projects / GitHub / MotorolaMobilityLLC / kernel-slsi.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
[COMMON]fimc-is2: modified OV12A10/OV12A10FF/OV16885/5E9 power on sequence delay
[GitHub/MotorolaMobilityLLC/kernel-slsi.git] / kernel / sys.c
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/utsname.h>
11 #include <linux/mman.h>
12 #include <linux/reboot.h>
13 #include <linux/prctl.h>
14 #include <linux/highuid.h>
15 #include <linux/fs.h>
16 #include <linux/kmod.h>
17 #include <linux/perf_event.h>
18 #include <linux/resource.h>
19 #include <linux/kernel.h>
20 #include <linux/workqueue.h>
21 #include <linux/capability.h>
22 #include <linux/device.h>
23 #include <linux/key.h>
24 #include <linux/times.h>
25 #include <linux/posix-timers.h>
26 #include <linux/security.h>
27 #include <linux/dcookies.h>
28 #include <linux/suspend.h>
29 #include <linux/tty.h>
30 #include <linux/signal.h>
31 #include <linux/cn_proc.h>
32 #include <linux/getcpu.h>
33 #include <linux/task_io_accounting_ops.h>
34 #include <linux/seccomp.h>
35 #include <linux/cpu.h>
36 #include <linux/personality.h>
37 #include <linux/ptrace.h>
38 #include <linux/fs_struct.h>
39 #include <linux/file.h>
40 #include <linux/mount.h>
41 #include <linux/gfp.h>
42 #include <linux/syscore_ops.h>
43 #include <linux/version.h>
44 #include <linux/ctype.h>
45 #include <linux/mm.h>
46 #include <linux/mempolicy.h>
47
48 #include <linux/compat.h>
49 #include <linux/syscalls.h>
50 #include <linux/kprobes.h>
51 #include <linux/user_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 #ifndef SET_UNALIGN_CTL
77 # define SET_UNALIGN_CTL(a, b) (-EINVAL)
78 #endif
79 #ifndef GET_UNALIGN_CTL
80 # define GET_UNALIGN_CTL(a, b) (-EINVAL)
81 #endif
82 #ifndef SET_FPEMU_CTL
83 # define SET_FPEMU_CTL(a, b) (-EINVAL)
84 #endif
85 #ifndef GET_FPEMU_CTL
86 # define GET_FPEMU_CTL(a, b) (-EINVAL)
87 #endif
88 #ifndef SET_FPEXC_CTL
89 # define SET_FPEXC_CTL(a, b) (-EINVAL)
90 #endif
91 #ifndef GET_FPEXC_CTL
92 # define GET_FPEXC_CTL(a, b) (-EINVAL)
93 #endif
94 #ifndef GET_ENDIAN
95 # define GET_ENDIAN(a, b) (-EINVAL)
96 #endif
97 #ifndef SET_ENDIAN
98 # define SET_ENDIAN(a, b) (-EINVAL)
99 #endif
100 #ifndef GET_TSC_CTL
101 # define GET_TSC_CTL(a) (-EINVAL)
102 #endif
103 #ifndef SET_TSC_CTL
104 # define SET_TSC_CTL(a) (-EINVAL)
105 #endif
106 #ifndef MPX_ENABLE_MANAGEMENT
107 # define MPX_ENABLE_MANAGEMENT() (-EINVAL)
108 #endif
109 #ifndef MPX_DISABLE_MANAGEMENT
110 # define MPX_DISABLE_MANAGEMENT() (-EINVAL)
111 #endif
112 #ifndef GET_FP_MODE
113 # define GET_FP_MODE(a) (-EINVAL)
114 #endif
115 #ifndef SET_FP_MODE
116 # define SET_FP_MODE(a,b) (-EINVAL)
117 #endif
118
119 /*
120 * this is where the system-wide overflow UID and GID are defined, for
121 * architectures that now have 32-bit UID/GID but didn't in the past
122 */
123
124 int overflowuid = DEFAULT_OVERFLOWUID;
125 int overflowgid = DEFAULT_OVERFLOWGID;
126
127 EXPORT_SYMBOL(overflowuid);
128 EXPORT_SYMBOL(overflowgid);
129
130 /*
131 * the same as above, but for filesystems which can only store a 16-bit
132 * UID and GID. as such, this is needed on all architectures
133 */
134
135 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
136 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
137
138 EXPORT_SYMBOL(fs_overflowuid);
139 EXPORT_SYMBOL(fs_overflowgid);
140
141 /*
142 * Returns true if current's euid is same as p's uid or euid,
143 * or has CAP_SYS_NICE to p's user_ns.
144 *
145 * Called with rcu_read_lock, creds are safe
146 */
147 static bool set_one_prio_perm(struct task_struct *p)
148 {
149 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
150
151 if (uid_eq(pcred->uid, cred->euid) ||
152 uid_eq(pcred->euid, cred->euid))
153 return true;
154 if (ns_capable(pcred->user_ns, CAP_SYS_NICE))
155 return true;
156 return false;
157 }
158
159 /*
160 * set the priority of a task
161 * - the caller must hold the RCU read lock
162 */
163 static int set_one_prio(struct task_struct *p, int niceval, int error)
164 {
165 int no_nice;
166
167 if (!set_one_prio_perm(p)) {
168 error = -EPERM;
169 goto out;
170 }
171 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
172 error = -EACCES;
173 goto out;
174 }
175 no_nice = security_task_setnice(p, niceval);
176 if (no_nice) {
177 error = no_nice;
178 goto out;
179 }
180 if (error == -ESRCH)
181 error = 0;
182 set_user_nice(p, niceval);
183 out:
184 return error;
185 }
186
187 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
188 {
189 struct task_struct *g, *p;
190 struct user_struct *user;
191 const struct cred *cred = current_cred();
192 int error = -EINVAL;
193 struct pid *pgrp;
194 kuid_t uid;
195
196 if (which > PRIO_USER || which < PRIO_PROCESS)
197 goto out;
198
199 /* normalize: avoid signed division (rounding problems) */
200 error = -ESRCH;
201 if (niceval < MIN_NICE)
202 niceval = MIN_NICE;
203 if (niceval > MAX_NICE)
204 niceval = MAX_NICE;
205
206 rcu_read_lock();
207 read_lock(&tasklist_lock);
208 switch (which) {
209 case PRIO_PROCESS:
210 if (who)
211 p = find_task_by_vpid(who);
212 else
213 p = current;
214 if (p)
215 error = set_one_prio(p, niceval, error);
216 break;
217 case PRIO_PGRP:
218 if (who)
219 pgrp = find_vpid(who);
220 else
221 pgrp = task_pgrp(current);
222 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
223 error = set_one_prio(p, niceval, error);
224 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
225 break;
226 case PRIO_USER:
227 uid = make_kuid(cred->user_ns, who);
228 user = cred->user;
229 if (!who)
230 uid = cred->uid;
231 else if (!uid_eq(uid, cred->uid)) {
232 user = find_user(uid);
233 if (!user)
234 goto out_unlock; /* No processes for this user */
235 }
236 do_each_thread(g, p) {
237 if (uid_eq(task_uid(p), uid) && task_pid_vnr(p))
238 error = set_one_prio(p, niceval, error);
239 } while_each_thread(g, p);
240 if (!uid_eq(uid, cred->uid))
241 free_uid(user); /* For find_user() */
242 break;
243 }
244 out_unlock:
245 read_unlock(&tasklist_lock);
246 rcu_read_unlock();
247 out:
248 return error;
249 }
250
251 /*
252 * Ugh. To avoid negative return values, "getpriority()" will
253 * not return the normal nice-value, but a negated value that
254 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
255 * to stay compatible.
256 */
257 SYSCALL_DEFINE2(getpriority, int, which, int, who)
258 {
259 struct task_struct *g, *p;
260 struct user_struct *user;
261 const struct cred *cred = current_cred();
262 long niceval, retval = -ESRCH;
263 struct pid *pgrp;
264 kuid_t uid;
265
266 if (which > PRIO_USER || which < PRIO_PROCESS)
267 return -EINVAL;
268
269 rcu_read_lock();
270 read_lock(&tasklist_lock);
271 switch (which) {
272 case PRIO_PROCESS:
273 if (who)
274 p = find_task_by_vpid(who);
275 else
276 p = current;
277 if (p) {
278 niceval = nice_to_rlimit(task_nice(p));
279 if (niceval > retval)
280 retval = niceval;
281 }
282 break;
283 case PRIO_PGRP:
284 if (who)
285 pgrp = find_vpid(who);
286 else
287 pgrp = task_pgrp(current);
288 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
289 niceval = nice_to_rlimit(task_nice(p));
290 if (niceval > retval)
291 retval = niceval;
292 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
293 break;
294 case PRIO_USER:
295 uid = make_kuid(cred->user_ns, who);
296 user = cred->user;
297 if (!who)
298 uid = cred->uid;
299 else if (!uid_eq(uid, cred->uid)) {
300 user = find_user(uid);
301 if (!user)
302 goto out_unlock; /* No processes for this user */
303 }
304 do_each_thread(g, p) {
305 if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) {
306 niceval = nice_to_rlimit(task_nice(p));
307 if (niceval > retval)
308 retval = niceval;
309 }
310 } while_each_thread(g, p);
311 if (!uid_eq(uid, cred->uid))
312 free_uid(user); /* for find_user() */
313 break;
314 }
315 out_unlock:
316 read_unlock(&tasklist_lock);
317 rcu_read_unlock();
318
319 return retval;
320 }
321
322 /*
323 * Unprivileged users may change the real gid to the effective gid
324 * or vice versa. (BSD-style)
325 *
326 * If you set the real gid at all, or set the effective gid to a value not
327 * equal to the real gid, then the saved gid is set to the new effective gid.
328 *
329 * This makes it possible for a setgid program to completely drop its
330 * privileges, which is often a useful assertion to make when you are doing
331 * a security audit over a program.
332 *
333 * The general idea is that a program which uses just setregid() will be
334 * 100% compatible with BSD. A program which uses just setgid() will be
335 * 100% compatible with POSIX with saved IDs.
336 *
337 * SMP: There are not races, the GIDs are checked only by filesystem
338 * operations (as far as semantic preservation is concerned).
339 */
340 #ifdef CONFIG_MULTIUSER
341 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
342 {
343 struct user_namespace *ns = current_user_ns();
344 const struct cred *old;
345 struct cred *new;
346 int retval;
347 kgid_t krgid, kegid;
348
349 krgid = make_kgid(ns, rgid);
350 kegid = make_kgid(ns, egid);
351
352 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
353 return -EINVAL;
354 if ((egid != (gid_t) -1) && !gid_valid(kegid))
355 return -EINVAL;
356
357 new = prepare_creds();
358 if (!new)
359 return -ENOMEM;
360 old = current_cred();
361
362 retval = -EPERM;
363 if (rgid != (gid_t) -1) {
364 if (gid_eq(old->gid, krgid) ||
365 gid_eq(old->egid, krgid) ||
366 ns_capable(old->user_ns, CAP_SETGID))
367 new->gid = krgid;
368 else
369 goto error;
370 }
371 if (egid != (gid_t) -1) {
372 if (gid_eq(old->gid, kegid) ||
373 gid_eq(old->egid, kegid) ||
374 gid_eq(old->sgid, kegid) ||
375 ns_capable(old->user_ns, CAP_SETGID))
376 new->egid = kegid;
377 else
378 goto error;
379 }
380
381 if (rgid != (gid_t) -1 ||
382 (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
383 new->sgid = new->egid;
384 new->fsgid = new->egid;
385
386 return commit_creds(new);
387
388 error:
389 abort_creds(new);
390 return retval;
391 }
392
393 /*
394 * setgid() is implemented like SysV w/ SAVED_IDS
395 *
396 * SMP: Same implicit races as above.
