projects/linux·blob 50c94de6 · refs

2778 lines · 65.0 KiB

1// SPDX-License-Identifier: GPL-2.0
2/*
3 * linux/kernel/sys.c
4 *
5 * Copyright (C) 1991, 1992 Linus Torvalds
6 */
7
8#include <linux/export.h>
9#include <linux/mm.h>
10#include <linux/mm_inline.h>
11#include <linux/utsname.h>
12#include <linux/mman.h>
13#include <linux/reboot.h>
14#include <linux/prctl.h>
15#include <linux/highuid.h>
16#include <linux/fs.h>
17#include <linux/kmod.h>
18#include <linux/perf_event.h>
19#include <linux/resource.h>
20#include <linux/kernel.h>
21#include <linux/workqueue.h>
22#include <linux/capability.h>
23#include <linux/device.h>
24#include <linux/key.h>
25#include <linux/times.h>
26#include <linux/posix-timers.h>
27#include <linux/security.h>
28#include <linux/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>
45#include <linux/syscall_user_dispatch.h>
46
47#include <linux/compat.h>
48#include <linux/syscalls.h>
49#include <linux/kprobes.h>
50#include <linux/user_namespace.h>
51#include <linux/time_namespace.h>
52#include <linux/binfmts.h>
53
54#include <linux/sched.h>
55#include <linux/sched/autogroup.h>
56#include <linux/sched/loadavg.h>
57#include <linux/sched/stat.h>
58#include <linux/sched/mm.h>
59#include <linux/sched/coredump.h>
60#include <linux/sched/task.h>
61#include <linux/sched/cputime.h>
62#include <linux/rcupdate.h>
63#include <linux/uidgid.h>
64#include <linux/cred.h>
65
66#include <linux/nospec.h>
67
68#include <linux/kmsg_dump.h>
69/* Move somewhere else to avoid recompiling? */
70#include <generated/utsrelease.h>
71
72#include <linux/uaccess.h>
73#include <asm/io.h>
74#include <asm/unistd.h>
75
76#include "uid16.h"
77
78#ifndef SET_UNALIGN_CTL
79# define SET_UNALIGN_CTL(a, b) (-EINVAL)
80#endif
81#ifndef GET_UNALIGN_CTL
82# define GET_UNALIGN_CTL(a, b) (-EINVAL)
83#endif
84#ifndef SET_FPEMU_CTL
85# define SET_FPEMU_CTL(a, b) (-EINVAL)
86#endif
87#ifndef GET_FPEMU_CTL
88# define GET_FPEMU_CTL(a, b) (-EINVAL)
89#endif
90#ifndef SET_FPEXC_CTL
91# define SET_FPEXC_CTL(a, b) (-EINVAL)
92#endif
93#ifndef GET_FPEXC_CTL
94# define GET_FPEXC_CTL(a, b) (-EINVAL)
95#endif
96#ifndef GET_ENDIAN
97# define GET_ENDIAN(a, b) (-EINVAL)
98#endif
99#ifndef SET_ENDIAN
100# define SET_ENDIAN(a, b) (-EINVAL)
101#endif
102#ifndef GET_TSC_CTL
103# define GET_TSC_CTL(a) (-EINVAL)
104#endif
105#ifndef SET_TSC_CTL
106# define SET_TSC_CTL(a) (-EINVAL)
107#endif
108#ifndef GET_FP_MODE
109# define GET_FP_MODE(a) (-EINVAL)
110#endif
111#ifndef SET_FP_MODE
112# define SET_FP_MODE(a,b) (-EINVAL)
113#endif
114#ifndef SVE_SET_VL
115# define SVE_SET_VL(a) (-EINVAL)
116#endif
117#ifndef SVE_GET_VL
118# define SVE_GET_VL() (-EINVAL)
119#endif
120#ifndef PAC_RESET_KEYS
121# define PAC_RESET_KEYS(a, b) (-EINVAL)
122#endif
123#ifndef PAC_SET_ENABLED_KEYS
124# define PAC_SET_ENABLED_KEYS(a, b, c) (-EINVAL)
125#endif
126#ifndef PAC_GET_ENABLED_KEYS
127# define PAC_GET_ENABLED_KEYS(a) (-EINVAL)
128#endif
129#ifndef SET_TAGGED_ADDR_CTRL
130# define SET_TAGGED_ADDR_CTRL(a) (-EINVAL)
131#endif
132#ifndef GET_TAGGED_ADDR_CTRL
133# define GET_TAGGED_ADDR_CTRL() (-EINVAL)
134#endif
135
136/*
137 * this is where the system-wide overflow UID and GID are defined, for
138 * architectures that now have 32-bit UID/GID but didn't in the past
139 */
140
141int overflowuid = DEFAULT_OVERFLOWUID;
142int overflowgid = DEFAULT_OVERFLOWGID;
143
144EXPORT_SYMBOL(overflowuid);
145EXPORT_SYMBOL(overflowgid);
146
147/*
148 * the same as above, but for filesystems which can only store a 16-bit
149 * UID and GID. as such, this is needed on all architectures
150 */
151
152int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
153int fs_overflowgid = DEFAULT_FS_OVERFLOWGID;
154
155EXPORT_SYMBOL(fs_overflowuid);
156EXPORT_SYMBOL(fs_overflowgid);
157
158/*
159 * Returns true if current's euid is same as p's uid or euid,
160 * or has CAP_SYS_NICE to p's user_ns.
161 *
162 * Called with rcu_read_lock, creds are safe
163 */
164static bool set_one_prio_perm(struct task_struct *p)
165{
166 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
167
168 if (uid_eq(pcred->uid, cred->euid) ||
169 uid_eq(pcred->euid, cred->euid))
170 return true;
171 if (ns_capable(pcred->user_ns, CAP_SYS_NICE))
172 return true;
173 return false;
174}
175
176/*
177 * set the priority of a task
178 * - the caller must hold the RCU read lock
179 */
180static int set_one_prio(struct task_struct *p, int niceval, int error)
181{
182 int no_nice;
183
184 if (!set_one_prio_perm(p)) {
185 error = -EPERM;
186 goto out;
187 }
188 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
189 error = -EACCES;
190 goto out;
191 }
192 no_nice = security_task_setnice(p, niceval);
193 if (no_nice) {
194 error = no_nice;
195 goto out;
196 }
197 if (error == -ESRCH)
198 error = 0;
199 set_user_nice(p, niceval);
200out:
201 return error;
202}
203
204SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
205{
206 struct task_struct *g, *p;
207 struct user_struct *user;
208 const struct cred *cred = current_cred();
209 int error = -EINVAL;
210 struct pid *pgrp;
211 kuid_t uid;
212
213 if (which > PRIO_USER || which < PRIO_PROCESS)
214 goto out;
215
216 /* normalize: avoid signed division (rounding problems) */
217 error = -ESRCH;
218 if (niceval < MIN_NICE)
219 niceval = MIN_NICE;
220 if (niceval > MAX_NICE)
221 niceval = MAX_NICE;
222
223 rcu_read_lock();
224 switch (which) {
225 case PRIO_PROCESS:
226 if (who)
227 p = find_task_by_vpid(who);
228 else
229 p = current;
230 if (p)
231 error = set_one_prio(p, niceval, error);
232 break;
233 case PRIO_PGRP:
234 if (who)
235 pgrp = find_vpid(who);
236 else
237 pgrp = task_pgrp(current);
238 read_lock(&tasklist_lock);
239 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
240 error = set_one_prio(p, niceval, error);
241 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
242 read_unlock(&tasklist_lock);
243 break;
244 case PRIO_USER:
245 uid = make_kuid(cred->user_ns, who);
246 user = cred->user;
247 if (!who)
248 uid = cred->uid;
249 else if (!uid_eq(uid, cred->uid)) {
250 user = find_user(uid);
251 if (!user)
252 goto out_unlock; /* No processes for this user */
253 }
254 for_each_process_thread(g, p) {
255 if (uid_eq(task_uid(p), uid) && task_pid_vnr(p))
256 error = set_one_prio(p, niceval, error);
257 }
258 if (!uid_eq(uid, cred->uid))
259 free_uid(user); /* For find_user() */
260 break;
261 }
262out_unlock:
263 rcu_read_unlock();
264out:
265 return error;
266}
267
268/*
269 * Ugh. To avoid negative return values, "getpriority()" will
270 * not return the normal nice-value, but a negated value that
271 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
272 * to stay compatible.
273 */
274SYSCALL_DEFINE2(getpriority, int, which, int, who)
275{
276 struct task_struct *g, *p;
277 struct user_struct *user;
278 const struct cred *cred = current_cred();
279 long niceval, retval = -ESRCH;
280 struct pid *pgrp;
281 kuid_t uid;
282
283 if (which > PRIO_USER || which < PRIO_PROCESS)
284 return -EINVAL;
285
286 rcu_read_lock();
287 switch (which) {
288 case PRIO_PROCESS:
289 if (who)
290 p = find_task_by_vpid(who);
291 else
292 p = current;
293 if (p) {
294 niceval = nice_to_rlimit(task_nice(p));
295 if (niceval > retval)
296 retval = niceval;
297 }
298 break;
299 case PRIO_PGRP:
300 if (who)
301 pgrp = find_vpid(who);
302 else
303 pgrp = task_pgrp(current);
304 read_lock(&tasklist_lock);
305 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
306 niceval = nice_to_rlimit(task_nice(p));
307 if (niceval > retval)
308 retval = niceval;
309 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
310 read_unlock(&tasklist_lock);
311 break;
312 case PRIO_USER:
313 uid = make_kuid(cred->user_ns, who);
314 user = cred->user;
315 if (!who)
316 uid = cred->uid;
317 else if (!uid_eq(uid, cred->uid)) {
318 user = find_user(uid);
319 if (!user)
320 goto out_unlock; /* No processes for this user */
321 }
322 for_each_process_thread(g, p) {
323 if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) {
324 niceval = nice_to_rlimit(task_nice(p));
325 if (niceval > retval)
326 retval = niceval;
327 }
328 }
329 if (!uid_eq(uid, cred->uid))
330 free_uid(user); /* for find_user() */
331 break;
332 }
333out_unlock:
334 rcu_read_unlock();
335
336 return retval;
337}
338
339/*
340 * Unprivileged users may change the real gid to the effective gid
341 * or vice versa. (BSD-style)
342 *
343 * If you set the real gid at all, or set the effective gid to a value not
344 * equal to the real gid, then the saved gid is set to the new effective gid.
345 *
346 * This makes it possible for a setgid program to completely drop its
347 * privileges, which is often a useful assertion to make when you are doing
348 * a security audit over a program.
349 *
350 * The general idea is that a program which uses just setregid() will be
351 * 100% compatible with BSD. A program which uses just setgid() will be
352 * 100% compatible with POSIX with saved IDs.
353 *
354 * SMP: There are not races, the GIDs are checked only by filesystem
355 * operations (as far as semantic preservation is concerned).
