projects / linux-2.6-block.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
Merge tag 'mm-hotfixes-stable-2025-07-11-16-16' of git://git.kernel.org/pub/scm/linux...
[linux-2.6-block.git] / kernel / sys.c
... / ...
CommitLineData
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/ksm.h>
19#include <linux/perf_event.h>
20#include <linux/resource.h>
21#include <linux/kernel.h>
22#include <linux/workqueue.h>
23#include <linux/capability.h>
24#include <linux/device.h>
25#include <linux/key.h>
26#include <linux/times.h>
27#include <linux/posix-timers.h>
28#include <linux/security.h>
29#include <linux/random.h>
30#include <linux/suspend.h>
31#include <linux/tty.h>
32#include <linux/signal.h>
33#include <linux/cn_proc.h>
34#include <linux/getcpu.h>
35#include <linux/task_io_accounting_ops.h>
36#include <linux/seccomp.h>
37#include <linux/cpu.h>
38#include <linux/personality.h>
39#include <linux/ptrace.h>
40#include <linux/fs_struct.h>
41#include <linux/file.h>
42#include <linux/mount.h>
43#include <linux/gfp.h>
44#include <linux/syscore_ops.h>
45#include <linux/version.h>
46#include <linux/ctype.h>
47#include <linux/syscall_user_dispatch.h>
48
49#include <linux/compat.h>
50#include <linux/syscalls.h>
51#include <linux/kprobes.h>
52#include <linux/user_namespace.h>
53#include <linux/time_namespace.h>
54#include <linux/binfmts.h>
55#include <linux/futex.h>
56
57#include <linux/sched.h>
58#include <linux/sched/autogroup.h>
59#include <linux/sched/loadavg.h>
60#include <linux/sched/stat.h>
61#include <linux/sched/mm.h>
62#include <linux/sched/coredump.h>
63#include <linux/sched/task.h>
64#include <linux/sched/cputime.h>
65#include <linux/rcupdate.h>
66#include <linux/uidgid.h>
67#include <linux/cred.h>
68
69#include <linux/nospec.h>
70
71#include <linux/kmsg_dump.h>
72/* Move somewhere else to avoid recompiling? */
73#include <generated/utsrelease.h>
74
75#include <linux/uaccess.h>
76#include <asm/io.h>
77#include <asm/unistd.h>
78
79#include <trace/events/task.h>
80
81#include "uid16.h"
82
83#ifndef SET_UNALIGN_CTL
84# define SET_UNALIGN_CTL(a, b) (-EINVAL)
85#endif
86#ifndef GET_UNALIGN_CTL
87# define GET_UNALIGN_CTL(a, b) (-EINVAL)
88#endif
89#ifndef SET_FPEMU_CTL
90# define SET_FPEMU_CTL(a, b) (-EINVAL)
91#endif
92#ifndef GET_FPEMU_CTL
93# define GET_FPEMU_CTL(a, b) (-EINVAL)
94#endif
95#ifndef SET_FPEXC_CTL
96# define SET_FPEXC_CTL(a, b) (-EINVAL)
97#endif
98#ifndef GET_FPEXC_CTL
99# define GET_FPEXC_CTL(a, b) (-EINVAL)
100#endif
101#ifndef GET_ENDIAN
102# define GET_ENDIAN(a, b) (-EINVAL)
103#endif
104#ifndef SET_ENDIAN
105# define SET_ENDIAN(a, b) (-EINVAL)
106#endif
107#ifndef GET_TSC_CTL
108# define GET_TSC_CTL(a) (-EINVAL)
109#endif
110#ifndef SET_TSC_CTL
111# define SET_TSC_CTL(a) (-EINVAL)
112#endif
113#ifndef GET_FP_MODE
114# define GET_FP_MODE(a) (-EINVAL)
115#endif
116#ifndef SET_FP_MODE
117# define SET_FP_MODE(a,b) (-EINVAL)
118#endif
119#ifndef SVE_SET_VL
120# define SVE_SET_VL(a) (-EINVAL)
121#endif
122#ifndef SVE_GET_VL
123# define SVE_GET_VL() (-EINVAL)
124#endif
125#ifndef SME_SET_VL
126# define SME_SET_VL(a) (-EINVAL)
127#endif
128#ifndef SME_GET_VL
129# define SME_GET_VL() (-EINVAL)
130#endif
131#ifndef PAC_RESET_KEYS
132# define PAC_RESET_KEYS(a, b) (-EINVAL)
133#endif
134#ifndef PAC_SET_ENABLED_KEYS
135# define PAC_SET_ENABLED_KEYS(a, b, c) (-EINVAL)
136#endif
137#ifndef PAC_GET_ENABLED_KEYS
138# define PAC_GET_ENABLED_KEYS(a) (-EINVAL)
139#endif
140#ifndef SET_TAGGED_ADDR_CTRL
141# define SET_TAGGED_ADDR_CTRL(a) (-EINVAL)
142#endif
143#ifndef GET_TAGGED_ADDR_CTRL
144# define GET_TAGGED_ADDR_CTRL() (-EINVAL)
145#endif
146#ifndef RISCV_V_SET_CONTROL
147# define RISCV_V_SET_CONTROL(a) (-EINVAL)
148#endif
149#ifndef RISCV_V_GET_CONTROL
150# define RISCV_V_GET_CONTROL() (-EINVAL)
151#endif
152#ifndef RISCV_SET_ICACHE_FLUSH_CTX
153# define RISCV_SET_ICACHE_FLUSH_CTX(a, b) (-EINVAL)
154#endif
155#ifndef PPC_GET_DEXCR_ASPECT
156# define PPC_GET_DEXCR_ASPECT(a, b) (-EINVAL)
157#endif
158#ifndef PPC_SET_DEXCR_ASPECT
159# define PPC_SET_DEXCR_ASPECT(a, b, c) (-EINVAL)
160#endif
161
162/*
163 * this is where the system-wide overflow UID and GID are defined, for
164 * architectures that now have 32-bit UID/GID but didn't in the past
165 */
166
167int overflowuid = DEFAULT_OVERFLOWUID;
168int overflowgid = DEFAULT_OVERFLOWGID;
169
170EXPORT_SYMBOL(overflowuid);
171EXPORT_SYMBOL(overflowgid);
172
173/*
174 * the same as above, but for filesystems which can only store a 16-bit
175 * UID and GID. as such, this is needed on all architectures
176 */
177
178int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
179int fs_overflowgid = DEFAULT_FS_OVERFLOWGID;
180
181EXPORT_SYMBOL(fs_overflowuid);
182EXPORT_SYMBOL(fs_overflowgid);
183
184/*
185 * Returns true if current's euid is same as p's uid or euid,
186 * or has CAP_SYS_NICE to p's user_ns.
187 *
188 * Called with rcu_read_lock, creds are safe
189 */
190static bool set_one_prio_perm(struct task_struct *p)
191{
192 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
193
194 if (uid_eq(pcred->uid, cred->euid) ||
195 uid_eq(pcred->euid, cred->euid))
196 return true;
197 if (ns_capable(pcred->user_ns, CAP_SYS_NICE))
198 return true;
199 return false;
200}
201
202/*
203 * set the priority of a task
204 * - the caller must hold the RCU read lock
205 */
206static int set_one_prio(struct task_struct *p, int niceval, int error)
207{
208 int no_nice;
209
210 if (!set_one_prio_perm(p)) {
211 error = -EPERM;
212 goto out;
213 }
214 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
215 error = -EACCES;
216 goto out;
217 }
218 no_nice = security_task_setnice(p, niceval);
219 if (no_nice) {
220 error = no_nice;
221 goto out;
222 }
223 if (error == -ESRCH)
224 error = 0;
225 set_user_nice(p, niceval);
226out:
227 return error;
228}
229
230SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
231{
232 struct task_struct *g, *p;
233 struct user_struct *user;
234 const struct cred *cred = current_cred();
235 int error = -EINVAL;
236 struct pid *pgrp;
237 kuid_t uid;
238
239 if (which > PRIO_USER || which < PRIO_PROCESS)
240 goto out;
241
242 /* normalize: avoid signed division (rounding problems) */
243 error = -ESRCH;
244 if (niceval < MIN_NICE)
245 niceval = MIN_NICE;
246 if (niceval > MAX_NICE)
247 niceval = MAX_NICE;
248
249 rcu_read_lock();
250 switch (which) {
251 case PRIO_PROCESS:
252 if (who)
253 p = find_task_by_vpid(who);
254 else
255 p = current;
256 if (p)
257 error = set_one_prio(p, niceval, error);
258 break;
259 case PRIO_PGRP:
260 if (who)
261 pgrp = find_vpid(who);
262 else
263 pgrp = task_pgrp(current);
264 read_lock(&tasklist_lock);
265 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
266 error = set_one_prio(p, niceval, error);
267 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
268 read_unlock(&tasklist_lock);
269 break;
270 case PRIO_USER:
271 uid = make_kuid(cred->user_ns, who);
272 user = cred->user;
273 if (!who)
274 uid = cred->uid;
275 else if (!uid_eq(uid, cred->uid)) {
276 user = find_user(uid);
277 if (!user)
278 goto out_unlock; /* No processes for this user */
279 }
280 for_each_process_thread(g, p) {
281 if (uid_eq(task_uid(p), uid) && task_pid_vnr(p))
282 error = set_one_prio(p, niceval, error);
283 }
284 if (!uid_eq(uid, cred->uid))
285 free_uid(user); /* For find_user() */
286 break;
287 }
288out_unlock:
289 rcu_read_unlock();
290out:
291 return error;
292}
293
294/*
295 * Ugh. To avoid negative return values, "getpriority()" will
296 * not return the normal nice-value, but a negated value that
297 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
298 * to stay compatible.
299 */
300SYSCALL_DEFINE2(getpriority, int, which, int, who)
301{
302 struct task_struct *g, *p;
303 struct user_struct *user;
304 const struct cred *cred = current_cred();
305 long niceval, retval = -ESRCH;
306 struct pid *pgrp;
307 kuid_t uid;
308
309 if (which > PRIO_USER || which < PRIO_PROCESS)
310 return -EINVAL;
311
312 rcu_read_lock();
313 switch (which) {
314 case PRIO_PROCESS:
315 if (who)
316 p = find_task_by_vpid(who);
317 else
318 p = current;
319 if (p) {
320 niceval = nice_to_rlimit(task_nice(p));
321 if (niceval > retval)
322 retval = niceval;
323 }
324 break;
325 case PRIO_PGRP:
326 if (who)
327 pgrp = find_vpid(who);
328 else
329 pgrp = task_pgrp(current);
330 read_lock(&tasklist_lock);
331 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
332 niceval = nice_to_rlimit(task_nice(p));
333 if (niceval > retval)
334 retval = niceval;
335 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
336 read_unlock(&tasklist_lock);
337 break;
338 case PRIO_USER:
339 uid = make_kuid(cred->user_ns, who);
340 user = cred->user;
341 if (!who)
342 uid = cred->uid;
343 else if (!uid_eq(uid, cred->uid)) {
344 user = find_user(uid);
345 if (!user)
346 goto out_unlock; /* No processes for this user */
347 }
348 for_each_process_thread(g, p) {
349 if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) {
350 niceval = nice_to_rlimit(task_nice(p));
351 if (niceval > retval)
352 retval = niceval;
353 }
354 }
355 if (!uid_eq(uid, cred->uid))
356 free_uid(user); /* for find_user() */
357 break;
358 }
359out_unlock:
360 rcu_read_unlock();
361
362 return retval;
363}
364
365/*
366 * Unprivileged users may change the real gid to the effective gid
367 * or vice versa. (BSD-style)
368 *
369 * If you set the real gid at all, or set the effective gid to a value not
370 * equal to the real gid, then the saved gid is set to the new effective gid.
371 *
372 * This makes it possible for a setgid program to completely drop its
373 * privileges, which is often a useful assertion to make when you are doing
374 * a security audit over a program.
375 *
376 * The general idea is that a program which uses just setregid() will be
377 * 100% compatible with BSD. A program which uses just setgid() will be
378 * 100% compatible with POSIX with saved IDs.
379 *
380 * SMP: There are not races, the GIDs are checked only by filesystem
381 * operations (as far as semantic preservation is concerned).
