid(user); /* for find_user() */
326 read_unlock(&tasklist_lock);
333 * Unprivileged users may change the real gid to the effective gid
334 * or vice versa. (BSD-style)
336 * If you set the real gid at all, or set the effective gid to a value not
337 * equal to the real gid, then the saved gid is set to the new effective gid.
339 * This makes it possible for a setgid program to completely drop its
340 * privileges, which is often a useful assertion to make when you are doing
341 * a security audit over a program.
343 * The general idea is that a program which uses just setregid() will be
344 * 100% compatible with BSD. A program which uses just setgid() will be
345 * 100% compatible with POSIX with saved IDs.
347 * SMP: There are not races, the GIDs are checked only by filesystem
348 * operations (as far as semantic preservation is concerned).
350 #ifdef CONFIG_MULTIUSER
351 long __sys_setregid(gid_t rgid, gid_t egid)
353 struct user_namespace *ns = current_user_ns();
354 const struct cred *old;
359 krgid = make_kgid(ns, rgid);
360 kegid = make_kgid(ns, egid);
362 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
364 if ((egid != (gid_t) -1) && !gid_valid(kegid))
367 new = prepare_creds();
370 old = current_cred();
373 if (rgid != (gid_t) -1) {
374 if (gid_eq(old->gid, krgid) ||
375 gid_eq(old->egid, krgid) ||
376 ns_capable(old->user_ns, CAP_SETGID))
381 if (egid != (gid_t) -1) {
382 if (gid_eq(old->gid, kegid) ||
383 gid_eq(old->egid, kegid) ||
384 gid_eq(old->sgid, kegid) ||
385 ns_capable(old->user_ns, CAP_SETGID))
391 if (rgid != (gid_t) -1 ||
392 (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
393 new->sgid = new->egid;
394 new->fsgid = new->egid;
396 return commit_creds(new);
403 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
405 return __sys_setregid(rgid, egid);
409 * setgid() is implemented like SysV w/ SAVED_IDS
411 * SMP: Same implicit races as above.
413 long __sys_setgid(gid_t gid)
415 struct user_namespace *ns = current_user_ns();
416 const struct cred *old;
421 kgid = make_kgid(ns, gid);
422 if (!gid_valid(kgid))
425 new = prepare_creds();
428 old = current_cred();
431 if (ns_capable(old->user_ns, CAP_SETGID))
432 new->gid = new->egid = new->sgid = new->fsgid = kgid;
433 else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
434 new->egid = new->fsgid = kgid;
438 return commit_creds(new);
445 SYSCALL_DEFINE1(setgid, gid_t, gid)
447 return __sys_setgid(gid);
451 * change the user struct in a credentials set to match the new UID
453 static int set_user(struct cred *new)
455 struct user_struct *new_user;
457 new_user = alloc_uid(new->uid);
462 * We don't fail in case of NPROC limit excess here because too many
463 * poorly written programs don't check set*uid() return code, assuming
464 * it never fails if called by root. We may still enforce NPROC limit
465 * for programs doing set*uid()+execve() by harmlessly deferring the
466 * failure to the execve() stage.
468 if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
469 new_user != INIT_USER)
470 current->flags |= PF_NPROC_EXCEEDED;
472 current->flags &= ~PF_NPROC_EXCEEDED;
475 new->user = new_user;
480 * Unprivileged users may change the real uid to the effective uid
481 * or vice versa. (BSD-style)
483 * If you set the real uid at all, or set the effective uid to a value not
484 * equal to the real uid, then the saved uid is set to the new effective uid.
486 * This makes it possible for a setuid program to completely drop its
487 * privileges, which is often a useful assertion to make when you are doing
488 * a security audit over a program.
490 * The general idea is that a program which uses just setreuid() will be
491 * 100% compatible with BSD. A program which uses just setuid() will be
492 * 100% compatible with POSIX with saved IDs.
494 long __sys_setreuid(uid_t ruid, uid_t euid)
496 struct user_namespace *ns = current_user_ns();
497 const struct cred *old;
502 kruid = make_kuid(ns, ruid);
503 keuid = make_kuid(ns, euid);
505 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
507 if ((euid != (uid_t) -1) && !uid_valid(keuid))
510 new = prepare_creds();
513 old = current_cred();
516 if (ruid != (uid_t) -1) {
518 if (!uid_eq(old->uid, kruid) &&
519 !uid_eq(old->euid, kruid) &&
520 !ns_capable_setid(old->user_ns, CAP_SETUID))
524 if (euid != (uid_t) -1) {
526 if (!uid_eq(old->uid, keuid) &&
527 !uid_eq(old->euid, keuid) &&
528 !uid_eq(old->suid, keuid) &&
529 !ns_capable_setid(old->user_ns, CAP_SETUID))
533 if (!uid_eq(new->uid, old->uid)) {
534 retval = set_user(new);
538 if (ruid != (uid_t) -1 ||
539 (euid != (uid_t) -1 && !uid_eq(keuid, old->uid)))
540 new->suid = new->euid;
541 new->fsuid = new->euid;
543 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
547 return commit_creds(new);
554 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
556 return __sys_setreuid(ruid, euid);
560 * setuid() is implemented like SysV with SAVED_IDS
562 * Note that SAVED_ID's is deficient in that a setuid root program
563 * like sendmail, for example, cannot set its uid to be a normal
564 * user and then switch back, because if you're root, setuid() sets
565 * the saved uid too. If you don't like this, blame the bright people
566 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
567 * will allow a root program to temporarily drop privileges and be able to
568 * regain them by swapping the real and effective uid.
570 long __sys_setuid(uid_t uid)
572 struct user_namespace *ns = current_user_ns();
573 const struct cred *old;
578 kuid = make_kuid(ns, uid);
579 if (!uid_valid(kuid))
582 new = prepare_creds();
585 old = current_cred();
588 if (ns_capable_setid(old->user_ns, CAP_SETUID)) {
589 new->suid = new->uid = kuid;
590 if (!uid_eq(kuid, old->uid)) {
591 retval = set_user(new);
595 } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) {
599 new->fsuid = new->euid = kuid;
601 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
605 return commit_creds(new);
612 SYSCALL_DEFINE1(setuid, uid_t, uid)
614 return __sys_setuid(uid);
619 * This function implements a generic ability to update ruid, euid,
620 * and suid. This allows you to implement the 4.4 compatible seteuid().
