4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/export.h>
9 #include <linux/utsname.h>
10 #include <linux/mman.h>
11 #include <linux/reboot.h>
12 #include <linux/prctl.h>
13 #include <linux/highuid.h>
15 #include <linux/kmod.h>
16 #include <linux/perf_event.h>
17 #include <linux/resource.h>
18 #include <linux/kernel.h>
19 #include <linux/workqueue.h>
20 #include <linux/capability.h>
21 #include <linux/device.h>
22 #include <linux/key.h>
23 #include <linux/times.h>
24 #include <linux/posix-timers.h>
25 #include <linux/security.h>
26 #include <linux/dcookies.h>
27 #include <linux/suspend.h>
28 #include <linux/tty.h>
29 #include <linux/signal.h>
30 #include <linux/cn_proc.h>
31 #include <linux/getcpu.h>
32 #include <linux/task_io_accounting_ops.h>
33 #include <linux/seccomp.h>
34 #include <linux/cpu.h>
35 #include <linux/personality.h>
36 #include <linux/ptrace.h>
37 #include <linux/fs_struct.h>
38 #include <linux/file.h>
39 #include <linux/mount.h>
40 #include <linux/gfp.h>
41 #include <linux/syscore_ops.h>
42 #include <linux/version.h>
43 #include <linux/ctype.h>
45 #include <linux/compat.h>
46 #include <linux/syscalls.h>
47 #include <linux/kprobes.h>
48 #include <linux/user_namespace.h>
49 #include <linux/binfmts.h>
51 #include <linux/sched.h>
52 #include <linux/rcupdate.h>
53 #include <linux/uidgid.h>
54 #include <linux/cred.h>
56 #include <linux/nospec.h>
58 #include <linux/kmsg_dump.h>
59 /* Move somewhere else to avoid recompiling? */
60 #include <generated/utsrelease.h>
62 #include <asm/uaccess.h>
64 #include <asm/unistd.h>
66 /* Hardening for Spectre-v1 */
67 #include <linux/nospec.h>
69 #ifndef SET_UNALIGN_CTL
70 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
72 #ifndef GET_UNALIGN_CTL
73 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
76 # define SET_FPEMU_CTL(a,b) (-EINVAL)
79 # define GET_FPEMU_CTL(a,b) (-EINVAL)
82 # define SET_FPEXC_CTL(a,b) (-EINVAL)
85 # define GET_FPEXC_CTL(a,b) (-EINVAL)
88 # define GET_ENDIAN(a,b) (-EINVAL)
91 # define SET_ENDIAN(a,b) (-EINVAL)
94 # define GET_TSC_CTL(a) (-EINVAL)
97 # define SET_TSC_CTL(a) (-EINVAL)
101 * this is where the system-wide overflow UID and GID are defined, for
102 * architectures that now have 32-bit UID/GID but didn't in the past
105 int overflowuid
= DEFAULT_OVERFLOWUID
;
106 int overflowgid
= DEFAULT_OVERFLOWGID
;
108 EXPORT_SYMBOL(overflowuid
);
109 EXPORT_SYMBOL(overflowgid
);
112 * the same as above, but for filesystems which can only store a 16-bit
113 * UID and GID. as such, this is needed on all architectures
116 int fs_overflowuid
= DEFAULT_FS_OVERFLOWUID
;
117 int fs_overflowgid
= DEFAULT_FS_OVERFLOWUID
;
119 EXPORT_SYMBOL(fs_overflowuid
);
120 EXPORT_SYMBOL(fs_overflowgid
);
123 * Returns true if current's euid is same as p's uid or euid,
124 * or has CAP_SYS_NICE to p's user_ns.
126 * Called with rcu_read_lock, creds are safe
128 static bool set_one_prio_perm(struct task_struct
*p
)
130 const struct cred
*cred
= current_cred(), *pcred
= __task_cred(p
);
132 if (uid_eq(pcred
->uid
, cred
->euid
) ||
133 uid_eq(pcred
->euid
, cred
->euid
))
135 if (ns_capable(pcred
->user_ns
, CAP_SYS_NICE
))
141 * set the priority of a task
142 * - the caller must hold the RCU read lock
144 static int set_one_prio(struct task_struct
*p
, int niceval
, int error
)
148 if (!set_one_prio_perm(p
)) {
152 if (niceval
< task_nice(p
) && !can_nice(p
, niceval
)) {
156 no_nice
= security_task_setnice(p
, niceval
);
163 set_user_nice(p
, niceval
);
168 SYSCALL_DEFINE3(setpriority
, int, which
, int, who
, int, niceval
)
170 struct task_struct
*g
, *p
;
171 struct user_struct
*user
;
172 const struct cred
*cred
= current_cred();
177 if (which
> PRIO_USER
|| which
< PRIO_PROCESS
)
180 /* normalize: avoid signed division (rounding problems) */
182 if (niceval
< MIN_NICE
)
184 if (niceval
> MAX_NICE
)
188 read_lock(&tasklist_lock
);
192 p
= find_task_by_vpid(who
);
196 error
= set_one_prio(p
, niceval
, error
);
200 pgrp
= find_vpid(who
);
202 pgrp
= task_pgrp(current
);
203 do_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
) {
204 error
= set_one_prio(p
, niceval
, error
);
205 } while_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
);
208 uid
= make_kuid(cred
->user_ns
, who
);
212 else if (!uid_eq(uid
, cred
->uid
) &&
213 !(user
= find_user(uid
)))
214 goto out_unlock
; /* No processes for this user */
216 do_each_thread(g
, p
) {
217 if (uid_eq(task_uid(p
), uid
))
218 error
= set_one_prio(p
, niceval
, error
);
219 } while_each_thread(g
, p
);
220 if (!uid_eq(uid
, cred
->uid
))
221 free_uid(user
); /* For find_user() */
225 read_unlock(&tasklist_lock
);
232 * Ugh. To avoid negative return values, "getpriority()" will
233 * not return the normal nice-value, but a negated value that
234 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
235 * to stay compatible.
237 SYSCALL_DEFINE2(getpriority
, int, which
, int, who
)
239 struct task_struct
*g
, *p
;
240 struct user_struct
*user
;
241 const struct cred
*cred
= current_cred();
242 long niceval
, retval
= -ESRCH
;
246 if (which
> PRIO_USER
|| which
< PRIO_PROCESS
)
250 read_lock(&tasklist_lock
);
254 p
= find_task_by_vpid(who
);
258 niceval
= nice_to_rlimit(task_nice(p
));
259 if (niceval
> retval
)
265 pgrp
= find_vpid(who
);
267 pgrp
= task_pgrp(current
);
268 do_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
) {
269 niceval
= nice_to_rlimit(task_nice(p
));
270 if (niceval
> retval
)
272 } while_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
);
275 uid
= make_kuid(cred
->user_ns
, who
);
279 else if (!uid_eq(uid
, cred
->uid
) &&
280 !(user
= find_user(uid
)))
281 goto out_unlock
; /* No processes for this user */
283 do_each_thread(g
, p
) {
284 if (uid_eq(task_uid(p
), uid
)) {
285 niceval
= nice_to_rlimit(task_nice(p
));
286 if (niceval
> retval
)
289 } while_each_thread(g
, p
);
290 if (!uid_eq(uid
, cred
->uid
))
291 free_uid(user
); /* for find_user() */
295 read_unlock(&tasklist_lock
);
302 * Unprivileged users may change the real gid to the effective gid
303 * or vice versa. (BSD-style)
305 * If you set the real gid at all, or set the effective gid to a value not
306 * equal to the real gid, then the saved gid is set to the new effective gid.
308 * This makes it possible for a setgid program to completely drop its
309 * privileges, which is often a useful assertion to make when you are doing
310 * a security audit over a program.
