]> git.proxmox.com Git - mirror_ubuntu-disco-kernel.git/blame - kernel/sys.c
projects / mirror_ubuntu-disco-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
UBUNTU: Ubuntu-5.0.0-29.31
[mirror_ubuntu-disco-kernel.git] / kernel / sys.c
CommitLineData
b2441318 1// SPDX-License-Identifier: GPL-2.0
1da177e4
LT		 2/*
3 * linux/kernel/sys.c
4 *
5 * Copyright (C) 1991, 1992 Linus Torvalds
6 */
7
9984de1a 8#include <linux/export.h>
1da177e4
LT		 9#include <linux/mm.h>
10#include <linux/utsname.h>
11#include <linux/mman.h>
1da177e4
LT		 12#include <linux/reboot.h>
13#include <linux/prctl.h>
1da177e4
LT		 14#include <linux/highuid.h>
15#include <linux/fs.h>
74da1ff7 16#include <linux/kmod.h>
cdd6c482 17#include <linux/perf_event.h>
3e88c553 18#include <linux/resource.h>
dc009d92 19#include <linux/kernel.h>
1da177e4 20#include <linux/workqueue.h>
c59ede7b 21#include <linux/capability.h>
1da177e4
LT		 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>
7ed20e1a 30#include <linux/signal.h>
9f46080c 31#include <linux/cn_proc.h>
3cfc348b 32#include <linux/getcpu.h>
6eaeeaba 33#include <linux/task_io_accounting_ops.h>
1d9d02fe 34#include <linux/seccomp.h>
4047727e 35#include <linux/cpu.h>
e28cbf22 36#include <linux/personality.h>
e3d5a27d 37#include <linux/ptrace.h>
5ad4e53b 38#include <linux/fs_struct.h>
b32dfe37
CG		 39#include <linux/file.h>
40#include <linux/mount.h>
5a0e3ad6 41#include <linux/gfp.h>
40dc166c 42#include <linux/syscore_ops.h>
be27425d
AK		 43#include <linux/version.h>
44#include <linux/ctype.h>
1da177e4
LT		 45
46#include <linux/compat.h>
47#include <linux/syscalls.h>
00d7c05a 48#include <linux/kprobes.h>
acce292c 49#include <linux/user_namespace.h>
7fe5e042 50#include <linux/binfmts.h>
1da177e4 51
4a22f166 52#include <linux/sched.h>
4eb5aaa3 53#include <linux/sched/autogroup.h>
4f17722c 54#include <linux/sched/loadavg.h>
03441a34 55#include <linux/sched/stat.h>
6e84f315 56#include <linux/sched/mm.h>
f7ccbae4 57#include <linux/sched/coredump.h>
29930025 58#include <linux/sched/task.h>
32ef5517 59#include <linux/sched/cputime.h>
4a22f166
SR		 60#include <linux/rcupdate.h>
61#include <linux/uidgid.h>
62#include <linux/cred.h>
63
b617cfc8
TG		 64#include <linux/nospec.h>
65
04c6862c 66#include <linux/kmsg_dump.h>
be27425d
AK		 67/* Move somewhere else to avoid recompiling? */
68#include <generated/utsrelease.h>
04c6862c 69
7c0f6ba6 70#include <linux/uaccess.h>
1da177e4
LT		 71#include <asm/io.h>
72#include <asm/unistd.h>
73
e530dca5
DB		 74#include "uid16.h"
75
1da177e4 76#ifndef SET_UNALIGN_CTL
ec94fc3d 77# define SET_UNALIGN_CTL(a, b) (-EINVAL)
1da177e4
LT		 78#endif
79#ifndef GET_UNALIGN_CTL
ec94fc3d 80# define GET_UNALIGN_CTL(a, b) (-EINVAL)
1da177e4
LT		 81#endif
82#ifndef SET_FPEMU_CTL
ec94fc3d 83# define SET_FPEMU_CTL(a, b) (-EINVAL)
1da177e4
LT		 84#endif
85#ifndef GET_FPEMU_CTL
ec94fc3d 86# define GET_FPEMU_CTL(a, b) (-EINVAL)
1da177e4
LT		 87#endif
88#ifndef SET_FPEXC_CTL
ec94fc3d 89# define SET_FPEXC_CTL(a, b) (-EINVAL)
1da177e4
LT		 90#endif
91#ifndef GET_FPEXC_CTL
ec94fc3d 92# define GET_FPEXC_CTL(a, b) (-EINVAL)
1da177e4 93#endif
651d765d 94#ifndef GET_ENDIAN
ec94fc3d 95# define GET_ENDIAN(a, b) (-EINVAL)
651d765d
AB		 96#endif
97#ifndef SET_ENDIAN
ec94fc3d 98# define SET_ENDIAN(a, b) (-EINVAL)
651d765d 99#endif
8fb402bc
EB		 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
fe3d197f 106#ifndef MPX_ENABLE_MANAGEMENT
46a6e0cf 107# define MPX_ENABLE_MANAGEMENT() (-EINVAL)
fe3d197f
DH		 108#endif
109#ifndef MPX_DISABLE_MANAGEMENT
46a6e0cf 110# define MPX_DISABLE_MANAGEMENT() (-EINVAL)
fe3d197f 111#endif
9791554b
PB		 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
2d2123bc
DM		 118#ifndef SVE_SET_VL
119# define SVE_SET_VL(a) (-EINVAL)
120#endif
121#ifndef SVE_GET_VL
122# define SVE_GET_VL() (-EINVAL)
123#endif
ba830885
KM		 124#ifndef PAC_RESET_KEYS
125# define PAC_RESET_KEYS(a, b) (-EINVAL)
126#endif
1da177e4
LT		 127
128/*
129 * this is where the system-wide overflow UID and GID are defined, for
130 * architectures that now have 32-bit UID/GID but didn't in the past
131 */
132
133int overflowuid = DEFAULT_OVERFLOWUID;
134int overflowgid = DEFAULT_OVERFLOWGID;
135
1da177e4
LT		 136EXPORT_SYMBOL(overflowuid);
137EXPORT_SYMBOL(overflowgid);
1da177e4
LT		 138
139/*
140 * the same as above, but for filesystems which can only store a 16-bit
141 * UID and GID. as such, this is needed on all architectures
142 */
143
144int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
8b2770a4 145int fs_overflowgid = DEFAULT_FS_OVERFLOWGID;
1da177e4
LT		 146
147EXPORT_SYMBOL(fs_overflowuid);
148EXPORT_SYMBOL(fs_overflowgid);
149
fc832ad3
SH		 150/*
151 * Returns true if current's euid is same as p's uid or euid,
152 * or has CAP_SYS_NICE to p's user_ns.
153 *
154 * Called with rcu_read_lock, creds are safe
155 */
156static bool set_one_prio_perm(struct task_struct *p)
157{
158 const struct cred *cred = current_cred(), *pcred = __task_cred(p);
159
5af66203
EB		 160 if (uid_eq(pcred->uid, cred->euid) ||
161 uid_eq(pcred->euid, cred->euid))
fc832ad3 162 return true;
c4a4d603 163 if (ns_capable(pcred->user_ns, CAP_SYS_NICE))
fc832ad3
SH		 164 return true;
165 return false;
166}
167
c69e8d9c
DH		 168/*
169 * set the priority of a task
170 * - the caller must hold the RCU read lock
171 */
1da177e4
LT		 172static int set_one_prio(struct task_struct *p, int niceval, int error)
173{
174 int no_nice;
175
fc832ad3 176 if (!set_one_prio_perm(p)) {
1da177e4
LT		 177 error = -EPERM;
178 goto out;
179 }
e43379f1 180 if (niceval < task_nice(p) && !can_nice(p, niceval)) {
1da177e4
LT		 181 error = -EACCES;
182 goto out;
183 }
184 no_nice = security_task_setnice(p, niceval);
185 if (no_nice) {
186 error = no_nice;
187 goto out;
188 }
189 if (error == -ESRCH)
190 error = 0;
191 set_user_nice(p, niceval);
192out:
193 return error;
194}
195
754fe8d2 196SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
1da177e4
LT		 197{
198 struct task_struct *g, *p;
199 struct user_struct *user;
86a264ab 200 const struct cred *cred = current_cred();
1da177e4 201 int error = -EINVAL;
41487c65 202 struct pid *pgrp;
7b44ab97 203 kuid_t uid;
1da177e4 204
3e88c553 205 if (which > PRIO_USER || which < PRIO_PROCESS)
1da177e4
LT		 206 goto out;
207
208 /* normalize: avoid signed division (rounding problems) */
209 error = -ESRCH;
c4a4d2f4
DY		 210 if (niceval < MIN_NICE)
211 niceval = MIN_NICE;
212 if (niceval > MAX_NICE)
213 niceval = MAX_NICE;
1da177e4 214
d4581a23 215 rcu_read_lock();
1da177e4
LT		 216 read_lock(&tasklist_lock);
217 switch (which) {
ec94fc3d 218 case PRIO_PROCESS:
219 if (who)
220 p = find_task_by_vpid(who);
221 else
222 p = current;
223 if (p)
224 error = set_one_prio(p, niceval, error);
225 break;
226 case PRIO_PGRP:
227 if (who)
228 pgrp = find_vpid(who);
229 else
230 pgrp = task_pgrp(current);
231 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
232 error = set_one_prio(p, niceval, error);
233 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
234 break;
235 case PRIO_USER:
236 uid = make_kuid(cred->user_ns, who);
237 user = cred->user;
238 if (!who)
239 uid = cred->uid;
240 else if (!uid_eq(uid, cred->uid)) {
241 user = find_user(uid);
242 if (!user)
86a264ab 243 goto out_unlock; /* No processes for this user */
ec94fc3d 244 }
245 do_each_thread(g, p) {
8639b461 246 if (uid_eq(task_uid(p), uid) && task_pid_vnr(p))
ec94fc3d 247 error = set_one_prio(p, niceval, error);
248 } while_each_thread(g, p);
249 if (!uid_eq(uid, cred->uid))
250 free_uid(user); /* For find_user() */
251 break;
1da177e4
LT		 252 }
253out_unlock:
254 read_unlock(&tasklist_lock);
d4581a23 255 rcu_read_unlock();
1da177e4
LT		 256out:
257 return error;
258}
259
260/*
261 * Ugh. To avoid negative return values, "getpriority()" will
262 * not return the normal nice-value, but a negated value that
263 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
264 * to stay compatible.
265 */
754fe8d2 266SYSCALL_DEFINE2(getpriority, int, which, int, who)
1da177e4
LT		 267{
268 struct task_struct *g, *p;
269 struct user_struct *user;
86a264ab 270 const struct cred *cred = current_cred();
1da177e4 271 long niceval, retval = -ESRCH;
41487c65 272 struct pid *pgrp;
7b44ab97 273 kuid_t uid;
1da177e4 274
3e88c553 275 if (which > PRIO_USER || which < PRIO_PROCESS)
1da177e4
LT		 276 return -EINVAL;
277
70118837 278 rcu_read_lock();
1da177e4
LT		 279 read_lock(&tasklist_lock);
280 switch (which) {
ec94fc3d 281 case PRIO_PROCESS:
282 if (who)
283 p = find_task_by_vpid(who);
284 else
285 p = current;
286 if (p) {
287 niceval = nice_to_rlimit(task_nice(p));
288 if (niceval > retval)
289 retval = niceval;
290 }
291 break;
292 case PRIO_PGRP:
293 if (who)
294 pgrp = find_vpid(who);
295 else
296 pgrp = task_pgrp(current);
297 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
298 niceval = nice_to_rlimit(task_nice(p));
299 if (niceval > retval)
300 retval = niceval;
301 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
302 break;
303 case PRIO_USER:
304 uid = make_kuid(cred->user_ns, who);
305 user = cred->user;
306 if (!who)
307 uid = cred->uid;
308 else if (!uid_eq(uid, cred->uid)) {
309 user = find_user(uid);
310 if (!user)
311 goto out_unlock; /* No processes for this user */
312 }
313 do_each_thread(g, p) {
8639b461 314 if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) {
7aa2c016 315 niceval = nice_to_rlimit(task_nice(p));
1da177e4
LT		 316 if (niceval > retval)
317 retval = niceval;
318 }
ec94fc3d 319 } while_each_thread(g, p);
320 if (!uid_eq(uid, cred->uid))
321 free_uid(user); /* for find_user() */
322 break;
1da177e4
LT		 323 }
324out_unlock:
325 read_unlock(&tasklist_lock);
70118837 326 rcu_read_unlock();
1da177e4
LT		 327
328 return retval;
329}
330
1da177e4
LT		 331/*
332 * Unprivileged users may change the real gid to the effective gid
333 * or vice versa. (BSD-style)
334 *
335 * If you set the real gid at all, or set the effective gid to a value not
336 * equal to the real gid, then the saved gid is set to the new effective gid.
337 *
338 * This makes it possible for a setgid program to completely drop its
339 * privileges, which is often a useful assertion to make when you are doing
340 * a security audit over a program.
