xref: /openbmc/linux/kernel/sys.c (revision 2fa5ebe3)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  *  linux/kernel/sys.c
4  *
5  *  Copyright (C) 1991, 1992  Linus Torvalds
6  */
7 
8 #include <linux/export.h>
9 #include <linux/mm.h>
10 #include <linux/mm_inline.h>
11 #include <linux/utsname.h>
12 #include <linux/mman.h>
13 #include <linux/reboot.h>
14 #include <linux/prctl.h>
15 #include <linux/highuid.h>
16 #include <linux/fs.h>
17 #include <linux/kmod.h>
18 #include <linux/perf_event.h>
19 #include <linux/resource.h>
20 #include <linux/kernel.h>
21 #include <linux/workqueue.h>
22 #include <linux/capability.h>
23 #include <linux/device.h>
24 #include <linux/key.h>
25 #include <linux/times.h>
26 #include <linux/posix-timers.h>
27 #include <linux/security.h>
28 #include <linux/random.h>
29 #include <linux/suspend.h>
30 #include <linux/tty.h>
31 #include <linux/signal.h>
32 #include <linux/cn_proc.h>
33 #include <linux/getcpu.h>
34 #include <linux/task_io_accounting_ops.h>
35 #include <linux/seccomp.h>
36 #include <linux/cpu.h>
37 #include <linux/personality.h>
38 #include <linux/ptrace.h>
39 #include <linux/fs_struct.h>
40 #include <linux/file.h>
41 #include <linux/mount.h>
42 #include <linux/gfp.h>
43 #include <linux/syscore_ops.h>
44 #include <linux/version.h>
45 #include <linux/ctype.h>
46 #include <linux/syscall_user_dispatch.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/time_namespace.h>
53 #include <linux/binfmts.h>
54 
55 #include <linux/sched.h>
56 #include <linux/sched/autogroup.h>
57 #include <linux/sched/loadavg.h>
58 #include <linux/sched/stat.h>
59 #include <linux/sched/mm.h>
60 #include <linux/sched/coredump.h>
61 #include <linux/sched/task.h>
62 #include <linux/sched/cputime.h>
63 #include <linux/rcupdate.h>
64 #include <linux/uidgid.h>
65 #include <linux/cred.h>
66 
67 #include <linux/nospec.h>
68 
69 #include <linux/kmsg_dump.h>
70 /* Move somewhere else to avoid recompiling? */
71 #include <generated/utsrelease.h>
72 
73 #include <linux/uaccess.h>
74 #include <asm/io.h>
75 #include <asm/unistd.h>
76 
77 #include "uid16.h"
78 
79 #ifndef SET_UNALIGN_CTL
80 # define SET_UNALIGN_CTL(a, b)	(-EINVAL)
81 #endif
82 #ifndef GET_UNALIGN_CTL
83 # define GET_UNALIGN_CTL(a, b)	(-EINVAL)
84 #endif
85 #ifndef SET_FPEMU_CTL
86 # define SET_FPEMU_CTL(a, b)	(-EINVAL)
87 #endif
88 #ifndef GET_FPEMU_CTL
89 # define GET_FPEMU_CTL(a, b)	(-EINVAL)
90 #endif
91 #ifndef SET_FPEXC_CTL
92 # define SET_FPEXC_CTL(a, b)	(-EINVAL)
93 #endif
94 #ifndef GET_FPEXC_CTL
95 # define GET_FPEXC_CTL(a, b)	(-EINVAL)
96 #endif
97 #ifndef GET_ENDIAN
98 # define GET_ENDIAN(a, b)	(-EINVAL)
99 #endif
100 #ifndef SET_ENDIAN
101 # define SET_ENDIAN(a, b)	(-EINVAL)
102 #endif
103 #ifndef GET_TSC_CTL
104 # define GET_TSC_CTL(a)		(-EINVAL)
105 #endif
106 #ifndef SET_TSC_CTL
107 # define SET_TSC_CTL(a)		(-EINVAL)
108 #endif
109 #ifndef GET_FP_MODE
110 # define GET_FP_MODE(a)		(-EINVAL)
111 #endif
112 #ifndef SET_FP_MODE
113 # define SET_FP_MODE(a,b)	(-EINVAL)
114 #endif
115 #ifndef SVE_SET_VL
116 # define SVE_SET_VL(a)		(-EINVAL)
117 #endif
118 #ifndef SVE_GET_VL
119 # define SVE_GET_VL()		(-EINVAL)
120 #endif
121 #ifndef SME_SET_VL
122 # define SME_SET_VL(a)		(-EINVAL)
123 #endif
124 #ifndef SME_GET_VL
125 # define SME_GET_VL()		(-EINVAL)
126 #endif
127 #ifndef PAC_RESET_KEYS
128 # define PAC_RESET_KEYS(a, b)	(-EINVAL)
129 #endif
130 #ifndef PAC_SET_ENABLED_KEYS
131 # define PAC_SET_ENABLED_KEYS(a, b, c)	(-EINVAL)
132 #endif
133 #ifndef PAC_GET_ENABLED_KEYS
134 # define PAC_GET_ENABLED_KEYS(a)	(-EINVAL)
135 #endif
136 #ifndef SET_TAGGED_ADDR_CTRL
137 # define SET_TAGGED_ADDR_CTRL(a)	(-EINVAL)
138 #endif
139 #ifndef GET_TAGGED_ADDR_CTRL
140 # define GET_TAGGED_ADDR_CTRL()		(-EINVAL)
141 #endif
142 
143 /*
144  * this is where the system-wide overflow UID and GID are defined, for
145  * architectures that now have 32-bit UID/GID but didn't in the past
146  */
147 
148 int overflowuid = DEFAULT_OVERFLOWUID;
149 int overflowgid = DEFAULT_OVERFLOWGID;
150 
151 EXPORT_SYMBOL(overflowuid);
152 EXPORT_SYMBOL(overflowgid);
153 
154 /*
155  * the same as above, but for filesystems which can only store a 16-bit
156  * UID and GID. as such, this is needed on all architectures
157  */
158 
159 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
160 int fs_overflowgid = DEFAULT_FS_OVERFLOWGID;
161 
162 EXPORT_SYMBOL(fs_overflowuid);
163 EXPORT_SYMBOL(fs_overflowgid);
164 
165 /*
166  * Returns true if current's euid is same as p's uid or euid,
167  * or has CAP_SYS_NICE to p's user_ns.
168  *
169  * Called with rcu_read_lock, creds are safe
170  */
171 static bool set_one_prio_perm(struct task_struct *p)
172 {
173 	const struct cred *cred = current_cred(), *pcred = __task_cred(p);
174 
175 	if (uid_eq(pcred->uid,  cred->euid) ||
176 	    uid_eq(pcred->euid, cred->euid))
177 		return true;
178 	if (ns_capable(pcred->user_ns, CAP_SYS_NICE))
179 		return true;
180 	return false;
181 }
182 
183 /*
184  * set the priority of a task
185  * - the caller must hold the RCU read lock
186  */
187 static int set_one_prio(struct task_struct *p, int niceval, int error)
188 {
189 	int no_nice;
190 
191 	if (!set_one_prio_perm(p)) {
192 		error = -EPERM;
193 		goto out;
194 	}
195 	if (niceval < task_nice(p) && !can_nice(p, niceval)) {
196 		error = -EACCES;
197 		goto out;
198 	}
199 	no_nice = security_task_setnice(p, niceval);
200 	if (no_nice) {
201 		error = no_nice;
202 		goto out;
203 	}
204 	if (error == -ESRCH)
205 		error = 0;
206 	set_user_nice(p, niceval);
207 out:
208 	return error;
209 }
210 
211 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
212 {
213 	struct task_struct *g, *p;
214 	struct user_struct *user;
215 	const struct cred *cred = current_cred();
216 	int error = -EINVAL;
217 	struct pid *pgrp;
218 	kuid_t uid;
219 
220 	if (which > PRIO_USER || which < PRIO_PROCESS)
221 		goto out;
222 
223 	/* normalize: avoid signed division (rounding problems) */
224 	error = -ESRCH;
225 	if (niceval < MIN_NICE)
226 		niceval = MIN_NICE;
227 	if (niceval > MAX_NICE)
228 		niceval = MAX_NICE;
229 
230 	rcu_read_lock();
231 	switch (which) {
232 	case PRIO_PROCESS:
233 		if (who)
234 			p = find_task_by_vpid(who);
235 		else
236 			p = current;
237 		if (p)
238 			error = set_one_prio(p, niceval, error);
239 		break;
240 	case PRIO_PGRP:
241 		if (who)
242 			pgrp = find_vpid(who);
243 		else
244 			pgrp = task_pgrp(current);
245 		read_lock(&tasklist_lock);
246 		do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
247 			error = set_one_prio(p, niceval, error);
248 		} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
249 		read_unlock(&tasklist_lock);
250 		break;
251 	case PRIO_USER:
252 		uid = make_kuid(cred->user_ns, who);
253 		user = cred->user;
254 		if (!who)
255 			uid = cred->uid;
256 		else if (!uid_eq(uid, cred->uid)) {
257 			user = find_user(uid);
258 			if (!user)
259 				goto out_unlock;	/* No processes for this user */
260 		}
261 		for_each_process_thread(g, p) {
262 			if (uid_eq(task_uid(p), uid) && task_pid_vnr(p))
263 				error = set_one_prio(p, niceval, error);
264 		}
265 		if (!uid_eq(uid, cred->uid))
266 			free_uid(user);		/* For find_user() */
267 		break;
268 	}
269 out_unlock:
270 	rcu_read_unlock();
271 out:
272 	return error;
273 }
274 
275 /*
276  * Ugh. To avoid negative return values, "getpriority()" will
277  * not return the normal nice-value, but a negated value that
278  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
279  * to stay compatible.
280  */
281 SYSCALL_DEFINE2(getpriority, int, which, int, who)
282 {
283 	struct task_struct *g, *p;
284 	struct user_struct *user;
285 	const struct cred *cred = current_cred();
286 	long niceval, retval = -ESRCH;
287 	struct pid *pgrp;
288 	kuid_t uid;
289 
290 	if (which > PRIO_USER || which < PRIO_PROCESS)
291 		return -EINVAL;
292 
293 	rcu_read_lock();
294 	switch (which) {
295 	case PRIO_PROCESS:
296 		if (who)
297 			p = find_task_by_vpid(who);
298 		else
299 			p = current;
300 		if (p) {
301 			niceval = nice_to_rlimit(task_nice(p));
302 			if (niceval > retval)
303 				retval = niceval;
304 		}
305 		break;
306 	case PRIO_PGRP:
307 		if (who)
308 			pgrp = find_vpid(who);
309 		else
310 			pgrp = task_pgrp(current);
311 		read_lock(&tasklist_lock);
312 		do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
313 			niceval = nice_to_rlimit(task_nice(p));
314 			if (niceval > retval)
315 				retval = niceval;
316 		} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
317 		read_unlock(&tasklist_lock);
318 		break;
319 	case PRIO_USER:
320 		uid = make_kuid(cred->user_ns, who);
321 		user = cred->user;
322 		if (!who)
323 			uid = cred->uid;
324 		else if (!uid_eq(uid, cred->uid)) {
325 			user = find_user(uid);
326 			if (!user)
327 				goto out_unlock;	/* No processes for this user */
328 		}
329 		for_each_process_thread(g, p) {
330 			if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) {
331 				niceval = nice_to_rlimit(task_nice(p));
332 				if (niceval > retval)
333 					retval = niceval;
334 			}
335 		}
336 		if (!uid_eq(uid, cred->uid))
337 			free_uid(user);		/* for find_user() */
338 		break;
339 	}
340 out_unlock:
341 	rcu_read_unlock();
342 
343 	return retval;
344 }
345 
346 /*
347  * Unprivileged users may change the real gid to the effective gid
348  * or vice versa.  (BSD-style)
349  *
350  * If you set the real gid at all, or set the effective gid to a value not
351  * equal to the real gid, then the saved gid is set to the new effective gid.
352  *
353  * This makes it possible for a setgid program to completely drop its
354  * privileges, which is often a useful assertion to make when you are doing
355  * a security audit over a program.
356  *
357  * The general idea is that a program which uses just setregid() will be
358  * 100% compatible with BSD.  A program which uses just setgid() will be
359  * 100% compatible with POSIX with saved IDs.
360  *
361  * SMP: There are not races, the GIDs are checked only by filesystem
362  *      operations (as far as semantic preservation is concerned).