397 */
398 SYSCALL_DEFINE1(setgid, gid_t, gid)
399 {
400 struct user_namespace *ns = current_user_ns();
401 const struct cred *old;
402 struct cred *new;
403 int retval;
404 kgid_t kgid;
405
406 kgid = make_kgid(ns, gid);
407 if (!gid_valid(kgid))
408 return -EINVAL;
409
410 new = prepare_creds();
411 if (!new)
412 return -ENOMEM;
413 old = current_cred();
414
415 retval = -EPERM;
416 if (ns_capable(old->user_ns, CAP_SETGID))
417 new->gid = new->egid = new->sgid = new->fsgid = kgid;
418 else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
419 new->egid = new->fsgid = kgid;
420 else
421 goto error;
422
423 return commit_creds(new);
424
425 error:
426 abort_creds(new);
427 return retval;
428 }
429
430 /*
431 * change the user struct in a credentials set to match the new UID
432 */
433 static int set_user(struct cred *new)
434 {
435 struct user_struct *new_user;
436
437 new_user = alloc_uid(new->uid);
438 if (!new_user)
439 return -EAGAIN;
440
441 /*
442 * We don't fail in case of NPROC limit excess here because too many
443 * poorly written programs don't check set*uid() return code, assuming
444 * it never fails if called by root. We may still enforce NPROC limit
445 * for programs doing set*uid()+execve() by harmlessly deferring the
446 * failure to the execve() stage.
447 */
448 if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
449 new_user != INIT_USER)
450 current->flags |= PF_NPROC_EXCEEDED;
451 else
452 current->flags &= ~PF_NPROC_EXCEEDED;
453
454 free_uid(new->user);
455 new->user = new_user;
456 return 0;
457 }
458
459 /*
460 * Unprivileged users may change the real uid to the effective uid
461 * or vice versa. (BSD-style)
462 *
463 * If you set the real uid at all, or set the effective uid to a value not
464 * equal to the real uid, then the saved uid is set to the new effective uid.
465 *
466 * This makes it possible for a setuid program to completely drop its
467 * privileges, which is often a useful assertion to make when you are doing
468 * a security audit over a program.
469 *
470 * The general idea is that a program which uses just setreuid() will be
471 * 100% compatible with BSD. A program which uses just setuid() will be
472 * 100% compatible with POSIX with saved IDs.
473 */
474 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
475 {
476 struct user_namespace *ns = current_user_ns();
477 const struct cred *old;
478 struct cred *new;
479 int retval;
480 kuid_t kruid, keuid;
481
482 kruid = make_kuid(ns, ruid);
483 keuid = make_kuid(ns, euid);
484
485 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
486 return -EINVAL;
487 if ((euid != (uid_t) -1) && !uid_valid(keuid))
488 return -EINVAL;
489
490 new = prepare_creds();
491 if (!new)
492 return -ENOMEM;
493 old = current_cred();
494
495 retval = -EPERM;
496 if (ruid != (uid_t) -1) {
497 new->uid = kruid;
498 if (!uid_eq(old->uid, kruid) &&
499 !uid_eq(old->euid, kruid) &&
500 !ns_capable(old->user_ns, CAP_SETUID))
501 goto error;
502 }
503
504 if (euid != (uid_t) -1) {
505 new->euid = keuid;
506 if (!uid_eq(old->uid, keuid) &&
507 !uid_eq(old->euid, keuid) &&
508 !uid_eq(old->suid, keuid) &&
509 !ns_capable(old->user_ns, CAP_SETUID))
510 goto error;
511 }
512
513 if (!uid_eq(new->uid, old->uid)) {
514 retval = set_user(new);
515 if (retval < 0)
516 goto error;
517 }
518 if (ruid != (uid_t) -1 ||
519 (euid != (uid_t) -1 && !uid_eq(keuid, old->uid)))
520 new->suid = new->euid;
521 new->fsuid = new->euid;
522
523 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
524 if (retval < 0)
525 goto error;
526
527 return commit_creds(new);
528
529 error:
530 abort_creds(new);
531 return retval;
532 }
533
534 /*
535 * setuid() is implemented like SysV with SAVED_IDS
536 *
537 * Note that SAVED_ID's is deficient in that a setuid root program
538 * like sendmail, for example, cannot set its uid to be a normal
539 * user and then switch back, because if you're root, setuid() sets
540 * the saved uid too. If you don't like this, blame the bright people
541 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
542 * will allow a root program to temporarily drop privileges and be able to
543 * regain them by swapping the real and effective uid.
544 */
545 SYSCALL_DEFINE1(setuid, uid_t, uid)
546 {
547 struct user_namespace *ns = current_user_ns();
548 const struct cred *old;
549 struct cred *new;
550 int retval;
551 kuid_t kuid;
552
553 kuid = make_kuid(ns, uid);
554 if (!uid_valid(kuid))
555 return -EINVAL;
556
557 new = prepare_creds();
558 if (!new)
559 return -ENOMEM;
560 old = current_cred();
561
562 retval = -EPERM;
563 if (ns_capable(old->user_ns, CAP_SETUID)) {
564 new->suid = new->uid = kuid;
565 if (!uid_eq(kuid, old->uid)) {
566 retval = set_user(new);
567 if (retval < 0)
568 goto error;
569 }
570 } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) {
571 goto error;
572 }
573
574 new->fsuid = new->euid = kuid;
575
576 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
577 if (retval < 0)
578 goto error;
579
580 return commit_creds(new);
581
582 error:
583 abort_creds(new);
584 return retval;
585 }
586
587
588 /*
589 * This function implements a generic ability to update ruid, euid,
590 * and suid. This allows you to implement the 4.4 compatible seteuid().
591 */
592 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
593 {
594 struct user_namespace *ns = current_user_ns();
595 const struct cred *old;
596 struct cred *new;
597 int retval;
598 kuid_t kruid, keuid, ksuid;
599
600 kruid = make_kuid(ns, ruid);
601 keuid = make_kuid(ns, euid);
602 ksuid = make_kuid(ns, suid);
603
604 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
605 return -EINVAL;
606
607 if ((euid != (uid_t) -1) && !uid_valid(keuid))
608 return -EINVAL;
609
610 if ((suid != (uid_t) -1) && !uid_valid(ksuid))
611 return -EINVAL;
612
613 new = prepare_creds();
614 if (!new)
615 return -ENOMEM;
616
617 old = current_cred();
618
619 retval = -EPERM;
620 if (!ns_capable(old->user_ns, CAP_SETUID)) {
621 if (ruid != (uid_t) -1 && !uid_eq(kruid, old->uid) &&
622 !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid))
623 goto error;
624 if (euid != (uid_t) -1 && !uid_eq(keuid, old->uid) &&
625 !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid))
626 goto error;
627 if (suid != (uid_t) -1 && !uid_eq(ksuid, old->uid) &&
628 !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid))
629 goto error;
630 }
631
632 if (ruid != (uid_t) -1) {
633 new->uid = kruid;
634 if (!uid_eq(kruid, old->uid)) {
635 retval = set_user(new);
636 if (retval < 0)
637 goto error;
638 }
639 }
640 if (euid != (uid_t) -1)
641 new->euid = keuid;
642 if (suid != (uid_t) -1)
643 new->suid = ksuid;
644 new->fsuid = new->euid;
645
646 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
647 if (retval < 0)
648 goto error;
649
650 return commit_creds(new);
651
652 error:
653 abort_creds(new);
654 return retval;
655 }
656
657 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
658 {
659 const struct cred *cred = current_cred();
660 int retval;
661 uid_t ruid, euid, suid;
662
663 ruid = from_kuid_munged(cred->user_ns, cred->uid);
664 euid = from_kuid_munged(cred->user_ns, cred->euid);
665 suid = from_kuid_munged(cred->user_ns, cred->suid);
666
667 retval = put_user(ruid, ruidp);
668 if (!retval) {
669 retval = put_user(euid, euidp);
670 if (!retval)
671 return put_user(suid, suidp);
672 }
673 return retval;
674 }
675
676 /*
677 * Same as above, but for rgid, egid, sgid.
678 */
679 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
680 {
681 struct user_namespace *ns = current_user_ns();
682 const struct cred *old;
683 struct cred *new;
684 int retval;
685 kgid_t krgid, kegid, ksgid;
686
687 krgid = make_kgid(ns, rgid);
688 kegid = make_kgid(ns, egid);
689 ksgid = make_kgid(ns, sgid);
690
691 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
692 return -EINVAL;
693 if ((egid != (gid_t) -1) && !gid_valid(kegid))
694 return -EINVAL;
695 if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
696 return -EINVAL;
697
698 new = prepare_creds();
699 if (!new)
700 return -ENOMEM;
701 old = current_cred();
702
703 retval = -EPERM;
704 if (!ns_capable(old->user_ns, CAP_SETGID)) {
705 if (rgid != (gid_t) -1 && !gid_eq(krgid, old->gid) &&
706 !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid))
707 goto error;
708 if (egid != (gid_t) -1 && !gid_eq(kegid, old->gid) &&
709 !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid))
710 goto error;
711 if (sgid != (gid_t) -1 && !gid_eq(ksgid, old->gid) &&
712 !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid))
713 goto error;
714 }
715
716 if (rgid != (gid_t) -1)
717 new->gid = krgid;
718 if (egid != (gid_t) -1)
719 new->egid = kegid;
720 if (sgid != (gid_t) -1)
721 new->sgid = ksgid;
722 new->fsgid = new->egid;
723
724 return commit_creds(new);
725
726 error:
727 abort_creds(new);
728 return retval;
729 }
730
731 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
732 {
733 const struct cred *cred = current_cred();
734 int retval;
735 gid_t rgid, egid, sgid;
736
737 rgid = from_kgid_munged(cred->user_ns, cred->gid);
738 egid = from_kgid_munged(cred->user_ns, cred->egid);
739 sgid = from_kgid_munged(cred->user_ns, cred->sgid);
740
741 retval = put_user(rgid, rgidp);
742 if (!retval) {
743 retval = put_user(egid, egidp);
744 if (!retval)
745 retval = put_user(sgid, sgidp);
746 }
747
748 return retval;
749 }
750
751
752 /*
753 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
754 * is used for "access()" and for the NFS daemon (letting nfsd stay at
755 * whatever uid it wants to). It normally shadows "euid", except when
756 * explicitly set by setfsuid() or for access..
