xref: /linux/kernel/sys.c (revision aaa25a2fa7964d94690f6de5edd7164ca7d76555)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  *  linux/kernel/sys.c
4  *
5  *  Copyright (C) 1991, 1992  Linus Torvalds
6  */
7 
8 #include <linux/export.h>
9 #include <linux/mm.h>
10 #include <linux/utsname.h>
11 #include <linux/mman.h>
12 #include <linux/reboot.h>
13 #include <linux/prctl.h>
14 #include <linux/highuid.h>
15 #include <linux/fs.h>
16 #include <linux/kmod.h>
17 #include <linux/perf_event.h>
18 #include <linux/resource.h>
19 #include <linux/kernel.h>
20 #include <linux/workqueue.h>
21 #include <linux/capability.h>
22 #include <linux/device.h>
23 #include <linux/key.h>
24 #include <linux/times.h>
25 #include <linux/posix-timers.h>
26 #include <linux/security.h>
27 #include <linux/suspend.h>
28 #include <linux/tty.h>
29 #include <linux/signal.h>
30 #include <linux/cn_proc.h>
31 #include <linux/getcpu.h>
32 #include <linux/task_io_accounting_ops.h>
33 #include <linux/seccomp.h>
34 #include <linux/cpu.h>
35 #include <linux/personality.h>
36 #include <linux/ptrace.h>
37 #include <linux/fs_struct.h>
38 #include <linux/file.h>
39 #include <linux/mount.h>
40 #include <linux/gfp.h>
41 #include <linux/syscore_ops.h>
42 #include <linux/version.h>
43 #include <linux/ctype.h>
44 #include <linux/syscall_user_dispatch.h>
45 
46 #include <linux/compat.h>
47 #include <linux/syscalls.h>
48 #include <linux/kprobes.h>
49 #include <linux/user_namespace.h>
50 #include <linux/time_namespace.h>
51 #include <linux/binfmts.h>
52 
53 #include <linux/sched.h>
54 #include <linux/sched/autogroup.h>
55 #include <linux/sched/loadavg.h>
56 #include <linux/sched/stat.h>
57 #include <linux/sched/mm.h>
58 #include <linux/sched/coredump.h>
59 #include <linux/sched/task.h>
60 #include <linux/sched/cputime.h>
61 #include <linux/rcupdate.h>
62 #include <linux/uidgid.h>
63 #include <linux/cred.h>
64 
65 #include <linux/nospec.h>
66 
67 #include <linux/kmsg_dump.h>
68 /* Move somewhere else to avoid recompiling? */
69 #include <generated/utsrelease.h>
70 
71 #include <linux/uaccess.h>
72 #include <asm/io.h>
73 #include <asm/unistd.h>
74 
75 #include "uid16.h"
76 
77 #ifndef SET_UNALIGN_CTL
78 # define SET_UNALIGN_CTL(a, b)	(-EINVAL)
79 #endif
80 #ifndef GET_UNALIGN_CTL
81 # define GET_UNALIGN_CTL(a, b)	(-EINVAL)
82 #endif
83 #ifndef SET_FPEMU_CTL
84 # define SET_FPEMU_CTL(a, b)	(-EINVAL)
85 #endif
86 #ifndef GET_FPEMU_CTL
87 # define GET_FPEMU_CTL(a, b)	(-EINVAL)
88 #endif
89 #ifndef SET_FPEXC_CTL
90 # define SET_FPEXC_CTL(a, b)	(-EINVAL)
91 #endif
92 #ifndef GET_FPEXC_CTL
93 # define GET_FPEXC_CTL(a, b)	(-EINVAL)
94 #endif
95 #ifndef GET_ENDIAN
96 # define GET_ENDIAN(a, b)	(-EINVAL)
97 #endif
98 #ifndef SET_ENDIAN
99 # define SET_ENDIAN(a, b)	(-EINVAL)
100 #endif
101 #ifndef GET_TSC_CTL
102 # define GET_TSC_CTL(a)		(-EINVAL)
103 #endif
104 #ifndef SET_TSC_CTL
105 # define SET_TSC_CTL(a)		(-EINVAL)
106 #endif
107 #ifndef GET_FP_MODE
108 # define GET_FP_MODE(a)		(-EINVAL)
109 #endif
110 #ifndef SET_FP_MODE
111 # define SET_FP_MODE(a,b)	(-EINVAL)
112 #endif
113 #ifndef SVE_SET_VL
114 # define SVE_SET_VL(a)		(-EINVAL)
115 #endif
116 #ifndef SVE_GET_VL
117 # define SVE_GET_VL()		(-EINVAL)
118 #endif
119 #ifndef PAC_RESET_KEYS
120 # define PAC_RESET_KEYS(a, b)	(-EINVAL)
121 #endif
122 #ifndef PAC_SET_ENABLED_KEYS
123 # define PAC_SET_ENABLED_KEYS(a, b, c)	(-EINVAL)
124 #endif
125 #ifndef PAC_GET_ENABLED_KEYS
126 # define PAC_GET_ENABLED_KEYS(a)	(-EINVAL)
127 #endif
128 #ifndef SET_TAGGED_ADDR_CTRL
129 # define SET_TAGGED_ADDR_CTRL(a)	(-EINVAL)
130 #endif
131 #ifndef GET_TAGGED_ADDR_CTRL
132 # define GET_TAGGED_ADDR_CTRL()		(-EINVAL)
133 #endif
134 
135 /*
136  * this is where the system-wide overflow UID and GID are defined, for
137  * architectures that now have 32-bit UID/GID but didn't in the past
138  */
139 
140 int overflowuid = DEFAULT_OVERFLOWUID;
141 int overflowgid = DEFAULT_OVERFLOWGID;
142 
143 EXPORT_SYMBOL(overflowuid);
144 EXPORT_SYMBOL(overflowgid);
145 
146 /*
147  * the same as above, but for filesystems which can only store a 16-bit
148  * UID and GID. as such, this is needed on all architectures
149  */
150 
151 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
152 int fs_overflowgid = DEFAULT_FS_OVERFLOWGID;
153 
154 EXPORT_SYMBOL(fs_overflowuid);
155 EXPORT_SYMBOL(fs_overflowgid);
156 
157 /*
158  * Returns true if current's euid is same as p's uid or euid,
159  * or has CAP_SYS_NICE to p's user_ns.