363  */
364 #ifdef CONFIG_MULTIUSER
365 long __sys_setregid(gid_t rgid, gid_t egid)
366 {
367 	struct user_namespace *ns = current_user_ns();
368 	const struct cred *old;
369 	struct cred *new;
370 	int retval;
371 	kgid_t krgid, kegid;
372 
373 	krgid = make_kgid(ns, rgid);
374 	kegid = make_kgid(ns, egid);
375 
376 	if ((rgid != (gid_t) -1) && !gid_valid(krgid))
377 		return -EINVAL;
378 	if ((egid != (gid_t) -1) && !gid_valid(kegid))
379 		return -EINVAL;
380 
381 	new = prepare_creds();
382 	if (!new)
383 		return -ENOMEM;
384 	old = current_cred();
385 
386 	retval = -EPERM;
387 	if (rgid != (gid_t) -1) {
388 		if (gid_eq(old->gid, krgid) ||
389 		    gid_eq(old->egid, krgid) ||
390 		    ns_capable_setid(old->user_ns, CAP_SETGID))
391 			new->gid = krgid;
392 		else
393 			goto error;
394 	}
395 	if (egid != (gid_t) -1) {
396 		if (gid_eq(old->gid, kegid) ||
397 		    gid_eq(old->egid, kegid) ||
398 		    gid_eq(old->sgid, kegid) ||
399 		    ns_capable_setid(old->user_ns, CAP_SETGID))
400 			new->egid = kegid;
401 		else
402 			goto error;
403 	}
404 
405 	if (rgid != (gid_t) -1 ||
406 	    (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
407 		new->sgid = new->egid;
408 	new->fsgid = new->egid;
409 
410 	retval = security_task_fix_setgid(new, old, LSM_SETID_RE);
411 	if (retval < 0)
412 		goto error;
413 
414 	return commit_creds(new);
415 
416 error:
417 	abort_creds(new);
418 	return retval;
419 }
420 
421 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
422 {
423 	return __sys_setregid(rgid, egid);
424 }
425 
426 /*
427  * setgid() is implemented like SysV w/ SAVED_IDS
428  *
429  * SMP: Same implicit races as above.
430  */
431 long __sys_setgid(gid_t gid)
432 {
433 	struct user_namespace *ns = current_user_ns();
434 	const struct cred *old;
435 	struct cred *new;
436 	int retval;
437 	kgid_t kgid;
438 
439 	kgid = make_kgid(ns, gid);
440 	if (!gid_valid(kgid))
441 		return -EINVAL;
442 
443 	new = prepare_creds();
444 	if (!new)
445 		return -ENOMEM;
446 	old = current_cred();
447 
448 	retval = -EPERM;
449 	if (ns_capable_setid(old->user_ns, CAP_SETGID))
450 		new->gid = new->egid = new->sgid = new->fsgid = kgid;
451 	else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
452 		new->egid = new->fsgid = kgid;
453 	else
454 		goto error;
455 
456 	retval = security_task_fix_setgid(new, old, LSM_SETID_ID);
457 	if (retval < 0)
458 		goto error;
459 
460 	return commit_creds(new);
461 
462 error:
463 	abort_creds(new);
464 	return retval;
465 }
466 
467 SYSCALL_DEFINE1(setgid, gid_t, gid)
468 {
469 	return __sys_setgid(gid);
470 }
471 
472 /*
473  * change the user struct in a credentials set to match the new UID
474  */
475 static int set_user(struct cred *new)
476 {
477 	struct user_struct *new_user;
478 
479 	new_user = alloc_uid(new->uid);
480 	if (!new_user)
481 		return -EAGAIN;
482 
483 	free_uid(new->user);
484 	new->user = new_user;
485 	return 0;
486 }
487 
488 static void flag_nproc_exceeded(struct cred *new)
489 {
490 	if (new->ucounts == current_ucounts())
491 		return;
492 
493 	/*
494 	 * We don't fail in case of NPROC limit excess here because too many
495 	 * poorly written programs don't check set*uid() return code, assuming
496 	 * it never fails if called by root.  We may still enforce NPROC limit
497 	 * for programs doing set*uid()+execve() by harmlessly deferring the
498 	 * failure to the execve() stage.
499 	 */
500 	if (is_rlimit_overlimit(new->ucounts, UCOUNT_RLIMIT_NPROC, rlimit(RLIMIT_NPROC)) &&
501 			new->user != INIT_USER)
502 		current->flags |= PF_NPROC_EXCEEDED;
503 	else
504 		current->flags &= ~PF_NPROC_EXCEEDED;
505 }
506 
507 /*
508  * Unprivileged users may change the real uid to the effective uid
509  * or vice versa.  (BSD-style)
510  *
511  * If you set the real uid at all, or set the effective uid to a value not
512  * equal to the real uid, then the saved uid is set to the new effective uid.
513  *
514  * This makes it possible for a setuid program to completely drop its
515  * privileges, which is often a useful assertion to make when you are doing
516  * a security audit over a program.
517  *
518  * The general idea is that a program which uses just setreuid() will be
519  * 100% compatible with BSD.  A program which uses just setuid() will be
520  * 100% compatible with POSIX with saved IDs.
521  */
522 long __sys_setreuid(uid_t ruid, uid_t euid)
523 {
524 	struct user_namespace *ns = current_user_ns();
525 	const struct cred *old;
526 	struct cred *new;
527 	int retval;
528 	kuid_t kruid, keuid;
529 
530 	kruid = make_kuid(ns, ruid);
531 	keuid = make_kuid(ns, euid);
532 
533 	if ((ruid != (uid_t) -1) && !uid_valid(kruid))
534 		return -EINVAL;
535 	if ((euid != (uid_t) -1) && !uid_valid(keuid))
536 		return -EINVAL;
537 
538 	new = prepare_creds();
539 	if (!new)
540 		return -ENOMEM;
541 	old = current_cred();
542 
543 	retval = -EPERM;
544 	if (ruid != (uid_t) -1) {
545 		new->uid = kruid;
546 		if (!uid_eq(old->uid, kruid) &&
547 		    !uid_eq(old->euid, kruid) &&
548 		    !ns_capable_setid(old->user_ns, CAP_SETUID))
549 			goto error;
550 	}
551 
552 	if (euid != (uid_t) -1) {
553 		new->euid = keuid;
554 		if (!uid_eq(old->uid, keuid) &&
555 		    !uid_eq(old->euid, keuid) &&
556 		    !uid_eq(old->suid, keuid) &&
557 		    !ns_capable_setid(old->user_ns, CAP_SETUID))
558 			goto error;
559 	}
560 
561 	if (!uid_eq(new->uid, old->uid)) {
562 		retval = set_user(new);
563 		if (retval < 0)
564 			goto error;
565 	}
566 	if (ruid != (uid_t) -1 ||
567 	    (euid != (uid_t) -1 && !uid_eq(keuid, old->uid)))
568 		new->suid = new->euid;
569 	new->fsuid = new->euid;
570 
571 	retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
572 	if (retval < 0)
573 		goto error;
574 
575 	retval = set_cred_ucounts(new);
576 	if (retval < 0)
577 		goto error;
578 
579 	flag_nproc_exceeded(new);
580 	return commit_creds(new);
581 
582 error:
583 	abort_creds(new);
584 	return retval;
585 }
586 
587 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
588 {
589 	return __sys_setreuid(ruid, euid);
590 }
591 
592 /*
593  * setuid() is implemented like SysV with SAVED_IDS
594  *
595  * Note that SAVED_ID's is deficient in that a setuid root program
596  * like sendmail, for example, cannot set its uid to be a normal
597  * user and then switch back, because if you're root, setuid() sets
598  * the saved uid too.  If you don't like this, blame the bright people
599  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
600  * will allow a root program to temporarily drop privileges and be able to
601  * regain them by swapping the real and effective uid.
602  */
603 long __sys_setuid(uid_t uid)
604 {
605 	struct user_namespace *ns = current_user_ns();
606 	const struct cred *old;
607 	struct cred *new;
608 	int retval;
609 	kuid_t kuid;
610 
611 	kuid = make_kuid(ns, uid);
612 	if (!uid_valid(kuid))
613 		return -EINVAL;
614 
615 	new = prepare_creds();
616 	if (!new)
617 		return -ENOMEM;
618 	old = current_cred();
619 
620 	retval = -EPERM;
621 	if (ns_capable_setid(old->user_ns, CAP_SETUID)) {
622 		new->suid = new->uid = kuid;
623 		if (!uid_eq(kuid, old->uid)) {
624 			retval = set_user(new);
625 			if (retval < 0)
626 				goto error;
627 		}
628 	} else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) {
629 		goto error;
630 	}
631 
632 	new->fsuid = new->euid = kuid;
633 
634 	retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
635 	if (retval < 0)
636 		goto error;
637 
638 	retval = set_cred_ucounts(new);
639 	if (retval < 0)
640 		goto error;
641 
642 	flag_nproc_exceeded(new);
643 	return commit_creds(new);
644 
645 error:
646 	abort_creds(new);
647 	return retval;
648 }
649 
650 SYSCALL_DEFINE1(setuid, uid_t, uid)
651 {
652 	return __sys_setuid(uid);
653 }
654 
655 
656 /*
657  * This function implements a generic ability to update ruid, euid,
658  * and suid.  This allows you to implement the 4.4 compatible seteuid().
659  */
660 long __sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
661 {
662 	struct user_namespace *ns = current_user_ns();
663 	const struct cred *old;
664 	struct cred *new;
665 	int retval;
666 	kuid_t kruid, keuid, ksuid;
667 	bool ruid_new, euid_new, suid_new;
668 
669 	kruid = make_kuid(ns, ruid);
670 	keuid = make_kuid(ns, euid);
671 	ksuid = make_kuid(ns, suid);
672 
673 	if ((ruid != (uid_t) -1) && !uid_valid(kruid))
674 		return -EINVAL;
675 
676 	if ((euid != (uid_t) -1) && !uid_valid(keuid))
677 		return -EINVAL;
678 
679 	if ((suid != (uid_t) -1) && !uid_valid(ksuid))
680 		return -EINVAL;
681 
682 	old = current_cred();
683 
684 	/* check for no-op */
685 	if ((ruid == (uid_t) -1 || uid_eq(kruid, old->uid)) &&
686 	    (euid == (uid_t) -1 || (uid_eq(keuid, old->euid) &&
687 				    uid_eq(keuid, old->fsuid))) &&
688 	    (suid == (uid_t) -1 || uid_eq(ksuid, old->suid)))
689 		return 0;
690 
691 	ruid_new = ruid != (uid_t) -1        && !uid_eq(kruid, old->uid) &&
692 		   !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid);
693 	euid_new = euid != (uid_t) -1        && !uid_eq(keuid, old->uid) &&
694 		   !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid);
695 	suid_new = suid != (uid_t) -1        && !uid_eq(ksuid, old->uid) &&
696 		   !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid);
697 	if ((ruid_new || euid_new || suid_new) &&
698 	    !ns_capable_setid(old->user_ns, CAP_SETUID))
699 		return -EPERM;
700 
701 	new = prepare_creds();
702 	if (!new)
703 		return -ENOMEM;
704 
705 	if (ruid != (uid_t) -1) {
706 		new->uid = kruid;
707 		if (!uid_eq(kruid, old->uid)) {
708 			retval = set_user(new);
709 			if (retval < 0)
710 				goto error;
711 		}
712 	}
713 	if (euid != (uid_t) -1)
714 		new->euid = keuid;
715 	if (suid != (uid_t) -1)
716 		new->suid = ksuid;
717 	new->fsuid = new->euid;
718 
719 	retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
720 	if (retval < 0)
721 		goto error;
722 
723 	retval = set_cred_ucounts(new);
724 	if (retval < 0)
725 		goto error;
726 
727 	flag_nproc_exceeded(new);
728 	return commit_creds(new);
729 
730 error:
731 	abort_creds(new);
732 	return retval;
733 }
734 
735 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
736 {
737 	return __sys_setresuid(ruid, euid, suid);
738 }
739 
740 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
741 {
742 	const struct cred *cred = current_cred();
743 	int retval;
744 	uid_t ruid, euid, suid;
745 
746 	ruid = from_kuid_munged(cred->user_ns, cred->uid);
747 	euid = from_kuid_munged(cred->user_ns, cred->euid);
748 	suid = from_kuid_munged(cred->user_ns, cred->suid);
749 
750 	retval = put_user(ruid, ruidp);
751 	if (!retval) {
752 		retval = put_user(euid, euidp);
753 		if (!retval)
754 			return put_user(suid, suidp);
755 	}
756 	return retval;
757 }
758 
759 /*
760  * Same as above, but for rgid, egid, sgid.