757 */
758 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
759 {
760 const struct cred *old;
761 struct cred *new;
762 uid_t old_fsuid;
763 kuid_t kuid;
764
765 old = current_cred();
766 old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
767
768 kuid = make_kuid(old->user_ns, uid);
769 if (!uid_valid(kuid))
770 return old_fsuid;
771
772 new = prepare_creds();
773 if (!new)
774 return old_fsuid;
775
776 if (uid_eq(kuid, old->uid) || uid_eq(kuid, old->euid) ||
777 uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
778 ns_capable(old->user_ns, CAP_SETUID)) {
779 if (!uid_eq(kuid, old->fsuid)) {
780 new->fsuid = kuid;
781 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
782 goto change_okay;
783 }
784 }
785
786 abort_creds(new);
787 return old_fsuid;
788
789 change_okay:
790 commit_creds(new);
791 return old_fsuid;
792 }
793
794 /*
795 * Samma på svenska..
796 */
797 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
798 {
799 const struct cred *old;
800 struct cred *new;
801 gid_t old_fsgid;
802 kgid_t kgid;
803
804 old = current_cred();
805 old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
806
807 kgid = make_kgid(old->user_ns, gid);
808 if (!gid_valid(kgid))
809 return old_fsgid;
810
811 new = prepare_creds();
812 if (!new)
813 return old_fsgid;
814
815 if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->egid) ||
816 gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
817 ns_capable(old->user_ns, CAP_SETGID)) {
818 if (!gid_eq(kgid, old->fsgid)) {
819 new->fsgid = kgid;
820 goto change_okay;
821 }
822 }
823
824 abort_creds(new);
825 return old_fsgid;
826
827 change_okay:
828 commit_creds(new);
829 return old_fsgid;
830 }
831 #endif /* CONFIG_MULTIUSER */
832
833 /**
834 * sys_getpid - return the thread group id of the current process
835 *
836 * Note, despite the name, this returns the tgid not the pid. The tgid and
837 * the pid are identical unless CLONE_THREAD was specified on clone() in
838 * which case the tgid is the same in all threads of the same group.
839 *
840 * This is SMP safe as current->tgid does not change.
841 */
842 SYSCALL_DEFINE0(getpid)
843 {
844 return task_tgid_vnr(current);
845 }
846
847 /* Thread ID - the internal kernel "pid" */
848 SYSCALL_DEFINE0(gettid)
849 {
850 return task_pid_vnr(current);
851 }
852
853 /*
854 * Accessing ->real_parent is not SMP-safe, it could
855 * change from under us. However, we can use a stale
856 * value of ->real_parent under rcu_read_lock(), see
857 * release_task()->call_rcu(delayed_put_task_struct).
858 */
859 SYSCALL_DEFINE0(getppid)
860 {
861 int pid;
862
863 rcu_read_lock();
864 pid = task_tgid_vnr(rcu_dereference(current->real_parent));
865 rcu_read_unlock();
866
867 return pid;
868 }
869
870 SYSCALL_DEFINE0(getuid)
871 {
872 /* Only we change this so SMP safe */
873 return from_kuid_munged(current_user_ns(), current_uid());
874 }
875
876 SYSCALL_DEFINE0(geteuid)
877 {
878 /* Only we change this so SMP safe */
879 return from_kuid_munged(current_user_ns(), current_euid());
880 }
881
882 SYSCALL_DEFINE0(getgid)
883 {
884 /* Only we change this so SMP safe */
885 return from_kgid_munged(current_user_ns(), current_gid());
886 }
887
888 SYSCALL_DEFINE0(getegid)
889 {
890 /* Only we change this so SMP safe */
891 return from_kgid_munged(current_user_ns(), current_egid());
892 }
893
894 static void do_sys_times(struct tms *tms)
895 {
896 u64 tgutime, tgstime, cutime, cstime;
897
898 thread_group_cputime_adjusted(current, &tgutime, &tgstime);
899 cutime = current->signal->cutime;
900 cstime = current->signal->cstime;
901 tms->tms_utime = nsec_to_clock_t(tgutime);
902 tms->tms_stime = nsec_to_clock_t(tgstime);
903 tms->tms_cutime = nsec_to_clock_t(cutime);
904 tms->tms_cstime = nsec_to_clock_t(cstime);
905 }
906
907 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
908 {
909 if (tbuf) {
910 struct tms tmp;
911
912 do_sys_times(&tmp);
913 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
914 return -EFAULT;
915 }
916 force_successful_syscall_return();
917 return (long) jiffies_64_to_clock_t(get_jiffies_64());
918 }
919
920 #ifdef CONFIG_COMPAT
921 static compat_clock_t clock_t_to_compat_clock_t(clock_t x)
922 {
923 return compat_jiffies_to_clock_t(clock_t_to_jiffies(x));
924 }
925
926 COMPAT_SYSCALL_DEFINE1(times, struct compat_tms __user *, tbuf)
927 {
928 if (tbuf) {
929 struct tms tms;
930 struct compat_tms tmp;
931
932 do_sys_times(&tms);
933 /* Convert our struct tms to the compat version. */
934 tmp.tms_utime = clock_t_to_compat_clock_t(tms.tms_utime);
935 tmp.tms_stime = clock_t_to_compat_clock_t(tms.tms_stime);
936 tmp.tms_cutime = clock_t_to_compat_clock_t(tms.tms_cutime);
937 tmp.tms_cstime = clock_t_to_compat_clock_t(tms.tms_cstime);
938 if (copy_to_user(tbuf, &tmp, sizeof(tmp)))
939 return -EFAULT;
940 }
941 force_successful_syscall_return();
942 return compat_jiffies_to_clock_t(jiffies);
943 }
944 #endif
945
946 /*
947 * This needs some heavy checking ...
948 * I just haven't the stomach for it. I also don't fully
949 * understand sessions/pgrp etc. Let somebody who does explain it.
950 *
951 * OK, I think I have the protection semantics right.... this is really
952 * only important on a multi-user system anyway, to make sure one user
953 * can't send a signal to a process owned by another. -TYT, 12/12/91
954 *
955 * !PF_FORKNOEXEC check to conform completely to POSIX.
956 */
957 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
958 {
959 struct task_struct *p;
960 struct task_struct *group_leader = current->group_leader;
961 struct pid *pgrp;
962 int err;
963
964 if (!pid)
965 pid = task_pid_vnr(group_leader);
966 if (!pgid)
967 pgid = pid;
968 if (pgid < 0)
969 return -EINVAL;
970 rcu_read_lock();
971
972 /* From this point forward we keep holding onto the tasklist lock
973 * so that our parent does not change from under us. -DaveM
974 */
975 write_lock_irq(&tasklist_lock);
976
977 err = -ESRCH;
978 p = find_task_by_vpid(pid);
979 if (!p)
980 goto out;
981
982 err = -EINVAL;
983 if (!thread_group_leader(p))
984 goto out;
985
986 if (same_thread_group(p->real_parent, group_leader)) {
987 err = -EPERM;
988 if (task_session(p) != task_session(group_leader))
989 goto out;
990 err = -EACCES;
991 if (!(p->flags & PF_FORKNOEXEC))
992 goto out;
993 } else {
994 err = -ESRCH;
995 if (p != group_leader)
996 goto out;
997 }
998
999 err = -EPERM;
1000 if (p->signal->leader)
1001 goto out;
1002
1003 pgrp = task_pid(p);
1004 if (pgid != pid) {
1005 struct task_struct *g;
1006
1007 pgrp = find_vpid(pgid);
1008 g = pid_task(pgrp, PIDTYPE_PGID);
1009 if (!g || task_session(g) != task_session(group_leader))
1010 goto out;
1011 }
1012
1013 err = security_task_setpgid(p, pgid);
1014 if (err)
1015 goto out;
1016
1017 if (task_pgrp(p) != pgrp)
1018 change_pid(p, PIDTYPE_PGID, pgrp);
1019
1020 err = 0;
1021 out:
1022 /* All paths lead to here, thus we are safe. -DaveM */
1023 write_unlock_irq(&tasklist_lock);
1024 rcu_read_unlock();
1025 return err;
1026 }
1027
1028 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1029 {
1030 struct task_struct *p;
1031 struct pid *grp;
1032 int retval;
1033
1034 rcu_read_lock();
1035 if (!pid)
1036 grp = task_pgrp(current);
1037 else {
1038 retval = -ESRCH;
1039 p = find_task_by_vpid(pid);
1040 if (!p)
1041 goto out;
1042 grp = task_pgrp(p);
1043 if (!grp)
1044 goto out;
1045
1046 retval = security_task_getpgid(p);
1047 if (retval)
1048 goto out;
1049 }
1050 retval = pid_vnr(grp);
1051 out:
1052 rcu_read_unlock();
1053 return retval;
1054 }
1055
1056 #ifdef __ARCH_WANT_SYS_GETPGRP
1057
1058 SYSCALL_DEFINE0(getpgrp)
1059 {
1060 return sys_getpgid(0);
1061 }
1062
1063 #endif
1064
1065 SYSCALL_DEFINE1(getsid, pid_t, pid)
1066 {
1067 struct task_struct *p;
1068 struct pid *sid;
1069 int retval;
1070
1071 rcu_read_lock();
1072 if (!pid)
1073 sid = task_session(current);
1074 else {
1075 retval = -ESRCH;
1076 p = find_task_by_vpid(pid);
1077 if (!p)
1078 goto out;
1079 sid = task_session(p);
1080 if (!sid)
1081 goto out;
1082
1083 retval = security_task_getsid(p);
1084 if (retval)
1085 goto out;
1086 }
1087 retval = pid_vnr(sid);
1088 out:
1089 rcu_read_unlock();
1090 return retval;
1091 }
1092
1093 static void set_special_pids(struct pid *pid)
1094 {
1095 struct task_struct *curr = current->group_leader;
1096
1097 if (task_session(curr) != pid)
1098 change_pid(curr, PIDTYPE_SID, pid);
1099
1100 if (task_pgrp(curr) != pid)
1101 change_pid(curr, PIDTYPE_PGID, pid);
1102 }
1103
1104 SYSCALL_DEFINE0(setsid)
1105 {
1106 struct task_struct *group_leader = current->group_leader;
1107 struct pid *sid = task_pid(group_leader);
1108 pid_t session = pid_vnr(sid);
1109 int err = -EPERM;
1110
1111 write_lock_irq(&tasklist_lock);
1112 /* Fail if I am already a session leader */
1113 if (group_leader->signal->leader)
1114 goto out;
1115
1116 /* Fail if a process group id already exists that equals the
1117 * proposed session id.
