xref: /openbmc/linux/kernel/sys.c (revision ea3de9ce)
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 
668 	kruid = make_kuid(ns, ruid);
669 	keuid = make_kuid(ns, euid);
670 	ksuid = make_kuid(ns, suid);
671 
672 	if ((ruid != (uid_t) -1) && !uid_valid(kruid))
673 		return -EINVAL;
674 
675 	if ((euid != (uid_t) -1) && !uid_valid(keuid))
676 		return -EINVAL;
677 
678 	if ((suid != (uid_t) -1) && !uid_valid(ksuid))
679 		return -EINVAL;
680 
681 	new = prepare_creds();
682 	if (!new)
683 		return -ENOMEM;
684 
685 	old = current_cred();
686 
687 	retval = -EPERM;
688 	if (!ns_capable_setid(old->user_ns, CAP_SETUID)) {
689 		if (ruid != (uid_t) -1        && !uid_eq(kruid, old->uid) &&
690 		    !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid))
691 			goto error;
692 		if (euid != (uid_t) -1        && !uid_eq(keuid, old->uid) &&
693 		    !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid))
694 			goto error;
695 		if (suid != (uid_t) -1        && !uid_eq(ksuid, old->uid) &&
696 		    !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid))
697 			goto error;
698 	}
699 
700 	if (ruid != (uid_t) -1) {
701 		new->uid = kruid;
702 		if (!uid_eq(kruid, old->uid)) {
703 			retval = set_user(new);
704 			if (retval < 0)
705 				goto error;
706 		}
707 	}
708 	if (euid != (uid_t) -1)
709 		new->euid = keuid;
710 	if (suid != (uid_t) -1)
711 		new->suid = ksuid;
712 	new->fsuid = new->euid;
713 
714 	retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
715 	if (retval < 0)
716 		goto error;
717 
718 	retval = set_cred_ucounts(new);
719 	if (retval < 0)
720 		goto error;
721 
722 	flag_nproc_exceeded(new);
723 	return commit_creds(new);
724 
725 error:
726 	abort_creds(new);
727 	return retval;
728 }
729 
730 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
731 {
732 	return __sys_setresuid(ruid, euid, suid);
733 }
734 
735 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
736 {
737 	const struct cred *cred = current_cred();
738 	int retval;
739 	uid_t ruid, euid, suid;
740 
741 	ruid = from_kuid_munged(cred->user_ns, cred->uid);
742 	euid = from_kuid_munged(cred->user_ns, cred->euid);
743 	suid = from_kuid_munged(cred->user_ns, cred->suid);
744 
745 	retval = put_user(ruid, ruidp);
746 	if (!retval) {
747 		retval = put_user(euid, euidp);
748 		if (!retval)
749 			return put_user(suid, suidp);
750 	}
751 	return retval;
752 }
753 
754 /*
755  * Same as above, but for rgid, egid, sgid.
756  */
757 long __sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
758 {
759 	struct user_namespace *ns = current_user_ns();
760 	const struct cred *old;
761 	struct cred *new;
762 	int retval;
763 	kgid_t krgid, kegid, ksgid;
764 
765 	krgid = make_kgid(ns, rgid);
766 	kegid = make_kgid(ns, egid);
767 	ksgid = make_kgid(ns, sgid);
768 
769 	if ((rgid != (gid_t) -1) && !gid_valid(krgid))
770 		return -EINVAL;
771 	if ((egid != (gid_t) -1) && !gid_valid(kegid))
772 		return -EINVAL;
773 	if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
774 		return -EINVAL;
775 
776 	new = prepare_creds();
777 	if (!new)
778 		return -ENOMEM;
779 	old = current_cred();
780 
781 	retval = -EPERM;
782 	if (!ns_capable_setid(old->user_ns, CAP_SETGID)) {
783 		if (rgid != (gid_t) -1        && !gid_eq(krgid, old->gid) &&
784 		    !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid))
785 			goto error;
786 		if (egid != (gid_t) -1        && !gid_eq(kegid, old->gid) &&
787 		    !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid))
788 			goto error;
789 		if (sgid != (gid_t) -1        && !gid_eq(ksgid, old->gid) &&
790 		    !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid))
791 			goto error;
792 	}
793 
794 	if (rgid != (gid_t) -1)
795 		new->gid = krgid;
796 	if (egid != (gid_t) -1)
797 		new->egid = kegid;
798 	if (sgid != (gid_t) -1)
799 		new->sgid = ksgid;
800 	new->fsgid = new->egid;
801 
802 	retval = security_task_fix_setgid(new, old, LSM_SETID_RES);
803 	if (retval < 0)
804 		goto error;
805 
806 	return commit_creds(new);
807 
808 error:
809 	abort_creds(new);
810 	return retval;
811 }
812 
813 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
814 {
815 	return __sys_setresgid(rgid, egid, sgid);
816 }
817 
818 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
819 {
820 	const struct cred *cred = current_cred();
821 	int retval;
822 	gid_t rgid, egid, sgid;
823 
824 	rgid = from_kgid_munged(cred->user_ns, cred->gid);
825 	egid = from_kgid_munged(cred->user_ns, cred->egid);
826 	sgid = from_kgid_munged(cred->user_ns, cred->sgid);
827 
828 	retval = put_user(rgid, rgidp);
829 	if (!retval) {
830 		retval = put_user(egid, egidp);
831 		if (!retval)
832 			retval = put_user(sgid, sgidp);
833 	}
834 
835 	return retval;
836 }
837 
838 
839 /*
840  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
841  * is used for "access()" and for the NFS daemon (letting nfsd stay at
842  * whatever uid it wants to). It normally shadows "euid", except when
843  * explicitly set by setfsuid() or for access..
844  */
845 long __sys_setfsuid(uid_t uid)
846 {
847 	const struct cred *old;
848 	struct cred *new;
849 	uid_t old_fsuid;
850 	kuid_t kuid;
851 
852 	old = current_cred();
853 	old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
854 
855 	kuid = make_kuid(old->user_ns, uid);
856 	if (!uid_valid(kuid))
857 		return old_fsuid;
858 
859 	new = prepare_creds();
860 	if (!new)
861 		return old_fsuid;
862 
863 	if (uid_eq(kuid, old->uid)  || uid_eq(kuid, old->euid)  ||
864 	    uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
865 	    ns_capable_setid(old->user_ns, CAP_SETUID)) {
866 		if (!uid_eq(kuid, old->fsuid)) {
867 			new->fsuid = kuid;
868 			if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
869 				goto change_okay;
870 		}
871 	}
872 
873 	abort_creds(new);
874 	return old_fsuid;
875 
876 change_okay:
877 	commit_creds(new);
878 	return old_fsuid;
879 }
880 
881 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
882 {
883 	return __sys_setfsuid(uid);
884 }
885 
886 /*
887  * Samma på svenska..
888  */
889 long __sys_setfsgid(gid_t gid)
890 {
891 	const struct cred *old;
892 	struct cred *new;
893 	gid_t old_fsgid;
894 	kgid_t kgid;
895 
896 	old = current_cred();
897 	old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
898 
899 	kgid = make_kgid(old->user_ns, gid);
900 	if (!gid_valid(kgid))
901 		return old_fsgid;
902 
903 	new = prepare_creds();
904 	if (!new)
905 		return old_fsgid;
906 
907 	if (gid_eq(kgid, old->gid)  || gid_eq(kgid, old->egid)  ||
908 	    gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
909 	    ns_capable_setid(old->user_ns, CAP_SETGID)) {
910 		if (!gid_eq(kgid, old->fsgid)) {
911 			new->fsgid = kgid;
912 			if (security_task_fix_setgid(new,old,LSM_SETID_FS) == 0)
913 				goto change_okay;
914 		}
915 	}
916 
917 	abort_creds(new);
918 	return old_fsgid;
919 
920 change_okay:
921 	commit_creds(new);
922 	return old_fsgid;
923 }
924 
925 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
926 {
927 	return __sys_setfsgid(gid);
928 }
929 #endif /* CONFIG_MULTIUSER */
930 
931 /**
932  * sys_getpid - return the thread group id of the current process
933  *
934  * Note, despite the name, this returns the tgid not the pid.  The tgid and
935  * the pid are identical unless CLONE_THREAD was specified on clone() in
936  * which case the tgid is the same in all threads of the same group.
937  *
938  * This is SMP safe as current->tgid does not change.
939  */
940 SYSCALL_DEFINE0(getpid)
941 {
942 	return task_tgid_vnr(current);
943 }
944 
945 /* Thread ID - the internal kernel "pid" */
946 SYSCALL_DEFINE0(gettid)
947 {
948 	return task_pid_vnr(current);
949 }
950 
951 /*
952  * Accessing ->real_parent is not SMP-safe, it could
953  * change from under us. However, we can use a stale
954  * value of ->real_parent under rcu_read_lock(), see
955  * release_task()->call_rcu(delayed_put_task_struct).