761  */
762 long __sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
763 {
764 	struct user_namespace *ns = current_user_ns();
765 	const struct cred *old;
766 	struct cred *new;
767 	int retval;
768 	kgid_t krgid, kegid, ksgid;
769 	bool rgid_new, egid_new, sgid_new;
770 
771 	krgid = make_kgid(ns, rgid);
772 	kegid = make_kgid(ns, egid);
773 	ksgid = make_kgid(ns, sgid);
774 
775 	if ((rgid != (gid_t) -1) && !gid_valid(krgid))
776 		return -EINVAL;
777 	if ((egid != (gid_t) -1) && !gid_valid(kegid))
778 		return -EINVAL;
779 	if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
780 		return -EINVAL;
781 
782 	old = current_cred();
783 
784 	/* check for no-op */
785 	if ((rgid == (gid_t) -1 || gid_eq(krgid, old->gid)) &&
786 	    (egid == (gid_t) -1 || (gid_eq(kegid, old->egid) &&
787 				    gid_eq(kegid, old->fsgid))) &&
788 	    (sgid == (gid_t) -1 || gid_eq(ksgid, old->sgid)))
789 		return 0;
790 
791 	rgid_new = rgid != (gid_t) -1        && !gid_eq(krgid, old->gid) &&
792 		   !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid);
793 	egid_new = egid != (gid_t) -1        && !gid_eq(kegid, old->gid) &&
794 		   !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid);
795 	sgid_new = sgid != (gid_t) -1        && !gid_eq(ksgid, old->gid) &&
796 		   !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid);
797 	if ((rgid_new || egid_new || sgid_new) &&
798 	    !ns_capable_setid(old->user_ns, CAP_SETGID))
799 		return -EPERM;
800 
801 	new = prepare_creds();
802 	if (!new)
803 		return -ENOMEM;
804 
805 	if (rgid != (gid_t) -1)
806 		new->gid = krgid;
807 	if (egid != (gid_t) -1)
808 		new->egid = kegid;
809 	if (sgid != (gid_t) -1)
810 		new->sgid = ksgid;
811 	new->fsgid = new->egid;
812 
813 	retval = security_task_fix_setgid(new, old, LSM_SETID_RES);
814 	if (retval < 0)
815 		goto error;
816 
817 	return commit_creds(new);
818 
819 error:
820 	abort_creds(new);
821 	return retval;
822 }
823 
824 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
825 {
826 	return __sys_setresgid(rgid, egid, sgid);
827 }
828 
829 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
830 {
831 	const struct cred *cred = current_cred();
832 	int retval;
833 	gid_t rgid, egid, sgid;
834 
835 	rgid = from_kgid_munged(cred->user_ns, cred->gid);
836 	egid = from_kgid_munged(cred->user_ns, cred->egid);
837 	sgid = from_kgid_munged(cred->user_ns, cred->sgid);
838 
839 	retval = put_user(rgid, rgidp);
840 	if (!retval) {
841 		retval = put_user(egid, egidp);
842 		if (!retval)
843 			retval = put_user(sgid, sgidp);
844 	}
845 
846 	return retval;
847 }
848 
849 
850 /*
851  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
852  * is used for "access()" and for the NFS daemon (letting nfsd stay at
853  * whatever uid it wants to). It normally shadows "euid", except when
854  * explicitly set by setfsuid() or for access..
855  */
856 long __sys_setfsuid(uid_t uid)
857 {
858 	const struct cred *old;
859 	struct cred *new;
860 	uid_t old_fsuid;
861 	kuid_t kuid;
862 
863 	old = current_cred();
864 	old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
865 
866 	kuid = make_kuid(old->user_ns, uid);
867 	if (!uid_valid(kuid))
868 		return old_fsuid;
869 
870 	new = prepare_creds();
871 	if (!new)
872 		return old_fsuid;
873 
874 	if (uid_eq(kuid, old->uid)  || uid_eq(kuid, old->euid)  ||
875 	    uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
876 	    ns_capable_setid(old->user_ns, CAP_SETUID)) {
877 		if (!uid_eq(kuid, old->fsuid)) {
878 			new->fsuid = kuid;
879 			if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
880 				goto change_okay;
881 		}
882 	}
883 
884 	abort_creds(new);
885 	return old_fsuid;
886 
887 change_okay:
888 	commit_creds(new);
889 	return old_fsuid;
890 }
891 
892 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
893 {
894 	return __sys_setfsuid(uid);
895 }
896 
897 /*
898  * Samma på svenska..
899  */
900 long __sys_setfsgid(gid_t gid)
901 {
902 	const struct cred *old;
903 	struct cred *new;
904 	gid_t old_fsgid;
905 	kgid_t kgid;
906 
907 	old = current_cred();
908 	old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
909 
910 	kgid = make_kgid(old->user_ns, gid);
911 	if (!gid_valid(kgid))
912 		return old_fsgid;
913 
914 	new = prepare_creds();
915 	if (!new)
916 		return old_fsgid;
917 
918 	if (gid_eq(kgid, old->gid)  || gid_eq(kgid, old->egid)  ||
919 	    gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
920 	    ns_capable_setid(old->user_ns, CAP_SETGID)) {
921 		if (!gid_eq(kgid, old->fsgid)) {
922 			new->fsgid = kgid;
923 			if (security_task_fix_setgid(new,old,LSM_SETID_FS) == 0)
924 				goto change_okay;
925 		}
926 	}
927 
928 	abort_creds(new);
929 	return old_fsgid;
930 
931 change_okay:
932 	commit_creds(new);
933 	return old_fsgid;
934 }
935 
936 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
937 {
938 	return __sys_setfsgid(gid);
939 }
940 #endif /* CONFIG_MULTIUSER */
941 
942 /**
943  * sys_getpid - return the thread group id of the current process
944  *
945  * Note, despite the name, this returns the tgid not the pid.  The tgid and
946  * the pid are identical unless CLONE_THREAD was specified on clone() in
947  * which case the tgid is the same in all threads of the same group.
948  *
949  * This is SMP safe as current->tgid does not change.
950  */
951 SYSCALL_DEFINE0(getpid)
952 {
953 	return task_tgid_vnr(current);
954 }
955 
956 /* Thread ID - the internal kernel "pid" */
957 SYSCALL_DEFINE0(gettid)
958 {
959 	return task_pid_vnr(current);
960 }
961 
962 /*
963  * Accessing ->real_parent is not SMP-safe, it could
964  * change from under us. However, we can use a stale
965  * value of ->real_parent under rcu_read_lock(), see
966  * release_task()->call_rcu(delayed_put_task_struct).
967  */
968 SYSCALL_DEFINE0(getppid)
969 {
970 	int pid;
971 
972 	rcu_read_lock();
973 	pid = task_tgid_vnr(rcu_dereference(current->real_parent));
974 	rcu_read_unlock();
975 
976 	return pid;
977 }
978 
979 SYSCALL_DEFINE0(getuid)
980 {
981 	/* Only we change this so SMP safe */
982 	return from_kuid_munged(current_user_ns(), current_uid());
983 }
984 
985 SYSCALL_DEFINE0(geteuid)
986 {
987 	/* Only we change this so SMP safe */
988 	return from_kuid_munged(current_user_ns(), current_euid());
989 }
990 
991 SYSCALL_DEFINE0(getgid)
992 {
993 	/* Only we change this so SMP safe */
994 	return from_kgid_munged(current_user_ns(), current_gid());
995 }
996 
997 SYSCALL_DEFINE0(getegid)
998 {
999 	/* Only we change this so SMP safe */
1000 	return from_kgid_munged(current_user_ns(), current_egid());
1001 }
1002 
1003 static void do_sys_times(struct tms *tms)
1004 {
1005 	u64 tgutime, tgstime, cutime, cstime;
1006 
1007 	thread_group_cputime_adjusted(current, &tgutime, &tgstime);
1008 	cutime = current->signal->cutime;
1009 	cstime = current->signal->cstime;
1010 	tms->tms_utime = nsec_to_clock_t(tgutime);
1011 	tms->tms_stime = nsec_to_clock_t(tgstime);
1012 	tms->tms_cutime = nsec_to_clock_t(cutime);
1013 	tms->tms_cstime = nsec_to_clock_t(cstime);
1014 }
1015 
1016 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
1017 {
1018 	if (tbuf) {
1019 		struct tms tmp;
1020 
1021 		do_sys_times(&tmp);
1022 		if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1023 			return -EFAULT;
1024 	}
1025 	force_successful_syscall_return();
1026 	return (long) jiffies_64_to_clock_t(get_jiffies_64());
1027 }
1028 
1029 #ifdef CONFIG_COMPAT
1030 static compat_clock_t clock_t_to_compat_clock_t(clock_t x)
1031 {
1032 	return compat_jiffies_to_clock_t(clock_t_to_jiffies(x));
1033 }
1034 
1035 COMPAT_SYSCALL_DEFINE1(times, struct compat_tms __user *, tbuf)
1036 {
1037 	if (tbuf) {
1038 		struct tms tms;
1039 		struct compat_tms tmp;
1040 
1041 		do_sys_times(&tms);
1042 		/* Convert our struct tms to the compat version. */
1043 		tmp.tms_utime = clock_t_to_compat_clock_t(tms.tms_utime);
1044 		tmp.tms_stime = clock_t_to_compat_clock_t(tms.tms_stime);
1045 		tmp.tms_cutime = clock_t_to_compat_clock_t(tms.tms_cutime);
1046 		tmp.tms_cstime = clock_t_to_compat_clock_t(tms.tms_cstime);
1047 		if (copy_to_user(tbuf, &tmp, sizeof(tmp)))
1048 			return -EFAULT;
1049 	}
1050 	force_successful_syscall_return();
1051 	return compat_jiffies_to_clock_t(jiffies);
1052 }
1053 #endif
1054 
1055 /*
1056  * This needs some heavy checking ...
1057  * I just haven't the stomach for it. I also don't fully
1058  * understand sessions/pgrp etc. Let somebody who does explain it.
1059  *
1060  * OK, I think I have the protection semantics right.... this is really
1061  * only important on a multi-user system anyway, to make sure one user
1062  * can't send a signal to a process owned by another.  -TYT, 12/12/91
1063  *
1064  * !PF_FORKNOEXEC check to conform completely to POSIX.
1065  */
1066 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
1067 {
1068 	struct task_struct *p;
1069 	struct task_struct *group_leader = current->group_leader;
1070 	struct pid *pgrp;
1071 	int err;
1072 
1073 	if (!pid)
1074 		pid = task_pid_vnr(group_leader);
1075 	if (!pgid)
1076 		pgid = pid;
1077 	if (pgid < 0)
1078 		return -EINVAL;
1079 	rcu_read_lock();
1080 
1081 	/* From this point forward we keep holding onto the tasklist lock
1082 	 * so that our parent does not change from under us. -DaveM
1083 	 */
1084 	write_lock_irq(&tasklist_lock);
1085 
1086 	err = -ESRCH;
1087 	p = find_task_by_vpid(pid);
1088 	if (!p)
1089 		goto out;
1090 
1091 	err = -EINVAL;
1092 	if (!thread_group_leader(p))
1093 		goto out;
1094 
1095 	if (same_thread_group(p->real_parent, group_leader)) {
1096 		err = -EPERM;
1097 		if (task_session(p) != task_session(group_leader))
1098 			goto out;
1099 		err = -EACCES;
1100 		if (!(p->flags & PF_FORKNOEXEC))
1101 			goto out;
1102 	} else {
1103 		err = -ESRCH;
1104 		if (p != group_leader)
1105 			goto out;
1106 	}
1107 
1108 	err = -EPERM;
1109 	if (p->signal->leader)
1110 		goto out;
1111 
1112 	pgrp = task_pid(p);
1113 	if (pgid != pid) {
1114 		struct task_struct *g;
1115 
1116 		pgrp = find_vpid(pgid);
1117 		g = pid_task(pgrp, PIDTYPE_PGID);
1118 		if (!g || task_session(g) != task_session(group_leader))
1119 			goto out;
1120 	}
1121 
1122 	err = security_task_setpgid(p, pgid);
1123 	if (err)
1124 		goto out;
1125 
1126 	if (task_pgrp(p) != pgrp)
1127 		change_pid(p, PIDTYPE_PGID, pgrp);
1128 
1129 	err = 0;
1130 out:
1131 	/* All paths lead to here, thus we are safe. -DaveM */
1132 	write_unlock_irq(&tasklist_lock);
1133 	rcu_read_unlock();
1134 	return err;
1135 }
1136 
1137 static int do_getpgid(pid_t pid)
1138 {
1139 	struct task_struct *p;
1140 	struct pid *grp;
1141 	int retval;
1142 
1143 	rcu_read_lock();
1144 	if (!pid)
1145 		grp = task_pgrp(current);
1146 	else {
1147 		retval = -ESRCH;
1148 		p = find_task_by_vpid(pid);
1149 		if (!p)
1150 			goto out;
1151 		grp = task_pgrp(p);
1152 		if (!grp)
1153 			goto out;
1154 
1155 		retval = security_task_getpgid(p);
1156 		if (retval)
1157 			goto out;
1158 	}
1159 	retval = pid_vnr(grp);
1160 out:
1161 	rcu_read_unlock();
1162 	return retval;
1163 }
1164 
1165 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1166 {
1167 	return do_getpgid(pid);
1168 }
1169 
1170 #ifdef __ARCH_WANT_SYS_GETPGRP
1171 
1172 SYSCALL_DEFINE0(getpgrp)
1173 {
1174 	return do_getpgid(0);
1175 }
1176 
1177 #endif
1178 
1179 SYSCALL_DEFINE1(getsid, pid_t, pid)
1180 {
1181 	struct task_struct *p;
1182 	struct pid *sid;
1183 	int retval;
1184 
1185 	rcu_read_lock();
1186 	if (!pid)
1187 		sid = task_session(current);
1188 	else {
1189 		retval = -ESRCH;
1190 		p = find_task_by_vpid(pid);
1191 		if (!p)
1192 			goto out;
1193 		sid = task_session(p);
1194 		if (!sid)
1195 			goto out;
1196 
1197 		retval = security_task_getsid(p);
1198 		if (retval)
1199 			goto out;
1200 	}
1201 	retval = pid_vnr(sid);
1202 out:
1203 	rcu_read_unlock();
1204 	return retval;
1205 }
1206 
1207 static void set_special_pids(struct pid *pid)
1208 {
1209 	struct task_struct *curr = current->group_leader;
1210 
1211 	if (task_session(curr) != pid)
1212 		change_pid(curr, PIDTYPE_SID, pid);
1213 
1214 	if (task_pgrp(curr) != pid)
1215 		change_pid(curr, PIDTYPE_PGID, pid);
1216 }
1217 
1218 int ksys_setsid(void)
1219 {
1220 	struct task_struct *group_leader = current->group_leader;
1221 	struct pid *sid = task_pid(group_leader);
1222 	pid_t session = pid_vnr(sid);
1223 	int err = -EPERM;
1224 
1225 	write_lock_irq(&tasklist_lock);
1226 	/* Fail if I am already a session leader */
1227 	if (group_leader->signal->leader)
1228 		goto out;
1229 
1230 	/* Fail if a process group id already exists that equals the
1231 	 * proposed session id.