1118 */
1119 if (pid_task(sid, PIDTYPE_PGID))
1120 goto out;
1121
1122 group_leader->signal->leader = 1;
1123 set_special_pids(sid);
1124
1125 proc_clear_tty(group_leader);
1126
1127 err = session;
1128 out:
1129 write_unlock_irq(&tasklist_lock);
1130 if (err > 0) {
1131 proc_sid_connector(group_leader);
1132 sched_autogroup_create_attach(group_leader);
1133 }
1134 return err;
1135 }
1136
1137 DECLARE_RWSEM(uts_sem);
1138
1139 #ifdef COMPAT_UTS_MACHINE
1140 #define override_architecture(name) \
1141 (personality(current->personality) == PER_LINUX32 && \
1142 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1143 sizeof(COMPAT_UTS_MACHINE)))
1144 #else
1145 #define override_architecture(name) 0
1146 #endif
1147
1148 /*
1149 * Work around broken programs that cannot handle "Linux 3.0".
1150 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1151 * And we map 4.x to 2.6.60+x, so 4.0 would be 2.6.60.
1152 */
1153 static int override_release(char __user *release, size_t len)
1154 {
1155 int ret = 0;
1156
1157 if (current->personality & UNAME26) {
1158 const char *rest = UTS_RELEASE;
1159 char buf[65] = { 0 };
1160 int ndots = 0;
1161 unsigned v;
1162 size_t copy;
1163
1164 while (*rest) {
1165 if (*rest == '.' && ++ndots >= 3)
1166 break;
1167 if (!isdigit(*rest) && *rest != '.')
1168 break;
1169 rest++;
1170 }
1171 v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 60;
1172 copy = clamp_t(size_t, len, 1, sizeof(buf));
1173 copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1174 ret = copy_to_user(release, buf, copy + 1);
1175 }
1176 return ret;
1177 }
1178
1179 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1180 {
1181 int errno = 0;
1182
1183 down_read(&uts_sem);
1184 if (copy_to_user(name, utsname(), sizeof *name))
1185 errno = -EFAULT;
1186 up_read(&uts_sem);
1187
1188 if (!errno && override_release(name->release, sizeof(name->release)))
1189 errno = -EFAULT;
1190 if (!errno && override_architecture(name))
1191 errno = -EFAULT;
1192 return errno;
1193 }
1194
1195 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1196 /*
1197 * Old cruft
1198 */
1199 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1200 {
1201 int error = 0;
1202
1203 if (!name)
1204 return -EFAULT;
1205
1206 down_read(&uts_sem);
1207 if (copy_to_user(name, utsname(), sizeof(*name)))
1208 error = -EFAULT;
1209 up_read(&uts_sem);
1210
1211 if (!error && override_release(name->release, sizeof(name->release)))
1212 error = -EFAULT;
1213 if (!error && override_architecture(name))
1214 error = -EFAULT;
1215 return error;
1216 }
1217
1218 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1219 {
1220 int error;
1221
1222 if (!name)
1223 return -EFAULT;
1224 if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
1225 return -EFAULT;
1226
1227 down_read(&uts_sem);
1228 error = __copy_to_user(&name->sysname, &utsname()->sysname,
1229 __OLD_UTS_LEN);
1230 error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
1231 error |= __copy_to_user(&name->nodename, &utsname()->nodename,
1232 __OLD_UTS_LEN);
1233 error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
1234 error |= __copy_to_user(&name->release, &utsname()->release,
1235 __OLD_UTS_LEN);
1236 error |= __put_user(0, name->release + __OLD_UTS_LEN);
1237 error |= __copy_to_user(&name->version, &utsname()->version,
1238 __OLD_UTS_LEN);
1239 error |= __put_user(0, name->version + __OLD_UTS_LEN);
1240 error |= __copy_to_user(&name->machine, &utsname()->machine,
1241 __OLD_UTS_LEN);
1242 error |= __put_user(0, name->machine + __OLD_UTS_LEN);
1243 up_read(&uts_sem);
1244
1245 if (!error && override_architecture(name))
1246 error = -EFAULT;
1247 if (!error && override_release(name->release, sizeof(name->release)))
1248 error = -EFAULT;
1249 return error ? -EFAULT : 0;
1250 }
1251 #endif
1252
1253 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1254 {
1255 int errno;
1256 char tmp[__NEW_UTS_LEN];
1257
1258 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1259 return -EPERM;
1260
1261 if (len < 0 || len > __NEW_UTS_LEN)
1262 return -EINVAL;
1263 down_write(&uts_sem);
1264 errno = -EFAULT;
1265 if (!copy_from_user(tmp, name, len)) {
1266 struct new_utsname *u = utsname();
1267
1268 memcpy(u->nodename, tmp, len);
1269 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1270 errno = 0;
1271 uts_proc_notify(UTS_PROC_HOSTNAME);
1272 }
1273 up_write(&uts_sem);
1274 return errno;
1275 }
1276
1277 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1278
1279 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1280 {
1281 int i, errno;
1282 struct new_utsname *u;
1283
1284 if (len < 0)
1285 return -EINVAL;
1286 down_read(&uts_sem);
1287 u = utsname();
1288 i = 1 + strlen(u->nodename);
1289 if (i > len)
1290 i = len;
1291 errno = 0;
1292 if (copy_to_user(name, u->nodename, i))
1293 errno = -EFAULT;
1294 up_read(&uts_sem);
1295 return errno;
1296 }
1297
1298 #endif
1299
1300 /*
1301 * Only setdomainname; getdomainname can be implemented by calling
1302 * uname()
1303 */
1304 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1305 {
1306 int errno;
1307 char tmp[__NEW_UTS_LEN];
1308
1309 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1310 return -EPERM;
1311 if (len < 0 || len > __NEW_UTS_LEN)
1312 return -EINVAL;
1313
1314 down_write(&uts_sem);
1315 errno = -EFAULT;
1316 if (!copy_from_user(tmp, name, len)) {
1317 struct new_utsname *u = utsname();
1318
1319 memcpy(u->domainname, tmp, len);
1320 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1321 errno = 0;
1322 uts_proc_notify(UTS_PROC_DOMAINNAME);
1323 }
1324 up_write(&uts_sem);
1325 return errno;
1326 }
1327
1328 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1329 {
1330 struct rlimit value;
1331 int ret;
1332
1333 ret = do_prlimit(current, resource, NULL, &value);
1334 if (!ret)
1335 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1336
1337 return ret;
1338 }
1339
1340 #ifdef CONFIG_COMPAT
1341
1342 COMPAT_SYSCALL_DEFINE2(setrlimit, unsigned int, resource,
1343 struct compat_rlimit __user *, rlim)
1344 {
1345 struct rlimit r;
1346 struct compat_rlimit r32;
1347
1348 if (copy_from_user(&r32, rlim, sizeof(struct compat_rlimit)))
1349 return -EFAULT;
1350
1351 if (r32.rlim_cur == COMPAT_RLIM_INFINITY)
1352 r.rlim_cur = RLIM_INFINITY;
1353 else
1354 r.rlim_cur = r32.rlim_cur;
1355 if (r32.rlim_max == COMPAT_RLIM_INFINITY)
1356 r.rlim_max = RLIM_INFINITY;
1357 else
1358 r.rlim_max = r32.rlim_max;
1359 return do_prlimit(current, resource, &r, NULL);
1360 }
1361
1362 COMPAT_SYSCALL_DEFINE2(getrlimit, unsigned int, resource,
1363 struct compat_rlimit __user *, rlim)
1364 {
1365 struct rlimit r;
1366 int ret;
1367
1368 ret = do_prlimit(current, resource, NULL, &r);
1369 if (!ret) {
1370 struct compat_rlimit r32;
1371 if (r.rlim_cur > COMPAT_RLIM_INFINITY)
1372 r32.rlim_cur = COMPAT_RLIM_INFINITY;
1373 else
1374 r32.rlim_cur = r.rlim_cur;
1375 if (r.rlim_max > COMPAT_RLIM_INFINITY)
1376 r32.rlim_max = COMPAT_RLIM_INFINITY;
1377 else
1378 r32.rlim_max = r.rlim_max;
1379
1380 if (copy_to_user(rlim, &r32, sizeof(struct compat_rlimit)))
1381 return -EFAULT;
1382 }
1383 return ret;
1384 }
1385
1386 #endif
1387
1388 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1389
1390 /*
1391 * Back compatibility for getrlimit. Needed for some apps.