956  */
957 SYSCALL_DEFINE0(getppid)
958 {
959 	int pid;
960 
961 	rcu_read_lock();
962 	pid = task_tgid_vnr(rcu_dereference(current->real_parent));
963 	rcu_read_unlock();
964 
965 	return pid;
966 }
967 
968 SYSCALL_DEFINE0(getuid)
969 {
970 	/* Only we change this so SMP safe */
971 	return from_kuid_munged(current_user_ns(), current_uid());
972 }
973 
974 SYSCALL_DEFINE0(geteuid)
975 {
976 	/* Only we change this so SMP safe */
977 	return from_kuid_munged(current_user_ns(), current_euid());
978 }
979 
980 SYSCALL_DEFINE0(getgid)
981 {
982 	/* Only we change this so SMP safe */
983 	return from_kgid_munged(current_user_ns(), current_gid());
984 }
985 
986 SYSCALL_DEFINE0(getegid)
987 {
988 	/* Only we change this so SMP safe */
989 	return from_kgid_munged(current_user_ns(), current_egid());
990 }
991 
992 static void do_sys_times(struct tms *tms)
993 {
994 	u64 tgutime, tgstime, cutime, cstime;
995 
996 	thread_group_cputime_adjusted(current, &tgutime, &tgstime);
997 	cutime = current->signal->cutime;
998 	cstime = current->signal->cstime;
999 	tms->tms_utime = nsec_to_clock_t(tgutime);
1000 	tms->tms_stime = nsec_to_clock_t(tgstime);
1001 	tms->tms_cutime = nsec_to_clock_t(cutime);
1002 	tms->tms_cstime = nsec_to_clock_t(cstime);
1003 }
1004 
1005 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
1006 {
1007 	if (tbuf) {
1008 		struct tms tmp;
1009 
1010 		do_sys_times(&tmp);
1011 		if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1012 			return -EFAULT;
1013 	}
1014 	force_successful_syscall_return();
1015 	return (long) jiffies_64_to_clock_t(get_jiffies_64());
1016 }
1017 
1018 #ifdef CONFIG_COMPAT
1019 static compat_clock_t clock_t_to_compat_clock_t(clock_t x)
1020 {
1021 	return compat_jiffies_to_clock_t(clock_t_to_jiffies(x));
1022 }
1023 
1024 COMPAT_SYSCALL_DEFINE1(times, struct compat_tms __user *, tbuf)
1025 {
1026 	if (tbuf) {
1027 		struct tms tms;
1028 		struct compat_tms tmp;
1029 
1030 		do_sys_times(&tms);
1031 		/* Convert our struct tms to the compat version. */
1032 		tmp.tms_utime = clock_t_to_compat_clock_t(tms.tms_utime);
1033 		tmp.tms_stime = clock_t_to_compat_clock_t(tms.tms_stime);
1034 		tmp.tms_cutime = clock_t_to_compat_clock_t(tms.tms_cutime);
1035 		tmp.tms_cstime = clock_t_to_compat_clock_t(tms.tms_cstime);
1036 		if (copy_to_user(tbuf, &tmp, sizeof(tmp)))
1037 			return -EFAULT;
1038 	}
1039 	force_successful_syscall_return();
1040 	return compat_jiffies_to_clock_t(jiffies);
1041 }
1042 #endif
1043 
1044 /*
1045  * This needs some heavy checking ...
1046  * I just haven't the stomach for it. I also don't fully
1047  * understand sessions/pgrp etc. Let somebody who does explain it.
1048  *
1049  * OK, I think I have the protection semantics right.... this is really
1050  * only important on a multi-user system anyway, to make sure one user
1051  * can't send a signal to a process owned by another.  -TYT, 12/12/91
1052  *
1053  * !PF_FORKNOEXEC check to conform completely to POSIX.
1054  */
1055 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
1056 {
1057 	struct task_struct *p;
1058 	struct task_struct *group_leader = current->group_leader;
1059 	struct pid *pgrp;
1060 	int err;
1061 
1062 	if (!pid)
1063 		pid = task_pid_vnr(group_leader);
1064 	if (!pgid)
1065 		pgid = pid;
1066 	if (pgid < 0)
1067 		return -EINVAL;
1068 	rcu_read_lock();
1069 
1070 	/* From this point forward we keep holding onto the tasklist lock
1071 	 * so that our parent does not change from under us. -DaveM
1072 	 */
1073 	write_lock_irq(&tasklist_lock);
1074 
1075 	err = -ESRCH;
1076 	p = find_task_by_vpid(pid);
1077 	if (!p)
1078 		goto out;
1079 
1080 	err = -EINVAL;
1081 	if (!thread_group_leader(p))
1082 		goto out;
1083 
1084 	if (same_thread_group(p->real_parent, group_leader)) {
1085 		err = -EPERM;
1086 		if (task_session(p) != task_session(group_leader))
1087 			goto out;
1088 		err = -EACCES;
1089 		if (!(p->flags & PF_FORKNOEXEC))
1090 			goto out;
1091 	} else {
1092 		err = -ESRCH;
1093 		if (p != group_leader)
1094 			goto out;
1095 	}
1096 
1097 	err = -EPERM;
1098 	if (p->signal->leader)
1099 		goto out;
1100 
1101 	pgrp = task_pid(p);
1102 	if (pgid != pid) {
1103 		struct task_struct *g;
1104 
1105 		pgrp = find_vpid(pgid);
1106 		g = pid_task(pgrp, PIDTYPE_PGID);
1107 		if (!g || task_session(g) != task_session(group_leader))
1108 			goto out;
1109 	}
1110 
1111 	err = security_task_setpgid(p, pgid);
1112 	if (err)
1113 		goto out;
1114 
1115 	if (task_pgrp(p) != pgrp)
1116 		change_pid(p, PIDTYPE_PGID, pgrp);
1117 
1118 	err = 0;
1119 out:
1120 	/* All paths lead to here, thus we are safe. -DaveM */
1121 	write_unlock_irq(&tasklist_lock);
1122 	rcu_read_unlock();
1123 	return err;
1124 }
1125 
1126 static int do_getpgid(pid_t pid)
1127 {
1128 	struct task_struct *p;
1129 	struct pid *grp;
1130 	int retval;
1131 
1132 	rcu_read_lock();
1133 	if (!pid)
1134 		grp = task_pgrp(current);
1135 	else {
1136 		retval = -ESRCH;
1137 		p = find_task_by_vpid(pid);
1138 		if (!p)
1139 			goto out;
1140 		grp = task_pgrp(p);
1141 		if (!grp)
1142 			goto out;
1143 
1144 		retval = security_task_getpgid(p);
1145 		if (retval)
1146 			goto out;
1147 	}
1148 	retval = pid_vnr(grp);
1149 out:
1150 	rcu_read_unlock();
1151 	return retval;
1152 }
1153 
1154 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1155 {
1156 	return do_getpgid(pid);
1157 }
1158 
1159 #ifdef __ARCH_WANT_SYS_GETPGRP
1160 
1161 SYSCALL_DEFINE0(getpgrp)
1162 {
1163 	return do_getpgid(0);
1164 }
1165 
1166 #endif
1167 
1168 SYSCALL_DEFINE1(getsid, pid_t, pid)
1169 {
1170 	struct task_struct *p;
1171 	struct pid *sid;
1172 	int retval;
1173 
1174 	rcu_read_lock();
1175 	if (!pid)
1176 		sid = task_session(current);
1177 	else {
1178 		retval = -ESRCH;
1179 		p = find_task_by_vpid(pid);
1180 		if (!p)
1181 			goto out;
1182 		sid = task_session(p);
1183 		if (!sid)
1184 			goto out;
1185 
1186 		retval = security_task_getsid(p);
1187 		if (retval)
1188 			goto out;
1189 	}
1190 	retval = pid_vnr(sid);
1191 out:
1192 	rcu_read_unlock();
1193 	return retval;
1194 }
1195 
1196 static void set_special_pids(struct pid *pid)
1197 {
1198 	struct task_struct *curr = current->group_leader;
1199 
1200 	if (task_session(curr) != pid)
1201 		change_pid(curr, PIDTYPE_SID, pid);
1202 
1203 	if (task_pgrp(curr) != pid)
1204 		change_pid(curr, PIDTYPE_PGID, pid);
1205 }
1206 
1207 int ksys_setsid(void)
1208 {
1209 	struct task_struct *group_leader = current->group_leader;
1210 	struct pid *sid = task_pid(group_leader);
1211 	pid_t session = pid_vnr(sid);
1212 	int err = -EPERM;
1213 
1214 	write_lock_irq(&tasklist_lock);
1215 	/* Fail if I am already a session leader */
1216 	if (group_leader->signal->leader)
1217 		goto out;
1218 
1219 	/* Fail if a process group id already exists that equals the
1220 	 * proposed session id.
1221 	 */
1222 	if (pid_task(sid, PIDTYPE_PGID))
1223 		goto out;
1224 
1225 	group_leader->signal->leader = 1;
1226 	set_special_pids(sid);
1227 
1228 	proc_clear_tty(group_leader);
1229 
1230 	err = session;
1231 out:
1232 	write_unlock_irq(&tasklist_lock);
1233 	if (err > 0) {
1234 		proc_sid_connector(group_leader);
1235 		sched_autogroup_create_attach(group_leader);
1236 	}
1237 	return err;
1238 }
1239 
1240 SYSCALL_DEFINE0(setsid)
1241 {
1242 	return ksys_setsid();
1243 }
1244 
1245 DECLARE_RWSEM(uts_sem);
1246 
1247 #ifdef COMPAT_UTS_MACHINE
1248 #define override_architecture(name) \
1249 	(personality(current->personality) == PER_LINUX32 && \
1250 	 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1251 		      sizeof(COMPAT_UTS_MACHINE)))
1252 #else
1253 #define override_architecture(name)	0
1254 #endif
1255 
1256 /*
1257  * Work around broken programs that cannot handle "Linux 3.0".
1258  * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1259  * And we map 4.x and later versions to 2.6.60+x, so 4.0/5.0/6.0/... would be
1260  * 2.6.60.
1261  */
1262 static int override_release(char __user *release, size_t len)
1263 {
1264 	int ret = 0;
1265 
1266 	if (current->personality & UNAME26) {
1267 		const char *rest = UTS_RELEASE;
1268 		char buf[65] = { 0 };
1269 		int ndots = 0;
1270 		unsigned v;
1271 		size_t copy;
1272 
1273 		while (*rest) {
1274 			if (*rest == '.' && ++ndots >= 3)
1275 				break;
1276 			if (!isdigit(*rest) && *rest != '.')