622 long __sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
624 struct user_namespace *ns = current_user_ns();
625 const struct cred *old;
628 kuid_t kruid, keuid, ksuid;
630 kruid = make_kuid(ns, ruid);
631 keuid = make_kuid(ns, euid);
632 ksuid = make_kuid(ns, suid);
634 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
637 if ((euid != (uid_t) -1) && !uid_valid(keuid))
640 if ((suid != (uid_t) -1) && !uid_valid(ksuid))
643 new = prepare_creds();
647 old = current_cred();
650 if (!ns_capable_setid(old->user_ns, CAP_SETUID)) {
651 if (ruid != (uid_t) -1 && !uid_eq(kruid, old->uid) &&
652 !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid))
654 if (euid != (uid_t) -1 && !uid_eq(keuid, old->uid) &&
655 !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid))
657 if (suid != (uid_t) -1 && !uid_eq(ksuid, old->uid) &&
658 !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid))
662 if (ruid != (uid_t) -1) {
664 if (!uid_eq(kruid, old->uid)) {
665 retval = set_user(new);
670 if (euid != (uid_t) -1)
672 if (suid != (uid_t) -1)
674 new->fsuid = new->euid;
676 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
680 return commit_creds(new);
687 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
689 return __sys_setresuid(ruid, euid, suid);
692 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
694 const struct cred *cred = current_cred();
696 uid_t ruid, euid, suid;
698 ruid = from_kuid_munged(cred->user_ns, cred->uid);
699 euid = from_kuid_munged(cred->user_ns, cred->euid);
700 suid = from_kuid_munged(cred->user_ns, cred->suid);
702 retval = put_user(ruid, ruidp);
704 retval = put_user(euid, euidp);
706 return put_user(suid, suidp);
712 * Same as above, but for rgid, egid, sgid.
714 long __sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
716 struct user_namespace *ns = current_user_ns();
717 const struct cred *old;
720 kgid_t krgid, kegid, ksgid;
722 krgid = make_kgid(ns, rgid);
723 kegid = make_kgid(ns, egid);
724 ksgid = make_kgid(ns, sgid);
726 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
728 if ((egid != (gid_t) -1) && !gid_valid(kegid))
730 if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
733 new = prepare_creds();
736 old = current_cred();
739 if (!ns_capable(old->user_ns, CAP_SETGID)) {
740 if (rgid != (gid_t) -1 && !gid_eq(krgid, old->gid) &&
741 !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid))
743 if (egid != (gid_t) -1 && !gid_eq(kegid, old->gid) &&
744 !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid))
746 if (sgid != (gid_t) -1 && !gid_eq(ksgid, old->gid) &&
747 !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid))
751 if (rgid != (gid_t) -1)
753 if (egid != (gid_t) -1)
755 if (sgid != (gid_t) -1)
757 new->fsgid = new->egid;
759 return commit_creds(new);
766 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
768 return __sys_setresgid(rgid, egid, sgid);
771 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
773 const struct cred *cred = current_cred();
775 gid_t rgid, egid, sgid;
777 rgid = from_kgid_munged(cred->user_ns, cred->gid);
778 egid = from_kgid_munged(cred->user_ns, cred->egid);
779 sgid = from_kgid_munged(cred->user_ns, cred->sgid);
781 retval = put_user(rgid, rgidp);
783 retval = put_user(egid, egidp);
785 retval = put_user(sgid, sgidp);
793 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
794 * is used for "access()" and for the NFS daemon (letting nfsd stay at
795 * whatever uid it wants to). It normally shadows "euid", except when
796 * explicitly set by setfsuid() or for access..
798 long __sys_setfsuid(uid_t uid)
800 const struct cred *old;
805 old = current_cred();
806 old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
808 kuid = make_kuid(old->user_ns, uid);
809 if (!uid_valid(kuid))
812 new = prepare_creds();
816 if (uid_eq(kuid, old->uid) || uid_eq(kuid, old->euid) ||
817 uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
818 ns_capable_setid(old->user_ns, CAP_SETUID)) {
819 if (!uid_eq(kuid, old->fsuid)) {
821 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
834 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
836 return __sys_setfsuid(uid);
840 * Samma på svenska..
842 long __sys_setfsgid(gid_t gid)
844 const struct cred *old;
849 old = current_cred();
850 old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
852 kgid = make_kgid(old->user_ns, gid);
853 if (!gid_valid(kgid))
856 new = prepare_creds();
860 if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->egid) ||
861 gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
862 ns_capable(old->user_ns, CAP_SETGID)) {
863 if (!gid_eq(kgid, old->fsgid)) {
877 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
879 return __sys_setfsgid(gid);
881 #endif /* CONFIG_MULTIUSER */
884 * sys_getpid - return the thread group id of the current process
886 * Note, despite the name, this returns the tgid not the pid. The tgid and
887 * the pid are identical unless CLONE_THREAD was specified on clone() in
888 * which case the tgid is the same in all threads of the same group.
890 * This is SMP safe as current->tgid does not change.
892 SYSCALL_DEFINE0(getpid)
894 return task_tgid_vnr(current);
897 /* Thread ID - the internal kernel "pid" */
898 SYSCALL_DEFINE0(gettid)
900 return task_pid_vnr(current);
904 * Accessing ->real_parent is not SMP-safe, it could
905 * change from under us. However, we can use a stale
906 * value of ->real_parent under rcu_read_lock(), see
907 * release_task()->call_rcu(delayed_put_task_struct).
909 SYSCALL_DEFINE0(getppid)
914 pid = task_tgid_vnr(rcu_dereference(current->real_parent));
920 SYSCALL_DEFINE0(getuid)
922 /* Only we change this so SMP safe */
923 return from_kuid_munged(current_user_ns(), current_uid());
926 SYSCALL_DEFINE0(geteuid)
928 /* Only we change this so SMP safe */
929 return from_kuid_munged(current_user_ns(), current_euid());
932 SYSCALL_DEFINE0(getgid)
934 /* Only we change this so SMP safe */
935 return from_kgid_munged(current_user_ns(), current_gid());
938 SYSCALL_DEFINE0(getegid)
940 /* Only we change this so SMP safe */
941 return from_kgid_munged(current_user_ns(), current_egid());
944 static void do_sys_times(struct tms *tms)
946 u64 tgutime, tgstime, cutime, cstime;
948 thread_group_cputime_adjusted(current, &tgutime, &tgstime);
949 cutime = current->signal->cutime;
950 cstime = current->signal->cstime;
951 tms->tms_utime = nsec_to_clock_t(tgutime);
952 tms->tms_stime = nsec_to_clock_t(tgstime);
953 tms->tms_cutime = nsec_to_clock_t(cutime);
954 tms->tms_cstime = nsec_to_clock_t(cstime);
957 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
963 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
966 force_successful_syscall_return();
967 return (long) jiffies_64_to_clock_t(get_jiffies_64());
971 static compat_clock_t clock_t_to_compat_clock_t(clock_t x)
973 return compat_jiffies_to_clock_t(clock_t_to_jiffies(x));
976 COMPAT_SYSCALL_DEFINE1(times, struct compat_tms __user *, tbuf)
980 struct compat_tms tmp;
983 /* Convert our struct tms to the compat version. */
984 tmp.tms_utime = clock_t_to_compat_clock_t(tms.tms_utime);
985 tmp.tms_stime = clock_t_to_compat_clock_t(tms.tms_stime);
986 tmp.tms_cutime = clock_t_to_compat_clock_t(tms.tms_cutime);
987 tmp.tms_cstime = clock_t_to_compat_clock_t(tms.tms_cstime);
988 if (copy_to_user(tbuf, &tmp, sizeof(tmp)))
991 force_successful_syscall_return();
992 return compat_jiffies_to_clock_t(jiffies);
997 * This needs some heavy checking ...