312 * The general idea is that a program which uses just setregid() will be
313 * 100% compatible with BSD. A program which uses just setgid() will be
314 * 100% compatible with POSIX with saved IDs.
316 * SMP: There are not races, the GIDs are checked only by filesystem
317 * operations (as far as semantic preservation is concerned).
319 SYSCALL_DEFINE2(setregid
, gid_t
, rgid
, gid_t
, egid
)
321 struct user_namespace
*ns
= current_user_ns();
322 const struct cred
*old
;
327 krgid
= make_kgid(ns
, rgid
);
328 kegid
= make_kgid(ns
, egid
);
330 if ((rgid
!= (gid_t
) -1) && !gid_valid(krgid
))
332 if ((egid
!= (gid_t
) -1) && !gid_valid(kegid
))
335 new = prepare_creds();
338 old
= current_cred();
341 if (rgid
!= (gid_t
) -1) {
342 if (gid_eq(old
->gid
, krgid
) ||
343 gid_eq(old
->egid
, krgid
) ||
344 ns_capable(old
->user_ns
, CAP_SETGID
))
349 if (egid
!= (gid_t
) -1) {
350 if (gid_eq(old
->gid
, kegid
) ||
351 gid_eq(old
->egid
, kegid
) ||
352 gid_eq(old
->sgid
, kegid
) ||
353 ns_capable(old
->user_ns
, CAP_SETGID
))
359 if (rgid
!= (gid_t
) -1 ||
360 (egid
!= (gid_t
) -1 && !gid_eq(kegid
, old
->gid
)))
361 new->sgid
= new->egid
;
362 new->fsgid
= new->egid
;
364 return commit_creds(new);
372 * setgid() is implemented like SysV w/ SAVED_IDS
374 * SMP: Same implicit races as above.
376 SYSCALL_DEFINE1(setgid
, gid_t
, gid
)
378 struct user_namespace
*ns
= current_user_ns();
379 const struct cred
*old
;
384 kgid
= make_kgid(ns
, gid
);
385 if (!gid_valid(kgid
))
388 new = prepare_creds();
391 old
= current_cred();
394 if (ns_capable(old
->user_ns
, CAP_SETGID
))
395 new->gid
= new->egid
= new->sgid
= new->fsgid
= kgid
;
396 else if (gid_eq(kgid
, old
->gid
) || gid_eq(kgid
, old
->sgid
))
397 new->egid
= new->fsgid
= kgid
;
401 return commit_creds(new);
409 * change the user struct in a credentials set to match the new UID
411 static int set_user(struct cred
*new)
413 struct user_struct
*new_user
;
415 new_user
= alloc_uid(new->uid
);
420 * We don't fail in case of NPROC limit excess here because too many
421 * poorly written programs don't check set*uid() return code, assuming
422 * it never fails if called by root. We may still enforce NPROC limit
423 * for programs doing set*uid()+execve() by harmlessly deferring the
424 * failure to the execve() stage.
426 if (atomic_read(&new_user
->processes
) >= rlimit(RLIMIT_NPROC
) &&
427 new_user
!= INIT_USER
)
428 current
->flags
|= PF_NPROC_EXCEEDED
;
430 current
->flags
&= ~PF_NPROC_EXCEEDED
;
433 new->user
= new_user
;
438 * Unprivileged users may change the real uid to the effective uid
439 * or vice versa. (BSD-style)
441 * If you set the real uid at all, or set the effective uid to a value not
442 * equal to the real uid, then the saved uid is set to the new effective uid.
444 * This makes it possible for a setuid program to completely drop its
445 * privileges, which is often a useful assertion to make when you are doing
446 * a security audit over a program.
448 * The general idea is that a program which uses just setreuid() will be
449 * 100% compatible with BSD. A program which uses just setuid() will be
450 * 100% compatible with POSIX with saved IDs.
452 SYSCALL_DEFINE2(setreuid
, uid_t
, ruid
, uid_t
, euid
)
454 struct user_namespace
*ns
= current_user_ns();
455 const struct cred
*old
;
460 kruid
= make_kuid(ns
, ruid
);
461 keuid
= make_kuid(ns
, euid
);
463 if ((ruid
!= (uid_t
) -1) && !uid_valid(kruid
))
465 if ((euid
!= (uid_t
) -1) && !uid_valid(keuid
))
468 new = prepare_creds();
471 old
= current_cred();
474 if (ruid
!= (uid_t
) -1) {
476 if (!uid_eq(old
->uid
, kruid
) &&
477 !uid_eq(old
->euid
, kruid
) &&
478 !ns_capable(old
->user_ns
, CAP_SETUID
))
482 if (euid
!= (uid_t
) -1) {
484 if (!uid_eq(old
->uid
, keuid
) &&
485 !uid_eq(old
->euid
, keuid
) &&
486 !uid_eq(old
->suid
, keuid
) &&
487 !ns_capable(old
->user_ns
, CAP_SETUID
))
491 if (!uid_eq(new->uid
, old
->uid
)) {
492 retval
= set_user(new);
496 if (ruid
!= (uid_t
) -1 ||
497 (euid
!= (uid_t
) -1 && !uid_eq(keuid
, old
->uid
)))
498 new->suid
= new->euid
;
499 new->fsuid
= new->euid
;
501 retval
= security_task_fix_setuid(new, old
, LSM_SETID_RE
);
505 return commit_creds(new);
513 * setuid() is implemented like SysV with SAVED_IDS
515 * Note that SAVED_ID's is deficient in that a setuid root program
516 * like sendmail, for example, cannot set its uid to be a normal
517 * user and then switch back, because if you're root, setuid() sets
518 * the saved uid too. If you don't like this, blame the bright people
519 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
520 * will allow a root program to temporarily drop privileges and be able to
521 * regain them by swapping the real and effective uid.
523 SYSCALL_DEFINE1(setuid
, uid_t
, uid
)
525 struct user_namespace
*ns
= current_user_ns();
526 const struct cred
*old
;
531 kuid
= make_kuid(ns
, uid
);
532 if (!uid_valid(kuid
))
535 new = prepare_creds();
538 old
= current_cred();
541 if (ns_capable(old
->user_ns
, CAP_SETUID
)) {
542 new->suid
= new->uid
= kuid
;
543 if (!uid_eq(kuid
, old
->uid
)) {
544 retval
= set_user(new);
548 } else if (!uid_eq(kuid
, old
->uid
) && !uid_eq(kuid
, new->suid
)) {
552 new->fsuid
= new->euid
= kuid
;
554 retval
= security_task_fix_setuid(new, old
, LSM_SETID_ID
);
558 return commit_creds(new);
567 * This function implements a generic ability to update ruid, euid,
568 * and suid. This allows you to implement the 4.4 compatible seteuid().