1277 				break;
1278 			rest++;
1279 		}
1280 		v = LINUX_VERSION_PATCHLEVEL + 60;
1281 		copy = clamp_t(size_t, len, 1, sizeof(buf));
1282 		copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1283 		ret = copy_to_user(release, buf, copy + 1);
1284 	}
1285 	return ret;
1286 }
1287 
1288 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1289 {
1290 	struct new_utsname tmp;
1291 
1292 	down_read(&uts_sem);
1293 	memcpy(&tmp, utsname(), sizeof(tmp));
1294 	up_read(&uts_sem);
1295 	if (copy_to_user(name, &tmp, sizeof(tmp)))
1296 		return -EFAULT;
1297 
1298 	if (override_release(name->release, sizeof(name->release)))
1299 		return -EFAULT;
1300 	if (override_architecture(name))
1301 		return -EFAULT;
1302 	return 0;
1303 }
1304 
1305 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1306 /*
1307  * Old cruft
1308  */
1309 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1310 {
1311 	struct old_utsname tmp;
1312 
1313 	if (!name)
1314 		return -EFAULT;
1315 
1316 	down_read(&uts_sem);
1317 	memcpy(&tmp, utsname(), sizeof(tmp));
1318 	up_read(&uts_sem);
1319 	if (copy_to_user(name, &tmp, sizeof(tmp)))
1320 		return -EFAULT;
1321 
1322 	if (override_release(name->release, sizeof(name->release)))
1323 		return -EFAULT;
1324 	if (override_architecture(name))
1325 		return -EFAULT;
1326 	return 0;
1327 }
1328 
1329 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1330 {
1331 	struct oldold_utsname tmp;
1332 
1333 	if (!name)
1334 		return -EFAULT;
1335 
1336 	memset(&tmp, 0, sizeof(tmp));
1337 
1338 	down_read(&uts_sem);
1339 	memcpy(&tmp.sysname, &utsname()->sysname, __OLD_UTS_LEN);
1340 	memcpy(&tmp.nodename, &utsname()->nodename, __OLD_UTS_LEN);
1341 	memcpy(&tmp.release, &utsname()->release, __OLD_UTS_LEN);
1342 	memcpy(&tmp.version, &utsname()->version, __OLD_UTS_LEN);
1343 	memcpy(&tmp.machine, &utsname()->machine, __OLD_UTS_LEN);
1344 	up_read(&uts_sem);
1345 	if (copy_to_user(name, &tmp, sizeof(tmp)))
1346 		return -EFAULT;
1347 
1348 	if (override_architecture(name))
1349 		return -EFAULT;
1350 	if (override_release(name->release, sizeof(name->release)))
1351 		return -EFAULT;
1352 	return 0;
1353 }
1354 #endif
1355 
1356 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1357 {
1358 	int errno;
1359 	char tmp[__NEW_UTS_LEN];
1360 
1361 	if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1362 		return -EPERM;
1363 
1364 	if (len < 0 || len > __NEW_UTS_LEN)
1365 		return -EINVAL;
1366 	errno = -EFAULT;
1367 	if (!copy_from_user(tmp, name, len)) {
1368 		struct new_utsname *u;
1369 
1370 		add_device_randomness(tmp, len);
1371 		down_write(&uts_sem);
1372 		u = utsname();
1373 		memcpy(u->nodename, tmp, len);
1374 		memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1375 		errno = 0;
1376 		uts_proc_notify(UTS_PROC_HOSTNAME);
1377 		up_write(&uts_sem);
1378 	}
1379 	return errno;
1380 }
1381 
1382 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1383 
1384 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1385 {
1386 	int i;
1387 	struct new_utsname *u;
1388 	char tmp[__NEW_UTS_LEN + 1];
1389 
1390 	if (len < 0)
1391 		return -EINVAL;
1392 	down_read(&uts_sem);
1393 	u = utsname();
1394 	i = 1 + strlen(u->nodename);
1395 	if (i > len)
1396 		i = len;
1397 	memcpy(tmp, u->nodename, i);
1398 	up_read(&uts_sem);
1399 	if (copy_to_user(name, tmp, i))
1400 		return -EFAULT;
1401 	return 0;
1402 }
1403 
1404 #endif
1405 
1406 /*
1407  * Only setdomainname; getdomainname can be implemented by calling
1408  * uname()
1409  */
1410 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1411 {
1412 	int errno;
1413 	char tmp[__NEW_UTS_LEN];
1414 
1415 	if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1416 		return -EPERM;
1417 	if (len < 0 || len > __NEW_UTS_LEN)
1418 		return -EINVAL;
1419 
1420 	errno = -EFAULT;
1421 	if (!copy_from_user(tmp, name, len)) {
1422 		struct new_utsname *u;
1423 
1424 		add_device_randomness(tmp, len);
1425 		down_write(&uts_sem);
1426 		u = utsname();
1427 		memcpy(u->domainname, tmp, len);
1428 		memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1429 		errno = 0;
1430 		uts_proc_notify(UTS_PROC_DOMAINNAME);
1431 		up_write(&uts_sem);
1432 	}
1433 	return errno;
1434 }
1435 
1436 /* make sure you are allowed to change @tsk limits before calling this */
1437 static int do_prlimit(struct task_struct *tsk, unsigned int resource,
1438 		      struct rlimit *new_rlim, struct rlimit *old_rlim)
1439 {
1440 	struct rlimit *rlim;
1441 	int retval = 0;
1442 
1443 	if (resource >= RLIM_NLIMITS)
1444 		return -EINVAL;
1445 	resource = array_index_nospec(resource, RLIM_NLIMITS);
1446 
1447 	if (new_rlim) {
1448 		if (new_rlim->rlim_cur > new_rlim->rlim_max)
1449 			return -EINVAL;
1450 		if (resource == RLIMIT_NOFILE &&
1451 				new_rlim->rlim_max > sysctl_nr_open)
1452 			return -EPERM;
1453 	}
1454 
1455 	/* Holding a refcount on tsk protects tsk->signal from disappearing. */
1456 	rlim = tsk->signal->rlim + resource;
1457 	task_lock(tsk->group_leader);
1458 	if (new_rlim) {
1459 		/*
1460 		 * Keep the capable check against init_user_ns until cgroups can
1461 		 * contain all limits.
1462 		 */
1463 		if (new_rlim->rlim_max > rlim->rlim_max &&
1464 				!capable(CAP_SYS_RESOURCE))
1465 			retval = -EPERM;
1466 		if (!retval)
1467 			retval = security_task_setrlimit(tsk, resource, new_rlim);
1468 	}
1469 	if (!retval) {
1470 		if (old_rlim)
1471 			*old_rlim = *rlim;
1472 		if (new_rlim)
1473 			*rlim = *new_rlim;
1474 	}
1475 	task_unlock(tsk->group_leader);
1476 
1477 	/*
1478 	 * RLIMIT_CPU handling. Arm the posix CPU timer if the limit is not
1479 	 * infinite. In case of RLIM_INFINITY the posix CPU timer code
1480 	 * ignores the rlimit.
1481 	 */
1482 	if (!retval && new_rlim && resource == RLIMIT_CPU &&
1483 	    new_rlim->rlim_cur != RLIM_INFINITY &&
1484 	    IS_ENABLED(CONFIG_POSIX_TIMERS)) {
1485 		/*
1486 		 * update_rlimit_cpu can fail if the task is exiting, but there
1487 		 * may be other tasks in the thread group that are not exiting,
1488 		 * and they need their cpu timers adjusted.
1489 		 *
1490 		 * The group_leader is the last task to be released, so if we
1491 		 * cannot update_rlimit_cpu on it, then the entire process is
1492 		 * exiting and we do not need to update at all.
1493 		 */
1494 		update_rlimit_cpu(tsk->group_leader, new_rlim->rlim_cur);
1495 	}
1496 
1497 	return retval;
1498 }
1499 
1500 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1501 {
1502 	struct rlimit value;
1503 	int ret;
1504 
1505 	ret = do_prlimit(current, resource, NULL, &value);
1506 	if (!ret)
1507 		ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1508 
1509 	return ret;
1510 }
1511 
1512 #ifdef CONFIG_COMPAT
1513 
1514 COMPAT_SYSCALL_DEFINE2(setrlimit, unsigned int, resource,
1515 		       struct compat_rlimit __user *, rlim)
1516 {
1517 	struct rlimit r;
1518 	struct compat_rlimit r32;
1519 
1520 	if (copy_from_user(&r32, rlim, sizeof(struct compat_rlimit)))
1521 		return -EFAULT;
1522 
1523 	if (r32.rlim_cur == COMPAT_RLIM_INFINITY)
1524 		r.rlim_cur = RLIM_INFINITY;
1525 	else
1526 		r.rlim_cur = r32.rlim_cur;
1527 	if (r32.rlim_max == COMPAT_RLIM_INFINITY)
1528 		r.rlim_max = RLIM_INFINITY;
1529 	else
1530 		r.rlim_max = r32.rlim_max;
1531 	return do_prlimit(current, resource, &r, NULL);
1532 }
1533 
1534 COMPAT_SYSCALL_DEFINE2(getrlimit, unsigned int, resource,
1535 		       struct compat_rlimit __user *, rlim)
1536 {
1537 	struct rlimit r;
1538 	int ret;
1539 
1540 	ret = do_prlimit(current, resource, NULL, &r);
1541 	if (!ret) {
1542 		struct compat_rlimit r32;
1543 		if (r.rlim_cur > COMPAT_RLIM_INFINITY)
1544 			r32.rlim_cur = COMPAT_RLIM_INFINITY;
1545 		else
1546 			r32.rlim_cur = r.rlim_cur;
1547 		if (r.rlim_max > COMPAT_RLIM_INFINITY)
1548 			r32.rlim_max = COMPAT_RLIM_INFINITY;
1549 		else
1550 			r32.rlim_max = r.rlim_max;
1551 
1552 		if (copy_to_user(rlim, &r32, sizeof(struct compat_rlimit)))
1553 			return -EFAULT;
1554 	}
1555 	return ret;
1556 }
1557 
1558 #endif
1559 
1560 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1561 
1562 /*
1563  *	Back compatibility for getrlimit. Needed for some apps.