160  *
161  * Called with rcu_read_lock, creds are safe
162  */
163 static bool set_one_prio_perm(struct task_struct *p)
164 {
165 	const struct cred *cred = current_cred(), *pcred = __task_cred(p);
166 
167 	if (uid_eq(pcred->uid,  cred->euid) ||
168 	    uid_eq(pcred->euid, cred->euid))
169 		return true;
170 	if (ns_capable(pcred->user_ns, CAP_SYS_NICE))
171 		return true;
172 	return false;
173 }
174 
175 /*
176  * set the priority of a task
177  * - the caller must hold the RCU read lock
178  */
179 static int set_one_prio(struct task_struct *p, int niceval, int error)
180 {
181 	int no_nice;
182 
183 	if (!set_one_prio_perm(p)) {
184 		error = -EPERM;
185 		goto out;
186 	}
187 	if (niceval < task_nice(p) && !can_nice(p, niceval)) {
188 		error = -EACCES;
189 		goto out;
190 	}
191 	no_nice = security_task_setnice(p, niceval);
192 	if (no_nice) {
193 		error = no_nice;
194 		goto out;
195 	}
196 	if (error == -ESRCH)
197 		error = 0;
198 	set_user_nice(p, niceval);
199 out:
200 	return error;
201 }
202 
203 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
204 {
205 	struct task_struct *g, *p;
206 	struct user_struct *user;
207 	const struct cred *cred = current_cred();
208 	int error = -EINVAL;
209 	struct pid *pgrp;
210 	kuid_t uid;
211 
212 	if (which > PRIO_USER || which < PRIO_PROCESS)
213 		goto out;
214 
215 	/* normalize: avoid signed division (rounding problems) */
216 	error = -ESRCH;
217 	if (niceval < MIN_NICE)
218 		niceval = MIN_NICE;
219 	if (niceval > MAX_NICE)
220 		niceval = MAX_NICE;
221 
222 	rcu_read_lock();
223 	switch (which) {
224 	case PRIO_PROCESS:
225 		if (who)
226 			p = find_task_by_vpid(who);
227 		else
228 			p = current;
229 		if (p)
230 			error = set_one_prio(p, niceval, error);
231 		break;
232 	case PRIO_PGRP:
233 		if (who)
234 			pgrp = find_vpid(who);
235 		else
236 			pgrp = task_pgrp(current);
237 		read_lock(&tasklist_lock);
238 		do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
239 			error = set_one_prio(p, niceval, error);
240 		} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
241 		read_unlock(&tasklist_lock);
242 		break;
243 	case PRIO_USER:
244 		uid = make_kuid(cred->user_ns, who);
245 		user = cred->user;
246 		if (!who)
247 			uid = cred->uid;
248 		else if (!uid_eq(uid, cred->uid)) {
249 			user = find_user(uid);
250 			if (!user)
251 				goto out_unlock;	/* No processes for this user */
252 		}
253 		for_each_process_thread(g, p) {
254 			if (uid_eq(task_uid(p), uid) && task_pid_vnr(p))
255 				error = set_one_prio(p, niceval, error);
256 		}
257 		if (!uid_eq(uid, cred->uid))
258 			free_uid(user);		/* For find_user() */
259 		break;
260 	}
261 out_unlock:
262 	rcu_read_unlock();
263 out:
264 	return error;
265 }
266 
267 /*
268  * Ugh. To avoid negative return values, "getpriority()" will
269  * not return the normal nice-value, but a negated value that
270  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
271  * to stay compatible.
272  */
273 SYSCALL_DEFINE2(getpriority, int, which, int, who)
274 {
275 	struct task_struct *g, *p;
276 	struct user_struct *user;
277 	const struct cred *cred = current_cred();
278 	long niceval, retval = -ESRCH;
279 	struct pid *pgrp;
280 	kuid_t uid;
281 
282 	if (which > PRIO_USER || which < PRIO_PROCESS)
283 		return -EINVAL;
284 
285 	rcu_read_lock();
286 	switch (which) {
287 	case PRIO_PROCESS:
288 		if (who)
289 			p = find_task_by_vpid(who);
290 		else
291 			p = current;
292 		if (p) {
293 			niceval = nice_to_rlimit(task_nice(p));
294 			if (niceval > retval)
295 				retval = niceval;
296 		}
297 		break;
298 	case PRIO_PGRP:
299 		if (who)
300 			pgrp = find_vpid(who);
301 		else
302 			pgrp = task_pgrp(current);
303 		read_lock(&tasklist_lock);
304 		do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
305 			niceval = nice_to_rlimit(task_nice(p));
306 			if (niceval > retval)
307 				retval = niceval;
308 		} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
309 		read_unlock(&tasklist_lock);
310 		break;
311 	case PRIO_USER:
312 		uid = make_kuid(cred->user_ns, who);
313 		user = cred->user;
314 		if (!who)
315 			uid = cred->uid;
316 		else if (!uid_eq(uid, cred->uid)) {
317 			user = find_user(uid);
318 			if (!user)
319 				goto out_unlock;	/* No processes for this user */
320 		}
321 		for_each_process_thread(g, p) {
322 			if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) {
323 				niceval = nice_to_rlimit(task_nice(p));
324 				if (niceval > retval)
325 					retval = niceval;
326 			}
327 		}
328 		if (!uid_eq(uid, cred->uid))
329 			free_uid(user);		/* for find_user() */
330 		break;
331 	}
332 out_unlock:
333 	rcu_read_unlock();
334 
335 	return retval;
336 }
337 
338 /*
339  * Unprivileged users may change the real gid to the effective gid
340  * or vice versa.  (BSD-style)
341  *
342  * If you set the real gid at all, or set the effective gid to a value not
343  * equal to the real gid, then the saved gid is set to the new effective gid.
344  *
345  * This makes it possible for a setgid program to completely drop its
346  * privileges, which is often a useful assertion to make when you are doing
347  * a security audit over a program.
348  *
349  * The general idea is that a program which uses just setregid() will be
350  * 100% compatible with BSD.  A program which uses just setgid() will be
351  * 100% compatible with POSIX with saved IDs.
352  *
353  * SMP: There are not races, the GIDs are checked only by filesystem
354  *      operations (as far as semantic preservation is concerned).
355  */
356 #ifdef CONFIG_MULTIUSER
357 long __sys_setregid(gid_t rgid, gid_t egid)
358 {
359 	struct user_namespace *ns = current_user_ns();
360 	const struct cred *old;
361 	struct cred *new;
362 	int retval;
363 	kgid_t krgid, kegid;
364 
365 	krgid = make_kgid(ns, rgid);
366 	kegid = make_kgid(ns, egid);
367 
368 	if ((rgid != (gid_t) -1) && !gid_valid(krgid))
369 		return -EINVAL;
370 	if ((egid != (gid_t) -1) && !gid_valid(kegid))
371 		return -EINVAL;
372 
373 	new = prepare_creds();
374 	if (!new)
375 		return -ENOMEM;
376 	old = current_cred();
377 
378 	retval = -EPERM;
379 	if (rgid != (gid_t) -1) {
380 		if (gid_eq(old->gid, krgid) ||
381 		    gid_eq(old->egid, krgid) ||
382 		    ns_capable_setid(old->user_ns, CAP_SETGID))
383 			new->gid = krgid;
384 		else
385 			goto error;
386 	}
387 	if (egid != (gid_t) -1) {
388 		if (gid_eq(old->gid, kegid) ||
389 		    gid_eq(old->egid, kegid) ||
390 		    gid_eq(old->sgid, kegid) ||
391 		    ns_capable_setid(old->user_ns, CAP_SETGID))
392 			new->egid = kegid;
393 		else
394 			goto error;
395 	}
396 
397 	if (rgid != (gid_t) -1 ||
398 	    (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
399 		new->sgid = new->egid;
400 	new->fsgid = new->egid;
401 
402 	retval = security_task_fix_setgid(new, old, LSM_SETID_RE);
403 	if (retval < 0)
404 		goto error;
405 
406 	return commit_creds(new);
407 
408 error:
409 	abort_creds(new);
410 	return retval;
411 }
412 
413 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
414 {
415 	return __sys_setregid(rgid, egid);
416 }
417 
418 /*
419  * setgid() is implemented like SysV w/ SAVED_IDS
420  *
421  * SMP: Same implicit races as above.