570 SYSCALL_DEFINE3(setresuid
, uid_t
, ruid
, uid_t
, euid
, uid_t
, suid
)
572 struct user_namespace
*ns
= current_user_ns();
573 const struct cred
*old
;
576 kuid_t kruid
, keuid
, ksuid
;
578 kruid
= make_kuid(ns
, ruid
);
579 keuid
= make_kuid(ns
, euid
);
580 ksuid
= make_kuid(ns
, suid
);
582 if ((ruid
!= (uid_t
) -1) && !uid_valid(kruid
))
585 if ((euid
!= (uid_t
) -1) && !uid_valid(keuid
))
588 if ((suid
!= (uid_t
) -1) && !uid_valid(ksuid
))
591 new = prepare_creds();
595 old
= current_cred();
598 if (!ns_capable(old
->user_ns
, CAP_SETUID
)) {
599 if (ruid
!= (uid_t
) -1 && !uid_eq(kruid
, old
->uid
) &&
600 !uid_eq(kruid
, old
->euid
) && !uid_eq(kruid
, old
->suid
))
602 if (euid
!= (uid_t
) -1 && !uid_eq(keuid
, old
->uid
) &&
603 !uid_eq(keuid
, old
->euid
) && !uid_eq(keuid
, old
->suid
))
605 if (suid
!= (uid_t
) -1 && !uid_eq(ksuid
, old
->uid
) &&
606 !uid_eq(ksuid
, old
->euid
) && !uid_eq(ksuid
, old
->suid
))
610 if (ruid
!= (uid_t
) -1) {
612 if (!uid_eq(kruid
, old
->uid
)) {
613 retval
= set_user(new);
618 if (euid
!= (uid_t
) -1)
620 if (suid
!= (uid_t
) -1)
622 new->fsuid
= new->euid
;
624 retval
= security_task_fix_setuid(new, old
, LSM_SETID_RES
);
628 return commit_creds(new);
635 SYSCALL_DEFINE3(getresuid
, uid_t __user
*, ruidp
, uid_t __user
*, euidp
, uid_t __user
*, suidp
)
637 const struct cred
*cred
= current_cred();
639 uid_t ruid
, euid
, suid
;
641 ruid
= from_kuid_munged(cred
->user_ns
, cred
->uid
);
642 euid
= from_kuid_munged(cred
->user_ns
, cred
->euid
);
643 suid
= from_kuid_munged(cred
->user_ns
, cred
->suid
);
645 if (!(retval
= put_user(ruid
, ruidp
)) &&
646 !(retval
= put_user(euid
, euidp
)))
647 retval
= put_user(suid
, suidp
);
653 * Same as above, but for rgid, egid, sgid.
655 SYSCALL_DEFINE3(setresgid
, gid_t
, rgid
, gid_t
, egid
, gid_t
, sgid
)
657 struct user_namespace
*ns
= current_user_ns();
658 const struct cred
*old
;
661 kgid_t krgid
, kegid
, ksgid
;
663 krgid
= make_kgid(ns
, rgid
);
664 kegid
= make_kgid(ns
, egid
);
665 ksgid
= make_kgid(ns
, sgid
);
667 if ((rgid
!= (gid_t
) -1) && !gid_valid(krgid
))
669 if ((egid
!= (gid_t
) -1) && !gid_valid(kegid
))
671 if ((sgid
!= (gid_t
) -1) && !gid_valid(ksgid
))
674 new = prepare_creds();
677 old
= current_cred();
680 if (!ns_capable(old
->user_ns
, CAP_SETGID
)) {
681 if (rgid
!= (gid_t
) -1 && !gid_eq(krgid
, old
->gid
) &&
682 !gid_eq(krgid
, old
->egid
) && !gid_eq(krgid
, old
->sgid
))
684 if (egid
!= (gid_t
) -1 && !gid_eq(kegid
, old
->gid
) &&
685 !gid_eq(kegid
, old
->egid
) && !gid_eq(kegid
, old
->sgid
))
687 if (sgid
!= (gid_t
) -1 && !gid_eq(ksgid
, old
->gid
) &&
688 !gid_eq(ksgid
, old
->egid
) && !gid_eq(ksgid
, old
->sgid
))
692 if (rgid
!= (gid_t
) -1)
694 if (egid
!= (gid_t
) -1)
696 if (sgid
!= (gid_t
) -1)
698 new->fsgid
= new->egid
;
700 return commit_creds(new);
707 SYSCALL_DEFINE3(getresgid
, gid_t __user
*, rgidp
, gid_t __user
*, egidp
, gid_t __user
*, sgidp
)
709 const struct cred
*cred
= current_cred();
711 gid_t rgid
, egid
, sgid
;
713 rgid
= from_kgid_munged(cred
->user_ns
, cred
->gid
);
714 egid
= from_kgid_munged(cred
->user_ns
, cred
->egid
);
715 sgid
= from_kgid_munged(cred
->user_ns
, cred
->sgid
);
717 if (!(retval
= put_user(rgid
, rgidp
)) &&
718 !(retval
= put_user(egid
, egidp
)))
719 retval
= put_user(sgid
, sgidp
);
726 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
727 * is used for "access()" and for the NFS daemon (letting nfsd stay at
728 * whatever uid it wants to). It normally shadows "euid", except when
729 * explicitly set by setfsuid() or for access..
731 SYSCALL_DEFINE1(setfsuid
, uid_t
, uid
)
733 const struct cred
*old
;
738 old
= current_cred();
739 old_fsuid
= from_kuid_munged(old
->user_ns
, old
->fsuid
);
741 kuid
= make_kuid(old
->user_ns
, uid
);
742 if (!uid_valid(kuid
))
745 new = prepare_creds();
749 if (uid_eq(kuid
, old
->uid
) || uid_eq(kuid
, old
->euid
) ||
750 uid_eq(kuid
, old
->suid
) || uid_eq(kuid
, old
->fsuid
) ||
751 ns_capable(old
->user_ns
, CAP_SETUID
)) {
752 if (!uid_eq(kuid
, old
->fsuid
)) {
754 if (security_task_fix_setuid(new, old
, LSM_SETID_FS
) == 0)
768 * Samma på svenska..
770 SYSCALL_DEFINE1(setfsgid
, gid_t
, gid
)
772 const struct cred
*old
;
777 old
= current_cred();
778 old_fsgid
= from_kgid_munged(old
->user_ns
, old
->fsgid
);
780 kgid
= make_kgid(old
->user_ns
, gid
);
781 if (!gid_valid(kgid
))
784 new = prepare_creds();
788 if (gid_eq(kgid
, old
->gid
) || gid_eq(kgid
, old
->egid
) ||
789 gid_eq(kgid
, old
->sgid
) || gid_eq(kgid
, old
->fsgid
) ||
790 ns_capable(old
->user_ns
, CAP_SETGID
)) {
791 if (!gid_eq(kgid
, old
->fsgid
)) {
806 * sys_getpid - return the thread group id of the current process
808 * Note, despite the name, this returns the tgid not the pid. The tgid and
809 * the pid are identical unless CLONE_THREAD was specified on clone() in
810 * which case the tgid is the same in all threads of the same group.
812 * This is SMP safe as current->tgid does not change.
814 SYSCALL_DEFINE0(getpid
)
816 return task_tgid_vnr(current
);
819 /* Thread ID - the internal kernel "pid" */
820 SYSCALL_DEFINE0(gettid
)
822 return task_pid_vnr(current
);
826 * Accessing ->real_parent is not SMP-safe, it could
827 * change from under us. However, we can use a stale
828 * value of ->real_parent under rcu_read_lock(), see
829 * release_task()->call_rcu(delayed_put_task_struct).