356 */
357#ifdef CONFIG_MULTIUSER
358long __sys_setregid(gid_t rgid, gid_t egid)
359{
360 struct user_namespace *ns = current_user_ns();
361 const struct cred *old;
362 struct cred *new;
363 int retval;
364 kgid_t krgid, kegid;
365
366 krgid = make_kgid(ns, rgid);
367 kegid = make_kgid(ns, egid);
368
369 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
370 return -EINVAL;
371 if ((egid != (gid_t) -1) && !gid_valid(kegid))
372 return -EINVAL;
373
374 new = prepare_creds();
375 if (!new)
376 return -ENOMEM;
377 old = current_cred();
378
379 retval = -EPERM;
380 if (rgid != (gid_t) -1) {
381 if (gid_eq(old->gid, krgid) ||
382 gid_eq(old->egid, krgid) ||
383 ns_capable_setid(old->user_ns, CAP_SETGID))
384 new->gid = krgid;
385 else
386 goto error;
387 }
388 if (egid != (gid_t) -1) {
389 if (gid_eq(old->gid, kegid) ||
390 gid_eq(old->egid, kegid) ||
391 gid_eq(old->sgid, kegid) ||
392 ns_capable_setid(old->user_ns, CAP_SETGID))
393 new->egid = kegid;
394 else
395 goto error;
396 }
397
398 if (rgid != (gid_t) -1 ||
399 (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
400 new->sgid = new->egid;
401 new->fsgid = new->egid;
402
403 retval = security_task_fix_setgid(new, old, LSM_SETID_RE);
404 if (retval < 0)
405 goto error;
406
407 return commit_creds(new);
408
409error:
410 abort_creds(new);
411 return retval;
412}
413
414SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
415{
416 return __sys_setregid(rgid, egid);
417}
418
419/*
420 * setgid() is implemented like SysV w/ SAVED_IDS
421 *
422 * SMP: Same implicit races as above.
423 */
424long __sys_setgid(gid_t gid)
425{
426 struct user_namespace *ns = current_user_ns();
427 const struct cred *old;
428 struct cred *new;
429 int retval;
430 kgid_t kgid;
431
432 kgid = make_kgid(ns, gid);
433 if (!gid_valid(kgid))
434 return -EINVAL;
435
436 new = prepare_creds();
437 if (!new)
438 return -ENOMEM;
439 old = current_cred();
440
441 retval = -EPERM;
442 if (ns_capable_setid(old->user_ns, CAP_SETGID))
443 new->gid = new->egid = new->sgid = new->fsgid = kgid;
444 else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
445 new->egid = new->fsgid = kgid;
446 else
447 goto error;
448
449 retval = security_task_fix_setgid(new, old, LSM_SETID_ID);
450 if (retval < 0)
451 goto error;
452
453 return commit_creds(new);
454
455error:
456 abort_creds(new);
457 return retval;
458}
459
460SYSCALL_DEFINE1(setgid, gid_t, gid)
461{
462 return __sys_setgid(gid);
463}
464
465/*
466 * change the user struct in a credentials set to match the new UID
467 */
468static int set_user(struct cred *new)
469{
470 struct user_struct *new_user;
471
472 new_user = alloc_uid(new->uid);
473 if (!new_user)
474 return -EAGAIN;
475
476 free_uid(new->user);
477 new->user = new_user;
478 return 0;
479}
480
481static void flag_nproc_exceeded(struct cred *new)
482{
483 if (new->ucounts == current_ucounts())
484 return;
485
486 /*
487 * We don't fail in case of NPROC limit excess here because too many
488 * poorly written programs don't check set*uid() return code, assuming
489 * it never fails if called by root. We may still enforce NPROC limit
490 * for programs doing set*uid()+execve() by harmlessly deferring the
491 * failure to the execve() stage.
492 */
493 if (is_ucounts_overlimit(new->ucounts, UCOUNT_RLIMIT_NPROC, rlimit(RLIMIT_NPROC)) &&
494 new->user != INIT_USER)
495 current->flags |= PF_NPROC_EXCEEDED;
496 else
497 current->flags &= ~PF_NPROC_EXCEEDED;
498}
499
500/*
501 * Unprivileged users may change the real uid to the effective uid
502 * or vice versa. (BSD-style)
503 *
504 * If you set the real uid at all, or set the effective uid to a value not
505 * equal to the real uid, then the saved uid is set to the new effective uid.
506 *
507 * This makes it possible for a setuid program to completely drop its
508 * privileges, which is often a useful assertion to make when you are doing
509 * a security audit over a program.
510 *
511 * The general idea is that a program which uses just setreuid() will be
512 * 100% compatible with BSD. A program which uses just setuid() will be
513 * 100% compatible with POSIX with saved IDs.
514 */
515long __sys_setreuid(uid_t ruid, uid_t euid)
516{
517 struct user_namespace *ns = current_user_ns();
518 const struct cred *old;
519 struct cred *new;
520 int retval;
521 kuid_t kruid, keuid;
522
523 kruid = make_kuid(ns, ruid);
524 keuid = make_kuid(ns, euid);
525
526 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
527 return -EINVAL;
528 if ((euid != (uid_t) -1) && !uid_valid(keuid))
529 return -EINVAL;
530
531 new = prepare_creds();
532 if (!new)
533 return -ENOMEM;
534 old = current_cred();
535
536 retval = -EPERM;
537 if (ruid != (uid_t) -1) {
538 new->uid = kruid;
539 if (!uid_eq(old->uid, kruid) &&
540 !uid_eq(old->euid, kruid) &&
541 !ns_capable_setid(old->user_ns, CAP_SETUID))
542 goto error;
543 }
544
545 if (euid != (uid_t) -1) {
546 new->euid = keuid;
547 if (!uid_eq(old->uid, keuid) &&
548 !uid_eq(old->euid, keuid) &&
549 !uid_eq(old->suid, keuid) &&
550 !ns_capable_setid(old->user_ns, CAP_SETUID))
551 goto error;
552 }
553
554 if (!uid_eq(new->uid, old->uid)) {
555 retval = set_user(new);
556 if (retval < 0)
557 goto error;
558 }
559 if (ruid != (uid_t) -1 ||
560 (euid != (uid_t) -1 && !uid_eq(keuid, old->uid)))
561 new->suid = new->euid;
562 new->fsuid = new->euid;
563
564 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
565 if (retval < 0)
566 goto error;
567
568 retval = set_cred_ucounts(new);
569 if (retval < 0)
570 goto error;
571
572 flag_nproc_exceeded(new);
573 return commit_creds(new);
574
575error:
576 abort_creds(new);
577 return retval;
578}
579
580SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
581{
582 return __sys_setreuid(ruid, euid);
583}
584
585/*
586 * setuid() is implemented like SysV with SAVED_IDS
587 *
588 * Note that SAVED_ID's is deficient in that a setuid root program
589 * like sendmail, for example, cannot set its uid to be a normal
590 * user and then switch back, because if you're root, setuid() sets
591 * the saved uid too. If you don't like this, blame the bright people
592 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
593 * will allow a root program to temporarily drop privileges and be able to
594 * regain them by swapping the real and effective uid.
595 */
596long __sys_setuid(uid_t uid)
597{
598 struct user_namespace *ns = current_user_ns();
599 const struct cred *old;
600 struct cred *new;
601 int retval;
602 kuid_t kuid;
603
604 kuid = make_kuid(ns, uid);
605 if (!uid_valid(kuid))
606 return -EINVAL;
607
608 new = prepare_creds();
609 if (!new)
610 return -ENOMEM;
611 old = current_cred();
612
613 retval = -EPERM;
614 if (ns_capable_setid(old->user_ns, CAP_SETUID)) {
615 new->suid = new->uid = kuid;
616 if (!uid_eq(kuid, old->uid)) {
617 retval = set_user(new);
618 if (retval < 0)
619 goto error;
620 }
621 } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) {
622 goto error;
623 }
624
625 new->fsuid = new->euid = kuid;
626
627 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
628 if (retval < 0)
629 goto error;
630
631 retval = set_cred_ucounts(new);
632 if (retval < 0)
633 goto error;
634
635 flag_nproc_exceeded(new);
636 return commit_creds(new);
637
638error:
639 abort_creds(new);
640 return retval;
641}
642
643SYSCALL_DEFINE1(setuid, uid_t, uid)
644{
645 return __sys_setuid(uid);
646}
647
648
649/*
650 * This function implements a generic ability to update ruid, euid,
651 * and suid. This allows you to implement the 4.4 compatible seteuid().
652 */
653long __sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
654{
655 struct user_namespace *ns = current_user_ns();
656 const struct cred *old;
657 struct cred *new;
658 int retval;
659 kuid_t kruid, keuid, ksuid;
660
661 kruid = make_kuid(ns, ruid);
662 keuid = make_kuid(ns, euid);
663 ksuid = make_kuid(ns, suid);
664
665 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
666 return -EINVAL;
667
668 if ((euid != (uid_t) -1) && !uid_valid(keuid))
669 return -EINVAL;
670
671 if ((suid != (uid_t) -1) && !uid_valid(ksuid))
672 return -EINVAL;
673
674 new = prepare_creds();
675 if (!new)
676 return -ENOMEM;
677
678 old = current_cred();
679
680 retval = -EPERM;
681 if (!ns_capable_setid(old->user_ns, CAP_SETUID)) {
682 if (ruid != (uid_t) -1 && !uid_eq(kruid, old->uid) &&
683 !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid))
684 goto error;
685 if (euid != (uid_t) -1 && !uid_eq(keuid, old->uid) &&
686 !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid))
687 goto error;
688 if (suid != (uid_t) -1 && !uid_eq(ksuid, old->uid) &&
689 !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid))
690 goto error;
691 }
692
693 if (ruid != (uid_t) -1) {
694 new->uid = kruid;
695 if (!uid_eq(kruid, old->uid)) {
696 retval = set_user(new);
697 if (retval < 0)
698 goto error;
699 }
700 }
701 if (euid != (uid_t) -1)
702 new->euid = keuid;
703 if (suid != (uid_t) -1)
704 new->suid = ksuid;
705 new->fsuid = new->euid;
706
707 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
708 if (retval < 0)
709 goto error;
710
711 retval = set_cred_ucounts(new);
712 if (retval < 0)
713 goto error;
714
715 flag_nproc_exceeded(new);
716 return commit_creds(new);
717
718error:
719 abort_creds(new);
720 return retval;
721}
722
723SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
724{
725 return __sys_setresuid(ruid, euid, suid);
726}
727
728SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
729{
730 const struct cred *cred = current_cred();
731 int retval;
732 uid_t ruid, euid, suid;
733
734 ruid = from_kuid_munged(cred->user_ns, cred->uid);
735 euid = from_kuid_munged(cred->user_ns, cred->euid);
736 suid = from_kuid_munged(cred->user_ns, cred->suid);
737
738 retval = put_user(ruid, ruidp);
739 if (!retval) {
740 retval = put_user(euid, euidp);
741 if (!retval)
742 return put_user(suid, suidp);
743 }
744 return retval;
745}
746
747/*
748 * Same as above, but for rgid, egid, sgid.