422  */
423 long __sys_setgid(gid_t gid)
424 {
425 	struct user_namespace *ns = current_user_ns();
426 	const struct cred *old;
427 	struct cred *new;
428 	int retval;
429 	kgid_t kgid;
430 
431 	kgid = make_kgid(ns, gid);
432 	if (!gid_valid(kgid))
433 		return -EINVAL;
434 
435 	new = prepare_creds();
436 	if (!new)
437 		return -ENOMEM;
438 	old = current_cred();
439 
440 	retval = -EPERM;
441 	if (ns_capable_setid(old->user_ns, CAP_SETGID))
442 		new->gid = new->egid = new->sgid = new->fsgid = kgid;
443 	else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
444 		new->egid = new->fsgid = kgid;
445 	else
446 		goto error;
447 
448 	retval = security_task_fix_setgid(new, old, LSM_SETID_ID);
449 	if (retval < 0)
450 		goto error;
451 
452 	return commit_creds(new);
453 
454 error:
455 	abort_creds(new);
456 	return retval;
457 }
458 
459 SYSCALL_DEFINE1(setgid, gid_t, gid)
460 {
461 	return __sys_setgid(gid);
462 }
463 
464 /*
465  * change the user struct in a credentials set to match the new UID
466  */
467 static int set_user(struct cred *new)
468 {
469 	struct user_struct *new_user;
470 
471 	new_user = alloc_uid(new->uid);
472 	if (!new_user)
473 		return -EAGAIN;
474 
475 	free_uid(new->user);
476 	new->user = new_user;
477 	return 0;
478 }
479 
480 static void flag_nproc_exceeded(struct cred *new)
481 {
482 	if (new->ucounts == current_ucounts())
483 		return;
484 
485 	/*
486 	 * We don't fail in case of NPROC limit excess here because too many
487 	 * poorly written programs don't check set*uid() return code, assuming
488 	 * it never fails if called by root.  We may still enforce NPROC limit
489 	 * for programs doing set*uid()+execve() by harmlessly deferring the
490 	 * failure to the execve() stage.
491 	 */
492 	if (is_ucounts_overlimit(new->ucounts, UCOUNT_RLIMIT_NPROC, rlimit(RLIMIT_NPROC)) &&
493 			new->user != INIT_USER)
494 		current->flags |= PF_NPROC_EXCEEDED;
495 	else
496 		current->flags &= ~PF_NPROC_EXCEEDED;
497 }
498 
499 /*
500  * Unprivileged users may change the real uid to the effective uid
501  * or vice versa.  (BSD-style)
502  *
503  * If you set the real uid at all, or set the effective uid to a value not
504  * equal to the real uid, then the saved uid is set to the new effective uid.
505  *
506  * This makes it possible for a setuid program to completely drop its
507  * privileges, which is often a useful assertion to make when you are doing
508  * a security audit over a program.
509  *
510  * The general idea is that a program which uses just setreuid() will be
511  * 100% compatible with BSD.  A program which uses just setuid() will be
512  * 100% compatible with POSIX with saved IDs.
513  */
514 long __sys_setreuid(uid_t ruid, uid_t euid)
515 {
516 	struct user_namespace *ns = current_user_ns();
517 	const struct cred *old;
518 	struct cred *new;
519 	int retval;
520 	kuid_t kruid, keuid;
521 
522 	kruid = make_kuid(ns, ruid);
523 	keuid = make_kuid(ns, euid);
524 
525 	if ((ruid != (uid_t) -1) && !uid_valid(kruid))
526 		return -EINVAL;
527 	if ((euid != (uid_t) -1) && !uid_valid(keuid))
528 		return -EINVAL;
529 
530 	new = prepare_creds();
531 	if (!new)
532 		return -ENOMEM;
533 	old = current_cred();
534 
535 	retval = -EPERM;
536 	if (ruid != (uid_t) -1) {
537 		new->uid = kruid;
538 		if (!uid_eq(old->uid, kruid) &&
539 		    !uid_eq(old->euid, kruid) &&
540 		    !ns_capable_setid(old->user_ns, CAP_SETUID))
541 			goto error;
542 	}
543 
544 	if (euid != (uid_t) -1) {
545 		new->euid = keuid;
546 		if (!uid_eq(old->uid, keuid) &&
547 		    !uid_eq(old->euid, keuid) &&
548 		    !uid_eq(old->suid, keuid) &&
549 		    !ns_capable_setid(old->user_ns, CAP_SETUID))
550 			goto error;
551 	}
552 
553 	if (!uid_eq(new->uid, old->uid)) {
554 		retval = set_user(new);
555 		if (retval < 0)
556 			goto error;
557 	}
558 	if (ruid != (uid_t) -1 ||
559 	    (euid != (uid_t) -1 && !uid_eq(keuid, old->uid)))
560 		new->suid = new->euid;
561 	new->fsuid = new->euid;
562 
563 	retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
564 	if (retval < 0)
565 		goto error;
566 
567 	retval = set_cred_ucounts(new);
568 	if (retval < 0)
569 		goto error;
570 
571 	flag_nproc_exceeded(new);
572 	return commit_creds(new);
573 
574 error:
575 	abort_creds(new);
576 	return retval;
577 }
578 
579 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
580 {
581 	return __sys_setreuid(ruid, euid);
582 }
583 
584 /*
585  * setuid() is implemented like SysV with SAVED_IDS
586  *
587  * Note that SAVED_ID's is deficient in that a setuid root program
588  * like sendmail, for example, cannot set its uid to be a normal
589  * user and then switch back, because if you're root, setuid() sets
590  * the saved uid too.  If you don't like this, blame the bright people
591  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
592  * will allow a root program to temporarily drop privileges and be able to
593  * regain them by swapping the real and effective uid.
594  */
595 long __sys_setuid(uid_t uid)
596 {
597 	struct user_namespace *ns = current_user_ns();
598 	const struct cred *old;
599 	struct cred *new;
600 	int retval;
601 	kuid_t kuid;
602 
603 	kuid = make_kuid(ns, uid);
604 	if (!uid_valid(kuid))
605 		return -EINVAL;
606 
607 	new = prepare_creds();
608 	if (!new)
609 		return -ENOMEM;
610 	old = current_cred();
611 
612 	retval = -EPERM;
613 	if (ns_capable_setid(old->user_ns, CAP_SETUID)) {
614 		new->suid = new->uid = kuid;
615 		if (!uid_eq(kuid, old->uid)) {
616 			retval = set_user(new);
617 			if (retval < 0)
618 				goto error;
619 		}
620 	} else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) {
621 		goto error;
622 	}
623 
624 	new->fsuid = new->euid = kuid;
625 
626 	retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
627 	if (retval < 0)
628 		goto error;
629 
630 	retval = set_cred_ucounts(new);
631 	if (retval < 0)
632 		goto error;
633 
634 	flag_nproc_exceeded(new);
635 	return commit_creds(new);
636 
637 error:
638 	abort_creds(new);
639 	return retval;
640 }
641 
642 SYSCALL_DEFINE1(setuid, uid_t, uid)
643 {
644 	return __sys_setuid(uid);
645 }
646 
647 
648 /*
649  * This function implements a generic ability to update ruid, euid,
650  * and suid.  This allows you to implement the 4.4 compatible seteuid().