382 */
383#ifdef CONFIG_MULTIUSER
384long __sys_setregid(gid_t rgid, gid_t egid)
385{
386 struct user_namespace *ns = current_user_ns();
387 const struct cred *old;
388 struct cred *new;
389 int retval;
390 kgid_t krgid, kegid;
391
392 krgid = make_kgid(ns, rgid);
393 kegid = make_kgid(ns, egid);
394
395 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
396 return -EINVAL;
397 if ((egid != (gid_t) -1) && !gid_valid(kegid))
398 return -EINVAL;
399
400 new = prepare_creds();
401 if (!new)
402 return -ENOMEM;
403 old = current_cred();
404
405 retval = -EPERM;
406 if (rgid != (gid_t) -1) {
407 if (gid_eq(old->gid, krgid) ||
408 gid_eq(old->egid, krgid) ||
409 ns_capable_setid(old->user_ns, CAP_SETGID))
410 new->gid = krgid;
411 else
412 goto error;
413 }
414 if (egid != (gid_t) -1) {
415 if (gid_eq(old->gid, kegid) ||
416 gid_eq(old->egid, kegid) ||
417 gid_eq(old->sgid, kegid) ||
418 ns_capable_setid(old->user_ns, CAP_SETGID))
419 new->egid = kegid;
420 else
421 goto error;
422 }
423
424 if (rgid != (gid_t) -1 ||
425 (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
426 new->sgid = new->egid;
427 new->fsgid = new->egid;
428
429 retval = security_task_fix_setgid(new, old, LSM_SETID_RE);
430 if (retval < 0)
431 goto error;
432
433 return commit_creds(new);
434
435error:
436 abort_creds(new);
437 return retval;
438}
439
440SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
441{
442 return __sys_setregid(rgid, egid);
443}
444
445/*
446 * setgid() is implemented like SysV w/ SAVED_IDS
447 *
448 * SMP: Same implicit races as above.
449 */
450long __sys_setgid(gid_t gid)
451{
452 struct user_namespace *ns = current_user_ns();
453 const struct cred *old;
454 struct cred *new;
455 int retval;
456 kgid_t kgid;
457
458 kgid = make_kgid(ns, gid);
459 if (!gid_valid(kgid))
460 return -EINVAL;
461
462 new = prepare_creds();
463 if (!new)
464 return -ENOMEM;
465 old = current_cred();
466
467 retval = -EPERM;
468 if (ns_capable_setid(old->user_ns, CAP_SETGID))
469 new->gid = new->egid = new->sgid = new->fsgid = kgid;
470 else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
471 new->egid = new->fsgid = kgid;
472 else
473 goto error;
474
475 retval = security_task_fix_setgid(new, old, LSM_SETID_ID);
476 if (retval < 0)
477 goto error;
478
479 return commit_creds(new);
480
481error:
482 abort_creds(new);
483 return retval;
484}
485
486SYSCALL_DEFINE1(setgid, gid_t, gid)
487{
488 return __sys_setgid(gid);
489}
490
491/*
492 * change the user struct in a credentials set to match the new UID
493 */
494static int set_user(struct cred *new)
495{
496 struct user_struct *new_user;
497
498 new_user = alloc_uid(new->uid);
499 if (!new_user)
500 return -EAGAIN;
501
502 free_uid(new->user);
503 new->user = new_user;
504 return 0;
505}
506
507static void flag_nproc_exceeded(struct cred *new)
508{
509 if (new->ucounts == current_ucounts())
510 return;
511
512 /*
513 * We don't fail in case of NPROC limit excess here because too many
514 * poorly written programs don't check set*uid() return code, assuming
515 * it never fails if called by root. We may still enforce NPROC limit
516 * for programs doing set*uid()+execve() by harmlessly deferring the
517 * failure to the execve() stage.
518 */
519 if (is_rlimit_overlimit(new->ucounts, UCOUNT_RLIMIT_NPROC, rlimit(RLIMIT_NPROC)) &&
520 new->user != INIT_USER)
521 current->flags |= PF_NPROC_EXCEEDED;
522 else
523 current->flags &= ~PF_NPROC_EXCEEDED;
524}
525
526/*
527 * Unprivileged users may change the real uid to the effective uid
528 * or vice versa. (BSD-style)
529 *
530 * If you set the real uid at all, or set the effective uid to a value not
531 * equal to the real uid, then the saved uid is set to the new effective uid.
532 *
533 * This makes it possible for a setuid program to completely drop its
534 * privileges, which is often a useful assertion to make when you are doing
535 * a security audit over a program.
536 *
537 * The general idea is that a program which uses just setreuid() will be
538 * 100% compatible with BSD. A program which uses just setuid() will be
539 * 100% compatible with POSIX with saved IDs.
540 */
541long __sys_setreuid(uid_t ruid, uid_t euid)
542{
543 struct user_namespace *ns = current_user_ns();
544 const struct cred *old;
545 struct cred *new;
546 int retval;
547 kuid_t kruid, keuid;
548
549 kruid = make_kuid(ns, ruid);
550 keuid = make_kuid(ns, euid);
551
552 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
553 return -EINVAL;
554 if ((euid != (uid_t) -1) && !uid_valid(keuid))
555 return -EINVAL;
556
557 new = prepare_creds();
558 if (!new)
559 return -ENOMEM;
560 old = current_cred();
561
562 retval = -EPERM;
563 if (ruid != (uid_t) -1) {
564 new->uid = kruid;
565 if (!uid_eq(old->uid, kruid) &&
566 !uid_eq(old->euid, kruid) &&
567 !ns_capable_setid(old->user_ns, CAP_SETUID))
568 goto error;
569 }
570
571 if (euid != (uid_t) -1) {
572 new->euid = keuid;
573 if (!uid_eq(old->uid, keuid) &&
574 !uid_eq(old->euid, keuid) &&
575 !uid_eq(old->suid, keuid) &&
576 !ns_capable_setid(old->user_ns, CAP_SETUID))
577 goto error;
578 }
579
580 if (!uid_eq(new->uid, old->uid)) {
581 retval = set_user(new);
582 if (retval < 0)
583 goto error;
584 }
585 if (ruid != (uid_t) -1 ||
586 (euid != (uid_t) -1 && !uid_eq(keuid, old->uid)))
587 new->suid = new->euid;
588 new->fsuid = new->euid;
589
590 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
591 if (retval < 0)
592 goto error;
593
594 retval = set_cred_ucounts(new);
595 if (retval < 0)
596 goto error;
597
598 flag_nproc_exceeded(new);
599 return commit_creds(new);
600
601error:
602 abort_creds(new);
603 return retval;
604}
605
606SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
607{
608 return __sys_setreuid(ruid, euid);
609}
610
611/*
612 * setuid() is implemented like SysV with SAVED_IDS
613 *
614 * Note that SAVED_ID's is deficient in that a setuid root program
615 * like sendmail, for example, cannot set its uid to be a normal
616 * user and then switch back, because if you're root, setuid() sets
617 * the saved uid too. If you don't like this, blame the bright people
618 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
619 * will allow a root program to temporarily drop privileges and be able to
620 * regain them by swapping the real and effective uid.
621 */
622long __sys_setuid(uid_t uid)
623{
624 struct user_namespace *ns = current_user_ns();
625 const struct cred *old;
626 struct cred *new;
627 int retval;
628 kuid_t kuid;
629
630 kuid = make_kuid(ns, uid);
631 if (!uid_valid(kuid))
632 return -EINVAL;
633
634 new = prepare_creds();
635 if (!new)
636 return -ENOMEM;
637 old = current_cred();
638
639 retval = -EPERM;
640 if (ns_capable_setid(old->user_ns, CAP_SETUID)) {
641 new->suid = new->uid = kuid;
642 if (!uid_eq(kuid, old->uid)) {
643 retval = set_user(new);
644 if (retval < 0)
645 goto error;
646 }
647 } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) {
648 goto error;
649 }
650
651 new->fsuid = new->euid = kuid;
652
653 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
654 if (retval < 0)
655 goto error;
656
657 retval = set_cred_ucounts(new);
658 if (retval < 0)
659 goto error;
660
661 flag_nproc_exceeded(new);
662 return commit_creds(new);
663
664error:
665 abort_creds(new);
666 return retval;
667}
668
669SYSCALL_DEFINE1(setuid, uid_t, uid)
670{
671 return __sys_setuid(uid);
672}
673
674
675/*
676 * This function implements a generic ability to update ruid, euid,
677 * and suid. This allows you to implement the 4.4 compatible seteuid().
678 */
679long __sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
680{
681 struct user_namespace *ns = current_user_ns();
682 const struct cred *old;
683 struct cred *new;
684 int retval;
685 kuid_t kruid, keuid, ksuid;
686 bool ruid_new, euid_new, suid_new;
687
688 kruid = make_kuid(ns, ruid);
689 keuid = make_kuid(ns, euid);
690 ksuid = make_kuid(ns, suid);
691
692 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
693 return -EINVAL;
694
695 if ((euid != (uid_t) -1) && !uid_valid(keuid))
696 return -EINVAL;
697
698 if ((suid != (uid_t) -1) && !uid_valid(ksuid))
699 return -EINVAL;
700
701 old = current_cred();
702
703 /* check for no-op */
704 if ((ruid == (uid_t) -1 || uid_eq(kruid, old->uid)) &&
705 (euid == (uid_t) -1 || (uid_eq(keuid, old->euid) &&
706 uid_eq(keuid, old->fsuid))) &&
707 (suid == (uid_t) -1 || uid_eq(ksuid, old->suid)))
708 return 0;
709
710 ruid_new = ruid != (uid_t) -1 && !uid_eq(kruid, old->uid) &&
711 !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid);
712 euid_new = euid != (uid_t) -1 && !uid_eq(keuid, old->uid) &&
713 !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid);
714 suid_new = suid != (uid_t) -1 && !uid_eq(ksuid, old->uid) &&
715 !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid);
716 if ((ruid_new || euid_new || suid_new) &&
717 !ns_capable_setid(old->user_ns, CAP_SETUID))
718 return -EPERM;
719
720 new = prepare_creds();
721 if (!new)
722 return -ENOMEM;
723
724 if (ruid != (uid_t) -1) {
725 new->uid = kruid;
726 if (!uid_eq(kruid, old->uid)) {
727 retval = set_user(new);
728 if (retval < 0)
729 goto error;
730 }
731 }
732 if (euid != (uid_t) -1)
733 new->euid = keuid;
734 if (suid != (uid_t) -1)
735 new->suid = ksuid;
736 new->fsuid = new->euid;
737
738 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
739 if (retval < 0)
740 goto error;
741
742 retval = set_cred_ucounts(new);
743 if (retval < 0)
744 goto error;
745
746 flag_nproc_exceeded(new);
747 return commit_creds(new);
748
749error:
750 abort_creds(new);
751 return retval;
752}
753
754SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
755{
756 return __sys_setresuid(ruid, euid, suid);
757}
758
759SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
760{
761 const struct cred *cred = current_cred();
762 int retval;
763 uid_t ruid, euid, suid;
764
765 ruid = from_kuid_munged(cred->user_ns, cred->uid);
766 euid = from_kuid_munged(cred->user_ns, cred->euid);
767 suid = from_kuid_munged(cred->user_ns, cred->suid);
768
769 retval = put_user(ruid, ruidp);
770 if (!retval) {
771 retval = put_user(euid, euidp);
772 if (!retval)
773 return put_user(suid, suidp);
774 }
775 return retval;
776}
777
778/*
779 * Same as above, but for rgid, egid, sgid.