749 */
750long __sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
751{
752 struct user_namespace *ns = current_user_ns();
753 const struct cred *old;
754 struct cred *new;
755 int retval;
756 kgid_t krgid, kegid, ksgid;
757
758 krgid = make_kgid(ns, rgid);
759 kegid = make_kgid(ns, egid);
760 ksgid = make_kgid(ns, sgid);
761
762 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
763 return -EINVAL;
764 if ((egid != (gid_t) -1) && !gid_valid(kegid))
765 return -EINVAL;
766 if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
767 return -EINVAL;
768
769 new = prepare_creds();
770 if (!new)
771 return -ENOMEM;
772 old = current_cred();
773
774 retval = -EPERM;
775 if (!ns_capable_setid(old->user_ns, CAP_SETGID)) {
776 if (rgid != (gid_t) -1 && !gid_eq(krgid, old->gid) &&
777 !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid))
778 goto error;
779 if (egid != (gid_t) -1 && !gid_eq(kegid, old->gid) &&
780 !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid))
781 goto error;
782 if (sgid != (gid_t) -1 && !gid_eq(ksgid, old->gid) &&
783 !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid))
784 goto error;
785 }
786
787 if (rgid != (gid_t) -1)
788 new->gid = krgid;
789 if (egid != (gid_t) -1)
790 new->egid = kegid;
791 if (sgid != (gid_t) -1)
792 new->sgid = ksgid;
793 new->fsgid = new->egid;
794
795 retval = security_task_fix_setgid(new, old, LSM_SETID_RES);
796 if (retval < 0)
797 goto error;
798
799 return commit_creds(new);
800
801error:
802 abort_creds(new);
803 return retval;
804}
805
806SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
807{
808 return __sys_setresgid(rgid, egid, sgid);
809}
810
811SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
812{
813 const struct cred *cred = current_cred();
814 int retval;
815 gid_t rgid, egid, sgid;
816
817 rgid = from_kgid_munged(cred->user_ns, cred->gid);
818 egid = from_kgid_munged(cred->user_ns, cred->egid);
819 sgid = from_kgid_munged(cred->user_ns, cred->sgid);
820
821 retval = put_user(rgid, rgidp);
822 if (!retval) {
823 retval = put_user(egid, egidp);
824 if (!retval)
825 retval = put_user(sgid, sgidp);
826 }
827
828 return retval;
829}
830
831
832/*
833 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
834 * is used for "access()" and for the NFS daemon (letting nfsd stay at
835 * whatever uid it wants to). It normally shadows "euid", except when
836 * explicitly set by setfsuid() or for access..
837 */
838long __sys_setfsuid(uid_t uid)
839{
840 const struct cred *old;
841 struct cred *new;
842 uid_t old_fsuid;
843 kuid_t kuid;
844
845 old = current_cred();
846 old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
847
848 kuid = make_kuid(old->user_ns, uid);
849 if (!uid_valid(kuid))
850 return old_fsuid;
851
852 new = prepare_creds();
853 if (!new)
854 return old_fsuid;
855
856 if (uid_eq(kuid, old->uid) || uid_eq(kuid, old->euid) ||
857 uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
858 ns_capable_setid(old->user_ns, CAP_SETUID)) {
859 if (!uid_eq(kuid, old->fsuid)) {
860 new->fsuid = kuid;
861 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
862 goto change_okay;
863 }
864 }
865
866 abort_creds(new);
867 return old_fsuid;
868
869change_okay:
870 commit_creds(new);
871 return old_fsuid;
872}
873
874SYSCALL_DEFINE1(setfsuid, uid_t, uid)
875{
876 return __sys_setfsuid(uid);
877}
878
879/*
880 * Samma på svenska..
881 */
882long __sys_setfsgid(gid_t gid)
883{
884 const struct cred *old;
885 struct cred *new;
886 gid_t old_fsgid;
887 kgid_t kgid;
888
889 old = current_cred();
890 old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
891
892 kgid = make_kgid(old->user_ns, gid);
893 if (!gid_valid(kgid))
894 return old_fsgid;
895
896 new = prepare_creds();
897 if (!new)
898 return old_fsgid;
899
900 if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->egid) ||
901 gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
902 ns_capable_setid(old->user_ns, CAP_SETGID)) {
903 if (!gid_eq(kgid, old->fsgid)) {
904 new->fsgid = kgid;
905 if (security_task_fix_setgid(new,old,LSM_SETID_FS) == 0)
906 goto change_okay;
907 }
908 }
909
910 abort_creds(new);
911 return old_fsgid;
912
913change_okay:
914 commit_creds(new);
915 return old_fsgid;
916}
917
918SYSCALL_DEFINE1(setfsgid, gid_t, gid)
919{
920 return __sys_setfsgid(gid);
921}
922#endif /* CONFIG_MULTIUSER */
923
924/**
925 * sys_getpid - return the thread group id of the current process
926 *
927 * Note, despite the name, this returns the tgid not the pid. The tgid and
928 * the pid are identical unless CLONE_THREAD was specified on clone() in
929 * which case the tgid is the same in all threads of the same group.
930 *
931 * This is SMP safe as current->tgid does not change.
932 */
933SYSCALL_DEFINE0(getpid)
934{
935 return task_tgid_vnr(current);
936}
937
938/* Thread ID - the internal kernel "pid" */
939SYSCALL_DEFINE0(gettid)
940{
941 return task_pid_vnr(current);
942}
943
944/*
945 * Accessing ->real_parent is not SMP-safe, it could
946 * change from under us. However, we can use a stale
947 * value of ->real_parent under rcu_read_lock(), see
948 * release_task()->call_rcu(delayed_put_task_struct).
949 */
950SYSCALL_DEFINE0(getppid)
951{
952 int pid;
953
954 rcu_read_lock();
955 pid = task_tgid_vnr(rcu_dereference(current->real_parent));
956 rcu_read_unlock();
957
958 return pid;
959}
960
961SYSCALL_DEFINE0(getuid)
962{
963 /* Only we change this so SMP safe */
964 return from_kuid_munged(current_user_ns(), current_uid());
965}
966
967SYSCALL_DEFINE0(geteuid)
968{
969 /* Only we change this so SMP safe */
970 return from_kuid_munged(current_user_ns(), current_euid());
971}
972
973SYSCALL_DEFINE0(getgid)
974{
975 /* Only we change this so SMP safe */
976 return from_kgid_munged(current_user_ns(), current_gid());
977}
978
979SYSCALL_DEFINE0(getegid)
980{
981 /* Only we change this so SMP safe */
982 return from_kgid_munged(current_user_ns(), current_egid());
983}
984
985static void do_sys_times(struct tms *tms)
986{
987 u64 tgutime, tgstime, cutime, cstime;
988
989 thread_group_cputime_adjusted(current, &tgutime, &tgstime);
990 cutime = current->signal->cutime;
991 cstime = current->signal->cstime;
992 tms->tms_utime = nsec_to_clock_t(tgutime);
993 tms->tms_stime = nsec_to_clock_t(tgstime);
994 tms->tms_cutime = nsec_to_clock_t(cutime);
995 tms->tms_cstime = nsec_to_clock_t(cstime);
996}
997
998SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
999{
1000 if (tbuf) {
1001 struct tms tmp;
1002
1003 do_sys_times(&tmp);
1004 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1005 return -EFAULT;
1006 }
1007 force_successful_syscall_return();
1008 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1009}
1010
1011#ifdef CONFIG_COMPAT
1012static compat_clock_t clock_t_to_compat_clock_t(clock_t x)
1013{
1014 return compat_jiffies_to_clock_t(clock_t_to_jiffies(x));
1015}
1016
1017COMPAT_SYSCALL_DEFINE1(times, struct compat_tms __user *, tbuf)
1018{
1019 if (tbuf) {
1020 struct tms tms;
1021 struct compat_tms tmp;
1022
1023 do_sys_times(&tms);
1024 /* Convert our struct tms to the compat version. */
1025 tmp.tms_utime = clock_t_to_compat_clock_t(tms.tms_utime);
1026 tmp.tms_stime = clock_t_to_compat_clock_t(tms.tms_stime);
1027 tmp.tms_cutime = clock_t_to_compat_clock_t(tms.tms_cutime);
1028 tmp.tms_cstime = clock_t_to_compat_clock_t(tms.tms_cstime);
1029 if (copy_to_user(tbuf, &tmp, sizeof(tmp)))
1030 return -EFAULT;
1031 }
1032 force_successful_syscall_return();
1033 return compat_jiffies_to_clock_t(jiffies);
1034}
1035#endif
1036
1037/*
1038 * This needs some heavy checking ...
1039 * I just haven't the stomach for it. I also don't fully
1040 * understand sessions/pgrp etc. Let somebody who does explain it.
1041 *
1042 * OK, I think I have the protection semantics right.... this is really
1043 * only important on a multi-user system anyway, to make sure one user
1044 * can't send a signal to a process owned by another. -TYT, 12/12/91
1045 *
1046 * !PF_FORKNOEXEC check to conform completely to POSIX.
1047 */
1048SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
1049{
1050 struct task_struct *p;
1051 struct task_struct *group_leader = current->group_leader;
1052 struct pid *pgrp;
1053 int err;
1054
1055 if (!pid)
1056 pid = task_pid_vnr(group_leader);
1057 if (!pgid)
1058 pgid = pid;
1059 if (pgid < 0)
1060 return -EINVAL;
1061 rcu_read_lock();
1062
1063 /* From this point forward we keep holding onto the tasklist lock
1064 * so that our parent does not change from under us. -DaveM
1065 */
1066 write_lock_irq(&tasklist_lock);
1067
1068 err = -ESRCH;
1069 p = find_task_by_vpid(pid);
1070 if (!p)
1071 goto out;
1072
1073 err = -EINVAL;
1074 if (!thread_group_leader(p))
1075 goto out;
1076
1077 if (same_thread_group(p->real_parent, group_leader)) {
1078 err = -EPERM;
1079 if (task_session(p) != task_session(group_leader))
1080 goto out;
1081 err = -EACCES;
1082 if (!(p->flags & PF_FORKNOEXEC))
1083 goto out;
1084 } else {
1085 err = -ESRCH;
1086 if (p != group_leader)
1087 goto out;
1088 }
1089
1090 err = -EPERM;
1091 if (p->signal->leader)
1092 goto out;
1093
1094 pgrp = task_pid(p);
1095 if (pgid != pid) {
1096 struct task_struct *g;
1097
1098 pgrp = find_vpid(pgid);
1099 g = pid_task(pgrp, PIDTYPE_PGID);
1100 if (!g || task_session(g) != task_session(group_leader))
1101 goto out;
1102 }
1103
1104 err = security_task_setpgid(p, pgid);
1105 if (err)
1106 goto out;
1107
1108 if (task_pgrp(p) != pgrp)
1109 change_pid(p, PIDTYPE_PGID, pgrp);
1110
1111 err = 0;
1112out:
1113 /* All paths lead to here, thus we are safe. -DaveM */
1114 write_unlock_irq(&tasklist_lock);
1115 rcu_read_unlock();
1116 return err;
1117}
1118
1119static int do_getpgid(pid_t pid)
1120{
1121 struct task_struct *p;
1122 struct pid *grp;
1123 int retval;
1124
1125 rcu_read_lock();
1126 if (!pid)
1127 grp = task_pgrp(current);
1128 else {
1129 retval = -ESRCH;
1130 p = find_task_by_vpid(pid);
1131 if (!p)
1132 goto out;
1133 grp = task_pgrp(p);
1134 if (!grp)
1135 goto out;
1136
1137 retval = security_task_getpgid(p);
1138 if (retval)
1139 goto out;
1140 }
1141 retval = pid_vnr(grp);
1142out:
1143 rcu_read_unlock();
1144 return retval;
1145}
1146
1147SYSCALL_DEFINE1(getpgid, pid_t, pid)
1148{
1149 return do_getpgid(pid);
1150}
1151
1152#ifdef __ARCH_WANT_SYS_GETPGRP
1153
1154SYSCALL_DEFINE0(getpgrp)
1155{
1156 return do_getpgid(0);
1157}
1158
1159#endif
1160
1161SYSCALL_DEFINE1(getsid, pid_t, pid)
1162{
1163 struct task_struct *p;
1164 struct pid *sid;
1165 int retval;
1166
1167 rcu_read_lock();
1168 if (!pid)
1169 sid = task_session(current);
1170 else {
1171 retval = -ESRCH;
1172 p = find_task_by_vpid(pid);
1173 if (!p)
1174 goto out;
1175 sid = task_session(p);
1176 if (!sid)
1177 goto out;
1178
1179 retval = security_task_getsid(p);
1180 if (retval)
1181 goto out;
1182 }
1183 retval = pid_vnr(sid);
1184out:
1185 rcu_read_unlock();
1186 return retval;
1187}
1188
1189static void set_special_pids(struct pid *pid)
1190{
1191 struct task_struct *curr = current->group_leader;
1192
1193 if (task_session(curr) != pid)
1194 change_pid(curr, PIDTYPE_SID, pid);
1195
1196 if (task_pgrp(curr) != pid)
1197 change_pid(curr, PIDTYPE_PGID, pid);
1198}
1199
1200int ksys_setsid(void)
1201{
1202 struct task_struct *group_leader = current->group_leader;
1203 struct pid *sid = task_pid(group_leader);
1204 pid_t session = pid_vnr(sid);
1205 int err = -EPERM;
1206
1207 write_lock_irq(&tasklist_lock);
1208 /* Fail if I am already a session leader */
1209 if (group_leader->signal->leader)
1210 goto out;
1211
1212 /* Fail if a process group id already exists that equals the
1213 * proposed session id.