831 SYSCALL_DEFINE0(getppid
)
836 pid
= task_tgid_vnr(rcu_dereference(current
->real_parent
));
842 SYSCALL_DEFINE0(getuid
)
844 /* Only we change this so SMP safe */
845 return from_kuid_munged(current_user_ns(), current_uid());
848 SYSCALL_DEFINE0(geteuid
)
850 /* Only we change this so SMP safe */
851 return from_kuid_munged(current_user_ns(), current_euid());
854 SYSCALL_DEFINE0(getgid
)
856 /* Only we change this so SMP safe */
857 return from_kgid_munged(current_user_ns(), current_gid());
860 SYSCALL_DEFINE0(getegid
)
862 /* Only we change this so SMP safe */
863 return from_kgid_munged(current_user_ns(), current_egid());
866 void do_sys_times(struct tms
*tms
)
868 cputime_t tgutime
, tgstime
, cutime
, cstime
;
870 spin_lock_irq(¤t
->sighand
->siglock
);
871 thread_group_cputime_adjusted(current
, &tgutime
, &tgstime
);
872 cutime
= current
->signal
->cutime
;
873 cstime
= current
->signal
->cstime
;
874 spin_unlock_irq(¤t
->sighand
->siglock
);
875 tms
->tms_utime
= cputime_to_clock_t(tgutime
);
876 tms
->tms_stime
= cputime_to_clock_t(tgstime
);
877 tms
->tms_cutime
= cputime_to_clock_t(cutime
);
878 tms
->tms_cstime
= cputime_to_clock_t(cstime
);
881 SYSCALL_DEFINE1(times
, struct tms __user
*, tbuf
)
887 if (copy_to_user(tbuf
, &tmp
, sizeof(struct tms
)))
890 force_successful_syscall_return();
891 return (long) jiffies_64_to_clock_t(get_jiffies_64());
895 * This needs some heavy checking ...
896 * I just haven't the stomach for it. I also don't fully
897 * understand sessions/pgrp etc. Let somebody who does explain it.
899 * OK, I think I have the protection semantics right.... this is really
900 * only important on a multi-user system anyway, to make sure one user
901 * can't send a signal to a process owned by another. -TYT, 12/12/91
903 * !PF_FORKNOEXEC check to conform completely to POSIX.
905 SYSCALL_DEFINE2(setpgid
, pid_t
, pid
, pid_t
, pgid
)
907 struct task_struct
*p
;
908 struct task_struct
*group_leader
= current
->group_leader
;
913 pid
= task_pid_vnr(group_leader
);
920 /* From this point forward we keep holding onto the tasklist lock
921 * so that our parent does not change from under us. -DaveM
923 write_lock_irq(&tasklist_lock
);
926 p
= find_task_by_vpid(pid
);
931 if (!thread_group_leader(p
))
934 if (same_thread_group(p
->real_parent
, group_leader
)) {
936 if (task_session(p
) != task_session(group_leader
))
939 if (!(p
->flags
& PF_FORKNOEXEC
))
943 if (p
!= group_leader
)
948 if (p
->signal
->leader
)
953 struct task_struct
*g
;
955 pgrp
= find_vpid(pgid
);
956 g
= pid_task(pgrp
, PIDTYPE_PGID
);
957 if (!g
|| task_session(g
) != task_session(group_leader
))
961 err
= security_task_setpgid(p
, pgid
);
965 if (task_pgrp(p
) != pgrp
)
966 change_pid(p
, PIDTYPE_PGID
, pgrp
);
970 /* All paths lead to here, thus we are safe. -DaveM */
971 write_unlock_irq(&tasklist_lock
);
976 SYSCALL_DEFINE1(getpgid
, pid_t
, pid
)
978 struct task_struct
*p
;
984 grp
= task_pgrp(current
);
987 p
= find_task_by_vpid(pid
);
994 retval
= security_task_getpgid(p
);
998 retval
= pid_vnr(grp
);
1004 #ifdef __ARCH_WANT_SYS_GETPGRP
1006 SYSCALL_DEFINE0(getpgrp
)
1008 return sys_getpgid(0);
1013 SYSCALL_DEFINE1(getsid
, pid_t
, pid
)
1015 struct task_struct
*p
;
1021 sid
= task_session(current
);
1024 p
= find_task_by_vpid(pid
);
1027 sid
= task_session(p
);
1031 retval
= security_task_getsid(p
);
1035 retval
= pid_vnr(sid
);
1041 static void set_special_pids(struct pid
*pid
)
1043 struct task_struct
*curr
= current
->group_leader
;
1045 if (task_session(curr
) != pid
)
1046 change_pid(curr
, PIDTYPE_SID
, pid
);
1048 if (task_pgrp(curr
) != pid
)
1049 change_pid(curr
, PIDTYPE_PGID
, pid
);
1052 SYSCALL_DEFINE0(setsid
)
1054 struct task_struct
*group_leader
= current
->group_leader
;
1055 struct pid
*sid
= task_pid(group_leader
);
1056 pid_t session
= pid_vnr(sid
);
1059 write_lock_irq(&tasklist_lock
);
1060 /* Fail if I am already a session leader */
1061 if (group_leader
->signal
->leader
)
1064 /* Fail if a process group id already exists that equals the
1065 * proposed session id.
1067 if (pid_task(sid
, PIDTYPE_PGID
))
1070 group_leader
->signal
->leader
= 1;
1071 set_special_pids(sid
);
1073 proc_clear_tty(group_leader
);
1077 write_unlock_irq(&tasklist_lock
);
1079 proc_sid_connector(group_leader
);
1080 sched_autogroup_create_attach(group_leader
);
1085 DECLARE_RWSEM(uts_sem
);
1087 #ifdef COMPAT_UTS_MACHINE
1088 #define override_architecture(name) \
1089 (personality(current->personality) == PER_LINUX32 && \
1090 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1091 sizeof(COMPAT_UTS_MACHINE)))
1093 #define override_architecture(name) 0
1097 * Work around broken programs that cannot handle "Linux 3.0".
1098 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1100 static int override_release(char __user
*release
, size_t len
)
1104 if (current
->personality
& UNAME26
) {
1105 const char *rest
= UTS_RELEASE
;
1106 char buf
[65] = { 0 };
1112 if (*rest
== '.' && ++ndots
>= 3)
1114 if (!isdigit(*rest
) && *rest
!= '.')