651  */
652 long __sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
653 {
654 	struct user_namespace *ns = current_user_ns();
655 	const struct cred *old;
656 	struct cred *new;
657 	int retval;
658 	kuid_t kruid, keuid, ksuid;
659 
660 	kruid = make_kuid(ns, ruid);
661 	keuid = make_kuid(ns, euid);
662 	ksuid = make_kuid(ns, suid);
663 
664 	if ((ruid != (uid_t) -1) && !uid_valid(kruid))
665 		return -EINVAL;
666 
667 	if ((euid != (uid_t) -1) && !uid_valid(keuid))
668 		return -EINVAL;
669 
670 	if ((suid != (uid_t) -1) && !uid_valid(ksuid))
671 		return -EINVAL;
672 
673 	new = prepare_creds();
674 	if (!new)
675 		return -ENOMEM;
676 
677 	old = current_cred();
678 
679 	retval = -EPERM;
680 	if (!ns_capable_setid(old->user_ns, CAP_SETUID)) {
681 		if (ruid != (uid_t) -1        && !uid_eq(kruid, old->uid) &&
682 		    !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid))
683 			goto error;
684 		if (euid != (uid_t) -1        && !uid_eq(keuid, old->uid) &&
685 		    !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid))
686 			goto error;
687 		if (suid != (uid_t) -1        && !uid_eq(ksuid, old->uid) &&
688 		    !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid))
689 			goto error;
690 	}
691 
692 	if (ruid != (uid_t) -1) {
693 		new->uid = kruid;
694 		if (!uid_eq(kruid, old->uid)) {
695 			retval = set_user(new);
696 			if (retval < 0)
697 				goto error;
698 		}
699 	}
700 	if (euid != (uid_t) -1)
701 		new->euid = keuid;
702 	if (suid != (uid_t) -1)
703 		new->suid = ksuid;
704 	new->fsuid = new->euid;
705 
706 	retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
707 	if (retval < 0)
708 		goto error;
709 
710 	retval = set_cred_ucounts(new);
711 	if (retval < 0)
712 		goto error;
713 
714 	flag_nproc_exceeded(new);
715 	return commit_creds(new);
716 
717 error:
718 	abort_creds(new);
719 	return retval;
720 }
721 
722 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
723 {
724 	return __sys_setresuid(ruid, euid, suid);
725 }
726 
727 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
728 {
729 	const struct cred *cred = current_cred();
730 	int retval;
731 	uid_t ruid, euid, suid;
732 
733 	ruid = from_kuid_munged(cred->user_ns, cred->uid);
734 	euid = from_kuid_munged(cred->user_ns, cred->euid);
735 	suid = from_kuid_munged(cred->user_ns, cred->suid);
736 
737 	retval = put_user(ruid, ruidp);
738 	if (!retval) {
739 		retval = put_user(euid, euidp);
740 		if (!retval)
741 			return put_user(suid, suidp);
742 	}
743 	return retval;
744 }
745 
746 /*
747  * Same as above, but for rgid, egid, sgid.
748  */
749 long __sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
750 {
751 	struct user_namespace *ns = current_user_ns();
752 	const struct cred *old;
753 	struct cred *new;
754 	int retval;
755 	kgid_t krgid, kegid, ksgid;
756 
757 	krgid = make_kgid(ns, rgid);
758 	kegid = make_kgid(ns, egid);
759 	ksgid = make_kgid(ns, sgid);
760 
761 	if ((rgid != (gid_t) -1) && !gid_valid(krgid))
762 		return -EINVAL;
763 	if ((egid != (gid_t) -1) && !gid_valid(kegid))
764 		return -EINVAL;
765 	if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
766 		return -EINVAL;
767 
768 	new = prepare_creds();
769 	if (!new)
770 		return -ENOMEM;
771 	old = current_cred();
772 
773 	retval = -EPERM;
774 	if (!ns_capable_setid(old->user_ns, CAP_SETGID)) {
775 		if (rgid != (gid_t) -1        && !gid_eq(krgid, old->gid) &&
776 		    !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid))
777 			goto error;
778 		if (egid != (gid_t) -1        && !gid_eq(kegid, old->gid) &&
779 		    !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid))
780 			goto error;
781 		if (sgid != (gid_t) -1        && !gid_eq(ksgid, old->gid) &&
782 		    !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid))
783 			goto error;
784 	}
785 
786 	if (rgid != (gid_t) -1)
787 		new->gid = krgid;
788 	if (egid != (gid_t) -1)
789 		new->egid = kegid;
790 	if (sgid != (gid_t) -1)
791 		new->sgid = ksgid;
792 	new->fsgid = new->egid;
793 
794 	retval = security_task_fix_setgid(new, old, LSM_SETID_RES);
795 	if (retval < 0)
796 		goto error;
797 
798 	return commit_creds(new);
799 
800 error:
801 	abort_creds(new);
802 	return retval;
803 }
804 
805 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
806 {
807 	return __sys_setresgid(rgid, egid, sgid);
808 }
809 
810 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
811 {
812 	const struct cred *cred = current_cred();
813 	int retval;
814 	gid_t rgid, egid, sgid;
815 
816 	rgid = from_kgid_munged(cred->user_ns, cred->gid);
817 	egid = from_kgid_munged(cred->user_ns, cred->egid);
818 	sgid = from_kgid_munged(cred->user_ns, cred->sgid);
819 
820 	retval = put_user(rgid, rgidp);
821 	if (!retval) {
822 		retval = put_user(egid, egidp);
823 		if (!retval)
824 			retval = put_user(sgid, sgidp);
825 	}
826 
827 	return retval;
828 }
829 
830 
831 /*
832  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
833  * is used for "access()" and for the NFS daemon (letting nfsd stay at
834  * whatever uid it wants to). It normally shadows "euid", except when
835  * explicitly set by setfsuid() or for access..
836  */
837 long __sys_setfsuid(uid_t uid)
838 {
839 	const struct cred *old;
840 	struct cred *new;
841 	uid_t old_fsuid;
842 	kuid_t kuid;
843 
844 	old = current_cred();
845 	old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
846 
847 	kuid = make_kuid(old->user_ns, uid);
848 	if (!uid_valid(kuid))
849 		return old_fsuid;
850 
851 	new = prepare_creds();
852 	if (!new)
853 		return old_fsuid;
854 
855 	if (uid_eq(kuid, old->uid)  || uid_eq(kuid, old->euid)  ||
856 	    uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
857 	    ns_capable_setid(old->user_ns, CAP_SETUID)) {
858 		if (!uid_eq(kuid, old->fsuid)) {
859 			new->fsuid = kuid;
860 			if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
861 				goto change_okay;
862 		}
863 	}
864 
865 	abort_creds(new);
866 	return old_fsuid;
867 
868 change_okay:
869 	commit_creds(new);
870 	return old_fsuid;
871 }
872 
873 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
874 {
875 	return __sys_setfsuid(uid);
876 }
877 
878 /*
879  * Samma på svenska..
880  */
881 long __sys_setfsgid(gid_t gid)
882 {
883 	const struct cred *old;
884 	struct cred *new;
885 	gid_t old_fsgid;
886 	kgid_t kgid;
887 
888 	old = current_cred();
889 	old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
890 
891 	kgid = make_kgid(old->user_ns, gid);
892 	if (!gid_valid(kgid))
893 		return old_fsgid;
894 
895 	new = prepare_creds();
896 	if (!new)
897 		return old_fsgid;
898 
899 	if (gid_eq(kgid, old->gid)  || gid_eq(kgid, old->egid)  ||
900 	    gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
901 	    ns_capable_setid(old->user_ns, CAP_SETGID)) {
902 		if (!gid_eq(kgid, old->fsgid)) {
903 			new->fsgid = kgid;
904 			if (security_task_fix_setgid(new,old,LSM_SETID_FS) == 0)
905 				goto change_okay;
906 		}
907 	}
908 
909 	abort_creds(new);
910 	return old_fsgid;
911 
912 change_okay:
913 	commit_creds(new);
914 	return old_fsgid;
915 }
916 
917 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
918 {
919 	return __sys_setfsgid(gid);
920 }
921 #endif /* CONFIG_MULTIUSER */
922 
923 /**
924  * sys_getpid - return the thread group id of the current process
925  *
926  * Note, despite the name, this returns the tgid not the pid.  The tgid and
927  * the pid are identical unless CLONE_THREAD was specified on clone() in
928  * which case the tgid is the same in all threads of the same group.