780 */
781long __sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
782{
783 struct user_namespace *ns = current_user_ns();
784 const struct cred *old;
785 struct cred *new;
786 int retval;
787 kgid_t krgid, kegid, ksgid;
788 bool rgid_new, egid_new, sgid_new;
789
790 krgid = make_kgid(ns, rgid);
791 kegid = make_kgid(ns, egid);
792 ksgid = make_kgid(ns, sgid);
793
794 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
795 return -EINVAL;
796 if ((egid != (gid_t) -1) && !gid_valid(kegid))
797 return -EINVAL;
798 if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
799 return -EINVAL;
800
801 old = current_cred();
802
803 /* check for no-op */
804 if ((rgid == (gid_t) -1 || gid_eq(krgid, old->gid)) &&
805 (egid == (gid_t) -1 || (gid_eq(kegid, old->egid) &&
806 gid_eq(kegid, old->fsgid))) &&
807 (sgid == (gid_t) -1 || gid_eq(ksgid, old->sgid)))
808 return 0;
809
810 rgid_new = rgid != (gid_t) -1 && !gid_eq(krgid, old->gid) &&
811 !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid);
812 egid_new = egid != (gid_t) -1 && !gid_eq(kegid, old->gid) &&
813 !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid);
814 sgid_new = sgid != (gid_t) -1 && !gid_eq(ksgid, old->gid) &&
815 !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid);
816 if ((rgid_new || egid_new || sgid_new) &&
817 !ns_capable_setid(old->user_ns, CAP_SETGID))
818 return -EPERM;
819
820 new = prepare_creds();
821 if (!new)
822 return -ENOMEM;
823
824 if (rgid != (gid_t) -1)
825 new->gid = krgid;
826 if (egid != (gid_t) -1)
827 new->egid = kegid;
828 if (sgid != (gid_t) -1)
829 new->sgid = ksgid;
830 new->fsgid = new->egid;
831
832 retval = security_task_fix_setgid(new, old, LSM_SETID_RES);
833 if (retval < 0)
834 goto error;
835
836 return commit_creds(new);
837
838error:
839 abort_creds(new);
840 return retval;
841}
842
843SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
844{
845 return __sys_setresgid(rgid, egid, sgid);
846}
847
848SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
849{
850 const struct cred *cred = current_cred();
851 int retval;
852 gid_t rgid, egid, sgid;
853
854 rgid = from_kgid_munged(cred->user_ns, cred->gid);
855 egid = from_kgid_munged(cred->user_ns, cred->egid);
856 sgid = from_kgid_munged(cred->user_ns, cred->sgid);
857
858 retval = put_user(rgid, rgidp);
859 if (!retval) {
860 retval = put_user(egid, egidp);
861 if (!retval)
862 retval = put_user(sgid, sgidp);
863 }
864
865 return retval;
866}
867
868
869/*
870 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
871 * is used for "access()" and for the NFS daemon (letting nfsd stay at
872 * whatever uid it wants to). It normally shadows "euid", except when
873 * explicitly set by setfsuid() or for access..
874 */
875long __sys_setfsuid(uid_t uid)
876{
877 const struct cred *old;
878 struct cred *new;
879 uid_t old_fsuid;
880 kuid_t kuid;
881
882 old = current_cred();
883 old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
884
885 kuid = make_kuid(old->user_ns, uid);
886 if (!uid_valid(kuid))
887 return old_fsuid;
888
889 new = prepare_creds();
890 if (!new)
891 return old_fsuid;
892
893 if (uid_eq(kuid, old->uid) || uid_eq(kuid, old->euid) ||
894 uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
895 ns_capable_setid(old->user_ns, CAP_SETUID)) {
896 if (!uid_eq(kuid, old->fsuid)) {
897 new->fsuid = kuid;
898 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
899 goto change_okay;
900 }
901 }
902
903 abort_creds(new);
904 return old_fsuid;
905
906change_okay:
907 commit_creds(new);
908 return old_fsuid;
909}
910
911SYSCALL_DEFINE1(setfsuid, uid_t, uid)
912{
913 return __sys_setfsuid(uid);
914}
915
916/*
917 * Samma på svenska..
918 */
919long __sys_setfsgid(gid_t gid)
920{
921 const struct cred *old;
922 struct cred *new;
923 gid_t old_fsgid;
924 kgid_t kgid;
925
926 old = current_cred();
927 old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
928
929 kgid = make_kgid(old->user_ns, gid);
930 if (!gid_valid(kgid))
931 return old_fsgid;
932
933 new = prepare_creds();
934 if (!new)
935 return old_fsgid;
936
937 if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->egid) ||
938 gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
939 ns_capable_setid(old->user_ns, CAP_SETGID)) {
940 if (!gid_eq(kgid, old->fsgid)) {
941 new->fsgid = kgid;
942 if (security_task_fix_setgid(new,old,LSM_SETID_FS) == 0)
943 goto change_okay;
944 }
945 }
946
947 abort_creds(new);
948 return old_fsgid;
949
950change_okay:
951 commit_creds(new);
952 return old_fsgid;
953}
954
955SYSCALL_DEFINE1(setfsgid, gid_t, gid)
956{
957 return __sys_setfsgid(gid);
958}
959#endif /* CONFIG_MULTIUSER */
960
961/**
962 * sys_getpid - return the thread group id of the current process
963 *
964 * Note, despite the name, this returns the tgid not the pid. The tgid and
965 * the pid are identical unless CLONE_THREAD was specified on clone() in
966 * which case the tgid is the same in all threads of the same group.
967 *
968 * This is SMP safe as current->tgid does not change.
969 */
970SYSCALL_DEFINE0(getpid)
971{
972 return task_tgid_vnr(current);
973}
974
975/* Thread ID - the internal kernel "pid" */
976SYSCALL_DEFINE0(gettid)
977{
978 return task_pid_vnr(current);
979}
980
981/*
982 * Accessing ->real_parent is not SMP-safe, it could
983 * change from under us. However, we can use a stale
984 * value of ->real_parent under rcu_read_lock(), see
985 * release_task()->call_rcu(delayed_put_task_struct).
986 */
987SYSCALL_DEFINE0(getppid)
988{
989 int pid;
990
991 rcu_read_lock();
992 pid = task_tgid_vnr(rcu_dereference(current->real_parent));
993 rcu_read_unlock();
994
995 return pid;
996}
997
998SYSCALL_DEFINE0(getuid)
999{
1000 /* Only we change this so SMP safe */
1001 return from_kuid_munged(current_user_ns(), current_uid());
1002}
1003
1004SYSCALL_DEFINE0(geteuid)
1005{
1006 /* Only we change this so SMP safe */
1007 return from_kuid_munged(current_user_ns(), current_euid());
1008}
1009
1010SYSCALL_DEFINE0(getgid)
1011{
1012 /* Only we change this so SMP safe */
1013 return from_kgid_munged(current_user_ns(), current_gid());
1014}
1015
1016SYSCALL_DEFINE0(getegid)
1017{
1018 /* Only we change this so SMP safe */
1019 return from_kgid_munged(current_user_ns(), current_egid());
1020}
1021
1022static void do_sys_times(struct tms *tms)
1023{
1024 u64 tgutime, tgstime, cutime, cstime;
1025
1026 thread_group_cputime_adjusted(current, &tgutime, &tgstime);
1027 cutime = current->signal->cutime;
1028 cstime = current->signal->cstime;
1029 tms->tms_utime = nsec_to_clock_t(tgutime);
1030 tms->tms_stime = nsec_to_clock_t(tgstime);
1031 tms->tms_cutime = nsec_to_clock_t(cutime);
1032 tms->tms_cstime = nsec_to_clock_t(cstime);
1033}
1034
1035SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
1036{
1037 if (tbuf) {
1038 struct tms tmp;
1039
1040 do_sys_times(&tmp);
1041 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1042 return -EFAULT;
1043 }
1044 force_successful_syscall_return();
1045 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1046}
1047
1048#ifdef CONFIG_COMPAT
1049static compat_clock_t clock_t_to_compat_clock_t(clock_t x)
1050{
1051 return compat_jiffies_to_clock_t(clock_t_to_jiffies(x));
1052}
1053
1054COMPAT_SYSCALL_DEFINE1(times, struct compat_tms __user *, tbuf)
1055{
1056 if (tbuf) {
1057 struct tms tms;
1058 struct compat_tms tmp;
1059
1060 do_sys_times(&tms);
1061 /* Convert our struct tms to the compat version. */
1062 tmp.tms_utime = clock_t_to_compat_clock_t(tms.tms_utime);
1063 tmp.tms_stime = clock_t_to_compat_clock_t(tms.tms_stime);
1064 tmp.tms_cutime = clock_t_to_compat_clock_t(tms.tms_cutime);
1065 tmp.tms_cstime = clock_t_to_compat_clock_t(tms.tms_cstime);
1066 if (copy_to_user(tbuf, &tmp, sizeof(tmp)))
1067 return -EFAULT;
1068 }
1069 force_successful_syscall_return();
1070 return compat_jiffies_to_clock_t(jiffies);
1071}
1072#endif
1073
1074/*
1075 * This needs some heavy checking ...
1076 * I just haven't the stomach for it. I also don't fully
1077 * understand sessions/pgrp etc. Let somebody who does explain it.
1078 *
1079 * OK, I think I have the protection semantics right.... this is really
1080 * only important on a multi-user system anyway, to make sure one user
1081 * can't send a signal to a process owned by another. -TYT, 12/12/91
1082 *
1083 * !PF_FORKNOEXEC check to conform completely to POSIX.
1084 */
1085SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
1086{
1087 struct task_struct *p;
1088 struct task_struct *group_leader = current->group_leader;
1089 struct pid *pids[PIDTYPE_MAX] = { 0 };
1090 struct pid *pgrp;
1091 int err;
1092
1093 if (!pid)
1094 pid = task_pid_vnr(group_leader);
1095 if (!pgid)
1096 pgid = pid;
1097 if (pgid < 0)
1098 return -EINVAL;
1099 rcu_read_lock();
1100
1101 /* From this point forward we keep holding onto the tasklist lock
1102 * so that our parent does not change from under us. -DaveM
1103 */
1104 write_lock_irq(&tasklist_lock);
1105
1106 err = -ESRCH;
1107 p = find_task_by_vpid(pid);
1108 if (!p)
1109 goto out;
1110
1111 err = -EINVAL;
1112 if (!thread_group_leader(p))
1113 goto out;
1114
1115 if (same_thread_group(p->real_parent, group_leader)) {
1116 err = -EPERM;
1117 if (task_session(p) != task_session(group_leader))
1118 goto out;
1119 err = -EACCES;
1120 if (!(p->flags & PF_FORKNOEXEC))
1121 goto out;
1122 } else {
1123 err = -ESRCH;
1124 if (p != group_leader)
1125 goto out;
1126 }
1127
1128 err = -EPERM;
1129 if (p->signal->leader)
1130 goto out;
1131
1132 pgrp = task_pid(p);
1133 if (pgid != pid) {
1134 struct task_struct *g;
1135
1136 pgrp = find_vpid(pgid);
1137 g = pid_task(pgrp, PIDTYPE_PGID);
1138 if (!g || task_session(g) != task_session(group_leader))
1139 goto out;
1140 }
1141
1142 err = security_task_setpgid(p, pgid);
1143 if (err)
1144 goto out;
1145
1146 if (task_pgrp(p) != pgrp)
1147 change_pid(pids, p, PIDTYPE_PGID, pgrp);
1148
1149 err = 0;
1150out:
1151 /* All paths lead to here, thus we are safe. -DaveM */
1152 write_unlock_irq(&tasklist_lock);
1153 rcu_read_unlock();
1154 free_pids(pids);
1155 return err;
1156}
1157
1158static int do_getpgid(pid_t pid)
1159{
1160 struct task_struct *p;
1161 struct pid *grp;
1162 int retval;
1163
1164 rcu_read_lock();
1165 if (!pid)
1166 grp = task_pgrp(current);
1167 else {
1168 retval = -ESRCH;
1169 p = find_task_by_vpid(pid);
1170 if (!p)
1171 goto out;
1172 grp = task_pgrp(p);
1173 if (!grp)
1174 goto out;
1175
1176 retval = security_task_getpgid(p);
1177 if (retval)
1178 goto out;
1179 }
1180 retval = pid_vnr(grp);
1181out:
1182 rcu_read_unlock();
1183 return retval;
1184}
1185
1186SYSCALL_DEFINE1(getpgid, pid_t, pid)
1187{
1188 return do_getpgid(pid);
1189}
1190
1191#ifdef __ARCH_WANT_SYS_GETPGRP
1192
1193SYSCALL_DEFINE0(getpgrp)
1194{
1195 return do_getpgid(0);
1196}
1197
1198#endif
1199
1200SYSCALL_DEFINE1(getsid, pid_t, pid)
1201{
1202 struct task_struct *p;
1203 struct pid *sid;
1204 int retval;
1205
1206 rcu_read_lock();
1207 if (!pid)
1208 sid = task_session(current);
1209 else {
1210 retval = -ESRCH;
1211 p = find_task_by_vpid(pid);
1212 if (!p)
1213 goto out;
1214 sid = task_session(p);
1215 if (!sid)
1216 goto out;
1217
1218 retval = security_task_getsid(p);
1219 if (retval)
1220 goto out;
1221 }
1222 retval = pid_vnr(sid);
1223out:
1224 rcu_read_unlock();
1225 return retval;
1226}
1227
1228static void set_special_pids(struct pid **pids, struct pid *pid)
1229{
1230 struct task_struct *curr = current->group_leader;
1231
1232 if (task_session(curr) != pid)
1233 change_pid(pids, curr, PIDTYPE_SID, pid);
1234
1235 if (task_pgrp(curr) != pid)
1236 change_pid(pids, curr, PIDTYPE_PGID, pid);
1237}
1238
1239int ksys_setsid(void)
1240{
1241 struct task_struct *group_leader = current->group_leader;
1242 struct pid *sid = task_pid(group_leader);
1243 struct pid *pids[PIDTYPE_MAX] = { 0 };
1244 pid_t session = pid_vnr(sid);
1245 int err = -EPERM;
1246
1247 write_lock_irq(&tasklist_lock);
1248 /* Fail if I am already a session leader */
1249 if (group_leader->signal->leader)
1250 goto out;
1251
1252 /* Fail if a process group id already exists that equals the
1253 * proposed session id.