1214 */
1215 if (pid_task(sid, PIDTYPE_PGID))
1216 goto out;
1217
1218 group_leader->signal->leader = 1;
1219 set_special_pids(sid);
1220
1221 proc_clear_tty(group_leader);
1222
1223 err = session;
1224out:
1225 write_unlock_irq(&tasklist_lock);
1226 if (err > 0) {
1227 proc_sid_connector(group_leader);
1228 sched_autogroup_create_attach(group_leader);
1229 }
1230 return err;
1231}
1232
1233SYSCALL_DEFINE0(setsid)
1234{
1235 return ksys_setsid();
1236}
1237
1238DECLARE_RWSEM(uts_sem);
1239
1240#ifdef COMPAT_UTS_MACHINE
1241#define override_architecture(name) \
1242 (personality(current->personality) == PER_LINUX32 && \
1243 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1244 sizeof(COMPAT_UTS_MACHINE)))
1245#else
1246#define override_architecture(name) 0
1247#endif
1248
1249/*
1250 * Work around broken programs that cannot handle "Linux 3.0".
1251 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1252 * And we map 4.x and later versions to 2.6.60+x, so 4.0/5.0/6.0/... would be
1253 * 2.6.60.
1254 */
1255static int override_release(char __user *release, size_t len)
1256{
1257 int ret = 0;
1258
1259 if (current->personality & UNAME26) {
1260 const char *rest = UTS_RELEASE;
1261 char buf[65] = { 0 };
1262 int ndots = 0;
1263 unsigned v;
1264 size_t copy;
1265
1266 while (*rest) {
1267 if (*rest == '.' && ++ndots >= 3)
1268 break;
1269 if (!isdigit(*rest) && *rest != '.')
1270 break;
1271 rest++;
1272 }
1273 v = LINUX_VERSION_PATCHLEVEL + 60;
1274 copy = clamp_t(size_t, len, 1, sizeof(buf));
1275 copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1276 ret = copy_to_user(release, buf, copy + 1);
1277 }
1278 return ret;
1279}
1280
1281SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1282{
1283 struct new_utsname tmp;
1284
1285 down_read(&uts_sem);
1286 memcpy(&tmp, utsname(), sizeof(tmp));
1287 up_read(&uts_sem);
1288 if (copy_to_user(name, &tmp, sizeof(tmp)))
1289 return -EFAULT;
1290
1291 if (override_release(name->release, sizeof(name->release)))
1292 return -EFAULT;
1293 if (override_architecture(name))
1294 return -EFAULT;
1295 return 0;
1296}
1297
1298#ifdef __ARCH_WANT_SYS_OLD_UNAME
1299/*
1300 * Old cruft
1301 */
1302SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1303{
1304 struct old_utsname tmp;
1305
1306 if (!name)
1307 return -EFAULT;
1308
1309 down_read(&uts_sem);
1310 memcpy(&tmp, utsname(), sizeof(tmp));
1311 up_read(&uts_sem);
1312 if (copy_to_user(name, &tmp, sizeof(tmp)))
1313 return -EFAULT;
1314
1315 if (override_release(name->release, sizeof(name->release)))
1316 return -EFAULT;
1317 if (override_architecture(name))
1318 return -EFAULT;
1319 return 0;
1320}
1321
1322SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1323{
1324 struct oldold_utsname tmp;
1325
1326 if (!name)
1327 return -EFAULT;
1328
1329 memset(&tmp, 0, sizeof(tmp));
1330
1331 down_read(&uts_sem);
1332 memcpy(&tmp.sysname, &utsname()->sysname, __OLD_UTS_LEN);
1333 memcpy(&tmp.nodename, &utsname()->nodename, __OLD_UTS_LEN);
1334 memcpy(&tmp.release, &utsname()->release, __OLD_UTS_LEN);
1335 memcpy(&tmp.version, &utsname()->version, __OLD_UTS_LEN);
1336 memcpy(&tmp.machine, &utsname()->machine, __OLD_UTS_LEN);
1337 up_read(&uts_sem);
1338 if (copy_to_user(name, &tmp, sizeof(tmp)))
1339 return -EFAULT;
1340
1341 if (override_architecture(name))
1342 return -EFAULT;
1343 if (override_release(name->release, sizeof(name->release)))
1344 return -EFAULT;
1345 return 0;
1346}
1347#endif
1348
1349SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1350{
1351 int errno;
1352 char tmp[__NEW_UTS_LEN];
1353
1354 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1355 return -EPERM;
1356
1357 if (len < 0 || len > __NEW_UTS_LEN)
1358 return -EINVAL;
1359 errno = -EFAULT;
1360 if (!copy_from_user(tmp, name, len)) {
1361 struct new_utsname *u;
1362
1363 down_write(&uts_sem);
1364 u = utsname();
1365 memcpy(u->nodename, tmp, len);
1366 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1367 errno = 0;
1368 uts_proc_notify(UTS_PROC_HOSTNAME);
1369 up_write(&uts_sem);
1370 }
1371 return errno;
1372}
1373
1374#ifdef __ARCH_WANT_SYS_GETHOSTNAME
1375
1376SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1377{
1378 int i;
1379 struct new_utsname *u;
1380 char tmp[__NEW_UTS_LEN + 1];
1381
1382 if (len < 0)
1383 return -EINVAL;
1384 down_read(&uts_sem);
1385 u = utsname();
1386 i = 1 + strlen(u->nodename);
1387 if (i > len)
1388 i = len;
1389 memcpy(tmp, u->nodename, i);
1390 up_read(&uts_sem);
1391 if (copy_to_user(name, tmp, i))
1392 return -EFAULT;
1393 return 0;
1394}
1395
1396#endif
1397
1398/*
1399 * Only setdomainname; getdomainname can be implemented by calling
1400 * uname()
1401 */
1402SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1403{
1404 int errno;
1405 char tmp[__NEW_UTS_LEN];
1406
1407 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1408 return -EPERM;
1409 if (len < 0 || len > __NEW_UTS_LEN)
1410 return -EINVAL;
1411
1412 errno = -EFAULT;
1413 if (!copy_from_user(tmp, name, len)) {
1414 struct new_utsname *u;
1415
1416 down_write(&uts_sem);
1417 u = utsname();
1418 memcpy(u->domainname, tmp, len);
1419 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1420 errno = 0;
1421 uts_proc_notify(UTS_PROC_DOMAINNAME);
1422 up_write(&uts_sem);
1423 }
1424 return errno;
1425}
1426
1427/* make sure you are allowed to change @tsk limits before calling this */
1428static int do_prlimit(struct task_struct *tsk, unsigned int resource,
1429 struct rlimit *new_rlim, struct rlimit *old_rlim)
1430{
1431 struct rlimit *rlim;
1432 int retval = 0;
1433
1434 if (resource >= RLIM_NLIMITS)
1435 return -EINVAL;
1436 if (new_rlim) {
1437 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1438 return -EINVAL;
1439 if (resource == RLIMIT_NOFILE &&
1440 new_rlim->rlim_max > sysctl_nr_open)
1441 return -EPERM;
1442 }
1443
1444 /* Holding a refcount on tsk protects tsk->signal from disappearing. */
1445 rlim = tsk->signal->rlim + resource;
1446 task_lock(tsk->group_leader);
1447 if (new_rlim) {
1448 /*
1449 * Keep the capable check against init_user_ns until cgroups can
1450 * contain all limits.
1451 */
1452 if (new_rlim->rlim_max > rlim->rlim_max &&
1453 !capable(CAP_SYS_RESOURCE))
1454 retval = -EPERM;
1455 if (!retval)
1456 retval = security_task_setrlimit(tsk, resource, new_rlim);
1457 }
1458 if (!retval) {
1459 if (old_rlim)
1460 *old_rlim = *rlim;
1461 if (new_rlim)
1462 *rlim = *new_rlim;
1463 }
1464 task_unlock(tsk->group_leader);
1465
1466 /*
1467 * RLIMIT_CPU handling. Arm the posix CPU timer if the limit is not
1468 * infinite. In case of RLIM_INFINITY the posix CPU timer code
1469 * ignores the rlimit.
1470 */
1471 if (!retval && new_rlim && resource == RLIMIT_CPU &&
1472 new_rlim->rlim_cur != RLIM_INFINITY &&
1473 IS_ENABLED(CONFIG_POSIX_TIMERS)) {
1474 /*
1475 * update_rlimit_cpu can fail if the task is exiting, but there
1476 * may be other tasks in the thread group that are not exiting,
1477 * and they need their cpu timers adjusted.
1478 *
1479 * The group_leader is the last task to be released, so if we
1480 * cannot update_rlimit_cpu on it, then the entire process is
1481 * exiting and we do not need to update at all.