1392 */
1393 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1394 struct rlimit __user *, rlim)
1395 {
1396 struct rlimit x;
1397 if (resource >= RLIM_NLIMITS)
1398 return -EINVAL;
1399
1400 resource = array_index_nospec(resource, RLIM_NLIMITS);
1401 task_lock(current->group_leader);
1402 x = current->signal->rlim[resource];
1403 task_unlock(current->group_leader);
1404 if (x.rlim_cur > 0x7FFFFFFF)
1405 x.rlim_cur = 0x7FFFFFFF;
1406 if (x.rlim_max > 0x7FFFFFFF)
1407 x.rlim_max = 0x7FFFFFFF;
1408 return copy_to_user(rlim, &x, sizeof(x)) ? -EFAULT : 0;
1409 }
1410
1411 #ifdef CONFIG_COMPAT
1412 COMPAT_SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1413 struct compat_rlimit __user *, rlim)
1414 {
1415 struct rlimit r;
1416
1417 if (resource >= RLIM_NLIMITS)
1418 return -EINVAL;
1419
1420 resource = array_index_nospec(resource, RLIM_NLIMITS);
1421 task_lock(current->group_leader);
1422 r = current->signal->rlim[resource];
1423 task_unlock(current->group_leader);
1424 if (r.rlim_cur > 0x7FFFFFFF)
1425 r.rlim_cur = 0x7FFFFFFF;
1426 if (r.rlim_max > 0x7FFFFFFF)
1427 r.rlim_max = 0x7FFFFFFF;
1428
1429 if (put_user(r.rlim_cur, &rlim->rlim_cur) ||
1430 put_user(r.rlim_max, &rlim->rlim_max))
1431 return -EFAULT;
1432 return 0;
1433 }
1434 #endif
1435
1436 #endif
1437
1438 static inline bool rlim64_is_infinity(__u64 rlim64)
1439 {
1440 #if BITS_PER_LONG < 64
1441 return rlim64 >= ULONG_MAX;
1442 #else
1443 return rlim64 == RLIM64_INFINITY;
1444 #endif
1445 }
1446
1447 static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1448 {
1449 if (rlim->rlim_cur == RLIM_INFINITY)
1450 rlim64->rlim_cur = RLIM64_INFINITY;
1451 else
1452 rlim64->rlim_cur = rlim->rlim_cur;
1453 if (rlim->rlim_max == RLIM_INFINITY)
1454 rlim64->rlim_max = RLIM64_INFINITY;
1455 else
1456 rlim64->rlim_max = rlim->rlim_max;
1457 }
1458
1459 static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1460 {
1461 if (rlim64_is_infinity(rlim64->rlim_cur))
1462 rlim->rlim_cur = RLIM_INFINITY;
1463 else
1464 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1465 if (rlim64_is_infinity(rlim64->rlim_max))
1466 rlim->rlim_max = RLIM_INFINITY;
1467 else
1468 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1469 }
1470
1471 /* make sure you are allowed to change @tsk limits before calling this */
1472 int do_prlimit(struct task_struct *tsk, unsigned int resource,
1473 struct rlimit *new_rlim, struct rlimit *old_rlim)
1474 {
1475 struct rlimit *rlim;
1476 int retval = 0;
1477
1478 if (resource >= RLIM_NLIMITS)
1479 return -EINVAL;
1480 if (new_rlim) {
1481 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1482 return -EINVAL;
1483 if (resource == RLIMIT_NOFILE &&
1484 new_rlim->rlim_max > sysctl_nr_open)
1485 return -EPERM;
1486 }
1487
1488 /* protect tsk->signal and tsk->sighand from disappearing */
1489 read_lock(&tasklist_lock);
1490 if (!tsk->sighand) {
1491 retval = -ESRCH;
1492 goto out;
1493 }
1494
1495 rlim = tsk->signal->rlim + resource;
1496 task_lock(tsk->group_leader);
1497 if (new_rlim) {
1498 /* Keep the capable check against init_user_ns until
1499 cgroups can contain all limits */
1500 if (new_rlim->rlim_max > rlim->rlim_max &&
1501 !capable(CAP_SYS_RESOURCE))
1502 retval = -EPERM;
1503 if (!retval)
1504 retval = security_task_setrlimit(tsk, resource, new_rlim);
1505 if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) {
1506 /*
1507 * The caller is asking for an immediate RLIMIT_CPU
1508 * expiry. But we use the zero value to mean "it was
1509 * never set". So let's cheat and make it one second
1510 * instead
1511 */
1512 new_rlim->rlim_cur = 1;
1513 }
1514 }
1515 if (!retval) {
1516 if (old_rlim)
1517 *old_rlim = *rlim;
1518 if (new_rlim)
1519 *rlim = *new_rlim;
1520 }
1521 task_unlock(tsk->group_leader);
1522
1523 /*
1524 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1525 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1526 * very long-standing error, and fixing it now risks breakage of
1527 * applications, so we live with it
1528 */
1529 if (!retval && new_rlim && resource == RLIMIT_CPU &&
1530 new_rlim->rlim_cur != RLIM_INFINITY &&
1531 IS_ENABLED(CONFIG_POSIX_TIMERS))
1532 update_rlimit_cpu(tsk, new_rlim->rlim_cur);
1533 out:
1534 read_unlock(&tasklist_lock);
1535 return retval;
1536 }
1537
1538 /* rcu lock must be held */
1539 static int check_prlimit_permission(struct task_struct *task,
1540 unsigned int flags)
1541 {
1542 const struct cred *cred = current_cred(), *tcred;
1543 bool id_match;
1544
1545 if (current == task)
1546 return 0;
1547
1548 tcred = __task_cred(task);
1549 id_match = (uid_eq(cred->uid, tcred->euid) &&
1550 uid_eq(cred->uid, tcred->suid) &&
1551 uid_eq(cred->uid, tcred->uid) &&
1552 gid_eq(cred->gid, tcred->egid) &&
1553 gid_eq(cred->gid, tcred->sgid) &&
1554 gid_eq(cred->gid, tcred->gid));
1555 if (!id_match && !ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
1556 return -EPERM;
1557
1558 return security_task_prlimit(cred, tcred, flags);
1559 }
1560
1561 SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1562 const struct rlimit64 __user *, new_rlim,
1563 struct rlimit64 __user *, old_rlim)
1564 {
1565 struct rlimit64 old64, new64;
1566 struct rlimit old, new;
1567 struct task_struct *tsk;
1568 unsigned int checkflags = 0;
1569 int ret;
1570
1571 if (old_rlim)
1572 checkflags |= LSM_PRLIMIT_READ;
1573
1574 if (new_rlim) {
1575 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1576 return -EFAULT;
1577 rlim64_to_rlim(&new64, &new);
1578 checkflags |= LSM_PRLIMIT_WRITE;
1579 }
1580
1581 rcu_read_lock();
1582 tsk = pid ? find_task_by_vpid(pid) : current;
1583 if (!tsk) {
1584 rcu_read_unlock();
1585 return -ESRCH;
1586 }
1587 ret = check_prlimit_permission(tsk, checkflags);
1588 if (ret) {
1589 rcu_read_unlock();
1590 return ret;
1591 }
1592 get_task_struct(tsk);
1593 rcu_read_unlock();
1594
1595 ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1596 old_rlim ? &old : NULL);
1597
1598 if (!ret && old_rlim) {
1599 rlim_to_rlim64(&old, &old64);
1600 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1601 ret = -EFAULT;
1602 }
1603
1604 put_task_struct(tsk);
1605 return ret;
1606 }
1607
1608 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1609 {
1610 struct rlimit new_rlim;
1611
1612 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1613 return -EFAULT;
1614 return do_prlimit(current, resource, &new_rlim, NULL);
1615 }
1616
1617 /*
1618 * It would make sense to put struct rusage in the task_struct,
1619 * except that would make the task_struct be *really big*. After
1620 * task_struct gets moved into malloc'ed memory, it would
1621 * make sense to do this. It will make moving the rest of the information
1622 * a lot simpler! (Which we're not doing right now because we're not
1623 * measuring them yet).
1624 *
1625 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1626 * races with threads incrementing their own counters. But since word
1627 * reads are atomic, we either get new values or old values and we don't
1628 * care which for the sums. We always take the siglock to protect reading
1629 * the c* fields from p->signal from races with exit.c updating those
1630 * fields when reaping, so a sample either gets all the additions of a
1631 * given child after it's reaped, or none so this sample is before reaping.
1632 *
1633 * Locking:
1634 * We need to take the siglock for CHILDEREN, SELF and BOTH
1635 * for the cases current multithreaded, non-current single threaded
1636 * non-current multithreaded. Thread traversal is now safe with
1637 * the siglock held.
1638 * Strictly speaking, we donot need to take the siglock if we are current and
1639 * single threaded, as no one else can take our signal_struct away, no one
1640 * else can reap the children to update signal->c* counters, and no one else
1641 * can race with the signal-> fields. If we do not take any lock, the
1642 * signal-> fields could be read out of order while another thread was just
1643 * exiting. So we should place a read memory barrier when we avoid the lock.
1644 * On the writer side, write memory barrier is implied in __exit_signal
1645 * as __exit_signal releases the siglock spinlock after updating the signal->
1646 * fields. But we don't do this yet to keep things simple.
1647 *
1648 */
1649
1650 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1651 {
1652 r->ru_nvcsw += t->nvcsw;
1653 r->ru_nivcsw += t->nivcsw;
1654 r->ru_minflt += t->min_flt;
1655 r->ru_majflt += t->maj_flt;
1656 r->ru_inblock += task_io_get_inblock(t);
1657 r->ru_oublock += task_io_get_oublock(t);
1658 }
1659
1660 void getrusage(struct task_struct *p, int who, struct rusage *r)
1661 {
1662 struct task_struct *t;
1663 unsigned long flags;
1664 u64 tgutime, tgstime, utime, stime;
1665 unsigned long maxrss = 0;
1666
1667 memset((char *)r, 0, sizeof (*r));
1668 utime = stime = 0;
1669
1670 if (who == RUSAGE_THREAD) {
1671 task_cputime_adjusted(current, &utime, &stime);
1672 accumulate_thread_rusage(p, r);
1673 maxrss = p->signal->maxrss;
1674 goto out;
1675 }
1676
1677 if (!lock_task_sighand(p, &flags))
1678 return;
1679
1680 switch (who) {
1681 case RUSAGE_BOTH:
1682 case RUSAGE_CHILDREN:
1683 utime = p->signal->cutime;
1684 stime = p->signal->cstime;
1685 r->ru_nvcsw = p->signal->cnvcsw;
1686 r->ru_nivcsw = p->signal->cnivcsw;
1687 r->ru_minflt = p->signal->cmin_flt;
1688 r->ru_majflt = p->signal->cmaj_flt;
1689 r->ru_inblock = p->signal->cinblock;
1690 r->ru_oublock = p->signal->coublock;
1691 maxrss = p->signal->cmaxrss;
1692
1693 if (who == RUSAGE_CHILDREN)
1694 break;
1695
1696 case RUSAGE_SELF:
1697 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1698 utime += tgutime;
1699 stime += tgstime;
1700 r->ru_nvcsw += p->signal->nvcsw;
1701 r->ru_nivcsw += p->signal->nivcsw;
1702 r->ru_minflt += p->signal->min_flt;
1703 r->ru_majflt += p->signal->maj_flt;
1704 r->ru_inblock += p->signal->inblock;
1705 r->ru_oublock += p->signal->oublock;
1706 if (maxrss < p->signal->maxrss)
1707 maxrss = p->signal->maxrss;
1708 t = p;
1709 do {
1710 accumulate_thread_rusage(t, r);
1711 } while_each_thread(p, t);
1712 break;
1713
1714 default:
1715 BUG();
1716 }
1717 unlock_task_sighand(p, &flags);
1718
1719 out:
1720 r->ru_utime = ns_to_timeval(utime);
1721 r->ru_stime = ns_to_timeval(stime);
1722
1723 if (who != RUSAGE_CHILDREN) {
1724 struct mm_struct *mm = get_task_mm(p);
1725
1726 if (mm) {
1727 setmax_mm_hiwater_rss(&maxrss, mm);
1728 mmput(mm);
1729 }
1730 }
1731 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1732 }
1733
1734 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1735 {
1736 struct rusage r;
1737
1738 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1739 who != RUSAGE_THREAD)
1740 return -EINVAL;
1741
1742 getrusage(current, who, &r);
1743 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1744 }
1745
1746 #ifdef CONFIG_COMPAT
1747 COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru)
1748 {
1749 struct rusage r;
1750
1751 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1752 who != RUSAGE_THREAD)
1753 return -EINVAL;
1754
1755 getrusage(current, who, &r);
1756 return put_compat_rusage(&r, ru);
1757 }
1758 #endif
1759
1760 SYSCALL_DEFINE1(umask, int, mask)
1761 {
1762 mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1763 return mask;
1764 }
1765
1766 static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1767 {
1768 struct fd exe;
1769 struct file *old_exe, *exe_file;
1770 struct inode *inode;
1771 int err;
1772
1773 exe = fdget(fd);
1774 if (!exe.file)
1775 return -EBADF;
1776
1777 inode = file_inode(exe.file);
1778
1779 /*
1780 * Because the original mm->exe_file points to executable file, make
1781 * sure that this one is executable as well, to avoid breaking an
1782 * overall picture.
1783 */
1784 err = -EACCES;
1785 if (!S_ISREG(inode->i_mode) || path_noexec(&exe.file->f_path))
1786 goto exit;
1787
1788 err = inode_permission(inode, MAY_EXEC);
1789 if (err)
1790 goto exit;
1791
1792 /*
1793 * Forbid mm->exe_file change if old file still mapped.