1254 */
1255 if (pid_task(sid, PIDTYPE_PGID))
1256 goto out;
1257
1258 group_leader->signal->leader = 1;
1259 set_special_pids(pids, sid);
1260
1261 proc_clear_tty(group_leader);
1262
1263 err = session;
1264out:
1265 write_unlock_irq(&tasklist_lock);
1266 free_pids(pids);
1267 if (err > 0) {
1268 proc_sid_connector(group_leader);
1269 sched_autogroup_create_attach(group_leader);
1270 }
1271 return err;
1272}
1273
1274SYSCALL_DEFINE0(setsid)
1275{
1276 return ksys_setsid();
1277}
1278
1279DECLARE_RWSEM(uts_sem);
1280
1281#ifdef COMPAT_UTS_MACHINE
1282#define override_architecture(name) \
1283 (personality(current->personality) == PER_LINUX32 && \
1284 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1285 sizeof(COMPAT_UTS_MACHINE)))
1286#else
1287#define override_architecture(name) 0
1288#endif
1289
1290/*
1291 * Work around broken programs that cannot handle "Linux 3.0".
1292 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1293 * And we map 4.x and later versions to 2.6.60+x, so 4.0/5.0/6.0/... would be
1294 * 2.6.60.
1295 */
1296static int override_release(char __user *release, size_t len)
1297{
1298 int ret = 0;
1299
1300 if (current->personality & UNAME26) {
1301 const char *rest = UTS_RELEASE;
1302 char buf[65] = { 0 };
1303 int ndots = 0;
1304 unsigned v;
1305 size_t copy;
1306
1307 while (*rest) {
1308 if (*rest == '.' && ++ndots >= 3)
1309 break;
1310 if (!isdigit(*rest) && *rest != '.')
1311 break;
1312 rest++;
1313 }
1314 v = LINUX_VERSION_PATCHLEVEL + 60;
1315 copy = clamp_t(size_t, len, 1, sizeof(buf));
1316 copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1317 ret = copy_to_user(release, buf, copy + 1);
1318 }
1319 return ret;
1320}
1321
1322SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1323{
1324 struct new_utsname tmp;
1325
1326 down_read(&uts_sem);
1327 memcpy(&tmp, utsname(), sizeof(tmp));
1328 up_read(&uts_sem);
1329 if (copy_to_user(name, &tmp, sizeof(tmp)))
1330 return -EFAULT;
1331
1332 if (override_release(name->release, sizeof(name->release)))
1333 return -EFAULT;
1334 if (override_architecture(name))
1335 return -EFAULT;
1336 return 0;
1337}
1338
1339#ifdef __ARCH_WANT_SYS_OLD_UNAME
1340/*
1341 * Old cruft
1342 */
1343SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1344{
1345 struct old_utsname tmp;
1346
1347 if (!name)
1348 return -EFAULT;
1349
1350 down_read(&uts_sem);
1351 memcpy(&tmp, utsname(), sizeof(tmp));
1352 up_read(&uts_sem);
1353 if (copy_to_user(name, &tmp, sizeof(tmp)))
1354 return -EFAULT;
1355
1356 if (override_release(name->release, sizeof(name->release)))
1357 return -EFAULT;
1358 if (override_architecture(name))
1359 return -EFAULT;
1360 return 0;
1361}
1362
1363SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1364{
1365 struct oldold_utsname tmp;
1366
1367 if (!name)
1368 return -EFAULT;
1369
1370 memset(&tmp, 0, sizeof(tmp));
1371
1372 down_read(&uts_sem);
1373 memcpy(&tmp.sysname, &utsname()->sysname, __OLD_UTS_LEN);
1374 memcpy(&tmp.nodename, &utsname()->nodename, __OLD_UTS_LEN);
1375 memcpy(&tmp.release, &utsname()->release, __OLD_UTS_LEN);
1376 memcpy(&tmp.version, &utsname()->version, __OLD_UTS_LEN);
1377 memcpy(&tmp.machine, &utsname()->machine, __OLD_UTS_LEN);
1378 up_read(&uts_sem);
1379 if (copy_to_user(name, &tmp, sizeof(tmp)))
1380 return -EFAULT;
1381
1382 if (override_architecture(name))
1383 return -EFAULT;
1384 if (override_release(name->release, sizeof(name->release)))
1385 return -EFAULT;
1386 return 0;
1387}
1388#endif
1389
1390SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1391{
1392 int errno;
1393 char tmp[__NEW_UTS_LEN];
1394
1395 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1396 return -EPERM;
1397
1398 if (len < 0 || len > __NEW_UTS_LEN)
1399 return -EINVAL;
1400 errno = -EFAULT;
1401 if (!copy_from_user(tmp, name, len)) {
1402 struct new_utsname *u;
1403
1404 add_device_randomness(tmp, len);
1405 down_write(&uts_sem);
1406 u = utsname();
1407 memcpy(u->nodename, tmp, len);
1408 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1409 errno = 0;
1410 uts_proc_notify(UTS_PROC_HOSTNAME);
1411 up_write(&uts_sem);
1412 }
1413 return errno;
1414}
1415
1416#ifdef __ARCH_WANT_SYS_GETHOSTNAME
1417
1418SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1419{
1420 int i;
1421 struct new_utsname *u;
1422 char tmp[__NEW_UTS_LEN + 1];
1423
1424 if (len < 0)
1425 return -EINVAL;
1426 down_read(&uts_sem);
1427 u = utsname();
1428 i = 1 + strlen(u->nodename);
1429 if (i > len)
1430 i = len;
1431 memcpy(tmp, u->nodename, i);
1432 up_read(&uts_sem);
1433 if (copy_to_user(name, tmp, i))
1434 return -EFAULT;
1435 return 0;
1436}
1437
1438#endif
1439
1440/*
1441 * Only setdomainname; getdomainname can be implemented by calling
1442 * uname()
1443 */
1444SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1445{
1446 int errno;
1447 char tmp[__NEW_UTS_LEN];
1448
1449 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1450 return -EPERM;
1451 if (len < 0 || len > __NEW_UTS_LEN)
1452 return -EINVAL;
1453
1454 errno = -EFAULT;
1455 if (!copy_from_user(tmp, name, len)) {
1456 struct new_utsname *u;
1457
1458 add_device_randomness(tmp, len);
1459 down_write(&uts_sem);
1460 u = utsname();
1461 memcpy(u->domainname, tmp, len);
1462 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1463 errno = 0;
1464 uts_proc_notify(UTS_PROC_DOMAINNAME);
1465 up_write(&uts_sem);
1466 }
1467 return errno;
1468}
1469
1470/* make sure you are allowed to change @tsk limits before calling this */
1471static int do_prlimit(struct task_struct *tsk, unsigned int resource,
1472 struct rlimit *new_rlim, struct rlimit *old_rlim)
1473{
1474 struct rlimit *rlim;
1475 int retval = 0;
1476
1477 if (resource >= RLIM_NLIMITS)
1478 return -EINVAL;
1479 resource = array_index_nospec(resource, RLIM_NLIMITS);
1480
1481 if (new_rlim) {
1482 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1483 return -EINVAL;
1484 if (resource == RLIMIT_NOFILE &&
1485 new_rlim->rlim_max > sysctl_nr_open)
1486 return -EPERM;
1487 }
1488
1489 /* Holding a refcount on tsk protects tsk->signal from disappearing. */
1490 rlim = tsk->signal->rlim + resource;
1491 task_lock(tsk->group_leader);
1492 if (new_rlim) {
1493 /*
1494 * Keep the capable check against init_user_ns until cgroups can
1495 * contain all limits.
1496 */
1497 if (new_rlim->rlim_max > rlim->rlim_max &&
1498 !capable(CAP_SYS_RESOURCE))
1499 retval = -EPERM;
1500 if (!retval)
1501 retval = security_task_setrlimit(tsk, resource, new_rlim);
1502 }
1503 if (!retval) {
1504 if (old_rlim)
1505 *old_rlim = *rlim;
1506 if (new_rlim)
1507 *rlim = *new_rlim;
1508 }
1509 task_unlock(tsk->group_leader);
1510
1511 /*
1512 * RLIMIT_CPU handling. Arm the posix CPU timer if the limit is not
1513 * infinite. In case of RLIM_INFINITY the posix CPU timer code
1514 * ignores the rlimit.
1515 */
1516 if (!retval && new_rlim && resource == RLIMIT_CPU &&
1517 new_rlim->rlim_cur != RLIM_INFINITY &&
1518 IS_ENABLED(CONFIG_POSIX_TIMERS)) {
1519 /*
1520 * update_rlimit_cpu can fail if the task is exiting, but there
1521 * may be other tasks in the thread group that are not exiting,
1522 * and they need their cpu timers adjusted.
1523 *
1524 * The group_leader is the last task to be released, so if we
1525 * cannot update_rlimit_cpu on it, then the entire process is
1526 * exiting and we do not need to update at all.
1527 */
1528 update_rlimit_cpu(tsk->group_leader, new_rlim->rlim_cur);
1529 }
1530
1531 return retval;
1532}
1533
1534SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1535{
1536 struct rlimit value;
1537 int ret;
1538
1539 ret = do_prlimit(current, resource, NULL, &value);
1540 if (!ret)
1541 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1542
1543 return ret;
1544}
1545
1546#ifdef CONFIG_COMPAT
1547
1548COMPAT_SYSCALL_DEFINE2(setrlimit, unsigned int, resource,
1549 struct compat_rlimit __user *, rlim)
1550{
1551 struct rlimit r;
1552 struct compat_rlimit r32;
1553
1554 if (copy_from_user(&r32, rlim, sizeof(struct compat_rlimit)))
1555 return -EFAULT;
1556
1557 if (r32.rlim_cur == COMPAT_RLIM_INFINITY)
1558 r.rlim_cur = RLIM_INFINITY;
1559 else
1560 r.rlim_cur = r32.rlim_cur;
1561 if (r32.rlim_max == COMPAT_RLIM_INFINITY)
1562 r.rlim_max = RLIM_INFINITY;
1563 else
1564 r.rlim_max = r32.rlim_max;
1565 return do_prlimit(current, resource, &r, NULL);
1566}
1567
1568COMPAT_SYSCALL_DEFINE2(getrlimit, unsigned int, resource,
1569 struct compat_rlimit __user *, rlim)
1570{
1571 struct rlimit r;
1572 int ret;
1573
1574 ret = do_prlimit(current, resource, NULL, &r);
1575 if (!ret) {
1576 struct compat_rlimit r32;
1577 if (r.rlim_cur > COMPAT_RLIM_INFINITY)
1578 r32.rlim_cur = COMPAT_RLIM_INFINITY;
1579 else
1580 r32.rlim_cur = r.rlim_cur;
1581 if (r.rlim_max > COMPAT_RLIM_INFINITY)
1582 r32.rlim_max = COMPAT_RLIM_INFINITY;
1583 else
1584 r32.rlim_max = r.rlim_max;
1585
1586 if (copy_to_user(rlim, &r32, sizeof(struct compat_rlimit)))
1587 return -EFAULT;
1588 }
1589 return ret;
1590}
1591
1592#endif
1593
1594#ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1595
1596/*
1597 * Back compatibility for getrlimit. Needed for some apps.