998 * I just haven't the stomach for it. I also don't fully
999 * understand sessions/pgrp etc. Let somebody who does explain it.
1001 * OK, I think I have the protection semantics right.... this is really
1002 * only important on a multi-user system anyway, to make sure one user
1003 * can't send a signal to a process owned by another. -TYT, 12/12/91
1005 * !PF_FORKNOEXEC check to conform completely to POSIX.
1007 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
1009 struct task_struct *p;
1010 struct task_struct *group_leader = current->group_leader;
1015 pid = task_pid_vnr(group_leader);
1022 /* From this point forward we keep holding onto the tasklist lock
1023 * so that our parent does not change from under us. -DaveM
1025 write_lock_irq(&tasklist_lock);
1028 p = find_task_by_vpid(pid);
1033 if (!thread_group_leader(p))
1036 if (same_thread_group(p->real_parent, group_leader)) {
1038 if (task_session(p) != task_session(group_leader))
1041 if (!(p->flags & PF_FORKNOEXEC))
1045 if (p != group_leader)
1050 if (p->signal->leader)
1055 struct task_struct *g;
1057 pgrp = find_vpid(pgid);
1058 g = pid_task(pgrp, PIDTYPE_PGID);
1059 if (!g || task_session(g) != task_session(group_leader))
1063 err = security_task_setpgid(p, pgid);
1067 if (task_pgrp(p) != pgrp)
1068 change_pid(p, PIDTYPE_PGID, pgrp);
1072 /* All paths lead to here, thus we are safe. -DaveM */
1073 write_unlock_irq(&tasklist_lock);
1078 static int do_getpgid(pid_t pid)
1080 struct task_struct *p;
1086 grp = task_pgrp(current);
1089 p = find_task_by_vpid(pid);
1096 retval = security_task_getpgid(p);
1100 retval = pid_vnr(grp);
1106 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1108 return do_getpgid(pid);
1111 #ifdef __ARCH_WANT_SYS_GETPGRP
1113 SYSCALL_DEFINE0(getpgrp)
1115 return do_getpgid(0);
1120 SYSCALL_DEFINE1(getsid, pid_t, pid)
1122 struct task_struct *p;
1128 sid = task_session(current);
1131 p = find_task_by_vpid(pid);
1134 sid = task_session(p);
1138 retval = security_task_getsid(p);
1142 retval = pid_vnr(sid);
1148 static void set_special_pids(struct pid *pid)
1150 struct task_struct *curr = current->group_leader;
1152 if (task_session(curr) != pid)
1153 change_pid(curr, PIDTYPE_SID, pid);
1155 if (task_pgrp(curr) != pid)
1156 change_pid(curr, PIDTYPE_PGID, pid);
1159 int ksys_setsid(void)
1161 struct task_struct *group_leader = current->group_leader;
1162 struct pid *sid = task_pid(group_leader);
1163 pid_t session = pid_vnr(sid);
1166 write_lock_irq(&tasklist_lock);
1167 /* Fail if I am already a session leader */
1168 if (group_leader->signal->leader)
1171 /* Fail if a process group id already exists that equals the
1172 * proposed session id.
1174 if (pid_task(sid, PIDTYPE_PGID))
1177 group_leader->signal->leader = 1;
1178 set_special_pids(sid);
1180 proc_clear_tty(group_leader);
1184 write_unlock_irq(&tasklist_lock);
1186 proc_sid_connector(group_leader);
1187 sched_autogroup_create_attach(group_leader);
1192 SYSCALL_DEFINE0(setsid)
1194 return ksys_setsid();
1197 DECLARE_RWSEM(uts_sem);
1199 #ifdef COMPAT_UTS_MACHINE
1200 #define override_architecture(name) \
1201 (personality(current->personality) == PER_LINUX32 && \
1202 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1203 sizeof(COMPAT_UTS_MACHINE)))
1205 #define override_architecture(name) 0
1209 * Work around broken programs that cannot handle "Linux 3.0".
1210 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1211 * And we map 4.x and later versions to 2.6.60+x, so 4.0/5.0/6.0/... would be
1214 static int override_release(char __user *release, size_t len)
1218 if (current->personality & UNAME26) {
1219 const char *rest = UTS_RELEASE;
1220 char buf[65] = { 0 };
1226 if (*rest == '.' && ++ndots >= 3)
1228 if (!isdigit(*rest) && *rest != '.')
1232 v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 60;
1233 copy = clamp_t(size_t, len, 1, sizeof(buf));
1234 copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1235 ret = copy_to_user(release, buf, copy + 1);
1240 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1242 struct new_utsname tmp;
1244 down_read(&uts_sem);
1245 memcpy(&tmp, utsname(), sizeof(tmp));
1247 if (copy_to_user(name, &tmp, sizeof(tmp)))
1250 if (override_release(name->release, sizeof(name->release)))
1252 if (override_architecture(name))
1257 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1261 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1263 struct old_utsname tmp;
1268 down_read(&uts_sem);
1269 memcpy(&tmp, utsname(), sizeof(tmp));
1271 if (copy_to_user(name, &tmp, sizeof(tmp)))
1274 if (override_release(name->release, sizeof(name->release)))
1276 if (override_architecture(name))
1281 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1283 struct oldold_utsname tmp;
1288 memset(&tmp, 0, sizeof(tmp));
1290 down_read(&uts_sem);
1291 memcpy(&tmp.sysname, &utsname()->sysname, __OLD_UTS_LEN);
1292 memcpy(&tmp.nodename, &utsname()->nodename, __OLD_UTS_LEN);
1293 memcpy(&tmp.release, &utsname()->release, __OLD_UTS_LEN);
1294 memcpy(&tmp.version, &utsname()->version, __OLD_UTS_LEN);
1295 memcpy(&tmp.machine, &utsname()->machine, __OLD_UTS_LEN);
1297 if (copy_to_user(name, &tmp, sizeof(tmp)))
1300 if (override_architecture(name))
1302 if (override_release(name->release, sizeof(name->release)))
1308 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1311 char tmp[__NEW_UTS_LEN];
1313 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1316 if (len < 0 || len > __NEW_UTS_LEN)
1319 if (!copy_from_user(tmp, name, len)) {
1320 struct new_utsname *u;
1322 down_write(&uts_sem);
1324 memcpy(u->nodename, tmp, len);
1325 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1327 uts_proc_notify(UTS_PROC_HOSTNAME);
1333 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1335 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1338 struct new_utsname *u;
1339 char tmp[__NEW_UTS_LEN + 1];
1343 down_read(&uts_sem);
1345 i = 1 + strlen(u->nodename);
1348 memcpy(tmp, u->nodename, i);
1350 if (copy_to_user(name, tmp, i))
1358 * Only setdomainname; getdomainname can be implemented by calling
1361 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1364 char tmp[__NEW_UTS_LEN];
1366 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1368 if (len < 0 || len > __NEW_UTS_LEN)
1372 if (!copy_from_user(tmp, name, len)) {
1373 struct new_utsname *u;
1375 down_write(&uts_sem);
1377 memcpy(u->domainname, tmp, len);
1378 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1380 uts_proc_notify(UTS_PROC_DOMAINNAME);
1386 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1388 struct rlimit value;
1391 ret = do_prlimit(current, resource, NULL, &value);
1393 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1398 #ifdef CONFIG_COMPAT
1400 COMPAT_SYSCALL_DEFINE2(setrlimit, unsigned int, resource,
1401 struct compat_rlimit __user *, rlim)
1404 struct compat_rlimit r32;
1406 if (copy_from_user(&r32, rlim, sizeof(struct compat_rlimit)))
1409 if (r32.rlim_cur == COMPAT_RLIM_INFINITY)
1410 r.rlim_cur = RLIM_INFINITY;
1412 r.rlim_cur = r32.rlim_cur;
1413 if (r32.rlim_max == COMPAT_RLIM_INFINITY)
1414 r.rlim_max = RLIM_INFINITY;
1416 r.rlim_max = r32.rlim_max;
1417 return do_prlimit(current, resource, &r, NULL);
1420 COMPAT_SYSCALL_DEFINE2(getrlimit, unsigned int, resource,
1421 struct compat_rlimit __user *, rlim)
1426 ret = do_prlimit(current, resource, NULL, &r);
1428 struct compat_rlimit r32;
1429 if (r.rlim_cur > COMPAT_RLIM_INFINITY)
1430 r32.rlim_cur = COMPAT_RLIM_INFINITY;
1432 r32.rlim_cur = r.rlim_cur;
1433 if (r.rlim_max > COMPAT_RLIM_INFINITY)
1434 r32.rlim_max = COMPAT_RLIM_INFINITY;
1436 r32.rlim_max = r.rlim_max;
1438 if (copy_to_user(rlim, &r32, sizeof(struct compat_rlimit)))
1446 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1449 * Back compatibility for getrlimit. Needed for some apps.