1794 */
1795 exe_file = get_mm_exe_file(mm);
1796 err = -EBUSY;
1797 if (exe_file) {
1798 struct vm_area_struct *vma;
1799
1800 down_read(&mm->mmap_sem);
1801 for (vma = mm->mmap; vma; vma = vma->vm_next) {
1802 if (!vma->vm_file)
1803 continue;
1804 if (path_equal(&vma->vm_file->f_path,
1805 &exe_file->f_path))
1806 goto exit_err;
1807 }
1808
1809 up_read(&mm->mmap_sem);
1810 fput(exe_file);
1811 }
1812
1813 err = 0;
1814 /* set the new file, lockless */
1815 get_file(exe.file);
1816 old_exe = xchg(&mm->exe_file, exe.file);
1817 if (old_exe)
1818 fput(old_exe);
1819 exit:
1820 fdput(exe);
1821 return err;
1822 exit_err:
1823 up_read(&mm->mmap_sem);
1824 fput(exe_file);
1825 goto exit;
1826 }
1827
1828 /*
1829 * WARNING: we don't require any capability here so be very careful
1830 * in what is allowed for modification from userspace.
1831 */
1832 static int validate_prctl_map(struct prctl_mm_map *prctl_map)
1833 {
1834 unsigned long mmap_max_addr = TASK_SIZE;
1835 struct mm_struct *mm = current->mm;
1836 int error = -EINVAL, i;
1837
1838 static const unsigned char offsets[] = {
1839 offsetof(struct prctl_mm_map, start_code),
1840 offsetof(struct prctl_mm_map, end_code),
1841 offsetof(struct prctl_mm_map, start_data),
1842 offsetof(struct prctl_mm_map, end_data),
1843 offsetof(struct prctl_mm_map, start_brk),
1844 offsetof(struct prctl_mm_map, brk),
1845 offsetof(struct prctl_mm_map, start_stack),
1846 offsetof(struct prctl_mm_map, arg_start),
1847 offsetof(struct prctl_mm_map, arg_end),
1848 offsetof(struct prctl_mm_map, env_start),
1849 offsetof(struct prctl_mm_map, env_end),
1850 };
1851
1852 /*
1853 * Make sure the members are not somewhere outside
1854 * of allowed address space.
1855 */
1856 for (i = 0; i < ARRAY_SIZE(offsets); i++) {
1857 u64 val = *(u64 *)((char *)prctl_map + offsets[i]);
1858
1859 if ((unsigned long)val >= mmap_max_addr ||
1860 (unsigned long)val < mmap_min_addr)
1861 goto out;
1862 }
1863
1864 /*
1865 * Make sure the pairs are ordered.
1866 */
1867 #define __prctl_check_order(__m1, __op, __m2) \
1868 ((unsigned long)prctl_map->__m1 __op \
1869 (unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1870 error = __prctl_check_order(start_code, <, end_code);
1871 error |= __prctl_check_order(start_data, <, end_data);
1872 error |= __prctl_check_order(start_brk, <=, brk);
1873 error |= __prctl_check_order(arg_start, <=, arg_end);
1874 error |= __prctl_check_order(env_start, <=, env_end);
1875 if (error)
1876 goto out;
1877 #undef __prctl_check_order
1878
1879 error = -EINVAL;
1880
1881 /*
1882 * @brk should be after @end_data in traditional maps.
1883 */
1884 if (prctl_map->start_brk <= prctl_map->end_data ||
1885 prctl_map->brk <= prctl_map->end_data)
1886 goto out;
1887
1888 /*
1889 * Neither we should allow to override limits if they set.
1890 */
1891 if (check_data_rlimit(rlimit(RLIMIT_DATA), prctl_map->brk,
1892 prctl_map->start_brk, prctl_map->end_data,
1893 prctl_map->start_data))
1894 goto out;
1895
1896 /*
1897 * Someone is trying to cheat the auxv vector.
1898 */
1899 if (prctl_map->auxv_size) {
1900 if (!prctl_map->auxv || prctl_map->auxv_size > sizeof(mm->saved_auxv))
1901 goto out;
1902 }
1903
1904 /*
1905 * Finally, make sure the caller has the rights to
1906 * change /proc/pid/exe link: only local sys admin should
1907 * be allowed to.
1908 */
1909 if (prctl_map->exe_fd != (u32)-1) {
1910 if (!ns_capable(current_user_ns(), CAP_SYS_ADMIN))
1911 goto out;
1912 }
1913
1914 error = 0;
1915 out:
1916 return error;
1917 }
1918
1919 #ifdef CONFIG_CHECKPOINT_RESTORE
1920 static int prctl_set_mm_map(int opt, const void __user *addr, unsigned long data_size)
1921 {
1922 struct prctl_mm_map prctl_map = { .exe_fd = (u32)-1, };
1923 unsigned long user_auxv[AT_VECTOR_SIZE];
1924 struct mm_struct *mm = current->mm;
1925 int error;
1926
1927 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
1928 BUILD_BUG_ON(sizeof(struct prctl_mm_map) > 256);
1929
1930 if (opt == PR_SET_MM_MAP_SIZE)
1931 return put_user((unsigned int)sizeof(prctl_map),
1932 (unsigned int __user *)addr);
1933
1934 if (data_size != sizeof(prctl_map))
1935 return -EINVAL;
1936
1937 if (copy_from_user(&prctl_map, addr, sizeof(prctl_map)))
1938 return -EFAULT;
1939
1940 error = validate_prctl_map(&prctl_map);
1941 if (error)
1942 return error;
1943
1944 if (prctl_map.auxv_size) {
1945 memset(user_auxv, 0, sizeof(user_auxv));
1946 if (copy_from_user(user_auxv,
1947 (const void __user *)prctl_map.auxv,
1948 prctl_map.auxv_size))
1949 return -EFAULT;
1950
1951 /* Last entry must be AT_NULL as specification requires */
1952 user_auxv[AT_VECTOR_SIZE - 2] = AT_NULL;
1953 user_auxv[AT_VECTOR_SIZE - 1] = AT_NULL;
1954 }
1955
1956 if (prctl_map.exe_fd != (u32)-1) {
1957 error = prctl_set_mm_exe_file(mm, prctl_map.exe_fd);
1958 if (error)
1959 return error;
1960 }
1961
1962 down_write(&mm->mmap_sem);
1963
1964 /*
1965 * We don't validate if these members are pointing to
1966 * real present VMAs because application may have correspond
1967 * VMAs already unmapped and kernel uses these members for statistics
1968 * output in procfs mostly, except
1969 *
1970 * - @start_brk/@brk which are used in do_brk but kernel lookups
1971 * for VMAs when updating these memvers so anything wrong written
1972 * here cause kernel to swear at userspace program but won't lead
1973 * to any problem in kernel itself
1974 */
1975
1976 mm->start_code = prctl_map.start_code;
1977 mm->end_code = prctl_map.end_code;
1978 mm->start_data = prctl_map.start_data;
1979 mm->end_data = prctl_map.end_data;
1980 mm->start_brk = prctl_map.start_brk;
1981 mm->brk = prctl_map.brk;
1982 mm->start_stack = prctl_map.start_stack;
1983 mm->arg_start = prctl_map.arg_start;
1984 mm->arg_end = prctl_map.arg_end;
1985 mm->env_start = prctl_map.env_start;
1986 mm->env_end = prctl_map.env_end;
1987
1988 /*
1989 * Note this update of @saved_auxv is lockless thus
1990 * if someone reads this member in procfs while we're
1991 * updating -- it may get partly updated results. It's
1992 * known and acceptable trade off: we leave it as is to
1993 * not introduce additional locks here making the kernel
1994 * more complex.
1995 */
1996 if (prctl_map.auxv_size)
1997 memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv));
1998
1999 up_write(&mm->mmap_sem);
2000 return 0;
2001 }
2002 #endif /* CONFIG_CHECKPOINT_RESTORE */
2003
2004 static int prctl_set_auxv(struct mm_struct *mm, unsigned long addr,
2005 unsigned long len)
2006 {
2007 /*
2008 * This doesn't move the auxiliary vector itself since it's pinned to
2009 * mm_struct, but it permits filling the vector with new values. It's
2010 * up to the caller to provide sane values here, otherwise userspace
2011 * tools which use this vector might be unhappy.
2012 */
2013 unsigned long user_auxv[AT_VECTOR_SIZE];
2014
2015 if (len > sizeof(user_auxv))
2016 return -EINVAL;
2017
2018 if (copy_from_user(user_auxv, (const void __user *)addr, len))
2019 return -EFAULT;
2020
2021 /* Make sure the last entry is always AT_NULL */
2022 user_auxv[AT_VECTOR_SIZE - 2] = 0;
2023 user_auxv[AT_VECTOR_SIZE - 1] = 0;
2024
2025 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
2026
2027 task_lock(current);
2028 memcpy(mm->saved_auxv, user_auxv, len);
2029 task_unlock(current);
2030
2031 return 0;
2032 }
2033
2034 static int prctl_set_mm(int opt, unsigned long addr,
2035 unsigned long arg4, unsigned long arg5)
2036 {
2037 struct mm_struct *mm = current->mm;
2038 struct prctl_mm_map prctl_map;
2039 struct vm_area_struct *vma;
2040 int error;
2041
2042 if (arg5 || (arg4 && (opt != PR_SET_MM_AUXV &&
2043 opt != PR_SET_MM_MAP &&
2044 opt != PR_SET_MM_MAP_SIZE)))
2045 return -EINVAL;
2046
2047 #ifdef CONFIG_CHECKPOINT_RESTORE
2048 if (opt == PR_SET_MM_MAP || opt == PR_SET_MM_MAP_SIZE)
2049 return prctl_set_mm_map(opt, (const void __user *)addr, arg4);
2050 #endif
2051
2052 if (!capable(CAP_SYS_RESOURCE))
2053 return -EPERM;
2054
2055 if (opt == PR_SET_MM_EXE_FILE)
2056 return prctl_set_mm_exe_file(mm, (unsigned int)addr);
2057
2058 if (opt == PR_SET_MM_AUXV)
2059 return prctl_set_auxv(mm, addr, arg4);
2060
2061 if (addr >= TASK_SIZE || addr < mmap_min_addr)
2062 return -EINVAL;
2063
2064 error = -EINVAL;
2065
2066 down_write(&mm->mmap_sem);
2067 vma = find_vma(mm, addr);
2068
2069 prctl_map.start_code = mm->start_code;
2070 prctl_map.end_code = mm->end_code;
2071 prctl_map.start_data = mm->start_data;
2072 prctl_map.end_data = mm->end_data;
2073 prctl_map.start_brk = mm->start_brk;
2074 prctl_map.brk = mm->brk;
2075 prctl_map.start_stack = mm->start_stack;
2076 prctl_map.arg_start = mm->arg_start;
2077 prctl_map.arg_end = mm->arg_end;
2078 prctl_map.env_start = mm->env_start;
2079 prctl_map.env_end = mm->env_end;
2080 prctl_map.auxv = NULL;
2081 prctl_map.auxv_size = 0;
2082 prctl_map.exe_fd = -1;
2083
2084 switch (opt) {
2085 case PR_SET_MM_START_CODE:
2086 prctl_map.start_code = addr;
2087 break;
2088 case PR_SET_MM_END_CODE:
2089 prctl_map.end_code = addr;
2090 break;
2091 case PR_SET_MM_START_DATA:
2092 prctl_map.start_data = addr;
2093 break;
2094 case PR_SET_MM_END_DATA:
2095 prctl_map.end_data = addr;
2096 break;
2097 case PR_SET_MM_START_STACK:
2098 prctl_map.start_stack = addr;
2099 break;
2100 case PR_SET_MM_START_BRK:
2101 prctl_map.start_brk = addr;
2102 break;
2103 case PR_SET_MM_BRK:
2104 prctl_map.brk = addr;
2105 break;
2106 case PR_SET_MM_ARG_START:
2107 prctl_map.arg_start = addr;
2108 break;
2109 case PR_SET_MM_ARG_END:
2110 prctl_map.arg_end = addr;
2111 break;
2112 case PR_SET_MM_ENV_START:
2113 prctl_map.env_start = addr;
2114 break;
2115 case PR_SET_MM_ENV_END:
2116 prctl_map.env_end = addr;
2117 break;
2118 default:
2119 goto out;
2120 }
2121
2122 error = validate_prctl_map(&prctl_map);
2123 if (error)
2124 goto out;
2125
2126 switch (opt) {
2127 /*
2128 * If command line arguments and environment
2129 * are placed somewhere else on stack, we can
2130 * set them up here, ARG_START/END to setup
2131 * command line argumets and ENV_START/END
2132 * for environment.