1482 */
1483 update_rlimit_cpu(tsk->group_leader, new_rlim->rlim_cur);
1484 }
1485
1486 return retval;
1487}
1488
1489SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1490{
1491 struct rlimit value;
1492 int ret;
1493
1494 ret = do_prlimit(current, resource, NULL, &value);
1495 if (!ret)
1496 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1497
1498 return ret;
1499}
1500
1501#ifdef CONFIG_COMPAT
1502
1503COMPAT_SYSCALL_DEFINE2(setrlimit, unsigned int, resource,
1504 struct compat_rlimit __user *, rlim)
1505{
1506 struct rlimit r;
1507 struct compat_rlimit r32;
1508
1509 if (copy_from_user(&r32, rlim, sizeof(struct compat_rlimit)))
1510 return -EFAULT;
1511
1512 if (r32.rlim_cur == COMPAT_RLIM_INFINITY)
1513 r.rlim_cur = RLIM_INFINITY;
1514 else
1515 r.rlim_cur = r32.rlim_cur;
1516 if (r32.rlim_max == COMPAT_RLIM_INFINITY)
1517 r.rlim_max = RLIM_INFINITY;
1518 else
1519 r.rlim_max = r32.rlim_max;
1520 return do_prlimit(current, resource, &r, NULL);
1521}
1522
1523COMPAT_SYSCALL_DEFINE2(getrlimit, unsigned int, resource,
1524 struct compat_rlimit __user *, rlim)
1525{
1526 struct rlimit r;
1527 int ret;
1528
1529 ret = do_prlimit(current, resource, NULL, &r);
1530 if (!ret) {
1531 struct compat_rlimit r32;
1532 if (r.rlim_cur > COMPAT_RLIM_INFINITY)
1533 r32.rlim_cur = COMPAT_RLIM_INFINITY;
1534 else
1535 r32.rlim_cur = r.rlim_cur;
1536 if (r.rlim_max > COMPAT_RLIM_INFINITY)
1537 r32.rlim_max = COMPAT_RLIM_INFINITY;
1538 else
1539 r32.rlim_max = r.rlim_max;
1540
1541 if (copy_to_user(rlim, &r32, sizeof(struct compat_rlimit)))
1542 return -EFAULT;
1543 }
1544 return ret;
1545}
1546
1547#endif
1548
1549#ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1550
1551/*
1552 * Back compatibility for getrlimit. Needed for some apps.
1553 */
1554SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1555 struct rlimit __user *, rlim)
1556{
1557 struct rlimit x;
1558 if (resource >= RLIM_NLIMITS)
1559 return -EINVAL;
1560
1561 resource = array_index_nospec(resource, RLIM_NLIMITS);
1562 task_lock(current->group_leader);
1563 x = current->signal->rlim[resource];
1564 task_unlock(current->group_leader);
1565 if (x.rlim_cur > 0x7FFFFFFF)
1566 x.rlim_cur = 0x7FFFFFFF;
1567 if (x.rlim_max > 0x7FFFFFFF)
1568 x.rlim_max = 0x7FFFFFFF;
1569 return copy_to_user(rlim, &x, sizeof(x)) ? -EFAULT : 0;
1570}
1571
1572#ifdef CONFIG_COMPAT
1573COMPAT_SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1574 struct compat_rlimit __user *, rlim)
1575{
1576 struct rlimit r;
1577
1578 if (resource >= RLIM_NLIMITS)
1579 return -EINVAL;
1580
1581 resource = array_index_nospec(resource, RLIM_NLIMITS);
1582 task_lock(current->group_leader);
1583 r = current->signal->rlim[resource];
1584 task_unlock(current->group_leader);
1585 if (r.rlim_cur > 0x7FFFFFFF)
1586 r.rlim_cur = 0x7FFFFFFF;
1587 if (r.rlim_max > 0x7FFFFFFF)
1588 r.rlim_max = 0x7FFFFFFF;
1589
1590 if (put_user(r.rlim_cur, &rlim->rlim_cur) ||
1591 put_user(r.rlim_max, &rlim->rlim_max))
1592 return -EFAULT;
1593 return 0;
1594}
1595#endif
1596
1597#endif
1598
1599static inline bool rlim64_is_infinity(__u64 rlim64)
1600{
1601#if BITS_PER_LONG < 64
1602 return rlim64 >= ULONG_MAX;
1603#else
1604 return rlim64 == RLIM64_INFINITY;
1605#endif
1606}
1607
1608static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1609{
1610 if (rlim->rlim_cur == RLIM_INFINITY)
1611 rlim64->rlim_cur = RLIM64_INFINITY;
1612 else
1613 rlim64->rlim_cur = rlim->rlim_cur;
1614 if (rlim->rlim_max == RLIM_INFINITY)
1615 rlim64->rlim_max = RLIM64_INFINITY;
1616 else
1617 rlim64->rlim_max = rlim->rlim_max;
1618}
1619
1620static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1621{
1622 if (rlim64_is_infinity(rlim64->rlim_cur))
1623 rlim->rlim_cur = RLIM_INFINITY;
1624 else
1625 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1626 if (rlim64_is_infinity(rlim64->rlim_max))
1627 rlim->rlim_max = RLIM_INFINITY;
1628 else
1629 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1630}
1631
1632/* rcu lock must be held */
1633static int check_prlimit_permission(struct task_struct *task,
1634 unsigned int flags)
1635{
1636 const struct cred *cred = current_cred(), *tcred;
1637 bool id_match;
1638
1639 if (current == task)
1640 return 0;
1641
1642 tcred = __task_cred(task);
1643 id_match = (uid_eq(cred->uid, tcred->euid) &&
1644 uid_eq(cred->uid, tcred->suid) &&
1645 uid_eq(cred->uid, tcred->uid) &&
1646 gid_eq(cred->gid, tcred->egid) &&
1647 gid_eq(cred->gid, tcred->sgid) &&
1648 gid_eq(cred->gid, tcred->gid));
1649 if (!id_match && !ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
1650 return -EPERM;
1651
1652 return security_task_prlimit(cred, tcred, flags);
1653}
1654
1655SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1656 const struct rlimit64 __user *, new_rlim,
1657 struct rlimit64 __user *, old_rlim)
1658{
1659 struct rlimit64 old64, new64;
1660 struct rlimit old, new;
1661 struct task_struct *tsk;
1662 unsigned int checkflags = 0;
1663 int ret;
1664
1665 if (old_rlim)
1666 checkflags |= LSM_PRLIMIT_READ;
1667
1668 if (new_rlim) {
1669 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1670 return -EFAULT;
1671 rlim64_to_rlim(&new64, &new);
1672 checkflags |= LSM_PRLIMIT_WRITE;
1673 }
1674
1675 rcu_read_lock();
1676 tsk = pid ? find_task_by_vpid(pid) : current;
1677 if (!tsk) {
1678 rcu_read_unlock();
1679 return -ESRCH;
1680 }
1681 ret = check_prlimit_permission(tsk, checkflags);
1682 if (ret) {
1683 rcu_read_unlock();
1684 return ret;
1685 }
1686 get_task_struct(tsk);
1687 rcu_read_unlock();
1688
1689 ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1690 old_rlim ? &old : NULL);
1691
1692 if (!ret && old_rlim) {
1693 rlim_to_rlim64(&old, &old64);
1694 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1695 ret = -EFAULT;
1696 }
1697
1698 put_task_struct(tsk);
1699 return ret;
1700}
1701
1702SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1703{
1704 struct rlimit new_rlim;
1705
1706 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1707 return -EFAULT;
1708 return do_prlimit(current, resource, &new_rlim, NULL);
1709}
1710
1711/*
1712 * It would make sense to put struct rusage in the task_struct,
1713 * except that would make the task_struct be *really big*. After
1714 * task_struct gets moved into malloc'ed memory, it would
1715 * make sense to do this. It will make moving the rest of the information
1716 * a lot simpler! (Which we're not doing right now because we're not
1717 * measuring them yet).
1718 *
1719 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1720 * races with threads incrementing their own counters. But since word
1721 * reads are atomic, we either get new values or old values and we don't
1722 * care which for the sums. We always take the siglock to protect reading
1723 * the c* fields from p->signal from races with exit.c updating those
1724 * fields when reaping, so a sample either gets all the additions of a
1725 * given child after it's reaped, or none so this sample is before reaping.
1726 *
1727 * Locking:
1728 * We need to take the siglock for CHILDEREN, SELF and BOTH
1729 * for the cases current multithreaded, non-current single threaded
1730 * non-current multithreaded. Thread traversal is now safe with
1731 * the siglock held.
1732 * Strictly speaking, we donot need to take the siglock if we are current and
1733 * single threaded, as no one else can take our signal_struct away, no one
1734 * else can reap the children to update signal->c* counters, and no one else
1735 * can race with the signal-> fields. If we do not take any lock, the
1736 * signal-> fields could be read out of order while another thread was just
1737 * exiting. So we should place a read memory barrier when we avoid the lock.
1738 * On the writer side, write memory barrier is implied in __exit_signal
1739 * as __exit_signal releases the siglock spinlock after updating the signal->
1740 * fields. But we don't do this yet to keep things simple.
1741 *
1742 */
1743
1744static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1745{
1746 r->ru_nvcsw += t->nvcsw;
1747 r->ru_nivcsw += t->nivcsw;
1748 r->ru_minflt += t->min_flt;
1749 r->ru_majflt += t->maj_flt;
1750 r->ru_inblock += task_io_get_inblock(t);
1751 r->ru_oublock += task_io_get_oublock(t);
1752}
1753
1754void getrusage(struct task_struct *p, int who, struct rusage *r)
1755{
1756 struct task_struct *t;
1757 unsigned long flags;
1758 u64 tgutime, tgstime, utime, stime;
1759 unsigned long maxrss = 0;
1760
1761 memset((char *)r, 0, sizeof (*r));
1762 utime = stime = 0;
1763
1764 if (who == RUSAGE_THREAD) {
1765 task_cputime_adjusted(current, &utime, &stime);
1766 accumulate_thread_rusage(p, r);
1767 maxrss = p->signal->maxrss;
1768 goto out;
1769 }
1770
1771 if (!lock_task_sighand(p, &flags))
1772 return;
1773
1774 switch (who) {
1775 case RUSAGE_BOTH:
1776 case RUSAGE_CHILDREN:
1777 utime = p->signal->cutime;
1778 stime = p->signal->cstime;
1779 r->ru_nvcsw = p->signal->cnvcsw;
1780 r->ru_nivcsw = p->signal->cnivcsw;
1781 r->ru_minflt = p->signal->cmin_flt;
1782 r->ru_majflt = p->signal->cmaj_flt;
1783 r->ru_inblock = p->signal->cinblock;
1784 r->ru_oublock = p->signal->coublock;
1785 maxrss = p->signal->cmaxrss;
1786
1787 if (who == RUSAGE_CHILDREN)
1788 break;
1789 fallthrough;
1790
1791 case RUSAGE_SELF:
1792 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1793 utime += tgutime;
1794 stime += tgstime;
1795 r->ru_nvcsw += p->signal->nvcsw;
1796 r->ru_nivcsw += p->signal->nivcsw;
1797 r->ru_minflt += p->signal->min_flt;
1798 r->ru_majflt += p->signal->maj_flt;
1799 r->ru_inblock += p->signal->inblock;
1800 r->ru_oublock += p->signal->oublock;
1801 if (maxrss < p->signal->maxrss)
1802 maxrss = p->signal->maxrss;
1803 t = p;
1804 do {
1805 accumulate_thread_rusage(t, r);
1806 } while_each_thread(p, t);
1807 break;
1808
1809 default:
1810 BUG();
1811 }
1812 unlock_task_sighand(p, &flags);
1813
1814out:
1815 r->ru_utime = ns_to_kernel_old_timeval(utime);
1816 r->ru_stime = ns_to_kernel_old_timeval(stime);
1817
1818 if (who != RUSAGE_CHILDREN) {
1819 struct mm_struct *mm = get_task_mm(p);
1820
1821 if (mm) {
1822 setmax_mm_hiwater_rss(&maxrss, mm);
1823 mmput(mm);
1824 }
1825 }
1826 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1827}
1828
1829SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1830{
1831 struct rusage r;
1832
1833 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1834 who != RUSAGE_THREAD)
1835 return -EINVAL;
1836
1837 getrusage(current, who, &r);
1838 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1839}
1840
1841#ifdef CONFIG_COMPAT
1842COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru)
1843{
1844 struct rusage r;
1845
1846 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1847 who != RUSAGE_THREAD)
1848 return -EINVAL;
1849
1850 getrusage(current, who, &r);
1851 return put_compat_rusage(&r, ru);
1852}
1853#endif
1854
1855SYSCALL_DEFINE1(umask, int, mask)
1856{
1857 mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1858 return mask;
1859}
1860
1861static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1862{
1863 struct fd exe;
1864 struct inode *inode;
1865 int err;
1866
1867 exe = fdget(fd);
1868 if (!exe.file)
1869 return -EBADF;
1870
1871 inode = file_inode(exe.file);
1872
1873 /*
1874 * Because the original mm->exe_file points to executable file, make
1875 * sure that this one is executable as well, to avoid breaking an
1876 * overall picture.