341 *
342 * The general idea is that a program which uses just setregid() will be
343 * 100% compatible with BSD. A program which uses just setgid() will be
ec94fc3d 344 * 100% compatible with POSIX with saved IDs.
1da177e4
LT		 345 *
346 * SMP: There are not races, the GIDs are checked only by filesystem
347 * operations (as far as semantic preservation is concerned).
348 */
2813893f 349#ifdef CONFIG_MULTIUSER
e530dca5 350long __sys_setregid(gid_t rgid, gid_t egid)
1da177e4 351{
a29c33f4 352 struct user_namespace *ns = current_user_ns();
d84f4f99
DH		 353 const struct cred *old;
354 struct cred *new;
1da177e4 355 int retval;
a29c33f4
EB		 356 kgid_t krgid, kegid;
357
358 krgid = make_kgid(ns, rgid);
359 kegid = make_kgid(ns, egid);
360
361 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
362 return -EINVAL;
363 if ((egid != (gid_t) -1) && !gid_valid(kegid))
364 return -EINVAL;
1da177e4 365
d84f4f99
DH		 366 new = prepare_creds();
367 if (!new)
368 return -ENOMEM;
369 old = current_cred();
370
d84f4f99 371 retval = -EPERM;
1da177e4 372 if (rgid != (gid_t) -1) {
a29c33f4
EB		 373 if (gid_eq(old->gid, krgid) ||
374 gid_eq(old->egid, krgid) ||
c7b96acf 375 ns_capable(old->user_ns, CAP_SETGID))
a29c33f4 376 new->gid = krgid;
1da177e4 377 else
d84f4f99 378 goto error;
1da177e4
LT		 379 }
380 if (egid != (gid_t) -1) {
a29c33f4
EB		 381 if (gid_eq(old->gid, kegid) ||
382 gid_eq(old->egid, kegid) ||
383 gid_eq(old->sgid, kegid) ||
c7b96acf 384 ns_capable(old->user_ns, CAP_SETGID))
a29c33f4 385 new->egid = kegid;
756184b7 386 else
d84f4f99 387 goto error;
1da177e4 388 }
d84f4f99 389
1da177e4 390 if (rgid != (gid_t) -1 ||
a29c33f4 391 (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
d84f4f99
DH		 392 new->sgid = new->egid;
393 new->fsgid = new->egid;
394
395 return commit_creds(new);
396
397error:
398 abort_creds(new);
399 return retval;
1da177e4
LT		 400}
401
e530dca5
DB		 402SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
403{
404 return __sys_setregid(rgid, egid);
405}
406
1da177e4 407/*
ec94fc3d 408 * setgid() is implemented like SysV w/ SAVED_IDS
1da177e4
LT		 409 *
410 * SMP: Same implicit races as above.
411 */
e530dca5 412long __sys_setgid(gid_t gid)
1da177e4 413{
a29c33f4 414 struct user_namespace *ns = current_user_ns();
d84f4f99
DH		 415 const struct cred *old;
416 struct cred *new;
1da177e4 417 int retval;
a29c33f4
EB		 418 kgid_t kgid;
419
420 kgid = make_kgid(ns, gid);
421 if (!gid_valid(kgid))
422 return -EINVAL;
1da177e4 423
d84f4f99
DH		 424 new = prepare_creds();
425 if (!new)
426 return -ENOMEM;
427 old = current_cred();
428
d84f4f99 429 retval = -EPERM;
c7b96acf 430 if (ns_capable(old->user_ns, CAP_SETGID))
a29c33f4
EB		 431 new->gid = new->egid = new->sgid = new->fsgid = kgid;
432 else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
433 new->egid = new->fsgid = kgid;
1da177e4 434 else
d84f4f99 435 goto error;
1da177e4 436
d84f4f99
DH		 437 return commit_creds(new);
438
439error:
440 abort_creds(new);
441 return retval;
1da177e4 442}
54e99124 443
e530dca5
DB		 444SYSCALL_DEFINE1(setgid, gid_t, gid)
445{
446 return __sys_setgid(gid);
447}
448
d84f4f99
DH		 449/*
450 * change the user struct in a credentials set to match the new UID
451 */
452static int set_user(struct cred *new)
1da177e4
LT		 453{
454 struct user_struct *new_user;
455
078de5f7 456 new_user = alloc_uid(new->uid);
1da177e4
LT		 457 if (!new_user)
458 return -EAGAIN;
459
72fa5997
VK		 460 /*
461 * We don't fail in case of NPROC limit excess here because too many
462 * poorly written programs don't check set*uid() return code, assuming
463 * it never fails if called by root. We may still enforce NPROC limit
464 * for programs doing set*uid()+execve() by harmlessly deferring the
465 * failure to the execve() stage.
466 */
78d7d407 467 if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
72fa5997
VK		 468 new_user != INIT_USER)
469 current->flags |= PF_NPROC_EXCEEDED;
470 else
471 current->flags &= ~PF_NPROC_EXCEEDED;
1da177e4 472
d84f4f99
DH		 473 free_uid(new->user);
474 new->user = new_user;
1da177e4
LT		 475 return 0;
476}
477
478/*
479 * Unprivileged users may change the real uid to the effective uid
480 * or vice versa. (BSD-style)
481 *
482 * If you set the real uid at all, or set the effective uid to a value not
483 * equal to the real uid, then the saved uid is set to the new effective uid.
484 *
485 * This makes it possible for a setuid program to completely drop its
486 * privileges, which is often a useful assertion to make when you are doing
487 * a security audit over a program.
488 *
489 * The general idea is that a program which uses just setreuid() will be
490 * 100% compatible with BSD. A program which uses just setuid() will be
ec94fc3d 491 * 100% compatible with POSIX with saved IDs.
1da177e4 492 */
e530dca5 493long __sys_setreuid(uid_t ruid, uid_t euid)
1da177e4 494{
a29c33f4 495 struct user_namespace *ns = current_user_ns();
d84f4f99
DH		 496 const struct cred *old;
497 struct cred *new;
1da177e4 498 int retval;
a29c33f4
EB		 499 kuid_t kruid, keuid;
500
501 kruid = make_kuid(ns, ruid);
502 keuid = make_kuid(ns, euid);
503
504 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
505 return -EINVAL;
506 if ((euid != (uid_t) -1) && !uid_valid(keuid))
507 return -EINVAL;
1da177e4 508
d84f4f99
DH		 509 new = prepare_creds();
510 if (!new)
511 return -ENOMEM;
512 old = current_cred();
513
d84f4f99 514 retval = -EPERM;
1da177e4 515 if (ruid != (uid_t) -1) {
a29c33f4
EB		 516 new->uid = kruid;
517 if (!uid_eq(old->uid, kruid) &&
518 !uid_eq(old->euid, kruid) &&
c7b96acf 519 !ns_capable(old->user_ns, CAP_SETUID))
d84f4f99 520 goto error;
1da177e4
LT		 521 }
522
523 if (euid != (uid_t) -1) {
a29c33f4
EB		 524 new->euid = keuid;
525 if (!uid_eq(old->uid, keuid) &&
526 !uid_eq(old->euid, keuid) &&
527 !uid_eq(old->suid, keuid) &&
c7b96acf 528 !ns_capable(old->user_ns, CAP_SETUID))
d84f4f99 529 goto error;
1da177e4
LT		 530 }
531
a29c33f4 532 if (!uid_eq(new->uid, old->uid)) {
54e99124
DG		 533 retval = set_user(new);
534 if (retval < 0)
535 goto error;
536 }
1da177e4 537 if (ruid != (uid_t) -1 ||
a29c33f4 538 (euid != (uid_t) -1 && !uid_eq(keuid, old->uid)))
d84f4f99
DH		 539 new->suid = new->euid;
540 new->fsuid = new->euid;
1da177e4 541
d84f4f99
DH		 542 retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
543 if (retval < 0)
544 goto error;
1da177e4 545
d84f4f99 546 return commit_creds(new);
1da177e4 547
d84f4f99
DH		 548error:
549 abort_creds(new);
550 return retval;
551}
ec94fc3d 552
e530dca5
DB		 553SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
554{
555 return __sys_setreuid(ruid, euid);
556}
557
1da177e4 558/*
ec94fc3d 559 * setuid() is implemented like SysV with SAVED_IDS
560 *
1da177e4 561 * Note that SAVED_ID's is deficient in that a setuid root program
ec94fc3d 562 * like sendmail, for example, cannot set its uid to be a normal
1da177e4
LT		 563 * user and then switch back, because if you're root, setuid() sets
564 * the saved uid too. If you don't like this, blame the bright people
565 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
566 * will allow a root program to temporarily drop privileges and be able to
ec94fc3d 567 * regain them by swapping the real and effective uid.
1da177e4 568 */
e530dca5 569long __sys_setuid(uid_t uid)
1da177e4 570{
a29c33f4 571 struct user_namespace *ns = current_user_ns();
d84f4f99
DH		 572 const struct cred *old;
573 struct cred *new;
1da177e4 574 int retval;
a29c33f4
EB		 575 kuid_t kuid;
576
577 kuid = make_kuid(ns, uid);
578 if (!uid_valid(kuid))
579 return -EINVAL;
1da177e4 580
d84f4f99
DH		 581 new = prepare_creds();
582 if (!new)
583 return -ENOMEM;
584 old = current_cred();
585
d84f4f99 586 retval = -EPERM;
c7b96acf 587 if (ns_capable(old->user_ns, CAP_SETUID)) {
a29c33f4
EB		 588 new->suid = new->uid = kuid;
589 if (!uid_eq(kuid, old->uid)) {
54e99124
DG		 590 retval = set_user(new);
591 if (retval < 0)
592 goto error;
d84f4f99 593 }
a29c33f4 594 } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) {
d84f4f99 595 goto error;
1da177e4 596 }
1da177e4 597
a29c33f4 598 new->fsuid = new->euid = kuid;
d84f4f99
DH		 599
600 retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
601 if (retval < 0)
602 goto error;
1da177e4 603
d84f4f99 604 return commit_creds(new);
1da177e4 605
d84f4f99
DH		 606error:
607 abort_creds(new);
608 return retval;
1da177e4
LT		 609}
610
e530dca5
DB		 611SYSCALL_DEFINE1(setuid, uid_t, uid)
612{
613 return __sys_setuid(uid);
614}
615
1da177e4
LT		 616
617/*
618 * This function implements a generic ability to update ruid, euid,
619 * and suid. This allows you to implement the 4.4 compatible seteuid().
620 */
e530dca5 621long __sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
1da177e4 622{
a29c33f4 623 struct user_namespace *ns = current_user_ns();
d84f4f99
DH		 624 const struct cred *old;
625 struct cred *new;
1da177e4 626 int retval;
a29c33f4
EB		 627 kuid_t kruid, keuid, ksuid;
628
629 kruid = make_kuid(ns, ruid);
630 keuid = make_kuid(ns, euid);
631 ksuid = make_kuid(ns, suid);
632
633 if ((ruid != (uid_t) -1) && !uid_valid(kruid))
634 return -EINVAL;
635
636 if ((euid != (uid_t) -1) && !uid_valid(keuid))
637 return -EINVAL;
638
639 if ((suid != (uid_t) -1) && !uid_valid(ksuid))
640 return -EINVAL;
1da177e4 641
d84f4f99
DH		 642 new = prepare_creds();
643 if (!new)
644 return -ENOMEM;
645
d84f4f99 646 old = current_cred();
1da177e4 647
d84f4f99 648 retval = -EPERM;
c7b96acf 649 if (!ns_capable(old->user_ns, CAP_SETUID)) {
a29c33f4
EB		 650 if (ruid != (uid_t) -1 && !uid_eq(kruid, old->uid) &&
651 !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid))
d84f4f99 652 goto error;
a29c33f4
EB		 653 if (euid != (uid_t) -1 && !uid_eq(keuid, old->uid) &&
654 !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid))
d84f4f99 655 goto error;
a29c33f4
EB		 656 if (suid != (uid_t) -1 && !uid_eq(ksuid, old->uid) &&
657 !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid))
d84f4f99 658 goto error;
1da177e4 659 }
d84f4f99 660
1da177e4 661 if (ruid != (uid_t) -1) {
a29c33f4
EB		 662 new->uid = kruid;
663 if (!uid_eq(kruid, old->uid)) {
54e99124
DG		 664 retval = set_user(new);
665 if (retval < 0)
666 goto error;
667 }
1da177e4 668 }
d84f4f99 669 if (euid != (uid_t) -1)
a29c33f4 670 new->euid = keuid;
1da177e4 671 if (suid != (uid_t) -1)
a29c33f4 672 new->suid = ksuid;
d84f4f99 673 new->fsuid = new->euid;
1da177e4 674
d84f4f99
DH		 675 retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
676 if (retval < 0)
677 goto error;
1da177e4 678
d84f4f99 679 return commit_creds(new);
1da177e4 680
d84f4f99
DH		 681error:
682 abort_creds(new);
683 return retval;
1da177e4
LT		 684}
685
e530dca5
DB		 686SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
687{
688 return __sys_setresuid(ruid, euid, suid);
689}
690
a29c33f4 691SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
1da177e4 692{
86a264ab 693 const struct cred *cred = current_cred();
1da177e4 694 int retval;
a29c33f4
EB		 695 uid_t ruid, euid, suid;
696
697 ruid = from_kuid_munged(cred->user_ns, cred->uid);
698 euid = from_kuid_munged(cred->user_ns, cred->euid);
699 suid = from_kuid_munged(cred->user_ns, cred->suid);
1da177e4 700
ec94fc3d 701 retval = put_user(ruid, ruidp);
702 if (!retval) {
703 retval = put_user(euid, euidp);
704 if (!retval)
705 return put_user(suid, suidp);
706 }
1da177e4
LT		 707 return retval;
708}
709
710/*
711 * Same as above, but for rgid, egid, sgid.
