xref: /openbmc/linux/kernel/sys.c (revision e40573a4)
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/suspend.h>
29 #include <linux/tty.h>
30 #include <linux/signal.h>
31 #include <linux/cn_proc.h>
32 #include <linux/getcpu.h>
33 #include <linux/task_io_accounting_ops.h>
34 #include <linux/seccomp.h>
35 #include <linux/cpu.h>
36 #include <linux/personality.h>
37 #include <linux/ptrace.h>
38 #include <linux/fs_struct.h>
39 #include <linux/file.h>
40 #include <linux/mount.h>
41 #include <linux/gfp.h>
42 #include <linux/syscore_ops.h>
43 #include <linux/version.h>
44 #include <linux/ctype.h>
45 #include <linux/syscall_user_dispatch.h>
46 
47 #include <linux/compat.h>
48 #include <linux/syscalls.h>
49 #include <linux/kprobes.h>
50 #include <linux/user_namespace.h>
51 #include <linux/time_namespace.h>
52 #include <linux/binfmts.h>
53 
54 #include <linux/sched.h>
55 #include <linux/sched/autogroup.h>
56 #include <linux/sched/loadavg.h>
57 #include <linux/sched/stat.h>
58 #include <linux/sched/mm.h>
59 #include <linux/sched/coredump.h>
60 #include <linux/sched/task.h>
61 #include <linux/sched/cputime.h>
62 #include <linux/rcupdate.h>
63 #include <linux/uidgid.h>
64 #include <linux/cred.h>
65 
66 #include <linux/nospec.h>
67 
68 #include <linux/kmsg_dump.h>
69 /* Move somewhere else to avoid recompiling? */
70 #include <generated/utsrelease.h>
71 
72 #include <linux/uaccess.h>
73 #include <asm/io.h>
74 #include <asm/unistd.h>
75 
76 #include "uid16.h"
77 
78 #ifndef SET_UNALIGN_CTL
79 # define SET_UNALIGN_CTL(a, b)	(-EINVAL)
80 #endif
81 #ifndef GET_UNALIGN_CTL
82 # define GET_UNALIGN_CTL(a, b)	(-EINVAL)
83 #endif
84 #ifndef SET_FPEMU_CTL
85 # define SET_FPEMU_CTL(a, b)	(-EINVAL)
86 #endif
87 #ifndef GET_FPEMU_CTL
88 # define GET_FPEMU_CTL(a, b)	(-EINVAL)
89 #endif
90 #ifndef SET_FPEXC_CTL
91 # define SET_FPEXC_CTL(a, b)	(-EINVAL)
92 #endif
93 #ifndef GET_FPEXC_CTL
94 # define GET_FPEXC_CTL(a, b)	(-EINVAL)
95 #endif
96 #ifndef GET_ENDIAN
97 # define GET_ENDIAN(a, b)	(-EINVAL)
98 #endif
99 #ifndef SET_ENDIAN
100 # define SET_ENDIAN(a, b)	(-EINVAL)
101 #endif
102 #ifndef GET_TSC_CTL
103 # define GET_TSC_CTL(a)		(-EINVAL)
104 #endif
105 #ifndef SET_TSC_CTL
106 # define SET_TSC_CTL(a)		(-EINVAL)
107 #endif
108 #ifndef GET_FP_MODE
109 # define GET_FP_MODE(a)		(-EINVAL)
110 #endif
111 #ifndef SET_FP_MODE
112 # define SET_FP_MODE(a,b)	(-EINVAL)
113 #endif
114 #ifndef SVE_SET_VL
115 # define SVE_SET_VL(a)		(-EINVAL)
116 #endif
117 #ifndef SVE_GET_VL
118 # define SVE_GET_VL()		(-EINVAL)
119 #endif
120 #ifndef SME_SET_VL
121 # define SME_SET_VL(a)		(-EINVAL)
122 #endif
123 #ifndef SME_GET_VL
124 # define SME_GET_VL()		(-EINVAL)
125 #endif
126 #ifndef PAC_RESET_KEYS
127 # define PAC_RESET_KEYS(a, b)	(-EINVAL)
128 #endif
129 #ifndef PAC_SET_ENABLED_KEYS
130 # define PAC_SET_ENABLED_KEYS(a, b, c)	(-EINVAL)
131 #endif
132 #ifndef PAC_GET_ENABLED_KEYS
133 # define PAC_GET_ENABLED_KEYS(a)	(-EINVAL)
134 #endif
135 #ifndef SET_TAGGED_ADDR_CTRL
136 # define SET_TAGGED_ADDR_CTRL(a)	(-EINVAL)
137 #endif
138 #ifndef GET_TAGGED_ADDR_CTRL
139 # define GET_TAGGED_ADDR_CTRL()		(-EINVAL)
140 #endif
141 
142 /*
143  * this is where the system-wide overflow UID and GID are defined, for
144  * architectures that now have 32-bit UID/GID but didn't in the past
145  */
146 
147 int overflowuid = DEFAULT_OVERFLOWUID;
148 int overflowgid = DEFAULT_OVERFLOWGID;
149 
150 EXPORT_SYMBOL(overflowuid);
151 EXPORT_SYMBOL(overflowgid);
152 
153 /*
154  * the same as above, but for filesystems which can only store a 16-bit
155  * UID and GID. as such, this is needed on all architectures
156  */
157 
158 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
159 int fs_overflowgid = DEFAULT_FS_OVERFLOWGID;
160 
161 EXPORT_SYMBOL(fs_overflowuid);
162 EXPORT_SYMBOL(fs_overflowgid);
163 
164 /*
165  * Returns true if current's euid is same as p's uid or euid,
166  * or has CAP_SYS_NICE to p's user_ns.
167  *
168  * Called with rcu_read_lock, creds are safe
169  */
170 static bool set_one_prio_perm(struct task_struct *p)
171 {
172 	const struct cred *cred = current_cred(), *pcred = __task_cred(p);
173 
174 	if (uid_eq(pcred->uid,  cred->euid) ||
175 	    uid_eq(pcred->euid, cred->euid))
176 		return true;
177 	if (ns_capable(pcred->user_ns, CAP_SYS_NICE))
178 		return true;
179 	return false;
180 }
181 
182 /*
183  * set the priority of a task
184  * - the caller must hold the RCU read lock
185  */
186 static int set_one_prio(struct task_struct *p, int niceval, int error)
187 {
188 	int no_nice;
189 
190 	if (!set_one_prio_perm(p)) {
191 		error = -EPERM;
192 		goto out;
193 	}
194 	if (niceval < task_nice(p) && !can_nice(p, niceval)) {
195 		error = -EACCES;
196 		goto out;
197 	}
198 	no_nice = security_task_setnice(p, niceval);
199 	if (no_nice) {
200 		error = no_nice;
201 		goto out;
202 	}
203 	if (error == -ESRCH)
204 		error = 0;
205 	set_user_nice(p, niceval);
206 out:
207 	return error;
208 }
209 
210 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
211 {
212 	struct task_struct *g, *p;
213 	struct user_struct *user;
214 	const struct cred *cred = current_cred();
215 	int error = -EINVAL;
216 	struct pid *pgrp;
217 	kuid_t uid;
218 
219 	if (which > PRIO_USER || which < PRIO_PROCESS)
220 		goto out;
221 
222 	/* normalize: avoid signed division (rounding problems) */
223 	error = -ESRCH;
224 	if (niceval < MIN_NICE)
225 		niceval = MIN_NICE;
226 	if (niceval > MAX_NICE)
227 		niceval = MAX_NICE;
228 
229 	rcu_read_lock();
230 	switch (which) {
231 	case PRIO_PROCESS:
232 		if (who)
233 			p = find_task_by_vpid(who);
234 		else
235 			p = current;
236 		if (p)
237 			error = set_one_prio(p, niceval, error);
238 		break;
239 	case PRIO_PGRP:
240 		if (who)
241 			pgrp = find_vpid(who);
242 		else
243 			pgrp = task_pgrp(current);
244 		read_lock(&tasklist_lock);
245 		do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
246 			error = set_one_prio(p, niceval, error);
247 		} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
248 		read_unlock(&tasklist_lock);
249 		break;
250 	case PRIO_USER:
251 		uid = make_kuid(cred->user_ns, who);
252 		user = cred->user;
253 		if (!who)
254 			uid = cred->uid;
255 		else if (!uid_eq(uid, cred->uid)) {
256 			user = find_user(uid);
257 			if (!user)
258 				goto out_unlock;	/* No processes for this user */
259 		}
260 		for_each_process_thread(g, p) {
261 			if (uid_eq(task_uid(p), uid) && task_pid_vnr(p))
262 				error = set_one_prio(p, niceval, error);
263 		}
264 		if (!uid_eq(uid, cred->uid))
265 			free_uid(user);		/* For find_user() */
266 		break;
267 	}
268 out_unlock:
269 	rcu_read_unlock();
270 out:
271 	return error;
272 }
273 
274 /*
275  * Ugh. To avoid negative return values, "getpriority()" will
276  * not return the normal nice-value, but a negated value that
277  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
278  * to stay compatible.
279  */
280 SYSCALL_DEFINE2(getpriority, int, which, int, who)
281 {
282 	struct task_struct *g, *p;
283 	struct user_struct *user;
284 	const struct cred *cred = current_cred();
285 	long niceval, retval = -ESRCH;
286 	struct pid *pgrp;
287 	kuid_t uid;
288 
289 	if (which > PRIO_USER || which < PRIO_PROCESS)
290 		return -EINVAL;
291 
292 	rcu_read_lock();
293 	switch (which) {
294 	case PRIO_PROCESS:
295 		if (who)
296 			p = find_task_by_vpid(who);
297 		else
298 			p = current;
299 		if (p) {
300 			niceval = nice_to_rlimit(task_nice(p));
301 			if (niceval > retval)
302 				retval = niceval;
303 		}
304 		break;
305 	case PRIO_PGRP:
306 		if (who)
307 			pgrp = find_vpid(who);
308 		else
309 			pgrp = task_pgrp(current);
310 		read_lock(&tasklist_lock);
311 		do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
312 			niceval = nice_to_rlimit(task_nice(p));
313 			if (niceval > retval)
314 				retval = niceval;
315 		} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
316 		read_unlock(&tasklist_lock);
317 		break;
318 	case PRIO_USER:
319 		uid = make_kuid(cred->user_ns, who);
320 		user = cred->user;
321 		if (!who)
322 			uid = cred->uid;
323 		else if (!uid_eq(uid, cred->uid)) {
324 			user = find_user(uid);
325 			if (!user)
326 				goto out_unlock;	/* No processes for this user */
327 		}
328 		for_each_process_thread(g, p) {
329 			if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) {
330 				niceval = nice_to_rlimit(task_nice(p));
331 				if (niceval > retval)
332 					retval = niceval;
333 			}
334 		}
335 		if (!uid_eq(uid, cred->uid))
336 			free_uid(user);		/* for find_user() */
337 		break;
338 	}
339 out_unlock:
340 	rcu_read_unlock();
341 
342 	return retval;
343 }
344 
345 /*
346  * Unprivileged users may change the real gid to the effective gid
347  * or vice versa.  (BSD-style)
348  *
349  * If you set the real gid at all, or set the effective gid to a value not
350  * equal to the real gid, then the saved gid is set to the new effective gid.
351  *
352  * This makes it possible for a setgid program to completely drop its
353  * privileges, which is often a useful assertion to make when you are doing
354  * a security audit over a program.
355  *
356  * The general idea is that a program which uses just setregid() will be
357  * 100% compatible with BSD.  A program which uses just setgid() will be
358  * 100% compatible with POSIX with saved IDs.
359  *
360  * SMP: There are not races, the GIDs are checked only by filesystem
361  *      operations (as far as semantic preservation is concerned).