1232 	 */
1233 	if (pid_task(sid, PIDTYPE_PGID))
1234 		goto out;
1235 
1236 	group_leader->signal->leader = 1;
1237 	set_special_pids(sid);
1238 
1239 	proc_clear_tty(group_leader);
1240 
1241 	err = session;
1242 out:
1243 	write_unlock_irq(&tasklist_lock);
1244 	if (err > 0) {
1245 		proc_sid_connector(group_leader);
1246 		sched_autogroup_create_attach(group_leader);
1247 	}
1248 	return err;
1249 }
1250 
1251 SYSCALL_DEFINE0(setsid)
1252 {
1253 	return ksys_setsid();
1254 }
1255 
1256 DECLARE_RWSEM(uts_sem);
1257 
1258 #ifdef COMPAT_UTS_MACHINE
1259 #define override_architecture(name) \
1260 	(personality(current->personality) == PER_LINUX32 && \
1261 	 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1262 		      sizeof(COMPAT_UTS_MACHINE)))
1263 #else
1264 #define override_architecture(name)	0
1265 #endif
1266 
1267 /*
1268  * Work around broken programs that cannot handle "Linux 3.0".
1269  * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1270  * And we map 4.x and later versions to 2.6.60+x, so 4.0/5.0/6.0/... would be
1271  * 2.6.60.
1272  */
1273 static int override_release(char __user *release, size_t len)
1274 {
1275 	int ret = 0;
1276 
1277 	if (current->personality & UNAME26) {
1278 		const char *rest = UTS_RELEASE;
1279 		char buf[65] = { 0 };
1280 		int ndots = 0;
1281 		unsigned v;
1282 		size_t copy;
1283 
1284 		while (*rest) {
1285 			if (*rest == '.' && ++ndots >= 3)
1286 				break;
1287 			if (!isdigit(*rest) && *rest != '.')
1288 				break;
1289 			rest++;
1290 		}
1291 		v = LINUX_VERSION_PATCHLEVEL + 60;
1292 		copy = clamp_t(size_t, len, 1, sizeof(buf));
1293 		copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1294 		ret = copy_to_user(release, buf, copy + 1);
1295 	}
1296 	return ret;
1297 }
1298 
1299 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1300 {
1301 	struct new_utsname tmp;
1302 
1303 	down_read(&uts_sem);
1304 	memcpy(&tmp, utsname(), sizeof(tmp));
1305 	up_read(&uts_sem);
1306 	if (copy_to_user(name, &tmp, sizeof(tmp)))
1307 		return -EFAULT;
1308 
1309 	if (override_release(name->release, sizeof(name->release)))
1310 		return -EFAULT;
1311 	if (override_architecture(name))
1312 		return -EFAULT;
1313 	return 0;
1314 }
1315 
1316 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1317 /*
1318  * Old cruft
1319  */
1320 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1321 {
1322 	struct old_utsname tmp;
1323 
1324 	if (!name)
1325 		return -EFAULT;
1326 
1327 	down_read(&uts_sem);
1328 	memcpy(&tmp, utsname(), sizeof(tmp));
1329 	up_read(&uts_sem);
1330 	if (copy_to_user(name, &tmp, sizeof(tmp)))
1331 		return -EFAULT;
1332 
1333 	if (override_release(name->release, sizeof(name->release)))
1334 		return -EFAULT;
1335 	if (override_architecture(name))
1336 		return -EFAULT;
1337 	return 0;
1338 }
1339 
1340 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1341 {
1342 	struct oldold_utsname tmp;
1343 
1344 	if (!name)
1345 		return -EFAULT;
1346 
1347 	memset(&tmp, 0, sizeof(tmp));
1348 
1349 	down_read(&uts_sem);
1350 	memcpy(&tmp.sysname, &utsname()->sysname, __OLD_UTS_LEN);
1351 	memcpy(&tmp.nodename, &utsname()->nodename, __OLD_UTS_LEN);
1352 	memcpy(&tmp.release, &utsname()->release, __OLD_UTS_LEN);
1353 	memcpy(&tmp.version, &utsname()->version, __OLD_UTS_LEN);
1354 	memcpy(&tmp.machine, &utsname()->machine, __OLD_UTS_LEN);
1355 	up_read(&uts_sem);
1356 	if (copy_to_user(name, &tmp, sizeof(tmp)))
1357 		return -EFAULT;
1358 
1359 	if (override_architecture(name))
1360 		return -EFAULT;
1361 	if (override_release(name->release, sizeof(name->release)))
1362 		return -EFAULT;
1363 	return 0;
1364 }
1365 #endif
1366 
1367 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1368 {
1369 	int errno;
1370 	char tmp[__NEW_UTS_LEN];
1371 
1372 	if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1373 		return -EPERM;
1374 
1375 	if (len < 0 || len > __NEW_UTS_LEN)
1376 		return -EINVAL;
1377 	errno = -EFAULT;
1378 	if (!copy_from_user(tmp, name, len)) {
1379 		struct new_utsname *u;
1380 
1381 		add_device_randomness(tmp, len);
1382 		down_write(&uts_sem);
1383 		u = utsname();
1384 		memcpy(u->nodename, tmp, len);
1385 		memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1386 		errno = 0;
1387 		uts_proc_notify(UTS_PROC_HOSTNAME);
1388 		up_write(&uts_sem);
1389 	}
1390 	return errno;
1391 }
1392 
1393 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1394 
1395 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1396 {
1397 	int i;
1398 	struct new_utsname *u;
1399 	char tmp[__NEW_UTS_LEN + 1];
1400 
1401 	if (len < 0)
1402 		return -EINVAL;
1403 	down_read(&uts_sem);
1404 	u = utsname();
1405 	i = 1 + strlen(u->nodename);
1406 	if (i > len)
1407 		i = len;
1408 	memcpy(tmp, u->nodename, i);
1409 	up_read(&uts_sem);
1410 	if (copy_to_user(name, tmp, i))
1411 		return -EFAULT;
1412 	return 0;
1413 }
1414 
1415 #endif
1416 
1417 /*
1418  * Only setdomainname; getdomainname can be implemented by calling
1419  * uname()
1420  */
1421 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1422 {
1423 	int errno;
1424 	char tmp[__NEW_UTS_LEN];
1425 
1426 	if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1427 		return -EPERM;
1428 	if (len < 0 || len > __NEW_UTS_LEN)
1429 		return -EINVAL;
1430 
1431 	errno = -EFAULT;
1432 	if (!copy_from_user(tmp, name, len)) {
1433 		struct new_utsname *u;
1434 
1435 		add_device_randomness(tmp, len);
1436 		down_write(&uts_sem);
1437 		u = utsname();
1438 		memcpy(u->domainname, tmp, len);
1439 		memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1440 		errno = 0;
1441 		uts_proc_notify(UTS_PROC_DOMAINNAME);
1442 		up_write(&uts_sem);
1443 	}
1444 	return errno;
1445 }
1446 
1447 /* make sure you are allowed to change @tsk limits before calling this */
1448 static int do_prlimit(struct task_struct *tsk, unsigned int resource,
1449 		      struct rlimit *new_rlim, struct rlimit *old_rlim)
1450 {
1451 	struct rlimit *rlim;
1452 	int retval = 0;
1453 
1454 	if (resource >= RLIM_NLIMITS)
1455 		return -EINVAL;
1456 	resource = array_index_nospec(resource, RLIM_NLIMITS);
1457 
1458 	if (new_rlim) {
1459 		if (new_rlim->rlim_cur > new_rlim->rlim_max)
1460 			return -EINVAL;
1461 		if (resource == RLIMIT_NOFILE &&
1462 				new_rlim->rlim_max > sysctl_nr_open)
1463 			return -EPERM;
1464 	}
1465 
1466 	/* Holding a refcount on tsk protects tsk->signal from disappearing. */
1467 	rlim = tsk->signal->rlim + resource;
1468 	task_lock(tsk->group_leader);
1469 	if (new_rlim) {
1470 		/*
1471 		 * Keep the capable check against init_user_ns until cgroups can
1472 		 * contain all limits.
1473 		 */
1474 		if (new_rlim->rlim_max > rlim->rlim_max &&
1475 				!capable(CAP_SYS_RESOURCE))
1476 			retval = -EPERM;
1477 		if (!retval)
1478 			retval = security_task_setrlimit(tsk, resource, new_rlim);
1479 	}
1480 	if (!retval) {
1481 		if (old_rlim)
1482 			*old_rlim = *rlim;
1483 		if (new_rlim)
1484 			*rlim = *new_rlim;
1485 	}
1486 	task_unlock(tsk->group_leader);
1487 
1488 	/*
1489 	 * RLIMIT_CPU handling. Arm the posix CPU timer if the limit is not
1490 	 * infinite. In case of RLIM_INFINITY the posix CPU timer code
1491 	 * ignores the rlimit.
1492 	 */
1493 	if (!retval && new_rlim && resource == RLIMIT_CPU &&
1494 	    new_rlim->rlim_cur != RLIM_INFINITY &&
1495 	    IS_ENABLED(CONFIG_POSIX_TIMERS)) {
1496 		/*
1497 		 * update_rlimit_cpu can fail if the task is exiting, but there
1498 		 * may be other tasks in the thread group that are not exiting,
1499 		 * and they need their cpu timers adjusted.
1500 		 *
1501 		 * The group_leader is the last task to be released, so if we
1502 		 * cannot update_rlimit_cpu on it, then the entire process is
1503 		 * exiting and we do not need to update at all.