1877 */
1878 err = -EACCES;
1879 if (!S_ISREG(inode->i_mode) || path_noexec(&exe.file->f_path))
1880 goto exit;
1881
1882 err = file_permission(exe.file, MAY_EXEC);
1883 if (err)
1884 goto exit;
1885
1886 err = replace_mm_exe_file(mm, exe.file);
1887exit:
1888 fdput(exe);
1889 return err;
1890}
1891
1892/*
1893 * Check arithmetic relations of passed addresses.
1894 *
1895 * WARNING: we don't require any capability here so be very careful
1896 * in what is allowed for modification from userspace.
1897 */
1898static int validate_prctl_map_addr(struct prctl_mm_map *prctl_map)
1899{
1900 unsigned long mmap_max_addr = TASK_SIZE;
1901 int error = -EINVAL, i;
1902
1903 static const unsigned char offsets[] = {
1904 offsetof(struct prctl_mm_map, start_code),
1905 offsetof(struct prctl_mm_map, end_code),
1906 offsetof(struct prctl_mm_map, start_data),
1907 offsetof(struct prctl_mm_map, end_data),
1908 offsetof(struct prctl_mm_map, start_brk),
1909 offsetof(struct prctl_mm_map, brk),
1910 offsetof(struct prctl_mm_map, start_stack),
1911 offsetof(struct prctl_mm_map, arg_start),
1912 offsetof(struct prctl_mm_map, arg_end),
1913 offsetof(struct prctl_mm_map, env_start),
1914 offsetof(struct prctl_mm_map, env_end),
1915 };
1916
1917 /*
1918 * Make sure the members are not somewhere outside
1919 * of allowed address space.
1920 */
1921 for (i = 0; i < ARRAY_SIZE(offsets); i++) {
1922 u64 val = *(u64 *)((char *)prctl_map + offsets[i]);
1923
1924 if ((unsigned long)val >= mmap_max_addr ||
1925 (unsigned long)val < mmap_min_addr)
1926 goto out;
1927 }
1928
1929 /*
1930 * Make sure the pairs are ordered.
1931 */
1932#define __prctl_check_order(__m1, __op, __m2) \
1933 ((unsigned long)prctl_map->__m1 __op \
1934 (unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1935 error = __prctl_check_order(start_code, <, end_code);
1936 error |= __prctl_check_order(start_data,<=, end_data);
1937 error |= __prctl_check_order(start_brk, <=, brk);
1938 error |= __prctl_check_order(arg_start, <=, arg_end);
1939 error |= __prctl_check_order(env_start, <=, env_end);
1940 if (error)
1941 goto out;
1942#undef __prctl_check_order
1943
1944 error = -EINVAL;
1945
1946 /*
1947 * Neither we should allow to override limits if they set.
1948 */
1949 if (check_data_rlimit(rlimit(RLIMIT_DATA), prctl_map->brk,
1950 prctl_map->start_brk, prctl_map->end_data,
1951 prctl_map->start_data))
1952 goto out;
1953
1954 error = 0;
1955out:
1956 return error;
1957}
1958
1959#ifdef CONFIG_CHECKPOINT_RESTORE
1960static int prctl_set_mm_map(int opt, const void __user *addr, unsigned long data_size)
1961{
1962 struct prctl_mm_map prctl_map = { .exe_fd = (u32)-1, };
1963 unsigned long user_auxv[AT_VECTOR_SIZE];
1964 struct mm_struct *mm = current->mm;
1965 int error;
1966
1967 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
1968 BUILD_BUG_ON(sizeof(struct prctl_mm_map) > 256);
1969
1970 if (opt == PR_SET_MM_MAP_SIZE)
1971 return put_user((unsigned int)sizeof(prctl_map),
1972 (unsigned int __user *)addr);
1973
1974 if (data_size != sizeof(prctl_map))
1975 return -EINVAL;
1976
1977 if (copy_from_user(&prctl_map, addr, sizeof(prctl_map)))
1978 return -EFAULT;
1979
1980 error = validate_prctl_map_addr(&prctl_map);
1981 if (error)
1982 return error;
1983
1984 if (prctl_map.auxv_size) {
1985 /*
1986 * Someone is trying to cheat the auxv vector.
1987 */
1988 if (!prctl_map.auxv ||
1989 prctl_map.auxv_size > sizeof(mm->saved_auxv))
1990 return -EINVAL;
1991
1992 memset(user_auxv, 0, sizeof(user_auxv));
1993 if (copy_from_user(user_auxv,
1994 (const void __user *)prctl_map.auxv,
1995 prctl_map.auxv_size))
1996 return -EFAULT;
1997
1998 /* Last entry must be AT_NULL as specification requires */
1999 user_auxv[AT_VECTOR_SIZE - 2] = AT_NULL;
2000 user_auxv[AT_VECTOR_SIZE - 1] = AT_NULL;
2001 }
2002
2003 if (prctl_map.exe_fd != (u32)-1) {
2004 /*
2005 * Check if the current user is checkpoint/restore capable.
2006 * At the time of this writing, it checks for CAP_SYS_ADMIN
2007 * or CAP_CHECKPOINT_RESTORE.
2008 * Note that a user with access to ptrace can masquerade an
2009 * arbitrary program as any executable, even setuid ones.
2010 * This may have implications in the tomoyo subsystem.
2011 */
2012 if (!checkpoint_restore_ns_capable(current_user_ns()))
2013 return -EPERM;
2014
2015 error = prctl_set_mm_exe_file(mm, prctl_map.exe_fd);
2016 if (error)
2017 return error;
2018 }
2019
2020 /*
2021 * arg_lock protects concurrent updates but we still need mmap_lock for
2022 * read to exclude races with sys_brk.
2023 */
2024 mmap_read_lock(mm);
2025
2026 /*
2027 * We don't validate if these members are pointing to
2028 * real present VMAs because application may have correspond
2029 * VMAs already unmapped and kernel uses these members for statistics
2030 * output in procfs mostly, except
2031 *
2032 * - @start_brk/@brk which are used in do_brk_flags but kernel lookups
2033 * for VMAs when updating these members so anything wrong written
2034 * here cause kernel to swear at userspace program but won't lead
2035 * to any problem in kernel itself
2036 */
2037
2038 spin_lock(&mm->arg_lock);
2039 mm->start_code = prctl_map.start_code;
2040 mm->end_code = prctl_map.end_code;
2041 mm->start_data = prctl_map.start_data;
2042 mm->end_data = prctl_map.end_data;
2043 mm->start_brk = prctl_map.start_brk;
2044 mm->brk = prctl_map.brk;
2045 mm->start_stack = prctl_map.start_stack;
2046 mm->arg_start = prctl_map.arg_start;
2047 mm->arg_end = prctl_map.arg_end;
2048 mm->env_start = prctl_map.env_start;
2049 mm->env_end = prctl_map.env_end;
2050 spin_unlock(&mm->arg_lock);
2051
2052 /*
2053 * Note this update of @saved_auxv is lockless thus
2054 * if someone reads this member in procfs while we're
2055 * updating -- it may get partly updated results. It's
2056 * known and acceptable trade off: we leave it as is to
2057 * not introduce additional locks here making the kernel
2058 * more complex.
2059 */
2060 if (prctl_map.auxv_size)
2061 memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv));
2062
2063 mmap_read_unlock(mm);
2064 return 0;
2065}
2066#endif /* CONFIG_CHECKPOINT_RESTORE */
2067
2068static int prctl_set_auxv(struct mm_struct *mm, unsigned long addr,
2069 unsigned long len)
2070{
2071 /*
2072 * This doesn't move the auxiliary vector itself since it's pinned to
2073 * mm_struct, but it permits filling the vector with new values. It's
2074 * up to the caller to provide sane values here, otherwise userspace
2075 * tools which use this vector might be unhappy.
2076 */
2077 unsigned long user_auxv[AT_VECTOR_SIZE] = {};
2078
2079 if (len > sizeof(user_auxv))
2080 return -EINVAL;
2081
2082 if (copy_from_user(user_auxv, (const void __user *)addr, len))
2083 return -EFAULT;
2084
2085 /* Make sure the last entry is always AT_NULL */
2086 user_auxv[AT_VECTOR_SIZE - 2] = 0;
2087 user_auxv[AT_VECTOR_SIZE - 1] = 0;
2088
2089 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
2090
2091 task_lock(current);
2092 memcpy(mm->saved_auxv, user_auxv, len);
2093 task_unlock(current);
2094
2095 return 0;
2096}
2097
2098static int prctl_set_mm(int opt, unsigned long addr,
2099 unsigned long arg4, unsigned long arg5)
2100{
2101 struct mm_struct *mm = current->mm;
2102 struct prctl_mm_map prctl_map = {
2103 .auxv = NULL,
2104 .auxv_size = 0,
2105 .exe_fd = -1,
2106 };
2107 struct vm_area_struct *vma;
2108 int error;
2109
2110 if (arg5 || (arg4 && (opt != PR_SET_MM_AUXV &&
2111 opt != PR_SET_MM_MAP &&
2112 opt != PR_SET_MM_MAP_SIZE)))
2113 return -EINVAL;
2114
2115#ifdef CONFIG_CHECKPOINT_RESTORE
2116 if (opt == PR_SET_MM_MAP || opt == PR_SET_MM_MAP_SIZE)
2117 return prctl_set_mm_map(opt, (const void __user *)addr, arg4);
2118#endif
2119
2120 if (!capable(CAP_SYS_RESOURCE))
2121 return -EPERM;
2122
2123 if (opt == PR_SET_MM_EXE_FILE)
2124 return prctl_set_mm_exe_file(mm, (unsigned int)addr);
2125
2126 if (opt == PR_SET_MM_AUXV)
2127 return prctl_set_auxv(mm, addr, arg4);
2128
2129 if (addr >= TASK_SIZE || addr < mmap_min_addr)
2130 return -EINVAL;
2131
2132 error = -EINVAL;
2133
2134 /*
2135 * arg_lock protects concurrent updates of arg boundaries, we need
2136 * mmap_lock for a) concurrent sys_brk, b) finding VMA for addr
2137 * validation.