1451 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1452 struct rlimit __user *, rlim)
1455 if (resource >= RLIM_NLIMITS)
1458 resource = array_index_nospec(resource, RLIM_NLIMITS);
1459 task_lock(current->group_leader);
1460 x = current->signal->rlim[resource];
1461 task_unlock(current->group_leader);
1462 if (x.rlim_cur > 0x7FFFFFFF)
1463 x.rlim_cur = 0x7FFFFFFF;
1464 if (x.rlim_max > 0x7FFFFFFF)
1465 x.rlim_max = 0x7FFFFFFF;
1466 return copy_to_user(rlim, &x, sizeof(x)) ? -EFAULT : 0;
1469 #ifdef CONFIG_COMPAT
1470 COMPAT_SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1471 struct compat_rlimit __user *, rlim)
1475 if (resource >= RLIM_NLIMITS)
1478 resource = array_index_nospec(resource, RLIM_NLIMITS);
1479 task_lock(current->group_leader);
1480 r = current->signal->rlim[resource];
1481 task_unlock(current->group_leader);
1482 if (r.rlim_cur > 0x7FFFFFFF)
1483 r.rlim_cur = 0x7FFFFFFF;
1484 if (r.rlim_max > 0x7FFFFFFF)
1485 r.rlim_max = 0x7FFFFFFF;
1487 if (put_user(r.rlim_cur, &rlim->rlim_cur) ||
1488 put_user(r.rlim_max, &rlim->rlim_max))
1496 static inline bool rlim64_is_infinity(__u64 rlim64)
1498 #if BITS_PER_LONG < 64
1499 return rlim64 >= ULONG_MAX;
1501 return rlim64 == RLIM64_INFINITY;
1505 static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1507 if (rlim->rlim_cur == RLIM_INFINITY)
1508 rlim64->rlim_cur = RLIM64_INFINITY;
1510 rlim64->rlim_cur = rlim->rlim_cur;
1511 if (rlim->rlim_max == RLIM_INFINITY)
1512 rlim64->rlim_max = RLIM64_INFINITY;
1514 rlim64->rlim_max = rlim->rlim_max;
1517 static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1519 if (rlim64_is_infinity(rlim64->rlim_cur))
1520 rlim->rlim_cur = RLIM_INFINITY;
1522 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1523 if (rlim64_is_infinity(rlim64->rlim_max))
1524 rlim->rlim_max = RLIM_INFINITY;
1526 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1529 /* make sure you are allowed to change @tsk limits before calling this */
1530 int do_prlimit(struct task_struct *tsk, unsigned int resource,
1531 struct rlimit *new_rlim, struct rlimit *old_rlim)
1533 struct rlimit *rlim;
1536 if (resource >= RLIM_NLIMITS)
1539 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1541 if (resource == RLIMIT_NOFILE &&
1542 new_rlim->rlim_max > sysctl_nr_open)
1546 /* protect tsk->signal and tsk->sighand from disappearing */
1547 read_lock(&tasklist_lock);
1548 if (!tsk->sighand) {
1553 rlim = tsk->signal->rlim + resource;
1554 task_lock(tsk->group_leader);
1556 /* Keep the capable check against init_user_ns until
1557 cgroups can contain all limits */
1558 if (new_rlim->rlim_max > rlim->rlim_max &&
1559 !capable(CAP_SYS_RESOURCE))
1562 retval = security_task_setrlimit(tsk, resource, new_rlim);
1570 task_unlock(tsk->group_leader);
1573 * RLIMIT_CPU handling. Arm the posix CPU timer if the limit is not
1574 * infite. In case of RLIM_INFINITY the posix CPU timer code
1575 * ignores the rlimit.
1577 if (!retval && new_rlim && resource == RLIMIT_CPU &&
1578 new_rlim->rlim_cur != RLIM_INFINITY &&
1579 IS_ENABLED(CONFIG_POSIX_TIMERS))
1580 update_rlimit_cpu(tsk, new_rlim->rlim_cur);
1582 read_unlock(&tasklist_lock);
1586 /* rcu lock must be held */
1587 static int check_prlimit_permission(struct task_struct *task,
1590 const struct cred *cred = current_cred(), *tcred;
1593 if (current == task)
1596 tcred = __task_cred(task);
1597 id_match = (uid_eq(cred->uid, tcred->euid) &&
1598 uid_eq(cred->uid, tcred->suid) &&
1599 uid_eq(cred->uid, tcred->uid) &&
1600 gid_eq(cred->gid, tcred->egid) &&
1601 gid_eq(cred->gid, tcred->sgid) &&
1602 gid_eq(cred->gid, tcred->gid));
1603 if (!id_match && !ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
1606 return security_task_prlimit(cred, tcred, flags);
1609 SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1610 const struct rlimit64 __user *, new_rlim,
1611 struct rlimit64 __user *, old_rlim)
1613 struct rlimit64 old64, new64;
1614 struct rlimit old, new;
1615 struct task_struct *tsk;
1616 unsigned int checkflags = 0;
1620 checkflags |= LSM_PRLIMIT_READ;
1623 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1625 rlim64_to_rlim(&new64, &new);
1626 checkflags |= LSM_PRLIMIT_WRITE;
1630 tsk = pid ? find_task_by_vpid(pid) : current;
1635 ret = check_prlimit_permission(tsk, checkflags);
1640 get_task_struct(tsk);
1643 ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1644 old_rlim ? &old : NULL);
1646 if (!ret && old_rlim) {
1647 rlim_to_rlim64(&old, &old64);
1648 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1652 put_task_struct(tsk);
1656 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1658 struct rlimit new_rlim;
1660 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1662 return do_prlimit(current, resource, &new_rlim, NULL);
1666 * It would make sense to put struct rusage in the task_struct,
1667 * except that would make the task_struct be *really big*. After
1668 * task_struct gets moved into malloc'ed memory, it would
1669 * make sense to do this. It will make moving the rest of the information
1670 * a lot simpler! (Which we're not doing right now because we're not
1671 * measuring them yet).