1564  */
1565 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1566 		struct rlimit __user *, rlim)
1567 {
1568 	struct rlimit x;
1569 	if (resource >= RLIM_NLIMITS)
1570 		return -EINVAL;
1571 
1572 	resource = array_index_nospec(resource, RLIM_NLIMITS);
1573 	task_lock(current->group_leader);
1574 	x = current->signal->rlim[resource];
1575 	task_unlock(current->group_leader);
1576 	if (x.rlim_cur > 0x7FFFFFFF)
1577 		x.rlim_cur = 0x7FFFFFFF;
1578 	if (x.rlim_max > 0x7FFFFFFF)
1579 		x.rlim_max = 0x7FFFFFFF;
1580 	return copy_to_user(rlim, &x, sizeof(x)) ? -EFAULT : 0;
1581 }
1582 
1583 #ifdef CONFIG_COMPAT
1584 COMPAT_SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1585 		       struct compat_rlimit __user *, rlim)
1586 {
1587 	struct rlimit r;
1588 
1589 	if (resource >= RLIM_NLIMITS)
1590 		return -EINVAL;
1591 
1592 	resource = array_index_nospec(resource, RLIM_NLIMITS);
1593 	task_lock(current->group_leader);
1594 	r = current->signal->rlim[resource];
1595 	task_unlock(current->group_leader);
1596 	if (r.rlim_cur > 0x7FFFFFFF)
1597 		r.rlim_cur = 0x7FFFFFFF;
1598 	if (r.rlim_max > 0x7FFFFFFF)
1599 		r.rlim_max = 0x7FFFFFFF;
1600 
1601 	if (put_user(r.rlim_cur, &rlim->rlim_cur) ||
1602 	    put_user(r.rlim_max, &rlim->rlim_max))
1603 		return -EFAULT;
1604 	return 0;
1605 }
1606 #endif
1607 
1608 #endif
1609 
1610 static inline bool rlim64_is_infinity(__u64 rlim64)
1611 {
1612 #if BITS_PER_LONG < 64
1613 	return rlim64 >= ULONG_MAX;
1614 #else
1615 	return rlim64 == RLIM64_INFINITY;
1616 #endif
1617 }
1618 
1619 static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1620 {
1621 	if (rlim->rlim_cur == RLIM_INFINITY)
1622 		rlim64->rlim_cur = RLIM64_INFINITY;
1623 	else
1624 		rlim64->rlim_cur = rlim->rlim_cur;
1625 	if (rlim->rlim_max == RLIM_INFINITY)
1626 		rlim64->rlim_max = RLIM64_INFINITY;
1627 	else
1628 		rlim64->rlim_max = rlim->rlim_max;
1629 }
1630 
1631 static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1632 {
1633 	if (rlim64_is_infinity(rlim64->rlim_cur))
1634 		rlim->rlim_cur = RLIM_INFINITY;
1635 	else
1636 		rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1637 	if (rlim64_is_infinity(rlim64->rlim_max))
1638 		rlim->rlim_max = RLIM_INFINITY;
1639 	else
1640 		rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1641 }
1642 
1643 /* rcu lock must be held */
1644 static int check_prlimit_permission(struct task_struct *task,
1645 				    unsigned int flags)
1646 {
1647 	const struct cred *cred = current_cred(), *tcred;
1648 	bool id_match;
1649 
1650 	if (current == task)
1651 		return 0;
1652 
1653 	tcred = __task_cred(task);
1654 	id_match = (uid_eq(cred->uid, tcred->euid) &&
1655 		    uid_eq(cred->uid, tcred->suid) &&
1656 		    uid_eq(cred->uid, tcred->uid)  &&
1657 		    gid_eq(cred->gid, tcred->egid) &&
1658 		    gid_eq(cred->gid, tcred->sgid) &&
1659 		    gid_eq(cred->gid, tcred->gid));
1660 	if (!id_match && !ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
1661 		return -EPERM;
1662 
1663 	return security_task_prlimit(cred, tcred, flags);
1664 }
1665 
1666 SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1667 		const struct rlimit64 __user *, new_rlim,
1668 		struct rlimit64 __user *, old_rlim)
1669 {
1670 	struct rlimit64 old64, new64;
1671 	struct rlimit old, new;
1672 	struct task_struct *tsk;
1673 	unsigned int checkflags = 0;
1674 	int ret;
1675 
1676 	if (old_rlim)
1677 		checkflags |= LSM_PRLIMIT_READ;
1678 
1679 	if (new_rlim) {
1680 		if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1681 			return -EFAULT;
1682 		rlim64_to_rlim(&new64, &new);
1683 		checkflags |= LSM_PRLIMIT_WRITE;
1684 	}
1685 
1686 	rcu_read_lock();
1687 	tsk = pid ? find_task_by_vpid(pid) : current;
1688 	if (!tsk) {
1689 		rcu_read_unlock();
1690 		return -ESRCH;
1691 	}
1692 	ret = check_prlimit_permission(tsk, checkflags);
1693 	if (ret) {
1694 		rcu_read_unlock();
1695 		return ret;
1696 	}
1697 	get_task_struct(tsk);
1698 	rcu_read_unlock();
1699 
1700 	ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1701 			old_rlim ? &old : NULL);
1702 
1703 	if (!ret && old_rlim) {
1704 		rlim_to_rlim64(&old, &old64);
1705 		if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1706 			ret = -EFAULT;
1707 	}
1708 
1709 	put_task_struct(tsk);
1710 	return ret;
1711 }
1712 
1713 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1714 {
1715 	struct rlimit new_rlim;
1716 
1717 	if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1718 		return -EFAULT;
1719 	return do_prlimit(current, resource, &new_rlim, NULL);
1720 }
1721 
1722 /*
1723  * It would make sense to put struct rusage in the task_struct,
1724  * except that would make the task_struct be *really big*.  After
1725  * task_struct gets moved into malloc'ed memory, it would
1726  * make sense to do this.  It will make moving the rest of the information
1727  * a lot simpler!  (Which we're not doing right now because we're not
1728  * measuring them yet).
1729  *
1730  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1731  * races with threads incrementing their own counters.  But since word
1732  * reads are atomic, we either get new values or old values and we don't
1733  * care which for the sums.  We always take the siglock to protect reading
1734  * the c* fields from p->signal from races with exit.c updating those
1735  * fields when reaping, so a sample either gets all the additions of a
1736  * given child after it's reaped, or none so this sample is before reaping.
1737  *
1738  * Locking:
1739  * We need to take the siglock for CHILDEREN, SELF and BOTH
1740  * for  the cases current multithreaded, non-current single threaded
1741  * non-current multithreaded.  Thread traversal is now safe with
1742  * the siglock held.
1743  * Strictly speaking, we donot need to take the siglock if we are current and
1744  * single threaded,  as no one else can take our signal_struct away, no one
1745  * else can  reap the  children to update signal->c* counters, and no one else
1746  * can race with the signal-> fields. If we do not take any lock, the
1747  * signal-> fields could be read out of order while another thread was just
1748  * exiting. So we should  place a read memory barrier when we avoid the lock.
1749  * On the writer side,  write memory barrier is implied in  __exit_signal
1750  * as __exit_signal releases  the siglock spinlock after updating the signal->
1751  * fields. But we don't do this yet to keep things simple.
1752  *
1753  */
1754 
1755 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1756 {
1757 	r->ru_nvcsw += t->nvcsw;
1758 	r->ru_nivcsw += t->nivcsw;
1759 	r->ru_minflt += t->min_flt;
1760 	r->ru_majflt += t->maj_flt;
1761 	r->ru_inblock += task_io_get_inblock(t);
1762 	r->ru_oublock += task_io_get_oublock(t);
1763 }
1764 
1765 void getrusage(struct task_struct *p, int who, struct rusage *r)
1766 {
1767 	struct task_struct *t;
1768 	unsigned long flags;
1769 	u64 tgutime, tgstime, utime, stime;
1770 	unsigned long maxrss = 0;
1771 
1772 	memset((char *)r, 0, sizeof (*r));
1773 	utime = stime = 0;
1774 
1775 	if (who == RUSAGE_THREAD) {
1776 		task_cputime_adjusted(current, &utime, &stime);
1777 		accumulate_thread_rusage(p, r);
1778 		maxrss = p->signal->maxrss;
1779 		goto out;
1780 	}
1781 
1782 	if (!lock_task_sighand(p, &flags))
1783 		return;
1784 
1785 	switch (who) {
1786 	case RUSAGE_BOTH:
1787 	case RUSAGE_CHILDREN:
1788 		utime = p->signal->cutime;
1789 		stime = p->signal->cstime;
1790 		r->ru_nvcsw = p->signal->cnvcsw;
1791 		r->ru_nivcsw = p->signal->cnivcsw;
1792 		r->ru_minflt = p->signal->cmin_flt;
1793 		r->ru_majflt = p->signal->cmaj_flt;
1794 		r->ru_inblock = p->signal->cinblock;
1795 		r->ru_oublock = p->signal->coublock;
1796 		maxrss = p->signal->cmaxrss;
1797 
1798 		if (who == RUSAGE_CHILDREN)
1799 			break;
1800 		fallthrough;
1801 
1802 	case RUSAGE_SELF:
1803 		thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1804 		utime += tgutime;
1805 		stime += tgstime;
1806 		r->ru_nvcsw += p->signal->nvcsw;
1807 		r->ru_nivcsw += p->signal->nivcsw;
1808 		r->ru_minflt += p->signal->min_flt;
1809 		r->ru_majflt += p->signal->maj_flt;
1810 		r->ru_inblock += p->signal->inblock;
1811 		r->ru_oublock += p->signal->oublock;
1812 		if (maxrss < p->signal->maxrss)
1813 			maxrss = p->signal->maxrss;
1814 		t = p;
1815 		do {
1816 			accumulate_thread_rusage(t, r);
1817 		} while_each_thread(p, t);
1818 		break;
1819 
1820 	default:
1821 		BUG();
1822 	}
1823 	unlock_task_sighand(p, &flags);
1824 
1825 out:
1826 	r->ru_utime = ns_to_kernel_old_timeval(utime);
1827 	r->ru_stime = ns_to_kernel_old_timeval(stime);
1828 
1829 	if (who != RUSAGE_CHILDREN) {
1830 		struct mm_struct *mm = get_task_mm(p);
1831 
1832 		if (mm) {
1833 			setmax_mm_hiwater_rss(&maxrss, mm);
1834 			mmput(mm);
1835 		}
1836 	}
1837 	r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1838 }
1839 
1840 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1841 {
1842 	struct rusage r;
1843 
1844 	if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1845 	    who != RUSAGE_THREAD)
1846 		return -EINVAL;
1847 
1848 	getrusage(current, who, &r);
1849 	return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1850 }
1851 
1852 #ifdef CONFIG_COMPAT
1853 COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru)
1854 {
1855 	struct rusage r;
1856 
1857 	if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1858 	    who != RUSAGE_THREAD)
1859 		return -EINVAL;
1860 
1861 	getrusage(current, who, &r);
1862 	return put_compat_rusage(&r, ru);
1863 }
1864 #endif
1865 
1866 SYSCALL_DEFINE1(umask, int, mask)
1867 {
1868 	mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1869 	return mask;
1870 }
1871 
1872 static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1873 {
1874 	struct fd exe;
1875 	struct inode *inode;
1876 	int err;
1877 
1878 	exe = fdget(fd);
1879 	if (!exe.file)
1880 		return -EBADF;
1881 
1882 	inode = file_inode(exe.file);
1883 
1884 	/*
1885 	 * Because the original mm->exe_file points to executable file, make
1886 	 * sure that this one is executable as well, to avoid breaking an
1887 	 * overall picture.