362  */
363 #ifdef CONFIG_MULTIUSER
364 long __sys_setregid(gid_t rgid, gid_t egid)
365 {
366 	struct user_namespace *ns = current_user_ns();
367 	const struct cred *old;
368 	struct cred *new;
369 	int retval;
370 	kgid_t krgid, kegid;
371 
372 	krgid = make_kgid(ns, rgid);
373 	kegid = make_kgid(ns, egid);
374 
375 	if ((rgid != (gid_t) -1) && !gid_valid(krgid))
376 		return -EINVAL;
377 	if ((egid != (gid_t) -1) && !gid_valid(kegid))
378 		return -EINVAL;
379 
380 	new = prepare_creds();
381 	if (!new)
382 		return -ENOMEM;
383 	old = current_cred();
384 
385 	retval = -EPERM;
386 	if (rgid != (gid_t) -1) {
387 		if (gid_eq(old->gid, krgid) ||
388 		    gid_eq(old->egid, krgid) ||
389 		    ns_capable_setid(old->user_ns, CAP_SETGID))
390 			new->gid = krgid;
391 		else
392 			goto error;
393 	}
394 	if (egid != (gid_t) -1) {
395 		if (gid_eq(old->gid, kegid) ||
396 		    gid_eq(old->egid, kegid) ||
397 		    gid_eq(old->sgid, kegid) ||
398 		    ns_capable_setid(old->user_ns, CAP_SETGID))
399 			new->egid = kegid;
400 		else
401 			goto error;
402 	}
403 
404 	if (rgid != (gid_t) -1 ||
405 	    (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
406 		new->sgid = new->egid;
407 	new->fsgid = new->egid;
408 
409 	retval = security_task_fix_setgid(new, old, LSM_SETID_RE);
410 	if (retval < 0)
411 		goto error;
412 
413 	return commit_creds(new);
414 
415 error:
416 	abort_creds(new);
417 	return retval;
418 }
419 
420 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
421 {
422 	return __sys_setregid(rgid, egid);
423 }
424 
425 /*
426  * setgid() is implemented like SysV w/ SAVED_IDS
427  *
428  * SMP: Same implicit races as above.
429  */
430 long __sys_setgid(gid_t gid)
431 {
432 	struct user_namespace *ns = current_user_ns();
433 	const struct cred *old;
434 	struct cred *new;
435 	int retval;
436 	kgid_t kgid;
437 
438 	kgid = make_kgid(ns, gid);
439 	if (!gid_valid(kgid))
440 		return -EINVAL;
441 
442 	new = prepare_creds();
443 	if (!new)
444 		return -ENOMEM;
445 	old = current_cred();
446 
447 	retval = -EPERM;
448 	if (ns_capable_setid(old->user_ns, CAP_SETGID))
449 		new->gid = new->egid = new->sgid = new->fsgid = kgid;
450 	else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
451 		new->egid = new->fsgid = kgid;
452 	else
453 		goto error;
454 
455 	retval = security_task_fix_setgid(new, old, LSM_SETID_ID);
456 	if (retval < 0)
457 		goto error;
458 
459 	return commit_creds(new);
460 
461 error:
462 	abort_creds(new);
463 	return retval;
464 }
465 
466 SYSCALL_DEFINE1(setgid, gid_t, gid)
467 {
468 	return __sys_setgid(gid);
469 }
470 
471 /*
472  * change the user struct in a credentials set to match the new UID
473  */
474 static int set_user(struct cred *new)
475 {
476 	struct user_struct *new_user;
477 
478 	new_user = alloc_uid(new->uid);
479 	if (!new_user)
480 		return -EAGAIN;
481 
482 	free_uid(new->user);
483 	new->user = new_user;
484 	return 0;
485 }
486 
487 static void flag_nproc_exceeded(struct cred *new)
488 {
489 	if (new->ucounts == current_ucounts())
490 		return;
491 
492 	/*
493 	 * We don't fail in case of NPROC limit excess here because too many
494 	 * poorly written programs don't check set*uid() return code, assuming
495 	 * it never fails if called by root.  We may still enforce NPROC limit
496 	 * for programs doing set*uid()+execve() by harmlessly deferring the
497 	 * failure to the execve() stage.
498 	 */
499 	if (is_ucounts_overlimit(new->ucounts, UCOUNT_RLIMIT_NPROC, rlimit(RLIMIT_NPROC)) &&
500 			new->user != INIT_USER)
501 		current->flags |= PF_NPROC_EXCEEDED;
502 	else
503 		current->flags &= ~PF_NPROC_EXCEEDED;
504 }
505 
506 /*
507  * Unprivileged users may change the real uid to the effective uid
508  * or vice versa.  (BSD-style)
509  *
510  * If you set the real uid at all, or set the effective uid to a value not
511  * equal to the real uid, then the saved uid is set to the new effective uid.
512  *
513  * This makes it possible for a setuid program to completely drop its
514  * privileges, which is often a useful assertion to make when you are doing
515  * a security audit over a program.
516  *
517  * The general idea is that a program which uses just setreuid() will be
518  * 100% compatible with BSD.  A program which uses just setuid() will be
519  * 100% compatible with POSIX with saved IDs.
520  */
521 long __sys_setreuid(uid_t ruid, uid_t euid)
522 {
523 	struct user_namespace *ns = current_user_ns();
524 	const struct cred *old;
525 	struct cred *new;
526 	int retval;
527 	kuid_t kruid, keuid;
528 
529 	kruid = make_kuid(ns, ruid);
530 	keuid = make_kuid(ns, euid);
531 
532 	if ((ruid != (uid_t) -1) && !uid_valid(kruid))
533 		return -EINVAL;
534 	if ((euid != (uid_t) -1) && !uid_valid(keuid))
535 		return -EINVAL;
536 
537 	new = prepare_creds();
538 	if (!new)
539 		return -ENOMEM;
540 	old = current_cred();
541 
542 	retval = -EPERM;
543 	if (ruid != (uid_t) -1) {
544 		new->uid = kruid;
545 		if (!uid_eq(old->uid, kruid) &&
546 		    !uid_eq(old->euid, kruid) &&
547 		    !ns_capable_setid(old->user_ns, CAP_SETUID))
548 			goto error;
549 	}
550 
551 	if (euid != (uid_t) -1) {
552 		new->euid = keuid;
553 		if (!uid_eq(old->uid, keuid) &&
554 		    !uid_eq(old->euid, keuid) &&
555 		    !uid_eq(old->suid, keuid) &&
556 		    !ns_capable_setid(old->user_ns, CAP_SETUID))
557 			goto error;
558 	}
559 
560 	if (!uid_eq(new->uid, old->uid)) {
561 		retval = set_user(new);
562 		if (retval < 0)
563 			goto error;
564 	}
565 	if (ruid != (uid_t) -1 ||
566 	    (euid != (uid_t) -1 && !uid_eq(keuid, old->uid)))
567 		new->suid = new->euid;
568 	new->fsuid = new->euid;
569 
570 	retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
571 	if (retval < 0)
572 		goto error;
573 
574 	retval = set_cred_ucounts(new);
575 	if (retval < 0)
576 		goto error;
577 
578 	flag_nproc_exceeded(new);
579 	return commit_creds(new);
580 
581 error:
582 	abort_creds(new);
583 	return retval;
584 }
585 
586 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
587 {
588 	return __sys_setreuid(ruid, euid);
589 }
590 
591 /*
592  * setuid() is implemented like SysV with SAVED_IDS
593  *
594  * Note that SAVED_ID's is deficient in that a setuid root program
595  * like sendmail, for example, cannot set its uid to be a normal
596  * user and then switch back, because if you're root, setuid() sets
597  * the saved uid too.  If you don't like this, blame the bright people
598  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
599  * will allow a root program to temporarily drop privileges and be able to
600  * regain them by swapping the real and effective uid.
601  */
602 long __sys_setuid(uid_t uid)
603 {
604 	struct user_namespace *ns = current_user_ns();
605 	const struct cred *old;
606 	struct cred *new;
607 	int retval;
608 	kuid_t kuid;
609 
610 	kuid = make_kuid(ns, uid);
611 	if (!uid_valid(kuid))
612 		return -EINVAL;
613 
614 	new = prepare_creds();
615 	if (!new)
616 		return -ENOMEM;
617 	old = current_cred();
618 
619 	retval = -EPERM;
620 	if (ns_capable_setid(old->user_ns, CAP_SETUID)) {
621 		new->suid = new->uid = kuid;
622 		if (!uid_eq(kuid, old->uid)) {
623 			retval = set_user(new);
624 			if (retval < 0)
625 				goto error;
626 		}
627 	} else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) {
628 		goto error;
629 	}
630 
631 	new->fsuid = new->euid = kuid;
632 
633 	retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
634 	if (retval < 0)
635 		goto error;
636 
637 	retval = set_cred_ucounts(new);
638 	if (retval < 0)
639 		goto error;
640 
641 	flag_nproc_exceeded(new);
642 	return commit_creds(new);
643 
644 error:
645 	abort_creds(new);
646 	return retval;
647 }
648 
649 SYSCALL_DEFINE1(setuid, uid_t, uid)
650 {
651 	return __sys_setuid(uid);
652 }
653 
654 
655 /*
656  * This function implements a generic ability to update ruid, euid,
657  * and suid.  This allows you to implement the 4.4 compatible seteuid().
658  */
659 long __sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
660 {
661 	struct user_namespace *ns = current_user_ns();
662 	const struct cred *old;
663 	struct cred *new;
664 	int retval;
665 	kuid_t kruid, keuid, ksuid;
666 
667 	kruid = make_kuid(ns, ruid);
668 	keuid = make_kuid(ns, euid);
669 	ksuid = make_kuid(ns, suid);
670 
671 	if ((ruid != (uid_t) -1) && !uid_valid(kruid))
672 		return -EINVAL;
673 
674 	if ((euid != (uid_t) -1) && !uid_valid(keuid))
675 		return -EINVAL;
676 
677 	if ((suid != (uid_t) -1) && !uid_valid(ksuid))
678 		return -EINVAL;
679 
680 	new = prepare_creds();
681 	if (!new)
682 		return -ENOMEM;
683 
684 	old = current_cred();
685 
686 	retval = -EPERM;
687 	if (!ns_capable_setid(old->user_ns, CAP_SETUID)) {
688 		if (ruid != (uid_t) -1        && !uid_eq(kruid, old->uid) &&
689 		    !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid))
690 			goto error;
691 		if (euid != (uid_t) -1        && !uid_eq(keuid, old->uid) &&
692 		    !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid))
693 			goto error;
694 		if (suid != (uid_t) -1        && !uid_eq(ksuid, old->uid) &&
695 		    !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid))
696 			goto error;
697 	}
698 
699 	if (ruid != (uid_t) -1) {
700 		new->uid = kruid;
701 		if (!uid_eq(kruid, old->uid)) {
702 			retval = set_user(new);
703 			if (retval < 0)
704 				goto error;
705 		}
706 	}
707 	if (euid != (uid_t) -1)
708 		new->euid = keuid;
709 	if (suid != (uid_t) -1)
710 		new->suid = ksuid;
711 	new->fsuid = new->euid;
712 
713 	retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
714 	if (retval < 0)
715 		goto error;
716 
717 	retval = set_cred_ucounts(new);
718 	if (retval < 0)
719 		goto error;
720 
721 	flag_nproc_exceeded(new);
722 	return commit_creds(new);
723 
724 error:
725 	abort_creds(new);
726 	return retval;
727 }
728 
729 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
730 {
731 	return __sys_setresuid(ruid, euid, suid);
732 }
733 
734 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
735 {
736 	const struct cred *cred = current_cred();
737 	int retval;
738 	uid_t ruid, euid, suid;
739 
740 	ruid = from_kuid_munged(cred->user_ns, cred->uid);
741 	euid = from_kuid_munged(cred->user_ns, cred->euid);
742 	suid = from_kuid_munged(cred->user_ns, cred->suid);
743 
744 	retval = put_user(ruid, ruidp);
745 	if (!retval) {
746 		retval = put_user(euid, euidp);
747 		if (!retval)
748 			return put_user(suid, suidp);
749 	}
750 	return retval;
751 }
752 
753 /*
754  * Same as above, but for rgid, egid, sgid.