1673 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1674 * races with threads incrementing their own counters. But since word
1675 * reads are atomic, we either get new values or old values and we don't
1676 * care which for the sums. We always take the siglock to protect reading
1677 * the c* fields from p->signal from races with exit.c updating those
1678 * fields when reaping, so a sample either gets all the additions of a
1679 * given child after it's reaped, or none so this sample is before reaping.
1682 * We need to take the siglock for CHILDEREN, SELF and BOTH
1683 * for the cases current multithreaded, non-current single threaded
1684 * non-current multithreaded. Thread traversal is now safe with
1686 * Strictly speaking, we donot need to take the siglock if we are current and
1687 * single threaded, as no one else can take our signal_struct away, no one
1688 * else can reap the children to update signal->c* counters, and no one else
1689 * can race with the signal-> fields. If we do not take any lock, the
1690 * signal-> fields could be read out of order while another thread was just
1691 * exiting. So we should place a read memory barrier when we avoid the lock.
1692 * On the writer side, write memory barrier is implied in __exit_signal
1693 * as __exit_signal releases the siglock spinlock after updating the signal->
1694 * fields. But we don't do this yet to keep things simple.
1698 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1700 r->ru_nvcsw += t->nvcsw;
1701 r->ru_nivcsw += t->nivcsw;
1702 r->ru_minflt += t->min_flt;
1703 r->ru_majflt += t->maj_flt;
1704 r->ru_inblock += task_io_get_inblock(t);
1705 r->ru_oublock += task_io_get_oublock(t);
1708 void getrusage(struct task_struct *p, int who, struct rusage *r)
1710 struct task_struct *t;
1711 unsigned long flags;
1712 u64 tgutime, tgstime, utime, stime;
1713 unsigned long maxrss = 0;
1715 memset((char *)r, 0, sizeof (*r));
1718 if (who == RUSAGE_THREAD) {
1719 task_cputime_adjusted(current, &utime, &stime);
1720 accumulate_thread_rusage(p, r);
1721 maxrss = p->signal->maxrss;
1725 if (!lock_task_sighand(p, &flags))
1730 case RUSAGE_CHILDREN:
1731 utime = p->signal->cutime;
1732 stime = p->signal->cstime;
1733 r->ru_nvcsw = p->signal->cnvcsw;
1734 r->ru_nivcsw = p->signal->cnivcsw;
1735 r->ru_minflt = p->signal->cmin_flt;
1736 r->ru_majflt = p->signal->cmaj_flt;
1737 r->ru_inblock = p->signal->cinblock;
1738 r->ru_oublock = p->signal->coublock;
1739 maxrss = p->signal->cmaxrss;
1741 if (who == RUSAGE_CHILDREN)
1746 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1749 r->ru_nvcsw += p->signal->nvcsw;
1750 r->ru_nivcsw += p->signal->nivcsw;
1751 r->ru_minflt += p->signal->min_flt;
1752 r->ru_majflt += p->signal->maj_flt;
1753 r->ru_inblock += p->signal->inblock;
1754 r->ru_oublock += p->signal->oublock;
1755 if (maxrss < p->signal->maxrss)
1756 maxrss = p->signal->maxrss;
1759 accumulate_thread_rusage(t, r);
1760 } while_each_thread(p, t);
1766 unlock_task_sighand(p, &flags);
1769 r->ru_utime = ns_to_kernel_old_timeval(utime);
1770 r->ru_stime = ns_to_kernel_old_timeval(stime);
1772 if (who != RUSAGE_CHILDREN) {
1773 struct mm_struct *mm = get_task_mm(p);
1776 setmax_mm_hiwater_rss(&maxrss, mm);
1780 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1783 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1787 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1788 who != RUSAGE_THREAD)
1791 getrusage(current, who, &r);
1792 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1795 #ifdef CONFIG_COMPAT
1796 COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru)
1800 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1801 who != RUSAGE_THREAD)
1804 getrusage(current, who, &r);
1805 return put_compat_rusage(&r, ru);
1809 SYSCALL_DEFINE1(umask, int, mask)
1811 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1815 static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1818 struct file *old_exe, *exe_file;
1819 struct inode *inode;
1826 inode = file_inode(exe.file);
1829 * Because the original mm->exe_file points to executable file, make
1830 * sure that this one is executable as well, to avoid breaking an
1834 if (!S_ISREG(inode->i_mode) || path_noexec(&exe.file->f_path))
1837 err = inode_permission(inode, MAY_EXEC);
1842 * Forbid mm->exe_file change if old file still mapped.
1844 exe_file = get_mm_exe_file(mm);
1847 struct vm_area_struct *vma;
1849 down_read(&mm->mmap_sem);
1850 for (vma = mm->mmap; vma; vma = vma->vm_next) {
1853 if (path_equal(&vma->vm_file->f_path,
1858 up_read(&mm->mmap_sem);
1863 /* set the new file, lockless */
1865 old_exe = xchg(&mm->exe_file, exe.file);
1872 up_read(&mm->mmap_sem);
1878 * Check arithmetic relations of passed addresses.
1880 * WARNING: we don't require any capability here so be very careful
1881 * in what is allowed for modification from userspace.
1883 static int validate_prctl_map_addr(struct prctl_mm_map *prctl_map)
1885 unsigned long mmap_max_addr = TASK_SIZE;
1886 int error = -EINVAL, i;
1888 static const unsigned char offsets[] = {
1889 offsetof(struct prctl_mm_map, start_code),
1890 offsetof(struct prctl_mm_map, end_code),
1891 offsetof(struct prctl_mm_map, start_data),
1892 offsetof(struct prctl_mm_map, end_data),
1893 offsetof(struct prctl_mm_map, start_brk),
1894 offsetof(struct prctl_mm_map, brk),
1895 offsetof(struct prctl_mm_map, start_stack),
1896 offsetof(struct prctl_mm_map, arg_start),
1897 offsetof(struct prctl_mm_map, arg_end),
1898 offsetof(struct prctl_mm_map, env_start),
1899 offsetof(struct prctl_mm_map, env_end),
1903 * Make sure the members are not somewhere outside
1904 * of allowed address space.