1504 		 */
1505 		update_rlimit_cpu(tsk->group_leader, new_rlim->rlim_cur);
1506 	}
1507 
1508 	return retval;
1509 }
1510 
1511 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1512 {
1513 	struct rlimit value;
1514 	int ret;
1515 
1516 	ret = do_prlimit(current, resource, NULL, &value);
1517 	if (!ret)
1518 		ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1519 
1520 	return ret;
1521 }
1522 
1523 #ifdef CONFIG_COMPAT
1524 
1525 COMPAT_SYSCALL_DEFINE2(setrlimit, unsigned int, resource,
1526 		       struct compat_rlimit __user *, rlim)
1527 {
1528 	struct rlimit r;
1529 	struct compat_rlimit r32;
1530 
1531 	if (copy_from_user(&r32, rlim, sizeof(struct compat_rlimit)))
1532 		return -EFAULT;
1533 
1534 	if (r32.rlim_cur == COMPAT_RLIM_INFINITY)
1535 		r.rlim_cur = RLIM_INFINITY;
1536 	else
1537 		r.rlim_cur = r32.rlim_cur;
1538 	if (r32.rlim_max == COMPAT_RLIM_INFINITY)
1539 		r.rlim_max = RLIM_INFINITY;
1540 	else
1541 		r.rlim_max = r32.rlim_max;
1542 	return do_prlimit(current, resource, &r, NULL);
1543 }
1544 
1545 COMPAT_SYSCALL_DEFINE2(getrlimit, unsigned int, resource,
1546 		       struct compat_rlimit __user *, rlim)
1547 {
1548 	struct rlimit r;
1549 	int ret;
1550 
1551 	ret = do_prlimit(current, resource, NULL, &r);
1552 	if (!ret) {
1553 		struct compat_rlimit r32;
1554 		if (r.rlim_cur > COMPAT_RLIM_INFINITY)
1555 			r32.rlim_cur = COMPAT_RLIM_INFINITY;
1556 		else
1557 			r32.rlim_cur = r.rlim_cur;
1558 		if (r.rlim_max > COMPAT_RLIM_INFINITY)
1559 			r32.rlim_max = COMPAT_RLIM_INFINITY;
1560 		else
1561 			r32.rlim_max = r.rlim_max;
1562 
1563 		if (copy_to_user(rlim, &r32, sizeof(struct compat_rlimit)))
1564 			return -EFAULT;
1565 	}
1566 	return ret;
1567 }
1568 
1569 #endif
1570 
1571 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1572 
1573 /*
1574  *	Back compatibility for getrlimit. Needed for some apps.
1575  */
1576 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1577 		struct rlimit __user *, rlim)
1578 {
1579 	struct rlimit x;
1580 	if (resource >= RLIM_NLIMITS)
1581 		return -EINVAL;
1582 
1583 	resource = array_index_nospec(resource, RLIM_NLIMITS);
1584 	task_lock(current->group_leader);
1585 	x = current->signal->rlim[resource];
1586 	task_unlock(current->group_leader);
1587 	if (x.rlim_cur > 0x7FFFFFFF)
1588 		x.rlim_cur = 0x7FFFFFFF;
1589 	if (x.rlim_max > 0x7FFFFFFF)
1590 		x.rlim_max = 0x7FFFFFFF;
1591 	return copy_to_user(rlim, &x, sizeof(x)) ? -EFAULT : 0;
1592 }
1593 
1594 #ifdef CONFIG_COMPAT
1595 COMPAT_SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1596 		       struct compat_rlimit __user *, rlim)
1597 {
1598 	struct rlimit r;
1599 
1600 	if (resource >= RLIM_NLIMITS)
1601 		return -EINVAL;
1602 
1603 	resource = array_index_nospec(resource, RLIM_NLIMITS);
1604 	task_lock(current->group_leader);
1605 	r = current->signal->rlim[resource];
1606 	task_unlock(current->group_leader);
1607 	if (r.rlim_cur > 0x7FFFFFFF)
1608 		r.rlim_cur = 0x7FFFFFFF;
1609 	if (r.rlim_max > 0x7FFFFFFF)
1610 		r.rlim_max = 0x7FFFFFFF;
1611 
1612 	if (put_user(r.rlim_cur, &rlim->rlim_cur) ||
1613 	    put_user(r.rlim_max, &rlim->rlim_max))
1614 		return -EFAULT;
1615 	return 0;
1616 }
1617 #endif
1618 
1619 #endif
1620 
1621 static inline bool rlim64_is_infinity(__u64 rlim64)
1622 {
1623 #if BITS_PER_LONG < 64
1624 	return rlim64 >= ULONG_MAX;
1625 #else
1626 	return rlim64 == RLIM64_INFINITY;
1627 #endif
1628 }
1629 
1630 static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1631 {
1632 	if (rlim->rlim_cur == RLIM_INFINITY)
1633 		rlim64->rlim_cur = RLIM64_INFINITY;
1634 	else
1635 		rlim64->rlim_cur = rlim->rlim_cur;
1636 	if (rlim->rlim_max == RLIM_INFINITY)
1637 		rlim64->rlim_max = RLIM64_INFINITY;
1638 	else
1639 		rlim64->rlim_max = rlim->rlim_max;
1640 }
1641 
1642 static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1643 {
1644 	if (rlim64_is_infinity(rlim64->rlim_cur))
1645 		rlim->rlim_cur = RLIM_INFINITY;
1646 	else
1647 		rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1648 	if (rlim64_is_infinity(rlim64->rlim_max))
1649 		rlim->rlim_max = RLIM_INFINITY;
1650 	else
1651 		rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1652 }
1653 
1654 /* rcu lock must be held */
1655 static int check_prlimit_permission(struct task_struct *task,
1656 				    unsigned int flags)
1657 {
1658 	const struct cred *cred = current_cred(), *tcred;
1659 	bool id_match;
1660 
1661 	if (current == task)
1662 		return 0;
1663 
1664 	tcred = __task_cred(task);
1665 	id_match = (uid_eq(cred->uid, tcred->euid) &&
1666 		    uid_eq(cred->uid, tcred->suid) &&
1667 		    uid_eq(cred->uid, tcred->uid)  &&
1668 		    gid_eq(cred->gid, tcred->egid) &&
1669 		    gid_eq(cred->gid, tcred->sgid) &&
1670 		    gid_eq(cred->gid, tcred->gid));
1671 	if (!id_match && !ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
1672 		return -EPERM;
1673 
1674 	return security_task_prlimit(cred, tcred, flags);
1675 }
1676 
1677 SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1678 		const struct rlimit64 __user *, new_rlim,
1679 		struct rlimit64 __user *, old_rlim)
1680 {
1681 	struct rlimit64 old64, new64;
1682 	struct rlimit old, new;
1683 	struct task_struct *tsk;
1684 	unsigned int checkflags = 0;
1685 	int ret;
1686 
1687 	if (old_rlim)
1688 		checkflags |= LSM_PRLIMIT_READ;
1689 
1690 	if (new_rlim) {
1691 		if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1692 			return -EFAULT;
1693 		rlim64_to_rlim(&new64, &new);
1694 		checkflags |= LSM_PRLIMIT_WRITE;
1695 	}
1696 
1697 	rcu_read_lock();
1698 	tsk = pid ? find_task_by_vpid(pid) : current;
1699 	if (!tsk) {
1700 		rcu_read_unlock();
1701 		return -ESRCH;
1702 	}
1703 	ret = check_prlimit_permission(tsk, checkflags);
1704 	if (ret) {
1705 		rcu_read_unlock();
1706 		return ret;
1707 	}
1708 	get_task_struct(tsk);
1709 	rcu_read_unlock();
1710 
1711 	ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1712 			old_rlim ? &old : NULL);
1713 
1714 	if (!ret && old_rlim) {
1715 		rlim_to_rlim64(&old, &old64);
1716 		if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1717 			ret = -EFAULT;
1718 	}
1719 
1720 	put_task_struct(tsk);
1721 	return ret;
1722 }
1723 
1724 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1725 {
1726 	struct rlimit new_rlim;
1727 
1728 	if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1729 		return -EFAULT;
1730 	return do_prlimit(current, resource, &new_rlim, NULL);
1731 }
1732 
1733 /*
1734  * It would make sense to put struct rusage in the task_struct,
1735  * except that would make the task_struct be *really big*.  After
1736  * task_struct gets moved into malloc'ed memory, it would
1737  * make sense to do this.  It will make moving the rest of the information
1738  * a lot simpler!  (Which we're not doing right now because we're not
1739  * measuring them yet).
1740  *
1741  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1742  * races with threads incrementing their own counters.  But since word
1743  * reads are atomic, we either get new values or old values and we don't
1744  * care which for the sums.  We always take the siglock to protect reading
1745  * the c* fields from p->signal from races with exit.c updating those
1746  * fields when reaping, so a sample either gets all the additions of a
1747  * given child after it's reaped, or none so this sample is before reaping.
1748  *
1749  * Locking:
1750  * We need to take the siglock for CHILDEREN, SELF and BOTH
1751  * for  the cases current multithreaded, non-current single threaded
1752  * non-current multithreaded.  Thread traversal is now safe with
1753  * the siglock held.
1754  * Strictly speaking, we donot need to take the siglock if we are current and
1755  * single threaded,  as no one else can take our signal_struct away, no one
1756  * else can  reap the  children to update signal->c* counters, and no one else
1757  * can race with the signal-> fields. If we do not take any lock, the
1758  * signal-> fields could be read out of order while another thread was just
1759  * exiting. So we should  place a read memory barrier when we avoid the lock.
1760  * On the writer side,  write memory barrier is implied in  __exit_signal
1761  * as __exit_signal releases  the siglock spinlock after updating the signal->
1762  * fields. But we don't do this yet to keep things simple.
1763  *
1764  */
1765 
1766 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1767 {
1768 	r->ru_nvcsw += t->nvcsw;
1769 	r->ru_nivcsw += t->nivcsw;
1770 	r->ru_minflt += t->min_flt;
1771 	r->ru_majflt += t->maj_flt;
1772 	r->ru_inblock += task_io_get_inblock(t);
1773 	r->ru_oublock += task_io_get_oublock(t);
1774 }
1775 
1776 void getrusage(struct task_struct *p, int who, struct rusage *r)
1777 {
1778 	struct task_struct *t;
1779 	unsigned long flags;
1780 	u64 tgutime, tgstime, utime, stime;
1781 	unsigned long maxrss = 0;
1782 
1783 	memset((char *)r, 0, sizeof (*r));
1784 	utime = stime = 0;
1785 
1786 	if (who == RUSAGE_THREAD) {
1787 		task_cputime_adjusted(current, &utime, &stime);
1788 		accumulate_thread_rusage(p, r);
1789 		maxrss = p->signal->maxrss;
1790 		goto out;
1791 	}
1792 
1793 	if (!lock_task_sighand(p, &flags))
1794 		return;
1795 
1796 	switch (who) {
1797 	case RUSAGE_BOTH:
1798 	case RUSAGE_CHILDREN:
1799 		utime = p->signal->cutime;
1800 		stime = p->signal->cstime;
1801 		r->ru_nvcsw = p->signal->cnvcsw;
1802 		r->ru_nivcsw = p->signal->cnivcsw;
1803 		r->ru_minflt = p->signal->cmin_flt;
1804 		r->ru_majflt = p->signal->cmaj_flt;
1805 		r->ru_inblock = p->signal->cinblock;
1806 		r->ru_oublock = p->signal->coublock;
1807 		maxrss = p->signal->cmaxrss;
1808 
1809 		if (who == RUSAGE_CHILDREN)
1810 			break;
1811 		fallthrough;
1812 
1813 	case RUSAGE_SELF:
1814 		thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1815 		utime += tgutime;
1816 		stime += tgstime;
1817 		r->ru_nvcsw += p->signal->nvcsw;
1818 		r->ru_nivcsw += p->signal->nivcsw;
1819 		r->ru_minflt += p->signal->min_flt;
1820 		r->ru_majflt += p->signal->maj_flt;
1821 		r->ru_inblock += p->signal->inblock;
1822 		r->ru_oublock += p->signal->oublock;
1823 		if (maxrss < p->signal->maxrss)
1824 			maxrss = p->signal->maxrss;
1825 		t = p;
1826 		do {
1827 			accumulate_thread_rusage(t, r);
1828 		} while_each_thread(p, t);
1829 		break;
1830 
1831 	default:
1832 		BUG();
1833 	}
1834 	unlock_task_sighand(p, &flags);
1835 
1836 out:
1837 	r->ru_utime = ns_to_kernel_old_timeval(utime);
1838 	r->ru_stime = ns_to_kernel_old_timeval(stime);
1839 
1840 	if (who != RUSAGE_CHILDREN) {
1841 		struct mm_struct *mm = get_task_mm(p);
1842 
1843 		if (mm) {
1844 			setmax_mm_hiwater_rss(&maxrss, mm);
1845 			mmput(mm);
1846 		}
1847 	}
1848 	r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1849 }
1850 
1851 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1852 {
1853 	struct rusage r;
1854 
1855 	if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1856 	    who != RUSAGE_THREAD)
1857 		return -EINVAL;
1858 
1859 	getrusage(current, who, &r);
1860 	return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1861 }
1862 
1863 #ifdef CONFIG_COMPAT
1864 COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru)
1865 {
1866 	struct rusage r;
1867 
1868 	if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1869 	    who != RUSAGE_THREAD)
1870 		return -EINVAL;
1871 
1872 	getrusage(current, who, &r);
1873 	return put_compat_rusage(&r, ru);
1874 }
1875 #endif
1876 
1877 SYSCALL_DEFINE1(umask, int, mask)
1878 {
1879 	mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1880 	return mask;
1881 }
1882 
1883 static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1884 {
1885 	struct fd exe;
1886 	struct inode *inode;
1887 	int err;
1888 
1889 	exe = fdget(fd);
1890 	if (!exe.file)
1891 		return -EBADF;
1892 
1893 	inode = file_inode(exe.file);
1894 
1895 	/*
1896 	 * Because the original mm->exe_file points to executable file, make
1897 	 * sure that this one is executable as well, to avoid breaking an
1898 	 * overall picture.