1598 */
1599SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1600 struct rlimit __user *, rlim)
1601{
1602 struct rlimit x;
1603 if (resource >= RLIM_NLIMITS)
1604 return -EINVAL;
1605
1606 resource = array_index_nospec(resource, RLIM_NLIMITS);
1607 task_lock(current->group_leader);
1608 x = current->signal->rlim[resource];
1609 task_unlock(current->group_leader);
1610 if (x.rlim_cur > 0x7FFFFFFF)
1611 x.rlim_cur = 0x7FFFFFFF;
1612 if (x.rlim_max > 0x7FFFFFFF)
1613 x.rlim_max = 0x7FFFFFFF;
1614 return copy_to_user(rlim, &x, sizeof(x)) ? -EFAULT : 0;
1615}
1616
1617#ifdef CONFIG_COMPAT
1618COMPAT_SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1619 struct compat_rlimit __user *, rlim)
1620{
1621 struct rlimit r;
1622
1623 if (resource >= RLIM_NLIMITS)
1624 return -EINVAL;
1625
1626 resource = array_index_nospec(resource, RLIM_NLIMITS);
1627 task_lock(current->group_leader);
1628 r = current->signal->rlim[resource];
1629 task_unlock(current->group_leader);
1630 if (r.rlim_cur > 0x7FFFFFFF)
1631 r.rlim_cur = 0x7FFFFFFF;
1632 if (r.rlim_max > 0x7FFFFFFF)
1633 r.rlim_max = 0x7FFFFFFF;
1634
1635 if (put_user(r.rlim_cur, &rlim->rlim_cur) ||
1636 put_user(r.rlim_max, &rlim->rlim_max))
1637 return -EFAULT;
1638 return 0;
1639}
1640#endif
1641
1642#endif
1643
1644static inline bool rlim64_is_infinity(__u64 rlim64)
1645{
1646#if BITS_PER_LONG < 64
1647 return rlim64 >= ULONG_MAX;
1648#else
1649 return rlim64 == RLIM64_INFINITY;
1650#endif
1651}
1652
1653static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1654{
1655 if (rlim->rlim_cur == RLIM_INFINITY)
1656 rlim64->rlim_cur = RLIM64_INFINITY;
1657 else
1658 rlim64->rlim_cur = rlim->rlim_cur;
1659 if (rlim->rlim_max == RLIM_INFINITY)
1660 rlim64->rlim_max = RLIM64_INFINITY;
1661 else
1662 rlim64->rlim_max = rlim->rlim_max;
1663}
1664
1665static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1666{
1667 if (rlim64_is_infinity(rlim64->rlim_cur))
1668 rlim->rlim_cur = RLIM_INFINITY;
1669 else
1670 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1671 if (rlim64_is_infinity(rlim64->rlim_max))
1672 rlim->rlim_max = RLIM_INFINITY;
1673 else
1674 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1675}
1676
1677/* rcu lock must be held */
1678static int check_prlimit_permission(struct task_struct *task,
1679 unsigned int flags)
1680{
1681 const struct cred *cred = current_cred(), *tcred;
1682 bool id_match;
1683
1684 if (current == task)
1685 return 0;
1686
1687 tcred = __task_cred(task);
1688 id_match = (uid_eq(cred->uid, tcred->euid) &&
1689 uid_eq(cred->uid, tcred->suid) &&
1690 uid_eq(cred->uid, tcred->uid) &&
1691 gid_eq(cred->gid, tcred->egid) &&
1692 gid_eq(cred->gid, tcred->sgid) &&
1693 gid_eq(cred->gid, tcred->gid));
1694 if (!id_match && !ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
1695 return -EPERM;
1696
1697 return security_task_prlimit(cred, tcred, flags);
1698}
1699
1700SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1701 const struct rlimit64 __user *, new_rlim,
1702 struct rlimit64 __user *, old_rlim)
1703{
1704 struct rlimit64 old64, new64;
1705 struct rlimit old, new;
1706 struct task_struct *tsk;
1707 unsigned int checkflags = 0;
1708 int ret;
1709
1710 if (old_rlim)
1711 checkflags |= LSM_PRLIMIT_READ;
1712
1713 if (new_rlim) {
1714 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1715 return -EFAULT;
1716 rlim64_to_rlim(&new64, &new);
1717 checkflags |= LSM_PRLIMIT_WRITE;
1718 }
1719
1720 rcu_read_lock();
1721 tsk = pid ? find_task_by_vpid(pid) : current;
1722 if (!tsk) {
1723 rcu_read_unlock();
1724 return -ESRCH;
1725 }
1726 ret = check_prlimit_permission(tsk, checkflags);
1727 if (ret) {
1728 rcu_read_unlock();
1729 return ret;
1730 }
1731 get_task_struct(tsk);
1732 rcu_read_unlock();
1733
1734 ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1735 old_rlim ? &old : NULL);
1736
1737 if (!ret && old_rlim) {
1738 rlim_to_rlim64(&old, &old64);
1739 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1740 ret = -EFAULT;
1741 }
1742
1743 put_task_struct(tsk);
1744 return ret;
1745}
1746
1747SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1748{
1749 struct rlimit new_rlim;
1750
1751 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1752 return -EFAULT;
1753 return do_prlimit(current, resource, &new_rlim, NULL);
1754}
1755
1756/*
1757 * It would make sense to put struct rusage in the task_struct,
1758 * except that would make the task_struct be *really big*. After
1759 * task_struct gets moved into malloc'ed memory, it would
1760 * make sense to do this. It will make moving the rest of the information
1761 * a lot simpler! (Which we're not doing right now because we're not
1762 * measuring them yet).
1763 *
1764 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1765 * races with threads incrementing their own counters. But since word
1766 * reads are atomic, we either get new values or old values and we don't
1767 * care which for the sums. We always take the siglock to protect reading
1768 * the c* fields from p->signal from races with exit.c updating those
1769 * fields when reaping, so a sample either gets all the additions of a
1770 * given child after it's reaped, or none so this sample is before reaping.
1771 *
1772 * Locking:
1773 * We need to take the siglock for CHILDEREN, SELF and BOTH
1774 * for the cases current multithreaded, non-current single threaded
1775 * non-current multithreaded. Thread traversal is now safe with
1776 * the siglock held.
1777 * Strictly speaking, we donot need to take the siglock if we are current and
1778 * single threaded, as no one else can take our signal_struct away, no one
1779 * else can reap the children to update signal->c* counters, and no one else
1780 * can race with the signal-> fields. If we do not take any lock, the
1781 * signal-> fields could be read out of order while another thread was just
1782 * exiting. So we should place a read memory barrier when we avoid the lock.
1783 * On the writer side, write memory barrier is implied in __exit_signal
1784 * as __exit_signal releases the siglock spinlock after updating the signal->
1785 * fields. But we don't do this yet to keep things simple.
1786 *
1787 */
1788
1789static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1790{
1791 r->ru_nvcsw += t->nvcsw;
1792 r->ru_nivcsw += t->nivcsw;
1793 r->ru_minflt += t->min_flt;
1794 r->ru_majflt += t->maj_flt;
1795 r->ru_inblock += task_io_get_inblock(t);
1796 r->ru_oublock += task_io_get_oublock(t);
1797}
1798
1799void getrusage(struct task_struct *p, int who, struct rusage *r)
1800{
1801 struct task_struct *t;
1802 unsigned long flags;
1803 u64 tgutime, tgstime, utime, stime;
1804 unsigned long maxrss;
1805 struct mm_struct *mm;
1806 struct signal_struct *sig = p->signal;
1807 unsigned int seq = 0;
1808
1809retry:
1810 memset(r, 0, sizeof(*r));
1811 utime = stime = 0;
1812 maxrss = 0;
1813
1814 if (who == RUSAGE_THREAD) {
1815 task_cputime_adjusted(current, &utime, &stime);
1816 accumulate_thread_rusage(p, r);
1817 maxrss = sig->maxrss;
1818 goto out_thread;
1819 }
1820
1821 flags = read_seqbegin_or_lock_irqsave(&sig->stats_lock, &seq);
1822
1823 switch (who) {
1824 case RUSAGE_BOTH:
1825 case RUSAGE_CHILDREN:
1826 utime = sig->cutime;
1827 stime = sig->cstime;
1828 r->ru_nvcsw = sig->cnvcsw;
1829 r->ru_nivcsw = sig->cnivcsw;
1830 r->ru_minflt = sig->cmin_flt;
1831 r->ru_majflt = sig->cmaj_flt;
1832 r->ru_inblock = sig->cinblock;
1833 r->ru_oublock = sig->coublock;
1834 maxrss = sig->cmaxrss;
1835
1836 if (who == RUSAGE_CHILDREN)
1837 break;
1838 fallthrough;
1839
1840 case RUSAGE_SELF:
1841 r->ru_nvcsw += sig->nvcsw;
1842 r->ru_nivcsw += sig->nivcsw;
1843 r->ru_minflt += sig->min_flt;
1844 r->ru_majflt += sig->maj_flt;
1845 r->ru_inblock += sig->inblock;
1846 r->ru_oublock += sig->oublock;
1847 if (maxrss < sig->maxrss)
1848 maxrss = sig->maxrss;
1849
1850 rcu_read_lock();
1851 __for_each_thread(sig, t)
1852 accumulate_thread_rusage(t, r);
1853 rcu_read_unlock();
1854
1855 break;
1856
1857 default:
1858 BUG();
1859 }
1860
1861 if (need_seqretry(&sig->stats_lock, seq)) {
1862 seq = 1;
1863 goto retry;
1864 }
1865 done_seqretry_irqrestore(&sig->stats_lock, seq, flags);
1866
1867 if (who == RUSAGE_CHILDREN)
1868 goto out_children;
1869
1870 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1871 utime += tgutime;
1872 stime += tgstime;
1873
1874out_thread:
1875 mm = get_task_mm(p);
1876 if (mm) {
1877 setmax_mm_hiwater_rss(&maxrss, mm);
1878 mmput(mm);
1879 }
1880
1881out_children:
1882 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1883 r->ru_utime = ns_to_kernel_old_timeval(utime);
1884 r->ru_stime = ns_to_kernel_old_timeval(stime);
1885}
1886
1887SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1888{
1889 struct rusage r;
1890
1891 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1892 who != RUSAGE_THREAD)
1893 return -EINVAL;
1894
1895 getrusage(current, who, &r);
1896 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1897}
1898
1899#ifdef CONFIG_COMPAT
1900COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru)
1901{
1902 struct rusage r;
1903
1904 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1905 who != RUSAGE_THREAD)
1906 return -EINVAL;
1907
1908 getrusage(current, who, &r);
1909 return put_compat_rusage(&r, ru);
1910}
1911#endif
1912
1913SYSCALL_DEFINE1(umask, int, mask)
1914{
1915 mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1916 return mask;
1917}
1918
1919static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1920{
1921 CLASS(fd, exe)(fd);
1922 struct inode *inode;
1923 int err;
1924
1925 if (fd_empty(exe))
1926 return -EBADF;
1927
1928 inode = file_inode(fd_file(exe));
1929
1930 /*
1931 * Because the original mm->exe_file points to executable file, make
1932 * sure that this one is executable as well, to avoid breaking an
1933 * overall picture.
1934 */
1935 if (!S_ISREG(inode->i_mode) || path_noexec(&fd_file(exe)->f_path))
1936 return -EACCES;
1937
1938 err = file_permission(fd_file(exe), MAY_EXEC);
1939 if (err)
1940 return err;
1941
1942 return replace_mm_exe_file(mm, fd_file(exe));
1943}
1944
1945/*
1946 * Check arithmetic relations of passed addresses.
1947 *
1948 * WARNING: we don't require any capability here so be very careful
1949 * in what is allowed for modification from userspace.
1950 */
1951static int validate_prctl_map_addr(struct prctl_mm_map *prctl_map)
1952{
1953 unsigned long mmap_max_addr = TASK_SIZE;
1954 int error = -EINVAL, i;
1955
1956 static const unsigned char offsets[] = {
1957 offsetof(struct prctl_mm_map, start_code),
1958 offsetof(struct prctl_mm_map, end_code),
1959 offsetof(struct prctl_mm_map, start_data),
1960 offsetof(struct prctl_mm_map, end_data),
1961 offsetof(struct prctl_mm_map, start_brk),
1962 offsetof(struct prctl_mm_map, brk),
1963 offsetof(struct prctl_mm_map, start_stack),
1964 offsetof(struct prctl_mm_map, arg_start),
1965 offsetof(struct prctl_mm_map, arg_end),
1966 offsetof(struct prctl_mm_map, env_start),
1967 offsetof(struct prctl_mm_map, env_end),
1968 };
1969
1970 /*
1971 * Make sure the members are not somewhere outside
1972 * of allowed address space.
1973 */
1974 for (i = 0; i < ARRAY_SIZE(offsets); i++) {
1975 u64 val = *(u64 *)((char *)prctl_map + offsets[i]);
1976
1977 if ((unsigned long)val >= mmap_max_addr ||
1978 (unsigned long)val < mmap_min_addr)
1979 goto out;
1980 }
1981
1982 /*
1983 * Make sure the pairs are ordered.