755  */
756 long __sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
757 {
758 	struct user_namespace *ns = current_user_ns();
759 	const struct cred *old;
760 	struct cred *new;
761 	int retval;
762 	kgid_t krgid, kegid, ksgid;
763 
764 	krgid = make_kgid(ns, rgid);
765 	kegid = make_kgid(ns, egid);
766 	ksgid = make_kgid(ns, sgid);
767 
768 	if ((rgid != (gid_t) -1) && !gid_valid(krgid))
769 		return -EINVAL;
770 	if ((egid != (gid_t) -1) && !gid_valid(kegid))
771 		return -EINVAL;
772 	if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
773 		return -EINVAL;
774 
775 	new = prepare_creds();
776 	if (!new)
777 		return -ENOMEM;
778 	old = current_cred();
779 
780 	retval = -EPERM;
781 	if (!ns_capable_setid(old->user_ns, CAP_SETGID)) {
782 		if (rgid != (gid_t) -1        && !gid_eq(krgid, old->gid) &&
783 		    !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid))
784 			goto error;
785 		if (egid != (gid_t) -1        && !gid_eq(kegid, old->gid) &&
786 		    !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid))
787 			goto error;
788 		if (sgid != (gid_t) -1        && !gid_eq(ksgid, old->gid) &&
789 		    !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid))
790 			goto error;
791 	}
792 
793 	if (rgid != (gid_t) -1)
794 		new->gid = krgid;
795 	if (egid != (gid_t) -1)
796 		new->egid = kegid;
797 	if (sgid != (gid_t) -1)
798 		new->sgid = ksgid;
799 	new->fsgid = new->egid;
800 
801 	retval = security_task_fix_setgid(new, old, LSM_SETID_RES);
802 	if (retval < 0)
803 		goto error;
804 
805 	return commit_creds(new);
806 
807 error:
808 	abort_creds(new);
809 	return retval;
810 }
811 
812 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
813 {
814 	return __sys_setresgid(rgid, egid, sgid);
815 }
816 
817 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
818 {
819 	const struct cred *cred = current_cred();
820 	int retval;
821 	gid_t rgid, egid, sgid;
822 
823 	rgid = from_kgid_munged(cred->user_ns, cred->gid);
824 	egid = from_kgid_munged(cred->user_ns, cred->egid);
825 	sgid = from_kgid_munged(cred->user_ns, cred->sgid);
826 
827 	retval = put_user(rgid, rgidp);
828 	if (!retval) {
829 		retval = put_user(egid, egidp);
830 		if (!retval)
831 			retval = put_user(sgid, sgidp);
832 	}
833 
834 	return retval;
835 }
836 
837 
838 /*
839  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
840  * is used for "access()" and for the NFS daemon (letting nfsd stay at
841  * whatever uid it wants to). It normally shadows "euid", except when
842  * explicitly set by setfsuid() or for access..
843  */
844 long __sys_setfsuid(uid_t uid)
845 {
846 	const struct cred *old;
847 	struct cred *new;
848 	uid_t old_fsuid;
849 	kuid_t kuid;
850 
851 	old = current_cred();
852 	old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
853 
854 	kuid = make_kuid(old->user_ns, uid);
855 	if (!uid_valid(kuid))
856 		return old_fsuid;
857 
858 	new = prepare_creds();
859 	if (!new)
860 		return old_fsuid;
861 
862 	if (uid_eq(kuid, old->uid)  || uid_eq(kuid, old->euid)  ||
863 	    uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
864 	    ns_capable_setid(old->user_ns, CAP_SETUID)) {
865 		if (!uid_eq(kuid, old->fsuid)) {
866 			new->fsuid = kuid;
867 			if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
868 				goto change_okay;
869 		}
870 	}
871 
872 	abort_creds(new);
873 	return old_fsuid;
874 
875 change_okay:
876 	commit_creds(new);
877 	return old_fsuid;
878 }
879 
880 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
881 {
882 	return __sys_setfsuid(uid);
883 }
884 
885 /*
886  * Samma på svenska..
887  */
888 long __sys_setfsgid(gid_t gid)
889 {
890 	const struct cred *old;
891 	struct cred *new;
892 	gid_t old_fsgid;
893 	kgid_t kgid;
894 
895 	old = current_cred();
896 	old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
897 
898 	kgid = make_kgid(old->user_ns, gid);
899 	if (!gid_valid(kgid))
900 		return old_fsgid;
901 
902 	new = prepare_creds();
903 	if (!new)
904 		return old_fsgid;
905 
906 	if (gid_eq(kgid, old->gid)  || gid_eq(kgid, old->egid)  ||
907 	    gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
908 	    ns_capable_setid(old->user_ns, CAP_SETGID)) {
909 		if (!gid_eq(kgid, old->fsgid)) {
910 			new->fsgid = kgid;
911 			if (security_task_fix_setgid(new,old,LSM_SETID_FS) == 0)
912 				goto change_okay;
913 		}
914 	}
915 
916 	abort_creds(new);
917 	return old_fsgid;
918 
919 change_okay:
920 	commit_creds(new);
921 	return old_fsgid;
922 }
923 
924 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
925 {
926 	return __sys_setfsgid(gid);
927 }
928 #endif /* CONFIG_MULTIUSER */
929 
930 /**
931  * sys_getpid - return the thread group id of the current process
932  *
933  * Note, despite the name, this returns the tgid not the pid.  The tgid and
934  * the pid are identical unless CLONE_THREAD was specified on clone() in
935  * which case the tgid is the same in all threads of the same group.
936  *
937  * This is SMP safe as current->tgid does not change.
938  */
939 SYSCALL_DEFINE0(getpid)
940 {
941 	return task_tgid_vnr(current);
942 }
943 
944 /* Thread ID - the internal kernel "pid" */
945 SYSCALL_DEFINE0(gettid)
946 {
947 	return task_pid_vnr(current);
948 }
949 
950 /*
951  * Accessing ->real_parent is not SMP-safe, it could
952  * change from under us. However, we can use a stale
953  * value of ->real_parent under rcu_read_lock(), see
954  * release_task()->call_rcu(delayed_put_task_struct).
955  */
956 SYSCALL_DEFINE0(getppid)
957 {
958 	int pid;
959 
960 	rcu_read_lock();
961 	pid = task_tgid_vnr(rcu_dereference(current->real_parent));
962 	rcu_read_unlock();
963 
964 	return pid;
965 }
966 
967 SYSCALL_DEFINE0(getuid)
968 {
969 	/* Only we change this so SMP safe */
970 	return from_kuid_munged(current_user_ns(), current_uid());
971 }
972 
973 SYSCALL_DEFINE0(geteuid)
974 {
975 	/* Only we change this so SMP safe */
976 	return from_kuid_munged(current_user_ns(), current_euid());
977 }
978 
979 SYSCALL_DEFINE0(getgid)
980 {
981 	/* Only we change this so SMP safe */
982 	return from_kgid_munged(current_user_ns(), current_gid());
983 }
984 
985 SYSCALL_DEFINE0(getegid)
986 {
987 	/* Only we change this so SMP safe */
988 	return from_kgid_munged(current_user_ns(), current_egid());
989 }
990 
991 static void do_sys_times(struct tms *tms)
992 {
993 	u64 tgutime, tgstime, cutime, cstime;
994 
995 	thread_group_cputime_adjusted(current, &tgutime, &tgstime);
996 	cutime = current->signal->cutime;
997 	cstime = current->signal->cstime;
998 	tms->tms_utime = nsec_to_clock_t(tgutime);
999 	tms->tms_stime = nsec_to_clock_t(tgstime);
1000 	tms->tms_cutime = nsec_to_clock_t(cutime);
1001 	tms->tms_cstime = nsec_to_clock_t(cstime);
1002 }
1003 
1004 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
1005 {
1006 	if (tbuf) {
1007 		struct tms tmp;
1008 
1009 		do_sys_times(&tmp);
1010 		if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1011 			return -EFAULT;
1012 	}
1013 	force_successful_syscall_return();
1014 	return (long) jiffies_64_to_clock_t(get_jiffies_64());
1015 }
1016 
1017 #ifdef CONFIG_COMPAT
1018 static compat_clock_t clock_t_to_compat_clock_t(clock_t x)
1019 {
1020 	return compat_jiffies_to_clock_t(clock_t_to_jiffies(x));
1021 }
1022 
1023 COMPAT_SYSCALL_DEFINE1(times, struct compat_tms __user *, tbuf)
1024 {
1025 	if (tbuf) {
1026 		struct tms tms;
1027 		struct compat_tms tmp;
1028 
1029 		do_sys_times(&tms);
1030 		/* Convert our struct tms to the compat version. */
1031 		tmp.tms_utime = clock_t_to_compat_clock_t(tms.tms_utime);
1032 		tmp.tms_stime = clock_t_to_compat_clock_t(tms.tms_stime);
1033 		tmp.tms_cutime = clock_t_to_compat_clock_t(tms.tms_cutime);
1034 		tmp.tms_cstime = clock_t_to_compat_clock_t(tms.tms_cstime);
1035 		if (copy_to_user(tbuf, &tmp, sizeof(tmp)))
1036 			return -EFAULT;
1037 	}
1038 	force_successful_syscall_return();
1039 	return compat_jiffies_to_clock_t(jiffies);
1040 }
1041 #endif
1042 
1043 /*
1044  * This needs some heavy checking ...
1045  * I just haven't the stomach for it. I also don't fully
1046  * understand sessions/pgrp etc. Let somebody who does explain it.
1047  *
1048  * OK, I think I have the protection semantics right.... this is really
1049  * only important on a multi-user system anyway, to make sure one user
1050  * can't send a signal to a process owned by another.  -TYT, 12/12/91
1051  *
1052  * !PF_FORKNOEXEC check to conform completely to POSIX.
1053  */
1054 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
1055 {
1056 	struct task_struct *p;
1057 	struct task_struct *group_leader = current->group_leader;
1058 	struct pid *pgrp;
1059 	int err;
1060 
1061 	if (!pid)
1062 		pid = task_pid_vnr(group_leader);
1063 	if (!pgid)
1064 		pgid = pid;
1065 	if (pgid < 0)
1066 		return -EINVAL;
1067 	rcu_read_lock();
1068 
1069 	/* From this point forward we keep holding onto the tasklist lock
1070 	 * so that our parent does not change from under us. -DaveM
1071 	 */
1072 	write_lock_irq(&tasklist_lock);
1073 
1074 	err = -ESRCH;
1075 	p = find_task_by_vpid(pid);
1076 	if (!p)
1077 		goto out;
1078 
1079 	err = -EINVAL;
1080 	if (!thread_group_leader(p))
1081 		goto out;
1082 
1083 	if (same_thread_group(p->real_parent, group_leader)) {
1084 		err = -EPERM;
1085 		if (task_session(p) != task_session(group_leader))
1086 			goto out;
1087 		err = -EACCES;
1088 		if (!(p->flags & PF_FORKNOEXEC))
1089 			goto out;
1090 	} else {
1091 		err = -ESRCH;
1092 		if (p != group_leader)
1093 			goto out;
1094 	}
1095 
1096 	err = -EPERM;
1097 	if (p->signal->leader)
1098 		goto out;
1099 
1100 	pgrp = task_pid(p);
1101 	if (pgid != pid) {
1102 		struct task_struct *g;
1103 
1104 		pgrp = find_vpid(pgid);
1105 		g = pid_task(pgrp, PIDTYPE_PGID);
1106 		if (!g || task_session(g) != task_session(group_leader))
1107 			goto out;
1108 	}
1109 
1110 	err = security_task_setpgid(p, pgid);
1111 	if (err)
1112 		goto out;
1113 
1114 	if (task_pgrp(p) != pgrp)
1115 		change_pid(p, PIDTYPE_PGID, pgrp);
1116 
1117 	err = 0;
1118 out:
1119 	/* All paths lead to here, thus we are safe. -DaveM */
1120 	write_unlock_irq(&tasklist_lock);
1121 	rcu_read_unlock();
1122 	return err;
1123 }
1124 
1125 static int do_getpgid(pid_t pid)
1126 {
1127 	struct task_struct *p;
1128 	struct pid *grp;
1129 	int retval;
1130 
1131 	rcu_read_lock();
1132 	if (!pid)
1133 		grp = task_pgrp(current);
1134 	else {
1135 		retval = -ESRCH;
1136 		p = find_task_by_vpid(pid);
1137 		if (!p)
1138 			goto out;
1139 		grp = task_pgrp(p);
1140 		if (!grp)
1141 			goto out;
1142 
1143 		retval = security_task_getpgid(p);
1144 		if (retval)
1145 			goto out;
1146 	}
1147 	retval = pid_vnr(grp);
1148 out:
1149 	rcu_read_unlock();
1150 	return retval;
1151 }
1152 
1153 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1154 {
1155 	return do_getpgid(pid);
1156 }
1157 
1158 #ifdef __ARCH_WANT_SYS_GETPGRP
1159 
1160 SYSCALL_DEFINE0(getpgrp)
1161 {
1162 	return do_getpgid(0);
1163 }
1164 
1165 #endif
1166 
1167 SYSCALL_DEFINE1(getsid, pid_t, pid)
1168 {
1169 	struct task_struct *p;
1170 	struct pid *sid;
1171 	int retval;
1172 
1173 	rcu_read_lock();
1174 	if (!pid)
1175 		sid = task_session(current);
1176 	else {
1177 		retval = -ESRCH;
1178 		p = find_task_by_vpid(pid);
1179 		if (!p)
1180 			goto out;
1181 		sid = task_session(p);
1182 		if (!sid)
1183 			goto out;
1184 
1185 		retval = security_task_getsid(p);
1186 		if (retval)
1187 			goto out;
1188 	}
1189 	retval = pid_vnr(sid);
1190 out:
1191 	rcu_read_unlock();
1192 	return retval;
1193 }
1194 
1195 static void set_special_pids(struct pid *pid)
1196 {
1197 	struct task_struct *curr = current->group_leader;
1198 
1199 	if (task_session(curr) != pid)
1200 		change_pid(curr, PIDTYPE_SID, pid);
1201 
1202 	if (task_pgrp(curr) != pid)
1203 		change_pid(curr, PIDTYPE_PGID, pid);
1204 }
1205 
1206 int ksys_setsid(void)
1207 {
1208 	struct task_struct *group_leader = current->group_leader;
1209 	struct pid *sid = task_pid(group_leader);
1210 	pid_t session = pid_vnr(sid);
1211 	int err = -EPERM;
1212 
1213 	write_lock_irq(&tasklist_lock);
1214 	/* Fail if I am already a session leader */
1215 	if (group_leader->signal->leader)
1216 		goto out;
1217 
1218 	/* Fail if a process group id already exists that equals the
1219 	 * proposed session id.