712 */
e530dca5 713long __sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
1da177e4 714{
a29c33f4 715 struct user_namespace *ns = current_user_ns();
d84f4f99
DH		 716 const struct cred *old;
717 struct cred *new;
1da177e4 718 int retval;
a29c33f4
EB		 719 kgid_t krgid, kegid, ksgid;
720
721 krgid = make_kgid(ns, rgid);
722 kegid = make_kgid(ns, egid);
723 ksgid = make_kgid(ns, sgid);
724
725 if ((rgid != (gid_t) -1) && !gid_valid(krgid))
726 return -EINVAL;
727 if ((egid != (gid_t) -1) && !gid_valid(kegid))
728 return -EINVAL;
729 if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
730 return -EINVAL;
1da177e4 731
d84f4f99
DH		 732 new = prepare_creds();
733 if (!new)
734 return -ENOMEM;
735 old = current_cred();
736
d84f4f99 737 retval = -EPERM;
c7b96acf 738 if (!ns_capable(old->user_ns, CAP_SETGID)) {
a29c33f4
EB		 739 if (rgid != (gid_t) -1 && !gid_eq(krgid, old->gid) &&
740 !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid))
d84f4f99 741 goto error;
a29c33f4
EB		 742 if (egid != (gid_t) -1 && !gid_eq(kegid, old->gid) &&
743 !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid))
d84f4f99 744 goto error;
a29c33f4
EB		 745 if (sgid != (gid_t) -1 && !gid_eq(ksgid, old->gid) &&
746 !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid))
d84f4f99 747 goto error;
1da177e4 748 }
d84f4f99 749
1da177e4 750 if (rgid != (gid_t) -1)
a29c33f4 751 new->gid = krgid;
d84f4f99 752 if (egid != (gid_t) -1)
a29c33f4 753 new->egid = kegid;
1da177e4 754 if (sgid != (gid_t) -1)
a29c33f4 755 new->sgid = ksgid;
d84f4f99 756 new->fsgid = new->egid;
1da177e4 757
d84f4f99
DH		 758 return commit_creds(new);
759
760error:
761 abort_creds(new);
762 return retval;
1da177e4
LT		 763}
764
e530dca5
DB		 765SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
766{
767 return __sys_setresgid(rgid, egid, sgid);
768}
769
a29c33f4 770SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
1da177e4 771{
86a264ab 772 const struct cred *cred = current_cred();
1da177e4 773 int retval;
a29c33f4
EB		 774 gid_t rgid, egid, sgid;
775
776 rgid = from_kgid_munged(cred->user_ns, cred->gid);
777 egid = from_kgid_munged(cred->user_ns, cred->egid);
778 sgid = from_kgid_munged(cred->user_ns, cred->sgid);
1da177e4 779
ec94fc3d 780 retval = put_user(rgid, rgidp);
781 if (!retval) {
782 retval = put_user(egid, egidp);
783 if (!retval)
784 retval = put_user(sgid, sgidp);
785 }
1da177e4
LT		 786
787 return retval;
788}
789
790
791/*
792 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
793 * is used for "access()" and for the NFS daemon (letting nfsd stay at
794 * whatever uid it wants to). It normally shadows "euid", except when
795 * explicitly set by setfsuid() or for access..
796 */
e530dca5 797long __sys_setfsuid(uid_t uid)
1da177e4 798{
d84f4f99
DH		 799 const struct cred *old;
800 struct cred *new;
801 uid_t old_fsuid;
a29c33f4
EB		 802 kuid_t kuid;
803
804 old = current_cred();
805 old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
806
807 kuid = make_kuid(old->user_ns, uid);
808 if (!uid_valid(kuid))
809 return old_fsuid;
1da177e4 810
d84f4f99
DH		 811 new = prepare_creds();
812 if (!new)
a29c33f4 813 return old_fsuid;
1da177e4 814
a29c33f4
EB		 815 if (uid_eq(kuid, old->uid) || uid_eq(kuid, old->euid) ||
816 uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
c7b96acf 817 ns_capable(old->user_ns, CAP_SETUID)) {
a29c33f4
EB		 818 if (!uid_eq(kuid, old->fsuid)) {
819 new->fsuid = kuid;
d84f4f99
DH		 820 if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
821 goto change_okay;
1da177e4 822 }
1da177e4
LT		 823 }
824
d84f4f99
DH		 825 abort_creds(new);
826 return old_fsuid;
1da177e4 827
d84f4f99
DH		 828change_okay:
829 commit_creds(new);
1da177e4
LT		 830 return old_fsuid;
831}
832
e530dca5
DB		 833SYSCALL_DEFINE1(setfsuid, uid_t, uid)
834{
835 return __sys_setfsuid(uid);
836}
837
1da177e4 838/*
f42df9e6 839 * Samma på svenska..
1da177e4 840 */
e530dca5 841long __sys_setfsgid(gid_t gid)
1da177e4 842{
d84f4f99
DH		 843 const struct cred *old;
844 struct cred *new;
845 gid_t old_fsgid;
a29c33f4
EB		 846 kgid_t kgid;
847
848 old = current_cred();
849 old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
850
851 kgid = make_kgid(old->user_ns, gid);
852 if (!gid_valid(kgid))
853 return old_fsgid;
d84f4f99
DH		 854
855 new = prepare_creds();
856 if (!new)
a29c33f4 857 return old_fsgid;
1da177e4 858
a29c33f4
EB		 859 if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->egid) ||
860 gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
c7b96acf 861 ns_capable(old->user_ns, CAP_SETGID)) {
a29c33f4
EB		 862 if (!gid_eq(kgid, old->fsgid)) {
863 new->fsgid = kgid;
d84f4f99 864 goto change_okay;
1da177e4 865 }
1da177e4 866 }
d84f4f99 867
d84f4f99
DH		 868 abort_creds(new);
869 return old_fsgid;
870
871change_okay:
872 commit_creds(new);
1da177e4
LT		 873 return old_fsgid;
874}
e530dca5
DB		 875
876SYSCALL_DEFINE1(setfsgid, gid_t, gid)
877{
878 return __sys_setfsgid(gid);
879}
2813893f 880#endif /* CONFIG_MULTIUSER */
1da177e4 881
4a22f166
SR		 882/**
883 * sys_getpid - return the thread group id of the current process
884 *
885 * Note, despite the name, this returns the tgid not the pid. The tgid and
886 * the pid are identical unless CLONE_THREAD was specified on clone() in
887 * which case the tgid is the same in all threads of the same group.
888 *
889 * This is SMP safe as current->tgid does not change.
890 */
891SYSCALL_DEFINE0(getpid)
892{
893 return task_tgid_vnr(current);
894}
895
896/* Thread ID - the internal kernel "pid" */
897SYSCALL_DEFINE0(gettid)
898{
899 return task_pid_vnr(current);
900}
901
902/*
903 * Accessing ->real_parent is not SMP-safe, it could
904 * change from under us. However, we can use a stale
905 * value of ->real_parent under rcu_read_lock(), see
906 * release_task()->call_rcu(delayed_put_task_struct).
907 */
908SYSCALL_DEFINE0(getppid)
909{
910 int pid;
911
912 rcu_read_lock();
913 pid = task_tgid_vnr(rcu_dereference(current->real_parent));
914 rcu_read_unlock();
915
916 return pid;
917}
918
919SYSCALL_DEFINE0(getuid)
920{
921 /* Only we change this so SMP safe */
922 return from_kuid_munged(current_user_ns(), current_uid());
923}
924
925SYSCALL_DEFINE0(geteuid)
926{
927 /* Only we change this so SMP safe */
928 return from_kuid_munged(current_user_ns(), current_euid());
929}
930
931SYSCALL_DEFINE0(getgid)
932{
933 /* Only we change this so SMP safe */
934 return from_kgid_munged(current_user_ns(), current_gid());
935}
936
937SYSCALL_DEFINE0(getegid)
938{
939 /* Only we change this so SMP safe */
940 return from_kgid_munged(current_user_ns(), current_egid());
941}
942
ca2406ed 943static void do_sys_times(struct tms *tms)
f06febc9 944{
5613fda9 945 u64 tgutime, tgstime, cutime, cstime;
f06febc9 946
e80d0a1a 947 thread_group_cputime_adjusted(current, &tgutime, &tgstime);
f06febc9
FM		 948 cutime = current->signal->cutime;
949 cstime = current->signal->cstime;
5613fda9
FW		 950 tms->tms_utime = nsec_to_clock_t(tgutime);
951 tms->tms_stime = nsec_to_clock_t(tgstime);
952 tms->tms_cutime = nsec_to_clock_t(cutime);
953 tms->tms_cstime = nsec_to_clock_t(cstime);
f06febc9
FM		 954}
955
58fd3aa2 956SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
1da177e4 957{
1da177e4
LT		 958 if (tbuf) {
959 struct tms tmp;
f06febc9
FM		 960
961 do_sys_times(&tmp);
1da177e4
LT		 962 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
963 return -EFAULT;
964 }
e3d5a27d 965 force_successful_syscall_return();
1da177e4
LT		 966 return (long) jiffies_64_to_clock_t(get_jiffies_64());
967}
968
ca2406ed
AV		 969#ifdef CONFIG_COMPAT
970static compat_clock_t clock_t_to_compat_clock_t(clock_t x)
971{
972 return compat_jiffies_to_clock_t(clock_t_to_jiffies(x));
973}
974
975COMPAT_SYSCALL_DEFINE1(times, struct compat_tms __user *, tbuf)
976{
977 if (tbuf) {
978 struct tms tms;
979 struct compat_tms tmp;
980
981 do_sys_times(&tms);
982 /* Convert our struct tms to the compat version. */
983 tmp.tms_utime = clock_t_to_compat_clock_t(tms.tms_utime);
984 tmp.tms_stime = clock_t_to_compat_clock_t(tms.tms_stime);
985 tmp.tms_cutime = clock_t_to_compat_clock_t(tms.tms_cutime);
986 tmp.tms_cstime = clock_t_to_compat_clock_t(tms.tms_cstime);
987 if (copy_to_user(tbuf, &tmp, sizeof(tmp)))
988 return -EFAULT;
989 }
990 force_successful_syscall_return();
991 return compat_jiffies_to_clock_t(jiffies);
992}
993#endif
994
1da177e4
LT		 995/*
996 * This needs some heavy checking ...
997 * I just haven't the stomach for it. I also don't fully
998 * understand sessions/pgrp etc. Let somebody who does explain it.
999 *
1000 * OK, I think I have the protection semantics right.... this is really
1001 * only important on a multi-user system anyway, to make sure one user
1002 * can't send a signal to a process owned by another. -TYT, 12/12/91
1003 *
98611e4e 1004 * !PF_FORKNOEXEC check to conform completely to POSIX.