2138 */
2139 mmap_read_lock(mm);
2140 vma = find_vma(mm, addr);
2141
2142 spin_lock(&mm->arg_lock);
2143 prctl_map.start_code = mm->start_code;
2144 prctl_map.end_code = mm->end_code;
2145 prctl_map.start_data = mm->start_data;
2146 prctl_map.end_data = mm->end_data;
2147 prctl_map.start_brk = mm->start_brk;
2148 prctl_map.brk = mm->brk;
2149 prctl_map.start_stack = mm->start_stack;
2150 prctl_map.arg_start = mm->arg_start;
2151 prctl_map.arg_end = mm->arg_end;
2152 prctl_map.env_start = mm->env_start;
2153 prctl_map.env_end = mm->env_end;
2154
2155 switch (opt) {
2156 case PR_SET_MM_START_CODE:
2157 prctl_map.start_code = addr;
2158 break;
2159 case PR_SET_MM_END_CODE:
2160 prctl_map.end_code = addr;
2161 break;
2162 case PR_SET_MM_START_DATA:
2163 prctl_map.start_data = addr;
2164 break;
2165 case PR_SET_MM_END_DATA:
2166 prctl_map.end_data = addr;
2167 break;
2168 case PR_SET_MM_START_STACK:
2169 prctl_map.start_stack = addr;
2170 break;
2171 case PR_SET_MM_START_BRK:
2172 prctl_map.start_brk = addr;
2173 break;
2174 case PR_SET_MM_BRK:
2175 prctl_map.brk = addr;
2176 break;
2177 case PR_SET_MM_ARG_START:
2178 prctl_map.arg_start = addr;
2179 break;
2180 case PR_SET_MM_ARG_END:
2181 prctl_map.arg_end = addr;
2182 break;
2183 case PR_SET_MM_ENV_START:
2184 prctl_map.env_start = addr;
2185 break;
2186 case PR_SET_MM_ENV_END:
2187 prctl_map.env_end = addr;
2188 break;
2189 default:
2190 goto out;
2191 }
2192
2193 error = validate_prctl_map_addr(&prctl_map);
2194 if (error)
2195 goto out;
2196
2197 switch (opt) {
2198 /*
2199 * If command line arguments and environment
2200 * are placed somewhere else on stack, we can
2201 * set them up here, ARG_START/END to setup
2202 * command line arguments and ENV_START/END
2203 * for environment.
2204 */
2205 case PR_SET_MM_START_STACK:
2206 case PR_SET_MM_ARG_START:
2207 case PR_SET_MM_ARG_END:
2208 case PR_SET_MM_ENV_START:
2209 case PR_SET_MM_ENV_END:
2210 if (!vma) {
2211 error = -EFAULT;
2212 goto out;
2213 }
2214 }
2215
2216 mm->start_code = prctl_map.start_code;
2217 mm->end_code = prctl_map.end_code;
2218 mm->start_data = prctl_map.start_data;
2219 mm->end_data = prctl_map.end_data;
2220 mm->start_brk = prctl_map.start_brk;
2221 mm->brk = prctl_map.brk;
2222 mm->start_stack = prctl_map.start_stack;
2223 mm->arg_start = prctl_map.arg_start;
2224 mm->arg_end = prctl_map.arg_end;
2225 mm->env_start = prctl_map.env_start;
2226 mm->env_end = prctl_map.env_end;
2227
2228 error = 0;
2229out:
2230 spin_unlock(&mm->arg_lock);
2231 mmap_read_unlock(mm);
2232 return error;
2233}
2234
2235#ifdef CONFIG_CHECKPOINT_RESTORE
2236static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr)
2237{
2238 return put_user(me->clear_child_tid, tid_addr);
2239}
2240#else
2241static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr)
2242{
2243 return -EINVAL;
2244}
2245#endif
2246
2247static int propagate_has_child_subreaper(struct task_struct *p, void *data)
2248{
2249 /*
2250 * If task has has_child_subreaper - all its descendants
2251 * already have these flag too and new descendants will
2252 * inherit it on fork, skip them.
2253 *
2254 * If we've found child_reaper - skip descendants in
2255 * it's subtree as they will never get out pidns.
2256 */
2257 if (p->signal->has_child_subreaper ||
2258 is_child_reaper(task_pid(p)))
2259 return 0;
2260
2261 p->signal->has_child_subreaper = 1;
2262 return 1;
2263}
2264
2265int __weak arch_prctl_spec_ctrl_get(struct task_struct *t, unsigned long which)
2266{
2267 return -EINVAL;
2268}
2269
2270int __weak arch_prctl_spec_ctrl_set(struct task_struct *t, unsigned long which,
2271 unsigned long ctrl)
2272{
2273 return -EINVAL;
2274}
2275
2276#define PR_IO_FLUSHER (PF_MEMALLOC_NOIO | PF_LOCAL_THROTTLE)
2277
2278#ifdef CONFIG_ANON_VMA_NAME
2279
2280#define ANON_VMA_NAME_MAX_LEN 80
2281#define ANON_VMA_NAME_INVALID_CHARS "\\`$[]"
2282
2283static inline bool is_valid_name_char(char ch)
2284{
2285 /* printable ascii characters, excluding ANON_VMA_NAME_INVALID_CHARS */
2286 return ch > 0x1f && ch < 0x7f &&
2287 !strchr(ANON_VMA_NAME_INVALID_CHARS, ch);
2288}
2289
2290static int prctl_set_vma(unsigned long opt, unsigned long addr,
2291 unsigned long size, unsigned long arg)
2292{
2293 struct mm_struct *mm = current->mm;
2294 const char __user *uname;
2295 struct anon_vma_name *anon_name = NULL;
2296 int error;
2297
2298 switch (opt) {
2299 case PR_SET_VMA_ANON_NAME:
2300 uname = (const char __user *)arg;
2301 if (uname) {
2302 char *name, *pch;
2303
2304 name = strndup_user(uname, ANON_VMA_NAME_MAX_LEN);
2305 if (IS_ERR(name))
2306 return PTR_ERR(name);
2307
2308 for (pch = name; *pch != '\0'; pch++) {
2309 if (!is_valid_name_char(*pch)) {
2310 kfree(name);
2311 return -EINVAL;
2312 }
2313 }
2314 /* anon_vma has its own copy */
2315 anon_name = anon_vma_name_alloc(name);
2316 kfree(name);
2317 if (!anon_name)
2318 return -ENOMEM;
2319
2320 }
2321
2322 mmap_write_lock(mm);
2323 error = madvise_set_anon_name(mm, addr, size, anon_name);
2324 mmap_write_unlock(mm);
2325 anon_vma_name_put(anon_name);
2326 break;
2327 default:
2328 error = -EINVAL;
2329 }
2330
2331 return error;
2332}
2333
2334#else /* CONFIG_ANON_VMA_NAME */
2335static int prctl_set_vma(unsigned long opt, unsigned long start,
2336 unsigned long size, unsigned long arg)
2337{
2338 return -EINVAL;
2339}
2340#endif /* CONFIG_ANON_VMA_NAME */
2341
2342SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2343 unsigned long, arg4, unsigned long, arg5)
2344{
2345 struct task_struct *me = current;
2346 unsigned char comm[sizeof(me->comm)];
2347 long error;
2348
2349 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2350 if (error != -ENOSYS)
2351 return error;
2352
2353 error = 0;
2354 switch (option) {
2355 case PR_SET_PDEATHSIG:
2356 if (!valid_signal(arg2)) {
2357 error = -EINVAL;
2358 break;
2359 }
2360 me->pdeath_signal = arg2;
2361 break;
2362 case PR_GET_PDEATHSIG:
2363 error = put_user(me->pdeath_signal, (int __user *)arg2);
2364 break;
2365 case PR_GET_DUMPABLE:
2366 error = get_dumpable(me->mm);
2367 break;
2368 case PR_SET_DUMPABLE:
2369 if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) {
2370 error = -EINVAL;
2371 break;
2372 }
2373 set_dumpable(me->mm, arg2);
2374 break;
2375
2376 case PR_SET_UNALIGN:
2377 error = SET_UNALIGN_CTL(me, arg2);
2378 break;
2379 case PR_GET_UNALIGN:
2380 error = GET_UNALIGN_CTL(me, arg2);
2381 break;
2382 case PR_SET_FPEMU:
2383 error = SET_FPEMU_CTL(me, arg2);
2384 break;
2385 case PR_GET_FPEMU:
2386 error = GET_FPEMU_CTL(me, arg2);
2387 break;
2388 case PR_SET_FPEXC:
2389 error = SET_FPEXC_CTL(me, arg2);
2390 break;
2391 case PR_GET_FPEXC:
2392 error = GET_FPEXC_CTL(me, arg2);
2393 break;
2394 case PR_GET_TIMING:
2395 error = PR_TIMING_STATISTICAL;
2396 break;
2397 case PR_SET_TIMING:
2398 if (arg2 != PR_TIMING_STATISTICAL)
2399 error = -EINVAL;
2400 break;
2401 case PR_SET_NAME:
2402 comm[sizeof(me->comm) - 1] = 0;
2403 if (strncpy_from_user(comm, (char __user *)arg2,
2404 sizeof(me->comm) - 1) < 0)
2405 return -EFAULT;
2406 set_task_comm(me, comm);
2407 proc_comm_connector(me);
2408 break;
2409 case PR_GET_NAME:
2410 get_task_comm(comm, me);
2411 if (copy_to_user((char __user *)arg2, comm, sizeof(comm)))
2412 return -EFAULT;
2413 break;
2414 case PR_GET_ENDIAN:
2415 error = GET_ENDIAN(me, arg2);
2416 break;
2417 case PR_SET_ENDIAN:
2418 error = SET_ENDIAN(me, arg2);
2419 break;
2420 case PR_GET_SECCOMP:
2421 error = prctl_get_seccomp();
2422 break;
2423 case PR_SET_SECCOMP:
2424 error = prctl_set_seccomp(arg2, (char __user *)arg3);
2425 break;
2426 case PR_GET_TSC:
2427 error = GET_TSC_CTL(arg2);
2428 break;
2429 case PR_SET_TSC:
2430 error = SET_TSC_CTL(arg2);
2431 break;
2432 case PR_TASK_PERF_EVENTS_DISABLE:
2433 error = perf_event_task_disable();
2434 break;
2435 case PR_TASK_PERF_EVENTS_ENABLE:
2436 error = perf_event_task_enable();
2437 break;
2438 case PR_GET_TIMERSLACK:
2439 if (current->timer_slack_ns > ULONG_MAX)
2440 error = ULONG_MAX;
2441 else
2442 error = current->timer_slack_ns;
2443 break;
2444 case PR_SET_TIMERSLACK:
2445 if (arg2 <= 0)
2446 current->timer_slack_ns =
2447 current->default_timer_slack_ns;
2448 else
2449 current->timer_slack_ns = arg2;
2450 break;
2451 case PR_MCE_KILL:
2452 if (arg4 | arg5)
2453 return -EINVAL;
2454 switch (arg2) {
2455 case PR_MCE_KILL_CLEAR:
2456 if (arg3 != 0)
2457 return -EINVAL;
2458 current->flags &= ~PF_MCE_PROCESS;
2459 break;
2460 case PR_MCE_KILL_SET:
2461 current->flags |= PF_MCE_PROCESS;
2462 if (arg3 == PR_MCE_KILL_EARLY)
2463 current->flags |= PF_MCE_EARLY;
2464 else if (arg3 == PR_MCE_KILL_LATE)
2465 current->flags &= ~PF_MCE_EARLY;
2466 else if (arg3 == PR_MCE_KILL_DEFAULT)
2467 current->flags &=
2468 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
2469 else
2470 return -EINVAL;
2471 break;
2472 default:
2473 return -EINVAL;
2474 }
2475 break;
2476 case PR_MCE_KILL_GET:
2477 if (arg2 | arg3 | arg4 | arg5)
2478 return -EINVAL;
2479 if (current->flags & PF_MCE_PROCESS)
2480 error = (current->flags & PF_MCE_EARLY) ?
