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 #include "uid16.h"
77
78 #include <trace/hooks/sys.h>
79
80 #ifndef SET_UNALIGN_CTL
81 # define SET_UNALIGN_CTL(a, b) (-EINVAL)
82 #endif
83 #ifndef GET_UNALIGN_CTL
84 # define GET_UNALIGN_CTL(a, b) (-EINVAL)
85 #endif
86 #ifndef SET_FPEMU_CTL
87 # define SET_FPEMU_CTL(a, b) (-EINVAL)
88 #endif
89 #ifndef GET_FPEMU_CTL
90 # define GET_FPEMU_CTL(a, b) (-EINVAL)
91 #endif
92 #ifndef SET_FPEXC_CTL
93 # define SET_FPEXC_CTL(a, b) (-EINVAL)
94 #endif
95 #ifndef GET_FPEXC_CTL
96 # define GET_FPEXC_CTL(a, b) (-EINVAL)
97 #endif
98 #ifndef GET_ENDIAN
99 # define GET_ENDIAN(a, b) (-EINVAL)
100 #endif
101 #ifndef SET_ENDIAN
102 # define SET_ENDIAN(a, b) (-EINVAL)
103 #endif
104 #ifndef GET_TSC_CTL
105 # define GET_TSC_CTL(a) (-EINVAL)
106 #endif
107 #ifndef SET_TSC_CTL
108 # define SET_TSC_CTL(a) (-EINVAL)
109 #endif
110 #ifndef GET_FP_MODE
111 # define GET_FP_MODE(a) (-EINVAL)
112 #endif
113 #ifndef SET_FP_MODE
114 # define SET_FP_MODE(a,b) (-EINVAL)
115 #endif
116 #ifndef SVE_SET_VL
117 # define SVE_SET_VL(a) (-EINVAL)
118 #endif
119 #ifndef SVE_GET_VL
120 # define SVE_GET_VL() (-EINVAL)
121 #endif
122 #ifndef PAC_RESET_KEYS
123 # define PAC_RESET_KEYS(a, b) (-EINVAL)
124 #endif
125 #ifndef SET_TAGGED_ADDR_CTRL
126 # define SET_TAGGED_ADDR_CTRL(a) (-EINVAL)
127 #endif
128 #ifndef GET_TAGGED_ADDR_CTRL
129 # define GET_TAGGED_ADDR_CTRL() (-EINVAL)
130 #endif
131
132 /*
133 * this is where the system-wide overflow UID and GID are defined, for
134 * architectures that now have 32-bit UID/GID but didn't in the past
135 */
136
137 int overflowuid = DEFAULT_OVERFLOWUID;
138 int overflowgid = DEFAULT_OVERFLOWGID;
139
140 EXPORT_SYMBOL(overflowuid);
141 EXPORT_SYMBOL(overflowgid);
142
143 /*
144 * the same as above, but for filesystems which can only store a 16-bit
145 * UID and GID. as such, this is needed on all architectures
146 */
147
148 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
149 int fs_overflowgid = DEFAULT_FS_OVERFLOWGID;
150
151 EXPORT_SYMBOL(fs_overflowuid);
152 EXPORT_SYMBOL(fs_overflowgid);
153
154 /*
155 * Returns true if current's euid is same as p's uid or euid,
156 * or has CAP_SYS_NICE to p's user_ns.
157 *
158 * Called with rcu_read_lock, creds are safe
159 */
set_one_prio_perm(struct task_struct * p)160 static bool set_one_prio_perm(struct task_struct *p)
161 {
162 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
163
164 if (uid_eq(pcred->uid, cred->euid) ||
165 uid_eq(pcred->euid, cred->euid))
166 return true;
167 if (ns_capable(pcred->user_ns, CAP_SYS_NICE))
168 return true;
169 return false;
170 }
171
172 /*
173 * set the priority of a task
174 * - the caller must hold the RCU read lock
175 */
set_one_prio(struct task_struct * p,int niceval,int error)176 static int set_one_prio(struct task_struct *p, int niceval, int error)
177 {
178 int no_nice;
179
180 if (!set_one_prio_perm(p)) {
181 error = -EPERM;
182 goto out;
183 }
184 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
185 error = -EACCES;
186 goto out;
187 }
188 no_nice = security_task_setnice(p, niceval);
189 if (no_nice) {
190 error = no_nice;
191 goto out;
192 }
193 if (error == -ESRCH)
194 error = 0;
195 set_user_nice(p, niceval);
196 out:
197 return error;
198 }
199
SYSCALL_DEFINE3(setpriority,int,which,int,who,int,niceval)200 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
201 {
202 struct task_struct *g, *p;
203 struct user_struct *user;
204 const struct cred *cred = current_cred();
205 int error = -EINVAL;
206 struct pid *pgrp;
207 kuid_t uid;
208
209 if (which > PRIO_USER || which < PRIO_PROCESS)
210 goto out;
211
212 /* normalize: avoid signed division (rounding problems) */
213 error = -ESRCH;
214 if (niceval < MIN_NICE)
215 niceval = MIN_NICE;
216 if (niceval > MAX_NICE)
217 niceval = MAX_NICE;
218
219 rcu_read_lock();
220 read_lock(&tasklist_lock);
221 switch (which) {
222 case PRIO_PROCESS:
223 if (who)
224 p = find_task_by_vpid(who);
225 else
226 p = current;
227 if (p)
228 error = set_one_prio(p, niceval, error);
229 break;
230 case PRIO_PGRP:
231 if (who)
232 pgrp = find_vpid(who);
233 else
234 pgrp = task_pgrp(current);
235 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
236 error = set_one_prio(p, niceval, error);
237 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
238 break;
239 case PRIO_USER:
240 uid = make_kuid(cred->user_ns, who);
241 user = cred->user;
242 if (!who)
243 uid = cred->uid;
244 else if (!uid_eq(uid, cred->uid)) {
245 user = find_user(uid);
246 if (!user)
247 goto out_unlock; /* No processes for this user */
248 }
249 do_each_thread(g, p) {
250 if (uid_eq(task_uid(p), uid) && task_pid_vnr(p))
251 error = set_one_prio(p, niceval, error);
252 } while_each_thread(g, p);
253 if (!uid_eq(uid, cred->uid))
254 free_uid(user); /* For find_user() */
255 break;
256 }
257 out_unlock:
258 read_unlock(&tasklist_lock);
259 rcu_read_unlock();
260 out:
261 return error;
262 }
263
264 /*
265 * Ugh. To avoid negative return values, "getpriority()" will
266 * not return the normal nice-value, but a negated value that
267 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
268 * to stay compatible.
269 */
SYSCALL_DEFINE2(getpriority,int,which,int,who)270 SYSCALL_DEFINE2(getpriority, int, which, int, who)
271 {
272 struct task_struct *g, *p;
273 struct user_struct *user;
274 const struct cred *cred = current_cred();
275 long niceval, retval = -ESRCH;
276 struct pid *pgrp;
277 kuid_t uid;
278
279 if (which > PRIO_USER || which < PRIO_PROCESS)
280 return -EINVAL;
281
282 rcu_read_lock();
283 read_lock(&tasklist_lock);
284 switch (which) {
285 case PRIO_PROCESS:
286 if (who)
287 p = find_task_by_vpid(who);
288 else
289 p = current;
290 if (p) {
291 niceval = nice_to_rlimit(task_nice(p));
292 if (niceval > retval)
293 retval = niceval;
294 }
295 break;
296 case PRIO_PGRP:
297 if (who)
298 pgrp = find_vpid(who);
299 else
300 pgrp = task_pgrp(current);
301 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
302 niceval = nice_to_rlimit(task_nice(p));
303 if (niceval > retval)
304 retval = niceval;
305 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
306 break;
307 case PRIO_USER:
308 uid = make_kuid(cred->user_ns, who);
309 user = cred->user;
310 if (!who)
311 uid = cred->uid;
312 else if (!uid_eq(uid, cred->uid)) {
313 user = find_user(uid);
314 if (!user)
315 goto out_unlock; /* No processes for this user */
316 }
317 do_each_thread(g, p) {
318 if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) {
319 niceval = nice_to_rlimit(task_nice(p));
320 if (niceval > retval)
321 retval = niceval;
322 }
323 } while_each_thread(g, p);
324 if (!uid_eq(uid, cred->uid))
325 free_uid(user); /* for find_user() */
326 break;
327 }
328 out_unlock:
329 read_unlock(&tasklist_lock);
330 rcu_read_unlock();
331
332 return retval;
333 }
334
335 /*
336 * Unprivileged users may change the real gid to the effective gid
337 * or vice versa. (BSD-style)
338 *
339 * If you set the real gid at all, or set the effective gid to a value not
340 * equal to the real gid, then the saved gid is set to the new effective gid.
341 *
342 * This makes it possible for a setgid program to completely drop its
343 * privileges, which is often a useful assertion to make when you are doing
344 * a security audit over a program.
345 *
346 * The general idea is that a program which uses just setregid() will be
347 * 100% compatible with BSD. A program which uses just setgid() will be
348 * 100% compatible with POSIX with saved IDs.
349 *
350 * SMP: There are not races, the GIDs are checked only by filesystem
351 * operations (as far as semantic preservation is concerned).
352 */
353 #ifdef CONFIG_MULTIUSER
__sys_setregid(gid_t rgid,gid_t egid)354 long __sys_setregid(gid_t rgid, gid_t egid)
355 {
356 struct user_namespace *ns = current_user_ns();
357 const struct cred *old;
358 struct cred *new;
359 int retval;
360 kgid_t krgid, kegid;
361
362 krgid = make_kgid(ns, rgid);
363 kegid = make_kgid(ns, egid);
364
365 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
366 return -EINVAL;
367 if ((egid != (gid_t) -1) && !gid_valid(kegid))
368 return -EINVAL;
369
370 new = prepare_creds();
371 if (!new)
372 return -ENOMEM;
373 old = current_cred();
374
375 retval = -EPERM;
376 if (rgid != (gid_t) -1) {
377 if (gid_eq(old->gid, krgid) ||
378 gid_eq(old->egid, krgid) ||
379 ns_capable(old->user_ns, CAP_SETGID))
380 new->gid = krgid;
381 else
382 goto error;
383 }
384 if (egid != (gid_t) -1) {
385 if (gid_eq(old->gid, kegid) ||
386 gid_eq(old->egid, kegid) ||
387 gid_eq(old->sgid, kegid) ||
388 ns_capable(old->user_ns, CAP_SETGID))
389 new->egid = kegid;
390 else
391 goto error;
392 }
393
394 if (rgid != (gid_t) -1 ||
395 (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
396 new->sgid = new->egid;
397 new->fsgid = new->egid;
398
399 return commit_creds(new);
400
401 error:
402 abort_creds(new);
403 return retval;
404 }
405
SYSCALL_DEFINE2(setregid,gid_t,rgid,gid_t,egid)406 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
407 {
408 return __sys_setregid(rgid, egid);
409 }
410
411 /*
412 * setgid() is implemented like SysV w/ SAVED_IDS
413 *
414 * SMP: Same implicit races as above.
415 */
__sys_setgid(gid_t gid)416 long __sys_setgid(gid_t gid)
417 {
418 struct user_namespace *ns = current_user_ns();
419 const struct cred *old;
420 struct cred *new;
421 int retval;
422 kgid_t kgid;
423
424 kgid = make_kgid(ns, gid);
425 if (!gid_valid(kgid))
426 return -EINVAL;
427
428 new = prepare_creds();
429 if (!new)
430 return -ENOMEM;
431 old = current_cred();
432
433 retval = -EPERM;
434 if (ns_capable(old->user_ns, CAP_SETGID))
435 new->gid = new->egid = new->sgid = new->fsgid = kgid;
436 else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
437 new->egid = new->fsgid = kgid;
438 else
439 goto error;
440
441 return commit_creds(new);
442
443 error:
444 abort_creds(new);
445 return retval;
446 }
447
SYSCALL_DEFINE1(setgid,gid_t,gid)448 SYSCALL_DEFINE1(setgid, gid_t, gid)
449 {
450 return __sys_setgid(gid);
451 }
452
453 /*
454 * change the user struct in a credentials set to match the new UID
455 */
set_user(struct cred * new)456 static int set_user(struct cred *new)
457 {
458 struct user_struct *new_user;
459
460 new_user = alloc_uid(new->uid);
461 if (!new_user)
462 return -EAGAIN;
463
464 /*
465 * We don't fail in case of NPROC limit excess here because too many
466 * poorly written programs don't check set*uid() return code, assuming
467 * it never fails if called by root. We may still enforce NPROC limit
468 * for programs doing set*uid()+execve() by harmlessly deferring the
469 * failure to the execve() stage.