1220 	 */
1221 	if (pid_task(sid, PIDTYPE_PGID))
1222 		goto out;
1223 
1224 	group_leader->signal->leader = 1;
1225 	set_special_pids(sid);
1226 
1227 	proc_clear_tty(group_leader);
1228 
1229 	err = session;
1230 out:
1231 	write_unlock_irq(&tasklist_lock);
1232 	if (err > 0) {
1233 		proc_sid_connector(group_leader);
1234 		sched_autogroup_create_attach(group_leader);
1235 	}
1236 	return err;
1237 }
1238 
1239 SYSCALL_DEFINE0(setsid)
1240 {
1241 	return ksys_setsid();
1242 }
1243 
1244 DECLARE_RWSEM(uts_sem);
1245 
1246 #ifdef COMPAT_UTS_MACHINE
1247 #define override_architecture(name) \
1248 	(personality(current->personality) == PER_LINUX32 && \
1249 	 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1250 		      sizeof(COMPAT_UTS_MACHINE)))
1251 #else
1252 #define override_architecture(name)	0
1253 #endif
1254 
1255 /*
1256  * Work around broken programs that cannot handle "Linux 3.0".
1257  * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1258  * And we map 4.x and later versions to 2.6.60+x, so 4.0/5.0/6.0/... would be
1259  * 2.6.60.
1260  */
1261 static int override_release(char __user *release, size_t len)
1262 {
1263 	int ret = 0;
1264 
1265 	if (current->personality & UNAME26) {
1266 		const char *rest = UTS_RELEASE;
1267 		char buf[65] = { 0 };
1268 		int ndots = 0;
1269 		unsigned v;
1270 		size_t copy;
1271 
1272 		while (*rest) {
1273 			if (*rest == '.' && ++ndots >= 3)
1274 				break;
1275 			if (!isdigit(*rest) && *rest != '.')
1276 				break;
1277 			rest++;
1278 		}
1279 		v = LINUX_VERSION_PATCHLEVEL + 60;
1280 		copy = clamp_t(size_t, len, 1, sizeof(buf));
1281 		copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1282 		ret = copy_to_user(release, buf, copy + 1);
1283 	}
1284 	return ret;
1285 }
1286 
1287 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1288 {
1289 	struct new_utsname tmp;
1290 
1291 	down_read(&uts_sem);
1292 	memcpy(&tmp, utsname(), sizeof(tmp));
1293 	up_read(&uts_sem);
1294 	if (copy_to_user(name, &tmp, sizeof(tmp)))
1295 		return -EFAULT;
1296 
1297 	if (override_release(name->release, sizeof(name->release)))
1298 		return -EFAULT;
1299 	if (override_architecture(name))
1300 		return -EFAULT;
1301 	return 0;
1302 }
1303 
1304 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1305 /*
1306  * Old cruft
1307  */
1308 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1309 {
1310 	struct old_utsname tmp;
1311 
1312 	if (!name)
1313 		return -EFAULT;
1314 
1315 	down_read(&uts_sem);
1316 	memcpy(&tmp, utsname(), sizeof(tmp));
1317 	up_read(&uts_sem);
1318 	if (copy_to_user(name, &tmp, sizeof(tmp)))
1319 		return -EFAULT;
1320 
1321 	if (override_release(name->release, sizeof(name->release)))
1322 		return -EFAULT;
1323 	if (override_architecture(name))
1324 		return -EFAULT;
1325 	return 0;
1326 }
1327 
1328 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1329 {
1330 	struct oldold_utsname tmp;
1331 
1332 	if (!name)
1333 		return -EFAULT;
1334 
1335 	memset(&tmp, 0, sizeof(tmp));
1336 
1337 	down_read(&uts_sem);
1338 	memcpy(&tmp.sysname, &utsname()->sysname, __OLD_UTS_LEN);
1339 	memcpy(&tmp.nodename, &utsname()->nodename, __OLD_UTS_LEN);
1340 	memcpy(&tmp.release, &utsname()->release, __OLD_UTS_LEN);
1341 	memcpy(&tmp.version, &utsname()->version, __OLD_UTS_LEN);
1342 	memcpy(&tmp.machine, &utsname()->machine, __OLD_UTS_LEN);
1343 	up_read(&uts_sem);
1344 	if (copy_to_user(name, &tmp, sizeof(tmp)))
1345 		return -EFAULT;
1346 
1347 	if (override_architecture(name))
1348 		return -EFAULT;
1349 	if (override_release(name->release, sizeof(name->release)))
1350 		return -EFAULT;
1351 	return 0;
1352 }
1353 #endif
1354 
1355 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1356 {
1357 	int errno;
1358 	char tmp[__NEW_UTS_LEN];
1359 
1360 	if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1361 		return -EPERM;
1362 
1363 	if (len < 0 || len > __NEW_UTS_LEN)
1364 		return -EINVAL;
1365 	errno = -EFAULT;
1366 	if (!copy_from_user(tmp, name, len)) {
1367 		struct new_utsname *u;
1368 
1369 		down_write(&uts_sem);
1370 		u = utsname();
1371 		memcpy(u->nodename, tmp, len);
1372 		memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1373 		errno = 0;
1374 		uts_proc_notify(UTS_PROC_HOSTNAME);
1375 		up_write(&uts_sem);
1376 	}
1377 	return errno;
1378 }
1379 
1380 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1381 
1382 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1383 {
1384 	int i;
1385 	struct new_utsname *u;
1386 	char tmp[__NEW_UTS_LEN + 1];
1387 
1388 	if (len < 0)
1389 		return -EINVAL;
1390 	down_read(&uts_sem);
1391 	u = utsname();
1392 	i = 1 + strlen(u->nodename);
1393 	if (i > len)
1394 		i = len;
1395 	memcpy(tmp, u->nodename, i);
1396 	up_read(&uts_sem);
1397 	if (copy_to_user(name, tmp, i))
1398 		return -EFAULT;
1399 	return 0;
1400 }
1401 
1402 #endif
1403 
1404 /*
1405  * Only setdomainname; getdomainname can be implemented by calling
1406  * uname()
1407  */
1408 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1409 {
1410 	int errno;
1411 	char tmp[__NEW_UTS_LEN];
1412 
1413 	if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1414 		return -EPERM;
1415 	if (len < 0 || len > __NEW_UTS_LEN)
1416 		return -EINVAL;
1417 
1418 	errno = -EFAULT;
1419 	if (!copy_from_user(tmp, name, len)) {
1420 		struct new_utsname *u;
1421 
1422 		down_write(&uts_sem);
1423 		u = utsname();
1424 		memcpy(u->domainname, tmp, len);
1425 		memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1426 		errno = 0;
1427 		uts_proc_notify(UTS_PROC_DOMAINNAME);
1428 		up_write(&uts_sem);
1429 	}
1430 	return errno;
1431 }
1432 
1433 /* make sure you are allowed to change @tsk limits before calling this */
1434 static int do_prlimit(struct task_struct *tsk, unsigned int resource,
1435 		      struct rlimit *new_rlim, struct rlimit *old_rlim)
1436 {
1437 	struct rlimit *rlim;
1438 	int retval = 0;
1439 
1440 	if (resource >= RLIM_NLIMITS)
1441 		return -EINVAL;
1442 	if (new_rlim) {
1443 		if (new_rlim->rlim_cur > new_rlim->rlim_max)
1444 			return -EINVAL;
1445 		if (resource == RLIMIT_NOFILE &&
1446 				new_rlim->rlim_max > sysctl_nr_open)
1447 			return -EPERM;
1448 	}
1449 
1450 	/* Holding a refcount on tsk protects tsk->signal from disappearing. */
1451 	rlim = tsk->signal->rlim + resource;
1452 	task_lock(tsk->group_leader);
1453 	if (new_rlim) {
1454 		/*
1455 		 * Keep the capable check against init_user_ns until cgroups can
1456 		 * contain all limits.
1457 		 */
1458 		if (new_rlim->rlim_max > rlim->rlim_max &&
1459 				!capable(CAP_SYS_RESOURCE))
1460 			retval = -EPERM;
1461 		if (!retval)
1462 			retval = security_task_setrlimit(tsk, resource, new_rlim);
1463 	}
1464 	if (!retval) {
1465 		if (old_rlim)
1466 			*old_rlim = *rlim;
1467 		if (new_rlim)
1468 			*rlim = *new_rlim;
1469 	}
1470 	task_unlock(tsk->group_leader);
1471 
1472 	/*
1473 	 * RLIMIT_CPU handling. Arm the posix CPU timer if the limit is not
1474 	 * infinite. In case of RLIM_INFINITY the posix CPU timer code
1475 	 * ignores the rlimit.
1476 	 */
1477 	if (!retval && new_rlim && resource == RLIMIT_CPU &&
1478 	    new_rlim->rlim_cur != RLIM_INFINITY &&
1479 	    IS_ENABLED(CONFIG_POSIX_TIMERS)) {
1480 		/*
1481 		 * update_rlimit_cpu can fail if the task is exiting, but there
1482 		 * may be other tasks in the thread group that are not exiting,
1483 		 * and they need their cpu timers adjusted.
1484 		 *
1485 		 * The group_leader is the last task to be released, so if we
1486 		 * cannot update_rlimit_cpu on it, then the entire process is
1487 		 * exiting and we do not need to update at all.