1899 	 */
1900 	err = -EACCES;
1901 	if (!S_ISREG(inode->i_mode) || path_noexec(&exe.file->f_path))
1902 		goto exit;
1903 
1904 	err = file_permission(exe.file, MAY_EXEC);
1905 	if (err)
1906 		goto exit;
1907 
1908 	err = replace_mm_exe_file(mm, exe.file);
1909 exit:
1910 	fdput(exe);
1911 	return err;
1912 }
1913 
1914 /*
1915  * Check arithmetic relations of passed addresses.
1916  *
1917  * WARNING: we don't require any capability here so be very careful
1918  * in what is allowed for modification from userspace.
1919  */
1920 static int validate_prctl_map_addr(struct prctl_mm_map *prctl_map)
1921 {
1922 	unsigned long mmap_max_addr = TASK_SIZE;
1923 	int error = -EINVAL, i;
1924 
1925 	static const unsigned char offsets[] = {
1926 		offsetof(struct prctl_mm_map, start_code),
1927 		offsetof(struct prctl_mm_map, end_code),
1928 		offsetof(struct prctl_mm_map, start_data),
1929 		offsetof(struct prctl_mm_map, end_data),
1930 		offsetof(struct prctl_mm_map, start_brk),
1931 		offsetof(struct prctl_mm_map, brk),
1932 		offsetof(struct prctl_mm_map, start_stack),
1933 		offsetof(struct prctl_mm_map, arg_start),
1934 		offsetof(struct prctl_mm_map, arg_end),
1935 		offsetof(struct prctl_mm_map, env_start),
1936 		offsetof(struct prctl_mm_map, env_end),
1937 	};
1938 
1939 	/*
1940 	 * Make sure the members are not somewhere outside
1941 	 * of allowed address space.
1942 	 */
1943 	for (i = 0; i < ARRAY_SIZE(offsets); i++) {
1944 		u64 val = *(u64 *)((char *)prctl_map + offsets[i]);
1945 
1946 		if ((unsigned long)val >= mmap_max_addr ||
1947 		    (unsigned long)val < mmap_min_addr)
1948 			goto out;
1949 	}
1950 
1951 	/*
1952 	 * Make sure the pairs are ordered.
1953 	 */
1954 #define __prctl_check_order(__m1, __op, __m2)				\
1955 	((unsigned long)prctl_map->__m1 __op				\
1956 	 (unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1957 	error  = __prctl_check_order(start_code, <, end_code);
1958 	error |= __prctl_check_order(start_data,<=, end_data);
1959 	error |= __prctl_check_order(start_brk, <=, brk);
1960 	error |= __prctl_check_order(arg_start, <=, arg_end);
1961 	error |= __prctl_check_order(env_start, <=, env_end);
1962 	if (error)
1963 		goto out;
1964 #undef __prctl_check_order
1965 
1966 	error = -EINVAL;
1967 
1968 	/*
1969 	 * Neither we should allow to override limits if they set.
1970 	 */
1971 	if (check_data_rlimit(rlimit(RLIMIT_DATA), prctl_map->brk,
1972 			      prctl_map->start_brk, prctl_map->end_data,
1973 			      prctl_map->start_data))
1974 			goto out;
1975 
1976 	error = 0;
1977 out:
1978 	return error;
1979 }
1980 
1981 #ifdef CONFIG_CHECKPOINT_RESTORE
1982 static int prctl_set_mm_map(int opt, const void __user *addr, unsigned long data_size)
1983 {
1984 	struct prctl_mm_map prctl_map = { .exe_fd = (u32)-1, };
1985 	unsigned long user_auxv[AT_VECTOR_SIZE];
1986 	struct mm_struct *mm = current->mm;
1987 	int error;
1988 
1989 	BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
1990 	BUILD_BUG_ON(sizeof(struct prctl_mm_map) > 256);
1991 
1992 	if (opt == PR_SET_MM_MAP_SIZE)
1993 		return put_user((unsigned int)sizeof(prctl_map),
1994 				(unsigned int __user *)addr);
1995 
1996 	if (data_size != sizeof(prctl_map))
1997 		return -EINVAL;
1998 
1999 	if (copy_from_user(&prctl_map, addr, sizeof(prctl_map)))
2000 		return -EFAULT;
2001 
2002 	error = validate_prctl_map_addr(&prctl_map);
2003 	if (error)
2004 		return error;
2005 
2006 	if (prctl_map.auxv_size) {
2007 		/*
2008 		 * Someone is trying to cheat the auxv vector.
2009 		 */
2010 		if (!prctl_map.auxv ||
2011 				prctl_map.auxv_size > sizeof(mm->saved_auxv))
2012 			return -EINVAL;
2013 
2014 		memset(user_auxv, 0, sizeof(user_auxv));
2015 		if (copy_from_user(user_auxv,
2016 				   (const void __user *)prctl_map.auxv,
2017 				   prctl_map.auxv_size))
2018 			return -EFAULT;
2019 
2020 		/* Last entry must be AT_NULL as specification requires */
2021 		user_auxv[AT_VECTOR_SIZE - 2] = AT_NULL;
2022 		user_auxv[AT_VECTOR_SIZE - 1] = AT_NULL;
2023 	}
2024 
2025 	if (prctl_map.exe_fd != (u32)-1) {
2026 		/*
2027 		 * Check if the current user is checkpoint/restore capable.
2028 		 * At the time of this writing, it checks for CAP_SYS_ADMIN
2029 		 * or CAP_CHECKPOINT_RESTORE.
2030 		 * Note that a user with access to ptrace can masquerade an
2031 		 * arbitrary program as any executable, even setuid ones.
2032 		 * This may have implications in the tomoyo subsystem.
2033 		 */
2034 		if (!checkpoint_restore_ns_capable(current_user_ns()))
2035 			return -EPERM;
2036 
2037 		error = prctl_set_mm_exe_file(mm, prctl_map.exe_fd);
2038 		if (error)
2039 			return error;
2040 	}
2041 
2042 	/*
2043 	 * arg_lock protects concurrent updates but we still need mmap_lock for
2044 	 * read to exclude races with sys_brk.
2045 	 */
2046 	mmap_read_lock(mm);
2047 
2048 	/*
2049 	 * We don't validate if these members are pointing to
2050 	 * real present VMAs because application may have correspond
2051 	 * VMAs already unmapped and kernel uses these members for statistics
2052 	 * output in procfs mostly, except
2053 	 *
2054 	 *  - @start_brk/@brk which are used in do_brk_flags but kernel lookups
2055 	 *    for VMAs when updating these members so anything wrong written
2056 	 *    here cause kernel to swear at userspace program but won't lead
2057 	 *    to any problem in kernel itself
2058 	 */
2059 
2060 	spin_lock(&mm->arg_lock);
2061 	mm->start_code	= prctl_map.start_code;
2062 	mm->end_code	= prctl_map.end_code;
2063 	mm->start_data	= prctl_map.start_data;
2064 	mm->end_data	= prctl_map.end_data;
2065 	mm->start_brk	= prctl_map.start_brk;
2066 	mm->brk		= prctl_map.brk;
2067 	mm->start_stack	= prctl_map.start_stack;
2068 	mm->arg_start	= prctl_map.arg_start;
2069 	mm->arg_end	= prctl_map.arg_end;
2070 	mm->env_start	= prctl_map.env_start;
2071 	mm->env_end	= prctl_map.env_end;
2072 	spin_unlock(&mm->arg_lock);
2073 
2074 	/*
2075 	 * Note this update of @saved_auxv is lockless thus
2076 	 * if someone reads this member in procfs while we're
2077 	 * updating -- it may get partly updated results. It's
2078 	 * known and acceptable trade off: we leave it as is to
2079 	 * not introduce additional locks here making the kernel
2080 	 * more complex.
2081 	 */
2082 	if (prctl_map.auxv_size)
2083 		memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv));
2084 
2085 	mmap_read_unlock(mm);
2086 	return 0;
2087 }
2088 #endif /* CONFIG_CHECKPOINT_RESTORE */
2089 
2090 static int prctl_set_auxv(struct mm_struct *mm, unsigned long addr,
2091 			  unsigned long len)
2092 {
2093 	/*
2094 	 * This doesn't move the auxiliary vector itself since it's pinned to
2095 	 * mm_struct, but it permits filling the vector with new values.  It's
2096 	 * up to the caller to provide sane values here, otherwise userspace
2097 	 * tools which use this vector might be unhappy.
2098 	 */
2099 	unsigned long user_auxv[AT_VECTOR_SIZE] = {};
2100 
2101 	if (len > sizeof(user_auxv))
2102 		return -EINVAL;
2103 
2104 	if (copy_from_user(user_auxv, (const void __user *)addr, len))
2105 		return -EFAULT;
2106 
2107 	/* Make sure the last entry is always AT_NULL */
2108 	user_auxv[AT_VECTOR_SIZE - 2] = 0;
2109 	user_auxv[AT_VECTOR_SIZE - 1] = 0;
2110 
2111 	BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
2112 
2113 	task_lock(current);
2114 	memcpy(mm->saved_auxv, user_auxv, len);
2115 	task_unlock(current);
2116 
2117 	return 0;
2118 }
2119 
2120 static int prctl_set_mm(int opt, unsigned long addr,
2121 			unsigned long arg4, unsigned long arg5)
2122 {
2123 	struct mm_struct *mm = current->mm;
2124 	struct prctl_mm_map prctl_map = {
2125 		.auxv = NULL,
2126 		.auxv_size = 0,
2127 		.exe_fd = -1,
2128 	};
2129 	struct vm_area_struct *vma;
2130 	int error;
2131 
2132 	if (arg5 || (arg4 && (opt != PR_SET_MM_AUXV &&
2133 			      opt != PR_SET_MM_MAP &&
2134 			      opt != PR_SET_MM_MAP_SIZE)))
2135 		return -EINVAL;
2136 
2137 #ifdef CONFIG_CHECKPOINT_RESTORE
2138 	if (opt == PR_SET_MM_MAP || opt == PR_SET_MM_MAP_SIZE)
2139 		return prctl_set_mm_map(opt, (const void __user *)addr, arg4);
2140 #endif
2141 
2142 	if (!capable(CAP_SYS_RESOURCE))
2143 		return -EPERM;
2144 
2145 	if (opt == PR_SET_MM_EXE_FILE)
2146 		return prctl_set_mm_exe_file(mm, (unsigned int)addr);
2147 
2148 	if (opt == PR_SET_MM_AUXV)
2149 		return prctl_set_auxv(mm, addr, arg4);
2150 
2151 	if (addr >= TASK_SIZE || addr < mmap_min_addr)
2152 		return -EINVAL;
2153 
2154 	error = -EINVAL;
2155 
2156 	/*
2157 	 * arg_lock protects concurrent updates of arg boundaries, we need
2158 	 * mmap_lock for a) concurrent sys_brk, b) finding VMA for addr
2159 	 * validation.