470 */
471 if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
472 new_user != INIT_USER)
473 current->flags |= PF_NPROC_EXCEEDED;
474 else
475 current->flags &= ~PF_NPROC_EXCEEDED;
476
477 free_uid(new->user);
478 new->user = new_user;
479 return 0;
480 }
481
482 /*
483 * Unprivileged users may change the real uid to the effective uid
484 * or vice versa. (BSD-style)
485 *
486 * If you set the real uid at all, or set the effective uid to a value not
487 * equal to the real uid, then the saved uid is set to the new effective uid.
488 *
489 * This makes it possible for a setuid program to completely drop its
490 * privileges, which is often a useful assertion to make when you are doing
491 * a security audit over a program.
492 *
493 * The general idea is that a program which uses just setreuid() will be
494 * 100% compatible with BSD. A program which uses just setuid() will be
495 * 100% compatible with POSIX with saved IDs.
496 */
__sys_setreuid(uid_t ruid,uid_t euid)497 long __sys_setreuid(uid_t ruid, uid_t euid)
498 {
499 struct user_namespace *ns = current_user_ns();
500 const struct cred *old;
501 struct cred *new;
502 int retval;
503 kuid_t kruid, keuid;
504
505 kruid = make_kuid(ns, ruid);
506 keuid = make_kuid(ns, euid);
507
508 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
509 return -EINVAL;
510 if ((euid != (uid_t) -1) && !uid_valid(keuid))
511 return -EINVAL;
512
513 new = prepare_creds();
514 if (!new)
515 return -ENOMEM;
516 old = current_cred();
517
518 retval = -EPERM;
519 if (ruid != (uid_t) -1) {
520 new->uid = kruid;
521 if (!uid_eq(old->uid, kruid) &&
522 !uid_eq(old->euid, kruid) &&
523 !ns_capable_setid(old->user_ns, CAP_SETUID))
524 goto error;
525 }
526
527 if (euid != (uid_t) -1) {
528 new->euid = keuid;
529 if (!uid_eq(old->uid, keuid) &&
530 !uid_eq(old->euid, keuid) &&
531 !uid_eq(old->suid, keuid) &&
532 !ns_capable_setid(old->user_ns, CAP_SETUID))
533 goto error;
534 }
535
536 if (!uid_eq(new->uid, old->uid)) {
537 retval = set_user(new);
538 if (retval < 0)
539 goto error;
540 }
541 if (ruid != (uid_t) -1 ||
542 (euid != (uid_t) -1 && !uid_eq(keuid, old->uid)))
543 new->suid = new->euid;
544 new->fsuid = new->euid;
545
546 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
547 if (retval < 0)
548 goto error;
549
550 return commit_creds(new);
551
552 error:
553 abort_creds(new);
554 return retval;
555 }
556
SYSCALL_DEFINE2(setreuid,uid_t,ruid,uid_t,euid)557 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
558 {
559 return __sys_setreuid(ruid, euid);
560 }
561
562 /*
563 * setuid() is implemented like SysV with SAVED_IDS
564 *
565 * Note that SAVED_ID's is deficient in that a setuid root program
566 * like sendmail, for example, cannot set its uid to be a normal
567 * user and then switch back, because if you're root, setuid() sets
568 * the saved uid too. If you don't like this, blame the bright people
569 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
570 * will allow a root program to temporarily drop privileges and be able to
571 * regain them by swapping the real and effective uid.
572 */
__sys_setuid(uid_t uid)573 long __sys_setuid(uid_t uid)
574 {
575 struct user_namespace *ns = current_user_ns();
576 const struct cred *old;
577 struct cred *new;
578 int retval;
579 kuid_t kuid;
580
581 kuid = make_kuid(ns, uid);
582 if (!uid_valid(kuid))
583 return -EINVAL;
584
585 new = prepare_creds();
586 if (!new)
587 return -ENOMEM;
588 old = current_cred();
589
590 retval = -EPERM;
591 if (ns_capable_setid(old->user_ns, CAP_SETUID)) {
592 new->suid = new->uid = kuid;
593 if (!uid_eq(kuid, old->uid)) {
594 retval = set_user(new);
595 if (retval < 0)
596 goto error;
597 }
598 } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) {
599 goto error;
600 }
601
602 new->fsuid = new->euid = kuid;
603
604 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
605 if (retval < 0)
606 goto error;
607
608 return commit_creds(new);
609
610 error:
611 abort_creds(new);
612 return retval;
613 }
614
SYSCALL_DEFINE1(setuid,uid_t,uid)615 SYSCALL_DEFINE1(setuid, uid_t, uid)
616 {
617 return __sys_setuid(uid);
618 }
619
620
621 /*
622 * This function implements a generic ability to update ruid, euid,
623 * and suid. This allows you to implement the 4.4 compatible seteuid().
624 */
__sys_setresuid(uid_t ruid,uid_t euid,uid_t suid)625 long __sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
626 {
627 struct user_namespace *ns = current_user_ns();
628 const struct cred *old;
629 struct cred *new;
630 int retval;
631 kuid_t kruid, keuid, ksuid;
632
633 kruid = make_kuid(ns, ruid);
634 keuid = make_kuid(ns, euid);
635 ksuid = make_kuid(ns, suid);
636
637 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
638 return -EINVAL;
639
640 if ((euid != (uid_t) -1) && !uid_valid(keuid))
641 return -EINVAL;
642
643 if ((suid != (uid_t) -1) && !uid_valid(ksuid))
644 return -EINVAL;
645
646 new = prepare_creds();
647 if (!new)
648 return -ENOMEM;
649
650 old = current_cred();
651
652 retval = -EPERM;
653 if (!ns_capable_setid(old->user_ns, CAP_SETUID)) {
654 if (ruid != (uid_t) -1 && !uid_eq(kruid, old->uid) &&
655 !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid))
656 goto error;
657 if (euid != (uid_t) -1 && !uid_eq(keuid, old->uid) &&
658 !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid))
659 goto error;
660 if (suid != (uid_t) -1 && !uid_eq(ksuid, old->uid) &&
661 !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid))
662 goto error;
663 }
664
665 if (ruid != (uid_t) -1) {
666 new->uid = kruid;
667 if (!uid_eq(kruid, old->uid)) {
668 retval = set_user(new);
669 if (retval < 0)
670 goto error;
671 }
672 }
673 if (euid != (uid_t) -1)
674 new->euid = keuid;
675 if (suid != (uid_t) -1)
676 new->suid = ksuid;
677 new->fsuid = new->euid;
678
679 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
680 if (retval < 0)
681 goto error;
682
683 return commit_creds(new);
684
685 error:
686 abort_creds(new);
687 return retval;
688 }
689
SYSCALL_DEFINE3(setresuid,uid_t,ruid,uid_t,euid,uid_t,suid)690 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
691 {
692 return __sys_setresuid(ruid, euid, suid);
693 }
694
SYSCALL_DEFINE3(getresuid,uid_t __user *,ruidp,uid_t __user *,euidp,uid_t __user *,suidp)695 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
696 {
697 const struct cred *cred = current_cred();
698 int retval;
699 uid_t ruid, euid, suid;
700
701 ruid = from_kuid_munged(cred->user_ns, cred->uid);
702 euid = from_kuid_munged(cred->user_ns, cred->euid);
703 suid = from_kuid_munged(cred->user_ns, cred->suid);
704
705 retval = put_user(ruid, ruidp);
706 if (!retval) {
707 retval = put_user(euid, euidp);
708 if (!retval)
709 return put_user(suid, suidp);
710 }
711 return retval;
712 }
713
714 /*
715 * Same as above, but for rgid, egid, sgid.
716 */
__sys_setresgid(gid_t rgid,gid_t egid,gid_t sgid)717 long __sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
718 {
719 struct user_namespace *ns = current_user_ns();
720 const struct cred *old;
721 struct cred *new;
722 int retval;
723 kgid_t krgid, kegid, ksgid;
724
725 krgid = make_kgid(ns, rgid);
726 kegid = make_kgid(ns, egid);
727 ksgid = make_kgid(ns, sgid);
728
729 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
730 return -EINVAL;
731 if ((egid != (gid_t) -1) && !gid_valid(kegid))
732 return -EINVAL;
733 if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
734 return -EINVAL;
735
736 new = prepare_creds();
737 if (!new)
738 return -ENOMEM;
739 old = current_cred();
740
741 retval = -EPERM;
742 if (!ns_capable(old->user_ns, CAP_SETGID)) {
743 if (rgid != (gid_t) -1 && !gid_eq(krgid, old->gid) &&
744 !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid))
745 goto error;
746 if (egid != (gid_t) -1 && !gid_eq(kegid, old->gid) &&
747 !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid))
748 goto error;
749 if (sgid != (gid_t) -1 && !gid_eq(ksgid, old->gid) &&
750 !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid))
751 goto error;
752 }
753
754 if (rgid != (gid_t) -1)
755 new->gid = krgid;
756 if (egid != (gid_t) -1)
757 new->egid = kegid;
758 if (sgid != (gid_t) -1)
759 new->sgid = ksgid;
760 new->fsgid = new->egid;
761
762 return commit_creds(new);
763
764 error:
765 abort_creds(new);
766 return retval;
767 }
768
SYSCALL_DEFINE3(setresgid,gid_t,rgid,gid_t,egid,gid_t,sgid)769 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
770 {
771 return __sys_setresgid(rgid, egid, sgid);
772 }
773
SYSCALL_DEFINE3(getresgid,gid_t __user *,rgidp,gid_t __user *,egidp,gid_t __user *,sgidp)774 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
775 {
776 const struct cred *cred = current_cred();
777 int retval;
778 gid_t rgid, egid, sgid;
779
780 rgid = from_kgid_munged(cred->user_ns, cred->gid);
781 egid = from_kgid_munged(cred->user_ns, cred->egid);
782 sgid = from_kgid_munged(cred->user_ns, cred->sgid);
783
784 retval = put_user(rgid, rgidp);
785 if (!retval) {
786 retval = put_user(egid, egidp);
787 if (!retval)
788 retval = put_user(sgid, sgidp);
789 }
790
791 return retval;
792 }
793
794
795 /*
796 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
797 * is used for "access()" and for the NFS daemon (letting nfsd stay at
798 * whatever uid it wants to). It normally shadows "euid", except when
799 * explicitly set by setfsuid() or for access..
800 */
__sys_setfsuid(uid_t uid)801 long __sys_setfsuid(uid_t uid)
802 {
803 const struct cred *old;
804 struct cred *new;
805 uid_t old_fsuid;
806 kuid_t kuid;
807
808 old = current_cred();
809 old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
810
811 kuid = make_kuid(old->user_ns, uid);
812 if (!uid_valid(kuid))
813 return old_fsuid;
814
815 new = prepare_creds();
816 if (!new)
817 return old_fsuid;
818
819 if (uid_eq(kuid, old->uid) || uid_eq(kuid, old->euid) ||
820 uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
821 ns_capable_setid(old->user_ns, CAP_SETUID)) {
822 if (!uid_eq(kuid, old->fsuid)) {
823 new->fsuid = kuid;
824 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
825 goto change_okay;
826 }
827 }
828
829 abort_creds(new);
830 return old_fsuid;
831
832 change_okay:
833 commit_creds(new);
834 return old_fsuid;
835 }
836
SYSCALL_DEFINE1(setfsuid,uid_t,uid)837 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
838 {
839 return __sys_setfsuid(uid);
840 }
841
842 /*
843 * Samma på svenska..