2481 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2482 else
2483 error = PR_MCE_KILL_DEFAULT;
2484 break;
2485 case PR_SET_MM:
2486 error = prctl_set_mm(arg2, arg3, arg4, arg5);
2487 break;
2488 case PR_GET_TID_ADDRESS:
2489 error = prctl_get_tid_address(me, (int __user * __user *)arg2);
2490 break;
2491 case PR_SET_CHILD_SUBREAPER:
2492 me->signal->is_child_subreaper = !!arg2;
2493 if (!arg2)
2494 break;
2495
2496 walk_process_tree(me, propagate_has_child_subreaper, NULL);
2497 break;
2498 case PR_GET_CHILD_SUBREAPER:
2499 error = put_user(me->signal->is_child_subreaper,
2500 (int __user *)arg2);
2501 break;
2502 case PR_SET_NO_NEW_PRIVS:
2503 if (arg2 != 1 || arg3 || arg4 || arg5)
2504 return -EINVAL;
2505
2506 task_set_no_new_privs(current);
2507 break;
2508 case PR_GET_NO_NEW_PRIVS:
2509 if (arg2 || arg3 || arg4 || arg5)
2510 return -EINVAL;
2511 return task_no_new_privs(current) ? 1 : 0;
2512 case PR_GET_THP_DISABLE:
2513 if (arg2 || arg3 || arg4 || arg5)
2514 return -EINVAL;
2515 error = !!test_bit(MMF_DISABLE_THP, &me->mm->flags);
2516 break;
2517 case PR_SET_THP_DISABLE:
2518 if (arg3 || arg4 || arg5)
2519 return -EINVAL;
2520 if (mmap_write_lock_killable(me->mm))
2521 return -EINTR;
2522 if (arg2)
2523 set_bit(MMF_DISABLE_THP, &me->mm->flags);
2524 else
2525 clear_bit(MMF_DISABLE_THP, &me->mm->flags);
2526 mmap_write_unlock(me->mm);
2527 break;
2528 case PR_MPX_ENABLE_MANAGEMENT:
2529 case PR_MPX_DISABLE_MANAGEMENT:
2530 /* No longer implemented: */
2531 return -EINVAL;
2532 case PR_SET_FP_MODE:
2533 error = SET_FP_MODE(me, arg2);
2534 break;
2535 case PR_GET_FP_MODE:
2536 error = GET_FP_MODE(me);
2537 break;
2538 case PR_SVE_SET_VL:
2539 error = SVE_SET_VL(arg2);
2540 break;
2541 case PR_SVE_GET_VL:
2542 error = SVE_GET_VL();
2543 break;
2544 case PR_GET_SPECULATION_CTRL:
2545 if (arg3 || arg4 || arg5)
2546 return -EINVAL;
2547 error = arch_prctl_spec_ctrl_get(me, arg2);
2548 break;
2549 case PR_SET_SPECULATION_CTRL:
2550 if (arg4 || arg5)
2551 return -EINVAL;
2552 error = arch_prctl_spec_ctrl_set(me, arg2, arg3);
2553 break;
2554 case PR_PAC_RESET_KEYS:
2555 if (arg3 || arg4 || arg5)
2556 return -EINVAL;
2557 error = PAC_RESET_KEYS(me, arg2);
2558 break;
2559 case PR_PAC_SET_ENABLED_KEYS:
2560 if (arg4 || arg5)
2561 return -EINVAL;
2562 error = PAC_SET_ENABLED_KEYS(me, arg2, arg3);
2563 break;
2564 case PR_PAC_GET_ENABLED_KEYS:
2565 if (arg2 || arg3 || arg4 || arg5)
2566 return -EINVAL;
2567 error = PAC_GET_ENABLED_KEYS(me);
2568 break;
2569 case PR_SET_TAGGED_ADDR_CTRL:
2570 if (arg3 || arg4 || arg5)
2571 return -EINVAL;
2572 error = SET_TAGGED_ADDR_CTRL(arg2);
2573 break;
2574 case PR_GET_TAGGED_ADDR_CTRL:
2575 if (arg2 || arg3 || arg4 || arg5)
2576 return -EINVAL;
2577 error = GET_TAGGED_ADDR_CTRL();
2578 break;
2579 case PR_SET_IO_FLUSHER:
2580 if (!capable(CAP_SYS_RESOURCE))
2581 return -EPERM;
2582
2583 if (arg3 || arg4 || arg5)
2584 return -EINVAL;
2585
2586 if (arg2 == 1)
2587 current->flags |= PR_IO_FLUSHER;
2588 else if (!arg2)
2589 current->flags &= ~PR_IO_FLUSHER;
2590 else
2591 return -EINVAL;
2592 break;
2593 case PR_GET_IO_FLUSHER:
2594 if (!capable(CAP_SYS_RESOURCE))
2595 return -EPERM;
2596
2597 if (arg2 || arg3 || arg4 || arg5)
2598 return -EINVAL;
2599
2600 error = (current->flags & PR_IO_FLUSHER) == PR_IO_FLUSHER;
2601 break;
2602 case PR_SET_SYSCALL_USER_DISPATCH:
2603 error = set_syscall_user_dispatch(arg2, arg3, arg4,
2604 (char __user *) arg5);
2605 break;
2606#ifdef CONFIG_SCHED_CORE
2607 case PR_SCHED_CORE:
2608 error = sched_core_share_pid(arg2, arg3, arg4, arg5);
2609 break;
2610#endif
2611 case PR_SET_VMA:
2612 error = prctl_set_vma(arg2, arg3, arg4, arg5);
2613 break;
2614 default:
2615 error = -EINVAL;
2616 break;
2617 }
2618 return error;
2619}
2620
2621SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2622 struct getcpu_cache __user *, unused)
2623{
2624 int err = 0;
2625 int cpu = raw_smp_processor_id();
2626
2627 if (cpup)
2628 err |= put_user(cpu, cpup);
2629 if (nodep)
2630 err |= put_user(cpu_to_node(cpu), nodep);
2631 return err ? -EFAULT : 0;
2632}
2633
2634/**
2635 * do_sysinfo - fill in sysinfo struct
2636 * @info: pointer to buffer to fill
2637 */
2638static int do_sysinfo(struct sysinfo *info)
2639{
2640 unsigned long mem_total, sav_total;
2641 unsigned int mem_unit, bitcount;
2642 struct timespec64 tp;
2643
2644 memset(info, 0, sizeof(struct sysinfo));
2645
2646 ktime_get_boottime_ts64(&tp);
2647 timens_add_boottime(&tp);
2648 info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
2649
2650 get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
2651
2652 info->procs = nr_threads;
2653
2654 si_meminfo(info);
2655 si_swapinfo(info);
2656
2657 /*
2658 * If the sum of all the available memory (i.e. ram + swap)
2659 * is less than can be stored in a 32 bit unsigned long then
2660 * we can be binary compatible with 2.2.x kernels. If not,
2661 * well, in that case 2.2.x was broken anyways...
2662 *
2663 * -Erik Andersen <[email protected]>
2664 */
2665
2666 mem_total = info->totalram + info->totalswap;
2667 if (mem_total < info->totalram || mem_total < info->totalswap)
2668 goto out;
2669 bitcount = 0;
2670 mem_unit = info->mem_unit;
2671 while (mem_unit > 1) {
2672 bitcount++;
2673 mem_unit >>= 1;
2674 sav_total = mem_total;
2675 mem_total <<= 1;
2676 if (mem_total < sav_total)
2677 goto out;
2678 }
2679
2680 /*
2681 * If mem_total did not overflow, multiply all memory values by
2682 * info->mem_unit and set it to 1. This leaves things compatible
2683 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2684 * kernels...
2685 */
2686
2687 info->mem_unit = 1;
2688 info->totalram <<= bitcount;
2689 info->freeram <<= bitcount;
2690 info->sharedram <<= bitcount;
2691 info->bufferram <<= bitcount;
2692 info->totalswap <<= bitcount;
2693 info->freeswap <<= bitcount;
2694 info->totalhigh <<= bitcount;
2695 info->freehigh <<= bitcount;
2696
2697out:
2698 return 0;
2699}
2700
2701SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
2702{
2703 struct sysinfo val;
2704
2705 do_sysinfo(&val);
2706
2707 if (copy_to_user(info, &val, sizeof(struct sysinfo)))
2708 return -EFAULT;
2709
2710 return 0;
2711}
2712
2713#ifdef CONFIG_COMPAT
2714struct compat_sysinfo {
2715 s32 uptime;
2716 u32 loads[3];
2717 u32 totalram;
2718 u32 freeram;
2719 u32 sharedram;
2720 u32 bufferram;
2721 u32 totalswap;
2722 u32 freeswap;
2723 u16 procs;
2724 u16 pad;
2725 u32 totalhigh;
2726 u32 freehigh;
2727 u32 mem_unit;
2728 char _f[20-2*sizeof(u32)-sizeof(int)];
2729};
2730
2731COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2732{
2733 struct sysinfo s;
2734 struct compat_sysinfo s_32;
2735
2736 do_sysinfo(&s);
2737
2738 /* Check to see if any memory value is too large for 32-bit and scale
2739 * down if needed
2740 */
2741 if (upper_32_bits(s.totalram) || upper_32_bits(s.totalswap)) {
2742 int bitcount = 0;
2743
2744 while (s.mem_unit < PAGE_SIZE) {
2745 s.mem_unit <<= 1;
2746 bitcount++;
2747 }
2748
2749 s.totalram >>= bitcount;
2750 s.freeram >>= bitcount;
2751 s.sharedram >>= bitcount;
2752 s.bufferram >>= bitcount;
2753 s.totalswap >>= bitcount;
2754 s.freeswap >>= bitcount;
2755 s.totalhigh >>= bitcount;
2756 s.freehigh >>= bitcount;
2757 }
2758
2759 memset(&s_32, 0, sizeof(s_32));
2760 s_32.uptime = s.uptime;
2761 s_32.loads[0] = s.loads[0];
2762 s_32.loads[1] = s.loads[1];
2763 s_32.loads[2] = s.loads[2];
2764 s_32.totalram = s.totalram;
2765 s_32.freeram = s.freeram;
2766 s_32.sharedram = s.sharedram;
2767 s_32.bufferram = s.bufferram;
2768 s_32.totalswap = s.totalswap;
2769 s_32.freeswap = s.freeswap;
2770 s_32.procs = s.procs;
2771 s_32.totalhigh = s.totalhigh;
2772 s_32.freehigh = s.freehigh;
2773 s_32.mem_unit = s.mem_unit;
2774 if (copy_to_user(info, &s_32, sizeof(s_32)))
2775 return -EFAULT;
2776 return 0;
2777}
2778#endif /* CONFIG_COMPAT */