1da177e4 1005 */
b290ebe2 1006SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
1da177e4
LT		 1007{
1008 struct task_struct *p;
ee0acf90 1009 struct task_struct *group_leader = current->group_leader;
4e021306
ON		 1010 struct pid *pgrp;
1011 int err;
1da177e4
LT		 1012
1013 if (!pid)
b488893a 1014 pid = task_pid_vnr(group_leader);
1da177e4
LT		 1015 if (!pgid)
1016 pgid = pid;
1017 if (pgid < 0)
1018 return -EINVAL;
950eaaca 1019 rcu_read_lock();
1da177e4
LT		 1020
1021 /* From this point forward we keep holding onto the tasklist lock
1022 * so that our parent does not change from under us. -DaveM
1023 */
1024 write_lock_irq(&tasklist_lock);
1025
1026 err = -ESRCH;
4e021306 1027 p = find_task_by_vpid(pid);
1da177e4
LT		 1028 if (!p)
1029 goto out;
1030
1031 err = -EINVAL;
1032 if (!thread_group_leader(p))
1033 goto out;
1034
4e021306 1035 if (same_thread_group(p->real_parent, group_leader)) {
1da177e4 1036 err = -EPERM;
41487c65 1037 if (task_session(p) != task_session(group_leader))
1da177e4
LT		 1038 goto out;
1039 err = -EACCES;
98611e4e 1040 if (!(p->flags & PF_FORKNOEXEC))
1da177e4
LT		 1041 goto out;
1042 } else {
1043 err = -ESRCH;
ee0acf90 1044 if (p != group_leader)
1da177e4
LT		 1045 goto out;
1046 }
1047
1048 err = -EPERM;
1049 if (p->signal->leader)
1050 goto out;
1051
4e021306 1052 pgrp = task_pid(p);
1da177e4 1053 if (pgid != pid) {
b488893a 1054 struct task_struct *g;
1da177e4 1055
4e021306
ON		 1056 pgrp = find_vpid(pgid);
1057 g = pid_task(pgrp, PIDTYPE_PGID);
41487c65 1058 if (!g || task_session(g) != task_session(group_leader))
f020bc46 1059 goto out;
1da177e4
LT		 1060 }
1061
1da177e4
LT		 1062 err = security_task_setpgid(p, pgid);
1063 if (err)
1064 goto out;
1065
1b0f7ffd 1066 if (task_pgrp(p) != pgrp)
83beaf3c 1067 change_pid(p, PIDTYPE_PGID, pgrp);
1da177e4
LT		 1068
1069 err = 0;
1070out:
1071 /* All paths lead to here, thus we are safe. -DaveM */
1072 write_unlock_irq(&tasklist_lock);
950eaaca 1073 rcu_read_unlock();
1da177e4
LT		 1074 return err;
1075}
1076
192c5807 1077static int do_getpgid(pid_t pid)
1da177e4 1078{
12a3de0a
ON		 1079 struct task_struct *p;
1080 struct pid *grp;
1081 int retval;
1082
1083 rcu_read_lock();
756184b7 1084 if (!pid)
12a3de0a 1085 grp = task_pgrp(current);
756184b7 1086 else {
1da177e4 1087 retval = -ESRCH;
12a3de0a
ON		 1088 p = find_task_by_vpid(pid);
1089 if (!p)
1090 goto out;
1091 grp = task_pgrp(p);
1092 if (!grp)
1093 goto out;
1094
1095 retval = security_task_getpgid(p);
1096 if (retval)
1097 goto out;
1da177e4 1098 }
12a3de0a
ON		 1099 retval = pid_vnr(grp);
1100out:
1101 rcu_read_unlock();
1102 return retval;
1da177e4
LT		 1103}
1104
192c5807
DB		 1105SYSCALL_DEFINE1(getpgid, pid_t, pid)
1106{
1107 return do_getpgid(pid);
1108}
1109
1da177e4
LT		 1110#ifdef __ARCH_WANT_SYS_GETPGRP
1111
dbf040d9 1112SYSCALL_DEFINE0(getpgrp)
1da177e4 1113{
192c5807 1114 return do_getpgid(0);
1da177e4
LT		 1115}
1116
1117#endif
1118
dbf040d9 1119SYSCALL_DEFINE1(getsid, pid_t, pid)
1da177e4 1120{
1dd768c0
ON		 1121 struct task_struct *p;
1122 struct pid *sid;
1123 int retval;
1124
1125 rcu_read_lock();
756184b7 1126 if (!pid)
1dd768c0 1127 sid = task_session(current);
756184b7 1128 else {
1da177e4 1129 retval = -ESRCH;
1dd768c0
ON		 1130 p = find_task_by_vpid(pid);
1131 if (!p)
1132 goto out;
1133 sid = task_session(p);
1134 if (!sid)
1135 goto out;
1136
1137 retval = security_task_getsid(p);
1138 if (retval)
1139 goto out;
1da177e4 1140 }
1dd768c0
ON		 1141 retval = pid_vnr(sid);
1142out:
1143 rcu_read_unlock();
1144 return retval;
1da177e4
LT		 1145}
1146
81dabb46
ON		 1147static void set_special_pids(struct pid *pid)
1148{
1149 struct task_struct *curr = current->group_leader;
1150
1151 if (task_session(curr) != pid)
1152 change_pid(curr, PIDTYPE_SID, pid);
1153
1154 if (task_pgrp(curr) != pid)
1155 change_pid(curr, PIDTYPE_PGID, pid);
1156}
1157
e2aaa9f4 1158int ksys_setsid(void)
1da177e4 1159{
e19f247a 1160 struct task_struct *group_leader = current->group_leader;
e4cc0a9c
ON		 1161 struct pid *sid = task_pid(group_leader);
1162 pid_t session = pid_vnr(sid);
1da177e4
LT		 1163 int err = -EPERM;
1164
1da177e4 1165 write_lock_irq(&tasklist_lock);
390e2ff0
EB		 1166 /* Fail if I am already a session leader */
1167 if (group_leader->signal->leader)
1168 goto out;
1169
430c6231
ON		 1170 /* Fail if a process group id already exists that equals the
1171 * proposed session id.
390e2ff0 1172 */
6806aac6 1173 if (pid_task(sid, PIDTYPE_PGID))
1da177e4
LT		 1174 goto out;
1175
e19f247a 1176 group_leader->signal->leader = 1;
81dabb46 1177 set_special_pids(sid);
24ec839c 1178
9c9f4ded 1179 proc_clear_tty(group_leader);
24ec839c 1180
e4cc0a9c 1181 err = session;
1da177e4
LT		 1182out:
1183 write_unlock_irq(&tasklist_lock);
5091faa4 1184 if (err > 0) {
0d0df599 1185 proc_sid_connector(group_leader);
5091faa4
MG		 1186 sched_autogroup_create_attach(group_leader);
1187 }
1da177e4
LT		 1188 return err;
1189}
1190
e2aaa9f4
DB		 1191SYSCALL_DEFINE0(setsid)
1192{
1193 return ksys_setsid();
1194}
1195
1da177e4
LT		 1196DECLARE_RWSEM(uts_sem);
1197
26ca697f
AW		 1198#ifdef COMPAT_UTS_MACHINE
1199static char compat_uts_machine[__OLD_UTS_LEN+1] = COMPAT_UTS_MACHINE;
1200
1201static int __init parse_compat_uts_machine(char *arg)
1202{
1203 strncpy(compat_uts_machine, arg, __OLD_UTS_LEN);
1204 compat_uts_machine[__OLD_UTS_LEN] = 0;
1205 return 0;
1206}
1207early_param("compat_uts_machine", parse_compat_uts_machine);
1208
1209#undef COMPAT_UTS_MACHINE
1210#define COMPAT_UTS_MACHINE compat_uts_machine
1211#endif
1212
e28cbf22
CH		 1213#ifdef COMPAT_UTS_MACHINE
1214#define override_architecture(name) \
46da2766 1215 (personality(current->personality) == PER_LINUX32 && \
e28cbf22
CH		 1216 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1217 sizeof(COMPAT_UTS_MACHINE)))
1218#else
1219#define override_architecture(name) 0
1220#endif
1221
be27425d
AK		 1222/*
1223 * Work around broken programs that cannot handle "Linux 3.0".
1224 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
b7285b42
JN		 1225 * And we map 4.x and later versions to 2.6.60+x, so 4.0/5.0/6.0/... would be
1226 * 2.6.60.
be27425d 1227 */
2702b152 1228static int override_release(char __user *release, size_t len)
be27425d
AK		 1229{
1230 int ret = 0;
be27425d
AK		 1231
1232 if (current->personality & UNAME26) {
2702b152
KC		 1233 const char *rest = UTS_RELEASE;
1234 char buf[65] = { 0 };
be27425d
AK		 1235 int ndots = 0;
1236 unsigned v;
2702b152 1237 size_t copy;
be27425d
AK		 1238
1239 while (*rest) {
1240 if (*rest == '.' && ++ndots >= 3)
1241 break;
1242 if (!isdigit(*rest) && *rest != '.')
1243 break;
1244 rest++;
1245 }
39afb5ee 1246 v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 60;
31fd84b9 1247 copy = clamp_t(size_t, len, 1, sizeof(buf));
2702b152
KC		 1248 copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1249 ret = copy_to_user(release, buf, copy + 1);
be27425d
AK		 1250 }
1251 return ret;
1252}
1253
e48fbb69 1254SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1da177e4 1255{
42a0cc34 1256 struct new_utsname tmp;
1da177e4
LT		 1257
1258 down_read(&uts_sem);
42a0cc34 1259 memcpy(&tmp, utsname(), sizeof(tmp));
1da177e4 1260 up_read(&uts_sem);
42a0cc34
JH		 1261 if (copy_to_user(name, &tmp, sizeof(tmp)))
1262 return -EFAULT;
e28cbf22 1263
42a0cc34
JH		 1264 if (override_release(name->release, sizeof(name->release)))
1265 return -EFAULT;
1266 if (override_architecture(name))
1267 return -EFAULT;
1268 return 0;
1da177e4
LT		 1269}
1270
5cacdb4a
CH		 1271#ifdef __ARCH_WANT_SYS_OLD_UNAME
1272/*
1273 * Old cruft
1274 */
1275SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1276{
42a0cc34 1277 struct old_utsname tmp;
5cacdb4a
CH		 1278
1279 if (!name)
1280 return -EFAULT;
1281
1282 down_read(&uts_sem);
42a0cc34 1283 memcpy(&tmp, utsname(), sizeof(tmp));
5cacdb4a 1284 up_read(&uts_sem);
42a0cc34
JH		 1285 if (copy_to_user(name, &tmp, sizeof(tmp)))
1286 return -EFAULT;
5cacdb4a 1287
42a0cc34
JH		 1288 if (override_release(name->release, sizeof(name->release)))
1289 return -EFAULT;
1290 if (override_architecture(name))
1291 return -EFAULT;
1292 return 0;
5cacdb4a
CH		 1293}
1294
1295SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1296{
42a0cc34 1297 struct oldold_utsname tmp = {};
5cacdb4a
CH		 1298
1299 if (!name)
1300 return -EFAULT;
5cacdb4a
CH		 1301
1302 down_read(&uts_sem);
42a0cc34
JH		 1303 memcpy(&tmp.sysname, &utsname()->sysname, __OLD_UTS_LEN);
1304 memcpy(&tmp.nodename, &utsname()->nodename, __OLD_UTS_LEN);
1305 memcpy(&tmp.release, &utsname()->release, __OLD_UTS_LEN);
1306 memcpy(&tmp.version, &utsname()->version, __OLD_UTS_LEN);
1307 memcpy(&tmp.machine, &utsname()->machine, __OLD_UTS_LEN);
5cacdb4a 1308 up_read(&uts_sem);
42a0cc34
JH		 1309 if (copy_to_user(name, &tmp, sizeof(tmp)))
1310 return -EFAULT;
5cacdb4a 1311
42a0cc34
JH		 1312 if (override_architecture(name))
1313 return -EFAULT;
1314 if (override_release(name->release, sizeof(name->release)))
1315 return -EFAULT;
1316 return 0;
5cacdb4a
CH		 1317}
1318#endif
1319
5a8a82b1 1320SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1da177e4
LT		 1321{
1322 int errno;
1323 char tmp[__NEW_UTS_LEN];
1324
bb96a6f5 1325 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1da177e4 1326 return -EPERM;
fc832ad3 1327
1da177e4
LT		 1328 if (len < 0 || len > __NEW_UTS_LEN)
1329 return -EINVAL;
1da177e4
LT		 1330 errno = -EFAULT;
1331 if (!copy_from_user(tmp, name, len)) {
42a0cc34 1332 struct new_utsname *u;
9679e4dd 1333
42a0cc34
JH		 1334 down_write(&uts_sem);
1335 u = utsname();
9679e4dd
AM		 1336 memcpy(u->nodename, tmp, len);
1337 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1da177e4 1338 errno = 0;
499eea6b 1339 uts_proc_notify(UTS_PROC_HOSTNAME);
42a0cc34 1340 up_write(&uts_sem);
1da177e4 1341 }
1da177e4
LT		 1342 return errno;
1343}
1344
1345#ifdef __ARCH_WANT_SYS_GETHOSTNAME
1346
5a8a82b1 1347SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1da177e4 1348{
42a0cc34 1349 int i;
9679e4dd 1350 struct new_utsname *u;
42a0cc34 1351 char tmp[__NEW_UTS_LEN + 1];
1da177e4
LT		 1352
1353 if (len < 0)
1354 return -EINVAL;
1355 down_read(&uts_sem);
9679e4dd
AM		 1356 u = utsname();
1357 i = 1 + strlen(u->nodename);
1da177e4
LT		 1358 if (i > len)
1359 i = len;
42a0cc34 1360 memcpy(tmp, u->nodename, i);
1da177e4 1361 up_read(&uts_sem);
42a0cc34
JH		 1362 if (copy_to_user(name, tmp, i))
1363 return -EFAULT;
1364 return 0;
1da177e4
LT		 1365}
1366
1367#endif
1368
1369/*
1370 * Only setdomainname; getdomainname can be implemented by calling
1371 * uname()
1372 */
5a8a82b1 1373SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1da177e4
LT		 1374{
1375 int errno;
1376 char tmp[__NEW_UTS_LEN];
1377
fc832ad3 1378 if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1da177e4
LT		 1379 return -EPERM;
1380 if (len < 0 || len > __NEW_UTS_LEN)
1381 return -EINVAL;
1382
1da177e4
LT		 1383 errno = -EFAULT;
1384 if (!copy_from_user(tmp, name, len)) {
42a0cc34 1385 struct new_utsname *u;
9679e4dd 1386
42a0cc34
JH		 1387 down_write(&uts_sem);
1388 u = utsname();
9679e4dd
AM		 1389 memcpy(u->domainname, tmp, len);
1390 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1da177e4 1391 errno = 0;
499eea6b 1392 uts_proc_notify(UTS_PROC_DOMAINNAME);
42a0cc34 1393 up_write(&uts_sem);
1da177e4 1394 }
1da177e4
LT		 1395 return errno;
1396}
1397
e48fbb69 1398SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1da177e4 1399{
b9518345
JS		 1400 struct rlimit value;
1401 int ret;
1402
1403 ret = do_prlimit(current, resource, NULL, &value);
1404 if (!ret)
1405 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1406
1407 return ret;
1da177e4
LT		 1408}
1409
d9e968cb
AV		 1410#ifdef CONFIG_COMPAT
1411
1412COMPAT_SYSCALL_DEFINE2(setrlimit, unsigned int, resource,
1413 struct compat_rlimit __user *, rlim)
1414{
1415 struct rlimit r;
1416 struct compat_rlimit r32;
1417
1418 if (copy_from_user(&r32, rlim, sizeof(struct compat_rlimit)))
1419 return -EFAULT;
1420
1421 if (r32.rlim_cur == COMPAT_RLIM_INFINITY)
1422 r.rlim_cur = RLIM_INFINITY;
1423 else
1424 r.rlim_cur = r32.rlim_cur;
1425 if (r32.rlim_max == COMPAT_RLIM_INFINITY)
1426 r.rlim_max = RLIM_INFINITY;
1427 else
1428 r.rlim_max = r32.rlim_max;
1429 return do_prlimit(current, resource, &r, NULL);
1430}
1431
1432COMPAT_SYSCALL_DEFINE2(getrlimit, unsigned int, resource,
1433 struct compat_rlimit __user *, rlim)
1434{
1435 struct rlimit r;
1436 int ret;
1437
1438 ret = do_prlimit(current, resource, NULL, &r);
1439 if (!ret) {
58c7ffc0 1440 struct compat_rlimit r32;
d9e968cb
AV		 1441 if (r.rlim_cur > COMPAT_RLIM_INFINITY)
1442 r32.rlim_cur = COMPAT_RLIM_INFINITY;
1443 else
1444 r32.rlim_cur = r.rlim_cur;
1445 if (r.rlim_max > COMPAT_RLIM_INFINITY)
1446 r32.rlim_max = COMPAT_RLIM_INFINITY;
1447 else
1448 r32.rlim_max = r.rlim_max;
1449
1450 if (copy_to_user(rlim, &r32, sizeof(struct compat_rlimit)))
1451 return -EFAULT;
1452 }
1453 return ret;
1454}
1455
1456#endif
1457
1da177e4
LT		 1458#ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1459
1460/*
1461 * Back compatibility for getrlimit. Needed for some apps.