1888 	 */
1889 	err = -EACCES;
1890 	if (!S_ISREG(inode->i_mode) || path_noexec(&exe.file->f_path))
1891 		goto exit;
1892 
1893 	err = file_permission(exe.file, MAY_EXEC);
1894 	if (err)
1895 		goto exit;
1896 
1897 	err = replace_mm_exe_file(mm, exe.file);
1898 exit:
1899 	fdput(exe);
1900 	return err;
1901 }
1902 
1903 /*
1904  * Check arithmetic relations of passed addresses.
1905  *
1906  * WARNING: we don't require any capability here so be very careful
1907  * in what is allowed for modification from userspace.
1908  */
1909 static int validate_prctl_map_addr(struct prctl_mm_map *prctl_map)
1910 {
1911 	unsigned long mmap_max_addr = TASK_SIZE;
1912 	int error = -EINVAL, i;
1913 
1914 	static const unsigned char offsets[] = {
1915 		offsetof(struct prctl_mm_map, start_code),
1916 		offsetof(struct prctl_mm_map, end_code),
1917 		offsetof(struct prctl_mm_map, start_data),
1918 		offsetof(struct prctl_mm_map, end_data),
1919 		offsetof(struct prctl_mm_map, start_brk),
1920 		offsetof(struct prctl_mm_map, brk),
1921 		offsetof(struct prctl_mm_map, start_stack),
1922 		offsetof(struct prctl_mm_map, arg_start),
1923 		offsetof(struct prctl_mm_map, arg_end),
1924 		offsetof(struct prctl_mm_map, env_start),
1925 		offsetof(struct prctl_mm_map, env_end),
1926 	};
1927 
1928 	/*
1929 	 * Make sure the members are not somewhere outside
1930 	 * of allowed address space.
1931 	 */
1932 	for (i = 0; i < ARRAY_SIZE(offsets); i++) {
1933 		u64 val = *(u64 *)((char *)prctl_map + offsets[i]);
1934 
1935 		if ((unsigned long)val >= mmap_max_addr ||
1936 		    (unsigned long)val < mmap_min_addr)
1937 			goto out;
1938 	}
1939 
1940 	/*
1941 	 * Make sure the pairs are ordered.
1942 	 */
1943 #define __prctl_check_order(__m1, __op, __m2)				\
1944 	((unsigned long)prctl_map->__m1 __op				\
1945 	 (unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1946 	error  = __prctl_check_order(start_code, <, end_code);
1947 	error |= __prctl_check_order(start_data,<=, end_data);
1948 	error |= __prctl_check_order(start_brk, <=, brk);
1949 	error |= __prctl_check_order(arg_start, <=, arg_end);
1950 	error |= __prctl_check_order(env_start, <=, env_end);
1951 	if (error)
1952 		goto out;
1953 #undef __prctl_check_order
1954 
1955 	error = -EINVAL;
1956 
1957 	/*
1958 	 * Neither we should allow to override limits if they set.
1959 	 */
1960 	if (check_data_rlimit(rlimit(RLIMIT_DATA), prctl_map->brk,
1961 			      prctl_map->start_brk, prctl_map->end_data,
1962 			      prctl_map->start_data))
1963 			goto out;
1964 
1965 	error = 0;
1966 out:
1967 	return error;
1968 }
1969 
1970 #ifdef CONFIG_CHECKPOINT_RESTORE
1971 static int prctl_set_mm_map(int opt, const void __user *addr, unsigned long data_size)
1972 {
1973 	struct prctl_mm_map prctl_map = { .exe_fd = (u32)-1, };
1974 	unsigned long user_auxv[AT_VECTOR_SIZE];
1975 	struct mm_struct *mm = current->mm;
1976 	int error;
1977 
1978 	BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
1979 	BUILD_BUG_ON(sizeof(struct prctl_mm_map) > 256);
1980 
1981 	if (opt == PR_SET_MM_MAP_SIZE)
1982 		return put_user((unsigned int)sizeof(prctl_map),
1983 				(unsigned int __user *)addr);
1984 
1985 	if (data_size != sizeof(prctl_map))
1986 		return -EINVAL;
1987 
1988 	if (copy_from_user(&prctl_map, addr, sizeof(prctl_map)))
1989 		return -EFAULT;
1990 
1991 	error = validate_prctl_map_addr(&prctl_map);
1992 	if (error)
1993 		return error;
1994 
1995 	if (prctl_map.auxv_size) {
1996 		/*
1997 		 * Someone is trying to cheat the auxv vector.
1998 		 */
1999 		if (!prctl_map.auxv ||
2000 				prctl_map.auxv_size > sizeof(mm->saved_auxv))
2001 			return -EINVAL;
2002 
2003 		memset(user_auxv, 0, sizeof(user_auxv));
2004 		if (copy_from_user(user_auxv,
2005 				   (const void __user *)prctl_map.auxv,
2006 				   prctl_map.auxv_size))
2007 			return -EFAULT;
2008 
2009 		/* Last entry must be AT_NULL as specification requires */
2010 		user_auxv[AT_VECTOR_SIZE - 2] = AT_NULL;
2011 		user_auxv[AT_VECTOR_SIZE - 1] = AT_NULL;
2012 	}
2013 
2014 	if (prctl_map.exe_fd != (u32)-1) {
2015 		/*
2016 		 * Check if the current user is checkpoint/restore capable.
2017 		 * At the time of this writing, it checks for CAP_SYS_ADMIN
2018 		 * or CAP_CHECKPOINT_RESTORE.
2019 		 * Note that a user with access to ptrace can masquerade an
2020 		 * arbitrary program as any executable, even setuid ones.
2021 		 * This may have implications in the tomoyo subsystem.
2022 		 */
2023 		if (!checkpoint_restore_ns_capable(current_user_ns()))
2024 			return -EPERM;
2025 
2026 		error = prctl_set_mm_exe_file(mm, prctl_map.exe_fd);
2027 		if (error)
2028 			return error;
2029 	}
2030 
2031 	/*
2032 	 * arg_lock protects concurrent updates but we still need mmap_lock for
2033 	 * read to exclude races with sys_brk.
2034 	 */
2035 	mmap_read_lock(mm);
2036 
2037 	/*
2038 	 * We don't validate if these members are pointing to
2039 	 * real present VMAs because application may have correspond
2040 	 * VMAs already unmapped and kernel uses these members for statistics
2041 	 * output in procfs mostly, except
2042 	 *
2043 	 *  - @start_brk/@brk which are used in do_brk_flags but kernel lookups
2044 	 *    for VMAs when updating these members so anything wrong written
2045 	 *    here cause kernel to swear at userspace program but won't lead
2046 	 *    to any problem in kernel itself
2047 	 */
2048 
2049 	spin_lock(&mm->arg_lock);
2050 	mm->start_code	= prctl_map.start_code;
2051 	mm->end_code	= prctl_map.end_code;
2052 	mm->start_data	= prctl_map.start_data;
2053 	mm->end_data	= prctl_map.end_data;
2054 	mm->start_brk	= prctl_map.start_brk;
2055 	mm->brk		= prctl_map.brk;
2056 	mm->start_stack	= prctl_map.start_stack;
2057 	mm->arg_start	= prctl_map.arg_start;
2058 	mm->arg_end	= prctl_map.arg_end;
2059 	mm->env_start	= prctl_map.env_start;
2060 	mm->env_end	= prctl_map.env_end;
2061 	spin_unlock(&mm->arg_lock);
2062 
2063 	/*
2064 	 * Note this update of @saved_auxv is lockless thus
2065 	 * if someone reads this member in procfs while we're
2066 	 * updating -- it may get partly updated results. It's
2067 	 * known and acceptable trade off: we leave it as is to
2068 	 * not introduce additional locks here making the kernel
2069 	 * more complex.