1984 */
1985#define __prctl_check_order(__m1, __op, __m2) \
1986 ((unsigned long)prctl_map->__m1 __op \
1987 (unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1988 error = __prctl_check_order(start_code, <, end_code);
1989 error |= __prctl_check_order(start_data,<=, end_data);
1990 error |= __prctl_check_order(start_brk, <=, brk);
1991 error |= __prctl_check_order(arg_start, <=, arg_end);
1992 error |= __prctl_check_order(env_start, <=, env_end);
1993 if (error)
1994 goto out;
1995#undef __prctl_check_order
1996
1997 error = -EINVAL;
1998
1999 /*
2000 * Neither we should allow to override limits if they set.
2001 */
2002 if (check_data_rlimit(rlimit(RLIMIT_DATA), prctl_map->brk,
2003 prctl_map->start_brk, prctl_map->end_data,
2004 prctl_map->start_data))
2005 goto out;
2006
2007 error = 0;
2008out:
2009 return error;
2010}
2011
2012#ifdef CONFIG_CHECKPOINT_RESTORE
2013static int prctl_set_mm_map(int opt, const void __user *addr, unsigned long data_size)
2014{
2015 struct prctl_mm_map prctl_map = { .exe_fd = (u32)-1, };
2016 unsigned long user_auxv[AT_VECTOR_SIZE];
2017 struct mm_struct *mm = current->mm;
2018 int error;
2019
2020 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
2021 BUILD_BUG_ON(sizeof(struct prctl_mm_map) > 256);
2022
2023 if (opt == PR_SET_MM_MAP_SIZE)
2024 return put_user((unsigned int)sizeof(prctl_map),
2025 (unsigned int __user *)addr);
2026
2027 if (data_size != sizeof(prctl_map))
2028 return -EINVAL;
2029
2030 if (copy_from_user(&prctl_map, addr, sizeof(prctl_map)))
2031 return -EFAULT;
2032
2033 error = validate_prctl_map_addr(&prctl_map);
2034 if (error)
2035 return error;
2036
2037 if (prctl_map.auxv_size) {
2038 /*
2039 * Someone is trying to cheat the auxv vector.
2040 */
2041 if (!prctl_map.auxv ||
2042 prctl_map.auxv_size > sizeof(mm->saved_auxv))
2043 return -EINVAL;
2044
2045 memset(user_auxv, 0, sizeof(user_auxv));
2046 if (copy_from_user(user_auxv,
2047 (const void __user *)prctl_map.auxv,
2048 prctl_map.auxv_size))
2049 return -EFAULT;
2050
2051 /* Last entry must be AT_NULL as specification requires */
2052 user_auxv[AT_VECTOR_SIZE - 2] = AT_NULL;
2053 user_auxv[AT_VECTOR_SIZE - 1] = AT_NULL;
2054 }
2055
2056 if (prctl_map.exe_fd != (u32)-1) {
2057 /*
2058 * Check if the current user is checkpoint/restore capable.
2059 * At the time of this writing, it checks for CAP_SYS_ADMIN
2060 * or CAP_CHECKPOINT_RESTORE.
2061 * Note that a user with access to ptrace can masquerade an
2062 * arbitrary program as any executable, even setuid ones.
2063 * This may have implications in the tomoyo subsystem.
2064 */
2065 if (!checkpoint_restore_ns_capable(current_user_ns()))
2066 return -EPERM;
2067
2068 error = prctl_set_mm_exe_file(mm, prctl_map.exe_fd);
2069 if (error)
2070 return error;
2071 }
2072
2073 /*
2074 * arg_lock protects concurrent updates but we still need mmap_lock for
2075 * read to exclude races with sys_brk.
2076 */
2077 mmap_read_lock(mm);
2078
2079 /*
2080 * We don't validate if these members are pointing to
2081 * real present VMAs because application may have correspond
2082 * VMAs already unmapped and kernel uses these members for statistics
2083 * output in procfs mostly, except
2084 *
2085 * - @start_brk/@brk which are used in do_brk_flags but kernel lookups
2086 * for VMAs when updating these members so anything wrong written
2087 * here cause kernel to swear at userspace program but won't lead
2088 * to any problem in kernel itself
2089 */
2090
2091 spin_lock(&mm->arg_lock);
2092 mm->start_code = prctl_map.start_code;
2093 mm->end_code = prctl_map.end_code;
2094 mm->start_data = prctl_map.start_data;
2095 mm->end_data = prctl_map.end_data;
2096 mm->start_brk = prctl_map.start_brk;
2097 mm->brk = prctl_map.brk;
2098 mm->start_stack = prctl_map.start_stack;
2099 mm->arg_start = prctl_map.arg_start;
2100 mm->arg_end = prctl_map.arg_end;
2101 mm->env_start = prctl_map.env_start;
2102 mm->env_end = prctl_map.env_end;
2103 spin_unlock(&mm->arg_lock);
2104
2105 /*
2106 * Note this update of @saved_auxv is lockless thus
2107 * if someone reads this member in procfs while we're
2108 * updating -- it may get partly updated results. It's
2109 * known and acceptable trade off: we leave it as is to
2110 * not introduce additional locks here making the kernel
2111 * more complex.
2112 */
2113 if (prctl_map.auxv_size)
2114 memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv));
2115
2116 mmap_read_unlock(mm);
2117 return 0;
2118}
2119#endif /* CONFIG_CHECKPOINT_RESTORE */
2120
2121static int prctl_set_auxv(struct mm_struct *mm, unsigned long addr,
2122 unsigned long len)
2123{
2124 /*
2125 * This doesn't move the auxiliary vector itself since it's pinned to
2126 * mm_struct, but it permits filling the vector with new values. It's
2127 * up to the caller to provide sane values here, otherwise userspace
2128 * tools which use this vector might be unhappy.
2129 */
2130 unsigned long user_auxv[AT_VECTOR_SIZE] = {};
2131
2132 if (len > sizeof(user_auxv))
2133 return -EINVAL;
2134
2135 if (copy_from_user(user_auxv, (const void __user *)addr, len))
2136 return -EFAULT;
2137
2138 /* Make sure the last entry is always AT_NULL */
2139 user_auxv[AT_VECTOR_SIZE - 2] = 0;
2140 user_auxv[AT_VECTOR_SIZE - 1] = 0;
2141
2142 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
2143
2144 task_lock(current);
2145 memcpy(mm->saved_auxv, user_auxv, len);
2146 task_unlock(current);
2147
2148 return 0;
2149}
2150
2151static int prctl_set_mm(int opt, unsigned long addr,
2152 unsigned long arg4, unsigned long arg5)
2153{
2154 struct mm_struct *mm = current->mm;
2155 struct prctl_mm_map prctl_map = {
2156 .auxv = NULL,
2157 .auxv_size = 0,
2158 .exe_fd = -1,
2159 };
2160 struct vm_area_struct *vma;
2161 int error;
2162
2163 if (arg5 || (arg4 && (opt != PR_SET_MM_AUXV &&
2164 opt != PR_SET_MM_MAP &&
2165 opt != PR_SET_MM_MAP_SIZE)))
2166 return -EINVAL;
2167
2168#ifdef CONFIG_CHECKPOINT_RESTORE
2169 if (opt == PR_SET_MM_MAP || opt == PR_SET_MM_MAP_SIZE)
2170 return prctl_set_mm_map(opt, (const void __user *)addr, arg4);
2171#endif
2172
2173 if (!capable(CAP_SYS_RESOURCE))
2174 return -EPERM;
2175
2176 if (opt == PR_SET_MM_EXE_FILE)
2177 return prctl_set_mm_exe_file(mm, (unsigned int)addr);
2178
2179 if (opt == PR_SET_MM_AUXV)
2180 return prctl_set_auxv(mm, addr, arg4);
2181
2182 if (addr >= TASK_SIZE || addr < mmap_min_addr)
2183 return -EINVAL;
2184
2185 error = -EINVAL;
2186
2187 /*
2188 * arg_lock protects concurrent updates of arg boundaries, we need
2189 * mmap_lock for a) concurrent sys_brk, b) finding VMA for addr
2190 * validation.
2191 */
2192 mmap_read_lock(mm);
2193 vma = find_vma(mm, addr);
2194
2195 spin_lock(&mm->arg_lock);
2196 prctl_map.start_code = mm->start_code;
2197 prctl_map.end_code = mm->end_code;
2198 prctl_map.start_data = mm->start_data;
2199 prctl_map.end_data = mm->end_data;
2200 prctl_map.start_brk = mm->start_brk;
2201 prctl_map.brk = mm->brk;
2202 prctl_map.start_stack = mm->start_stack;
2203 prctl_map.arg_start = mm->arg_start;
2204 prctl_map.arg_end = mm->arg_end;
2205 prctl_map.env_start = mm->env_start;
2206 prctl_map.env_end = mm->env_end;
2207
2208 switch (opt) {
2209 case PR_SET_MM_START_CODE:
2210 prctl_map.start_code = addr;
2211 break;
2212 case PR_SET_MM_END_CODE:
2213 prctl_map.end_code = addr;
2214 break;
2215 case PR_SET_MM_START_DATA:
2216 prctl_map.start_data = addr;
2217 break;
2218 case PR_SET_MM_END_DATA:
2219 prctl_map.end_data = addr;
2220 break;
2221 case PR_SET_MM_START_STACK:
2222 prctl_map.start_stack = addr;
2223 break;
2224 case PR_SET_MM_START_BRK:
2225 prctl_map.start_brk = addr;
2226 break;
2227 case PR_SET_MM_BRK:
2228 prctl_map.brk = addr;
2229 break;
2230 case PR_SET_MM_ARG_START:
2231 prctl_map.arg_start = addr;
2232 break;
2233 case PR_SET_MM_ARG_END:
2234 prctl_map.arg_end = addr;
2235 break;
2236 case PR_SET_MM_ENV_START:
2237 prctl_map.env_start = addr;
2238 break;
2239 case PR_SET_MM_ENV_END:
2240 prctl_map.env_end = addr;
2241 break;
2242 default:
2243 goto out;
2244 }
2245
2246 error = validate_prctl_map_addr(&prctl_map);
2247 if (error)
2248 goto out;
2249
2250 switch (opt) {
2251 /*
2252 * If command line arguments and environment
2253 * are placed somewhere else on stack, we can
2254 * set them up here, ARG_START/END to setup
2255 * command line arguments and ENV_START/END
2256 * for environment.
2257 */
2258 case PR_SET_MM_START_STACK:
2259 case PR_SET_MM_ARG_START:
2260 case PR_SET_MM_ARG_END:
2261 case PR_SET_MM_ENV_START:
2262 case PR_SET_MM_ENV_END:
2263 if (!vma) {
2264 error = -EFAULT;
2265 goto out;
2266 }
2267 }
2268
2269 mm->start_code = prctl_map.start_code;
2270 mm->end_code = prctl_map.end_code;
2271 mm->start_data = prctl_map.start_data;
2272 mm->end_data = prctl_map.end_data;
2273 mm->start_brk = prctl_map.start_brk;
2274 mm->brk = prctl_map.brk;
2275 mm->start_stack = prctl_map.start_stack;
2276 mm->arg_start = prctl_map.arg_start;
2277 mm->arg_end = prctl_map.arg_end;
2278 mm->env_start = prctl_map.env_start;
2279 mm->env_end = prctl_map.env_end;
2280
2281 error = 0;
2282out:
2283 spin_unlock(&mm->arg_lock);
2284 mmap_read_unlock(mm);
2285 return error;
2286}
2287
2288#ifdef CONFIG_CHECKPOINT_RESTORE
2289static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr)
2290{
2291 return put_user(me->clear_child_tid, tid_addr);
2292}
2293#else
2294static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr)
2295{
2296 return -EINVAL;
2297}
2298#endif
2299
2300static int propagate_has_child_subreaper(struct task_struct *p, void *data)
2301{
2302 /*
2303 * If task has has_child_subreaper - all its descendants
2304 * already have these flag too and new descendants will
2305 * inherit it on fork, skip them.
2306 *
2307 * If we've found child_reaper - skip descendants in
2308 * it's subtree as they will never get out pidns.