1906 for (i = 0; i < ARRAY_SIZE(offsets); i++) {
1907 u64 val = *(u64 *)((char *)prctl_map + offsets[i]);
1909 if ((unsigned long)val >= mmap_max_addr ||
1910 (unsigned long)val < mmap_min_addr)
1915 * Make sure the pairs are ordered.
1917 #define __prctl_check_order(__m1, __op, __m2) \
1918 ((unsigned long)prctl_map->__m1 __op \
1919 (unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1920 error = __prctl_check_order(start_code, <, end_code);
1921 error |= __prctl_check_order(start_data,<=, end_data);
1922 error |= __prctl_check_order(start_brk, <=, brk);
1923 error |= __prctl_check_order(arg_start, <=, arg_end);
1924 error |= __prctl_check_order(env_start, <=, env_end);
1927 #undef __prctl_check_order
1932 * @brk should be after @end_data in traditional maps.
1934 if (prctl_map->start_brk <= prctl_map->end_data ||
1935 prctl_map->brk <= prctl_map->end_data)
1939 * Neither we should allow to override limits if they set.
1941 if (check_data_rlimit(rlimit(RLIMIT_DATA), prctl_map->brk,
1942 prctl_map->start_brk, prctl_map->end_data,
1943 prctl_map->start_data))
1951 #ifdef CONFIG_CHECKPOINT_RESTORE
1952 static int prctl_set_mm_map(int opt, const void __user *addr, unsigned long data_size)
1954 struct prctl_mm_map prctl_map = { .exe_fd = (u32)-1, };
1955 unsigned long user_auxv[AT_VECTOR_SIZE];
1956 struct mm_struct *mm = current->mm;
1959 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
1960 BUILD_BUG_ON(sizeof(struct prctl_mm_map) > 256);
1962 if (opt == PR_SET_MM_MAP_SIZE)
1963 return put_user((unsigned int)sizeof(prctl_map),
1964 (unsigned int __user *)addr);
1966 if (data_size != sizeof(prctl_map))
1969 if (copy_from_user(&prctl_map, addr, sizeof(prctl_map)))
1972 error = validate_prctl_map_addr(&prctl_map);
1976 if (prctl_map.auxv_size) {
1978 * Someone is trying to cheat the auxv vector.
1980 if (!prctl_map.auxv ||
1981 prctl_map.auxv_size > sizeof(mm->saved_auxv))
1984 memset(user_auxv, 0, sizeof(user_auxv));
1985 if (copy_from_user(user_auxv,
1986 (const void __user *)prctl_map.auxv,
1987 prctl_map.auxv_size))
1990 /* Last entry must be AT_NULL as specification requires */
1991 user_auxv[AT_VECTOR_SIZE - 2] = AT_NULL;
1992 user_auxv[AT_VECTOR_SIZE - 1] = AT_NULL;
1995 if (prctl_map.exe_fd != (u32)-1) {
1997 * Make sure the caller has the rights to
1998 * change /proc/pid/exe link: only local sys admin should
2001 if (!ns_capable(current_user_ns(), CAP_SYS_ADMIN))
2004 error = prctl_set_mm_exe_file(mm, prctl_map.exe_fd);
2010 * arg_lock protects concurent updates but we still need mmap_sem for
2011 * read to exclude races with sys_brk.
2013 down_read(&mm->mmap_sem);
2016 * We don't validate if these members are pointing to
2017 * real present VMAs because application may have correspond
2018 * VMAs already unmapped and kernel uses these members for statistics
2019 * output in procfs mostly, except
2021 * - @start_brk/@brk which are used in do_brk but kernel lookups
2022 * for VMAs when updating these memvers so anything wrong written
2023 * here cause kernel to swear at userspace program but won't lead
2024 * to any problem in kernel itself
2027 spin_lock(&mm->arg_lock);
2028 mm->start_code = prctl_map.start_code;
2029 mm->end_code = prctl_map.end_code;
2030 mm->start_data = prctl_map.start_data;
2031 mm->end_data = prctl_map.end_data;
2032 mm->start_brk = prctl_map.start_brk;
2033 mm->brk = prctl_map.brk;
2034 mm->start_stack = prctl_map.start_stack;
2035 mm->arg_start = prctl_map.arg_start;
2036 mm->arg_end = prctl_map.arg_end;
2037 mm->env_start = prctl_map.env_start;
2038 mm->env_end = prctl_map.env_end;
2039 spin_unlock(&mm->arg_lock);
2042 * Note this update of @saved_auxv is lockless thus
2043 * if someone reads this member in procfs while we're
2044 * updating -- it may get partly updated results. It's
2045 * known and acceptable trade off: we leave it as is to
2046 * not introduce additional locks here making the kernel
2049 if (prctl_map.auxv_size)
2050 memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv));
2052 up_read(&mm->mmap_sem);
2055 #endif /* CONFIG_CHECKPOINT_RESTORE */
2057 static int prctl_set_auxv(struct mm_struct *mm, unsigned long addr,
2061 * This doesn't move the auxiliary vector itself since it's pinned to
2062 * mm_struct, but it permits filling the vector with new values. It's
2063 * up to the caller to provide sane values here, otherwise userspace
2064 * tools which use this vector might be unhappy.