1462 */
e48fbb69
HC		 1463SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1464 struct rlimit __user *, rlim)
1da177e4
LT		 1465{
1466 struct rlimit x;
1467 if (resource >= RLIM_NLIMITS)
1468 return -EINVAL;
1469
23d6aef7 1470 resource = array_index_nospec(resource, RLIM_NLIMITS);
1da177e4
LT		 1471 task_lock(current->group_leader);
1472 x = current->signal->rlim[resource];
1473 task_unlock(current->group_leader);
756184b7 1474 if (x.rlim_cur > 0x7FFFFFFF)
1da177e4 1475 x.rlim_cur = 0x7FFFFFFF;
756184b7 1476 if (x.rlim_max > 0x7FFFFFFF)
1da177e4 1477 x.rlim_max = 0x7FFFFFFF;
ec94fc3d 1478 return copy_to_user(rlim, &x, sizeof(x)) ? -EFAULT : 0;
1da177e4
LT		 1479}
1480
613763a1
AV		 1481#ifdef CONFIG_COMPAT
1482COMPAT_SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1483 struct compat_rlimit __user *, rlim)
1484{
1485 struct rlimit r;
1486
1487 if (resource >= RLIM_NLIMITS)
1488 return -EINVAL;
1489
23d6aef7 1490 resource = array_index_nospec(resource, RLIM_NLIMITS);
613763a1
AV		 1491 task_lock(current->group_leader);
1492 r = current->signal->rlim[resource];
1493 task_unlock(current->group_leader);
1494 if (r.rlim_cur > 0x7FFFFFFF)
1495 r.rlim_cur = 0x7FFFFFFF;
1496 if (r.rlim_max > 0x7FFFFFFF)
1497 r.rlim_max = 0x7FFFFFFF;
1498
1499 if (put_user(r.rlim_cur, &rlim->rlim_cur) ||
1500 put_user(r.rlim_max, &rlim->rlim_max))
1501 return -EFAULT;
1502 return 0;
1503}
1504#endif
1505
1da177e4
LT		 1506#endif
1507
c022a0ac
JS		 1508static inline bool rlim64_is_infinity(__u64 rlim64)
1509{
1510#if BITS_PER_LONG < 64
1511 return rlim64 >= ULONG_MAX;
1512#else
1513 return rlim64 == RLIM64_INFINITY;
1514#endif
1515}
1516
1517static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1518{
1519 if (rlim->rlim_cur == RLIM_INFINITY)
1520 rlim64->rlim_cur = RLIM64_INFINITY;
1521 else
1522 rlim64->rlim_cur = rlim->rlim_cur;
1523 if (rlim->rlim_max == RLIM_INFINITY)
1524 rlim64->rlim_max = RLIM64_INFINITY;
1525 else
1526 rlim64->rlim_max = rlim->rlim_max;
1527}
1528
1529static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1530{
1531 if (rlim64_is_infinity(rlim64->rlim_cur))
1532 rlim->rlim_cur = RLIM_INFINITY;
1533 else
1534 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1535 if (rlim64_is_infinity(rlim64->rlim_max))
1536 rlim->rlim_max = RLIM_INFINITY;
1537 else
1538 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1539}
1540
1c1e618d 1541/* make sure you are allowed to change @tsk limits before calling this */
5b41535a
JS		 1542int do_prlimit(struct task_struct *tsk, unsigned int resource,
1543 struct rlimit *new_rlim, struct rlimit *old_rlim)
1da177e4 1544{
5b41535a 1545 struct rlimit *rlim;
86f162f4 1546 int retval = 0;
1da177e4
LT		 1547
1548 if (resource >= RLIM_NLIMITS)
1549 return -EINVAL;
5b41535a
JS		 1550 if (new_rlim) {
1551 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1552 return -EINVAL;
1553 if (resource == RLIMIT_NOFILE &&
1554 new_rlim->rlim_max > sysctl_nr_open)
1555 return -EPERM;
1556 }
1da177e4 1557
1c1e618d
JS		 1558 /* protect tsk->signal and tsk->sighand from disappearing */
1559 read_lock(&tasklist_lock);
1560 if (!tsk->sighand) {
1561 retval = -ESRCH;
1562 goto out;
1563 }
1564
5b41535a 1565 rlim = tsk->signal->rlim + resource;
86f162f4 1566 task_lock(tsk->group_leader);
5b41535a 1567 if (new_rlim) {
fc832ad3
SH		 1568 /* Keep the capable check against init_user_ns until
1569 cgroups can contain all limits */
5b41535a
JS		 1570 if (new_rlim->rlim_max > rlim->rlim_max &&
1571 !capable(CAP_SYS_RESOURCE))
1572 retval = -EPERM;
1573 if (!retval)
cad4ea54 1574 retval = security_task_setrlimit(tsk, resource, new_rlim);
5b41535a
JS		 1575 if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) {
1576 /*
1577 * The caller is asking for an immediate RLIMIT_CPU
1578 * expiry. But we use the zero value to mean "it was
1579 * never set". So let's cheat and make it one second
1580 * instead
1581 */
1582 new_rlim->rlim_cur = 1;
1583 }
1584 }
1585 if (!retval) {
1586 if (old_rlim)
1587 *old_rlim = *rlim;
1588 if (new_rlim)
1589 *rlim = *new_rlim;
9926e4c7 1590 }
7855c35d 1591 task_unlock(tsk->group_leader);
1da177e4 1592
d3561f78
AM		 1593 /*
1594 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1595 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1596 * very long-standing error, and fixing it now risks breakage of
1597 * applications, so we live with it
1598 */
5b41535a 1599 if (!retval && new_rlim && resource == RLIMIT_CPU &&
baa73d9e
NP		 1600 new_rlim->rlim_cur != RLIM_INFINITY &&
1601 IS_ENABLED(CONFIG_POSIX_TIMERS))
5b41535a 1602 update_rlimit_cpu(tsk, new_rlim->rlim_cur);
ec9e16ba 1603out:
1c1e618d 1604 read_unlock(&tasklist_lock);
2fb9d268 1605 return retval;
1da177e4
LT		 1606}
1607
c022a0ac 1608/* rcu lock must be held */
791ec491
SS		 1609static int check_prlimit_permission(struct task_struct *task,
1610 unsigned int flags)
c022a0ac
JS		 1611{
1612 const struct cred *cred = current_cred(), *tcred;
791ec491 1613 bool id_match;
c022a0ac 1614
fc832ad3
SH		 1615 if (current == task)
1616 return 0;
c022a0ac 1617
fc832ad3 1618 tcred = __task_cred(task);
791ec491
SS		 1619 id_match = (uid_eq(cred->uid, tcred->euid) &&
1620 uid_eq(cred->uid, tcred->suid) &&
1621 uid_eq(cred->uid, tcred->uid) &&
1622 gid_eq(cred->gid, tcred->egid) &&
1623 gid_eq(cred->gid, tcred->sgid) &&
1624 gid_eq(cred->gid, tcred->gid));
1625 if (!id_match && !ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
1626 return -EPERM;
fc832ad3 1627
791ec491 1628 return security_task_prlimit(cred, tcred, flags);
c022a0ac
JS		 1629}
1630
1631SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1632 const struct rlimit64 __user *, new_rlim,
1633 struct rlimit64 __user *, old_rlim)
1634{
1635 struct rlimit64 old64, new64;
1636 struct rlimit old, new;
1637 struct task_struct *tsk;
791ec491 1638 unsigned int checkflags = 0;
c022a0ac
JS		 1639 int ret;
1640
791ec491
SS		 1641 if (old_rlim)
1642 checkflags |= LSM_PRLIMIT_READ;
1643
c022a0ac
JS		 1644 if (new_rlim) {
1645 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1646 return -EFAULT;
1647 rlim64_to_rlim(&new64, &new);
791ec491 1648 checkflags |= LSM_PRLIMIT_WRITE;
c022a0ac
JS		 1649 }
1650
1651 rcu_read_lock();
1652 tsk = pid ? find_task_by_vpid(pid) : current;
1653 if (!tsk) {
1654 rcu_read_unlock();
1655 return -ESRCH;
1656 }
791ec491 1657 ret = check_prlimit_permission(tsk, checkflags);
c022a0ac
JS		 1658 if (ret) {
1659 rcu_read_unlock();
1660 return ret;
1661 }
1662 get_task_struct(tsk);
1663 rcu_read_unlock();
1664
1665 ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1666 old_rlim ? &old : NULL);
1667
1668 if (!ret && old_rlim) {
1669 rlim_to_rlim64(&old, &old64);
1670 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1671 ret = -EFAULT;
1672 }
1673
1674 put_task_struct(tsk);
1675 return ret;
1676}
1677
7855c35d
JS		 1678SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1679{
1680 struct rlimit new_rlim;
1681
1682 if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1683 return -EFAULT;
5b41535a 1684 return do_prlimit(current, resource, &new_rlim, NULL);
7855c35d
JS		 1685}
1686
1da177e4
LT		 1687/*
1688 * It would make sense to put struct rusage in the task_struct,
1689 * except that would make the task_struct be *really big*. After
1690 * task_struct gets moved into malloc'ed memory, it would
1691 * make sense to do this. It will make moving the rest of the information
1692 * a lot simpler! (Which we're not doing right now because we're not
1693 * measuring them yet).
1694 *
1da177e4
LT		 1695 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1696 * races with threads incrementing their own counters. But since word
1697 * reads are atomic, we either get new values or old values and we don't
1698 * care which for the sums. We always take the siglock to protect reading
1699 * the c* fields from p->signal from races with exit.c updating those
1700 * fields when reaping, so a sample either gets all the additions of a
1701 * given child after it's reaped, or none so this sample is before reaping.
2dd0ebcd 1702 *
de047c1b
RT		 1703 * Locking:
1704 * We need to take the siglock for CHILDEREN, SELF and BOTH
1705 * for the cases current multithreaded, non-current single threaded
1706 * non-current multithreaded. Thread traversal is now safe with
1707 * the siglock held.