2070 	 */
2071 	if (prctl_map.auxv_size)
2072 		memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv));
2073 
2074 	mmap_read_unlock(mm);
2075 	return 0;
2076 }
2077 #endif /* CONFIG_CHECKPOINT_RESTORE */
2078 
2079 static int prctl_set_auxv(struct mm_struct *mm, unsigned long addr,
2080 			  unsigned long len)
2081 {
2082 	/*
2083 	 * This doesn't move the auxiliary vector itself since it's pinned to
2084 	 * mm_struct, but it permits filling the vector with new values.  It's
2085 	 * up to the caller to provide sane values here, otherwise userspace
2086 	 * tools which use this vector might be unhappy.
2087 	 */
2088 	unsigned long user_auxv[AT_VECTOR_SIZE] = {};
2089 
2090 	if (len > sizeof(user_auxv))
2091 		return -EINVAL;
2092 
2093 	if (copy_from_user(user_auxv, (const void __user *)addr, len))
2094 		return -EFAULT;
2095 
2096 	/* Make sure the last entry is always AT_NULL */
2097 	user_auxv[AT_VECTOR_SIZE - 2] = 0;
2098 	user_auxv[AT_VECTOR_SIZE - 1] = 0;
2099 
2100 	BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
2101 
2102 	task_lock(current);
2103 	memcpy(mm->saved_auxv, user_auxv, len);
2104 	task_unlock(current);
2105 
2106 	return 0;
2107 }
2108 
2109 static int prctl_set_mm(int opt, unsigned long addr,
2110 			unsigned long arg4, unsigned long arg5)
2111 {
2112 	struct mm_struct *mm = current->mm;
2113 	struct prctl_mm_map prctl_map = {
2114 		.auxv = NULL,
2115 		.auxv_size = 0,
2116 		.exe_fd = -1,
2117 	};
2118 	struct vm_area_struct *vma;
2119 	int error;
2120 
2121 	if (arg5 || (arg4 && (opt != PR_SET_MM_AUXV &&
2122 			      opt != PR_SET_MM_MAP &&
2123 			      opt != PR_SET_MM_MAP_SIZE)))
2124 		return -EINVAL;
2125 
2126 #ifdef CONFIG_CHECKPOINT_RESTORE
2127 	if (opt == PR_SET_MM_MAP || opt == PR_SET_MM_MAP_SIZE)
2128 		return prctl_set_mm_map(opt, (const void __user *)addr, arg4);
2129 #endif
2130 
2131 	if (!capable(CAP_SYS_RESOURCE))
2132 		return -EPERM;
2133 
2134 	if (opt == PR_SET_MM_EXE_FILE)
2135 		return prctl_set_mm_exe_file(mm, (unsigned int)addr);
2136 
2137 	if (opt == PR_SET_MM_AUXV)
2138 		return prctl_set_auxv(mm, addr, arg4);
2139 
2140 	if (addr >= TASK_SIZE || addr < mmap_min_addr)
2141 		return -EINVAL;
2142 
2143 	error = -EINVAL;
2144 
2145 	/*
2146 	 * arg_lock protects concurrent updates of arg boundaries, we need
2147 	 * mmap_lock for a) concurrent sys_brk, b) finding VMA for addr
2148 	 * validation.
2149 	 */
2150 	mmap_read_lock(mm);
2151 	vma = find_vma(mm, addr);
2152 
2153 	spin_lock(&mm->arg_lock);
2154 	prctl_map.start_code	= mm->start_code;
2155 	prctl_map.end_code	= mm->end_code;
2156 	prctl_map.start_data	= mm->start_data;
2157 	prctl_map.end_data	= mm->end_data;
2158 	prctl_map.start_brk	= mm->start_brk;
2159 	prctl_map.brk		= mm->brk;
2160 	prctl_map.start_stack	= mm->start_stack;
2161 	prctl_map.arg_start	= mm->arg_start;
2162 	prctl_map.arg_end	= mm->arg_end;
2163 	prctl_map.env_start	= mm->env_start;
2164 	prctl_map.env_end	= mm->env_end;
2165 
2166 	switch (opt) {
2167 	case PR_SET_MM_START_CODE:
2168 		prctl_map.start_code = addr;
2169 		break;
2170 	case PR_SET_MM_END_CODE:
2171 		prctl_map.end_code = addr;
2172 		break;
2173 	case PR_SET_MM_START_DATA:
2174 		prctl_map.start_data = addr;
2175 		break;
2176 	case PR_SET_MM_END_DATA:
2177 		prctl_map.end_data = addr;
2178 		break;
2179 	case PR_SET_MM_START_STACK:
2180 		prctl_map.start_stack = addr;
2181 		break;
2182 	case PR_SET_MM_START_BRK:
2183 		prctl_map.start_brk = addr;
2184 		break;
2185 	case PR_SET_MM_BRK:
2186 		prctl_map.brk = addr;
2187 		break;
2188 	case PR_SET_MM_ARG_START:
2189 		prctl_map.arg_start = addr;
2190 		break;
2191 	case PR_SET_MM_ARG_END:
2192 		prctl_map.arg_end = addr;
2193 		break;
2194 	case PR_SET_MM_ENV_START:
2195 		prctl_map.env_start = addr;
2196 		break;
2197 	case PR_SET_MM_ENV_END:
2198 		prctl_map.env_end = addr;
2199 		break;
2200 	default:
2201 		goto out;
2202 	}
2203 
2204 	error = validate_prctl_map_addr(&prctl_map);
2205 	if (error)
2206 		goto out;
2207 
2208 	switch (opt) {
2209 	/*
2210 	 * If command line arguments and environment
2211 	 * are placed somewhere else on stack, we can
2212 	 * set them up here, ARG_START/END to setup
2213 	 * command line arguments and ENV_START/END
2214 	 * for environment.
2215 	 */
2216 	case PR_SET_MM_START_STACK:
2217 	case PR_SET_MM_ARG_START:
2218 	case PR_SET_MM_ARG_END:
2219 	case PR_SET_MM_ENV_START:
2220 	case PR_SET_MM_ENV_END:
2221 		if (!vma) {
2222 			error = -EFAULT;
2223 			goto out;
2224 		}
2225 	}
2226 
2227 	mm->start_code	= prctl_map.start_code;
2228 	mm->end_code	= prctl_map.end_code;
2229 	mm->start_data	= prctl_map.start_data;
2230 	mm->end_data	= prctl_map.end_data;
2231 	mm->start_brk	= prctl_map.start_brk;
2232 	mm->brk		= prctl_map.brk;
2233 	mm->start_stack	= prctl_map.start_stack;
2234 	mm->arg_start	= prctl_map.arg_start;
2235 	mm->arg_end	= prctl_map.arg_end;
2236 	mm->env_start	= prctl_map.env_start;
2237 	mm->env_end	= prctl_map.env_end;
2238 
2239 	error = 0;
2240 out:
2241 	spin_unlock(&mm->arg_lock);
2242 	mmap_read_unlock(mm);
2243 	return error;
2244 }
2245 
2246 #ifdef CONFIG_CHECKPOINT_RESTORE
2247 static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr)
2248 {
2249 	return put_user(me->clear_child_tid, tid_addr);
2250 }
2251 #else
2252 static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr)
2253 {
2254 	return -EINVAL;
2255 }
2256 #endif
2257 
2258 static int propagate_has_child_subreaper(struct task_struct *p, void *data)
2259 {
2260 	/*
2261 	 * If task has has_child_subreaper - all its descendants
2262 	 * already have these flag too and new descendants will
2263 	 * inherit it on fork, skip them.
2264 	 *
2265 	 * If we've found child_reaper - skip descendants in
2266 	 * it's subtree as they will never get out pidns.
2267 	 */
2268 	if (p->signal->has_child_subreaper ||
2269 	    is_child_reaper(task_pid(p)))
2270 		return 0;
2271 
2272 	p->signal->has_child_subreaper = 1;
2273 	return 1;
2274 }
2275 
2276 int __weak arch_prctl_spec_ctrl_get(struct task_struct *t, unsigned long which)
2277 {
2278 	return -EINVAL;
2279 }
2280 
2281 int __weak arch_prctl_spec_ctrl_set(struct task_struct *t, unsigned long which,
2282 				    unsigned long ctrl)
2283 {
2284 	return -EINVAL;
2285 }
2286 
2287 #define PR_IO_FLUSHER (PF_MEMALLOC_NOIO | PF_LOCAL_THROTTLE)
2288 
2289 #ifdef CONFIG_ANON_VMA_NAME
2290 
2291 #define ANON_VMA_NAME_MAX_LEN		80
2292 #define ANON_VMA_NAME_INVALID_CHARS	"\\`$[]"
2293 
2294 static inline bool is_valid_name_char(char ch)
2295 {
2296 	/* printable ascii characters, excluding ANON_VMA_NAME_INVALID_CHARS */
2297 	return ch > 0x1f && ch < 0x7f &&
2298 		!strchr(ANON_VMA_NAME_INVALID_CHARS, ch);
2299 }
2300 
2301 static int prctl_set_vma(unsigned long opt, unsigned long addr,
2302 			 unsigned long size, unsigned long arg)
2303 {
2304 	struct mm_struct *mm = current->mm;
2305 	const char __user *uname;
2306 	struct anon_vma_name *anon_name = NULL;
2307 	int error;
2308 
2309 	switch (opt) {
2310 	case PR_SET_VMA_ANON_NAME:
2311 		uname = (const char __user *)arg;
2312 		if (uname) {
2313 			char *name, *pch;
2314 
2315 			name = strndup_user(uname, ANON_VMA_NAME_MAX_LEN);
2316 			if (IS_ERR(name))
2317 				return PTR_ERR(name);
2318 
2319 			for (pch = name; *pch != '\0'; pch++) {
2320 				if (!is_valid_name_char(*pch)) {
2321 					kfree(name);
2322 					return -EINVAL;
2323 				}
2324 			}
2325 			/* anon_vma has its own copy */
2326 			anon_name = anon_vma_name_alloc(name);
2327 			kfree(name);
2328 			if (!anon_name)
2329 				return -ENOMEM;
2330 
2331 		}
2332 
2333 		mmap_write_lock(mm);
2334 		error = madvise_set_anon_name(mm, addr, size, anon_name);
2335 		mmap_write_unlock(mm);
2336 		anon_vma_name_put(anon_name);
2337 		break;
2338 	default:
2339 		error = -EINVAL;
2340 	}
2341 
2342 	return error;
2343 }
2344 
2345 #else /* CONFIG_ANON_VMA_NAME */
2346 static int prctl_set_vma(unsigned long opt, unsigned long start,
2347 			 unsigned long size, unsigned long arg)
2348 {
2349 	return -EINVAL;
2350 }
2351 #endif /* CONFIG_ANON_VMA_NAME */
2352 
2353 static inline int prctl_set_mdwe(unsigned long bits, unsigned long arg3,
2354 				 unsigned long arg4, unsigned long arg5)
2355 {
2356 	if (arg3 || arg4 || arg5)
2357 		return -EINVAL;
2358 
2359 	if (bits & ~(PR_MDWE_REFUSE_EXEC_GAIN))
2360 		return -EINVAL;
2361 
2362 	if (bits & PR_MDWE_REFUSE_EXEC_GAIN)
2363 		set_bit(MMF_HAS_MDWE, &current->mm->flags);
2364 	else if (test_bit(MMF_HAS_MDWE, &current->mm->flags))
2365 		return -EPERM; /* Cannot unset the flag */
2366 
2367 	return 0;
2368 }
2369 
2370 static inline int prctl_get_mdwe(unsigned long arg2, unsigned long arg3,
2371 				 unsigned long arg4, unsigned long arg5)
2372 {
2373 	if (arg2 || arg3 || arg4 || arg5)
2374 		return -EINVAL;
2375 
2376 	return test_bit(MMF_HAS_MDWE, &current->mm->flags) ?