844 */
__sys_setfsgid(gid_t gid)845 long __sys_setfsgid(gid_t gid)
846 {
847 const struct cred *old;
848 struct cred *new;
849 gid_t old_fsgid;
850 kgid_t kgid;
851
852 old = current_cred();
853 old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
854
855 kgid = make_kgid(old->user_ns, gid);
856 if (!gid_valid(kgid))
857 return old_fsgid;
858
859 new = prepare_creds();
860 if (!new)
861 return old_fsgid;
862
863 if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->egid) ||
864 gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
865 ns_capable(old->user_ns, CAP_SETGID)) {
866 if (!gid_eq(kgid, old->fsgid)) {
867 new->fsgid = kgid;
868 goto change_okay;
869 }
870 }
871
872 abort_creds(new);
873 return old_fsgid;
874
875 change_okay:
876 commit_creds(new);
877 return old_fsgid;
878 }
879
SYSCALL_DEFINE1(setfsgid,gid_t,gid)880 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
881 {
882 return __sys_setfsgid(gid);
883 }
884 #endif /* CONFIG_MULTIUSER */
885
886 /**
887 * sys_getpid - return the thread group id of the current process
888 *
889 * Note, despite the name, this returns the tgid not the pid. The tgid and
890 * the pid are identical unless CLONE_THREAD was specified on clone() in
891 * which case the tgid is the same in all threads of the same group.
892 *
893 * This is SMP safe as current->tgid does not change.
894 */
SYSCALL_DEFINE0(getpid)895 SYSCALL_DEFINE0(getpid)
896 {
897 return task_tgid_vnr(current);
898 }
899
900 /* Thread ID - the internal kernel "pid" */
SYSCALL_DEFINE0(gettid)901 SYSCALL_DEFINE0(gettid)
902 {
903 return task_pid_vnr(current);
904 }
905
906 /*
907 * Accessing ->real_parent is not SMP-safe, it could
908 * change from under us. However, we can use a stale
909 * value of ->real_parent under rcu_read_lock(), see
910 * release_task()->call_rcu(delayed_put_task_struct).
911 */
SYSCALL_DEFINE0(getppid)912 SYSCALL_DEFINE0(getppid)
913 {
914 int pid;
915
916 rcu_read_lock();
917 pid = task_tgid_vnr(rcu_dereference(current->real_parent));
918 rcu_read_unlock();
919
920 return pid;
921 }
922
SYSCALL_DEFINE0(getuid)923 SYSCALL_DEFINE0(getuid)
924 {
925 /* Only we change this so SMP safe */
926 return from_kuid_munged(current_user_ns(), current_uid());
927 }
928
SYSCALL_DEFINE0(geteuid)929 SYSCALL_DEFINE0(geteuid)
930 {
931 /* Only we change this so SMP safe */
932 return from_kuid_munged(current_user_ns(), current_euid());
933 }
934
SYSCALL_DEFINE0(getgid)935 SYSCALL_DEFINE0(getgid)
936 {
937 /* Only we change this so SMP safe */
938 return from_kgid_munged(current_user_ns(), current_gid());
939 }
940
SYSCALL_DEFINE0(getegid)941 SYSCALL_DEFINE0(getegid)
942 {
943 /* Only we change this so SMP safe */
944 return from_kgid_munged(current_user_ns(), current_egid());
945 }
946
do_sys_times(struct tms * tms)947 static void do_sys_times(struct tms *tms)
948 {
949 u64 tgutime, tgstime, cutime, cstime;
950
951 thread_group_cputime_adjusted(current, &tgutime, &tgstime);
952 cutime = current->signal->cutime;
953 cstime = current->signal->cstime;
954 tms->tms_utime = nsec_to_clock_t(tgutime);
955 tms->tms_stime = nsec_to_clock_t(tgstime);
956 tms->tms_cutime = nsec_to_clock_t(cutime);
957 tms->tms_cstime = nsec_to_clock_t(cstime);
958 }
959
SYSCALL_DEFINE1(times,struct tms __user *,tbuf)960 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
961 {
962 if (tbuf) {
963 struct tms tmp;
964
965 do_sys_times(&tmp);
966 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
967 return -EFAULT;
968 }
969 force_successful_syscall_return();
970 return (long) jiffies_64_to_clock_t(get_jiffies_64());
971 }
972
973 #ifdef CONFIG_COMPAT
clock_t_to_compat_clock_t(clock_t x)974 static compat_clock_t clock_t_to_compat_clock_t(clock_t x)
975 {
976 return compat_jiffies_to_clock_t(clock_t_to_jiffies(x));
977 }
978
COMPAT_SYSCALL_DEFINE1(times,struct compat_tms __user *,tbuf)979 COMPAT_SYSCALL_DEFINE1(times, struct compat_tms __user *, tbuf)
980 {
981 if (tbuf) {
982 struct tms tms;
983 struct compat_tms tmp;
984
985 do_sys_times(&tms);
986 /* Convert our struct tms to the compat version. */
987 tmp.tms_utime = clock_t_to_compat_clock_t(tms.tms_utime);
988 tmp.tms_stime = clock_t_to_compat_clock_t(tms.tms_stime);
989 tmp.tms_cutime = clock_t_to_compat_clock_t(tms.tms_cutime);
990 tmp.tms_cstime = clock_t_to_compat_clock_t(tms.tms_cstime);
991 if (copy_to_user(tbuf, &tmp, sizeof(tmp)))
992 return -EFAULT;
993 }
994 force_successful_syscall_return();
995 return compat_jiffies_to_clock_t(jiffies);
996 }
997 #endif
998
999 /*
1000 * This needs some heavy checking ...
1001 * I just haven't the stomach for it. I also don't fully
1002 * understand sessions/pgrp etc. Let somebody who does explain it.
1003 *
1004 * OK, I think I have the protection semantics right.... this is really
1005 * only important on a multi-user system anyway, to make sure one user
1006 * can't send a signal to a process owned by another. -TYT, 12/12/91
1007 *
1008 * !PF_FORKNOEXEC check to conform completely to POSIX.
1009 */
SYSCALL_DEFINE2(setpgid,pid_t,pid,pid_t,pgid)1010 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
1011 {
1012 struct task_struct *p;
1013 struct task_struct *group_leader = current->group_leader;
1014 struct pid *pgrp;
1015 int err;
1016
1017 if (!pid)
1018 pid = task_pid_vnr(group_leader);
1019 if (!pgid)
1020 pgid = pid;
1021 if (pgid < 0)
1022 return -EINVAL;
1023 rcu_read_lock();
1024
1025 /* From this point forward we keep holding onto the tasklist lock
1026 * so that our parent does not change from under us. -DaveM
1027 */
1028 write_lock_irq(&tasklist_lock);
1029
1030 err = -ESRCH;
1031 p = find_task_by_vpid(pid);
1032 if (!p)
1033 goto out;
1034
1035 err = -EINVAL;
1036 if (!thread_group_leader(p))
1037 goto out;
1038
1039 if (same_thread_group(p->real_parent, group_leader)) {
1040 err = -EPERM;
1041 if (task_session(p) != task_session(group_leader))
1042 goto out;
1043 err = -EACCES;
1044 if (!(p->flags & PF_FORKNOEXEC))
1045 goto out;
1046 } else {
1047 err = -ESRCH;
1048 if (p != group_leader)
1049 goto out;
1050 }
1051
1052 err = -EPERM;
1053 if (p->signal->leader)
1054 goto out;
1055
1056 pgrp = task_pid(p);
1057 if (pgid != pid) {
1058 struct task_struct *g;
1059
1060 pgrp = find_vpid(pgid);
1061 g = pid_task(pgrp, PIDTYPE_PGID);
1062 if (!g || task_session(g) != task_session(group_leader))
1063 goto out;
1064 }
1065
1066 err = security_task_setpgid(p, pgid);
1067 if (err)
1068 goto out;
1069
1070 if (task_pgrp(p) != pgrp)
1071 change_pid(p, PIDTYPE_PGID, pgrp);
1072
1073 err = 0;
1074 out:
1075 /* All paths lead to here, thus we are safe. -DaveM */
1076 write_unlock_irq(&tasklist_lock);
1077 rcu_read_unlock();
1078 return err;
1079 }
1080
do_getpgid(pid_t pid)1081 static int do_getpgid(pid_t pid)
1082 {
1083 struct task_struct *p;
1084 struct pid *grp;
1085 int retval;
1086
1087 rcu_read_lock();
1088 if (!pid)
1089 grp = task_pgrp(current);
1090 else {
1091 retval = -ESRCH;
1092 p = find_task_by_vpid(pid);
1093 if (!p)
1094 goto out;
1095 grp = task_pgrp(p);
1096 if (!grp)
1097 goto out;
1098
1099 retval = security_task_getpgid(p);
1100 if (retval)
1101 goto out;
1102 }
1103 retval = pid_vnr(grp);
1104 out:
1105 rcu_read_unlock();
1106 return retval;
1107 }
1108
SYSCALL_DEFINE1(getpgid,pid_t,pid)1109 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1110 {
1111 return do_getpgid(pid);
1112 }
1113
1114 #ifdef __ARCH_WANT_SYS_GETPGRP
1115
SYSCALL_DEFINE0(getpgrp)1116 SYSCALL_DEFINE0(getpgrp)
1117 {
1118 return do_getpgid(0);
1119 }
1120
1121 #endif
1122
SYSCALL_DEFINE1(getsid,pid_t,pid)1123 SYSCALL_DEFINE1(getsid, pid_t, pid)
1124 {
1125 struct task_struct *p;
1126 struct pid *sid;
1127 int retval;
1128
1129 rcu_read_lock();
1130 if (!pid)
1131 sid = task_session(current);
1132 else {
1133 retval = -ESRCH;
1134 p = find_task_by_vpid(pid);
1135 if (!p)
1136 goto out;
1137 sid = task_session(p);
1138 if (!sid)
1139 goto out;
1140
1141 retval = security_task_getsid(p);
1142 if (retval)
1143 goto out;
1144 }
1145 retval = pid_vnr(sid);
1146 out:
1147 rcu_read_unlock();
1148 return retval;
1149 }
1150
set_special_pids(struct pid * pid)1151 static void set_special_pids(struct pid *pid)
1152 {
1153 struct task_struct *curr = current->group_leader;
1154
1155 if (task_session(curr) != pid)
1156 change_pid(curr, PIDTYPE_SID, pid);
1157
1158 if (task_pgrp(curr) != pid)
1159 change_pid(curr, PIDTYPE_PGID, pid);
1160 }
1161
ksys_setsid(void)1162 int ksys_setsid(void)
1163 {
1164 struct task_struct *group_leader = current->group_leader;
1165 struct pid *sid = task_pid(group_leader);
1166 pid_t session = pid_vnr(sid);
1167 int err = -EPERM;
1168
1169 write_lock_irq(&tasklist_lock);
1170 /* Fail if I am already a session leader */
1171 if (group_leader->signal->leader)
1172 goto out;
1173
1174 /* Fail if a process group id already exists that equals the
1175 * proposed session id.
1176 */
1177 if (pid_task(sid, PIDTYPE_PGID))
1178 goto out;
1179
1180 group_leader->signal->leader = 1;
1181 set_special_pids(sid);
1182
1183 proc_clear_tty(group_leader);
1184
1185 err = session;
1186 out:
1187 write_unlock_irq(&tasklist_lock);
1188 if (err > 0) {
1189 proc_sid_connector(group_leader);
1190 sched_autogroup_create_attach(group_leader);
1191 }
1192 return err;
1193 }
1194
SYSCALL_DEFINE0(setsid)1195 SYSCALL_DEFINE0(setsid)
1196 {
1197 return ksys_setsid();
1198 }
1199
1200 DECLARE_RWSEM(uts_sem);
1201
1202 #ifdef COMPAT_UTS_MACHINE
1203 #define override_architecture(name) \
1204 (personality(current->personality) == PER_LINUX32 && \
1205 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1206 sizeof(COMPAT_UTS_MACHINE)))
1207 #else
1208 #define override_architecture(name) 0
1209 #endif
1210
1211 /*
1212 * Work around broken programs that cannot handle "Linux 3.0".