1708 * Strictly speaking, we donot need to take the siglock if we are current and
1709 * single threaded, as no one else can take our signal_struct away, no one
1710 * else can reap the children to update signal->c* counters, and no one else
1711 * can race with the signal-> fields. If we do not take any lock, the
1712 * signal-> fields could be read out of order while another thread was just
1713 * exiting. So we should place a read memory barrier when we avoid the lock.
1714 * On the writer side, write memory barrier is implied in __exit_signal
1715 * as __exit_signal releases the siglock spinlock after updating the signal->
1716 * fields. But we don't do this yet to keep things simple.
2dd0ebcd 1717 *
1da177e4
LT		 1718 */
1719
f06febc9 1720static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
679c9cd4 1721{
679c9cd4
SK		 1722 r->ru_nvcsw += t->nvcsw;
1723 r->ru_nivcsw += t->nivcsw;
1724 r->ru_minflt += t->min_flt;
1725 r->ru_majflt += t->maj_flt;
1726 r->ru_inblock += task_io_get_inblock(t);
1727 r->ru_oublock += task_io_get_oublock(t);
1728}
1729
ce72a16f 1730void getrusage(struct task_struct *p, int who, struct rusage *r)
1da177e4
LT		 1731{
1732 struct task_struct *t;
1733 unsigned long flags;
5613fda9 1734 u64 tgutime, tgstime, utime, stime;
1f10206c 1735 unsigned long maxrss = 0;
1da177e4 1736
ec94fc3d 1737 memset((char *)r, 0, sizeof (*r));
64861634 1738 utime = stime = 0;
1da177e4 1739
679c9cd4 1740 if (who == RUSAGE_THREAD) {
e80d0a1a 1741 task_cputime_adjusted(current, &utime, &stime);
f06febc9 1742 accumulate_thread_rusage(p, r);
1f10206c 1743 maxrss = p->signal->maxrss;
679c9cd4
SK		 1744 goto out;
1745 }
1746
d6cf723a 1747 if (!lock_task_sighand(p, &flags))
de047c1b 1748 return;
0f59cc4a 1749
1da177e4 1750 switch (who) {
ec94fc3d 1751 case RUSAGE_BOTH:
1752 case RUSAGE_CHILDREN:
1753 utime = p->signal->cutime;
1754 stime = p->signal->cstime;
1755 r->ru_nvcsw = p->signal->cnvcsw;
1756 r->ru_nivcsw = p->signal->cnivcsw;
1757 r->ru_minflt = p->signal->cmin_flt;
1758 r->ru_majflt = p->signal->cmaj_flt;
1759 r->ru_inblock = p->signal->cinblock;
1760 r->ru_oublock = p->signal->coublock;
1761 maxrss = p->signal->cmaxrss;
1762
1763 if (who == RUSAGE_CHILDREN)
1da177e4 1764 break;
0f59cc4a 1765
ec94fc3d 1766 case RUSAGE_SELF:
1767 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1768 utime += tgutime;
1769 stime += tgstime;
1770 r->ru_nvcsw += p->signal->nvcsw;
1771 r->ru_nivcsw += p->signal->nivcsw;
1772 r->ru_minflt += p->signal->min_flt;
1773 r->ru_majflt += p->signal->maj_flt;
1774 r->ru_inblock += p->signal->inblock;
1775 r->ru_oublock += p->signal->oublock;
1776 if (maxrss < p->signal->maxrss)
1777 maxrss = p->signal->maxrss;
1778 t = p;
1779 do {
1780 accumulate_thread_rusage(t, r);
1781 } while_each_thread(p, t);
1782 break;
1783
1784 default:
1785 BUG();
1da177e4 1786 }
de047c1b 1787 unlock_task_sighand(p, &flags);
de047c1b 1788
679c9cd4 1789out:
5613fda9
FW		 1790 r->ru_utime = ns_to_timeval(utime);
1791 r->ru_stime = ns_to_timeval(stime);
1f10206c
JP		 1792
1793 if (who != RUSAGE_CHILDREN) {
1794 struct mm_struct *mm = get_task_mm(p);
ec94fc3d 1795
1f10206c
JP		 1796 if (mm) {
1797 setmax_mm_hiwater_rss(&maxrss, mm);
1798 mmput(mm);
1799 }
1800 }
1801 r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1da177e4
LT		 1802}
1803
ce72a16f 1804SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1da177e4
LT		 1805{
1806 struct rusage r;
ec94fc3d 1807
679c9cd4
SK		 1808 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1809 who != RUSAGE_THREAD)
1da177e4 1810 return -EINVAL;
ce72a16f
AV		 1811
1812 getrusage(current, who, &r);
1813 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1da177e4
LT		 1814}
1815
8d2d5c4a
AV		 1816#ifdef CONFIG_COMPAT
1817COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru)
1818{
1819 struct rusage r;
1820
1821 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1822 who != RUSAGE_THREAD)
1823 return -EINVAL;
1824
ce72a16f 1825 getrusage(current, who, &r);
8d2d5c4a
AV		 1826 return put_compat_rusage(&r, ru);
1827}
1828#endif
1829
e48fbb69 1830SYSCALL_DEFINE1(umask, int, mask)
1da177e4
LT		 1831{
1832 mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1833 return mask;
1834}
3b7391de 1835
6e399cd1 1836static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
b32dfe37 1837{
2903ff01 1838 struct fd exe;
6e399cd1 1839 struct file *old_exe, *exe_file;
496ad9aa 1840 struct inode *inode;
2903ff01 1841 int err;
b32dfe37 1842
2903ff01
AV		 1843 exe = fdget(fd);
1844 if (!exe.file)
b32dfe37
CG		 1845 return -EBADF;
1846
496ad9aa 1847 inode = file_inode(exe.file);
b32dfe37
CG		 1848
1849 /*
1850 * Because the original mm->exe_file points to executable file, make
1851 * sure that this one is executable as well, to avoid breaking an
1852 * overall picture.
1853 */
1854 err = -EACCES;
90f8572b 1855 if (!S_ISREG(inode->i_mode) || path_noexec(&exe.file->f_path))
b32dfe37
CG		 1856 goto exit;
1857
496ad9aa 1858 err = inode_permission(inode, MAY_EXEC);
b32dfe37
CG		 1859 if (err)
1860 goto exit;
1861
bafb282d 1862 /*
4229fb1d 1863 * Forbid mm->exe_file change if old file still mapped.
bafb282d 1864 */
6e399cd1 1865 exe_file = get_mm_exe_file(mm);
bafb282d 1866 err = -EBUSY;
6e399cd1 1867 if (exe_file) {
4229fb1d
KK		 1868 struct vm_area_struct *vma;
1869
6e399cd1
DB		 1870 down_read(&mm->mmap_sem);
1871 for (vma = mm->mmap; vma; vma = vma->vm_next) {
1872 if (!vma->vm_file)
1873 continue;
1874 if (path_equal(&vma->vm_file->f_path,
1875 &exe_file->f_path))
1876 goto exit_err;
1877 }
1878
1879 up_read(&mm->mmap_sem);
1880 fput(exe_file);
bafb282d
KK		 1881 }
1882
4229fb1d 1883 err = 0;
6e399cd1
DB		 1884 /* set the new file, lockless */
1885 get_file(exe.file);
1886 old_exe = xchg(&mm->exe_file, exe.file);
1887 if (old_exe)
1888 fput(old_exe);
b32dfe37 1889exit:
2903ff01 1890 fdput(exe);
b32dfe37 1891 return err;
6e399cd1
DB		 1892exit_err:
1893 up_read(&mm->mmap_sem);
1894 fput(exe_file);
1895 goto exit;
b32dfe37
CG		 1896}
1897
f606b77f
CG		 1898/*
1899 * WARNING: we don't require any capability here so be very careful
1900 * in what is allowed for modification from userspace.
1901 */
1902static int validate_prctl_map(struct prctl_mm_map *prctl_map)
1903{
1904 unsigned long mmap_max_addr = TASK_SIZE;
1905 struct mm_struct *mm = current->mm;
1906 int error = -EINVAL, i;
1907
1908 static const unsigned char offsets[] = {
1909 offsetof(struct prctl_mm_map, start_code),
1910 offsetof(struct prctl_mm_map, end_code),
1911 offsetof(struct prctl_mm_map, start_data),
1912 offsetof(struct prctl_mm_map, end_data),
1913 offsetof(struct prctl_mm_map, start_brk),
1914 offsetof(struct prctl_mm_map, brk),
1915 offsetof(struct prctl_mm_map, start_stack),
1916 offsetof(struct prctl_mm_map, arg_start),
1917 offsetof(struct prctl_mm_map, arg_end),
1918 offsetof(struct prctl_mm_map, env_start),
1919 offsetof(struct prctl_mm_map, env_end),
1920 };
1921
1922 /*
1923 * Make sure the members are not somewhere outside
1924 * of allowed address space.
1925 */
1926 for (i = 0; i < ARRAY_SIZE(offsets); i++) {
1927 u64 val = *(u64 *)((char *)prctl_map + offsets[i]);
1928
1929 if ((unsigned long)val >= mmap_max_addr ||
1930 (unsigned long)val < mmap_min_addr)
1931 goto out;
1932 }
1933
1934 /*
1935 * Make sure the pairs are ordered.
1936 */
1937#define __prctl_check_order(__m1, __op, __m2) \
1938 ((unsigned long)prctl_map->__m1 __op \
1939 (unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1940 error = __prctl_check_order(start_code, <, end_code);
1941 error |= __prctl_check_order(start_data, <, end_data);
1942 error |= __prctl_check_order(start_brk, <=, brk);
1943 error |= __prctl_check_order(arg_start, <=, arg_end);
1944 error |= __prctl_check_order(env_start, <=, env_end);
1945 if (error)
1946 goto out;
1947#undef __prctl_check_order
1948
1949 error = -EINVAL;
1950
1951 /*
1952 * @brk should be after @end_data in traditional maps.
1953 */
1954 if (prctl_map->start_brk <= prctl_map->end_data ||
1955 prctl_map->brk <= prctl_map->end_data)
1956 goto out;
1957
1958 /*
1959 * Neither we should allow to override limits if they set.
1960 */
1961 if (check_data_rlimit(rlimit(RLIMIT_DATA), prctl_map->brk,
1962 prctl_map->start_brk, prctl_map->end_data,
1963 prctl_map->start_data))
1964 goto out;
1965
1966 /*
1967 * Someone is trying to cheat the auxv vector.
1968 */
1969 if (prctl_map->auxv_size) {
1970 if (!prctl_map->auxv || prctl_map->auxv_size > sizeof(mm->saved_auxv))
1971 goto out;
1972 }
1973
1974 /*
1975 * Finally, make sure the caller has the rights to
4d28df61 1976 * change /proc/pid/exe link: only local sys admin should
f606b77f
CG		 1977 * be allowed to.
1978 */
1979 if (prctl_map->exe_fd != (u32)-1) {
4d28df61 1980 if (!ns_capable(current_user_ns(), CAP_SYS_ADMIN))
f606b77f
CG		 1981 goto out;
1982 }
1983
1984 error = 0;
1985out:
1986 return error;
1987}
1988
4a00e9df 1989#ifdef CONFIG_CHECKPOINT_RESTORE
f606b77f
CG		 1990static int prctl_set_mm_map(int opt, const void __user *addr, unsigned long data_size)
1991{
1992 struct prctl_mm_map prctl_map = { .exe_fd = (u32)-1, };
1993 unsigned long user_auxv[AT_VECTOR_SIZE];
1994 struct mm_struct *mm = current->mm;
1995 int error;
1996
1997 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
1998 BUILD_BUG_ON(sizeof(struct prctl_mm_map) > 256);
1999
2000 if (opt == PR_SET_MM_MAP_SIZE)
2001 return put_user((unsigned int)sizeof(prctl_map),
2002 (unsigned int __user *)addr);
2003
2004 if (data_size != sizeof(prctl_map))
2005 return -EINVAL;
2006
2007 if (copy_from_user(&prctl_map, addr, sizeof(prctl_map)))
2008 return -EFAULT;
2009
2010 error = validate_prctl_map(&prctl_map);
2011 if (error)
2012 return error;
2013
2014 if (prctl_map.auxv_size) {
2015 memset(user_auxv, 0, sizeof(user_auxv));
2016 if (copy_from_user(user_auxv,
2017 (const void __user *)prctl_map.auxv,
2018 prctl_map.auxv_size))
2019 return -EFAULT;
2020
2021 /* Last entry must be AT_NULL as specification requires */
2022 user_auxv[AT_VECTOR_SIZE - 2] = AT_NULL;
2023 user_auxv[AT_VECTOR_SIZE - 1] = AT_NULL;
2024 }
2025
ddf1d398 2026 if (prctl_map.exe_fd != (u32)-1) {
6e399cd1 2027 error = prctl_set_mm_exe_file(mm, prctl_map.exe_fd);
ddf1d398
MG		 2028 if (error)
2029 return error;
2030 }
2031
88aa7cc6
YS		 2032 /*
2033 * arg_lock protects concurent updates but we still need mmap_sem for
2034 * read to exclude races with sys_brk.