2133 */
2134 case PR_SET_MM_START_STACK:
2135 case PR_SET_MM_ARG_START:
2136 case PR_SET_MM_ARG_END:
2137 case PR_SET_MM_ENV_START:
2138 case PR_SET_MM_ENV_END:
2139 if (!vma) {
2140 error = -EFAULT;
2141 goto out;
2142 }
2143 }
2144
2145 mm->start_code = prctl_map.start_code;
2146 mm->end_code = prctl_map.end_code;
2147 mm->start_data = prctl_map.start_data;
2148 mm->end_data = prctl_map.end_data;
2149 mm->start_brk = prctl_map.start_brk;
2150 mm->brk = prctl_map.brk;
2151 mm->start_stack = prctl_map.start_stack;
2152 mm->arg_start = prctl_map.arg_start;
2153 mm->arg_end = prctl_map.arg_end;
2154 mm->env_start = prctl_map.env_start;
2155 mm->env_end = prctl_map.env_end;
2156
2157 error = 0;
2158 out:
2159 up_write(&mm->mmap_sem);
2160 return error;
2161 }
2162
2163 #ifdef CONFIG_CHECKPOINT_RESTORE
2164 static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2165 {
2166 return put_user(me->clear_child_tid, tid_addr);
2167 }
2168 #else
2169 static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2170 {
2171 return -EINVAL;
2172 }
2173 #endif
2174
2175 static int propagate_has_child_subreaper(struct task_struct *p, void *data)
2176 {
2177 /*
2178 * If task has has_child_subreaper - all its decendants
2179 * already have these flag too and new decendants will
2180 * inherit it on fork, skip them.
2181 *
2182 * If we've found child_reaper - skip descendants in
2183 * it's subtree as they will never get out pidns.
2184 */
2185 if (p->signal->has_child_subreaper ||
2186 is_child_reaper(task_pid(p)))
2187 return 0;
2188
2189 p->signal->has_child_subreaper = 1;
2190 return 1;
2191 }
2192
2193 #ifdef CONFIG_MMU
2194 static int prctl_update_vma_anon_name(struct vm_area_struct *vma,
2195 struct vm_area_struct **prev,
2196 unsigned long start, unsigned long end,
2197 const char __user *name_addr)
2198 {
2199 struct mm_struct *mm = vma->vm_mm;
2200 int error = 0;
2201 pgoff_t pgoff;
2202
2203 if (name_addr == vma_get_anon_name(vma)) {
2204 *prev = vma;
2205 goto out;
2206 }
2207
2208 pgoff = vma->vm_pgoff + ((start - vma->vm_start) >> PAGE_SHIFT);
2209 *prev = vma_merge(mm, *prev, start, end, vma->vm_flags, vma->anon_vma,
2210 vma->vm_file, pgoff, vma_policy(vma),
2211 vma->vm_userfaultfd_ctx, name_addr);
2212 if (*prev) {
2213 vma = *prev;
2214 goto success;
2215 }
2216
2217 *prev = vma;
2218
2219 if (start != vma->vm_start) {
2220 error = split_vma(mm, vma, start, 1);
2221 if (error)
2222 goto out;
2223 }
2224
2225 if (end != vma->vm_end) {
2226 error = split_vma(mm, vma, end, 0);
2227 if (error)
2228 goto out;
2229 }
2230
2231 success:
2232 if (!vma->vm_file)
2233 vma->anon_name = name_addr;
2234
2235 out:
2236 if (error == -ENOMEM)
2237 error = -EAGAIN;
2238 return error;
2239 }
2240
2241 static int prctl_set_vma_anon_name(unsigned long start, unsigned long end,
2242 unsigned long arg)
2243 {
2244 unsigned long tmp;
2245 struct vm_area_struct *vma, *prev;
2246 int unmapped_error = 0;
2247 int error = -EINVAL;
2248
2249 /*
2250 * If the interval [start,end) covers some unmapped address
2251 * ranges, just ignore them, but return -ENOMEM at the end.
2252 * - this matches the handling in madvise.
2253 */
2254 vma = find_vma_prev(current->mm, start, &prev);
2255 if (vma && start > vma->vm_start)
2256 prev = vma;
2257
2258 for (;;) {
2259 /* Still start < end. */
2260 error = -ENOMEM;
2261 if (!vma)
2262 return error;
2263
2264 /* Here start < (end|vma->vm_end). */
2265 if (start < vma->vm_start) {
2266 unmapped_error = -ENOMEM;
2267 start = vma->vm_start;
2268 if (start >= end)
2269 return error;
2270 }
2271
2272 /* Here vma->vm_start <= start < (end|vma->vm_end) */
2273 tmp = vma->vm_end;
2274 if (end < tmp)
2275 tmp = end;
2276
2277 /* Here vma->vm_start <= start < tmp <= (end|vma->vm_end). */
2278 error = prctl_update_vma_anon_name(vma, &prev, start, tmp,
2279 (const char __user *)arg);
2280 if (error)
2281 return error;
2282 start = tmp;
2283 if (prev && start < prev->vm_end)
2284 start = prev->vm_end;
2285 error = unmapped_error;
2286 if (start >= end)
2287 return error;
2288 if (prev)
2289 vma = prev->vm_next;
2290 else /* madvise_remove dropped mmap_sem */
2291 vma = find_vma(current->mm, start);
2292 }
2293 }
2294
2295 static int prctl_set_vma(unsigned long opt, unsigned long start,
2296 unsigned long len_in, unsigned long arg)
2297 {
2298 struct mm_struct *mm = current->mm;
2299 int error;
2300 unsigned long len;
2301 unsigned long end;
2302
2303 if (start & ~PAGE_MASK)
2304 return -EINVAL;
2305 len = (len_in + ~PAGE_MASK) & PAGE_MASK;
2306
2307 /* Check to see whether len was rounded up from small -ve to zero */
2308 if (len_in && !len)
2309 return -EINVAL;
2310
2311 end = start + len;
2312 if (end < start)
2313 return -EINVAL;
2314
2315 if (end == start)
2316 return 0;
2317
2318 down_write(&mm->mmap_sem);
2319
2320 switch (opt) {
2321 case PR_SET_VMA_ANON_NAME:
2322 error = prctl_set_vma_anon_name(start, end, arg);
2323 break;
2324 default:
2325 error = -EINVAL;
2326 }
2327
2328 up_write(&mm->mmap_sem);
2329
2330 return error;
2331 }
2332 #else /* CONFIG_MMU */
2333 static int prctl_set_vma(unsigned long opt, unsigned long start,
2334 unsigned long len_in, unsigned long arg)
2335 {
2336 return -EINVAL;
2337 }
2338 #endif
2339
2340 int __weak arch_prctl_spec_ctrl_get(struct task_struct *t, unsigned long which)
2341 {
2342 return -EINVAL;
2343 }
2344
2345 int __weak arch_prctl_spec_ctrl_set(struct task_struct *t, unsigned long which,
2346 unsigned long ctrl)
2347 {
2348 return -EINVAL;
2349 }
2350
2351 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2352 unsigned long, arg4, unsigned long, arg5)
2353 {
2354 struct task_struct *me = current;
2355 unsigned char comm[sizeof(me->comm)];
2356 long error;
2357
2358 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2359 if (error != -ENOSYS)
2360 return error;
2361
2362 error = 0;
2363 switch (option) {
2364 case PR_SET_PDEATHSIG:
2365 if (!valid_signal(arg2)) {
2366 error = -EINVAL;
2367 break;
2368 }
2369 me->pdeath_signal = arg2;
2370 break;
2371 case PR_GET_PDEATHSIG:
2372 error = put_user(me->pdeath_signal, (int __user *)arg2);
2373 break;
2374 case PR_GET_DUMPABLE:
2375 error = get_dumpable(me->mm);
2376 break;
2377 case PR_SET_DUMPABLE:
2378 if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) {
2379 error = -EINVAL;
2380 break;
2381 }
2382 set_dumpable(me->mm, arg2);
2383 break;
2384
2385 case PR_SET_UNALIGN:
2386 error = SET_UNALIGN_CTL(me, arg2);
2387 break;
2388 case PR_GET_UNALIGN:
2389 error = GET_UNALIGN_CTL(me, arg2);
2390 break;
2391 case PR_SET_FPEMU:
2392 error = SET_FPEMU_CTL(me, arg2);
2393 break;
2394 case PR_GET_FPEMU:
2395 error = GET_FPEMU_CTL(me, arg2);
2396 break;
2397 case PR_SET_FPEXC:
2398 error = SET_FPEXC_CTL(me, arg2);
2399 break;
2400 case PR_GET_FPEXC:
2401 error = GET_FPEXC_CTL(me, arg2);
2402 break;
2403 case PR_GET_TIMING:
2404 error = PR_TIMING_STATISTICAL;
2405 break;
2406 case PR_SET_TIMING:
2407 if (arg2 != PR_TIMING_STATISTICAL)
2408 error = -EINVAL;
2409 break;
2410 case PR_SET_NAME:
2411 comm[sizeof(me->comm) - 1] = 0;
2412 if (strncpy_from_user(comm, (char __user *)arg2,
2413 sizeof(me->comm) - 1) < 0)
2414 return -EFAULT;
2415 set_task_comm(me, comm);
2416 proc_comm_connector(me);
2417 break;
2418 case PR_GET_NAME:
2419 get_task_comm(comm, me);
2420 if (copy_to_user((char __user *)arg2, comm, sizeof(comm)))
2421 return -EFAULT;
2422 break;
2423 case PR_GET_ENDIAN:
2424 error = GET_ENDIAN(me, arg2);
2425 break;
2426 case PR_SET_ENDIAN:
2427 error = SET_ENDIAN(me, arg2);
2428 break;
2429 case PR_GET_SECCOMP:
2430 error = prctl_get_seccomp();
2431 break;
2432 case PR_SET_SECCOMP:
2433 error = prctl_set_seccomp(arg2, (char __user *)arg3);
2434 break;
2435 case PR_GET_TSC:
2436 error = GET_TSC_CTL(arg2);
2437 break;
2438 case PR_SET_TSC:
2439 error = SET_TSC_CTL(arg2);
2440 break;
2441 case PR_TASK_PERF_EVENTS_DISABLE:
2442 error = perf_event_task_disable();
2443 break;
2444 case PR_TASK_PERF_EVENTS_ENABLE:
2445 error = perf_event_task_enable();
2446 break;
2447 case PR_GET_TIMERSLACK:
2448 if (current->timer_slack_ns > ULONG_MAX)
2449 error = ULONG_MAX;
2450 else
2451 error = current->timer_slack_ns;
2452 break;
2453 case PR_SET_TIMERSLACK:
2454 if (arg2 <= 0)
2455 current->timer_slack_ns =
2456 current->default_timer_slack_ns;
2457 else
2458 current->timer_slack_ns = arg2;
2459 break;
2460 case PR_MCE_KILL:
2461 if (arg4 | arg5)
2462 return -EINVAL;
2463 switch (arg2) {
2464 case PR_MCE_KILL_CLEAR:
2465 if (arg3 != 0)
2466 return -EINVAL;
2467 current->flags &= ~PF_MCE_PROCESS;
2468 break;
2469 case PR_MCE_KILL_SET:
2470 current->flags |= PF_MCE_PROCESS;
2471 if (arg3 == PR_MCE_KILL_EARLY)
2472 current->flags |= PF_MCE_EARLY;
2473 else if (arg3 == PR_MCE_KILL_LATE)
2474 current->flags &= ~PF_MCE_EARLY;
2475 else if (arg3 == PR_MCE_KILL_DEFAULT)
2476 current->flags &=
2477 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
2478 else
2479 return -EINVAL;
2480 break;
2481 default:
2482 return -EINVAL;
2483 }
2484 break;
2485 case PR_MCE_KILL_GET:
2486 if (arg2 | arg3 | arg4 | arg5)
2487 return -EINVAL;
2488 if (current->flags & PF_MCE_PROCESS)
2489 error = (current->flags & PF_MCE_EARLY) ?