2309 */
2310 if (p->signal->has_child_subreaper ||
2311 is_child_reaper(task_pid(p)))
2312 return 0;
2313
2314 p->signal->has_child_subreaper = 1;
2315 return 1;
2316}
2317
2318int __weak arch_prctl_spec_ctrl_get(struct task_struct *t, unsigned long which)
2319{
2320 return -EINVAL;
2321}
2322
2323int __weak arch_prctl_spec_ctrl_set(struct task_struct *t, unsigned long which,
2324 unsigned long ctrl)
2325{
2326 return -EINVAL;
2327}
2328
2329int __weak arch_get_shadow_stack_status(struct task_struct *t, unsigned long __user *status)
2330{
2331 return -EINVAL;
2332}
2333
2334int __weak arch_set_shadow_stack_status(struct task_struct *t, unsigned long status)
2335{
2336 return -EINVAL;
2337}
2338
2339int __weak arch_lock_shadow_stack_status(struct task_struct *t, unsigned long status)
2340{
2341 return -EINVAL;
2342}
2343
2344#define PR_IO_FLUSHER (PF_MEMALLOC_NOIO | PF_LOCAL_THROTTLE)
2345
2346#ifdef CONFIG_ANON_VMA_NAME
2347
2348#define ANON_VMA_NAME_MAX_LEN 80
2349#define ANON_VMA_NAME_INVALID_CHARS "\\`$[]"
2350
2351static inline bool is_valid_name_char(char ch)
2352{
2353 /* printable ascii characters, excluding ANON_VMA_NAME_INVALID_CHARS */
2354 return ch > 0x1f && ch < 0x7f &&
2355 !strchr(ANON_VMA_NAME_INVALID_CHARS, ch);
2356}
2357
2358static int prctl_set_vma(unsigned long opt, unsigned long addr,
2359 unsigned long size, unsigned long arg)
2360{
2361 struct mm_struct *mm = current->mm;
2362 const char __user *uname;
2363 struct anon_vma_name *anon_name = NULL;
2364 int error;
2365
2366 switch (opt) {
2367 case PR_SET_VMA_ANON_NAME:
2368 uname = (const char __user *)arg;
2369 if (uname) {
2370 char *name, *pch;
2371
2372 name = strndup_user(uname, ANON_VMA_NAME_MAX_LEN);
2373 if (IS_ERR(name))
2374 return PTR_ERR(name);
2375
2376 for (pch = name; *pch != '\0'; pch++) {
2377 if (!is_valid_name_char(*pch)) {
2378 kfree(name);
2379 return -EINVAL;
2380 }
2381 }
2382 /* anon_vma has its own copy */
2383 anon_name = anon_vma_name_alloc(name);
2384 kfree(name);
2385 if (!anon_name)
2386 return -ENOMEM;
2387
2388 }
2389
2390 mmap_write_lock(mm);
2391 error = madvise_set_anon_name(mm, addr, size, anon_name);
2392 mmap_write_unlock(mm);
2393 anon_vma_name_put(anon_name);
2394 break;
2395 default:
2396 error = -EINVAL;
2397 }
2398
2399 return error;
2400}
2401
2402#else /* CONFIG_ANON_VMA_NAME */
2403static int prctl_set_vma(unsigned long opt, unsigned long start,
2404 unsigned long size, unsigned long arg)
2405{
2406 return -EINVAL;
2407}
2408#endif /* CONFIG_ANON_VMA_NAME */
2409
2410static inline unsigned long get_current_mdwe(void)
2411{
2412 unsigned long ret = 0;
2413
2414 if (test_bit(MMF_HAS_MDWE, &current->mm->flags))
2415 ret |= PR_MDWE_REFUSE_EXEC_GAIN;
2416 if (test_bit(MMF_HAS_MDWE_NO_INHERIT, &current->mm->flags))
2417 ret |= PR_MDWE_NO_INHERIT;
2418
2419 return ret;
2420}
2421
2422static inline int prctl_set_mdwe(unsigned long bits, unsigned long arg3,
2423 unsigned long arg4, unsigned long arg5)
2424{
2425 unsigned long current_bits;
2426
2427 if (arg3 || arg4 || arg5)
2428 return -EINVAL;
2429
2430 if (bits & ~(PR_MDWE_REFUSE_EXEC_GAIN | PR_MDWE_NO_INHERIT))
2431 return -EINVAL;
2432
2433 /* NO_INHERIT only makes sense with REFUSE_EXEC_GAIN */
2434 if (bits & PR_MDWE_NO_INHERIT && !(bits & PR_MDWE_REFUSE_EXEC_GAIN))
2435 return -EINVAL;
2436
2437 /*
2438 * EOPNOTSUPP might be more appropriate here in principle, but
2439 * existing userspace depends on EINVAL specifically.
2440 */
2441 if (!arch_memory_deny_write_exec_supported())
2442 return -EINVAL;
2443
2444 current_bits = get_current_mdwe();
2445 if (current_bits && current_bits != bits)
2446 return -EPERM; /* Cannot unset the flags */
2447
2448 if (bits & PR_MDWE_NO_INHERIT)
2449 set_bit(MMF_HAS_MDWE_NO_INHERIT, &current->mm->flags);
2450 if (bits & PR_MDWE_REFUSE_EXEC_GAIN)
2451 set_bit(MMF_HAS_MDWE, &current->mm->flags);
2452
2453 return 0;
2454}
2455
2456static inline int prctl_get_mdwe(unsigned long arg2, unsigned long arg3,
2457 unsigned long arg4, unsigned long arg5)
2458{
2459 if (arg2 || arg3 || arg4 || arg5)
2460 return -EINVAL;
2461 return get_current_mdwe();
2462}
2463
2464static int prctl_get_auxv(void __user *addr, unsigned long len)
2465{
2466 struct mm_struct *mm = current->mm;
2467 unsigned long size = min_t(unsigned long, sizeof(mm->saved_auxv), len);
2468
2469 if (size && copy_to_user(addr, mm->saved_auxv, size))
2470 return -EFAULT;
2471 return sizeof(mm->saved_auxv);
2472}
2473
2474SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2475 unsigned long, arg4, unsigned long, arg5)
2476{
2477 struct task_struct *me = current;
2478 unsigned char comm[sizeof(me->comm)];
2479 long error;
2480
2481 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2482 if (error != -ENOSYS)
2483 return error;
2484
2485 error = 0;
2486 switch (option) {
2487 case PR_SET_PDEATHSIG:
2488 if (!valid_signal(arg2)) {
2489 error = -EINVAL;
2490 break;
2491 }
2492 me->pdeath_signal = arg2;
2493 break;
2494 case PR_GET_PDEATHSIG:
2495 error = put_user(me->pdeath_signal, (int __user *)arg2);
2496 break;
2497 case PR_GET_DUMPABLE:
2498 error = get_dumpable(me->mm);
2499 break;
2500 case PR_SET_DUMPABLE:
2501 if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) {
2502 error = -EINVAL;
2503 break;
2504 }
2505 set_dumpable(me->mm, arg2);
2506 break;
2507
2508 case PR_SET_UNALIGN:
2509 error = SET_UNALIGN_CTL(me, arg2);
2510 break;
2511 case PR_GET_UNALIGN:
2512 error = GET_UNALIGN_CTL(me, arg2);
2513 break;
2514 case PR_SET_FPEMU:
2515 error = SET_FPEMU_CTL(me, arg2);
2516 break;
2517 case PR_GET_FPEMU:
2518 error = GET_FPEMU_CTL(me, arg2);
2519 break;
2520 case PR_SET_FPEXC:
2521 error = SET_FPEXC_CTL(me, arg2);
2522 break;
2523 case PR_GET_FPEXC:
2524 error = GET_FPEXC_CTL(me, arg2);
2525 break;
2526 case PR_GET_TIMING:
2527 error = PR_TIMING_STATISTICAL;
2528 break;
2529 case PR_SET_TIMING:
2530 if (arg2 != PR_TIMING_STATISTICAL)
2531 error = -EINVAL;
2532 break;
2533 case PR_SET_NAME:
2534 comm[sizeof(me->comm) - 1] = 0;
2535 if (strncpy_from_user(comm, (char __user *)arg2,
2536 sizeof(me->comm) - 1) < 0)
2537 return -EFAULT;
2538 set_task_comm(me, comm);
2539 proc_comm_connector(me);
2540 break;
2541 case PR_GET_NAME:
2542 get_task_comm(comm, me);
2543 if (copy_to_user((char __user *)arg2, comm, sizeof(comm)))
2544 return -EFAULT;
2545 break;
2546 case PR_GET_ENDIAN:
2547 error = GET_ENDIAN(me, arg2);
2548 break;
2549 case PR_SET_ENDIAN:
2550 error = SET_ENDIAN(me, arg2);
2551 break;
2552 case PR_GET_SECCOMP:
2553 error = prctl_get_seccomp();
2554 break;
2555 case PR_SET_SECCOMP:
2556 error = prctl_set_seccomp(arg2, (char __user *)arg3);
2557 break;
2558 case PR_GET_TSC:
2559 error = GET_TSC_CTL(arg2);
2560 break;
2561 case PR_SET_TSC:
2562 error = SET_TSC_CTL(arg2);
2563 break;
2564 case PR_TASK_PERF_EVENTS_DISABLE:
2565 error = perf_event_task_disable();
2566 break;
2567 case PR_TASK_PERF_EVENTS_ENABLE:
2568 error = perf_event_task_enable();
2569 break;
2570 case PR_GET_TIMERSLACK:
2571 if (current->timer_slack_ns > ULONG_MAX)
2572 error = ULONG_MAX;
2573 else
2574 error = current->timer_slack_ns;
2575 break;
2576 case PR_SET_TIMERSLACK:
2577 if (rt_or_dl_task_policy(current))
2578 break;
2579 if (arg2 <= 0)
2580 current->timer_slack_ns =
2581 current->default_timer_slack_ns;
2582 else
2583 current->timer_slack_ns = arg2;
2584 break;
2585 case PR_MCE_KILL:
2586 if (arg4 | arg5)
2587 return -EINVAL;
2588 switch (arg2) {
2589 case PR_MCE_KILL_CLEAR:
2590 if (arg3 != 0)
2591 return -EINVAL;
2592 current->flags &= ~PF_MCE_PROCESS;
2593 break;
2594 case PR_MCE_KILL_SET:
2595 current->flags |= PF_MCE_PROCESS;
2596 if (arg3 == PR_MCE_KILL_EARLY)
2597 current->flags |= PF_MCE_EARLY;
2598 else if (arg3 == PR_MCE_KILL_LATE)
2599 current->flags &= ~PF_MCE_EARLY;
2600 else if (arg3 == PR_MCE_KILL_DEFAULT)
2601 current->flags &=
2602 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
2603 else
2604 return -EINVAL;
2605 break;
2606 default:
2607 return -EINVAL;
2608 }
2609 break;
2610 case PR_MCE_KILL_GET:
2611 if (arg2 | arg3 | arg4 | arg5)
2612 return -EINVAL;
2613 if (current->flags & PF_MCE_PROCESS)
2614 error = (current->flags & PF_MCE_EARLY) ?