1213 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1214 * And we map 4.x and later versions to 2.6.60+x, so 4.0/5.0/6.0/... would be
1215 * 2.6.60.
1216 */
override_release(char __user * release,size_t len)1217 static int override_release(char __user *release, size_t len)
1218 {
1219 int ret = 0;
1220
1221 if (current->personality & UNAME26) {
1222 const char *rest = UTS_RELEASE;
1223 char buf[65] = { 0 };
1224 int ndots = 0;
1225 unsigned v;
1226 size_t copy;
1227
1228 while (*rest) {
1229 if (*rest == '.' && ++ndots >= 3)
1230 break;
1231 if (!isdigit(*rest) && *rest != '.')
1232 break;
1233 rest++;
1234 }
1235 v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 60;
1236 copy = clamp_t(size_t, len, 1, sizeof(buf));
1237 copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1238 ret = copy_to_user(release, buf, copy + 1);
1239 }
1240 return ret;
1241 }
1242
SYSCALL_DEFINE1(newuname,struct new_utsname __user *,name)1243 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1244 {
1245 struct new_utsname tmp;
1246
1247 down_read(&uts_sem);
1248 memcpy(&tmp, utsname(), sizeof(tmp));
1249 up_read(&uts_sem);
1250 if (copy_to_user(name, &tmp, sizeof(tmp)))
1251 return -EFAULT;
1252
1253 if (override_release(name->release, sizeof(name->release)))
1254 return -EFAULT;
1255 if (override_architecture(name))
1256 return -EFAULT;
1257 return 0;
1258 }
1259
1260 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1261 /*
1262 * Old cruft
1263 */
SYSCALL_DEFINE1(uname,struct old_utsname __user *,name)1264 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1265 {
1266 struct old_utsname tmp;
1267
1268 if (!name)
1269 return -EFAULT;
1270
1271 down_read(&uts_sem);
1272 memcpy(&tmp, utsname(), sizeof(tmp));
1273 up_read(&uts_sem);
1274 if (copy_to_user(name, &tmp, sizeof(tmp)))
1275 return -EFAULT;
1276
1277 if (override_release(name->release, sizeof(name->release)))
1278 return -EFAULT;
1279 if (override_architecture(name))
1280 return -EFAULT;
1281 return 0;
1282 }
1283
SYSCALL_DEFINE1(olduname,struct oldold_utsname __user *,name)1284 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1285 {
1286 struct oldold_utsname tmp;
1287
1288 if (!name)
1289 return -EFAULT;
1290
1291 memset(&tmp, 0, sizeof(tmp));
1292
1293 down_read(&uts_sem);
1294 memcpy(&tmp.sysname, &utsname()->sysname, __OLD_UTS_LEN);
1295 memcpy(&tmp.nodename, &utsname()->nodename, __OLD_UTS_LEN);
1296 memcpy(&tmp.release, &utsname()->release, __OLD_UTS_LEN);
1297 memcpy(&tmp.version, &utsname()->version, __OLD_UTS_LEN);
1298 memcpy(&tmp.machine, &utsname()->machine, __OLD_UTS_LEN);
1299 up_read(&uts_sem);
1300 if (copy_to_user(name, &tmp, sizeof(tmp)))
1301 return -EFAULT;
1302
1303 if (override_architecture(name))
1304 return -EFAULT;
1305 if (override_release(name->release, sizeof(name->release)))
1306 return -EFAULT;
1307 return 0;
1308 }
1309 #endif
1310
SYSCALL_DEFINE2(sethostname,char __user *,name,int,len)1311 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1312 {
1313 int errno;
1314 char tmp[__NEW_UTS_LEN];
1315
1316 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1317 return -EPERM;
1318
1319 if (len < 0 || len > __NEW_UTS_LEN)
1320 return -EINVAL;
1321 errno = -EFAULT;
1322 if (!copy_from_user(tmp, name, len)) {
1323 struct new_utsname *u;
1324
1325 down_write(&uts_sem);
1326 u = utsname();
1327 memcpy(u->nodename, tmp, len);
1328 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1329 errno = 0;
1330 uts_proc_notify(UTS_PROC_HOSTNAME);
1331 up_write(&uts_sem);
1332 }
1333 return errno;
1334 }
1335
1336 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1337
SYSCALL_DEFINE2(gethostname,char __user *,name,int,len)1338 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1339 {
1340 int i;
1341 struct new_utsname *u;
1342 char tmp[__NEW_UTS_LEN + 1];
1343
1344 if (len < 0)
1345 return -EINVAL;
1346 down_read(&uts_sem);
1347 u = utsname();
1348 i = 1 + strlen(u->nodename);
1349 if (i > len)
1350 i = len;
1351 memcpy(tmp, u->nodename, i);
1352 up_read(&uts_sem);
1353 if (copy_to_user(name, tmp, i))
1354 return -EFAULT;
1355 return 0;
1356 }
1357
1358 #endif
1359
1360 /*
1361 * Only setdomainname; getdomainname can be implemented by calling
1362 * uname()
1363 */
SYSCALL_DEFINE2(setdomainname,char __user *,name,int,len)1364 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1365 {
1366 int errno;
1367 char tmp[__NEW_UTS_LEN];
1368
1369 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1370 return -EPERM;
1371 if (len < 0 || len > __NEW_UTS_LEN)
1372 return -EINVAL;
1373
1374 errno = -EFAULT;
1375 if (!copy_from_user(tmp, name, len)) {
1376 struct new_utsname *u;
1377
1378 down_write(&uts_sem);
1379 u = utsname();
1380 memcpy(u->domainname, tmp, len);
1381 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1382 errno = 0;
1383 uts_proc_notify(UTS_PROC_DOMAINNAME);
1384 up_write(&uts_sem);
1385 }
1386 return errno;
1387 }
1388
SYSCALL_DEFINE2(getrlimit,unsigned int,resource,struct rlimit __user *,rlim)1389 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1390 {
1391 struct rlimit value;
1392 int ret;
1393
1394 ret = do_prlimit(current, resource, NULL, &value);
1395 if (!ret)
1396 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1397
1398 return ret;
1399 }
1400
1401 #ifdef CONFIG_COMPAT
1402
COMPAT_SYSCALL_DEFINE2(setrlimit,unsigned int,resource,struct compat_rlimit __user *,rlim)1403 COMPAT_SYSCALL_DEFINE2(setrlimit, unsigned int, resource,
1404 struct compat_rlimit __user *, rlim)
1405 {
1406 struct rlimit r;
1407 struct compat_rlimit r32;
1408
1409 if (copy_from_user(&r32, rlim, sizeof(struct compat_rlimit)))
1410 return -EFAULT;
1411
1412 if (r32.rlim_cur == COMPAT_RLIM_INFINITY)
1413 r.rlim_cur = RLIM_INFINITY;
1414 else
1415 r.rlim_cur = r32.rlim_cur;
1416 if (r32.rlim_max == COMPAT_RLIM_INFINITY)
1417 r.rlim_max = RLIM_INFINITY;
1418 else
1419 r.rlim_max = r32.rlim_max;
1420 return do_prlimit(current, resource, &r, NULL);
1421 }
1422
COMPAT_SYSCALL_DEFINE2(getrlimit,unsigned int,resource,struct compat_rlimit __user *,rlim)1423 COMPAT_SYSCALL_DEFINE2(getrlimit, unsigned int, resource,
1424 struct compat_rlimit __user *, rlim)
1425 {
1426 struct rlimit r;
1427 int ret;
1428
1429 ret = do_prlimit(current, resource, NULL, &r);
1430 if (!ret) {
1431 struct compat_rlimit r32;
1432 if (r.rlim_cur > COMPAT_RLIM_INFINITY)
1433 r32.rlim_cur = COMPAT_RLIM_INFINITY;
1434 else
1435 r32.rlim_cur = r.rlim_cur;
1436 if (r.rlim_max > COMPAT_RLIM_INFINITY)
1437 r32.rlim_max = COMPAT_RLIM_INFINITY;
1438 else
1439 r32.rlim_max = r.rlim_max;
1440
1441 if (copy_to_user(rlim, &r32, sizeof(struct compat_rlimit)))
1442 return -EFAULT;
1443 }
1444 return ret;
1445 }
1446
1447 #endif
1448
1449 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1450
1451 /*
1452 * Back compatibility for getrlimit. Needed for some apps.
1453 */
SYSCALL_DEFINE2(old_getrlimit,unsigned int,resource,struct rlimit __user *,rlim)1454 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1455 struct rlimit __user *, rlim)
1456 {
1457 struct rlimit x;
1458 if (resource >= RLIM_NLIMITS)
1459 return -EINVAL;
1460
1461 resource = array_index_nospec(resource, RLIM_NLIMITS);
1462 task_lock(current->group_leader);
1463 x = current->signal->rlim[resource];
1464 task_unlock(current->group_leader);
1465 if (x.rlim_cur > 0x7FFFFFFF)
1466 x.rlim_cur = 0x7FFFFFFF;
1467 if (x.rlim_max > 0x7FFFFFFF)
1468 x.rlim_max = 0x7FFFFFFF;
1469 return copy_to_user(rlim, &x, sizeof(x)) ? -EFAULT : 0;
1470 }
1471
1472 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE2(old_getrlimit,unsigned int,resource,struct compat_rlimit __user *,rlim)1473 COMPAT_SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1474 struct compat_rlimit __user *, rlim)
1475 {
1476 struct rlimit r;
1477
1478 if (resource >= RLIM_NLIMITS)
1479 return -EINVAL;
1480
1481 resource = array_index_nospec(resource, RLIM_NLIMITS);
1482 task_lock(current->group_leader);
1483 r = current->signal->rlim[resource];
1484 task_unlock(current->group_leader);
1485 if (r.rlim_cur > 0x7FFFFFFF)
1486 r.rlim_cur = 0x7FFFFFFF;
1487 if (r.rlim_max > 0x7FFFFFFF)
1488 r.rlim_max = 0x7FFFFFFF;
1489
1490 if (put_user(r.rlim_cur, &rlim->rlim_cur) ||
1491 put_user(r.rlim_max, &rlim->rlim_max))
1492 return -EFAULT;
1493 return 0;
1494 }
1495 #endif
1496
1497 #endif
1498
rlim64_is_infinity(__u64 rlim64)1499 static inline bool rlim64_is_infinity(__u64 rlim64)
1500 {
1501 #if BITS_PER_LONG < 64
1502 return rlim64 >= ULONG_MAX;
1503 #else
1504 return rlim64 == RLIM64_INFINITY;
1505 #endif
1506 }
1507
rlim_to_rlim64(const struct rlimit * rlim,struct rlimit64 * rlim64)1508 static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1509 {
1510 if (rlim->rlim_cur == RLIM_INFINITY)
1511 rlim64->rlim_cur = RLIM64_INFINITY;
1512 else
1513 rlim64->rlim_cur = rlim->rlim_cur;
1514 if (rlim->rlim_max == RLIM_INFINITY)
1515 rlim64->rlim_max = RLIM64_INFINITY;
1516 else
1517 rlim64->rlim_max = rlim->rlim_max;
1518 }
1519
rlim64_to_rlim(const struct rlimit64 * rlim64,struct rlimit * rlim)1520 static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1521 {
1522 if (rlim64_is_infinity(rlim64->rlim_cur))
1523 rlim->rlim_cur = RLIM_INFINITY;
1524 else
1525 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1526 if (rlim64_is_infinity(rlim64->rlim_max))
1527 rlim->rlim_max = RLIM_INFINITY;
1528 else
1529 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1530 }
1531
1532 /* make sure you are allowed to change @tsk limits before calling this */
do_prlimit(struct task_struct * tsk,unsigned int resource,struct rlimit * new_rlim,struct rlimit * old_rlim)1533 int do_prlimit(struct task_struct *tsk, unsigned int resource,
1534 struct rlimit *new_rlim, struct rlimit *old_rlim)
1535 {
1536 struct rlimit *rlim;
1537 int retval = 0;
1538
1539 if (resource >= RLIM_NLIMITS)
1540 return -EINVAL;
1541 resource = array_index_nospec(resource, RLIM_NLIMITS);
1542
1543 if (new_rlim) {
1544 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1545 return -EINVAL;
1546 if (resource == RLIMIT_NOFILE &&
1547 new_rlim->rlim_max > sysctl_nr_open)
1548 return -EPERM;
1549 }
1550
1551 /* protect tsk->signal and tsk->sighand from disappearing */
1552 read_lock(&tasklist_lock);
1553 if (!tsk->sighand) {
1554 retval = -ESRCH;
1555 goto out;
1556 }
1557
1558 rlim = tsk->signal->rlim + resource;
1559 task_lock(tsk->group_leader);
1560 if (new_rlim) {
1561 /* Keep the capable check against init_user_ns until
1562 cgroups can contain all limits */
1563 if (new_rlim->rlim_max > rlim->rlim_max &&
1564 !capable(CAP_SYS_RESOURCE))
1565 retval = -EPERM;
1566 if (!retval)
1567 retval = security_task_setrlimit(tsk, resource, new_rlim);
1568 }
1569 if (!retval) {
1570 if (old_rlim)
1571 *old_rlim = *rlim;
1572 if (new_rlim)
1573 *rlim = *new_rlim;
1574 }
1575 task_unlock(tsk->group_leader);
1576
1577 /*
1578 * RLIMIT_CPU handling. Arm the posix CPU timer if the limit is not
1579 * infite. In case of RLIM_INFINITY the posix CPU timer code
1580 * ignores the rlimit.