1118 v
= ((LINUX_VERSION_CODE
>> 8) & 0xff) + 40;
1119 copy
= clamp_t(size_t, len
, 1, sizeof(buf
));
1120 copy
= scnprintf(buf
, copy
, "2.6.%u%s", v
, rest
);
1121 ret
= copy_to_user(release
, buf
, copy
+ 1);
1126 SYSCALL_DEFINE1(newuname
, struct new_utsname __user
*, name
)
1130 down_read(&uts_sem
);
1131 if (copy_to_user(name
, utsname(), sizeof *name
))
1135 if (!errno
&& override_release(name
->release
, sizeof(name
->release
)))
1137 if (!errno
&& override_architecture(name
))
1142 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1146 SYSCALL_DEFINE1(uname
, struct old_utsname __user
*, name
)
1153 down_read(&uts_sem
);
1154 if (copy_to_user(name
, utsname(), sizeof(*name
)))
1158 if (!error
&& override_release(name
->release
, sizeof(name
->release
)))
1160 if (!error
&& override_architecture(name
))
1165 SYSCALL_DEFINE1(olduname
, struct oldold_utsname __user
*, name
)
1171 if (!access_ok(VERIFY_WRITE
, name
, sizeof(struct oldold_utsname
)))
1174 down_read(&uts_sem
);
1175 error
= __copy_to_user(&name
->sysname
, &utsname()->sysname
,
1177 error
|= __put_user(0, name
->sysname
+ __OLD_UTS_LEN
);
1178 error
|= __copy_to_user(&name
->nodename
, &utsname()->nodename
,
1180 error
|= __put_user(0, name
->nodename
+ __OLD_UTS_LEN
);
1181 error
|= __copy_to_user(&name
->release
, &utsname()->release
,
1183 error
|= __put_user(0, name
->release
+ __OLD_UTS_LEN
);
1184 error
|= __copy_to_user(&name
->version
, &utsname()->version
,
1186 error
|= __put_user(0, name
->version
+ __OLD_UTS_LEN
);
1187 error
|= __copy_to_user(&name
->machine
, &utsname()->machine
,
1189 error
|= __put_user(0, name
->machine
+ __OLD_UTS_LEN
);
1192 if (!error
&& override_architecture(name
))
1194 if (!error
&& override_release(name
->release
, sizeof(name
->release
)))
1196 return error
? -EFAULT
: 0;
1200 SYSCALL_DEFINE2(sethostname
, char __user
*, name
, int, len
)
1203 char tmp
[__NEW_UTS_LEN
];
1205 if (!ns_capable(current
->nsproxy
->uts_ns
->user_ns
, CAP_SYS_ADMIN
))
1208 if (len
< 0 || len
> __NEW_UTS_LEN
)
1210 down_write(&uts_sem
);
1212 if (!copy_from_user(tmp
, name
, len
)) {
1213 struct new_utsname
*u
= utsname();
1215 memcpy(u
->nodename
, tmp
, len
);
1216 memset(u
->nodename
+ len
, 0, sizeof(u
->nodename
) - len
);
1218 uts_proc_notify(UTS_PROC_HOSTNAME
);
1224 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1226 SYSCALL_DEFINE2(gethostname
, char __user
*, name
, int, len
)
1229 struct new_utsname
*u
;
1233 down_read(&uts_sem
);
1235 i
= 1 + strlen(u
->nodename
);
1239 if (copy_to_user(name
, u
->nodename
, i
))
1248 * Only setdomainname; getdomainname can be implemented by calling
1251 SYSCALL_DEFINE2(setdomainname
, char __user
*, name
, int, len
)
1254 char tmp
[__NEW_UTS_LEN
];
1256 if (!ns_capable(current
->nsproxy
->uts_ns
->user_ns
, CAP_SYS_ADMIN
))
1258 if (len
< 0 || len
> __NEW_UTS_LEN
)
1261 down_write(&uts_sem
);
1263 if (!copy_from_user(tmp
, name
, len
)) {
1264 struct new_utsname
*u
= utsname();
1266 memcpy(u
->domainname
, tmp
, len
);
1267 memset(u
->domainname
+ len
, 0, sizeof(u
->domainname
) - len
);
1269 uts_proc_notify(UTS_PROC_DOMAINNAME
);
1275 SYSCALL_DEFINE2(getrlimit
, unsigned int, resource
, struct rlimit __user
*, rlim
)
1277 struct rlimit value
;
1280 ret
= do_prlimit(current
, resource
, NULL
, &value
);
1282 ret
= copy_to_user(rlim
, &value
, sizeof(*rlim
)) ? -EFAULT
: 0;
1287 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1290 * Back compatibility for getrlimit. Needed for some apps.
1293 SYSCALL_DEFINE2(old_getrlimit
, unsigned int, resource
,
1294 struct rlimit __user
*, rlim
)
1297 if (resource
>= RLIM_NLIMITS
)
1300 resource
= array_index_nospec(resource
, RLIM_NLIMITS
);
1301 task_lock(current
->group_leader
);
1302 x
= current
->signal
->rlim
[resource
];
1303 task_unlock(current
->group_leader
);
1304 if (x
.rlim_cur
> 0x7FFFFFFF)
1305 x
.rlim_cur
= 0x7FFFFFFF;
1306 if (x
.rlim_max
> 0x7FFFFFFF)
1307 x
.rlim_max
= 0x7FFFFFFF;
1308 return copy_to_user(rlim
, &x
, sizeof(x
))?-EFAULT
:0;
1313 static inline bool rlim64_is_infinity(__u64 rlim64
)
1315 #if BITS_PER_LONG < 64
1316 return rlim64
>= ULONG_MAX
;
1318 return rlim64
== RLIM64_INFINITY
;
1322 static void rlim_to_rlim64(const struct rlimit
*rlim
, struct rlimit64
*rlim64
)
1324 if (rlim
->rlim_cur
== RLIM_INFINITY
)
1325 rlim64
->rlim_cur
= RLIM64_INFINITY
;
1327 rlim64
->rlim_cur
= rlim
->rlim_cur
;
1328 if (rlim
->rlim_max
== RLIM_INFINITY
)
1329 rlim64
->rlim_max
= RLIM64_INFINITY
;
1331 rlim64
->rlim_max
= rlim
->rlim_max
;
1334 static void rlim64_to_rlim(const struct rlimit64
*rlim64
, struct rlimit
*rlim
)
1336 if (rlim64_is_infinity(rlim64
->rlim_cur
))
1337 rlim
->rlim_cur
= RLIM_INFINITY
;
1339 rlim
->rlim_cur
= (unsigned long)rlim64
->rlim_cur
;
1340 if (rlim64_is_infinity(rlim64
->rlim_max
))
1341 rlim
->rlim_max
= RLIM_INFINITY
;
1343 rlim
->rlim_max
= (unsigned long)rlim64
->rlim_max
;
1346 /* make sure you are allowed to change @tsk limits before calling this */
1347 int do_prlimit(struct task_struct
*tsk
, unsigned int resource
,
1348 struct rlimit
*new_rlim
, struct rlimit
*old_rlim
)
1350 struct rlimit
*rlim
;
1353 if (resource
>= RLIM_NLIMITS
)
1356 if (new_rlim
->rlim_cur
> new_rlim
->rlim_max
)
1358 if (resource
== RLIMIT_NOFILE
&&
1359 new_rlim
->rlim_max
> sysctl_nr_open
)
1363 /* protect tsk->signal and tsk->sighand from disappearing */
1364 read_lock(&tasklist_lock
);
1365 if (!tsk
->sighand
) {
1370 rlim
= tsk
->signal
->rlim
+ resource
;
1371 task_lock(tsk
->group_leader
);
1373 /* Keep the capable check against init_user_ns until
1374 cgroups can contain all limits */
1375 if (new_rlim
->rlim_max
> rlim
->rlim_max
&&
1376 !capable(CAP_SYS_RESOURCE
))
1379 retval
= security_task_setrlimit(tsk
->group_leader
,
1380 resource
, new_rlim
);
1381 if (resource
== RLIMIT_CPU
&& new_rlim
->rlim_cur
== 0) {
1383 * The caller is asking for an immediate RLIMIT_CPU
1384 * expiry. But we use the zero value to mean "it was
1385 * never set". So let's cheat and make it one second
1388 new_rlim
->rlim_cur
= 1;
1397 task_unlock(tsk
->group_leader
);
1400 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1401 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1402 * very long-standing error, and fixing it now risks breakage of
1403 * applications, so we live with it
1405 if (!retval
&& new_rlim
&& resource
== RLIMIT_CPU
&&
1406 new_rlim
->rlim_cur
!= RLIM_INFINITY
)
1407 update_rlimit_cpu(tsk
, new_rlim
->rlim_cur
);
1409 read_unlock(&tasklist_lock
);
1413 /* rcu lock must be held */
1414 static int check_prlimit_permission(struct task_struct
*task
)
1416 const struct cred
*cred
= current_cred(), *tcred
;
1418 if (current
== task
)
1421 tcred
= __task_cred(task
);
1422 if (uid_eq(cred
->uid
, tcred
->euid
) &&
1423 uid_eq(cred
->uid
, tcred
->suid
) &&
1424 uid_eq(cred
->uid
, tcred
->uid
) &&
1425 gid_eq(cred
->gid
, tcred
->egid
) &&
1426 gid_eq(cred
->gid
, tcred
->sgid
) &&
1427 gid_eq(cred
->gid
, tcred
->gid
))
1429 if (ns_capable(tcred
->user_ns
, CAP_SYS_RESOURCE
))
1435 SYSCALL_DEFINE4(prlimit64
, pid_t
, pid
, unsigned int, resource
,
1436 const struct rlimit64 __user
*, new_rlim
,
1437 struct rlimit64 __user
*, old_rlim
)
1439 struct rlimit64 old64
, new64
;
1440 struct rlimit old
, new;
1441 struct task_struct
*tsk
;
1445 if (copy_from_user(&new64
, new_rlim
, sizeof(new64
)))
1447 rlim64_to_rlim(&new64
, &new);
1451 tsk
= pid
? find_task_by_vpid(pid
) : current
;
1456 ret
= check_prlimit_permission(tsk
);
1461 get_task_struct(tsk
);
1464 ret
= do_prlimit(tsk
, resource
, new_rlim
? &new : NULL
,
1465 old_rlim
? &old
: NULL
);
1467 if (!ret
&& old_rlim
) {
1468 rlim_to_rlim64(&old
, &old64
);
1469 if (copy_to_user(old_rlim
, &old64
, sizeof(old64
)))
1473 put_task_struct(tsk
);
1477 SYSCALL_DEFINE2(setrlimit
, unsigned int, resource
, struct rlimit __user
*, rlim
)
1479 struct rlimit new_rlim
;
1481 if (copy_from_user(&new_rlim
, rlim
, sizeof(*rlim
)))
1483 return do_prlimit(current
, resource
, &new_rlim
, NULL
);
1487 * It would make sense to put struct rusage in the task_struct,
1488 * except that would make the task_struct be *really big*. After
1489 * task_struct gets moved into malloc'ed memory, it would
1490 * make sense to do this. It will make moving the rest of the information
1491 * a lot simpler! (Which we're not doing right now because we're not
1492 * measuring them yet).