2035 */
2036 down_read(&mm->mmap_sem);
f606b77f
CG		 2037
2038 /*
2039 * We don't validate if these members are pointing to
2040 * real present VMAs because application may have correspond
2041 * VMAs already unmapped and kernel uses these members for statistics
2042 * output in procfs mostly, except
2043 *
2044 * - @start_brk/@brk which are used in do_brk but kernel lookups
2045 * for VMAs when updating these memvers so anything wrong written
2046 * here cause kernel to swear at userspace program but won't lead
2047 * to any problem in kernel itself
2048 */
2049
88aa7cc6 2050 spin_lock(&mm->arg_lock);
f606b77f
CG		 2051 mm->start_code = prctl_map.start_code;
2052 mm->end_code = prctl_map.end_code;
2053 mm->start_data = prctl_map.start_data;
2054 mm->end_data = prctl_map.end_data;
2055 mm->start_brk = prctl_map.start_brk;
2056 mm->brk = prctl_map.brk;
2057 mm->start_stack = prctl_map.start_stack;
2058 mm->arg_start = prctl_map.arg_start;
2059 mm->arg_end = prctl_map.arg_end;
2060 mm->env_start = prctl_map.env_start;
2061 mm->env_end = prctl_map.env_end;
88aa7cc6 2062 spin_unlock(&mm->arg_lock);
f606b77f
CG		 2063
2064 /*
2065 * Note this update of @saved_auxv is lockless thus
2066 * if someone reads this member in procfs while we're
2067 * updating -- it may get partly updated results. It's
2068 * known and acceptable trade off: we leave it as is to
2069 * not introduce additional locks here making the kernel
2070 * more complex.
2071 */
2072 if (prctl_map.auxv_size)
2073 memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv));
2074
88aa7cc6 2075 up_read(&mm->mmap_sem);
ddf1d398 2076 return 0;
f606b77f
CG		 2077}
2078#endif /* CONFIG_CHECKPOINT_RESTORE */
2079
4a00e9df
AD		 2080static int prctl_set_auxv(struct mm_struct *mm, unsigned long addr,
2081 unsigned long len)
2082{
2083 /*
2084 * This doesn't move the auxiliary vector itself since it's pinned to
2085 * mm_struct, but it permits filling the vector with new values. It's
2086 * up to the caller to provide sane values here, otherwise userspace
2087 * tools which use this vector might be unhappy.
2088 */
2089 unsigned long user_auxv[AT_VECTOR_SIZE];
2090
2091 if (len > sizeof(user_auxv))
2092 return -EINVAL;
2093
2094 if (copy_from_user(user_auxv, (const void __user *)addr, len))
2095 return -EFAULT;
2096
2097 /* Make sure the last entry is always AT_NULL */
2098 user_auxv[AT_VECTOR_SIZE - 2] = 0;
2099 user_auxv[AT_VECTOR_SIZE - 1] = 0;
2100
2101 BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
2102
2103 task_lock(current);
2104 memcpy(mm->saved_auxv, user_auxv, len);
2105 task_unlock(current);
2106
2107 return 0;
2108}
2109
028ee4be
CG		 2110static int prctl_set_mm(int opt, unsigned long addr,
2111 unsigned long arg4, unsigned long arg5)
2112{
028ee4be 2113 struct mm_struct *mm = current->mm;
4a00e9df 2114 struct prctl_mm_map prctl_map;
fe8c7f5c
CG		 2115 struct vm_area_struct *vma;
2116 int error;
028ee4be 2117
f606b77f
CG		 2118 if (arg5 || (arg4 && (opt != PR_SET_MM_AUXV &&
2119 opt != PR_SET_MM_MAP &&
2120 opt != PR_SET_MM_MAP_SIZE)))
028ee4be
CG		 2121 return -EINVAL;
2122
f606b77f
CG		 2123#ifdef CONFIG_CHECKPOINT_RESTORE
2124 if (opt == PR_SET_MM_MAP || opt == PR_SET_MM_MAP_SIZE)
2125 return prctl_set_mm_map(opt, (const void __user *)addr, arg4);
2126#endif
2127
79f0713d 2128 if (!capable(CAP_SYS_RESOURCE))
028ee4be
CG		 2129 return -EPERM;
2130
6e399cd1
DB		 2131 if (opt == PR_SET_MM_EXE_FILE)
2132 return prctl_set_mm_exe_file(mm, (unsigned int)addr);
b32dfe37 2133
4a00e9df
AD		 2134 if (opt == PR_SET_MM_AUXV)
2135 return prctl_set_auxv(mm, addr, arg4);
2136
1ad75b9e 2137 if (addr >= TASK_SIZE || addr < mmap_min_addr)
028ee4be
CG		 2138 return -EINVAL;
2139
fe8c7f5c
CG		 2140 error = -EINVAL;
2141
ddf1d398 2142 down_write(&mm->mmap_sem);
028ee4be
CG		 2143 vma = find_vma(mm, addr);
2144
4a00e9df
AD		 2145 prctl_map.start_code = mm->start_code;
2146 prctl_map.end_code = mm->end_code;
2147 prctl_map.start_data = mm->start_data;
2148 prctl_map.end_data = mm->end_data;
2149 prctl_map.start_brk = mm->start_brk;
2150 prctl_map.brk = mm->brk;
2151 prctl_map.start_stack = mm->start_stack;
2152 prctl_map.arg_start = mm->arg_start;
2153 prctl_map.arg_end = mm->arg_end;
2154 prctl_map.env_start = mm->env_start;
2155 prctl_map.env_end = mm->env_end;
2156 prctl_map.auxv = NULL;
2157 prctl_map.auxv_size = 0;
2158 prctl_map.exe_fd = -1;
2159
028ee4be
CG		 2160 switch (opt) {
2161 case PR_SET_MM_START_CODE:
4a00e9df 2162 prctl_map.start_code = addr;
fe8c7f5c 2163 break;
028ee4be 2164 case PR_SET_MM_END_CODE:
4a00e9df 2165 prctl_map.end_code = addr;
028ee4be 2166 break;
028ee4be 2167 case PR_SET_MM_START_DATA:
4a00e9df 2168 prctl_map.start_data = addr;
028ee4be 2169 break;
fe8c7f5c 2170 case PR_SET_MM_END_DATA:
4a00e9df
AD		 2171 prctl_map.end_data = addr;
2172 break;
2173 case PR_SET_MM_START_STACK:
2174 prctl_map.start_stack = addr;
028ee4be 2175 break;
028ee4be 2176 case PR_SET_MM_START_BRK:
4a00e9df 2177 prctl_map.start_brk = addr;
028ee4be 2178 break;
028ee4be 2179 case PR_SET_MM_BRK:
4a00e9df 2180 prctl_map.brk = addr;
028ee4be 2181 break;
4a00e9df
AD		 2182 case PR_SET_MM_ARG_START:
2183 prctl_map.arg_start = addr;
2184 break;
2185 case PR_SET_MM_ARG_END:
2186 prctl_map.arg_end = addr;
2187 break;
2188 case PR_SET_MM_ENV_START:
2189 prctl_map.env_start = addr;
2190 break;
2191 case PR_SET_MM_ENV_END:
2192 prctl_map.env_end = addr;
2193 break;
2194 default:
2195 goto out;
2196 }
2197
2198 error = validate_prctl_map(&prctl_map);
2199 if (error)
2200 goto out;
028ee4be 2201
4a00e9df 2202 switch (opt) {
fe8c7f5c
CG		 2203 /*
2204 * If command line arguments and environment
2205 * are placed somewhere else on stack, we can
2206 * set them up here, ARG_START/END to setup
2207 * command line argumets and ENV_START/END
2208 * for environment.
2209 */
2210 case PR_SET_MM_START_STACK:
2211 case PR_SET_MM_ARG_START:
2212 case PR_SET_MM_ARG_END:
2213 case PR_SET_MM_ENV_START:
2214 case PR_SET_MM_ENV_END:
2215 if (!vma) {
2216 error = -EFAULT;
2217 goto out;
2218 }
028ee4be
CG		 2219 }
2220
4a00e9df
AD		 2221 mm->start_code = prctl_map.start_code;
2222 mm->end_code = prctl_map.end_code;
2223 mm->start_data = prctl_map.start_data;
2224 mm->end_data = prctl_map.end_data;
2225 mm->start_brk = prctl_map.start_brk;
2226 mm->brk = prctl_map.brk;
2227 mm->start_stack = prctl_map.start_stack;
2228 mm->arg_start = prctl_map.arg_start;
2229 mm->arg_end = prctl_map.arg_end;
2230 mm->env_start = prctl_map.env_start;
2231 mm->env_end = prctl_map.env_end;
2232
028ee4be 2233 error = 0;
028ee4be 2234out:
ddf1d398 2235 up_write(&mm->mmap_sem);
028ee4be
CG		 2236 return error;
2237}
300f786b 2238
52b36941 2239#ifdef CONFIG_CHECKPOINT_RESTORE
300f786b
CG		 2240static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2241{
2242 return put_user(me->clear_child_tid, tid_addr);
2243}
52b36941 2244#else
300f786b
CG		 2245static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2246{
2247 return -EINVAL;
2248}
028ee4be
CG		 2249#endif
2250
749860ce
PT		 2251static int propagate_has_child_subreaper(struct task_struct *p, void *data)
2252{
2253 /*
2254 * If task has has_child_subreaper - all its decendants
2255 * already have these flag too and new decendants will
2256 * inherit it on fork, skip them.
2257 *
2258 * If we've found child_reaper - skip descendants in
2259 * it's subtree as they will never get out pidns.
2260 */
2261 if (p->signal->has_child_subreaper ||
2262 is_child_reaper(task_pid(p)))
2263 return 0;
2264
2265 p->signal->has_child_subreaper = 1;
2266 return 1;
2267}
2268
7bbf1373 2269int __weak arch_prctl_spec_ctrl_get(struct task_struct *t, unsigned long which)
b617cfc8
TG		 2270{
2271 return -EINVAL;
2272}
2273
7bbf1373
KC		 2274int __weak arch_prctl_spec_ctrl_set(struct task_struct *t, unsigned long which,
2275 unsigned long ctrl)
b617cfc8
TG		 2276{
2277 return -EINVAL;
2278}
2279
c4ea37c2
HC		 2280SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2281 unsigned long, arg4, unsigned long, arg5)
1da177e4 2282{
b6dff3ec
DH		 2283 struct task_struct *me = current;
2284 unsigned char comm[sizeof(me->comm)];
2285 long error;
1da177e4 2286
d84f4f99
DH		 2287 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2288 if (error != -ENOSYS)
1da177e4
LT		 2289 return error;
2290
d84f4f99 2291 error = 0;
1da177e4 2292 switch (option) {
f3cbd435
AM		 2293 case PR_SET_PDEATHSIG:
2294 if (!valid_signal(arg2)) {
2295 error = -EINVAL;
1da177e4 2296 break;
f3cbd435
AM		 2297 }
2298 me->pdeath_signal = arg2;
2299 break;
2300 case PR_GET_PDEATHSIG:
2301 error = put_user(me->pdeath_signal, (int __user *)arg2);
2302 break;
2303 case PR_GET_DUMPABLE:
2304 error = get_dumpable(me->mm);
2305 break;
2306 case PR_SET_DUMPABLE:
2307 if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) {
2308 error = -EINVAL;
1da177e4 2309 break;
f3cbd435
AM		 2310 }
2311 set_dumpable(me->mm, arg2);
2312 break;
1da177e4 2313
f3cbd435
AM		 2314 case PR_SET_UNALIGN:
2315 error = SET_UNALIGN_CTL(me, arg2);
2316 break;
2317 case PR_GET_UNALIGN:
2318 error = GET_UNALIGN_CTL(me, arg2);
2319 break;
2320 case PR_SET_FPEMU:
2321 error = SET_FPEMU_CTL(me, arg2);
2322 break;
2323 case PR_GET_FPEMU:
2324 error = GET_FPEMU_CTL(me, arg2);
2325 break;
2326 case PR_SET_FPEXC:
2327 error = SET_FPEXC_CTL(me, arg2);
2328 break;
2329 case PR_GET_FPEXC:
2330 error = GET_FPEXC_CTL(me, arg2);
2331 break;
2332 case PR_GET_TIMING:
2333 error = PR_TIMING_STATISTICAL;
2334 break;
2335 case PR_SET_TIMING:
2336 if (arg2 != PR_TIMING_STATISTICAL)
2337 error = -EINVAL;
2338 break;
2339 case PR_SET_NAME:
2340 comm[sizeof(me->comm) - 1] = 0;
2341 if (strncpy_from_user(comm, (char __user *)arg2,
2342 sizeof(me->comm) - 1) < 0)
2343 return -EFAULT;
2344 set_task_comm(me, comm);
2345 proc_comm_connector(me);
2346 break;
2347 case PR_GET_NAME:
2348 get_task_comm(comm, me);
2349 if (copy_to_user((char __user *)arg2, comm, sizeof(comm)))
2350 return -EFAULT;
2351 break;
2352 case PR_GET_ENDIAN:
2353 error = GET_ENDIAN(me, arg2);
2354 break;
2355 case PR_SET_ENDIAN:
2356 error = SET_ENDIAN(me, arg2);
2357 break;
2358 case PR_GET_SECCOMP:
2359 error = prctl_get_seccomp();
2360 break;
2361 case PR_SET_SECCOMP:
2362 error = prctl_set_seccomp(arg2, (char __user *)arg3);
2363 break;
2364 case PR_GET_TSC:
2365 error = GET_TSC_CTL(arg2);
2366 break;
2367 case PR_SET_TSC:
2368 error = SET_TSC_CTL(arg2);
2369 break;
2370 case PR_TASK_PERF_EVENTS_DISABLE:
2371 error = perf_event_task_disable();
2372 break;
2373 case PR_TASK_PERF_EVENTS_ENABLE:
2374 error = perf_event_task_enable();
2375 break;
2376 case PR_GET_TIMERSLACK:
da8b44d5
JS		 2377 if (current->timer_slack_ns > ULONG_MAX)
2378 error = ULONG_MAX;
2379 else
2380 error = current->timer_slack_ns;
f3cbd435
AM		 2381 break;
2382 case PR_SET_TIMERSLACK:
2383 if (arg2 <= 0)
2384 current->timer_slack_ns =
6976675d 2385 current->default_timer_slack_ns;
f3cbd435
AM		 2386 else
2387 current->timer_slack_ns = arg2;
2388 break;
2389 case PR_MCE_KILL:
2390 if (arg4 | arg5)
2391 return -EINVAL;
2392 switch (arg2) {
2393 case PR_MCE_KILL_CLEAR:
2394 if (arg3 != 0)
4db96cf0 2395 return -EINVAL;
f3cbd435 2396 current->flags &= ~PF_MCE_PROCESS;
4db96cf0 2397 break;
f3cbd435
AM		 2398 case PR_MCE_KILL_SET:
2399 current->flags |= PF_MCE_PROCESS;
2400 if (arg3 == PR_MCE_KILL_EARLY)
2401 current->flags |= PF_MCE_EARLY;
2402 else if (arg3 == PR_MCE_KILL_LATE)
2403 current->flags &= ~PF_MCE_EARLY;
2404 else if (arg3 == PR_MCE_KILL_DEFAULT)
2405 current->flags &=
2406 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
1087e9b4 2407 else
259e5e6c 2408 return -EINVAL;
259e5e6c 2409 break;
1da177e4 2410 default:
f3cbd435
AM		 2411 return -EINVAL;
2412 }
2413 break;
2414 case PR_MCE_KILL_GET:
2415 if (arg2 | arg3 | arg4 | arg5)
2416 return -EINVAL;
2417 if (current->flags & PF_MCE_PROCESS)
2418 error = (current->flags & PF_MCE_EARLY) ?