1488 		 */
1489 		update_rlimit_cpu(tsk->group_leader, new_rlim->rlim_cur);
1490 	}
1491 
1492 	return retval;
1493 }
1494 
1495 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1496 {
1497 	struct rlimit value;
1498 	int ret;
1499 
1500 	ret = do_prlimit(current, resource, NULL, &value);
1501 	if (!ret)
1502 		ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1503 
1504 	return ret;
1505 }
1506 
1507 #ifdef CONFIG_COMPAT
1508 
1509 COMPAT_SYSCALL_DEFINE2(setrlimit, unsigned int, resource,
1510 		       struct compat_rlimit __user *, rlim)
1511 {
1512 	struct rlimit r;
1513 	struct compat_rlimit r32;
1514 
1515 	if (copy_from_user(&r32, rlim, sizeof(struct compat_rlimit)))
1516 		return -EFAULT;
1517 
1518 	if (r32.rlim_cur == COMPAT_RLIM_INFINITY)
1519 		r.rlim_cur = RLIM_INFINITY;
1520 	else
1521 		r.rlim_cur = r32.rlim_cur;
1522 	if (r32.rlim_max == COMPAT_RLIM_INFINITY)
1523 		r.rlim_max = RLIM_INFINITY;
1524 	else
1525 		r.rlim_max = r32.rlim_max;
1526 	return do_prlimit(current, resource, &r, NULL);
1527 }
1528 
1529 COMPAT_SYSCALL_DEFINE2(getrlimit, unsigned int, resource,
1530 		       struct compat_rlimit __user *, rlim)
1531 {
1532 	struct rlimit r;
1533 	int ret;
1534 
1535 	ret = do_prlimit(current, resource, NULL, &r);
1536 	if (!ret) {
1537 		struct compat_rlimit r32;
1538 		if (r.rlim_cur > COMPAT_RLIM_INFINITY)
1539 			r32.rlim_cur = COMPAT_RLIM_INFINITY;
1540 		else
1541 			r32.rlim_cur = r.rlim_cur;
1542 		if (r.rlim_max > COMPAT_RLIM_INFINITY)
1543 			r32.rlim_max = COMPAT_RLIM_INFINITY;
1544 		else
1545 			r32.rlim_max = r.rlim_max;
1546 
1547 		if (copy_to_user(rlim, &r32, sizeof(struct compat_rlimit)))
1548 			return -EFAULT;
1549 	}
1550 	return ret;
1551 }
1552 
1553 #endif
1554 
1555 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1556 
1557 /*
1558  *	Back compatibility for getrlimit. Needed for some apps.
1559  */
1560 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1561 		struct rlimit __user *, rlim)
1562 {
1563 	struct rlimit x;
1564 	if (resource >= RLIM_NLIMITS)
1565 		return -EINVAL;
1566 
1567 	resource = array_index_nospec(resource, RLIM_NLIMITS);
1568 	task_lock(current->group_leader);
1569 	x = current->signal->rlim[resource];
1570 	task_unlock(current->group_leader);
1571 	if (x.rlim_cur > 0x7FFFFFFF)
1572 		x.rlim_cur = 0x7FFFFFFF;
1573 	if (x.rlim_max > 0x7FFFFFFF)
1574 		x.rlim_max = 0x7FFFFFFF;
1575 	return copy_to_user(rlim, &x, sizeof(x)) ? -EFAULT : 0;
1576 }
1577 
1578 #ifdef CONFIG_COMPAT
1579 COMPAT_SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1580 		       struct compat_rlimit __user *, rlim)
1581 {
1582 	struct rlimit r;
1583 
1584 	if (resource >= RLIM_NLIMITS)
1585 		return -EINVAL;
1586 
1587 	resource = array_index_nospec(resource, RLIM_NLIMITS);
1588 	task_lock(current->group_leader);
1589 	r = current->signal->rlim[resource];
1590 	task_unlock(current->group_leader);
1591 	if (r.rlim_cur > 0x7FFFFFFF)
1592 		r.rlim_cur = 0x7FFFFFFF;
1593 	if (r.rlim_max > 0x7FFFFFFF)
1594 		r.rlim_max = 0x7FFFFFFF;
1595 
1596 	if (put_user(r.rlim_cur, &rlim->rlim_cur) ||
1597 	    put_user(r.rlim_max, &rlim->rlim_max))
1598 		return -EFAULT;
1599 	return 0;
1600 }
1601 #endif
1602 
1603 #endif
1604 
1605 static inline bool rlim64_is_infinity(__u64 rlim64)
1606 {
1607 #if BITS_PER_LONG < 64
1608 	return rlim64 >= ULONG_MAX;
1609 #else
1610 	return rlim64 == RLIM64_INFINITY;
1611 #endif
1612 }
1613 
1614 static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1615 {
1616 	if (rlim->rlim_cur == RLIM_INFINITY)
1617 		rlim64->rlim_cur = RLIM64_INFINITY;
1618 	else
1619 		rlim64->rlim_cur = rlim->rlim_cur;
1620 	if (rlim->rlim_max == RLIM_INFINITY)
1621 		rlim64->rlim_max = RLIM64_INFINITY;
1622 	else
1623 		rlim64->rlim_max = rlim->rlim_max;
1624 }
1625 
1626 static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1627 {
1628 	if (rlim64_is_infinity(rlim64->rlim_cur))
1629 		rlim->rlim_cur = RLIM_INFINITY;
1630 	else
1631 		rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1632 	if (rlim64_is_infinity(rlim64->rlim_max))
1633 		rlim->rlim_max = RLIM_INFINITY;
1634 	else
1635 		rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1636 }
1637 
1638 /* rcu lock must be held */
1639 static int check_prlimit_permission(struct task_struct *task,
1640 				    unsigned int flags)
1641 {
1642 	const struct cred *cred = current_cred(), *tcred;
1643 	bool id_match;
1644 
1645 	if (current == task)
1646 		return 0;
1647 
1648 	tcred = __task_cred(task);
1649 	id_match = (uid_eq(cred->uid, tcred->euid) &&
1650 		    uid_eq(cred->uid, tcred->suid) &&
1651 		    uid_eq(cred->uid, tcred->uid)  &&
1652 		    gid_eq(cred->gid, tcred->egid) &&
1653 		    gid_eq(cred->gid, tcred->sgid) &&
1654 		    gid_eq(cred->gid, tcred->gid));
1655 	if (!id_match && !ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
1656 		return -EPERM;
1657 
1658 	return security_task_prlimit(cred, tcred, flags);
1659 }
1660 
1661 SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1662 		const struct rlimit64 __user *, new_rlim,
1663 		struct rlimit64 __user *, old_rlim)
1664 {
1665 	struct rlimit64 old64, new64;
1666 	struct rlimit old, new;
1667 	struct task_struct *tsk;
1668 	unsigned int checkflags = 0;
1669 	int ret;
1670 
1671 	if (old_rlim)
1672 		checkflags |= LSM_PRLIMIT_READ;
1673 
1674 	if (new_rlim) {
1675 		if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1676 			return -EFAULT;
1677 		rlim64_to_rlim(&new64, &new);
1678 		checkflags |= LSM_PRLIMIT_WRITE;
1679 	}
1680 
1681 	rcu_read_lock();
1682 	tsk = pid ? find_task_by_vpid(pid) : current;
1683 	if (!tsk) {
1684 		rcu_read_unlock();
1685 		return -ESRCH;
1686 	}
1687 	ret = check_prlimit_permission(tsk, checkflags);
1688 	if (ret) {
1689 		rcu_read_unlock();
1690 		return ret;
1691 	}
1692 	get_task_struct(tsk);
1693 	rcu_read_unlock();
1694 
1695 	ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1696 			old_rlim ? &old : NULL);
1697 
1698 	if (!ret && old_rlim) {
1699 		rlim_to_rlim64(&old, &old64);
1700 		if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1701 			ret = -EFAULT;
1702 	}
1703 
1704 	put_task_struct(tsk);
1705 	return ret;
1706 }
1707 
1708 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1709 {
1710 	struct rlimit new_rlim;
1711 
1712 	if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1713 		return -EFAULT;
1714 	return do_prlimit(current, resource, &new_rlim, NULL);
1715 }
1716 
1717 /*
1718  * It would make sense to put struct rusage in the task_struct,
1719  * except that would make the task_struct be *really big*.  After
1720  * task_struct gets moved into malloc'ed memory, it would
1721  * make sense to do this.  It will make moving the rest of the information
1722  * a lot simpler!  (Which we're not doing right now because we're not
1723  * measuring them yet).
1724  *
1725  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1726  * races with threads incrementing their own counters.  But since word
1727  * reads are atomic, we either get new values or old values and we don't
1728  * care which for the sums.  We always take the siglock to protect reading
1729  * the c* fields from p->signal from races with exit.c updating those
1730  * fields when reaping, so a sample either gets all the additions of a
1731  * given child after it's reaped, or none so this sample is before reaping.
1732  *
1733  * Locking:
1734  * We need to take the siglock for CHILDEREN, SELF and BOTH
1735  * for  the cases current multithreaded, non-current single threaded
1736  * non-current multithreaded.  Thread traversal is now safe with
1737  * the siglock held.
1738  * Strictly speaking, we donot need to take the siglock if we are current and
1739  * single threaded,  as no one else can take our signal_struct away, no one
1740  * else can  reap the  children to update signal->c* counters, and no one else
1741  * can race with the signal-> fields. If we do not take any lock, the
1742  * signal-> fields could be read out of order while another thread was just
1743  * exiting. So we should  place a read memory barrier when we avoid the lock.
1744  * On the writer side,  write memory barrier is implied in  __exit_signal
1745  * as __exit_signal releases  the siglock spinlock after updating the signal->
1746  * fields. But we don't do this yet to keep things simple.
1747  *
1748  */
1749 
1750 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1751 {
1752 	r->ru_nvcsw += t->nvcsw;
1753 	r->ru_nivcsw += t->nivcsw;
1754 	r->ru_minflt += t->min_flt;
1755 	r->ru_majflt += t->maj_flt;
1756 	r->ru_inblock += task_io_get_inblock(t);
1757 	r->ru_oublock += task_io_get_oublock(t);
1758 }
1759 
1760 void getrusage(struct task_struct *p, int who, struct rusage *r)
1761 {
1762 	struct task_struct *t;
1763 	unsigned long flags;
1764 	u64 tgutime, tgstime, utime, stime;
1765 	unsigned long maxrss = 0;
1766 
1767 	memset((char *)r, 0, sizeof (*r));
1768 	utime = stime = 0;
1769 
1770 	if (who == RUSAGE_THREAD) {
1771 		task_cputime_adjusted(current, &utime, &stime);
1772 		accumulate_thread_rusage(p, r);
1773 		maxrss = p->signal->maxrss;
1774 		goto out;
1775 	}
1776 
1777 	if (!lock_task_sighand(p, &flags))
1778 		return;
1779 
1780 	switch (who) {
1781 	case RUSAGE_BOTH:
1782 	case RUSAGE_CHILDREN:
1783 		utime = p->signal->cutime;
1784 		stime = p->signal->cstime;
1785 		r->ru_nvcsw = p->signal->cnvcsw;
1786 		r->ru_nivcsw = p->signal->cnivcsw;
1787 		r->ru_minflt = p->signal->cmin_flt;
1788 		r->ru_majflt = p->signal->cmaj_flt;
1789 		r->ru_inblock = p->signal->cinblock;
1790 		r->ru_oublock = p->signal->coublock;
1791 		maxrss = p->signal->cmaxrss;
1792 
1793 		if (who == RUSAGE_CHILDREN)
1794 			break;
1795 		fallthrough;
1796 
1797 	case RUSAGE_SELF:
1798 		thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1799 		utime += tgutime;
1800 		stime += tgstime;
1801 		r->ru_nvcsw += p->signal->nvcsw;
1802 		r->ru_nivcsw += p->signal->nivcsw;
1803 		r->ru_minflt += p->signal->min_flt;
1804 		r->ru_majflt += p->signal->maj_flt;
1805 		r->ru_inblock += p->signal->inblock;
1806 		r->ru_oublock += p->signal->oublock;
1807 		if (maxrss < p->signal->maxrss)
1808 			maxrss = p->signal->maxrss;
1809 		t = p;
1810 		do {
1811 			accumulate_thread_rusage(t, r);
1812 		} while_each_thread(p, t);
1813 		break;
1814 
1815 	default:
1816 		BUG();
1817 	}
1818 	unlock_task_sighand(p, &flags);
1819 
1820 out:
1821 	r->ru_utime = ns_to_kernel_old_timeval(utime);
1822 	r->ru_stime = ns_to_kernel_old_timeval(stime);
1823 
1824 	if (who != RUSAGE_CHILDREN) {
1825 		struct mm_struct *mm = get_task_mm(p);
1826 
1827 		if (mm) {
1828 			setmax_mm_hiwater_rss(&maxrss, mm);
1829 			mmput(mm);
1830 		}
1831 	}
1832 	r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1833 }
1834 
1835 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1836 {
1837 	struct rusage r;
1838 
1839 	if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1840 	    who != RUSAGE_THREAD)
1841 		return -EINVAL;
1842 
1843 	getrusage(current, who, &r);
1844 	return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1845 }
1846 
1847 #ifdef CONFIG_COMPAT
1848 COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru)
1849 {
1850 	struct rusage r;
1851 
1852 	if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1853 	    who != RUSAGE_THREAD)
1854 		return -EINVAL;
1855 
1856 	getrusage(current, who, &r);
1857 	return put_compat_rusage(&r, ru);
1858 }
1859 #endif
1860 
1861 SYSCALL_DEFINE1(umask, int, mask)
1862 {
1863 	mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1864 	return mask;
1865 }
1866 
1867 static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1868 {
1869 	struct fd exe;
1870 	struct inode *inode;
1871 	int err;
1872 
1873 	exe = fdget(fd);
1874 	if (!exe.file)
1875 		return -EBADF;
1876 
1877 	inode = file_inode(exe.file);
1878 
1879 	/*
1880 	 * Because the original mm->exe_file points to executable file, make
1881 	 * sure that this one is executable as well, to avoid breaking an
1882 	 * overall picture.