2615 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2616 else
2617 error = PR_MCE_KILL_DEFAULT;
2618 break;
2619 case PR_SET_MM:
2620 error = prctl_set_mm(arg2, arg3, arg4, arg5);
2621 break;
2622 case PR_GET_TID_ADDRESS:
2623 error = prctl_get_tid_address(me, (int __user * __user *)arg2);
2624 break;
2625 case PR_SET_CHILD_SUBREAPER:
2626 me->signal->is_child_subreaper = !!arg2;
2627 if (!arg2)
2628 break;
2629
2630 walk_process_tree(me, propagate_has_child_subreaper, NULL);
2631 break;
2632 case PR_GET_CHILD_SUBREAPER:
2633 error = put_user(me->signal->is_child_subreaper,
2634 (int __user *)arg2);
2635 break;
2636 case PR_SET_NO_NEW_PRIVS:
2637 if (arg2 != 1 || arg3 || arg4 || arg5)
2638 return -EINVAL;
2639
2640 task_set_no_new_privs(current);
2641 break;
2642 case PR_GET_NO_NEW_PRIVS:
2643 if (arg2 || arg3 || arg4 || arg5)
2644 return -EINVAL;
2645 return task_no_new_privs(current) ? 1 : 0;
2646 case PR_GET_THP_DISABLE:
2647 if (arg2 || arg3 || arg4 || arg5)
2648 return -EINVAL;
2649 error = !!test_bit(MMF_DISABLE_THP, &me->mm->flags);
2650 break;
2651 case PR_SET_THP_DISABLE:
2652 if (arg3 || arg4 || arg5)
2653 return -EINVAL;
2654 if (mmap_write_lock_killable(me->mm))
2655 return -EINTR;
2656 if (arg2)
2657 set_bit(MMF_DISABLE_THP, &me->mm->flags);
2658 else
2659 clear_bit(MMF_DISABLE_THP, &me->mm->flags);
2660 mmap_write_unlock(me->mm);
2661 break;
2662 case PR_MPX_ENABLE_MANAGEMENT:
2663 case PR_MPX_DISABLE_MANAGEMENT:
2664 /* No longer implemented: */
2665 return -EINVAL;
2666 case PR_SET_FP_MODE:
2667 error = SET_FP_MODE(me, arg2);
2668 break;
2669 case PR_GET_FP_MODE:
2670 error = GET_FP_MODE(me);
2671 break;
2672 case PR_SVE_SET_VL:
2673 error = SVE_SET_VL(arg2);
2674 break;
2675 case PR_SVE_GET_VL:
2676 error = SVE_GET_VL();
2677 break;
2678 case PR_SME_SET_VL:
2679 error = SME_SET_VL(arg2);
2680 break;
2681 case PR_SME_GET_VL:
2682 error = SME_GET_VL();
2683 break;
2684 case PR_GET_SPECULATION_CTRL:
2685 if (arg3 || arg4 || arg5)
2686 return -EINVAL;
2687 error = arch_prctl_spec_ctrl_get(me, arg2);
2688 break;
2689 case PR_SET_SPECULATION_CTRL:
2690 if (arg4 || arg5)
2691 return -EINVAL;
2692 error = arch_prctl_spec_ctrl_set(me, arg2, arg3);
2693 break;
2694 case PR_PAC_RESET_KEYS:
2695 if (arg3 || arg4 || arg5)
2696 return -EINVAL;
2697 error = PAC_RESET_KEYS(me, arg2);
2698 break;
2699 case PR_PAC_SET_ENABLED_KEYS:
2700 if (arg4 || arg5)
2701 return -EINVAL;
2702 error = PAC_SET_ENABLED_KEYS(me, arg2, arg3);
2703 break;
2704 case PR_PAC_GET_ENABLED_KEYS:
2705 if (arg2 || arg3 || arg4 || arg5)
2706 return -EINVAL;
2707 error = PAC_GET_ENABLED_KEYS(me);
2708 break;
2709 case PR_SET_TAGGED_ADDR_CTRL:
2710 if (arg3 || arg4 || arg5)
2711 return -EINVAL;
2712 error = SET_TAGGED_ADDR_CTRL(arg2);
2713 break;
2714 case PR_GET_TAGGED_ADDR_CTRL:
2715 if (arg2 || arg3 || arg4 || arg5)
2716 return -EINVAL;
2717 error = GET_TAGGED_ADDR_CTRL();
2718 break;
2719 case PR_SET_IO_FLUSHER:
2720 if (!capable(CAP_SYS_RESOURCE))
2721 return -EPERM;
2722
2723 if (arg3 || arg4 || arg5)
2724 return -EINVAL;
2725
2726 if (arg2 == 1)
2727 current->flags |= PR_IO_FLUSHER;
2728 else if (!arg2)
2729 current->flags &= ~PR_IO_FLUSHER;
2730 else
2731 return -EINVAL;
2732 break;
2733 case PR_GET_IO_FLUSHER:
2734 if (!capable(CAP_SYS_RESOURCE))
2735 return -EPERM;
2736
2737 if (arg2 || arg3 || arg4 || arg5)
2738 return -EINVAL;
2739
2740 error = (current->flags & PR_IO_FLUSHER) == PR_IO_FLUSHER;
2741 break;
2742 case PR_SET_SYSCALL_USER_DISPATCH:
2743 error = set_syscall_user_dispatch(arg2, arg3, arg4,
2744 (char __user *) arg5);
2745 break;
2746#ifdef CONFIG_SCHED_CORE
2747 case PR_SCHED_CORE:
2748 error = sched_core_share_pid(arg2, arg3, arg4, arg5);
2749 break;
2750#endif
2751 case PR_SET_MDWE:
2752 error = prctl_set_mdwe(arg2, arg3, arg4, arg5);
2753 break;
2754 case PR_GET_MDWE:
2755 error = prctl_get_mdwe(arg2, arg3, arg4, arg5);
2756 break;
2757 case PR_PPC_GET_DEXCR:
2758 if (arg3 || arg4 || arg5)
2759 return -EINVAL;
2760 error = PPC_GET_DEXCR_ASPECT(me, arg2);
2761 break;
2762 case PR_PPC_SET_DEXCR:
2763 if (arg4 || arg5)
2764 return -EINVAL;
2765 error = PPC_SET_DEXCR_ASPECT(me, arg2, arg3);
2766 break;
2767 case PR_SET_VMA:
2768 error = prctl_set_vma(arg2, arg3, arg4, arg5);
2769 break;
2770 case PR_GET_AUXV:
2771 if (arg4 || arg5)
2772 return -EINVAL;
2773 error = prctl_get_auxv((void __user *)arg2, arg3);
2774 break;
2775#ifdef CONFIG_KSM
2776 case PR_SET_MEMORY_MERGE:
2777 if (arg3 || arg4 || arg5)
2778 return -EINVAL;
2779 if (mmap_write_lock_killable(me->mm))
2780 return -EINTR;
2781
2782 if (arg2)
2783 error = ksm_enable_merge_any(me->mm);
2784 else
2785 error = ksm_disable_merge_any(me->mm);
2786 mmap_write_unlock(me->mm);
2787 break;
2788 case PR_GET_MEMORY_MERGE:
2789 if (arg2 || arg3 || arg4 || arg5)
2790 return -EINVAL;
2791
2792 error = !!test_bit(MMF_VM_MERGE_ANY, &me->mm->flags);
2793 break;
2794#endif
2795 case PR_RISCV_V_SET_CONTROL:
2796 error = RISCV_V_SET_CONTROL(arg2);
2797 break;
2798 case PR_RISCV_V_GET_CONTROL:
2799 error = RISCV_V_GET_CONTROL();
2800 break;
2801 case PR_RISCV_SET_ICACHE_FLUSH_CTX:
2802 error = RISCV_SET_ICACHE_FLUSH_CTX(arg2, arg3);
2803 break;
2804 case PR_GET_SHADOW_STACK_STATUS:
2805 if (arg3 || arg4 || arg5)
2806 return -EINVAL;
2807 error = arch_get_shadow_stack_status(me, (unsigned long __user *) arg2);
2808 break;
2809 case PR_SET_SHADOW_STACK_STATUS:
2810 if (arg3 || arg4 || arg5)
2811 return -EINVAL;
2812 error = arch_set_shadow_stack_status(me, arg2);
2813 break;
2814 case PR_LOCK_SHADOW_STACK_STATUS:
2815 if (arg3 || arg4 || arg5)
2816 return -EINVAL;
2817 error = arch_lock_shadow_stack_status(me, arg2);
2818 break;
2819 case PR_TIMER_CREATE_RESTORE_IDS:
2820 if (arg3 || arg4 || arg5)
2821 return -EINVAL;
2822 error = posixtimer_create_prctl(arg2);
2823 break;
2824 case PR_FUTEX_HASH:
2825 error = futex_hash_prctl(arg2, arg3, arg4);
2826 break;
2827 default:
2828 trace_task_prctl_unknown(option, arg2, arg3, arg4, arg5);
2829 error = -EINVAL;
2830 break;
2831 }
2832 return error;
2833}
2834
2835SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2836 struct getcpu_cache __user *, unused)
2837{
2838 int err = 0;
2839 int cpu = raw_smp_processor_id();
2840
2841 if (cpup)
2842 err |= put_user(cpu, cpup);
2843 if (nodep)
2844 err |= put_user(cpu_to_node(cpu), nodep);
2845 return err ? -EFAULT : 0;
2846}
2847
2848/**
2849 * do_sysinfo - fill in sysinfo struct
2850 * @info: pointer to buffer to fill
2851 */
2852static int do_sysinfo(struct sysinfo *info)
2853{
2854 unsigned long mem_total, sav_total;
2855 unsigned int mem_unit, bitcount;
2856 struct timespec64 tp;
2857
2858 memset(info, 0, sizeof(struct sysinfo));
2859
2860 ktime_get_boottime_ts64(&tp);
2861 timens_add_boottime(&tp);
2862 info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
2863
2864 get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
2865
2866 info->procs = nr_threads;
2867
2868 si_meminfo(info);
2869 si_swapinfo(info);
2870
2871 /*
2872 * If the sum of all the available memory (i.e. ram + swap)
2873 * is less than can be stored in a 32 bit unsigned long then
2874 * we can be binary compatible with 2.2.x kernels. If not,
2875 * well, in that case 2.2.x was broken anyways...
2876 *
2877 * -Erik Andersen <andersee@debian.org>
2878 */
2879
2880 mem_total = info->totalram + info->totalswap;
2881 if (mem_total < info->totalram || mem_total < info->totalswap)
2882 goto out;
2883 bitcount = 0;
2884 mem_unit = info->mem_unit;
2885 while (mem_unit > 1) {
2886 bitcount++;
2887 mem_unit >>= 1;
2888 sav_total = mem_total;
2889 mem_total <<= 1;
2890 if (mem_total < sav_total)
2891 goto out;
2892 }
2893
2894 /*
2895 * If mem_total did not overflow, multiply all memory values by
2896 * info->mem_unit and set it to 1. This leaves things compatible
2897 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2898 * kernels...
2899 */
2900
2901 info->mem_unit = 1;
2902 info->totalram <<= bitcount;
2903 info->freeram <<= bitcount;
2904 info->sharedram <<= bitcount;
2905 info->bufferram <<= bitcount;
2906 info->totalswap <<= bitcount;
2907 info->freeswap <<= bitcount;
2908 info->totalhigh <<= bitcount;
2909 info->freehigh <<= bitcount;
2910
2911out:
2912 return 0;
2913}
2914
2915SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
2916{
2917 struct sysinfo val;
2918
2919 do_sysinfo(&val);
2920
2921 if (copy_to_user(info, &val, sizeof(struct sysinfo)))
2922 return -EFAULT;
2923
2924 return 0;
2925}
2926
2927#ifdef CONFIG_COMPAT
2928struct compat_sysinfo {
2929 s32 uptime;
2930 u32 loads[3];
2931 u32 totalram;
2932 u32 freeram;
2933 u32 sharedram;
2934 u32 bufferram;
2935 u32 totalswap;
2936 u32 freeswap;
2937 u16 procs;
2938 u16 pad;
2939 u32 totalhigh;
2940 u32 freehigh;
2941 u32 mem_unit;
2942 char _f[20-2*sizeof(u32)-sizeof(int)];
2943};
2944
2945COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2946{
2947 struct sysinfo s;
2948 struct compat_sysinfo s_32;
2949
2950 do_sysinfo(&s);
2951
2952 /* Check to see if any memory value is too large for 32-bit and scale
2953 * down if needed
2954 */
2955 if (upper_32_bits(s.totalram) || upper_32_bits(s.totalswap)) {
2956 int bitcount = 0;
2957
2958 while (s.mem_unit < PAGE_SIZE) {
2959 s.mem_unit <<= 1;
2960 bitcount++;
2961 }
2962
2963 s.totalram >>= bitcount;
2964 s.freeram >>= bitcount;
2965 s.sharedram >>= bitcount;
2966 s.bufferram >>= bitcount;
2967 s.totalswap >>= bitcount;
2968 s.freeswap >>= bitcount;
2969 s.totalhigh >>= bitcount;
2970 s.freehigh >>= bitcount;
2971 }
2972
2973 memset(&s_32, 0, sizeof(s_32));
2974 s_32.uptime = s.uptime;
2975 s_32.loads[0] = s.loads[0];
2976 s_32.loads[1] = s.loads[1];
2977 s_32.loads[2] = s.loads[2];
2978 s_32.totalram = s.totalram;
2979 s_32.freeram = s.freeram;
2980 s_32.sharedram = s.sharedram;
2981 s_32.bufferram = s.bufferram;
2982 s_32.totalswap = s.totalswap;
2983 s_32.freeswap = s.freeswap;
2984 s_32.procs = s.procs;
2985 s_32.totalhigh = s.totalhigh;
2986 s_32.freehigh = s.freehigh;
2987 s_32.mem_unit = s.mem_unit;
2988 if (copy_to_user(info, &s_32, sizeof(s_32)))
2989 return -EFAULT;
2990 return 0;
2991}
2992#endif /* CONFIG_COMPAT */
Linux 6.x block layer and io_uring tree(s)