2377 		PR_MDWE_REFUSE_EXEC_GAIN : 0;
2378 }
2379 
2380 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2381 		unsigned long, arg4, unsigned long, arg5)
2382 {
2383 	struct task_struct *me = current;
2384 	unsigned char comm[sizeof(me->comm)];
2385 	long error;
2386 
2387 	error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2388 	if (error != -ENOSYS)
2389 		return error;
2390 
2391 	error = 0;
2392 	switch (option) {
2393 	case PR_SET_PDEATHSIG:
2394 		if (!valid_signal(arg2)) {
2395 			error = -EINVAL;
2396 			break;
2397 		}
2398 		me->pdeath_signal = arg2;
2399 		break;
2400 	case PR_GET_PDEATHSIG:
2401 		error = put_user(me->pdeath_signal, (int __user *)arg2);
2402 		break;
2403 	case PR_GET_DUMPABLE:
2404 		error = get_dumpable(me->mm);
2405 		break;
2406 	case PR_SET_DUMPABLE:
2407 		if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) {
2408 			error = -EINVAL;
2409 			break;
2410 		}
2411 		set_dumpable(me->mm, arg2);
2412 		break;
2413 
2414 	case PR_SET_UNALIGN:
2415 		error = SET_UNALIGN_CTL(me, arg2);
2416 		break;
2417 	case PR_GET_UNALIGN:
2418 		error = GET_UNALIGN_CTL(me, arg2);
2419 		break;
2420 	case PR_SET_FPEMU:
2421 		error = SET_FPEMU_CTL(me, arg2);
2422 		break;
2423 	case PR_GET_FPEMU:
2424 		error = GET_FPEMU_CTL(me, arg2);
2425 		break;
2426 	case PR_SET_FPEXC:
2427 		error = SET_FPEXC_CTL(me, arg2);
2428 		break;
2429 	case PR_GET_FPEXC:
2430 		error = GET_FPEXC_CTL(me, arg2);
2431 		break;
2432 	case PR_GET_TIMING:
2433 		error = PR_TIMING_STATISTICAL;
2434 		break;
2435 	case PR_SET_TIMING:
2436 		if (arg2 != PR_TIMING_STATISTICAL)
2437 			error = -EINVAL;
2438 		break;
2439 	case PR_SET_NAME:
2440 		comm[sizeof(me->comm) - 1] = 0;
2441 		if (strncpy_from_user(comm, (char __user *)arg2,
2442 				      sizeof(me->comm) - 1) < 0)
2443 			return -EFAULT;
2444 		set_task_comm(me, comm);
2445 		proc_comm_connector(me);
2446 		break;
2447 	case PR_GET_NAME:
2448 		get_task_comm(comm, me);
2449 		if (copy_to_user((char __user *)arg2, comm, sizeof(comm)))
2450 			return -EFAULT;
2451 		break;
2452 	case PR_GET_ENDIAN:
2453 		error = GET_ENDIAN(me, arg2);
2454 		break;
2455 	case PR_SET_ENDIAN:
2456 		error = SET_ENDIAN(me, arg2);
2457 		break;
2458 	case PR_GET_SECCOMP:
2459 		error = prctl_get_seccomp();
2460 		break;
2461 	case PR_SET_SECCOMP:
2462 		error = prctl_set_seccomp(arg2, (char __user *)arg3);
2463 		break;
2464 	case PR_GET_TSC:
2465 		error = GET_TSC_CTL(arg2);
2466 		break;
2467 	case PR_SET_TSC:
2468 		error = SET_TSC_CTL(arg2);
2469 		break;
2470 	case PR_TASK_PERF_EVENTS_DISABLE:
2471 		error = perf_event_task_disable();
2472 		break;
2473 	case PR_TASK_PERF_EVENTS_ENABLE:
2474 		error = perf_event_task_enable();
2475 		break;
2476 	case PR_GET_TIMERSLACK:
2477 		if (current->timer_slack_ns > ULONG_MAX)
2478 			error = ULONG_MAX;
2479 		else
2480 			error = current->timer_slack_ns;
2481 		break;
2482 	case PR_SET_TIMERSLACK:
2483 		if (arg2 <= 0)
2484 			current->timer_slack_ns =
2485 					current->default_timer_slack_ns;
2486 		else
2487 			current->timer_slack_ns = arg2;
2488 		break;
2489 	case PR_MCE_KILL:
2490 		if (arg4 | arg5)
2491 			return -EINVAL;
2492 		switch (arg2) {
2493 		case PR_MCE_KILL_CLEAR:
2494 			if (arg3 != 0)
2495 				return -EINVAL;
2496 			current->flags &= ~PF_MCE_PROCESS;
2497 			break;
2498 		case PR_MCE_KILL_SET:
2499 			current->flags |= PF_MCE_PROCESS;
2500 			if (arg3 == PR_MCE_KILL_EARLY)
2501 				current->flags |= PF_MCE_EARLY;
2502 			else if (arg3 == PR_MCE_KILL_LATE)
2503 				current->flags &= ~PF_MCE_EARLY;
2504 			else if (arg3 == PR_MCE_KILL_DEFAULT)
2505 				current->flags &=
2506 						~(PF_MCE_EARLY|PF_MCE_PROCESS);
2507 			else
2508 				return -EINVAL;
2509 			break;
2510 		default:
2511 			return -EINVAL;
2512 		}
2513 		break;
2514 	case PR_MCE_KILL_GET:
2515 		if (arg2 | arg3 | arg4 | arg5)
2516 			return -EINVAL;
2517 		if (current->flags & PF_MCE_PROCESS)
2518 			error = (current->flags & PF_MCE_EARLY) ?