1581 */
1582 if (!retval && new_rlim && resource == RLIMIT_CPU &&
1583 new_rlim->rlim_cur != RLIM_INFINITY &&
1584 IS_ENABLED(CONFIG_POSIX_TIMERS))
1585 update_rlimit_cpu(tsk, new_rlim->rlim_cur);
1586 out:
1587 read_unlock(&tasklist_lock);
1588 return retval;
1589 }
1590
1591 /* rcu lock must be held */
check_prlimit_permission(struct task_struct * task,unsigned int flags)1592 static int check_prlimit_permission(struct task_struct *task,
1593 unsigned int flags)
1594 {
1595 const struct cred *cred = current_cred(), *tcred;
1596 bool id_match;
1597
1598 if (current == task)
1599 return 0;
1600
1601 tcred = __task_cred(task);
1602 id_match = (uid_eq(cred->uid, tcred->euid) &&
1603 uid_eq(cred->uid, tcred->suid) &&
1604 uid_eq(cred->uid, tcred->uid) &&
1605 gid_eq(cred->gid, tcred->egid) &&
1606 gid_eq(cred->gid, tcred->sgid) &&
1607 gid_eq(cred->gid, tcred->gid));
1608 if (!id_match && !ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
1609 return -EPERM;
1610
1611 return security_task_prlimit(cred, tcred, flags);
1612 }
1613
SYSCALL_DEFINE4(prlimit64,pid_t,pid,unsigned int,resource,const struct rlimit64 __user *,new_rlim,struct rlimit64 __user *,old_rlim)1614 SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1615 const struct rlimit64 __user *, new_rlim,
1616 struct rlimit64 __user *, old_rlim)
1617 {
1618 struct rlimit64 old64, new64;
1619 struct rlimit old, new;
1620 struct task_struct *tsk;
1621 unsigned int checkflags = 0;
1622 int ret;
1623
1624 if (old_rlim)
1625 checkflags |= LSM_PRLIMIT_READ;
1626
1627 if (new_rlim) {
1628 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1629 return -EFAULT;
1630 rlim64_to_rlim(&new64, &new);
1631 checkflags |= LSM_PRLIMIT_WRITE;
1632 }
1633
1634 rcu_read_lock();
1635 tsk = pid ? find_task_by_vpid(pid) : current;
1636 if (!tsk) {
1637 rcu_read_unlock();
1638 return -ESRCH;
1639 }
1640 ret = check_prlimit_permission(tsk, checkflags);
1641 if (ret) {
1642 rcu_read_unlock();
1643 return ret;
1644 }
1645 get_task_struct(tsk);
1646 rcu_read_unlock();
1647
1648 ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1649 old_rlim ? &old : NULL);
1650
1651 if (!ret && old_rlim) {
1652 rlim_to_rlim64(&old, &old64);
1653 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1654 ret = -EFAULT;
1655 }
1656
1657 put_task_struct(tsk);
1658 return ret;
1659 }
1660
SYSCALL_DEFINE2(setrlimit,unsigned int,resource,struct rlimit __user *,rlim)1661 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1662 {
1663 struct rlimit new_rlim;
1664
1665 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1666 return -EFAULT;
1667 return do_prlimit(current, resource, &new_rlim, NULL);
1668 }
1669
1670 /*
1671 * It would make sense to put struct rusage in the task_struct,
1672 * except that would make the task_struct be *really big*. After
1673 * task_struct gets moved into malloc'ed memory, it would
1674 * make sense to do this. It will make moving the rest of the information
1675 * a lot simpler! (Which we're not doing right now because we're not
1676 * measuring them yet).
1677 *
1678 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1679 * races with threads incrementing their own counters. But since word
1680 * reads are atomic, we either get new values or old values and we don't
1681 * care which for the sums. We always take the siglock to protect reading
1682 * the c* fields from p->signal from races with exit.c updating those
1683 * fields when reaping, so a sample either gets all the additions of a
1684 * given child after it's reaped, or none so this sample is before reaping.
1685 *
1686 * Locking:
1687 * We need to take the siglock for CHILDEREN, SELF and BOTH
1688 * for the cases current multithreaded, non-current single threaded
1689 * non-current multithreaded. Thread traversal is now safe with
1690 * the siglock held.
1691 * Strictly speaking, we donot need to take the siglock if we are current and
1692 * single threaded, as no one else can take our signal_struct away, no one
1693 * else can reap the children to update signal->c* counters, and no one else
1694 * can race with the signal-> fields. If we do not take any lock, the
1695 * signal-> fields could be read out of order while another thread was just
1696 * exiting. So we should place a read memory barrier when we avoid the lock.
1697 * On the writer side, write memory barrier is implied in __exit_signal
1698 * as __exit_signal releases the siglock spinlock after updating the signal->
1699 * fields. But we don't do this yet to keep things simple.
1700 *
1701 */
1702
accumulate_thread_rusage(struct task_struct * t,struct rusage * r)1703 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1704 {
1705 r->ru_nvcsw += t->nvcsw;
1706 r->ru_nivcsw += t->nivcsw;
1707 r->ru_minflt += t->min_flt;
1708 r->ru_majflt += t->maj_flt;
1709 r->ru_inblock += task_io_get_inblock(t);
1710 r->ru_oublock += task_io_get_oublock(t);
1711 }
1712
getrusage(struct task_struct * p,int who,struct rusage * r)1713 void getrusage(struct task_struct *p, int who, struct rusage *r)
1714 {
1715 struct task_struct *t;
1716 unsigned long flags;
1717 u64 tgutime, tgstime, utime, stime;
1718 unsigned long maxrss;
1719 struct mm_struct *mm;
1720 struct signal_struct *sig = p->signal;
1721 unsigned int seq = 0;
1722
1723 retry:
1724 memset(r, 0, sizeof(*r));
1725 utime = stime = 0;
1726 maxrss = 0;
1727
1728 if (who == RUSAGE_THREAD) {
1729 task_cputime_adjusted(current, &utime, &stime);
1730 accumulate_thread_rusage(p, r);
1731 maxrss = sig->maxrss;
1732 goto out_thread;
1733 }
1734
1735 flags = read_seqbegin_or_lock_irqsave(&sig->stats_lock, &seq);
1736
1737 switch (who) {
1738 case RUSAGE_BOTH:
1739 case RUSAGE_CHILDREN:
1740 utime = sig->cutime;
1741 stime = sig->cstime;
1742 r->ru_nvcsw = sig->cnvcsw;
1743 r->ru_nivcsw = sig->cnivcsw;
1744 r->ru_minflt = sig->cmin_flt;
1745 r->ru_majflt = sig->cmaj_flt;
1746 r->ru_inblock = sig->cinblock;
1747 r->ru_oublock = sig->coublock;
1748 maxrss = sig->cmaxrss;
1749
1750 if (who == RUSAGE_CHILDREN)
1751 break;
1752 /* fall through */
1753
1754 case RUSAGE_SELF:
1755 r->ru_nvcsw += sig->nvcsw;
1756 r->ru_nivcsw += sig->nivcsw;
1757 r->ru_minflt += sig->min_flt;
1758 r->ru_majflt += sig->maj_flt;
1759 r->ru_inblock += sig->inblock;
1760 r->ru_oublock += sig->oublock;
1761 if (maxrss < sig->maxrss)
1762 maxrss = sig->maxrss;
1763
1764 rcu_read_lock();
1765 __for_each_thread(sig, t)
1766 accumulate_thread_rusage(t, r);
1767 rcu_read_unlock();
1768
1769 break;
1770
1771 default:
1772 BUG();
1773 }
1774
1775 if (need_seqretry(&sig->stats_lock, seq)) {
1776 seq = 1;
1777 goto retry;
1778 }
1779 done_seqretry_irqrestore(&sig->stats_lock, seq, flags);
1780
1781 if (who == RUSAGE_CHILDREN)
1782 goto out_children;
1783
1784 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1785 utime += tgutime;
1786 stime += tgstime;
1787
1788 out_thread:
1789 mm = get_task_mm(p);
1790 if (mm) {
1791 setmax_mm_hiwater_rss(&maxrss, mm);
1792 mmput(mm);
1793 }
1794
1795 out_children:
1796 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1797 r->ru_utime = ns_to_kernel_old_timeval(utime);
1798 r->ru_stime = ns_to_kernel_old_timeval(stime);
1799 }
1800
SYSCALL_DEFINE2(getrusage,int,who,struct rusage __user *,ru)1801 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1802 {
1803 struct rusage r;
1804
1805 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1806 who != RUSAGE_THREAD)
1807 return -EINVAL;
1808
1809 getrusage(current, who, &r);
1810 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1811 }
1812
1813 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE2(getrusage,int,who,struct compat_rusage __user *,ru)1814 COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru)
1815 {
1816 struct rusage r;
1817
1818 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1819 who != RUSAGE_THREAD)
1820 return -EINVAL;
1821
1822 getrusage(current, who, &r);
1823 return put_compat_rusage(&r, ru);
1824 }
1825 #endif
1826
SYSCALL_DEFINE1(umask,int,mask)1827 SYSCALL_DEFINE1(umask, int, mask)
1828 {
1829 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1830 return mask;
1831 }
1832
prctl_set_mm_exe_file(struct mm_struct * mm,unsigned int fd)1833 static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1834 {
1835 struct fd exe;
1836 struct file *old_exe, *exe_file;
1837 struct inode *inode;
1838 int err;
1839
1840 exe = fdget(fd);
1841 if (!exe.file)
1842 return -EBADF;
1843
1844 inode = file_inode(exe.file);
1845
1846 /*
1847 * Because the original mm->exe_file points to executable file, make
1848 * sure that this one is executable as well, to avoid breaking an
1849 * overall picture.
1850 */
1851 err = -EACCES;
1852 if (!S_ISREG(inode->i_mode) || path_noexec(&exe.file->f_path))
1853 goto exit;
1854
1855 err = inode_permission(inode, MAY_EXEC);
1856 if (err)
1857 goto exit;
1858
1859 /*
1860 * Forbid mm->exe_file change if old file still mapped.