1883 	 */
1884 	err = -EACCES;
1885 	if (!S_ISREG(inode->i_mode) || path_noexec(&exe.file->f_path))
1886 		goto exit;
1887 
1888 	err = file_permission(exe.file, MAY_EXEC);
1889 	if (err)
1890 		goto exit;
1891 
1892 	err = replace_mm_exe_file(mm, exe.file);
1893 exit:
1894 	fdput(exe);
1895 	return err;
1896 }
1897 
1898 /*
1899  * Check arithmetic relations of passed addresses.
1900  *
1901  * WARNING: we don't require any capability here so be very careful
1902  * in what is allowed for modification from userspace.
1903  */
1904 static int validate_prctl_map_addr(struct prctl_mm_map *prctl_map)
1905 {
1906 	unsigned long mmap_max_addr = TASK_SIZE;
1907 	int error = -EINVAL, i;
1908 
1909 	static const unsigned char offsets[] = {
1910 		offsetof(struct prctl_mm_map, start_code),
1911 		offsetof(struct prctl_mm_map, end_code),
1912 		offsetof(struct prctl_mm_map, start_data),
1913 		offsetof(struct prctl_mm_map, end_data),
1914 		offsetof(struct prctl_mm_map, start_brk),
1915 		offsetof(struct prctl_mm_map, brk),
1916 		offsetof(struct prctl_mm_map, start_stack),
1917 		offsetof(struct prctl_mm_map, arg_start),
1918 		offsetof(struct prctl_mm_map, arg_end),
1919 		offsetof(struct prctl_mm_map, env_start),
1920 		offsetof(struct prctl_mm_map, env_end),
1921 	};
1922 
1923 	/*
1924 	 * Make sure the members are not somewhere outside
1925 	 * of allowed address space.
1926 	 */
1927 	for (i = 0; i < ARRAY_SIZE(offsets); i++) {
1928 		u64 val = *(u64 *)((char *)prctl_map + offsets[i]);
1929 
1930 		if ((unsigned long)val >= mmap_max_addr ||
1931 		    (unsigned long)val < mmap_min_addr)
1932 			goto out;
1933 	}
1934 
1935 	/*
1936 	 * Make sure the pairs are ordered.
1937 	 */
1938 #define __prctl_check_order(__m1, __op, __m2)				\
1939 	((unsigned long)prctl_map->__m1 __op				\
1940 	 (unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1941 	error  = __prctl_check_order(start_code, <, end_code);
1942 	error |= __prctl_check_order(start_data,<=, end_data);
1943 	error |= __prctl_check_order(start_brk, <=, brk);
1944 	error |= __prctl_check_order(arg_start, <=, arg_end);
1945 	error |= __prctl_check_order(env_start, <=, env_end);
1946 	if (error)
1947 		goto out;
1948 #undef __prctl_check_order
1949 
1950 	error = -EINVAL;
1951 
1952 	/*
1953 	 * Neither we should allow to override limits if they set.
1954 	 */
1955 	if (check_data_rlimit(rlimit(RLIMIT_DATA), prctl_map->brk,
1956 			      prctl_map->start_brk, prctl_map->end_data,
1957 			      prctl_map->start_data))
1958 			goto out;
1959 
1960 	error = 0;
1961 out:
1962 	return error;
1963 }
1964 
1965 #ifdef CONFIG_CHECKPOINT_RESTORE
1966 static int prctl_set_mm_map(int opt, const void __user *addr, unsigned long data_size)
1967 {
1968 	struct prctl_mm_map prctl_map = { .exe_fd = (u32)-1, };
1969 	unsigned long user_auxv[AT_VECTOR_SIZE];
1970 	struct mm_struct *mm = current->mm;
1971 	int error;
1972 
1973 	BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
1974 	BUILD_BUG_ON(sizeof(struct prctl_mm_map) > 256);
1975 
1976 	if (opt == PR_SET_MM_MAP_SIZE)
1977 		return put_user((unsigned int)sizeof(prctl_map),
1978 				(unsigned int __user *)addr);
1979 
1980 	if (data_size != sizeof(prctl_map))
1981 		return -EINVAL;
1982 
1983 	if (copy_from_user(&prctl_map, addr, sizeof(prctl_map)))
1984 		return -EFAULT;
1985 
1986 	error = validate_prctl_map_addr(&prctl_map);
1987 	if (error)
1988 		return error;
1989 
1990 	if (prctl_map.auxv_size) {
1991 		/*
1992 		 * Someone is trying to cheat the auxv vector.
1993 		 */
1994 		if (!prctl_map.auxv ||
1995 				prctl_map.auxv_size > sizeof(mm->saved_auxv))
1996 			return -EINVAL;
1997 
1998 		memset(user_auxv, 0, sizeof(user_auxv));
1999 		if (copy_from_user(user_auxv,
2000 				   (const void __user *)prctl_map.auxv,
2001 				   prctl_map.auxv_size))
2002 			return -EFAULT;
2003 
2004 		/* Last entry must be AT_NULL as specification requires */
2005 		user_auxv[AT_VECTOR_SIZE - 2] = AT_NULL;
2006 		user_auxv[AT_VECTOR_SIZE - 1] = AT_NULL;
2007 	}
2008 
2009 	if (prctl_map.exe_fd != (u32)-1) {
2010 		/*
2011 		 * Check if the current user is checkpoint/restore capable.
2012 		 * At the time of this writing, it checks for CAP_SYS_ADMIN
2013 		 * or CAP_CHECKPOINT_RESTORE.
2014 		 * Note that a user with access to ptrace can masquerade an
2015 		 * arbitrary program as any executable, even setuid ones.
2016 		 * This may have implications in the tomoyo subsystem.
2017 		 */
2018 		if (!checkpoint_restore_ns_capable(current_user_ns()))
2019 			return -EPERM;
2020 
2021 		error = prctl_set_mm_exe_file(mm, prctl_map.exe_fd);
2022 		if (error)
2023 			return error;
2024 	}
2025 
2026 	/*
2027 	 * arg_lock protects concurrent updates but we still need mmap_lock for
2028 	 * read to exclude races with sys_brk.
2029 	 */
2030 	mmap_read_lock(mm);
2031 
2032 	/*
2033 	 * We don't validate if these members are pointing to
2034 	 * real present VMAs because application may have correspond
2035 	 * VMAs already unmapped and kernel uses these members for statistics
2036 	 * output in procfs mostly, except
2037 	 *
2038 	 *  - @start_brk/@brk which are used in do_brk_flags but kernel lookups
2039 	 *    for VMAs when updating these members so anything wrong written
2040 	 *    here cause kernel to swear at userspace program but won't lead
2041 	 *    to any problem in kernel itself
2042 	 */
2043 
2044 	spin_lock(&mm->arg_lock);
2045 	mm->start_code	= prctl_map.start_code;
2046 	mm->end_code	= prctl_map.end_code;
2047 	mm->start_data	= prctl_map.start_data;
2048 	mm->end_data	= prctl_map.end_data;
2049 	mm->start_brk	= prctl_map.start_brk;
2050 	mm->brk		= prctl_map.brk;
2051 	mm->start_stack	= prctl_map.start_stack;
2052 	mm->arg_start	= prctl_map.arg_start;
2053 	mm->arg_end	= prctl_map.arg_end;
2054 	mm->env_start	= prctl_map.env_start;
2055 	mm->env_end	= prctl_map.env_end;
2056 	spin_unlock(&mm->arg_lock);
2057 
2058 	/*
2059 	 * Note this update of @saved_auxv is lockless thus
2060 	 * if someone reads this member in procfs while we're
2061 	 * updating -- it may get partly updated results. It's
2062 	 * known and acceptable trade off: we leave it as is to
2063 	 * not introduce additional locks here making the kernel
2064 	 * more complex.
2065 	 */
2066 	if (prctl_map.auxv_size)
2067 		memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv));
2068 
2069 	mmap_read_unlock(mm);
2070 	return 0;
2071 }
2072 #endif /* CONFIG_CHECKPOINT_RESTORE */
2073 
2074 static int prctl_set_auxv(struct mm_struct *mm, unsigned long addr,
2075 			  unsigned long len)
2076 {
2077 	/*
2078 	 * This doesn't move the auxiliary vector itself since it's pinned to
2079 	 * mm_struct, but it permits filling the vector with new values.  It's
2080 	 * up to the caller to provide sane values here, otherwise userspace
2081 	 * tools which use this vector might be unhappy.
2082 	 */
2083 	unsigned long user_auxv[AT_VECTOR_SIZE] = {};
2084 
2085 	if (len > sizeof(user_auxv))
2086 		return -EINVAL;
2087 
2088 	if (copy_from_user(user_auxv, (const void __user *)addr, len))
2089 		return -EFAULT;
2090 
2091 	/* Make sure the last entry is always AT_NULL */
2092 	user_auxv[AT_VECTOR_SIZE - 2] = 0;
2093 	user_auxv[AT_VECTOR_SIZE - 1] = 0;
2094 
2095 	BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
2096 
2097 	task_lock(current);
2098 	memcpy(mm->saved_auxv, user_auxv, len);
2099 	task_unlock(current);
2100 
2101 	return 0;
2102 }
2103 
2104 static int prctl_set_mm(int opt, unsigned long addr,
2105 			unsigned long arg4, unsigned long arg5)
2106 {
2107 	struct mm_struct *mm = current->mm;
2108 	struct prctl_mm_map prctl_map = {
2109 		.auxv = NULL,
2110 		.auxv_size = 0,
2111 		.exe_fd = -1,
2112 	};
2113 	struct vm_area_struct *vma;
2114 	int error;
2115 
2116 	if (arg5 || (arg4 && (opt != PR_SET_MM_AUXV &&
2117 			      opt != PR_SET_MM_MAP &&
2118 			      opt != PR_SET_MM_MAP_SIZE)))
2119 		return -EINVAL;
2120 
2121 #ifdef CONFIG_CHECKPOINT_RESTORE
2122 	if (opt == PR_SET_MM_MAP || opt == PR_SET_MM_MAP_SIZE)
2123 		return prctl_set_mm_map(opt, (const void __user *)addr, arg4);
2124 #endif
2125 
2126 	if (!capable(CAP_SYS_RESOURCE))
2127 		return -EPERM;
2128 
2129 	if (opt == PR_SET_MM_EXE_FILE)
2130 		return prctl_set_mm_exe_file(mm, (unsigned int)addr);
2131 
2132 	if (opt == PR_SET_MM_AUXV)
2133 		return prctl_set_auxv(mm, addr, arg4);
2134 
2135 	if (addr >= TASK_SIZE || addr < mmap_min_addr)
2136 		return -EINVAL;
2137 
2138 	error = -EINVAL;
2139 
2140 	/*
2141 	 * arg_lock protects concurrent updates of arg boundaries, we need
2142 	 * mmap_lock for a) concurrent sys_brk, b) finding VMA for addr
2143 	 * validation.