2066 unsigned long user_auxv[AT_VECTOR_SIZE];
2068 if (len > sizeof(user_auxv))
2071 if (copy_from_user(user_auxv, (const void __user *)addr, len))
2074 /* Make sure the last entry is always AT_NULL */
2075 user_auxv[AT_VECTOR_SIZE - 2] = 0;
2076 user_auxv[AT_VECTOR_SIZE - 1] = 0;
2078 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
2081 memcpy(mm->saved_auxv, user_auxv, len);
2082 task_unlock(current);
2087 static int prctl_set_mm(int opt, unsigned long addr,
2088 unsigned long arg4, unsigned long arg5)
2090 struct mm_struct *mm = current->mm;
2091 struct prctl_mm_map prctl_map = {
2096 struct vm_area_struct *vma;
2099 if (arg5 || (arg4 && (opt != PR_SET_MM_AUXV &&
2100 opt != PR_SET_MM_MAP &&
2101 opt != PR_SET_MM_MAP_SIZE)))
2104 #ifdef CONFIG_CHECKPOINT_RESTORE
2105 if (opt == PR_SET_MM_MAP || opt == PR_SET_MM_MAP_SIZE)
2106 return prctl_set_mm_map(opt, (const void __user *)addr, arg4);
2109 if (!capable(CAP_SYS_RESOURCE))
2112 if (opt == PR_SET_MM_EXE_FILE)
2113 return prctl_set_mm_exe_file(mm, (unsigned int)addr);
2115 if (opt == PR_SET_MM_AUXV)
2116 return prctl_set_auxv(mm, addr, arg4);
2118 if (addr >= TASK_SIZE || addr < mmap_min_addr)
2124 * arg_lock protects concurent updates of arg boundaries, we need
2125 * mmap_sem for a) concurrent sys_brk, b) finding VMA for addr
2128 down_read(&mm->mmap_sem);
2129 vma = find_vma(mm, addr);
2131 spin_lock(&mm->arg_lock);
2132 prctl_map.start_code = mm->start_code;
2133 prctl_map.end_code = mm->end_code;
2134 prctl_map.start_data = mm->start_data;
2135 prctl_map.end_data = mm->end_data;
2136 prctl_map.start_brk = mm->start_brk;
2137 prctl_map.brk = mm->brk;
2138 prctl_map.start_stack = mm->start_stack;
2139 prctl_map.arg_start = mm->arg_start;
2140 prctl_map.arg_end = mm->arg_end;
2141 prctl_map.env_start = mm->env_start;
2142 prctl_map.env_end = mm->env_end;
2145 case PR_SET_MM_START_CODE:
2146 prctl_map.start_code = addr;
2148 case PR_SET_MM_END_CODE:
2149 prctl_map.end_code = addr;
2151 case PR_SET_MM_START_DATA:
2152 prctl_map.start_data = addr;
2154 case PR_SET_MM_END_DATA:
2155 prctl_map.end_data = addr;
2157 case PR_SET_MM_START_STACK:
2158 prctl_map.start_stack = addr;
2160 case PR_SET_MM_START_BRK:
2161 prctl_map.start_brk = addr;
2164 prctl_map.brk = addr;
2166 case PR_SET_MM_ARG_START:
2167 prctl_map.arg_start = addr;
2169 case PR_SET_MM_ARG_END:
2170 prctl_map.arg_end = addr;
2172 case PR_SET_MM_ENV_START:
2173 prctl_map.env_start = addr;
2175 case PR_SET_MM_ENV_END:
2176 prctl_map.env_end = addr;
2182 error = validate_prctl_map_addr(&prctl_map);
2188 * If command line arguments and environment
2189 * are placed somewhere else on stack, we can
2190 * set them up here, ARG_START/END to setup
2191 * command line argumets and ENV_START/END
2194 case PR_SET_MM_START_STACK:
2195 case PR_SET_MM_ARG_START:
2196 case PR_SET_MM_ARG_END:
2197 case PR_SET_MM_ENV_START:
2198 case PR_SET_MM_ENV_END:
2205 mm->start_code = prctl_map.start_code;
2206 mm->end_code = prctl_map.end_code;
2207 mm->start_data = prctl_map.start_data;
2208 mm->end_data = prctl_map.end_data;
2209 mm->start_brk = prctl_map.start_brk;
2210 mm->brk = prctl_map.brk;
2211 mm->start_stack = prctl_map.start_stack;
2212 mm->arg_start = prctl_map.arg_start;
2213 mm->arg_end = prctl_map.arg_end;
2214 mm->env_start = prctl_map.env_start;
2215 mm->env_end = prctl_map.env_end;
2219 spin_unlock(&mm->arg_lock);
2220 up_read(&mm->mmap_sem);
2224 #ifdef CONFIG_CHECKPOINT_RESTORE
2225 static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2227 return put_user(me->clear_child_tid, tid_addr);
2230 static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2236 static int propagate_has_child_subreaper(struct task_struct *p, void *data)
2239 * If task has has_child_subreaper - all its decendants
2240 * already have these flag too and new decendants will
2241 * inherit it on fork, skip them.
2243 * If we've found child_reaper - skip descendants in
2244 * it's subtree as they will never get out pidns.
2246 if (p->signal->has_child_subreaper ||
2247 is_child_reaper(task_pid(p)))
2250 p->signal->has_child_subreaper = 1;
2254 int __weak arch_prctl_spec_ctrl_get(struct task_struct *t, unsigned long which)
2259 int __weak arch_prctl_spec_ctrl_set(struct task_struct *t, unsigned long which,
2265 #define PR_IO_FLUSHER (PF_MEMALLOC_NOIO | PF_LOCAL_THROTTLE)
2267 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2268 unsigned long, arg4, unsigned long, arg5)
2270 struct task_struct *me = current;
2271 unsigned char comm[sizeof(me->comm)];
2274 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2275 if (error != -ENOSYS)
2280 case PR_SET_PDEATHSIG:
2281 if (!valid_signal(arg2)) {
2285 me->pdeath_signal = arg2;
2287 case PR_GET_PDEATHSIG:
2288 error = put_user(me->pdeath_signal, (int __user *)arg2);
2290 case PR_GET_DUMPABLE:
2291 error = get_dumpable(me->mm);
2293 case PR_SET_DUMPABLE:
2294 if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) {
2298 set_dumpable(me->mm, arg2);
2301 case PR_SET_UNALIGN:
2302 error = SET_UNALIGN_CTL(me, arg2);
2304 case PR_GET_UNALIGN:
2305 error = GET_UNALIGN_CTL(me, arg2);
2308 error = SET_FPEMU_CTL(me, arg2);
2311 error = GET_FPEMU_CTL(me, arg2);
2314 error = SET_FPEXC_CTL(me, arg2);
2317 error = GET_FPEXC_CTL(me, arg2);
2320 error = PR_TIMING_STATISTICAL;
2323 if (arg2 != PR_TIMING_STATISTICAL)
2327 comm[sizeof(me->comm) - 1] = 0;
2328 if (strncpy_from_user(comm, (char __user *)arg2,
2329 sizeof(me->comm) - 1) < 0)
2331 set_task_comm(me, comm);
2332 proc_comm_connector(me);
2335 get_task_comm(comm, me);
2336 if (copy_to_user((char __user *)arg2, comm, sizeof(comm)))
2340 error = GET_ENDIAN(me, arg2);
2343 error = SET_ENDIAN(me, arg2);
2345 case PR_GET_SECCOMP:
2346 error = prctl_get_seccomp();
2348 case PR_SET_SECCOMP:
2349 error = prctl_set_seccomp(arg2, (char __user *)arg3);
2352 error = GET_TSC_CTL(arg2);
2355 error = SET_TSC_CTL(arg2);
2357 case PR_TASK_PERF_EVENTS_DISABLE:
2358 error = perf_event_task_disable();
2360 case PR_TASK_PERF_EVENTS_ENABLE:
2361 error = perf_event_task_enable();
2363 case PR_GET_TIMERSLACK:
2364 if (current->timer_slack_ns > ULONG_MAX)
2367 error = current->timer_slack_ns;
2369 case PR_SET_TIMERSLACK:
2371 current->timer_slack_ns =
2372 current->default_timer_slack_ns;
2374 current->timer_slack_ns = arg2;
2380 case PR_MCE_KILL_CLEAR:
2383 current->flags &= ~PF_MCE_PROCESS;
2385 case PR_MCE_KILL_SET:
2386 current->flags |= PF_MCE_PROCESS;
2387 if (arg3 == PR_MCE_KILL_EARLY)
2388 current->flags |= PF_MCE_EARLY;
2389 else if (arg3 == PR_MCE_KILL_LATE)
2390 current->flags &= ~PF_MCE_EARLY;
2391 else if (arg3 == PR_MCE_KILL_DEFAULT)
2393 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
2401 case PR_MCE_KILL_GET:
2402 if (arg2 | arg3 | arg4 | arg5)
2404 if (current->flags & PF_MCE_PROCESS)
2405 error = (current->flags & PF_MCE_EARLY) ?