2490 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2491 else
2492 error = PR_MCE_KILL_DEFAULT;
2493 break;
2494 case PR_SET_MM:
2495 error = prctl_set_mm(arg2, arg3, arg4, arg5);
2496 break;
2497 case PR_GET_TID_ADDRESS:
2498 error = prctl_get_tid_address(me, (int __user **)arg2);
2499 break;
2500 case PR_SET_CHILD_SUBREAPER:
2501 me->signal->is_child_subreaper = !!arg2;
2502 if (!arg2)
2503 break;
2504
2505 walk_process_tree(me, propagate_has_child_subreaper, NULL);
2506 break;
2507 case PR_GET_CHILD_SUBREAPER:
2508 error = put_user(me->signal->is_child_subreaper,
2509 (int __user *)arg2);
2510 break;
2511 case PR_SET_NO_NEW_PRIVS:
2512 if (arg2 != 1 || arg3 || arg4 || arg5)
2513 return -EINVAL;
2514
2515 task_set_no_new_privs(current);
2516 break;
2517 case PR_GET_NO_NEW_PRIVS:
2518 if (arg2 || arg3 || arg4 || arg5)
2519 return -EINVAL;
2520 return task_no_new_privs(current) ? 1 : 0;
2521 case PR_GET_THP_DISABLE:
2522 if (arg2 || arg3 || arg4 || arg5)
2523 return -EINVAL;
2524 error = !!test_bit(MMF_DISABLE_THP, &me->mm->flags);
2525 break;
2526 case PR_SET_THP_DISABLE:
2527 if (arg3 || arg4 || arg5)
2528 return -EINVAL;
2529 if (down_write_killable(&me->mm->mmap_sem))
2530 return -EINTR;
2531 if (arg2)
2532 set_bit(MMF_DISABLE_THP, &me->mm->flags);
2533 else
2534 clear_bit(MMF_DISABLE_THP, &me->mm->flags);
2535 up_write(&me->mm->mmap_sem);
2536 break;
2537 case PR_MPX_ENABLE_MANAGEMENT:
2538 if (arg2 || arg3 || arg4 || arg5)
2539 return -EINVAL;
2540 error = MPX_ENABLE_MANAGEMENT();
2541 break;
2542 case PR_MPX_DISABLE_MANAGEMENT:
2543 if (arg2 || arg3 || arg4 || arg5)
2544 return -EINVAL;
2545 error = MPX_DISABLE_MANAGEMENT();
2546 break;
2547 case PR_SET_FP_MODE:
2548 error = SET_FP_MODE(me, arg2);
2549 break;
2550 case PR_GET_FP_MODE:
2551 error = GET_FP_MODE(me);
2552 break;
2553 case PR_GET_SPECULATION_CTRL:
2554 if (arg3 || arg4 || arg5)
2555 return -EINVAL;
2556 error = arch_prctl_spec_ctrl_get(me, arg2);
2557 break;
2558 case PR_SET_SPECULATION_CTRL:
2559 if (arg4 || arg5)
2560 return -EINVAL;
2561 error = arch_prctl_spec_ctrl_set(me, arg2, arg3);
2562 break;
2563 case PR_SET_VMA:
2564 error = prctl_set_vma(arg2, arg3, arg4, arg5);
2565 break;
2566 default:
2567 error = -EINVAL;
2568 break;
2569 }
2570 return error;
2571 }
2572
2573 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2574 struct getcpu_cache __user *, unused)
2575 {
2576 int err = 0;
2577 int cpu = raw_smp_processor_id();
2578
2579 if (cpup)
2580 err |= put_user(cpu, cpup);
2581 if (nodep)
2582 err |= put_user(cpu_to_node(cpu), nodep);
2583 return err ? -EFAULT : 0;
2584 }
2585
2586 /**
2587 * do_sysinfo - fill in sysinfo struct
2588 * @info: pointer to buffer to fill
2589 */
2590 static int do_sysinfo(struct sysinfo *info)
2591 {
2592 unsigned long mem_total, sav_total;
2593 unsigned int mem_unit, bitcount;
2594 struct timespec tp;
2595
2596 memset(info, 0, sizeof(struct sysinfo));
2597
2598 get_monotonic_boottime(&tp);
2599 info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
2600
2601 get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
2602
2603 info->procs = nr_threads;
2604
2605 si_meminfo(info);
2606 si_swapinfo(info);
2607
2608 /*
2609 * If the sum of all the available memory (i.e. ram + swap)
2610 * is less than can be stored in a 32 bit unsigned long then
2611 * we can be binary compatible with 2.2.x kernels. If not,
2612 * well, in that case 2.2.x was broken anyways...
2613 *
2614 * -Erik Andersen <andersee@debian.org>
2615 */
2616
2617 mem_total = info->totalram + info->totalswap;
2618 if (mem_total < info->totalram || mem_total < info->totalswap)
2619 goto out;
2620 bitcount = 0;
2621 mem_unit = info->mem_unit;
2622 while (mem_unit > 1) {
2623 bitcount++;
2624 mem_unit >>= 1;
2625 sav_total = mem_total;
2626 mem_total <<= 1;
2627 if (mem_total < sav_total)
2628 goto out;
2629 }
2630
2631 /*
2632 * If mem_total did not overflow, multiply all memory values by
2633 * info->mem_unit and set it to 1. This leaves things compatible
2634 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2635 * kernels...
2636 */
2637
2638 info->mem_unit = 1;
2639 info->totalram <<= bitcount;
2640 info->freeram <<= bitcount;
2641 info->sharedram <<= bitcount;
2642 info->bufferram <<= bitcount;
2643 info->totalswap <<= bitcount;
2644 info->freeswap <<= bitcount;
2645 info->totalhigh <<= bitcount;
2646 info->freehigh <<= bitcount;
2647
2648 out:
2649 return 0;
2650 }
2651
2652 SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
2653 {
2654 struct sysinfo val;
2655
2656 do_sysinfo(&val);
2657
2658 if (copy_to_user(info, &val, sizeof(struct sysinfo)))
2659 return -EFAULT;
2660
2661 return 0;
2662 }
2663
2664 #ifdef CONFIG_COMPAT
2665 struct compat_sysinfo {
2666 s32 uptime;
2667 u32 loads[3];
2668 u32 totalram;
2669 u32 freeram;
2670 u32 sharedram;
2671 u32 bufferram;
2672 u32 totalswap;
2673 u32 freeswap;
2674 u16 procs;
2675 u16 pad;
2676 u32 totalhigh;
2677 u32 freehigh;
2678 u32 mem_unit;
2679 char _f[20-2*sizeof(u32)-sizeof(int)];
2680 };
2681
2682 COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2683 {
2684 struct sysinfo s;
2685
2686 do_sysinfo(&s);
2687
2688 /* Check to see if any memory value is too large for 32-bit and scale
2689 * down if needed
2690 */
2691 if (upper_32_bits(s.totalram) || upper_32_bits(s.totalswap)) {
2692 int bitcount = 0;
2693
2694 while (s.mem_unit < PAGE_SIZE) {
2695 s.mem_unit <<= 1;
2696 bitcount++;
2697 }
2698
2699 s.totalram >>= bitcount;
2700 s.freeram >>= bitcount;
2701 s.sharedram >>= bitcount;
2702 s.bufferram >>= bitcount;
2703 s.totalswap >>= bitcount;
2704 s.freeswap >>= bitcount;
2705 s.totalhigh >>= bitcount;
2706 s.freehigh >>= bitcount;
2707 }
2708
2709 if (!access_ok(VERIFY_WRITE, info, sizeof(struct compat_sysinfo)) ||
2710 __put_user(s.uptime, &info->uptime) ||
2711 __put_user(s.loads[0], &info->loads[0]) ||
2712 __put_user(s.loads[1], &info->loads[1]) ||
2713 __put_user(s.loads[2], &info->loads[2]) ||
2714 __put_user(s.totalram, &info->totalram) ||
2715 __put_user(s.freeram, &info->freeram) ||
2716 __put_user(s.sharedram, &info->sharedram) ||
2717 __put_user(s.bufferram, &info->bufferram) ||
2718 __put_user(s.totalswap, &info->totalswap) ||
2719 __put_user(s.freeswap, &info->freeswap) ||
2720 __put_user(s.procs, &info->procs) ||
2721 __put_user(s.totalhigh, &info->totalhigh) ||
2722 __put_user(s.freehigh, &info->freehigh) ||
2723 __put_user(s.mem_unit, &info->mem_unit))
2724 return -EFAULT;
2725
2726 return 0;
2727 }
2728 #endif /* CONFIG_COMPAT */
Motorola kernel-slsi