2144 	 */
2145 	mmap_read_lock(mm);
2146 	vma = find_vma(mm, addr);
2147 
2148 	spin_lock(&mm->arg_lock);
2149 	prctl_map.start_code	= mm->start_code;
2150 	prctl_map.end_code	= mm->end_code;
2151 	prctl_map.start_data	= mm->start_data;
2152 	prctl_map.end_data	= mm->end_data;
2153 	prctl_map.start_brk	= mm->start_brk;
2154 	prctl_map.brk		= mm->brk;
2155 	prctl_map.start_stack	= mm->start_stack;
2156 	prctl_map.arg_start	= mm->arg_start;
2157 	prctl_map.arg_end	= mm->arg_end;
2158 	prctl_map.env_start	= mm->env_start;
2159 	prctl_map.env_end	= mm->env_end;
2160 
2161 	switch (opt) {
2162 	case PR_SET_MM_START_CODE:
2163 		prctl_map.start_code = addr;
2164 		break;
2165 	case PR_SET_MM_END_CODE:
2166 		prctl_map.end_code = addr;
2167 		break;
2168 	case PR_SET_MM_START_DATA:
2169 		prctl_map.start_data = addr;
2170 		break;
2171 	case PR_SET_MM_END_DATA:
2172 		prctl_map.end_data = addr;
2173 		break;
2174 	case PR_SET_MM_START_STACK:
2175 		prctl_map.start_stack = addr;
2176 		break;
2177 	case PR_SET_MM_START_BRK:
2178 		prctl_map.start_brk = addr;
2179 		break;
2180 	case PR_SET_MM_BRK:
2181 		prctl_map.brk = addr;
2182 		break;
2183 	case PR_SET_MM_ARG_START:
2184 		prctl_map.arg_start = addr;
2185 		break;
2186 	case PR_SET_MM_ARG_END:
2187 		prctl_map.arg_end = addr;
2188 		break;
2189 	case PR_SET_MM_ENV_START:
2190 		prctl_map.env_start = addr;
2191 		break;
2192 	case PR_SET_MM_ENV_END:
2193 		prctl_map.env_end = addr;
2194 		break;
2195 	default:
2196 		goto out;
2197 	}
2198 
2199 	error = validate_prctl_map_addr(&prctl_map);
2200 	if (error)
2201 		goto out;
2202 
2203 	switch (opt) {
2204 	/*
2205 	 * If command line arguments and environment
2206 	 * are placed somewhere else on stack, we can
2207 	 * set them up here, ARG_START/END to setup
2208 	 * command line arguments and ENV_START/END
2209 	 * for environment.
2210 	 */
2211 	case PR_SET_MM_START_STACK:
2212 	case PR_SET_MM_ARG_START:
2213 	case PR_SET_MM_ARG_END:
2214 	case PR_SET_MM_ENV_START:
2215 	case PR_SET_MM_ENV_END:
2216 		if (!vma) {
2217 			error = -EFAULT;
2218 			goto out;
2219 		}
2220 	}
2221 
2222 	mm->start_code	= prctl_map.start_code;
2223 	mm->end_code	= prctl_map.end_code;
2224 	mm->start_data	= prctl_map.start_data;
2225 	mm->end_data	= prctl_map.end_data;
2226 	mm->start_brk	= prctl_map.start_brk;
2227 	mm->brk		= prctl_map.brk;
2228 	mm->start_stack	= prctl_map.start_stack;
2229 	mm->arg_start	= prctl_map.arg_start;
2230 	mm->arg_end	= prctl_map.arg_end;
2231 	mm->env_start	= prctl_map.env_start;
2232 	mm->env_end	= prctl_map.env_end;
2233 
2234 	error = 0;
2235 out:
2236 	spin_unlock(&mm->arg_lock);
2237 	mmap_read_unlock(mm);
2238 	return error;
2239 }
2240 
2241 #ifdef CONFIG_CHECKPOINT_RESTORE
2242 static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr)
2243 {
2244 	return put_user(me->clear_child_tid, tid_addr);
2245 }
2246 #else
2247 static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr)
2248 {
2249 	return -EINVAL;
2250 }
2251 #endif
2252 
2253 static int propagate_has_child_subreaper(struct task_struct *p, void *data)
2254 {
2255 	/*
2256 	 * If task has has_child_subreaper - all its descendants
2257 	 * already have these flag too and new descendants will
2258 	 * inherit it on fork, skip them.
2259 	 *
2260 	 * If we've found child_reaper - skip descendants in
2261 	 * it's subtree as they will never get out pidns.
2262 	 */
2263 	if (p->signal->has_child_subreaper ||
2264 	    is_child_reaper(task_pid(p)))
2265 		return 0;
2266 
2267 	p->signal->has_child_subreaper = 1;
2268 	return 1;
2269 }
2270 
2271 int __weak arch_prctl_spec_ctrl_get(struct task_struct *t, unsigned long which)
2272 {
2273 	return -EINVAL;
2274 }
2275 
2276 int __weak arch_prctl_spec_ctrl_set(struct task_struct *t, unsigned long which,
2277 				    unsigned long ctrl)
2278 {
2279 	return -EINVAL;
2280 }
2281 
2282 #define PR_IO_FLUSHER (PF_MEMALLOC_NOIO | PF_LOCAL_THROTTLE)
2283 
2284 #ifdef CONFIG_ANON_VMA_NAME
2285 
2286 #define ANON_VMA_NAME_MAX_LEN		80
2287 #define ANON_VMA_NAME_INVALID_CHARS	"\\`$[]"
2288 
2289 static inline bool is_valid_name_char(char ch)
2290 {
2291 	/* printable ascii characters, excluding ANON_VMA_NAME_INVALID_CHARS */
2292 	return ch > 0x1f && ch < 0x7f &&
2293 		!strchr(ANON_VMA_NAME_INVALID_CHARS, ch);
2294 }
2295 
2296 static int prctl_set_vma(unsigned long opt, unsigned long addr,
2297 			 unsigned long size, unsigned long arg)
2298 {
2299 	struct mm_struct *mm = current->mm;
2300 	const char __user *uname;
2301 	struct anon_vma_name *anon_name = NULL;
2302 	int error;
2303 
2304 	switch (opt) {
2305 	case PR_SET_VMA_ANON_NAME:
2306 		uname = (const char __user *)arg;
2307 		if (uname) {
2308 			char *name, *pch;
2309 
2310 			name = strndup_user(uname, ANON_VMA_NAME_MAX_LEN);
2311 			if (IS_ERR(name))
2312 				return PTR_ERR(name);
2313 
2314 			for (pch = name; *pch != '\0'; pch++) {
2315 				if (!is_valid_name_char(*pch)) {
2316 					kfree(name);
2317 					return -EINVAL;
2318 				}
2319 			}
2320 			/* anon_vma has its own copy */
2321 			anon_name = anon_vma_name_alloc(name);
2322 			kfree(name);
2323 			if (!anon_name)
2324 				return -ENOMEM;
2325 
2326 		}
2327 
2328 		mmap_write_lock(mm);
2329 		error = madvise_set_anon_name(mm, addr, size, anon_name);
2330 		mmap_write_unlock(mm);
2331 		anon_vma_name_put(anon_name);
2332 		break;
2333 	default:
2334 		error = -EINVAL;
2335 	}
2336 
2337 	return error;
2338 }
2339 
2340 #else /* CONFIG_ANON_VMA_NAME */
2341 static int prctl_set_vma(unsigned long opt, unsigned long start,
2342 			 unsigned long size, unsigned long arg)
2343 {
2344 	return -EINVAL;
2345 }
2346 #endif /* CONFIG_ANON_VMA_NAME */
2347 
2348 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2349 		unsigned long, arg4, unsigned long, arg5)
2350 {
2351 	struct task_struct *me = current;
2352 	unsigned char comm[sizeof(me->comm)];
2353 	long error;
2354 
2355 	error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2356 	if (error != -ENOSYS)
2357 		return error;
2358 
2359 	error = 0;
2360 	switch (option) {
2361 	case PR_SET_PDEATHSIG:
2362 		if (!valid_signal(arg2)) {
2363 			error = -EINVAL;
2364 			break;
2365 		}
2366 		me->pdeath_signal = arg2;
2367 		break;
2368 	case PR_GET_PDEATHSIG:
2369 		error = put_user(me->pdeath_signal, (int __user *)arg2);
2370 		break;
2371 	case PR_GET_DUMPABLE:
2372 		error = get_dumpable(me->mm);
2373 		break;
2374 	case PR_SET_DUMPABLE:
2375 		if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) {
2376 			error = -EINVAL;
2377 			break;
2378 		}
2379 		set_dumpable(me->mm, arg2);
2380 		break;
2381 
2382 	case PR_SET_UNALIGN:
2383 		error = SET_UNALIGN_CTL(me, arg2);
2384 		break;
2385 	case PR_GET_UNALIGN:
2386 		error = GET_UNALIGN_CTL(me, arg2);
2387 		break;
2388 	case PR_SET_FPEMU:
2389 		error = SET_FPEMU_CTL(me, arg2);
2390 		break;
2391 	case PR_GET_FPEMU:
2392 		error = GET_FPEMU_CTL(me, arg2);
2393 		break;
2394 	case PR_SET_FPEXC:
2395 		error = SET_FPEXC_CTL(me, arg2);
2396 		break;
2397 	case PR_GET_FPEXC:
2398 		error = GET_FPEXC_CTL(me, arg2);
2399 		break;
2400 	case PR_GET_TIMING:
2401 		error = PR_TIMING_STATISTICAL;
2402 		break;
2403 	case PR_SET_TIMING:
2404 		if (arg2 != PR_TIMING_STATISTICAL)
2405 			error = -EINVAL;
2406 		break;
2407 	case PR_SET_NAME:
2408 		comm[sizeof(me->comm) - 1] = 0;
2409 		if (strncpy_from_user(comm, (char __user *)arg2,
2410 				      sizeof(me->comm) - 1) < 0)
2411 			return -EFAULT;
2412 		set_task_comm(me, comm);
2413 		proc_comm_connector(me);
2414 		break;
2415 	case PR_GET_NAME:
2416 		get_task_comm(comm, me);
2417 		if (copy_to_user((char __user *)arg2, comm, sizeof(comm)))
2418 			return -EFAULT;
2419 		break;
2420 	case PR_GET_ENDIAN:
2421 		error = GET_ENDIAN(me, arg2);
2422 		break;
2423 	case PR_SET_ENDIAN:
2424 		error = SET_ENDIAN(me, arg2);
2425 		break;
2426 	case PR_GET_SECCOMP:
2427 		error = prctl_get_seccomp();
2428 		break;
2429 	case PR_SET_SECCOMP:
2430 		error = prctl_set_seccomp(arg2, (char __user *)arg3);
2431 		break;
2432 	case PR_GET_TSC:
2433 		error = GET_TSC_CTL(arg2);
2434 		break;
2435 	case PR_SET_TSC:
2436 		error = SET_TSC_CTL(arg2);
2437 		break;
2438 	case PR_TASK_PERF_EVENTS_DISABLE:
2439 		error = perf_event_task_disable();
2440 		break;
2441 	case PR_TASK_PERF_EVENTS_ENABLE:
2442 		error = perf_event_task_enable();
2443 		break;
2444 	case PR_GET_TIMERSLACK:
2445 		if (current->timer_slack_ns > ULONG_MAX)
2446 			error = ULONG_MAX;
2447 		else
2448 			error = current->timer_slack_ns;
2449 		break;
2450 	case PR_SET_TIMERSLACK:
2451 		if (arg2 <= 0)
2452 			current->timer_slack_ns =
2453 					current->default_timer_slack_ns;
2454 		else
2455 			current->timer_slack_ns = arg2;
2456 		break;
2457 	case PR_MCE_KILL:
2458 		if (arg4 | arg5)
2459 			return -EINVAL;
2460 		switch (arg2) {
2461 		case PR_MCE_KILL_CLEAR:
2462 			if (arg3 != 0)
2463 				return -EINVAL;
2464 			current->flags &= ~PF_MCE_PROCESS;
2465 			break;
2466 		case PR_MCE_KILL_SET:
2467 			current->flags |= PF_MCE_PROCESS;
2468 			if (arg3 == PR_MCE_KILL_EARLY)
2469 				current->flags |= PF_MCE_EARLY;
2470 			else if (arg3 == PR_MCE_KILL_LATE)
2471 				current->flags &= ~PF_MCE_EARLY;
2472 			else if (arg3 == PR_MCE_KILL_DEFAULT)
2473 				current->flags &=
2474 						~(PF_MCE_EARLY|PF_MCE_PROCESS);
2475 			else
2476 				return -EINVAL;
2477 			break;
2478 		default:
2479 			return -EINVAL;
2480 		}
2481 		break;
2482 	case PR_MCE_KILL_GET:
2483 		if (arg2 | arg3 | arg4 | arg5)
2484 			return -EINVAL;
2485 		if (current->flags & PF_MCE_PROCESS)
2486 			error = (current->flags & PF_MCE_EARLY) ?