2419 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2420 else
2421 error = PR_MCE_KILL_DEFAULT;
2422 break;
2423 case PR_SET_MM:
2424 error = prctl_set_mm(arg2, arg3, arg4, arg5);
2425 break;
2426 case PR_GET_TID_ADDRESS:
2427 error = prctl_get_tid_address(me, (int __user **)arg2);
2428 break;
2429 case PR_SET_CHILD_SUBREAPER:
2430 me->signal->is_child_subreaper = !!arg2;
749860ce
PT		 2431 if (!arg2)
2432 break;
2433
2434 walk_process_tree(me, propagate_has_child_subreaper, NULL);
f3cbd435
AM		 2435 break;
2436 case PR_GET_CHILD_SUBREAPER:
2437 error = put_user(me->signal->is_child_subreaper,
2438 (int __user *)arg2);
2439 break;
2440 case PR_SET_NO_NEW_PRIVS:
2441 if (arg2 != 1 || arg3 || arg4 || arg5)
2442 return -EINVAL;
2443
1d4457f9 2444 task_set_no_new_privs(current);
f3cbd435
AM		 2445 break;
2446 case PR_GET_NO_NEW_PRIVS:
2447 if (arg2 || arg3 || arg4 || arg5)
2448 return -EINVAL;
1d4457f9 2449 return task_no_new_privs(current) ? 1 : 0;
a0715cc2
AT		 2450 case PR_GET_THP_DISABLE:
2451 if (arg2 || arg3 || arg4 || arg5)
2452 return -EINVAL;
18600332 2453 error = !!test_bit(MMF_DISABLE_THP, &me->mm->flags);
a0715cc2
AT		 2454 break;
2455 case PR_SET_THP_DISABLE:
2456 if (arg3 || arg4 || arg5)
2457 return -EINVAL;
17b0573d
MH		 2458 if (down_write_killable(&me->mm->mmap_sem))
2459 return -EINTR;
a0715cc2 2460 if (arg2)
18600332 2461 set_bit(MMF_DISABLE_THP, &me->mm->flags);
a0715cc2 2462 else
18600332 2463 clear_bit(MMF_DISABLE_THP, &me->mm->flags);
a0715cc2
AT		 2464 up_write(&me->mm->mmap_sem);
2465 break;
fe3d197f 2466 case PR_MPX_ENABLE_MANAGEMENT:
e9d1b4f3
DH		 2467 if (arg2 || arg3 || arg4 || arg5)
2468 return -EINVAL;
46a6e0cf 2469 error = MPX_ENABLE_MANAGEMENT();
fe3d197f
DH		 2470 break;
2471 case PR_MPX_DISABLE_MANAGEMENT:
e9d1b4f3
DH		 2472 if (arg2 || arg3 || arg4 || arg5)
2473 return -EINVAL;
46a6e0cf 2474 error = MPX_DISABLE_MANAGEMENT();
fe3d197f 2475 break;
9791554b
PB		 2476 case PR_SET_FP_MODE:
2477 error = SET_FP_MODE(me, arg2);
2478 break;
2479 case PR_GET_FP_MODE:
2480 error = GET_FP_MODE(me);
2481 break;
2d2123bc
DM		 2482 case PR_SVE_SET_VL:
2483 error = SVE_SET_VL(arg2);
2484 break;
2485 case PR_SVE_GET_VL:
2486 error = SVE_GET_VL();
2487 break;
b617cfc8
TG		 2488 case PR_GET_SPECULATION_CTRL:
2489 if (arg3 || arg4 || arg5)
2490 return -EINVAL;
7bbf1373 2491 error = arch_prctl_spec_ctrl_get(me, arg2);
b617cfc8
TG		 2492 break;
2493 case PR_SET_SPECULATION_CTRL:
2494 if (arg4 || arg5)
2495 return -EINVAL;
7bbf1373 2496 error = arch_prctl_spec_ctrl_set(me, arg2, arg3);
b617cfc8 2497 break;
ba830885
KM		 2498 case PR_PAC_RESET_KEYS:
2499 if (arg3 || arg4 || arg5)
2500 return -EINVAL;
2501 error = PAC_RESET_KEYS(me, arg2);
2502 break;
f3cbd435
AM		 2503 default:
2504 error = -EINVAL;
2505 break;
1da177e4
LT		 2506 }
2507 return error;
2508}
3cfc348b 2509
836f92ad
HC		 2510SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2511 struct getcpu_cache __user *, unused)
3cfc348b
AK		 2512{
2513 int err = 0;
2514 int cpu = raw_smp_processor_id();
ec94fc3d 2515
3cfc348b
AK		 2516 if (cpup)
2517 err |= put_user(cpu, cpup);
2518 if (nodep)
2519 err |= put_user(cpu_to_node(cpu), nodep);
3cfc348b
AK		 2520 return err ? -EFAULT : 0;
2521}
10a0a8d4 2522
4a22f166
SR		 2523/**
2524 * do_sysinfo - fill in sysinfo struct
2525 * @info: pointer to buffer to fill
2526 */
2527static int do_sysinfo(struct sysinfo *info)
2528{
2529 unsigned long mem_total, sav_total;
2530 unsigned int mem_unit, bitcount;
dc1b7b6c 2531 struct timespec64 tp;
4a22f166
SR		 2532
2533 memset(info, 0, sizeof(struct sysinfo));
2534
dc1b7b6c 2535 ktime_get_boottime_ts64(&tp);
4a22f166
SR		 2536 info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
2537
2538 get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
2539
2540 info->procs = nr_threads;
2541
2542 si_meminfo(info);
2543 si_swapinfo(info);
2544
2545 /*
2546 * If the sum of all the available memory (i.e. ram + swap)
2547 * is less than can be stored in a 32 bit unsigned long then
2548 * we can be binary compatible with 2.2.x kernels. If not,
2549 * well, in that case 2.2.x was broken anyways...
2550 *
2551 * -Erik Andersen <andersee@debian.org>
2552 */
2553
2554 mem_total = info->totalram + info->totalswap;
2555 if (mem_total < info->totalram || mem_total < info->totalswap)
2556 goto out;
2557 bitcount = 0;
2558 mem_unit = info->mem_unit;
2559 while (mem_unit > 1) {
2560 bitcount++;
2561 mem_unit >>= 1;
2562 sav_total = mem_total;
2563 mem_total <<= 1;
2564 if (mem_total < sav_total)
2565 goto out;
2566 }
2567
2568 /*
2569 * If mem_total did not overflow, multiply all memory values by
2570 * info->mem_unit and set it to 1. This leaves things compatible
2571 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2572 * kernels...
2573 */
2574
2575 info->mem_unit = 1;
2576 info->totalram <<= bitcount;
2577 info->freeram <<= bitcount;
2578 info->sharedram <<= bitcount;
2579 info->bufferram <<= bitcount;
2580 info->totalswap <<= bitcount;
2581 info->freeswap <<= bitcount;
2582 info->totalhigh <<= bitcount;
2583 info->freehigh <<= bitcount;
2584
2585out:
2586 return 0;
2587}
2588
2589SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
2590{
2591 struct sysinfo val;
2592
2593 do_sysinfo(&val);
2594
2595 if (copy_to_user(info, &val, sizeof(struct sysinfo)))
2596 return -EFAULT;
2597
2598 return 0;
2599}
2600
2601#ifdef CONFIG_COMPAT
2602struct compat_sysinfo {
2603 s32 uptime;
2604 u32 loads[3];
2605 u32 totalram;
2606 u32 freeram;
2607 u32 sharedram;
2608 u32 bufferram;
2609 u32 totalswap;
2610 u32 freeswap;
2611 u16 procs;
2612 u16 pad;
2613 u32 totalhigh;
2614 u32 freehigh;
2615 u32 mem_unit;
2616 char _f[20-2*sizeof(u32)-sizeof(int)];
2617};
2618
2619COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2620{
2621 struct sysinfo s;
2622
2623 do_sysinfo(&s);
2624
2625 /* Check to see if any memory value is too large for 32-bit and scale
2626 * down if needed
2627 */
0baae41e 2628 if (upper_32_bits(s.totalram) || upper_32_bits(s.totalswap)) {
4a22f166
SR		 2629 int bitcount = 0;
2630
2631 while (s.mem_unit < PAGE_SIZE) {
2632 s.mem_unit <<= 1;
2633 bitcount++;
2634 }
2635
2636 s.totalram >>= bitcount;
2637 s.freeram >>= bitcount;
2638 s.sharedram >>= bitcount;
2639 s.bufferram >>= bitcount;
2640 s.totalswap >>= bitcount;
2641 s.freeswap >>= bitcount;
2642 s.totalhigh >>= bitcount;
2643 s.freehigh >>= bitcount;
2644 }
2645
96d4f267 2646 if (!access_ok(info, sizeof(struct compat_sysinfo)) ||
4a22f166
SR		 2647 __put_user(s.uptime, &info->uptime) ||
2648 __put_user(s.loads[0], &info->loads[0]) ||
2649 __put_user(s.loads[1], &info->loads[1]) ||
2650 __put_user(s.loads[2], &info->loads[2]) ||
2651 __put_user(s.totalram, &info->totalram) ||
2652 __put_user(s.freeram, &info->freeram) ||
2653 __put_user(s.sharedram, &info->sharedram) ||
2654 __put_user(s.bufferram, &info->bufferram) ||
2655 __put_user(s.totalswap, &info->totalswap) ||
2656 __put_user(s.freeswap, &info->freeswap) ||
2657 __put_user(s.procs, &info->procs) ||
2658 __put_user(s.totalhigh, &info->totalhigh) ||
2659 __put_user(s.freehigh, &info->freehigh) ||
2660 __put_user(s.mem_unit, &info->mem_unit))
2661 return -EFAULT;
2662
2663 return 0;
2664}
2665#endif /* CONFIG_COMPAT */
Ubuntu Disco kernel mirror for Proxmox VE and PMG