929  *
930  * This is SMP safe as current->tgid does not change.
931  */
932 SYSCALL_DEFINE0(getpid)
933 {
934 	return task_tgid_vnr(current);
935 }
936 
937 /* Thread ID - the internal kernel "pid" */
938 SYSCALL_DEFINE0(gettid)
939 {
940 	return task_pid_vnr(current);
941 }
942 
943 /*
944  * Accessing ->real_parent is not SMP-safe, it could
945  * change from under us. However, we can use a stale
946  * value of ->real_parent under rcu_read_lock(), see
947  * release_task()->call_rcu(delayed_put_task_struct).
948  */
949 SYSCALL_DEFINE0(getppid)
950 {
951 	int pid;
952 
953 	rcu_read_lock();
954 	pid = task_tgid_vnr(rcu_dereference(current->real_parent));
955 	rcu_read_unlock();
956 
957 	return pid;
958 }
959 
960 SYSCALL_DEFINE0(getuid)
961 {
962 	/* Only we change this so SMP safe */
963 	return from_kuid_munged(current_user_ns(), current_uid());
964 }
965 
966 SYSCALL_DEFINE0(geteuid)
967 {
968 	/* Only we change this so SMP safe */
969 	return from_kuid_munged(current_user_ns(), current_euid());
970 }
971 
972 SYSCALL_DEFINE0(getgid)
973 {
974 	/* Only we change this so SMP safe */
975 	return from_kgid_munged(current_user_ns(), current_gid());
976 }
977 
978 SYSCALL_DEFINE0(getegid)
979 {
980 	/* Only we change this so SMP safe */
981 	return from_kgid_munged(current_user_ns(), current_egid());
982 }
983 
984 static void do_sys_times(struct tms *tms)
985 {
986 	u64 tgutime, tgstime, cutime, cstime;
987 
988 	thread_group_cputime_adjusted(current, &tgutime, &tgstime);
989 	cutime = current->signal->cutime;
990 	cstime = current->signal->cstime;
991 	tms->tms_utime = nsec_to_clock_t(tgutime);
992 	tms->tms_stime = nsec_to_clock_t(tgstime);
993 	tms->tms_cutime = nsec_to_clock_t(cutime);
994 	tms->tms_cstime = nsec_to_clock_t(cstime);
995 }
996 
997 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
998 {
999 	if (tbuf) {
1000 		struct tms tmp;
1001 
1002 		do_sys_times(&tmp);
1003 		if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1004 			return -EFAULT;
1005 	}
1006 	force_successful_syscall_return();
1007 	return (long) jiffies_64_to_clock_t(get_jiffies_64());
1008 }
1009 
1010 #ifdef CONFIG_COMPAT
1011 static compat_clock_t clock_t_to_compat_clock_t(clock_t x)
1012 {
1013 	return compat_jiffies_to_clock_t(clock_t_to_jiffies(x));
1014 }
1015 
1016 COMPAT_SYSCALL_DEFINE1(times, struct compat_tms __user *, tbuf)
1017 {
1018 	if (tbuf) {
1019 		struct tms tms;
1020 		struct compat_tms tmp;
1021 
1022 		do_sys_times(&tms);
1023 		/* Convert our struct tms to the compat version. */
1024 		tmp.tms_utime = clock_t_to_compat_clock_t(tms.tms_utime);
1025 		tmp.tms_stime = clock_t_to_compat_clock_t(tms.tms_stime);
1026 		tmp.tms_cutime = clock_t_to_compat_clock_t(tms.tms_cutime);
1027 		tmp.tms_cstime = clock_t_to_compat_clock_t(tms.tms_cstime);
1028 		if (copy_to_user(tbuf, &tmp, sizeof(tmp)))
1029 			return -EFAULT;
1030 	}
1031 	force_successful_syscall_return();
1032 	return compat_jiffies_to_clock_t(jiffies);
1033 }
1034 #endif
1035 
1036 /*
1037  * This needs some heavy checking ...
1038  * I just haven't the stomach for it. I also don't fully
1039  * understand sessions/pgrp etc. Let somebody who does explain it.
1040  *
1041  * OK, I think I have the protection semantics right.... this is really
1042  * only important on a multi-user system anyway, to make sure one user
1043  * can't send a signal to a process owned by another.  -TYT, 12/12/91
1044  *
1045  * !PF_FORKNOEXEC check to conform completely to POSIX.
1046  */
1047 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
1048 {
1049 	struct task_struct *p;
1050 	struct task_struct *group_leader = current->group_leader;
1051 	struct pid *pgrp;
1052 	int err;
1053 
1054 	if (!pid)
1055 		pid = task_pid_vnr(group_leader);
1056 	if (!pgid)
1057 		pgid = pid;
1058 	if (pgid < 0)
1059 		return -EINVAL;
1060 	rcu_read_lock();
1061 
1062 	/* From this point forward we keep holding onto the tasklist lock
1063 	 * so that our parent does not change from under us. -DaveM
1064 	 */
1065 	write_lock_irq(&tasklist_lock);
1066 
1067 	err = -ESRCH;
1068 	p = find_task_by_vpid(pid);
1069 	if (!p)
1070 		goto out;
1071 
1072 	err = -EINVAL;
1073 	if (!thread_group_leader(p))
1074 		goto out;
1075 
1076 	if (same_thread_group(p->real_parent, group_leader)) {
1077 		err = -EPERM;
1078 		if (task_session(p) != task_session(group_leader))
1079 			goto out;
1080 		err = -EACCES;
1081 		if (!(p->flags & PF_FORKNOEXEC))
1082 			goto out;
1083 	} else {
1084 		err = -ESRCH;
1085 		if (p != group_leader)
1086 			goto out;
1087 	}
1088 
1089 	err = -EPERM;
1090 	if (p->signal->leader)
1091 		goto out;
1092 
1093 	pgrp = task_pid(p);
1094 	if (pgid != pid) {
1095 		struct task_struct *g;
1096 
1097 		pgrp = find_vpid(pgid);
1098 		g = pid_task(pgrp, PIDTYPE_PGID);
1099 		if (!g || task_session(g) != task_session(group_leader))
1100 			goto out;
1101 	}
1102 
1103 	err = security_task_setpgid(p, pgid);
1104 	if (err)
1105 		goto out;
1106 
1107 	if (task_pgrp(p) != pgrp)
1108 		change_pid(p, PIDTYPE_PGID, pgrp);
1109 
1110 	err = 0;
1111 out:
1112 	/* All paths lead to here, thus we are safe. -DaveM */
1113 	write_unlock_irq(&tasklist_lock);
1114 	rcu_read_unlock();
1115 	return err;
1116 }
1117 
1118 static int do_getpgid(pid_t pid)
1119 {
1120 	struct task_struct *p;
1121 	struct pid *grp;
1122 	int retval;
1123 
1124 	rcu_read_lock();
1125 	if (!pid)
1126 		grp = task_pgrp(current);
1127 	else {
1128 		retval = -ESRCH;
1129 		p = find_task_by_vpid(pid);
1130 		if (!p)
1131 			goto out;
1132 		grp = task_pgrp(p);
1133 		if (!grp)
1134 			goto out;
1135 
1136 		retval = security_task_getpgid(p);
1137 		if (retval)
1138 			goto out;
1139 	}
1140 	retval = pid_vnr(grp);
1141 out:
1142 	rcu_read_unlock();
1143 	return retval;
1144 }
1145 
1146 SYSCALL_DEFINE1(getpgid, pid_t, pid)
1147 {
1148 	return do_getpgid(pid);
1149 }
1150 
1151 #ifdef __ARCH_WANT_SYS_GETPGRP
1152 
1153 SYSCALL_DEFINE0(getpgrp)
1154 {
1155 	return do_getpgid(0);
1156 }
1157 
1158 #endif
1159 
1160 SYSCALL_DEFINE1(getsid, pid_t, pid)
1161 {
1162 	struct task_struct *p;
1163 	struct pid *sid;
1164 	int retval;
1165 
1166 	rcu_read_lock();