1494 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1495 * races with threads incrementing their own counters. But since word
1496 * reads are atomic, we either get new values or old values and we don't
1497 * care which for the sums. We always take the siglock to protect reading
1498 * the c* fields from p->signal from races with exit.c updating those
1499 * fields when reaping, so a sample either gets all the additions of a
1500 * given child after it's reaped, or none so this sample is before reaping.
1503 * We need to take the siglock for CHILDEREN, SELF and BOTH
1504 * for the cases current multithreaded, non-current single threaded
1505 * non-current multithreaded. Thread traversal is now safe with
1507 * Strictly speaking, we donot need to take the siglock if we are current and
1508 * single threaded, as no one else can take our signal_struct away, no one
1509 * else can reap the children to update signal->c* counters, and no one else
1510 * can race with the signal-> fields. If we do not take any lock, the
1511 * signal-> fields could be read out of order while another thread was just
1512 * exiting. So we should place a read memory barrier when we avoid the lock.
1513 * On the writer side, write memory barrier is implied in __exit_signal
1514 * as __exit_signal releases the siglock spinlock after updating the signal->
1515 * fields. But we don't do this yet to keep things simple.
1519 static void accumulate_thread_rusage(struct task_struct
*t
, struct rusage
*r
)
1521 r
->ru_nvcsw
+= t
->nvcsw
;
1522 r
->ru_nivcsw
+= t
->nivcsw
;
1523 r
->ru_minflt
+= t
->min_flt
;
1524 r
->ru_majflt
+= t
->maj_flt
;
1525 r
->ru_inblock
+= task_io_get_inblock(t
);
1526 r
->ru_oublock
+= task_io_get_oublock(t
);
1529 static void k_getrusage(struct task_struct
*p
, int who
, struct rusage
*r
)
1531 struct task_struct
*t
;
1532 unsigned long flags
;
1533 cputime_t tgutime
, tgstime
, utime
, stime
;
1534 unsigned long maxrss
= 0;
1536 memset((char *) r
, 0, sizeof *r
);
1539 if (who
== RUSAGE_THREAD
) {
1540 task_cputime_adjusted(current
, &utime
, &stime
);
1541 accumulate_thread_rusage(p
, r
);
1542 maxrss
= p
->signal
->maxrss
;
1546 if (!lock_task_sighand(p
, &flags
))
1551 case RUSAGE_CHILDREN
:
1552 utime
= p
->signal
->cutime
;
1553 stime
= p
->signal
->cstime
;
1554 r
->ru_nvcsw
= p
->signal
->cnvcsw
;
1555 r
->ru_nivcsw
= p
->signal
->cnivcsw
;
1556 r
->ru_minflt
= p
->signal
->cmin_flt
;
1557 r
->ru_majflt
= p
->signal
->cmaj_flt
;
1558 r
->ru_inblock
= p
->signal
->cinblock
;
1559 r
->ru_oublock
= p
->signal
->coublock
;
1560 maxrss
= p
->signal
->cmaxrss
;
1562 if (who
== RUSAGE_CHILDREN
)
1566 thread_group_cputime_adjusted(p
, &tgutime
, &tgstime
);
1569 r
->ru_nvcsw
+= p
->signal
->nvcsw
;
1570 r
->ru_nivcsw
+= p
->signal
->nivcsw
;
1571 r
->ru_minflt
+= p
->signal
->min_flt
;
1572 r
->ru_majflt
+= p
->signal
->maj_flt
;
1573 r
->ru_inblock
+= p
->signal
->inblock
;
1574 r
->ru_oublock
+= p
->signal
->oublock
;
1575 if (maxrss
< p
->signal
->maxrss
)
1576 maxrss
= p
->signal
->maxrss
;
1579 accumulate_thread_rusage(t
, r
);
1580 } while_each_thread(p
, t
);
1586 unlock_task_sighand(p
, &flags
);
1589 cputime_to_timeval(utime
, &r
->ru_utime
);
1590 cputime_to_timeval(stime
, &r
->ru_stime
);
1592 if (who
!= RUSAGE_CHILDREN
) {
1593 struct mm_struct
*mm
= get_task_mm(p
);
1595 setmax_mm_hiwater_rss(&maxrss
, mm
);
1599 r
->ru_maxrss
= maxrss
* (PAGE_SIZE
/ 1024); /* convert pages to KBs */
1602 int getrusage(struct task_struct
*p
, int who
, struct rusage __user
*ru
)
1605 k_getrusage(p
, who
, &r
);
1606 return copy_to_user(ru
, &r
, sizeof(r
)) ? -EFAULT
: 0;
1609 SYSCALL_DEFINE2(getrusage
, int, who
, struct rusage __user
*, ru
)
1611 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
&&
1612 who
!= RUSAGE_THREAD
)
1614 return getrusage(current
, who
, ru
);
1617 #ifdef CONFIG_COMPAT
1618 COMPAT_SYSCALL_DEFINE2(getrusage
, int, who
, struct compat_rusage __user
*, ru
)
1622 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
&&
1623 who
!= RUSAGE_THREAD
)
1626 k_getrusage(current
, who
, &r
);
1627 return put_compat_rusage(&r
, ru
);
1631 SYSCALL_DEFINE1(umask
, int, mask
)
1633 mask
= xchg(¤t
->fs
->umask
, mask
& S_IRWXUGO
);
1637 static int prctl_set_mm_exe_file(struct mm_struct
*mm
, unsigned int fd
)
1640 struct inode
*inode
;
1647 inode
= file_inode(exe
.file
);
1650 * Because the original mm->exe_file points to executable file, make
1651 * sure that this one is executable as well, to avoid breaking an
1655 if (!S_ISREG(inode
->i_mode
) || path_noexec(&exe
.file
->f_path
))
1658 err
= inode_permission(inode
, MAY_EXEC
);
1662 down_write(&mm
->mmap_sem
);
1665 * Forbid mm->exe_file change if old file still mapped.
1669 struct vm_area_struct
*vma
;
1671 for (vma
= mm
->mmap
; vma
; vma
= vma
->vm_next
)
1673 path_equal(&vma
->vm_file
->f_path
,
1674 &mm
->exe_file
->f_path
))
1679 * The symlink can be changed only once, just to disallow arbitrary
1680 * transitions malicious software might bring in. This means one
1681 * could make a snapshot over all processes running and monitor
1682 * /proc/pid/exe changes to notice unusual activity if needed.