1861 */
1862 exe_file = get_mm_exe_file(mm);
1863 err = -EBUSY;
1864 if (exe_file) {
1865 struct vm_area_struct *vma;
1866
1867 down_read(&mm->mmap_sem);
1868 for (vma = mm->mmap; vma; vma = vma->vm_next) {
1869 if (!vma->vm_file)
1870 continue;
1871 if (path_equal(&vma->vm_file->f_path,
1872 &exe_file->f_path))
1873 goto exit_err;
1874 }
1875
1876 up_read(&mm->mmap_sem);
1877 fput(exe_file);
1878 }
1879
1880 err = 0;
1881 /* set the new file, lockless */
1882 get_file(exe.file);
1883 old_exe = xchg(&mm->exe_file, exe.file);
1884 if (old_exe)
1885 fput(old_exe);
1886 exit:
1887 fdput(exe);
1888 return err;
1889 exit_err:
1890 up_read(&mm->mmap_sem);
1891 fput(exe_file);
1892 goto exit;
1893 }
1894
1895 /*
1896 * Check arithmetic relations of passed addresses.
1897 *
1898 * WARNING: we don't require any capability here so be very careful
1899 * in what is allowed for modification from userspace.
1900 */
validate_prctl_map_addr(struct prctl_mm_map * prctl_map)1901 static int validate_prctl_map_addr(struct prctl_mm_map *prctl_map)
1902 {
1903 unsigned long mmap_max_addr = TASK_SIZE;
1904 int error = -EINVAL, i;
1905
1906 static const unsigned char offsets[] = {
1907 offsetof(struct prctl_mm_map, start_code),
1908 offsetof(struct prctl_mm_map, end_code),
1909 offsetof(struct prctl_mm_map, start_data),
1910 offsetof(struct prctl_mm_map, end_data),
1911 offsetof(struct prctl_mm_map, start_brk),
1912 offsetof(struct prctl_mm_map, brk),
1913 offsetof(struct prctl_mm_map, start_stack),
1914 offsetof(struct prctl_mm_map, arg_start),
1915 offsetof(struct prctl_mm_map, arg_end),
1916 offsetof(struct prctl_mm_map, env_start),
1917 offsetof(struct prctl_mm_map, env_end),
1918 };
1919
1920 /*
1921 * Make sure the members are not somewhere outside
1922 * of allowed address space.
1923 */
1924 for (i = 0; i < ARRAY_SIZE(offsets); i++) {
1925 u64 val = *(u64 *)((char *)prctl_map + offsets[i]);
1926
1927 if ((unsigned long)val >= mmap_max_addr ||
1928 (unsigned long)val < mmap_min_addr)
1929 goto out;
1930 }
1931
1932 /*
1933 * Make sure the pairs are ordered.
1934 */
1935 #define __prctl_check_order(__m1, __op, __m2) \
1936 ((unsigned long)prctl_map->__m1 __op \
1937 (unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1938 error = __prctl_check_order(start_code, <, end_code);
1939 error |= __prctl_check_order(start_data,<=, end_data);
1940 error |= __prctl_check_order(start_brk, <=, brk);
1941 error |= __prctl_check_order(arg_start, <=, arg_end);
1942 error |= __prctl_check_order(env_start, <=, env_end);
1943 if (error)
1944 goto out;
1945 #undef __prctl_check_order
1946
1947 error = -EINVAL;
1948
1949 /*
1950 * Neither we should allow to override limits if they set.
1951 */
1952 if (check_data_rlimit(rlimit(RLIMIT_DATA), prctl_map->brk,
1953 prctl_map->start_brk, prctl_map->end_data,
1954 prctl_map->start_data))
1955 goto out;
1956
1957 error = 0;
1958 out:
1959 return error;
1960 }
1961
1962 #ifdef CONFIG_CHECKPOINT_RESTORE
prctl_set_mm_map(int opt,const void __user * addr,unsigned long data_size)1963 static int prctl_set_mm_map(int opt, const void __user *addr, unsigned long data_size)
1964 {
1965 struct prctl_mm_map prctl_map = { .exe_fd = (u32)-1, };
1966 unsigned long user_auxv[AT_VECTOR_SIZE];
1967 struct mm_struct *mm = current->mm;
1968 int error;
1969
1970 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
1971 BUILD_BUG_ON(sizeof(struct prctl_mm_map) > 256);
1972
1973 if (opt == PR_SET_MM_MAP_SIZE)
1974 return put_user((unsigned int)sizeof(prctl_map),
1975 (unsigned int __user *)addr);
1976
1977 if (data_size != sizeof(prctl_map))
1978 return -EINVAL;
1979
1980 if (copy_from_user(&prctl_map, addr, sizeof(prctl_map)))
1981 return -EFAULT;
1982
1983 error = validate_prctl_map_addr(&prctl_map);
1984 if (error)
1985 return error;
1986
1987 if (prctl_map.auxv_size) {
1988 /*
1989 * Someone is trying to cheat the auxv vector.
1990 */
1991 if (!prctl_map.auxv ||
1992 prctl_map.auxv_size > sizeof(mm->saved_auxv))
1993 return -EINVAL;
1994
1995 memset(user_auxv, 0, sizeof(user_auxv));
1996 if (copy_from_user(user_auxv,
1997 (const void __user *)prctl_map.auxv,
1998 prctl_map.auxv_size))
1999 return -EFAULT;
2000
2001 /* Last entry must be AT_NULL as specification requires */
2002 user_auxv[AT_VECTOR_SIZE - 2] = AT_NULL;
2003 user_auxv[AT_VECTOR_SIZE - 1] = AT_NULL;
2004 }
2005
2006 if (prctl_map.exe_fd != (u32)-1) {
2007 /*
2008 * Make sure the caller has the rights to
2009 * change /proc/pid/exe link: only local sys admin should
2010 * be allowed to.
2011 */
2012 if (!ns_capable(current_user_ns(), CAP_SYS_ADMIN))
2013 return -EINVAL;
2014
2015 error = prctl_set_mm_exe_file(mm, prctl_map.exe_fd);
2016 if (error)
2017 return error;
2018 }
2019
2020 /*
2021 * arg_lock protects concurent updates but we still need mmap_sem for
2022 * read to exclude races with sys_brk.
2023 */
2024 down_read(&mm->mmap_sem);
2025
2026 /*
2027 * We don't validate if these members are pointing to
2028 * real present VMAs because application may have correspond
2029 * VMAs already unmapped and kernel uses these members for statistics
2030 * output in procfs mostly, except
2031 *
2032 * - @start_brk/@brk which are used in do_brk but kernel lookups
2033 * for VMAs when updating these memvers so anything wrong written
2034 * here cause kernel to swear at userspace program but won't lead
2035 * to any problem in kernel itself
2036 */
2037
2038 spin_lock(&mm->arg_lock);
2039 mm->start_code = prctl_map.start_code;
2040 mm->end_code = prctl_map.end_code;
2041 mm->start_data = prctl_map.start_data;
2042 mm->end_data = prctl_map.end_data;
2043 mm->start_brk = prctl_map.start_brk;
2044 mm->brk = prctl_map.brk;
2045 mm->start_stack = prctl_map.start_stack;
2046 mm->arg_start = prctl_map.arg_start;
2047 mm->arg_end = prctl_map.arg_end;
2048 mm->env_start = prctl_map.env_start;
2049 mm->env_end = prctl_map.env_end;
2050 spin_unlock(&mm->arg_lock);
2051
2052 /*
2053 * Note this update of @saved_auxv is lockless thus
2054 * if someone reads this member in procfs while we're
2055 * updating -- it may get partly updated results. It's
2056 * known and acceptable trade off: we leave it as is to
2057 * not introduce additional locks here making the kernel
2058 * more complex.
2059 */
2060 if (prctl_map.auxv_size)
2061 memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv));
2062
2063 up_read(&mm->mmap_sem);
2064 return 0;
2065 }
2066 #endif /* CONFIG_CHECKPOINT_RESTORE */
2067
prctl_set_auxv(struct mm_struct * mm,unsigned long addr,unsigned long len)2068 static int prctl_set_auxv(struct mm_struct *mm, unsigned long addr,
2069 unsigned long len)
2070 {
2071 /*
2072 * This doesn't move the auxiliary vector itself since it's pinned to
2073 * mm_struct, but it permits filling the vector with new values. It's
2074 * up to the caller to provide sane values here, otherwise userspace
2075 * tools which use this vector might be unhappy.
2076 */
2077 unsigned long user_auxv[AT_VECTOR_SIZE];
2078
2079 if (len > sizeof(user_auxv))
2080 return -EINVAL;
2081
2082 if (copy_from_user(user_auxv, (const void __user *)addr, len))
2083 return -EFAULT;
2084
2085 /* Make sure the last entry is always AT_NULL */
2086 user_auxv[AT_VECTOR_SIZE - 2] = 0;
2087 user_auxv[AT_VECTOR_SIZE - 1] = 0;
2088
2089 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
2090
2091 task_lock(current);
2092 memcpy(mm->saved_auxv, user_auxv, len);
2093 task_unlock(current);
2094
2095 return 0;
2096 }
2097
prctl_set_mm(int opt,unsigned long addr,unsigned long arg4,unsigned long arg5)2098 static int prctl_set_mm(int opt, unsigned long addr,
2099 unsigned long arg4, unsigned long arg5)
2100 {
2101 struct mm_struct *mm = current->mm;
2102 struct prctl_mm_map prctl_map = {
2103 .auxv = NULL,
2104 .auxv_size = 0,
2105 .exe_fd = -1,
2106 };
2107 struct vm_area_struct *vma;
2108 int error;
2109
2110 if (arg5 || (arg4 && (opt != PR_SET_MM_AUXV &&
2111 opt != PR_SET_MM_MAP &&
2112 opt != PR_SET_MM_MAP_SIZE)))
2113 return -EINVAL;
2114
2115 #ifdef CONFIG_CHECKPOINT_RESTORE
2116 if (opt == PR_SET_MM_MAP || opt == PR_SET_MM_MAP_SIZE)
2117 return prctl_set_mm_map(opt, (const void __user *)addr, arg4);
2118 #endif
2119
2120 if (!capable(CAP_SYS_RESOURCE))
2121 return -EPERM;
2122
2123 if (opt == PR_SET_MM_EXE_FILE)
2124 return prctl_set_mm_exe_file(mm, (unsigned int)addr);
2125
2126 if (opt == PR_SET_MM_AUXV)
2127 return prctl_set_auxv(mm, addr, arg4);
2128
2129 if (addr >= TASK_SIZE || addr < mmap_min_addr)
2130 return -EINVAL;
2131
2132 error = -EINVAL;
2133
2134 /*
2135 * arg_lock protects concurent updates of arg boundaries, we need
2136 * mmap_sem for a) concurrent sys_brk, b) finding VMA for addr
2137 * validation.