1167 	if (!pid)
1168 		sid = task_session(current);
1169 	else {
1170 		retval = -ESRCH;
1171 		p = find_task_by_vpid(pid);
1172 		if (!p)
1173 			goto out;
1174 		sid = task_session(p);
1175 		if (!sid)
1176 			goto out;
1177 
1178 		retval = security_task_getsid(p);
1179 		if (retval)
1180 			goto out;
1181 	}
1182 	retval = pid_vnr(sid);
1183 out:
1184 	rcu_read_unlock();
1185 	return retval;
1186 }
1187 
1188 static void set_special_pids(struct pid *pid)
1189 {
1190 	struct task_struct *curr = current->group_leader;
1191 
1192 	if (task_session(curr) != pid)
1193 		change_pid(curr, PIDTYPE_SID, pid);
1194 
1195 	if (task_pgrp(curr) != pid)
1196 		change_pid(curr, PIDTYPE_PGID, pid);
1197 }
1198 
1199 int ksys_setsid(void)
1200 {
1201 	struct task_struct *group_leader = current->group_leader;
1202 	struct pid *sid = task_pid(group_leader);
1203 	pid_t session = pid_vnr(sid);
1204 	int err = -EPERM;
1205 
1206 	write_lock_irq(&tasklist_lock);
1207 	/* Fail if I am already a session leader */
1208 	if (group_leader->signal->leader)
1209 		goto out;
1210 
1211 	/* Fail if a process group id already exists that equals the
1212 	 * proposed session id.
1213 	 */
1214 	if (pid_task(sid, PIDTYPE_PGID))
1215 		goto out;
1216 
1217 	group_leader->signal->leader = 1;
1218 	set_special_pids(sid);
1219 
1220 	proc_clear_tty(group_leader);
1221 
1222 	err = session;
1223 out:
1224 	write_unlock_irq(&tasklist_lock);
1225 	if (err > 0) {
1226 		proc_sid_connector(group_leader);
1227 		sched_autogroup_create_attach(group_leader);
1228 	}
1229 	return err;
1230 }
1231 
1232 SYSCALL_DEFINE0(setsid)
1233 {
1234 	return ksys_setsid();
1235 }
1236 
1237 DECLARE_RWSEM(uts_sem);
1238 
1239 #ifdef COMPAT_UTS_MACHINE
1240 #define override_architecture(name) \
1241 	(personality(current->personality) == PER_LINUX32 && \
1242 	 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1243 		      sizeof(COMPAT_UTS_MACHINE)))
1244 #else
1245 #define override_architecture(name)	0
1246 #endif
1247 
1248 /*
1249  * Work around broken programs that cannot handle "Linux 3.0".
1250  * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1251  * And we map 4.x and later versions to 2.6.60+x, so 4.0/5.0/6.0/... would be
1252  * 2.6.60.
1253  */
1254 static int override_release(char __user *release, size_t len)
1255 {
1256 	int ret = 0;
1257 
1258 	if (current->personality & UNAME26) {
1259 		const char *rest = UTS_RELEASE;
1260 		char buf[65] = { 0 };
1261 		int ndots = 0;
1262 		unsigned v;
1263 		size_t copy;
1264 
1265 		while (*rest) {
1266 			if (*rest == '.' && ++ndots >= 3)
1267 				break;
1268 			if (!isdigit(*rest) && *rest != '.')
1269 				break;
1270 			rest++;
1271 		}
1272 		v = LINUX_VERSION_PATCHLEVEL + 60;
1273 		copy = clamp_t(size_t, len, 1, sizeof(buf));
1274 		copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1275 		ret = copy_to_user(release, buf, copy + 1);
1276 	}
1277 	return ret;
1278 }
1279 
1280 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1281 {
1282 	struct new_utsname tmp;
1283 
1284 	down_read(&uts_sem);
1285 	memcpy(&tmp, utsname(), sizeof(tmp));
1286 	up_read(&uts_sem);
1287 	if (copy_to_user(name, &tmp, sizeof(tmp)))
1288 		return -EFAULT;
1289 
1290 	if (override_release(name->release, sizeof(name->release)))
1291 		return -EFAULT;
1292 	if (override_architecture(name))
1293 		return -EFAULT;
1294 	return 0;
1295 }
1296 
1297 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1298 /*
1299  * Old cruft
1300  */
1301 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1302 {
1303 	struct old_utsname tmp;
1304 
1305 	if (!name)
1306 		return -EFAULT;
1307 
1308 	down_read(&uts_sem);
1309 	memcpy(&tmp, utsname(), sizeof(tmp));
1310 	up_read(&uts_sem);
1311 	if (copy_to_user(name, &tmp, sizeof(tmp)))
1312 		return -EFAULT;
1313 
1314 	if (override_release(name->release, sizeof(name->release)))
1315 		return -EFAULT;
1316 	if (override_architecture(name))
1317 		return -EFAULT;
1318 	return 0;
1319 }
1320 
1321 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1322 {
1323 	struct oldold_utsname tmp;
1324 
1325 	if (!name)
1326 		return -EFAULT;
1327 
1328 	memset(&tmp, 0, sizeof(tmp));
1329 
1330 	down_read(&uts_sem);
1331 	memcpy(&tmp.sysname, &utsname()->sysname, __OLD_UTS_LEN);
1332 	memcpy(&tmp.nodename, &utsname()->nodename, __OLD_UTS_LEN);
1333 	memcpy(&tmp.release, &utsname()->release, __OLD_UTS_LEN);
1334 	memcpy(&tmp.version, &utsname()->version, __OLD_UTS_LEN);
1335 	memcpy(&tmp.machine, &utsname()->machine, __OLD_UTS_LEN);
1336 	up_read(&uts_sem);
1337 	if (copy_to_user(name, &tmp, sizeof(tmp)))
1338 		return -EFAULT;
1339 
1340 	if (override_architecture(name))
1341 		return -EFAULT;
1342 	if (override_release(name->release, sizeof(name->release)))
1343 		return -EFAULT;
1344 	return 0;
1345 }
1346 #endif
1347 
1348 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1349 {
1350 	int errno;
1351 	char tmp[__NEW_UTS_LEN];
1352 
1353 	if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1354 		return -EPERM;
1355 
1356 	if (len < 0 || len > __NEW_UTS_LEN)
1357 		return -EINVAL;
1358 	errno = -EFAULT;
1359 	if (!copy_from_user(tmp, name, len)) {
1360 		struct new_utsname *u;
1361 
1362 		down_write(&uts_sem);
1363 		u = utsname();
1364 		memcpy(u->nodename, tmp, len);
1365 		memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1366 		errno = 0;
1367 		uts_proc_notify(UTS_PROC_HOSTNAME);
1368 		up_write(&uts_sem);
1369 	}
1370 	return errno;
1371 }
1372 
1373 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1374 
1375 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1376 {
1377 	int i;
1378 	struct new_utsname *u;
1379 	char tmp[__NEW_UTS_LEN + 1];
1380 
1381 	if (len < 0)
1382 		return -EINVAL;
1383 	down_read(&uts_sem);
1384 	u = utsname();
1385 	i = 1 + strlen(u->nodename);
1386 	if (i > len)
1387 		i = len;
1388 	memcpy(tmp, u->nodename, i);
1389 	up_read(&uts_sem);
1390 	if (copy_to_user(name, tmp, i))
1391 		return -EFAULT;
1392 	return 0;
1393 }
1394 
1395 #endif
1396 
1397 /*
1398  * Only setdomainname; getdomainname can be implemented by calling
1399  * uname()
1400  */
1401 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1402 {