2160 	 */
2161 	mmap_read_lock(mm);
2162 	vma = find_vma(mm, addr);
2163 
2164 	spin_lock(&mm->arg_lock);
2165 	prctl_map.start_code	= mm->start_code;
2166 	prctl_map.end_code	= mm->end_code;
2167 	prctl_map.start_data	= mm->start_data;
2168 	prctl_map.end_data	= mm->end_data;
2169 	prctl_map.start_brk	= mm->start_brk;
2170 	prctl_map.brk		= mm->brk;
2171 	prctl_map.start_stack	= mm->start_stack;
2172 	prctl_map.arg_start	= mm->arg_start;
2173 	prctl_map.arg_end	= mm->arg_end;
2174 	prctl_map.env_start	= mm->env_start;
2175 	prctl_map.env_end	= mm->env_end;
2176 
2177 	switch (opt) {
2178 	case PR_SET_MM_START_CODE:
2179 		prctl_map.start_code = addr;
2180 		break;
2181 	case PR_SET_MM_END_CODE:
2182 		prctl_map.end_code = addr;
2183 		break;
2184 	case PR_SET_MM_START_DATA:
2185 		prctl_map.start_data = addr;
2186 		break;
2187 	case PR_SET_MM_END_DATA:
2188 		prctl_map.end_data = addr;
2189 		break;
2190 	case PR_SET_MM_START_STACK:
2191 		prctl_map.start_stack = addr;
2192 		break;
2193 	case PR_SET_MM_START_BRK:
2194 		prctl_map.start_brk = addr;
2195 		break;
2196 	case PR_SET_MM_BRK:
2197 		prctl_map.brk = addr;
2198 		break;
2199 	case PR_SET_MM_ARG_START:
2200 		prctl_map.arg_start = addr;
2201 		break;
2202 	case PR_SET_MM_ARG_END:
2203 		prctl_map.arg_end = addr;
2204 		break;
2205 	case PR_SET_MM_ENV_START:
2206 		prctl_map.env_start = addr;
2207 		break;
2208 	case PR_SET_MM_ENV_END:
2209 		prctl_map.env_end = addr;
2210 		break;
2211 	default:
2212 		goto out;
2213 	}
2214 
2215 	error = validate_prctl_map_addr(&prctl_map);
2216 	if (error)
2217 		goto out;
2218 
2219 	switch (opt) {
2220 	/*
2221 	 * If command line arguments and environment
2222 	 * are placed somewhere else on stack, we can
2223 	 * set them up here, ARG_START/END to setup
2224 	 * command line arguments and ENV_START/END
2225 	 * for environment.
2226 	 */
2227 	case PR_SET_MM_START_STACK:
2228 	case PR_SET_MM_ARG_START:
2229 	case PR_SET_MM_ARG_END:
2230 	case PR_SET_MM_ENV_START:
2231 	case PR_SET_MM_ENV_END:
2232 		if (!vma) {
2233 			error = -EFAULT;
2234 			goto out;
2235 		}
2236 	}
2237 
2238 	mm->start_code	= prctl_map.start_code;
2239 	mm->end_code	= prctl_map.end_code;
2240 	mm->start_data	= prctl_map.start_data;
2241 	mm->end_data	= prctl_map.end_data;
2242 	mm->start_brk	= prctl_map.start_brk;
2243 	mm->brk		= prctl_map.brk;
2244 	mm->start_stack	= prctl_map.start_stack;
2245 	mm->arg_start	= prctl_map.arg_start;
2246 	mm->arg_end	= prctl_map.arg_end;
2247 	mm->env_start	= prctl_map.env_start;
2248 	mm->env_end	= prctl_map.env_end;
2249 
2250 	error = 0;
2251 out:
2252 	spin_unlock(&mm->arg_lock);
2253 	mmap_read_unlock(mm);
2254 	return error;
2255 }
2256 
2257 #ifdef CONFIG_CHECKPOINT_RESTORE
2258 static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr)
2259 {
2260 	return put_user(me->clear_child_tid, tid_addr);
2261 }
2262 #else
2263 static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr)
2264 {
2265 	return -EINVAL;
2266 }
2267 #endif
2268 
2269 static int propagate_has_child_subreaper(struct task_struct *p, void *data)
2270 {
2271 	/*
2272 	 * If task has has_child_subreaper - all its descendants
2273 	 * already have these flag too and new descendants will
2274 	 * inherit it on fork, skip them.
2275 	 *
2276 	 * If we've found child_reaper - skip descendants in
2277 	 * it's subtree as they will never get out pidns.
2278 	 */
2279 	if (p->signal->has_child_subreaper ||
2280 	    is_child_reaper(task_pid(p)))
2281 		return 0;
2282 
2283 	p->signal->has_child_subreaper = 1;
2284 	return 1;
2285 }
2286 
2287 int __weak arch_prctl_spec_ctrl_get(struct task_struct *t, unsigned long which)
2288 {
2289 	return -EINVAL;
2290 }
2291 
2292 int __weak arch_prctl_spec_ctrl_set(struct task_struct *t, unsigned long which,
2293 				    unsigned long ctrl)
2294 {
2295 	return -EINVAL;
2296 }
2297 
2298 #define PR_IO_FLUSHER (PF_MEMALLOC_NOIO | PF_LOCAL_THROTTLE)
2299 
2300 #ifdef CONFIG_ANON_VMA_NAME
2301 
2302 #define ANON_VMA_NAME_MAX_LEN		80
2303 #define ANON_VMA_NAME_INVALID_CHARS	"\\`$[]"
2304 
2305 static inline bool is_valid_name_char(char ch)
2306 {
2307 	/* printable ascii characters, excluding ANON_VMA_NAME_INVALID_CHARS */
2308 	return ch > 0x1f && ch < 0x7f &&
2309 		!strchr(ANON_VMA_NAME_INVALID_CHARS, ch);
2310 }
2311 
2312 static int prctl_set_vma(unsigned long opt, unsigned long addr,
2313 			 unsigned long size, unsigned long arg)
2314 {
2315 	struct mm_struct *mm = current->mm;
2316 	const char __user *uname;
2317 	struct anon_vma_name *anon_name = NULL;
2318 	int error;
2319 
2320 	switch (opt) {
2321 	case PR_SET_VMA_ANON_NAME:
2322 		uname = (const char __user *)arg;
2323 		if (uname) {
2324 			char *name, *pch;
2325 
2326 			name = strndup_user(uname, ANON_VMA_NAME_MAX_LEN);
2327 			if (IS_ERR(name))
2328 				return PTR_ERR(name);
2329 
2330 			for (pch = name; *pch != '\0'; pch++) {
2331 				if (!is_valid_name_char(*pch)) {
2332 					kfree(name);
2333 					return -EINVAL;
2334 				}
2335 			}
2336 			/* anon_vma has its own copy */
2337 			anon_name = anon_vma_name_alloc(name);
2338 			kfree(name);
2339 			if (!anon_name)
2340 				return -ENOMEM;
2341 
2342 		}
2343 
2344 		mmap_write_lock(mm);
2345 		error = madvise_set_anon_name(mm, addr, size, anon_name);
2346 		mmap_write_unlock(mm);
2347 		anon_vma_name_put(anon_name);
2348 		break;
2349 	default:
2350 		error = -EINVAL;
2351 	}
2352 
2353 	return error;
2354 }
2355 
2356 #else /* CONFIG_ANON_VMA_NAME */
2357 static int prctl_set_vma(unsigned long opt, unsigned long start,
2358 			 unsigned long size, unsigned long arg)
2359 {
2360 	return -EINVAL;
2361 }
2362 #endif /* CONFIG_ANON_VMA_NAME */
2363 
2364 static inline int prctl_set_mdwe(unsigned long bits, unsigned long arg3,
2365 				 unsigned long arg4, unsigned long arg5)
2366 {
2367 	if (arg3 || arg4 || arg5)
2368 		return -EINVAL;
2369 
2370 	if (bits & ~(PR_MDWE_REFUSE_EXEC_GAIN))
2371 		return -EINVAL;
2372 
2373 	if (bits & PR_MDWE_REFUSE_EXEC_GAIN)
2374 		set_bit(MMF_HAS_MDWE, &current->mm->flags);
2375 	else if (test_bit(MMF_HAS_MDWE, &current->mm->flags))
2376 		return -EPERM; /* Cannot unset the flag */
2377 
2378 	return 0;
2379 }
2380 
2381 static inline int prctl_get_mdwe(unsigned long arg2, unsigned long arg3,
2382 				 unsigned long arg4, unsigned long arg5)
2383 {
2384 	if (arg2 || arg3 || arg4 || arg5)
2385 		return -EINVAL;
2386 
2387 	return test_bit(MMF_HAS_MDWE, &current->mm->flags) ?
2388 		PR_MDWE_REFUSE_EXEC_GAIN : 0;
2389 }
2390 
2391 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2392 		unsigned long, arg4, unsigned long, arg5)
2393 {
2394 	struct task_struct *me = current;
2395 	unsigned char comm[sizeof(me->comm)];
2396 	long error;
2397 
2398 	error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2399 	if (error != -ENOSYS)
2400 		return error;
2401 
2402 	error = 0;
2403 	switch (option) {
2404 	case PR_SET_PDEATHSIG:
2405 		if (!valid_signal(arg2)) {
2406 			error = -EINVAL;
2407 			break;
2408 		}
2409 		me->pdeath_signal = arg2;
2410 		break;
2411 	case PR_GET_PDEATHSIG:
2412 		error = put_user(me->pdeath_signal, (int __user *)arg2);
2413 		break;
2414 	case PR_GET_DUMPABLE:
2415 		error = get_dumpable(me->mm);
2416 		break;
2417 	case PR_SET_DUMPABLE:
2418 		if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) {
2419 			error = -EINVAL;
2420 			break;
2421 		}
2422 		set_dumpable(me->mm, arg2);
2423 		break;
2424 
2425 	case PR_SET_UNALIGN:
2426 		error = SET_UNALIGN_CTL(me, arg2);
2427 		break;
2428 	case PR_GET_UNALIGN:
2429 		error = GET_UNALIGN_CTL(me, arg2);
2430 		break;
2431 	case PR_SET_FPEMU:
2432 		error = SET_FPEMU_CTL(me, arg2);
2433 		break;
2434 	case PR_GET_FPEMU:
2435 		error = GET_FPEMU_CTL(me, arg2);
2436 		break;
2437 	case PR_SET_FPEXC:
2438 		error = SET_FPEXC_CTL(me, arg2);
2439 		break;
2440 	case PR_GET_FPEXC:
2441 		error = GET_FPEXC_CTL(me, arg2);
2442 		break;
2443 	case PR_GET_TIMING:
2444 		error = PR_TIMING_STATISTICAL;
2445 		break;
2446 	case PR_SET_TIMING:
2447 		if (arg2 != PR_TIMING_STATISTICAL)
2448 			error = -EINVAL;
2449 		break;
2450 	case PR_SET_NAME:
2451 		comm[sizeof(me->comm) - 1] = 0;
2452 		if (strncpy_from_user(comm, (char __user *)arg2,
2453 				      sizeof(me->comm) - 1) < 0)
2454 			return -EFAULT;
2455 		set_task_comm(me, comm);
2456 		proc_comm_connector(me);
2457 		break;
2458 	case PR_GET_NAME:
2459 		get_task_comm(comm, me);
2460 		if (copy_to_user((char __user *)arg2, comm, sizeof(comm)))
2461 			return -EFAULT;
2462 		break;
2463 	case PR_GET_ENDIAN:
2464 		error = GET_ENDIAN(me, arg2);
2465 		break;
2466 	case PR_SET_ENDIAN:
2467 		error = SET_ENDIAN(me, arg2);
2468 		break;
2469 	case PR_GET_SECCOMP:
2470 		error = prctl_get_seccomp();
2471 		break;
2472 	case PR_SET_SECCOMP:
2473 		error = prctl_set_seccomp(arg2, (char __user *)arg3);
2474 		break;
2475 	case PR_GET_TSC:
2476 		error = GET_TSC_CTL(arg2);
2477 		break;
2478 	case PR_SET_TSC:
2479 		error = SET_TSC_CTL(arg2);
2480 		break;
2481 	case PR_TASK_PERF_EVENTS_DISABLE:
2482 		error = perf_event_task_disable();
2483 		break;
2484 	case PR_TASK_PERF_EVENTS_ENABLE:
2485 		error = perf_event_task_enable();
2486 		break;
2487 	case PR_GET_TIMERSLACK:
2488 		if (current->timer_slack_ns > ULONG_MAX)
2489 			error = ULONG_MAX;
2490 		else
2491 			error = current->timer_slack_ns;
2492 		break;
2493 	case PR_SET_TIMERSLACK:
2494 		if (arg2 <= 0)
2495 			current->timer_slack_ns =
2496 					current->default_timer_slack_ns;
2497 		else
2498 			current->timer_slack_ns = arg2;
2499 		break;
2500 	case PR_MCE_KILL:
2501 		if (arg4 | arg5)
2502 			return -EINVAL;
2503 		switch (arg2) {
2504 		case PR_MCE_KILL_CLEAR:
2505 			if (arg3 != 0)
2506 				return -EINVAL;
2507 			current->flags &= ~PF_MCE_PROCESS;
2508 			break;
2509 		case PR_MCE_KILL_SET:
2510 			current->flags |= PF_MCE_PROCESS;
2511 			if (arg3 == PR_MCE_KILL_EARLY)
2512 				current->flags |= PF_MCE_EARLY;
2513 			else if (arg3 == PR_MCE_KILL_LATE)
2514 				current->flags &= ~PF_MCE_EARLY;
2515 			else if (arg3 == PR_MCE_KILL_DEFAULT)
2516 				current->flags &=
2517 						~(PF_MCE_EARLY|PF_MCE_PROCESS);
2518 			else
2519 				return -EINVAL;
2520 			break;
2521 		default:
2522 			return -EINVAL;
2523 		}
2524 		break;
2525 	case PR_MCE_KILL_GET:
2526 		if (arg2 | arg3 | arg4 | arg5)
2527 			return -EINVAL;
2528 		if (current->flags & PF_MCE_PROCESS)
2529 			error = (current->flags & PF_MCE_EARLY) ?