2519 				PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2520 		else
2521 			error = PR_MCE_KILL_DEFAULT;
2522 		break;
2523 	case PR_SET_MM:
2524 		error = prctl_set_mm(arg2, arg3, arg4, arg5);
2525 		break;
2526 	case PR_GET_TID_ADDRESS:
2527 		error = prctl_get_tid_address(me, (int __user * __user *)arg2);
2528 		break;
2529 	case PR_SET_CHILD_SUBREAPER:
2530 		me->signal->is_child_subreaper = !!arg2;
2531 		if (!arg2)
2532 			break;
2533 
2534 		walk_process_tree(me, propagate_has_child_subreaper, NULL);
2535 		break;
2536 	case PR_GET_CHILD_SUBREAPER:
2537 		error = put_user(me->signal->is_child_subreaper,
2538 				 (int __user *)arg2);
2539 		break;
2540 	case PR_SET_NO_NEW_PRIVS:
2541 		if (arg2 != 1 || arg3 || arg4 || arg5)
2542 			return -EINVAL;
2543 
2544 		task_set_no_new_privs(current);
2545 		break;
2546 	case PR_GET_NO_NEW_PRIVS:
2547 		if (arg2 || arg3 || arg4 || arg5)
2548 			return -EINVAL;
2549 		return task_no_new_privs(current) ? 1 : 0;
2550 	case PR_GET_THP_DISABLE:
2551 		if (arg2 || arg3 || arg4 || arg5)
2552 			return -EINVAL;
2553 		error = !!test_bit(MMF_DISABLE_THP, &me->mm->flags);
2554 		break;
2555 	case PR_SET_THP_DISABLE:
2556 		if (arg3 || arg4 || arg5)
2557 			return -EINVAL;
2558 		if (mmap_write_lock_killable(me->mm))
2559 			return -EINTR;
2560 		if (arg2)
2561 			set_bit(MMF_DISABLE_THP, &me->mm->flags);
2562 		else
2563 			clear_bit(MMF_DISABLE_THP, &me->mm->flags);
2564 		mmap_write_unlock(me->mm);
2565 		break;
2566 	case PR_MPX_ENABLE_MANAGEMENT:
2567 	case PR_MPX_DISABLE_MANAGEMENT:
2568 		/* No longer implemented: */
2569 		return -EINVAL;
2570 	case PR_SET_FP_MODE:
2571 		error = SET_FP_MODE(me, arg2);
2572 		break;
2573 	case PR_GET_FP_MODE:
2574 		error = GET_FP_MODE(me);
2575 		break;
2576 	case PR_SVE_SET_VL:
2577 		error = SVE_SET_VL(arg2);
2578 		break;
2579 	case PR_SVE_GET_VL:
2580 		error = SVE_GET_VL();
2581 		break;
2582 	case PR_SME_SET_VL:
2583 		error = SME_SET_VL(arg2);
2584 		break;
2585 	case PR_SME_GET_VL:
2586 		error = SME_GET_VL();
2587 		break;
2588 	case PR_GET_SPECULATION_CTRL:
2589 		if (arg3 || arg4 || arg5)
2590 			return -EINVAL;
2591 		error = arch_prctl_spec_ctrl_get(me, arg2);
2592 		break;
2593 	case PR_SET_SPECULATION_CTRL:
2594 		if (arg4 || arg5)
2595 			return -EINVAL;
2596 		error = arch_prctl_spec_ctrl_set(me, arg2, arg3);
2597 		break;
2598 	case PR_PAC_RESET_KEYS:
2599 		if (arg3 || arg4 || arg5)
2600 			return -EINVAL;
2601 		error = PAC_RESET_KEYS(me, arg2);
2602 		break;
2603 	case PR_PAC_SET_ENABLED_KEYS:
2604 		if (arg4 || arg5)
2605 			return -EINVAL;
2606 		error = PAC_SET_ENABLED_KEYS(me, arg2, arg3);
2607 		break;
2608 	case PR_PAC_GET_ENABLED_KEYS:
2609 		if (arg2 || arg3 || arg4 || arg5)
2610 			return -EINVAL;
2611 		error = PAC_GET_ENABLED_KEYS(me);
2612 		break;
2613 	case PR_SET_TAGGED_ADDR_CTRL:
2614 		if (arg3 || arg4 || arg5)
2615 			return -EINVAL;
2616 		error = SET_TAGGED_ADDR_CTRL(arg2);
2617 		break;
2618 	case PR_GET_TAGGED_ADDR_CTRL:
2619 		if (arg2 || arg3 || arg4 || arg5)
2620 			return -EINVAL;
2621 		error = GET_TAGGED_ADDR_CTRL();
2622 		break;
2623 	case PR_SET_IO_FLUSHER:
2624 		if (!capable(CAP_SYS_RESOURCE))
2625 			return -EPERM;
2626 
2627 		if (arg3 || arg4 || arg5)
2628 			return -EINVAL;
2629 
2630 		if (arg2 == 1)
2631 			current->flags |= PR_IO_FLUSHER;
2632 		else if (!arg2)
2633 			current->flags &= ~PR_IO_FLUSHER;
2634 		else
2635 			return -EINVAL;
2636 		break;
2637 	case PR_GET_IO_FLUSHER:
2638 		if (!capable(CAP_SYS_RESOURCE))
2639 			return -EPERM;
2640 
2641 		if (arg2 || arg3 || arg4 || arg5)
2642 			return -EINVAL;
2643 
2644 		error = (current->flags & PR_IO_FLUSHER) == PR_IO_FLUSHER;
2645 		break;
2646 	case PR_SET_SYSCALL_USER_DISPATCH:
2647 		error = set_syscall_user_dispatch(arg2, arg3, arg4,
2648 						  (char __user *) arg5);
2649 		break;
2650 #ifdef CONFIG_SCHED_CORE
2651 	case PR_SCHED_CORE:
2652 		error = sched_core_share_pid(arg2, arg3, arg4, arg5);
2653 		break;
2654 #endif
2655 	case PR_SET_MDWE:
2656 		error = prctl_set_mdwe(arg2, arg3, arg4, arg5);
2657 		break;
2658 	case PR_GET_MDWE:
2659 		error = prctl_get_mdwe(arg2, arg3, arg4, arg5);
2660 		break;
2661 	case PR_SET_VMA:
2662 		error = prctl_set_vma(arg2, arg3, arg4, arg5);
2663 		break;
2664 	default:
2665 		error = -EINVAL;
2666 		break;
2667 	}
2668 	return error;
2669 }
2670 
2671 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2672 		struct getcpu_cache __user *, unused)
2673 {
2674 	int err = 0;
2675 	int cpu = raw_smp_processor_id();
2676 
2677 	if (cpup)
2678 		err |= put_user(cpu, cpup);
2679 	if (nodep)
2680 		err |= put_user(cpu_to_node(cpu), nodep);
2681 	return err ? -EFAULT : 0;
2682 }
2683 
2684 /**
2685  * do_sysinfo - fill in sysinfo struct
2686  * @info: pointer to buffer to fill
2687  */
2688 static int do_sysinfo(struct sysinfo *info)
2689 {
2690 	unsigned long mem_total, sav_total;
2691 	unsigned int mem_unit, bitcount;
2692 	struct timespec64 tp;
2693 
2694 	memset(info, 0, sizeof(struct sysinfo));
2695 
2696 	ktime_get_boottime_ts64(&tp);
2697 	timens_add_boottime(&tp);
2698 	info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
2699 
2700 	get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
2701 
2702 	info->procs = nr_threads;
2703 
2704 	si_meminfo(info);
2705 	si_swapinfo(info);
2706 
2707 	/*
2708 	 * If the sum of all the available memory (i.e. ram + swap)
2709 	 * is less than can be stored in a 32 bit unsigned long then
2710 	 * we can be binary compatible with 2.2.x kernels.  If not,
2711 	 * well, in that case 2.2.x was broken anyways...
2712 	 *
2713 	 *  -Erik Andersen <andersee@debian.org>
2714 	 */
2715 
2716 	mem_total = info->totalram + info->totalswap;
2717 	if (mem_total < info->totalram || mem_total < info->totalswap)
2718 		goto out;
2719 	bitcount = 0;
2720 	mem_unit = info->mem_unit;
2721 	while (mem_unit > 1) {
2722 		bitcount++;
2723 		mem_unit >>= 1;
2724 		sav_total = mem_total;
2725 		mem_total <<= 1;
2726 		if (mem_total < sav_total)
2727 			goto out;
2728 	}
2729 
2730 	/*
2731 	 * If mem_total did not overflow, multiply all memory values by
2732 	 * info->mem_unit and set it to 1.  This leaves things compatible
2733 	 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2734 	 * kernels...
2735 	 */
2736 
2737 	info->mem_unit = 1;
2738 	info->totalram <<= bitcount;
2739 	info->freeram <<= bitcount;
2740 	info->sharedram <<= bitcount;
2741 	info->bufferram <<= bitcount;
2742 	info->totalswap <<= bitcount;
2743 	info->freeswap <<= bitcount;
2744 	info->totalhigh <<= bitcount;
2745 	info->freehigh <<= bitcount;
2746 
2747 out:
2748 	return 0;
2749 }
2750 
2751 SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
2752 {
2753 	struct sysinfo val;
2754 
2755 	do_sysinfo(&val);
2756 
2757 	if (copy_to_user(info, &val, sizeof(struct sysinfo)))
2758 		return -EFAULT;
2759 
2760 	return 0;
2761 }
2762 
2763 #ifdef CONFIG_COMPAT
2764 struct compat_sysinfo {
2765 	s32 uptime;
2766 	u32 loads[3];
2767 	u32 totalram;
2768 	u32 freeram;
2769 	u32 sharedram;
2770 	u32 bufferram;
2771 	u32 totalswap;
2772 	u32 freeswap;
2773 	u16 procs;
2774 	u16 pad;
2775 	u32 totalhigh;
2776 	u32 freehigh;
2777 	u32 mem_unit;
2778 	char _f[20-2*sizeof(u32)-sizeof(int)];
2779 };
2780 
2781 COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2782 {
2783 	struct sysinfo s;
2784 	struct compat_sysinfo s_32;
2785 
2786 	do_sysinfo(&s);
2787 
2788 	/* Check to see if any memory value is too large for 32-bit and scale
2789 	 *  down if needed
2790 	 */
2791 	if (upper_32_bits(s.totalram) || upper_32_bits(s.totalswap)) {
2792 		int bitcount = 0;
2793 
2794 		while (s.mem_unit < PAGE_SIZE) {
2795 			s.mem_unit <<= 1;
2796 			bitcount++;
2797 		}
2798 
2799 		s.totalram >>= bitcount;
2800 		s.freeram >>= bitcount;
2801 		s.sharedram >>= bitcount;
2802 		s.bufferram >>= bitcount;
2803 		s.totalswap >>= bitcount;
2804 		s.freeswap >>= bitcount;
2805 		s.totalhigh >>= bitcount;
2806 		s.freehigh >>= bitcount;
2807 	}
2808 
2809 	memset(&s_32, 0, sizeof(s_32));
2810 	s_32.uptime = s.uptime;
2811 	s_32.loads[0] = s.loads[0];
2812 	s_32.loads[1] = s.loads[1];
2813 	s_32.loads[2] = s.loads[2];
2814 	s_32.totalram = s.totalram;
2815 	s_32.freeram = s.freeram;
2816 	s_32.sharedram = s.sharedram;
2817 	s_32.bufferram = s.bufferram;
2818 	s_32.totalswap = s.totalswap;
2819 	s_32.freeswap = s.freeswap;
2820 	s_32.procs = s.procs;
2821 	s_32.totalhigh = s.totalhigh;
2822 	s_32.freehigh = s.freehigh;
2823 	s_32.mem_unit = s.mem_unit;
2824 	if (copy_to_user(info, &s_32, sizeof(s_32)))
2825 		return -EFAULT;
2826 	return 0;
2827 }
2828 #endif /* CONFIG_COMPAT */
2829