1403 	int errno;
1404 	char tmp[__NEW_UTS_LEN];
1405 
1406 	if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1407 		return -EPERM;
1408 	if (len < 0 || len > __NEW_UTS_LEN)
1409 		return -EINVAL;
1410 
1411 	errno = -EFAULT;
1412 	if (!copy_from_user(tmp, name, len)) {
1413 		struct new_utsname *u;
1414 
1415 		down_write(&uts_sem);
1416 		u = utsname();
1417 		memcpy(u->domainname, tmp, len);
1418 		memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1419 		errno = 0;
1420 		uts_proc_notify(UTS_PROC_DOMAINNAME);
1421 		up_write(&uts_sem);
1422 	}
1423 	return errno;
1424 }
1425 
1426 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1427 {
1428 	struct rlimit value;
1429 	int ret;
1430 
1431 	ret = do_prlimit(current, resource, NULL, &value);
1432 	if (!ret)
1433 		ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1434 
1435 	return ret;
1436 }
1437 
1438 #ifdef CONFIG_COMPAT
1439 
1440 COMPAT_SYSCALL_DEFINE2(setrlimit, unsigned int, resource,
1441 		       struct compat_rlimit __user *, rlim)
1442 {
1443 	struct rlimit r;
1444 	struct compat_rlimit r32;
1445 
1446 	if (copy_from_user(&r32, rlim, sizeof(struct compat_rlimit)))
1447 		return -EFAULT;
1448 
1449 	if (r32.rlim_cur == COMPAT_RLIM_INFINITY)
1450 		r.rlim_cur = RLIM_INFINITY;
1451 	else
1452 		r.rlim_cur = r32.rlim_cur;
1453 	if (r32.rlim_max == COMPAT_RLIM_INFINITY)
1454 		r.rlim_max = RLIM_INFINITY;
1455 	else
1456 		r.rlim_max = r32.rlim_max;
1457 	return do_prlimit(current, resource, &r, NULL);
1458 }
1459 
1460 COMPAT_SYSCALL_DEFINE2(getrlimit, unsigned int, resource,
1461 		       struct compat_rlimit __user *, rlim)
1462 {
1463 	struct rlimit r;
1464 	int ret;
1465 
1466 	ret = do_prlimit(current, resource, NULL, &r);
1467 	if (!ret) {
1468 		struct compat_rlimit r32;
1469 		if (r.rlim_cur > COMPAT_RLIM_INFINITY)
1470 			r32.rlim_cur = COMPAT_RLIM_INFINITY;
1471 		else
1472 			r32.rlim_cur = r.rlim_cur;
1473 		if (r.rlim_max > COMPAT_RLIM_INFINITY)
1474 			r32.rlim_max = COMPAT_RLIM_INFINITY;
1475 		else
1476 			r32.rlim_max = r.rlim_max;
1477 
1478 		if (copy_to_user(rlim, &r32, sizeof(struct compat_rlimit)))
1479 			return -EFAULT;
1480 	}
1481 	return ret;
1482 }
1483 
1484 #endif
1485 
1486 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1487 
1488 /*
1489  *	Back compatibility for getrlimit. Needed for some apps.
1490  */
1491 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1492 		struct rlimit __user *, rlim)
1493 {
1494 	struct rlimit x;
1495 	if (resource >= RLIM_NLIMITS)
1496 		return -EINVAL;
1497 
1498 	resource = array_index_nospec(resource, RLIM_NLIMITS);
1499 	task_lock(current->group_leader);
1500 	x = current->signal->rlim[resource];
1501 	task_unlock(current->group_leader);
1502 	if (x.rlim_cur > 0x7FFFFFFF)
1503 		x.rlim_cur = 0x7FFFFFFF;
1504 	if (x.rlim_max > 0x7FFFFFFF)
1505 		x.rlim_max = 0x7FFFFFFF;
1506 	return copy_to_user(rlim, &x, sizeof(x)) ? -EFAULT : 0;
1507 }
1508 
1509 #ifdef CONFIG_COMPAT
1510 COMPAT_SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1511 		       struct compat_rlimit __user *, rlim)
1512 {
1513 	struct rlimit r;
1514 
1515 	if (resource >= RLIM_NLIMITS)
1516 		return -EINVAL;
1517 
1518 	resource = array_index_nospec(resource, RLIM_NLIMITS);
1519 	task_lock(current->group_leader);
1520 	r = current->signal->rlim[resource];
1521 	task_unlock(current->group_leader);
1522 	if (r.rlim_cur > 0x7FFFFFFF)
1523 		r.rlim_cur = 0x7FFFFFFF;
1524 	if (r.rlim_max > 0x7FFFFFFF)
1525 		r.rlim_max = 0x7FFFFFFF;
1526 
1527 	if (put_user(r.rlim_cur, &rlim->rlim_cur) ||
1528 	    put_user(r.rlim_max, &rlim->rlim_max))
1529 		return -EFAULT;
1530 	return 0;
1531 }
1532 #endif
1533 
1534 #endif
1535 
1536 static inline bool rlim64_is_infinity(__u64 rlim64)
1537 {
1538 #if BITS_PER_LONG < 64
1539 	return rlim64 >= ULONG_MAX;
1540 #else
1541 	return rlim64 == RLIM64_INFINITY;
1542 #endif
1543 }
1544 
1545 static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1546 {
1547 	if (rlim->rlim_cur == RLIM_INFINITY)
1548 		rlim64->rlim_cur = RLIM64_INFINITY;
1549 	else
1550 		rlim64->rlim_cur = rlim->rlim_cur;
1551 	if (rlim->rlim_max == RLIM_INFINITY)
1552 		rlim64->rlim_max = RLIM64_INFINITY;
1553 	else
1554 		rlim64->rlim_max = rlim->rlim_max;
1555 }
1556 
1557 static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1558 {
1559 	if (rlim64_is_infinity(rlim64->rlim_cur))
1560 		rlim->rlim_cur = RLIM_INFINITY;
1561 	else
1562 		rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1563 	if (rlim64_is_infinity(rlim64->rlim_max))
1564 		rlim->rlim_max = RLIM_INFINITY;
1565 	else
1566 		rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1567 }
1568 
1569 /* make sure you are allowed to change @tsk limits before calling this */
1570 int do_prlimit(struct task_struct *tsk, unsigned int resource,
1571 		struct rlimit *new_rlim, struct rlimit *old_rlim)
1572 {
1573 	struct rlimit *rlim;
1574 	int retval = 0;
1575 
1576 	if (resource >= RLIM_NLIMITS)
1577 		return -EINVAL;
1578 	if (new_rlim) {
1579 		if (new_rlim->rlim_cur > new_rlim->rlim_max)
1580 			return -EINVAL;
1581 		if (resource == RLIMIT_NOFILE &&
1582 				new_rlim->rlim_max > sysctl_nr_open)
1583 			return -EPERM;
1584 	}
1585 
1586 	/* protect tsk->signal and tsk->sighand from disappearing */
1587 	read_lock(&tasklist_lock);
1588 	if (!tsk->sighand) {
1589 		retval = -ESRCH;
1590 		goto out;
1591 	}
1592 
1593 	rlim = tsk->signal->rlim + resource;
1594 	task_lock(tsk->group_leader);
1595 	if (new_rlim) {
1596 		/* Keep the capable check against init_user_ns until
1597 		   cgroups can contain all limits */
1598 		if (new_rlim->rlim_max > rlim->rlim_max &&
1599 				!capable(CAP_SYS_RESOURCE))
1600 			retval = -EPERM;
1601 		if (!retval)
1602 			retval = security_task_setrlimit(tsk, resource, new_rlim);
1603 	}
1604 	if (!retval) {
1605 		if (old_rlim)
1606 			*old_rlim = *rlim;
1607 		if (new_rlim)
1608 			*rlim = *new_rlim;
1609 	}
1610 	task_unlock(tsk->group_leader);
1611 
1612 	/*
1613 	 * RLIMIT_CPU handling. Arm the posix CPU timer if the limit is not
1614 	 * infinite. In case of RLIM_INFINITY the posix CPU timer code
1615 	 * ignores the rlimit.