1685 if (test_and_set_bit(MMF_EXE_FILE_CHANGED
, &mm
->flags
))
1689 set_mm_exe_file(mm
, exe
.file
); /* this grabs a reference to exe.file */
1691 up_write(&mm
->mmap_sem
);
1698 static int prctl_set_mm(int opt
, unsigned long addr
,
1699 unsigned long arg4
, unsigned long arg5
)
1701 unsigned long rlim
= rlimit(RLIMIT_DATA
);
1702 struct mm_struct
*mm
= current
->mm
;
1703 struct vm_area_struct
*vma
;
1706 if (arg5
|| (arg4
&& opt
!= PR_SET_MM_AUXV
))
1709 if (!capable(CAP_SYS_RESOURCE
))
1712 if (opt
== PR_SET_MM_EXE_FILE
)
1713 return prctl_set_mm_exe_file(mm
, (unsigned int)addr
);
1715 if (addr
>= TASK_SIZE
|| addr
< mmap_min_addr
)
1720 down_read(&mm
->mmap_sem
);
1721 vma
= find_vma(mm
, addr
);
1724 case PR_SET_MM_START_CODE
:
1725 mm
->start_code
= addr
;
1727 case PR_SET_MM_END_CODE
:
1728 mm
->end_code
= addr
;
1730 case PR_SET_MM_START_DATA
:
1731 mm
->start_data
= addr
;
1733 case PR_SET_MM_END_DATA
:
1734 mm
->end_data
= addr
;
1737 case PR_SET_MM_START_BRK
:
1738 if (addr
<= mm
->end_data
)
1741 if (rlim
< RLIM_INFINITY
&&
1743 (mm
->end_data
- mm
->start_data
) > rlim
)
1746 mm
->start_brk
= addr
;
1750 if (addr
<= mm
->end_data
)
1753 if (rlim
< RLIM_INFINITY
&&
1754 (addr
- mm
->start_brk
) +
1755 (mm
->end_data
- mm
->start_data
) > rlim
)
1762 * If command line arguments and environment
1763 * are placed somewhere else on stack, we can
1764 * set them up here, ARG_START/END to setup
1765 * command line argumets and ENV_START/END
1768 case PR_SET_MM_START_STACK
:
1769 case PR_SET_MM_ARG_START
:
1770 case PR_SET_MM_ARG_END
:
1771 case PR_SET_MM_ENV_START
:
1772 case PR_SET_MM_ENV_END
:
1777 if (opt
== PR_SET_MM_START_STACK
)
1778 mm
->start_stack
= addr
;
1779 else if (opt
== PR_SET_MM_ARG_START
)
1780 mm
->arg_start
= addr
;
1781 else if (opt
== PR_SET_MM_ARG_END
)
1783 else if (opt
== PR_SET_MM_ENV_START
)
1784 mm
->env_start
= addr
;
1785 else if (opt
== PR_SET_MM_ENV_END
)
1790 * This doesn't move auxiliary vector itself
1791 * since it's pinned to mm_struct, but allow
1792 * to fill vector with new values. It's up
1793 * to a caller to provide sane values here
1794 * otherwise user space tools which use this
1795 * vector might be unhappy.
1797 case PR_SET_MM_AUXV
: {
1798 unsigned long user_auxv
[AT_VECTOR_SIZE
];
1800 if (arg4
> sizeof(user_auxv
))
1802 up_read(&mm
->mmap_sem
);
1804 if (copy_from_user(user_auxv
, (const void __user
*)addr
, arg4
))
1807 /* Make sure the last entry is always AT_NULL */
1808 user_auxv
[AT_VECTOR_SIZE
- 2] = 0;
1809 user_auxv
[AT_VECTOR_SIZE
- 1] = 0;
1811 BUILD_BUG_ON(sizeof(user_auxv
) != sizeof(mm
->saved_auxv
));
1814 memcpy(mm
->saved_auxv
, user_auxv
, arg4
);
1815 task_unlock(current
);
1825 up_read(&mm
->mmap_sem
);
1829 #ifdef CONFIG_CHECKPOINT_RESTORE
1830 static int prctl_get_tid_address(struct task_struct
*me
, int __user
**tid_addr
)
1832 return put_user(me
->clear_child_tid
, tid_addr
);
1835 static int prctl_get_tid_address(struct task_struct
*me
, int __user
**tid_addr
)
1841 int __weak
arch_prctl_spec_ctrl_get(struct task_struct
*t
, unsigned long which
)
1846 int __weak
arch_prctl_spec_ctrl_set(struct task_struct
*t
, unsigned long which
,
1852 SYSCALL_DEFINE5(prctl
, int, option
, unsigned long, arg2
, unsigned long, arg3
,
1853 unsigned long, arg4
, unsigned long, arg5
)
1855 struct task_struct
*me
= current
;
1856 unsigned char comm
[sizeof(me
->comm
)];
1859 error
= security_task_prctl(option
, arg2
, arg3
, arg4
, arg5
);
1860 if (error
!= -ENOSYS
)
1865 case PR_SET_PDEATHSIG
:
1866 if (!valid_signal(arg2
)) {
1870 me
->pdeath_signal
= arg2
;
1872 case PR_GET_PDEATHSIG
:
1873 error
= put_user(me
->pdeath_signal
, (int __user
*)arg2
);
1875 case PR_GET_DUMPABLE
:
1876 error
= get_dumpable(me
->mm
);
1878 case PR_SET_DUMPABLE
:
1879 if (arg2
!= SUID_DUMP_DISABLE
&& arg2
!= SUID_DUMP_USER
) {
1883 set_dumpable(me
->mm
, arg2
);
1886 case PR_SET_UNALIGN
:
1887 error
= SET_UNALIGN_CTL(me
, arg2
);
1889 case PR_GET_UNALIGN
:
1890 error
= GET_UNALIGN_CTL(me
, arg2
);
1893 error
= SET_FPEMU_CTL(me
, arg2
);
1896 error
= GET_FPEMU_CTL(me
, arg2
);
1899 error
= SET_FPEXC_CTL(me
, arg2
);
1902 error
= GET_FPEXC_CTL(me
, arg2
);
1905 error
= PR_TIMING_STATISTICAL
;
1908 if (arg2
!= PR_TIMING_STATISTICAL
)
1912 comm
[sizeof(me
->comm
) - 1] = 0;
1913 if (strncpy_from_user(comm
, (char __user
*)arg2
,
1914 sizeof(me
->comm
) - 1) < 0)
1916 set_task_comm(me
, comm
);
1917 proc_comm_connector(me
);
1920 get_task_comm(comm
, me
);
1921 if (copy_to_user((char __user
*)arg2
, comm
, sizeof(comm
)))
1925 error
= GET_ENDIAN(me
, arg2
);
1928 error
= SET_ENDIAN(me
, arg2
);
1930 case PR_GET_SECCOMP
:
1931 error
= prctl_get_seccomp();
1933 case PR_SET_SECCOMP
:
1934 error
= prctl_set_seccomp(arg2
, (char __user
*)arg3
);
1937 error
= GET_TSC_CTL(arg2
);
1940 error
= SET_TSC_CTL(arg2
);
1942 case PR_TASK_PERF_EVENTS_DISABLE
:
1943 error
= perf_event_task_disable();
1945 case PR_TASK_PERF_EVENTS_ENABLE
:
1946 error
= perf_event_task_enable();
1948 case PR_GET_TIMERSLACK
:
1949 error
= current
->timer_slack_ns
;
1951 case PR_SET_TIMERSLACK
:
1953 current
->timer_slack_ns
=
1954 current
->default_timer_slack_ns
;
1956 current
->timer_slack_ns
= arg2
;
1962 case PR_MCE_KILL_CLEAR
:
1965 current
->flags
&= ~PF_MCE_PROCESS
;
1967 case PR_MCE_KILL_SET
:
1968 current
->flags
|= PF_MCE_PROCESS
;
1969 if (arg3
== PR_MCE_KILL_EARLY
)
1970 current
->flags
|= PF_MCE_EARLY
;
1971 else if (arg3
== PR_MCE_KILL_LATE
)
1972 current
->flags
&= ~PF_MCE_EARLY
;
1973 else if (arg3
== PR_MCE_KILL_DEFAULT
)
1975 ~(PF_MCE_EARLY
|PF_MCE_PROCESS
);
1983 case PR_MCE_KILL_GET
:
1984 if (arg2
| arg3
| arg4
| arg5
)
1986 if (current
->flags
& PF_MCE_PROCESS
)
1987 error
= (current
->flags
& PF_MCE_EARLY
) ?