2406 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2408 error = PR_MCE_KILL_DEFAULT;
2411 error = prctl_set_mm(arg2, arg3, arg4, arg5);
2413 case PR_GET_TID_ADDRESS:
2414 error = prctl_get_tid_address(me, (int __user **)arg2);
2416 case PR_SET_CHILD_SUBREAPER:
2417 me->signal->is_child_subreaper = !!arg2;
2421 walk_process_tree(me, propagate_has_child_subreaper, NULL);
2423 case PR_GET_CHILD_SUBREAPER:
2424 error = put_user(me->signal->is_child_subreaper,
2425 (int __user *)arg2);
2427 case PR_SET_NO_NEW_PRIVS:
2428 if (arg2 != 1 || arg3 || arg4 || arg5)
2431 task_set_no_new_privs(current);
2433 case PR_GET_NO_NEW_PRIVS:
2434 if (arg2 || arg3 || arg4 || arg5)
2436 return task_no_new_privs(current) ? 1 : 0;
2437 case PR_GET_THP_DISABLE:
2438 if (arg2 || arg3 || arg4 || arg5)
2440 error = !!test_bit(MMF_DISABLE_THP, &me->mm->flags);
2442 case PR_SET_THP_DISABLE:
2443 if (arg3 || arg4 || arg5)
2445 if (down_write_killable(&me->mm->mmap_sem))
2448 set_bit(MMF_DISABLE_THP, &me->mm->flags);
2450 clear_bit(MMF_DISABLE_THP, &me->mm->flags);
2451 up_write(&me->mm->mmap_sem);
2453 case PR_MPX_ENABLE_MANAGEMENT:
2454 case PR_MPX_DISABLE_MANAGEMENT:
2455 /* No longer implemented: */
2457 case PR_SET_FP_MODE:
2458 error = SET_FP_MODE(me, arg2);
2460 case PR_GET_FP_MODE:
2461 error = GET_FP_MODE(me);
2464 error = SVE_SET_VL(arg2);
2467 error = SVE_GET_VL();
2469 case PR_GET_SPECULATION_CTRL:
2470 if (arg3 || arg4 || arg5)
2472 error = arch_prctl_spec_ctrl_get(me, arg2);
2474 case PR_SET_SPECULATION_CTRL:
2477 error = arch_prctl_spec_ctrl_set(me, arg2, arg3);
2479 case PR_PAC_RESET_KEYS:
2480 if (arg3 || arg4 || arg5)
2482 error = PAC_RESET_KEYS(me, arg2);
2484 case PR_SET_TAGGED_ADDR_CTRL:
2485 if (arg3 || arg4 || arg5)
2487 error = SET_TAGGED_ADDR_CTRL(arg2);
2489 case PR_GET_TAGGED_ADDR_CTRL:
2490 if (arg2 || arg3 || arg4 || arg5)
2492 error = GET_TAGGED_ADDR_CTRL();
2494 case PR_SET_IO_FLUSHER:
2495 if (!capable(CAP_SYS_RESOURCE))
2498 if (arg3 || arg4 || arg5)
2502 current->flags |= PR_IO_FLUSHER;
2504 current->flags &= ~PR_IO_FLUSHER;
2508 case PR_GET_IO_FLUSHER:
2509 if (!capable(CAP_SYS_RESOURCE))
2512 if (arg2 || arg3 || arg4 || arg5)
2515 error = (current->flags & PR_IO_FLUSHER) == PR_IO_FLUSHER;
2524 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2525 struct getcpu_cache __user *, unused)
2528 int cpu = raw_smp_processor_id();
2531 err |= put_user(cpu, cpup);
2533 err |= put_user(cpu_to_node(cpu), nodep);
2534 return err ? -EFAULT : 0;
2538 * do_sysinfo - fill in sysinfo struct
2539 * @info: pointer to buffer to fill
2541 static int do_sysinfo(struct sysinfo *info)
2543 unsigned long mem_total, sav_total;
2544 unsigned int mem_unit, bitcount;
2545 struct timespec64 tp;
2547 memset(info, 0, sizeof(struct sysinfo));
2549 ktime_get_boottime_ts64(&tp);
2550 timens_add_boottime(&tp);
2551 info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
2553 get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
2555 info->procs = nr_threads;
2561 * If the sum of all the available memory (i.e. ram + swap)
2562 * is less than can be stored in a 32 bit unsigned long then
2563 * we can be binary compatible with 2.2.x kernels. If not,
2564 * well, in that case 2.2.x was broken anyways...
2566 * -Erik Andersen <andersee@debian.org>
2569 mem_total = info->totalram + info->totalswap;
2570 if (mem_total < info->totalram || mem_total < info->totalswap)
2573 mem_unit = info->mem_unit;
2574 while (mem_unit > 1) {
2577 sav_total = mem_total;
2579 if (mem_total < sav_total)
2584 * If mem_total did not overflow, multiply all memory values by
2585 * info->mem_unit and set it to 1. This leaves things compatible
2586 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2591 info->totalram <<= bitcount;
2592 info->freeram <<= bitcount;
2593 info->sharedram <<= bitcount;
2594 info->bufferram <<= bitcount;
2595 info->totalswap <<= bitcount;
2596 info->freeswap <<= bitcount;
2597 info->totalhigh <<= bitcount;
2598 info->freehigh <<= bitcount;
2604 SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
2610 if (copy_to_user(info, &val, sizeof(struct sysinfo)))
2616 #ifdef CONFIG_COMPAT
2617 struct compat_sysinfo {
2631 char _f[20-2*sizeof(u32)-sizeof(int)];
2634 COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2637 struct compat_sysinfo s_32;
2641 /* Check to see if any memory value is too large for 32-bit and scale
2644 if (upper_32_bits(s.totalram) || upper_32_bits(s.totalswap)) {
2647 while (s.mem_unit < PAGE_SIZE) {
2652 s.totalram >>= bitcount;
2653 s.freeram >>= bitcount;
2654 s.sharedram >>= bitcount;
2655 s.bufferram >>= bitcount;
2656 s.totalswap >>= bitcount;
2657 s.freeswap >>= bitcount;
2658 s.totalhigh >>= bitcount;
2659 s.freehigh >>= bitcount;
2662 memset(&s_32, 0, sizeof(s_32));
2663 s_32.uptime = s.uptime;
2664 s_32.loads[0] = s.loads[0];
2665 s_32.loads[1] = s.loads[1];
2666 s_32.loads[2] = s.loads[2];
2667 s_32.totalram = s.totalram;
2668 s_32.freeram = s.freeram;
2669 s_32.sharedram = s.sharedram;
2670 s_32.bufferram = s.bufferram;
2671 s_32.totalswap = s.totalswap;
2672 s_32.freeswap = s.freeswap;
2673 s_32.procs = s.procs;
2674 s_32.totalhigh = s.totalhigh;
2675 s_32.freehigh = s.freehigh;
2676 s_32.mem_unit = s.mem_unit;
2677 if (copy_to_user(info, &s_32, sizeof(s_32)))
2681 #endif /* CONFIG_COMPAT */