2530 				PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2531 		else
2532 			error = PR_MCE_KILL_DEFAULT;
2533 		break;
2534 	case PR_SET_MM:
2535 		error = prctl_set_mm(arg2, arg3, arg4, arg5);
2536 		break;
2537 	case PR_GET_TID_ADDRESS:
2538 		error = prctl_get_tid_address(me, (int __user * __user *)arg2);
2539 		break;
2540 	case PR_SET_CHILD_SUBREAPER:
2541 		me->signal->is_child_subreaper = !!arg2;
2542 		if (!arg2)
2543 			break;
2544 
2545 		walk_process_tree(me, propagate_has_child_subreaper, NULL);
2546 		break;
2547 	case PR_GET_CHILD_SUBREAPER:
2548 		error = put_user(me->signal->is_child_subreaper,
2549 				 (int __user *)arg2);
2550 		break;
2551 	case PR_SET_NO_NEW_PRIVS:
2552 		if (arg2 != 1 || arg3 || arg4 || arg5)
2553 			return -EINVAL;
2554 
2555 		task_set_no_new_privs(current);
2556 		break;
2557 	case PR_GET_NO_NEW_PRIVS:
2558 		if (arg2 || arg3 || arg4 || arg5)
2559 			return -EINVAL;
2560 		return task_no_new_privs(current) ? 1 : 0;
2561 	case PR_GET_THP_DISABLE:
2562 		if (arg2 || arg3 || arg4 || arg5)
2563 			return -EINVAL;
2564 		error = !!test_bit(MMF_DISABLE_THP, &me->mm->flags);
2565 		break;
2566 	case PR_SET_THP_DISABLE:
2567 		if (arg3 || arg4 || arg5)
2568 			return -EINVAL;
2569 		if (mmap_write_lock_killable(me->mm))
2570 			return -EINTR;
2571 		if (arg2)
2572 			set_bit(MMF_DISABLE_THP, &me->mm->flags);
2573 		else
2574 			clear_bit(MMF_DISABLE_THP, &me->mm->flags);
2575 		mmap_write_unlock(me->mm);
2576 		break;
2577 	case PR_MPX_ENABLE_MANAGEMENT:
2578 	case PR_MPX_DISABLE_MANAGEMENT:
2579 		/* No longer implemented: */
2580 		return -EINVAL;
2581 	case PR_SET_FP_MODE:
2582 		error = SET_FP_MODE(me, arg2);
2583 		break;
2584 	case PR_GET_FP_MODE:
2585 		error = GET_FP_MODE(me);
2586 		break;
2587 	case PR_SVE_SET_VL:
2588 		error = SVE_SET_VL(arg2);
2589 		break;
2590 	case PR_SVE_GET_VL:
2591 		error = SVE_GET_VL();
2592 		break;
2593 	case PR_SME_SET_VL:
2594 		error = SME_SET_VL(arg2);
2595 		break;
2596 	case PR_SME_GET_VL:
2597 		error = SME_GET_VL();
2598 		break;
2599 	case PR_GET_SPECULATION_CTRL:
2600 		if (arg3 || arg4 || arg5)
2601 			return -EINVAL;
2602 		error = arch_prctl_spec_ctrl_get(me, arg2);
2603 		break;
2604 	case PR_SET_SPECULATION_CTRL:
2605 		if (arg4 || arg5)
2606 			return -EINVAL;
2607 		error = arch_prctl_spec_ctrl_set(me, arg2, arg3);
2608 		break;
2609 	case PR_PAC_RESET_KEYS:
2610 		if (arg3 || arg4 || arg5)
2611 			return -EINVAL;
2612 		error = PAC_RESET_KEYS(me, arg2);
2613 		break;
2614 	case PR_PAC_SET_ENABLED_KEYS:
2615 		if (arg4 || arg5)
2616 			return -EINVAL;
2617 		error = PAC_SET_ENABLED_KEYS(me, arg2, arg3);
2618 		break;
2619 	case PR_PAC_GET_ENABLED_KEYS:
2620 		if (arg2 || arg3 || arg4 || arg5)
2621 			return -EINVAL;
2622 		error = PAC_GET_ENABLED_KEYS(me);
2623 		break;
2624 	case PR_SET_TAGGED_ADDR_CTRL:
2625 		if (arg3 || arg4 || arg5)
2626 			return -EINVAL;
2627 		error = SET_TAGGED_ADDR_CTRL(arg2);
2628 		break;
2629 	case PR_GET_TAGGED_ADDR_CTRL:
2630 		if (arg2 || arg3 || arg4 || arg5)
2631 			return -EINVAL;
2632 		error = GET_TAGGED_ADDR_CTRL();
2633 		break;
2634 	case PR_SET_IO_FLUSHER:
2635 		if (!capable(CAP_SYS_RESOURCE))
2636 			return -EPERM;
2637 
2638 		if (arg3 || arg4 || arg5)
2639 			return -EINVAL;
2640 
2641 		if (arg2 == 1)
2642 			current->flags |= PR_IO_FLUSHER;
2643 		else if (!arg2)
2644 			current->flags &= ~PR_IO_FLUSHER;
2645 		else
2646 			return -EINVAL;
2647 		break;
2648 	case PR_GET_IO_FLUSHER:
2649 		if (!capable(CAP_SYS_RESOURCE))
2650 			return -EPERM;
2651 
2652 		if (arg2 || arg3 || arg4 || arg5)
2653 			return -EINVAL;
2654 
2655 		error = (current->flags & PR_IO_FLUSHER) == PR_IO_FLUSHER;
2656 		break;
2657 	case PR_SET_SYSCALL_USER_DISPATCH:
2658 		error = set_syscall_user_dispatch(arg2, arg3, arg4,
2659 						  (char __user *) arg5);
2660 		break;
2661 #ifdef CONFIG_SCHED_CORE
2662 	case PR_SCHED_CORE:
2663 		error = sched_core_share_pid(arg2, arg3, arg4, arg5);
2664 		break;
2665 #endif
2666 	case PR_SET_MDWE:
2667 		error = prctl_set_mdwe(arg2, arg3, arg4, arg5);
2668 		break;
2669 	case PR_GET_MDWE:
2670 		error = prctl_get_mdwe(arg2, arg3, arg4, arg5);
2671 		break;
2672 	case PR_SET_VMA:
2673 		error = prctl_set_vma(arg2, arg3, arg4, arg5);
2674 		break;
2675 	default:
2676 		error = -EINVAL;
2677 		break;
2678 	}
2679 	return error;
2680 }
2681 
2682 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2683 		struct getcpu_cache __user *, unused)
2684 {
2685 	int err = 0;
2686 	int cpu = raw_smp_processor_id();
2687 
2688 	if (cpup)
2689 		err |= put_user(cpu, cpup);
2690 	if (nodep)
2691 		err |= put_user(cpu_to_node(cpu), nodep);
2692 	return err ? -EFAULT : 0;
2693 }
2694 
2695 /**
2696  * do_sysinfo - fill in sysinfo struct
2697  * @info: pointer to buffer to fill
2698  */
2699 static int do_sysinfo(struct sysinfo *info)
2700 {
2701 	unsigned long mem_total, sav_total;
2702 	unsigned int mem_unit, bitcount;
2703 	struct timespec64 tp;
2704 
2705 	memset(info, 0, sizeof(struct sysinfo));
2706 
2707 	ktime_get_boottime_ts64(&tp);
2708 	timens_add_boottime(&tp);
2709 	info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
2710 
2711 	get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
2712 
2713 	info->procs = nr_threads;
2714 
2715 	si_meminfo(info);
2716 	si_swapinfo(info);
2717 
2718 	/*
2719 	 * If the sum of all the available memory (i.e. ram + swap)
2720 	 * is less than can be stored in a 32 bit unsigned long then
2721 	 * we can be binary compatible with 2.2.x kernels.  If not,
2722 	 * well, in that case 2.2.x was broken anyways...
2723 	 *
2724 	 *  -Erik Andersen <andersee@debian.org>
2725 	 */
2726 
2727 	mem_total = info->totalram + info->totalswap;
2728 	if (mem_total < info->totalram || mem_total < info->totalswap)
2729 		goto out;
2730 	bitcount = 0;
2731 	mem_unit = info->mem_unit;
2732 	while (mem_unit > 1) {
2733 		bitcount++;
2734 		mem_unit >>= 1;
2735 		sav_total = mem_total;
2736 		mem_total <<= 1;
2737 		if (mem_total < sav_total)
2738 			goto out;
2739 	}
2740 
2741 	/*
2742 	 * If mem_total did not overflow, multiply all memory values by
2743 	 * info->mem_unit and set it to 1.  This leaves things compatible
2744 	 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2745 	 * kernels...
2746 	 */
2747 
2748 	info->mem_unit = 1;
2749 	info->totalram <<= bitcount;
2750 	info->freeram <<= bitcount;
2751 	info->sharedram <<= bitcount;
2752 	info->bufferram <<= bitcount;
2753 	info->totalswap <<= bitcount;
2754 	info->freeswap <<= bitcount;
2755 	info->totalhigh <<= bitcount;
2756 	info->freehigh <<= bitcount;
2757 
2758 out:
2759 	return 0;
2760 }
2761 
2762 SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
2763 {
2764 	struct sysinfo val;
2765 
2766 	do_sysinfo(&val);
2767 
2768 	if (copy_to_user(info, &val, sizeof(struct sysinfo)))
2769 		return -EFAULT;
2770 
2771 	return 0;
2772 }
2773 
2774 #ifdef CONFIG_COMPAT
2775 struct compat_sysinfo {
2776 	s32 uptime;
2777 	u32 loads[3];
2778 	u32 totalram;
2779 	u32 freeram;
2780 	u32 sharedram;
2781 	u32 bufferram;
2782 	u32 totalswap;
2783 	u32 freeswap;
2784 	u16 procs;
2785 	u16 pad;
2786 	u32 totalhigh;
2787 	u32 freehigh;
2788 	u32 mem_unit;
2789 	char _f[20-2*sizeof(u32)-sizeof(int)];
2790 };
2791 
2792 COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2793 {
2794 	struct sysinfo s;
2795 	struct compat_sysinfo s_32;
2796 
2797 	do_sysinfo(&s);
2798 
2799 	/* Check to see if any memory value is too large for 32-bit and scale
2800 	 *  down if needed
2801 	 */
2802 	if (upper_32_bits(s.totalram) || upper_32_bits(s.totalswap)) {
2803 		int bitcount = 0;
2804 
2805 		while (s.mem_unit < PAGE_SIZE) {
2806 			s.mem_unit <<= 1;
2807 			bitcount++;
2808 		}
2809 
2810 		s.totalram >>= bitcount;
2811 		s.freeram >>= bitcount;
2812 		s.sharedram >>= bitcount;
2813 		s.bufferram >>= bitcount;
2814 		s.totalswap >>= bitcount;
2815 		s.freeswap >>= bitcount;
2816 		s.totalhigh >>= bitcount;
2817 		s.freehigh >>= bitcount;
2818 	}
2819 
2820 	memset(&s_32, 0, sizeof(s_32));
2821 	s_32.uptime = s.uptime;
2822 	s_32.loads[0] = s.loads[0];
2823 	s_32.loads[1] = s.loads[1];
2824 	s_32.loads[2] = s.loads[2];
2825 	s_32.totalram = s.totalram;
2826 	s_32.freeram = s.freeram;
2827 	s_32.sharedram = s.sharedram;
2828 	s_32.bufferram = s.bufferram;
2829 	s_32.totalswap = s.totalswap;
2830 	s_32.freeswap = s.freeswap;
2831 	s_32.procs = s.procs;
2832 	s_32.totalhigh = s.totalhigh;
2833 	s_32.freehigh = s.freehigh;
2834 	s_32.mem_unit = s.mem_unit;
2835 	if (copy_to_user(info, &s_32, sizeof(s_32)))
2836 		return -EFAULT;
2837 	return 0;
2838 }
2839 #endif /* CONFIG_COMPAT */
2840