2138 */
2139 down_read(&mm->mmap_sem);
2140 vma = find_vma(mm, addr);
2141
2142 spin_lock(&mm->arg_lock);
2143 prctl_map.start_code = mm->start_code;
2144 prctl_map.end_code = mm->end_code;
2145 prctl_map.start_data = mm->start_data;
2146 prctl_map.end_data = mm->end_data;
2147 prctl_map.start_brk = mm->start_brk;
2148 prctl_map.brk = mm->brk;
2149 prctl_map.start_stack = mm->start_stack;
2150 prctl_map.arg_start = mm->arg_start;
2151 prctl_map.arg_end = mm->arg_end;
2152 prctl_map.env_start = mm->env_start;
2153 prctl_map.env_end = mm->env_end;
2154
2155 switch (opt) {
2156 case PR_SET_MM_START_CODE:
2157 prctl_map.start_code = addr;
2158 break;
2159 case PR_SET_MM_END_CODE:
2160 prctl_map.end_code = addr;
2161 break;
2162 case PR_SET_MM_START_DATA:
2163 prctl_map.start_data = addr;
2164 break;
2165 case PR_SET_MM_END_DATA:
2166 prctl_map.end_data = addr;
2167 break;
2168 case PR_SET_MM_START_STACK:
2169 prctl_map.start_stack = addr;
2170 break;
2171 case PR_SET_MM_START_BRK:
2172 prctl_map.start_brk = addr;
2173 break;
2174 case PR_SET_MM_BRK:
2175 prctl_map.brk = addr;
2176 break;
2177 case PR_SET_MM_ARG_START:
2178 prctl_map.arg_start = addr;
2179 break;
2180 case PR_SET_MM_ARG_END:
2181 prctl_map.arg_end = addr;
2182 break;
2183 case PR_SET_MM_ENV_START:
2184 prctl_map.env_start = addr;
2185 break;
2186 case PR_SET_MM_ENV_END:
2187 prctl_map.env_end = addr;
2188 break;
2189 default:
2190 goto out;
2191 }
2192
2193 error = validate_prctl_map_addr(&prctl_map);
2194 if (error)
2195 goto out;
2196
2197 switch (opt) {
2198 /*
2199 * If command line arguments and environment
2200 * are placed somewhere else on stack, we can
2201 * set them up here, ARG_START/END to setup
2202 * command line argumets and ENV_START/END
2203 * for environment.
2204 */
2205 case PR_SET_MM_START_STACK:
2206 case PR_SET_MM_ARG_START:
2207 case PR_SET_MM_ARG_END:
2208 case PR_SET_MM_ENV_START:
2209 case PR_SET_MM_ENV_END:
2210 if (!vma) {
2211 error = -EFAULT;
2212 goto out;
2213 }
2214 }
2215
2216 mm->start_code = prctl_map.start_code;
2217 mm->end_code = prctl_map.end_code;
2218 mm->start_data = prctl_map.start_data;
2219 mm->end_data = prctl_map.end_data;
2220 mm->start_brk = prctl_map.start_brk;
2221 mm->brk = prctl_map.brk;
2222 mm->start_stack = prctl_map.start_stack;
2223 mm->arg_start = prctl_map.arg_start;
2224 mm->arg_end = prctl_map.arg_end;
2225 mm->env_start = prctl_map.env_start;
2226 mm->env_end = prctl_map.env_end;
2227
2228 error = 0;
2229 out:
2230 spin_unlock(&mm->arg_lock);
2231 up_read(&mm->mmap_sem);
2232 return error;
2233 }
2234
2235 #ifdef CONFIG_CHECKPOINT_RESTORE
prctl_get_tid_address(struct task_struct * me,int __user ** tid_addr)2236 static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2237 {
2238 return put_user(me->clear_child_tid, tid_addr);
2239 }
2240 #else
prctl_get_tid_address(struct task_struct * me,int __user ** tid_addr)2241 static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2242 {
2243 return -EINVAL;
2244 }
2245 #endif
2246
propagate_has_child_subreaper(struct task_struct * p,void * data)2247 static int propagate_has_child_subreaper(struct task_struct *p, void *data)
2248 {
2249 /*
2250 * If task has has_child_subreaper - all its decendants
2251 * already have these flag too and new decendants will
2252 * inherit it on fork, skip them.
2253 *
2254 * If we've found child_reaper - skip descendants in
2255 * it's subtree as they will never get out pidns.
2256 */
2257 if (p->signal->has_child_subreaper ||
2258 is_child_reaper(task_pid(p)))
2259 return 0;
2260
2261 p->signal->has_child_subreaper = 1;
2262 return 1;
2263 }
2264
arch_prctl_spec_ctrl_get(struct task_struct * t,unsigned long which)2265 int __weak arch_prctl_spec_ctrl_get(struct task_struct *t, unsigned long which)
2266 {
2267 return -EINVAL;
2268 }
2269
arch_prctl_spec_ctrl_set(struct task_struct * t,unsigned long which,unsigned long ctrl)2270 int __weak arch_prctl_spec_ctrl_set(struct task_struct *t, unsigned long which,
2271 unsigned long ctrl)
2272 {
2273 return -EINVAL;
2274 }
2275
2276 #ifdef CONFIG_MMU
prctl_update_vma_anon_name(struct vm_area_struct * vma,struct vm_area_struct ** prev,unsigned long start,unsigned long end,const char __user * name_addr)2277 static int prctl_update_vma_anon_name(struct vm_area_struct *vma,
2278 struct vm_area_struct **prev,
2279 unsigned long start, unsigned long end,
2280 const char __user *name_addr)
2281 {
2282 struct mm_struct *mm = vma->vm_mm;
2283 int error = 0;
2284 pgoff_t pgoff;
2285
2286 if (name_addr == vma_get_anon_name(vma)) {
2287 *prev = vma;
2288 goto out;
2289 }
2290
2291 pgoff = vma->vm_pgoff + ((start - vma->vm_start) >> PAGE_SHIFT);
2292 *prev = vma_merge(mm, *prev, start, end, vma->vm_flags, vma->anon_vma,
2293 vma->vm_file, pgoff, vma_policy(vma),
2294 vma->vm_userfaultfd_ctx, name_addr);
2295 if (*prev) {
2296 vma = *prev;
2297 goto success;
2298 }
2299
2300 *prev = vma;
2301
2302 if (start != vma->vm_start) {
2303 error = split_vma(mm, vma, start, 1);
2304 if (error)
2305 goto out;
2306 }
2307
2308 if (end != vma->vm_end) {
2309 error = split_vma(mm, vma, end, 0);
2310 if (error)
2311 goto out;
2312 }
2313
2314 success:
2315 if (!vma->vm_file)
2316 vma->anon_name = name_addr;
2317
2318 out:
2319 if (error == -ENOMEM)
2320 error = -EAGAIN;
2321 return error;
2322 }
2323
prctl_set_vma_anon_name(unsigned long start,unsigned long end,unsigned long arg)2324 static int prctl_set_vma_anon_name(unsigned long start, unsigned long end,
2325 unsigned long arg)
2326 {
2327 unsigned long tmp;
2328 struct vm_area_struct *vma, *prev;
2329 int unmapped_error = 0;
2330 int error = -EINVAL;
2331
2332 /*
2333 * If the interval [start,end) covers some unmapped address
2334 * ranges, just ignore them, but return -ENOMEM at the end.
2335 * - this matches the handling in madvise.
2336 */
2337 vma = find_vma_prev(current->mm, start, &prev);
2338 if (vma && start > vma->vm_start)
2339 prev = vma;
2340
2341 for (;;) {
2342 /* Still start < end. */
2343 error = -ENOMEM;
2344 if (!vma)
2345 return error;
2346
2347 /* Here start < (end|vma->vm_end). */
2348 if (start < vma->vm_start) {
2349 unmapped_error = -ENOMEM;
2350 start = vma->vm_start;
2351 if (start >= end)
2352 return error;
2353 }
2354
2355 /* Here vma->vm_start <= start < (end|vma->vm_end) */
2356 tmp = vma->vm_end;
2357 if (end < tmp)
2358 tmp = end;
2359
2360 /* Here vma->vm_start <= start < tmp <= (end|vma->vm_end). */
2361 error = prctl_update_vma_anon_name(vma, &prev, start, tmp,
2362 (const char __user *)arg);
2363 if (error)
2364 return error;
2365 start = tmp;
2366 if (prev && start < prev->vm_end)
2367 start = prev->vm_end;
2368 error = unmapped_error;
2369 if (start >= end)
2370 return error;
2371 if (prev)
2372 vma = prev->vm_next;
2373 else /* madvise_remove dropped mmap_sem */
2374 vma = find_vma(current->mm, start);
2375 }
2376 }
2377
prctl_set_vma(unsigned long opt,unsigned long start,unsigned long len_in,unsigned long arg)2378 static int prctl_set_vma(unsigned long opt, unsigned long start,
2379 unsigned long len_in, unsigned long arg)
2380 {
2381 struct mm_struct *mm = current->mm;
2382 int error;
2383 unsigned long len;
2384 unsigned long end;
2385
2386 if (start & ~PAGE_MASK)
2387 return -EINVAL;
2388 len = (len_in + ~PAGE_MASK) & PAGE_MASK;
2389
2390 /* Check to see whether len was rounded up from small -ve to zero */
2391 if (len_in && !len)
2392 return -EINVAL;
2393
2394 end = start + len;
2395 if (end < start)
2396 return -EINVAL;
2397
2398 if (end == start)
2399 return 0;
2400
2401 down_write(&mm->mmap_sem);
2402
2403 switch (opt) {
2404 case PR_SET_VMA_ANON_NAME:
2405 error = prctl_set_vma_anon_name(start, end, arg);
2406 break;
2407 default:
2408 error = -EINVAL;
2409 }
2410
2411 up_write(&mm->mmap_sem);
2412
2413 return error;
2414 }
2415 #else /* CONFIG_MMU */
prctl_set_vma(unsigned long opt,unsigned long start,unsigned long len_in,unsigned long arg)2416 static int prctl_set_vma(unsigned long opt, unsigned long start,
2417 unsigned long len_in, unsigned long arg)
2418 {
2419 return -EINVAL;
2420 }
2421 #endif
2422
SYSCALL_DEFINE5(prctl,int,option,unsigned long,arg2,unsigned long,arg3,unsigned long,arg4,unsigned long,arg5)2423 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2424 unsigned long, arg4, unsigned long, arg5)
2425 {
2426 struct task_struct *me = current;
2427 unsigned char comm[sizeof(me->comm)];
2428 long error;
2429
2430 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2431 if (error != -ENOSYS)
2432 return error;
2433
2434 error = 0;
2435 switch (option) {
2436 case PR_SET_PDEATHSIG:
2437 if (!valid_signal(arg2)) {
2438 error = -EINVAL;
2439 break;
2440 }
2441 me->pdeath_signal = arg2;
2442 break;
2443 case PR_GET_PDEATHSIG:
2444 error = put_user(me->pdeath_signal, (int __user *)arg2);
2445 break;
2446 case PR_GET_DUMPABLE:
2447 error = get_dumpable(me->mm);
2448 break;
2449 case PR_SET_DUMPABLE:
2450 if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) {
2451 error = -EINVAL;
2452 break;
2453 }
2454 set_dumpable(me->mm, arg2);
2455 break;
2456
2457 case PR_SET_UNALIGN:
2458 error = SET_UNALIGN_CTL(me, arg2);
2459 break;
2460 case PR_GET_UNALIGN:
2461 error = GET_UNALIGN_CTL(me, arg2);
2462 break;
2463 case PR_SET_FPEMU:
2464 error = SET_FPEMU_CTL(me, arg2);
2465 break;
2466 case PR_GET_FPEMU:
2467 error = GET_FPEMU_CTL(me, arg2);
2468 break;
2469 case PR_SET_FPEXC:
2470 error = SET_FPEXC_CTL(me, arg2);
2471 break;
2472 case PR_GET_FPEXC:
2473 error = GET_FPEXC_CTL(me, arg2);
2474 break;
2475 case PR_GET_TIMING:
2476 error = PR_TIMING_STATISTICAL;
2477 break;
2478 case PR_SET_TIMING:
2479 if (arg2 != PR_TIMING_STATISTICAL)
2480 error = -EINVAL;
2481 break;
2482 case PR_SET_NAME:
2483 comm[sizeof(me->comm) - 1] = 0;
2484 if (strncpy_from_user(comm, (char __user *)arg2,
2485 sizeof(me->comm) - 1) < 0)
2486 return -EFAULT;
2487 set_task_comm(me, comm);
2488 proc_comm_connector(me);
2489 break;
2490 case PR_GET_NAME:
2491 get_task_comm(comm, me);
2492 if (copy_to_user((char __user *)arg2, comm, sizeof(comm)))
2493 return -EFAULT;
2494 break;
2495 case PR_GET_ENDIAN:
2496 error = GET_ENDIAN(me, arg2);
2497 break;
2498 case PR_SET_ENDIAN:
2499 error = SET_ENDIAN(me, arg2);
2500 break;
2501 case PR_GET_SECCOMP:
2502 error = prctl_get_seccomp();
2503 break;
2504 case PR_SET_SECCOMP:
2505 error = prctl_set_seccomp(arg2, (char __user *)arg3);
2506 break;
2507 case PR_GET_TSC:
2508 error = GET_TSC_CTL(arg2);
2509 break;
2510 case PR_SET_TSC:
2511 error = SET_TSC_CTL(arg2);
2512 break;
2513 case PR_TASK_PERF_EVENTS_DISABLE:
2514 error = perf_event_task_disable();
2515 break;
2516 case PR_TASK_PERF_EVENTS_ENABLE:
2517 error = perf_event_task_enable();
2518 break;
2519 case PR_GET_TIMERSLACK:
2520 if (current->timer_slack_ns > ULONG_MAX)
2521 error = ULONG_MAX;
2522 else
2523 error = current->timer_slack_ns;
2524 break;
2525 case PR_SET_TIMERSLACK:
2526 if (arg2 <= 0)
2527 current->timer_slack_ns =
2528 current->default_timer_slack_ns;
2529 else
2530 current->timer_slack_ns = arg2;
2531 break;
2532 case PR_MCE_KILL:
2533 if (arg4 | arg5)
2534 return -EINVAL;
2535 switch (arg2) {
2536 case PR_MCE_KILL_CLEAR:
2537 if (arg3 != 0)
2538 return -EINVAL;
2539 current->flags &= ~PF_MCE_PROCESS;
2540 break;
2541 case PR_MCE_KILL_SET:
2542 current->flags |= PF_MCE_PROCESS;
2543 if (arg3 == PR_MCE_KILL_EARLY)
2544 current->flags |= PF_MCE_EARLY;
2545 else if (arg3 == PR_MCE_KILL_LATE)
2546 current->flags &= ~PF_MCE_EARLY;
2547 else if (arg3 == PR_MCE_KILL_DEFAULT)
2548 current->flags &=
2549 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
2550 else
2551 return -EINVAL;
2552 break;
2553 default:
2554 return -EINVAL;
2555 }
2556 break;
2557 case PR_MCE_KILL_GET:
2558 if (arg2 | arg3 | arg4 | arg5)
2559 return -EINVAL;
2560 if (current->flags & PF_MCE_PROCESS)
2561 error = (current->flags & PF_MCE_EARLY) ?