2487 				PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2488 		else
2489 			error = PR_MCE_KILL_DEFAULT;
2490 		break;
2491 	case PR_SET_MM:
2492 		error = prctl_set_mm(arg2, arg3, arg4, arg5);
2493 		break;
2494 	case PR_GET_TID_ADDRESS:
2495 		error = prctl_get_tid_address(me, (int __user * __user *)arg2);
2496 		break;
2497 	case PR_SET_CHILD_SUBREAPER:
2498 		me->signal->is_child_subreaper = !!arg2;
2499 		if (!arg2)
2500 			break;
2501 
2502 		walk_process_tree(me, propagate_has_child_subreaper, NULL);
2503 		break;
2504 	case PR_GET_CHILD_SUBREAPER:
2505 		error = put_user(me->signal->is_child_subreaper,
2506 				 (int __user *)arg2);
2507 		break;
2508 	case PR_SET_NO_NEW_PRIVS:
2509 		if (arg2 != 1 || arg3 || arg4 || arg5)
2510 			return -EINVAL;
2511 
2512 		task_set_no_new_privs(current);
2513 		break;
2514 	case PR_GET_NO_NEW_PRIVS:
2515 		if (arg2 || arg3 || arg4 || arg5)
2516 			return -EINVAL;
2517 		return task_no_new_privs(current) ? 1 : 0;
2518 	case PR_GET_THP_DISABLE:
2519 		if (arg2 || arg3 || arg4 || arg5)
2520 			return -EINVAL;
2521 		error = !!test_bit(MMF_DISABLE_THP, &me->mm->flags);
2522 		break;
2523 	case PR_SET_THP_DISABLE:
2524 		if (arg3 || arg4 || arg5)
2525 			return -EINVAL;
2526 		if (mmap_write_lock_killable(me->mm))
2527 			return -EINTR;
2528 		if (arg2)
2529 			set_bit(MMF_DISABLE_THP, &me->mm->flags);
2530 		else
2531 			clear_bit(MMF_DISABLE_THP, &me->mm->flags);
2532 		mmap_write_unlock(me->mm);
2533 		break;
2534 	case PR_MPX_ENABLE_MANAGEMENT:
2535 	case PR_MPX_DISABLE_MANAGEMENT:
2536 		/* No longer implemented: */
2537 		return -EINVAL;
2538 	case PR_SET_FP_MODE:
2539 		error = SET_FP_MODE(me, arg2);
2540 		break;
2541 	case PR_GET_FP_MODE:
2542 		error = GET_FP_MODE(me);
2543 		break;
2544 	case PR_SVE_SET_VL:
2545 		error = SVE_SET_VL(arg2);
2546 		break;
2547 	case PR_SVE_GET_VL:
2548 		error = SVE_GET_VL();
2549 		break;
2550 	case PR_SME_SET_VL:
2551 		error = SME_SET_VL(arg2);
2552 		break;
2553 	case PR_SME_GET_VL:
2554 		error = SME_GET_VL();
2555 		break;
2556 	case PR_GET_SPECULATION_CTRL:
2557 		if (arg3 || arg4 || arg5)
2558 			return -EINVAL;
2559 		error = arch_prctl_spec_ctrl_get(me, arg2);
2560 		break;
2561 	case PR_SET_SPECULATION_CTRL:
2562 		if (arg4 || arg5)
2563 			return -EINVAL;
2564 		error = arch_prctl_spec_ctrl_set(me, arg2, arg3);
2565 		break;
2566 	case PR_PAC_RESET_KEYS:
2567 		if (arg3 || arg4 || arg5)
2568 			return -EINVAL;
2569 		error = PAC_RESET_KEYS(me, arg2);
2570 		break;
2571 	case PR_PAC_SET_ENABLED_KEYS:
2572 		if (arg4 || arg5)
2573 			return -EINVAL;
2574 		error = PAC_SET_ENABLED_KEYS(me, arg2, arg3);
2575 		break;
2576 	case PR_PAC_GET_ENABLED_KEYS:
2577 		if (arg2 || arg3 || arg4 || arg5)
2578 			return -EINVAL;
2579 		error = PAC_GET_ENABLED_KEYS(me);
2580 		break;
2581 	case PR_SET_TAGGED_ADDR_CTRL:
2582 		if (arg3 || arg4 || arg5)
2583 			return -EINVAL;
2584 		error = SET_TAGGED_ADDR_CTRL(arg2);
2585 		break;
2586 	case PR_GET_TAGGED_ADDR_CTRL:
2587 		if (arg2 || arg3 || arg4 || arg5)
2588 			return -EINVAL;
2589 		error = GET_TAGGED_ADDR_CTRL();
2590 		break;
2591 	case PR_SET_IO_FLUSHER:
2592 		if (!capable(CAP_SYS_RESOURCE))
2593 			return -EPERM;
2594 
2595 		if (arg3 || arg4 || arg5)
2596 			return -EINVAL;
2597 
2598 		if (arg2 == 1)
2599 			current->flags |= PR_IO_FLUSHER;
2600 		else if (!arg2)
2601 			current->flags &= ~PR_IO_FLUSHER;
2602 		else
2603 			return -EINVAL;
2604 		break;
2605 	case PR_GET_IO_FLUSHER:
2606 		if (!capable(CAP_SYS_RESOURCE))
2607 			return -EPERM;
2608 
2609 		if (arg2 || arg3 || arg4 || arg5)
2610 			return -EINVAL;
2611 
2612 		error = (current->flags & PR_IO_FLUSHER) == PR_IO_FLUSHER;
2613 		break;
2614 	case PR_SET_SYSCALL_USER_DISPATCH:
2615 		error = set_syscall_user_dispatch(arg2, arg3, arg4,
2616 						  (char __user *) arg5);
2617 		break;
2618 #ifdef CONFIG_SCHED_CORE
2619 	case PR_SCHED_CORE:
2620 		error = sched_core_share_pid(arg2, arg3, arg4, arg5);
2621 		break;
2622 #endif
2623 	case PR_SET_VMA:
2624 		error = prctl_set_vma(arg2, arg3, arg4, arg5);
2625 		break;
2626 	default:
2627 		error = -EINVAL;
2628 		break;
2629 	}
2630 	return error;
2631 }
2632 
2633 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2634 		struct getcpu_cache __user *, unused)
2635 {
2636 	int err = 0;
2637 	int cpu = raw_smp_processor_id();
2638 
2639 	if (cpup)
2640 		err |= put_user(cpu, cpup);
2641 	if (nodep)
2642 		err |= put_user(cpu_to_node(cpu), nodep);
2643 	return err ? -EFAULT : 0;
2644 }
2645 
2646 /**
2647  * do_sysinfo - fill in sysinfo struct
2648  * @info: pointer to buffer to fill
2649  */
2650 static int do_sysinfo(struct sysinfo *info)
2651 {
2652 	unsigned long mem_total, sav_total;
2653 	unsigned int mem_unit, bitcount;
2654 	struct timespec64 tp;
2655 
2656 	memset(info, 0, sizeof(struct sysinfo));
2657 
2658 	ktime_get_boottime_ts64(&tp);
2659 	timens_add_boottime(&tp);
2660 	info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
2661 
2662 	get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
2663 
2664 	info->procs = nr_threads;
2665 
2666 	si_meminfo(info);
2667 	si_swapinfo(info);
2668 
2669 	/*
2670 	 * If the sum of all the available memory (i.e. ram + swap)
2671 	 * is less than can be stored in a 32 bit unsigned long then
2672 	 * we can be binary compatible with 2.2.x kernels.  If not,
2673 	 * well, in that case 2.2.x was broken anyways...
2674 	 *
2675 	 *  -Erik Andersen <andersee@debian.org>
2676 	 */
2677 
2678 	mem_total = info->totalram + info->totalswap;
2679 	if (mem_total < info->totalram || mem_total < info->totalswap)
2680 		goto out;
2681 	bitcount = 0;
2682 	mem_unit = info->mem_unit;
2683 	while (mem_unit > 1) {
2684 		bitcount++;
2685 		mem_unit >>= 1;
2686 		sav_total = mem_total;
2687 		mem_total <<= 1;
2688 		if (mem_total < sav_total)
2689 			goto out;
2690 	}
2691 
2692 	/*
2693 	 * If mem_total did not overflow, multiply all memory values by
2694 	 * info->mem_unit and set it to 1.  This leaves things compatible
2695 	 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2696 	 * kernels...
2697 	 */
2698 
2699 	info->mem_unit = 1;
2700 	info->totalram <<= bitcount;
2701 	info->freeram <<= bitcount;
2702 	info->sharedram <<= bitcount;
2703 	info->bufferram <<= bitcount;
2704 	info->totalswap <<= bitcount;
2705 	info->freeswap <<= bitcount;
2706 	info->totalhigh <<= bitcount;
2707 	info->freehigh <<= bitcount;
2708 
2709 out:
2710 	return 0;
2711 }
2712 
2713 SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
2714 {
2715 	struct sysinfo val;
2716 
2717 	do_sysinfo(&val);
2718 
2719 	if (copy_to_user(info, &val, sizeof(struct sysinfo)))
2720 		return -EFAULT;
2721 
2722 	return 0;
2723 }
2724 
2725 #ifdef CONFIG_COMPAT
2726 struct compat_sysinfo {
2727 	s32 uptime;
2728 	u32 loads[3];
2729 	u32 totalram;
2730 	u32 freeram;
2731 	u32 sharedram;
2732 	u32 bufferram;
2733 	u32 totalswap;
2734 	u32 freeswap;
2735 	u16 procs;
2736 	u16 pad;
2737 	u32 totalhigh;
2738 	u32 freehigh;
2739 	u32 mem_unit;
2740 	char _f[20-2*sizeof(u32)-sizeof(int)];
2741 };
2742 
2743 COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2744 {
2745 	struct sysinfo s;
2746 	struct compat_sysinfo s_32;
2747 
2748 	do_sysinfo(&s);
2749 
2750 	/* Check to see if any memory value is too large for 32-bit and scale
2751 	 *  down if needed
2752 	 */
2753 	if (upper_32_bits(s.totalram) || upper_32_bits(s.totalswap)) {
2754 		int bitcount = 0;
2755 
2756 		while (s.mem_unit < PAGE_SIZE) {
2757 			s.mem_unit <<= 1;
2758 			bitcount++;
2759 		}
2760 
2761 		s.totalram >>= bitcount;
2762 		s.freeram >>= bitcount;
2763 		s.sharedram >>= bitcount;
2764 		s.bufferram >>= bitcount;
2765 		s.totalswap >>= bitcount;
2766 		s.freeswap >>= bitcount;
2767 		s.totalhigh >>= bitcount;
2768 		s.freehigh >>= bitcount;
2769 	}
2770 
2771 	memset(&s_32, 0, sizeof(s_32));
2772 	s_32.uptime = s.uptime;
2773 	s_32.loads[0] = s.loads[0];
2774 	s_32.loads[1] = s.loads[1];
2775 	s_32.loads[2] = s.loads[2];
2776 	s_32.totalram = s.totalram;
2777 	s_32.freeram = s.freeram;
2778 	s_32.sharedram = s.sharedram;
2779 	s_32.bufferram = s.bufferram;
2780 	s_32.totalswap = s.totalswap;
2781 	s_32.freeswap = s.freeswap;
2782 	s_32.procs = s.procs;
2783 	s_32.totalhigh = s.totalhigh;
2784 	s_32.freehigh = s.freehigh;
2785 	s_32.mem_unit = s.mem_unit;
2786 	if (copy_to_user(info, &s_32, sizeof(s_32)))
2787 		return -EFAULT;
2788 	return 0;
2789 }
2790 #endif /* CONFIG_COMPAT */
2791