1988 PR_MCE_KILL_EARLY
: PR_MCE_KILL_LATE
;
1990 error
= PR_MCE_KILL_DEFAULT
;
1993 error
= prctl_set_mm(arg2
, arg3
, arg4
, arg5
);
1995 case PR_GET_TID_ADDRESS
:
1996 error
= prctl_get_tid_address(me
, (int __user
**)arg2
);
1998 case PR_SET_CHILD_SUBREAPER
:
1999 me
->signal
->is_child_subreaper
= !!arg2
;
2001 case PR_GET_CHILD_SUBREAPER
:
2002 error
= put_user(me
->signal
->is_child_subreaper
,
2003 (int __user
*)arg2
);
2005 case PR_SET_NO_NEW_PRIVS
:
2006 if (arg2
!= 1 || arg3
|| arg4
|| arg5
)
2009 task_set_no_new_privs(current
);
2011 case PR_GET_NO_NEW_PRIVS
:
2012 if (arg2
|| arg3
|| arg4
|| arg5
)
2014 return task_no_new_privs(current
) ? 1 : 0;
2015 case PR_GET_THP_DISABLE
:
2016 if (arg2
|| arg3
|| arg4
|| arg5
)
2018 error
= !!(me
->mm
->def_flags
& VM_NOHUGEPAGE
);
2020 case PR_SET_THP_DISABLE
:
2021 if (arg3
|| arg4
|| arg5
)
2023 down_write(&me
->mm
->mmap_sem
);
2025 me
->mm
->def_flags
|= VM_NOHUGEPAGE
;
2027 me
->mm
->def_flags
&= ~VM_NOHUGEPAGE
;
2028 up_write(&me
->mm
->mmap_sem
);
2030 case PR_GET_SPECULATION_CTRL
:
2031 if (arg3
|| arg4
|| arg5
)
2033 error
= arch_prctl_spec_ctrl_get(me
, arg2
);
2035 case PR_SET_SPECULATION_CTRL
:
2038 error
= arch_prctl_spec_ctrl_set(me
, arg2
, arg3
);
2047 SYSCALL_DEFINE3(getcpu
, unsigned __user
*, cpup
, unsigned __user
*, nodep
,
2048 struct getcpu_cache __user
*, unused
)
2051 int cpu
= raw_smp_processor_id();
2053 err
|= put_user(cpu
, cpup
);
2055 err
|= put_user(cpu_to_node(cpu
), nodep
);
2056 return err
? -EFAULT
: 0;
2060 * do_sysinfo - fill in sysinfo struct
2061 * @info: pointer to buffer to fill
2063 static int do_sysinfo(struct sysinfo
*info
)
2065 unsigned long mem_total
, sav_total
;
2066 unsigned int mem_unit
, bitcount
;
2069 memset(info
, 0, sizeof(struct sysinfo
));
2071 get_monotonic_boottime(&tp
);
2072 info
->uptime
= tp
.tv_sec
+ (tp
.tv_nsec
? 1 : 0);
2074 get_avenrun(info
->loads
, 0, SI_LOAD_SHIFT
- FSHIFT
);
2076 info
->procs
= nr_threads
;
2082 * If the sum of all the available memory (i.e. ram + swap)
2083 * is less than can be stored in a 32 bit unsigned long then
2084 * we can be binary compatible with 2.2.x kernels. If not,
2085 * well, in that case 2.2.x was broken anyways...
2087 * -Erik Andersen <andersee@debian.org>
2090 mem_total
= info
->totalram
+ info
->totalswap
;
2091 if (mem_total
< info
->totalram
|| mem_total
< info
->totalswap
)
2094 mem_unit
= info
->mem_unit
;
2095 while (mem_unit
> 1) {
2098 sav_total
= mem_total
;
2100 if (mem_total
< sav_total
)
2105 * If mem_total did not overflow, multiply all memory values by
2106 * info->mem_unit and set it to 1. This leaves things compatible
2107 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2112 info
->totalram
<<= bitcount
;
2113 info
->freeram
<<= bitcount
;
2114 info
->sharedram
<<= bitcount
;
2115 info
->bufferram
<<= bitcount
;
2116 info
->totalswap
<<= bitcount
;
2117 info
->freeswap
<<= bitcount
;
2118 info
->totalhigh
<<= bitcount
;
2119 info
->freehigh
<<= bitcount
;
2125 SYSCALL_DEFINE1(sysinfo
, struct sysinfo __user
*, info
)
2131 if (copy_to_user(info
, &val
, sizeof(struct sysinfo
)))
2137 #ifdef CONFIG_COMPAT
2138 struct compat_sysinfo
{
2152 char _f
[20-2*sizeof(u32
)-sizeof(int)];
2155 COMPAT_SYSCALL_DEFINE1(sysinfo
, struct compat_sysinfo __user
*, info
)
2161 /* Check to see if any memory value is too large for 32-bit and scale
2164 if ((s
.totalram
>> 32) || (s
.totalswap
>> 32)) {
2167 while (s
.mem_unit
< PAGE_SIZE
) {
2172 s
.totalram
>>= bitcount
;
2173 s
.freeram
>>= bitcount
;
2174 s
.sharedram
>>= bitcount
;
2175 s
.bufferram
>>= bitcount
;
2176 s
.totalswap
>>= bitcount
;
2177 s
.freeswap
>>= bitcount
;
2178 s
.totalhigh
>>= bitcount
;
2179 s
.freehigh
>>= bitcount
;
2182 if (!access_ok(VERIFY_WRITE
, info
, sizeof(struct compat_sysinfo
)) ||
2183 __put_user(s
.uptime
, &info
->uptime
) ||
2184 __put_user(s
.loads
[0], &info
->loads
[0]) ||
2185 __put_user(s
.loads
[1], &info
->loads
[1]) ||
2186 __put_user(s
.loads
[2], &info
->loads
[2]) ||
2187 __put_user(s
.totalram
, &info
->totalram
) ||
2188 __put_user(s
.freeram
, &info
->freeram
) ||
2189 __put_user(s
.sharedram
, &info
->sharedram
) ||
2190 __put_user(s
.bufferram
, &info
->bufferram
) ||
2191 __put_user(s
.totalswap
, &info
->totalswap
) ||
2192 __put_user(s
.freeswap
, &info
->freeswap
) ||
2193 __put_user(s
.procs
, &info
->procs
) ||
2194 __put_user(s
.totalhigh
, &info
->totalhigh
) ||
2195 __put_user(s
.freehigh
, &info
->freehigh
) ||
2196 __put_user(s
.mem_unit
, &info
->mem_unit
))
2201 #endif /* CONFIG_COMPAT */