2562 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2563 else
2564 error = PR_MCE_KILL_DEFAULT;
2565 break;
2566 case PR_SET_MM:
2567 error = prctl_set_mm(arg2, arg3, arg4, arg5);
2568 break;
2569 case PR_GET_TID_ADDRESS:
2570 error = prctl_get_tid_address(me, (int __user **)arg2);
2571 break;
2572 case PR_SET_CHILD_SUBREAPER:
2573 me->signal->is_child_subreaper = !!arg2;
2574 if (!arg2)
2575 break;
2576
2577 walk_process_tree(me, propagate_has_child_subreaper, NULL);
2578 break;
2579 case PR_GET_CHILD_SUBREAPER:
2580 error = put_user(me->signal->is_child_subreaper,
2581 (int __user *)arg2);
2582 break;
2583 case PR_SET_NO_NEW_PRIVS:
2584 if (arg2 != 1 || arg3 || arg4 || arg5)
2585 return -EINVAL;
2586
2587 task_set_no_new_privs(current);
2588 break;
2589 case PR_GET_NO_NEW_PRIVS:
2590 if (arg2 || arg3 || arg4 || arg5)
2591 return -EINVAL;
2592 return task_no_new_privs(current) ? 1 : 0;
2593 case PR_GET_THP_DISABLE:
2594 if (arg2 || arg3 || arg4 || arg5)
2595 return -EINVAL;
2596 error = !!test_bit(MMF_DISABLE_THP, &me->mm->flags);
2597 break;
2598 case PR_SET_THP_DISABLE:
2599 if (arg3 || arg4 || arg5)
2600 return -EINVAL;
2601 if (down_write_killable(&me->mm->mmap_sem))
2602 return -EINTR;
2603 if (arg2)
2604 set_bit(MMF_DISABLE_THP, &me->mm->flags);
2605 else
2606 clear_bit(MMF_DISABLE_THP, &me->mm->flags);
2607 up_write(&me->mm->mmap_sem);
2608 break;
2609 case PR_MPX_ENABLE_MANAGEMENT:
2610 case PR_MPX_DISABLE_MANAGEMENT:
2611 /* No longer implemented: */
2612 return -EINVAL;
2613 case PR_SET_FP_MODE:
2614 error = SET_FP_MODE(me, arg2);
2615 break;
2616 case PR_GET_FP_MODE:
2617 error = GET_FP_MODE(me);
2618 break;
2619 case PR_SVE_SET_VL:
2620 error = SVE_SET_VL(arg2);
2621 break;
2622 case PR_SVE_GET_VL:
2623 error = SVE_GET_VL();
2624 break;
2625 case PR_GET_SPECULATION_CTRL:
2626 if (arg3 || arg4 || arg5)
2627 return -EINVAL;
2628 error = arch_prctl_spec_ctrl_get(me, arg2);
2629 break;
2630 case PR_SET_SPECULATION_CTRL:
2631 if (arg4 || arg5)
2632 return -EINVAL;
2633 error = arch_prctl_spec_ctrl_set(me, arg2, arg3);
2634 break;
2635 case PR_SET_VMA:
2636 error = prctl_set_vma(arg2, arg3, arg4, arg5);
2637 break;
2638 case PR_PAC_RESET_KEYS:
2639 if (arg3 || arg4 || arg5)
2640 return -EINVAL;
2641 error = PAC_RESET_KEYS(me, arg2);
2642 break;
2643 case PR_SET_TAGGED_ADDR_CTRL:
2644 if (arg3 || arg4 || arg5)
2645 return -EINVAL;
2646 error = SET_TAGGED_ADDR_CTRL(arg2);
2647 break;
2648 case PR_GET_TAGGED_ADDR_CTRL:
2649 if (arg2 || arg3 || arg4 || arg5)
2650 return -EINVAL;
2651 error = GET_TAGGED_ADDR_CTRL();
2652 break;
2653 default:
2654 error = -EINVAL;
2655 break;
2656 }
2657 trace_android_vh_syscall_prctl_finished(option, me);
2658 return error;
2659 }
2660
SYSCALL_DEFINE3(getcpu,unsigned __user *,cpup,unsigned __user *,nodep,struct getcpu_cache __user *,unused)2661 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2662 struct getcpu_cache __user *, unused)
2663 {
2664 int err = 0;
2665 int cpu = raw_smp_processor_id();
2666
2667 if (cpup)
2668 err |= put_user(cpu, cpup);
2669 if (nodep)
2670 err |= put_user(cpu_to_node(cpu), nodep);
2671 return err ? -EFAULT : 0;
2672 }
2673
2674 /**
2675 * do_sysinfo - fill in sysinfo struct
2676 * @info: pointer to buffer to fill
2677 */
do_sysinfo(struct sysinfo * info)2678 static int do_sysinfo(struct sysinfo *info)
2679 {
2680 unsigned long mem_total, sav_total;
2681 unsigned int mem_unit, bitcount;
2682 struct timespec64 tp;
2683
2684 memset(info, 0, sizeof(struct sysinfo));
2685
2686 ktime_get_boottime_ts64(&tp);
2687 info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
2688
2689 get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
2690
2691 info->procs = nr_threads;
2692
2693 si_meminfo(info);
2694 si_swapinfo(info);
2695
2696 /*
2697 * If the sum of all the available memory (i.e. ram + swap)
2698 * is less than can be stored in a 32 bit unsigned long then
2699 * we can be binary compatible with 2.2.x kernels. If not,
2700 * well, in that case 2.2.x was broken anyways...
2701 *
2702 * -Erik Andersen <andersee@debian.org>
2703 */
2704
2705 mem_total = info->totalram + info->totalswap;
2706 if (mem_total < info->totalram || mem_total < info->totalswap)
2707 goto out;
2708 bitcount = 0;
2709 mem_unit = info->mem_unit;
2710 while (mem_unit > 1) {
2711 bitcount++;
2712 mem_unit >>= 1;
2713 sav_total = mem_total;
2714 mem_total <<= 1;
2715 if (mem_total < sav_total)
2716 goto out;
2717 }
2718
2719 /*
2720 * If mem_total did not overflow, multiply all memory values by
2721 * info->mem_unit and set it to 1. This leaves things compatible
2722 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2723 * kernels...
2724 */
2725
2726 info->mem_unit = 1;
2727 info->totalram <<= bitcount;
2728 info->freeram <<= bitcount;
2729 info->sharedram <<= bitcount;
2730 info->bufferram <<= bitcount;
2731 info->totalswap <<= bitcount;
2732 info->freeswap <<= bitcount;
2733 info->totalhigh <<= bitcount;
2734 info->freehigh <<= bitcount;
2735
2736 out:
2737 return 0;
2738 }
2739
SYSCALL_DEFINE1(sysinfo,struct sysinfo __user *,info)2740 SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
2741 {
2742 struct sysinfo val;
2743
2744 do_sysinfo(&val);
2745
2746 if (copy_to_user(info, &val, sizeof(struct sysinfo)))
2747 return -EFAULT;
2748
2749 return 0;
2750 }
2751
2752 #ifdef CONFIG_COMPAT
2753 struct compat_sysinfo {
2754 s32 uptime;
2755 u32 loads[3];
2756 u32 totalram;
2757 u32 freeram;
2758 u32 sharedram;
2759 u32 bufferram;
2760 u32 totalswap;
2761 u32 freeswap;
2762 u16 procs;
2763 u16 pad;
2764 u32 totalhigh;
2765 u32 freehigh;
2766 u32 mem_unit;
2767 char _f[20-2*sizeof(u32)-sizeof(int)];
2768 };
2769
COMPAT_SYSCALL_DEFINE1(sysinfo,struct compat_sysinfo __user *,info)2770 COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2771 {
2772 struct sysinfo s;
2773
2774 do_sysinfo(&s);
2775
2776 /* Check to see if any memory value is too large for 32-bit and scale
2777 * down if needed
2778 */
2779 if (upper_32_bits(s.totalram) || upper_32_bits(s.totalswap)) {
2780 int bitcount = 0;
2781
2782 while (s.mem_unit < PAGE_SIZE) {
2783 s.mem_unit <<= 1;
2784 bitcount++;
2785 }
2786
2787 s.totalram >>= bitcount;
2788 s.freeram >>= bitcount;
2789 s.sharedram >>= bitcount;
2790 s.bufferram >>= bitcount;
2791 s.totalswap >>= bitcount;
2792 s.freeswap >>= bitcount;
2793 s.totalhigh >>= bitcount;
2794 s.freehigh >>= bitcount;
2795 }
2796
2797 if (!access_ok(info, sizeof(struct compat_sysinfo)) ||
2798 __put_user(s.uptime, &info->uptime) ||
2799 __put_user(s.loads[0], &info->loads[0]) ||
2800 __put_user(s.loads[1], &info->loads[1]) ||
2801 __put_user(s.loads[2], &info->loads[2]) ||
2802 __put_user(s.totalram, &info->totalram) ||
2803 __put_user(s.freeram, &info->freeram) ||
2804 __put_user(s.sharedram, &info->sharedram) ||
2805 __put_user(s.bufferram, &info->bufferram) ||
2806 __put_user(s.totalswap, &info->totalswap) ||
2807 __put_user(s.freeswap, &info->freeswap) ||
2808 __put_user(s.procs, &info->procs) ||
2809 __put_user(s.totalhigh, &info->totalhigh) ||
2810 __put_user(s.freehigh, &info->freehigh) ||
2811 __put_user(s.mem_unit, &info->mem_unit))
2812 return -EFAULT;
2813
2814 return 0;
2815 }
2816 #endif /* CONFIG_COMPAT */
2817