1616 	 */
1617 	 if (!retval && new_rlim && resource == RLIMIT_CPU &&
1618 	     new_rlim->rlim_cur != RLIM_INFINITY &&
1619 	     IS_ENABLED(CONFIG_POSIX_TIMERS))
1620 		update_rlimit_cpu(tsk, new_rlim->rlim_cur);
1621 out:
1622 	read_unlock(&tasklist_lock);
1623 	return retval;
1624 }
1625 
1626 /* rcu lock must be held */
1627 static int check_prlimit_permission(struct task_struct *task,
1628 				    unsigned int flags)
1629 {
1630 	const struct cred *cred = current_cred(), *tcred;
1631 	bool id_match;
1632 
1633 	if (current == task)
1634 		return 0;
1635 
1636 	tcred = __task_cred(task);
1637 	id_match = (uid_eq(cred->uid, tcred->euid) &&
1638 		    uid_eq(cred->uid, tcred->suid) &&
1639 		    uid_eq(cred->uid, tcred->uid)  &&
1640 		    gid_eq(cred->gid, tcred->egid) &&
1641 		    gid_eq(cred->gid, tcred->sgid) &&
1642 		    gid_eq(cred->gid, tcred->gid));
1643 	if (!id_match && !ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
1644 		return -EPERM;
1645 
1646 	return security_task_prlimit(cred, tcred, flags);
1647 }
1648 
1649 SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1650 		const struct rlimit64 __user *, new_rlim,
1651 		struct rlimit64 __user *, old_rlim)
1652 {
1653 	struct rlimit64 old64, new64;
1654 	struct rlimit old, new;
1655 	struct task_struct *tsk;
1656 	unsigned int checkflags = 0;
1657 	int ret;
1658 
1659 	if (old_rlim)
1660 		checkflags |= LSM_PRLIMIT_READ;
1661 
1662 	if (new_rlim) {
1663 		if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1664 			return -EFAULT;
1665 		rlim64_to_rlim(&new64, &new);
1666 		checkflags |= LSM_PRLIMIT_WRITE;
1667 	}
1668 
1669 	rcu_read_lock();
1670 	tsk = pid ? find_task_by_vpid(pid) : current;
1671 	if (!tsk) {
1672 		rcu_read_unlock();
1673 		return -ESRCH;
1674 	}
1675 	ret = check_prlimit_permission(tsk, checkflags);
1676 	if (ret) {
1677 		rcu_read_unlock();
1678 		return ret;
1679 	}
1680 	get_task_struct(tsk);
1681 	rcu_read_unlock();
1682 
1683 	ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1684 			old_rlim ? &old : NULL);
1685 
1686 	if (!ret && old_rlim) {
1687 		rlim_to_rlim64(&old, &old64);
1688 		if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1689 			ret = -EFAULT;
1690 	}
1691 
1692 	put_task_struct(tsk);
1693 	return ret;
1694 }
1695 
1696 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1697 {
1698 	struct rlimit new_rlim;
1699 
1700 	if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1701 		return -EFAULT;
1702 	return do_prlimit(current, resource, &new_rlim, NULL);
1703 }
1704 
1705 /*
1706  * It would make sense to put struct rusage in the task_struct,
1707  * except that would make the task_struct be *really big*.  After
1708  * task_struct gets moved into malloc'ed memory, it would
1709  * make sense to do this.  It will make moving the rest of the information
1710  * a lot simpler!  (Which we're not doing right now because we're not
1711  * measuring them yet).
1712  *
1713  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1714  * races with threads incrementing their own counters.  But since word
1715  * reads are atomic, we either get new values or old values and we don't
1716  * care which for the sums.  We always take the siglock to protect reading
1717  * the c* fields from p->signal from races with exit.c updating those
1718  * fields when reaping, so a sample either gets all the additions of a
1719  * given child after it's reaped, or none so this sample is before reaping.
1720  *
1721  * Locking:
1722  * We need to take the siglock for CHILDEREN, SELF and BOTH
1723  * for  the cases current multithreaded, non-current single threaded
1724  * non-current multithreaded.  Thread traversal is now safe with
1725  * the siglock held.
1726  * Strictly speaking, we donot need to take the siglock if we are current and
1727  * single threaded,  as no one else can take our signal_struct away, no one
1728  * else can  reap the  children to update signal->c* counters, and no one else
1729  * can race with the signal-> fields. If we do not take any lock, the
1730  * signal-> fields could be read out of order while another thread was just
1731  * exiting. So we should  place a read memory barrier when we avoid the lock.
1732  * On the writer side,  write memory barrier is implied in  __exit_signal
1733  * as __exit_signal releases  the siglock spinlock after updating the signal->
1734  * fields. But we don't do this yet to keep things simple.
1735  *
1736  */
1737 
1738 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1739 {
1740 	r->ru_nvcsw += t->nvcsw;
1741 	r->ru_nivcsw += t->nivcsw;
1742 	r->ru_minflt += t->min_flt;
1743 	r->ru_majflt += t->maj_flt;
1744 	r->ru_inblock += task_io_get_inblock(t);
1745 	r->ru_oublock += task_io_get_oublock(t);
1746 }
1747 
1748 void getrusage(struct task_struct *p, int who, struct rusage *r)
1749 {
1750 	struct task_struct *t;
1751 	unsigned long flags;
1752 	u64 tgutime, tgstime, utime, stime;
1753 	unsigned long maxrss = 0;
1754 
1755 	memset((char *)r, 0, sizeof (*r));
1756 	utime = stime = 0;
1757 
1758 	if (who == RUSAGE_THREAD) {
1759 		task_cputime_adjusted(current, &utime, &stime);
1760 		accumulate_thread_rusage(p, r);
1761 		maxrss = p->signal->maxrss;
1762 		goto out;
1763 	}
1764 
1765 	if (!lock_task_sighand(p, &flags))
1766 		return;
1767 
1768 	switch (who) {
1769 	case RUSAGE_BOTH:
1770 	case RUSAGE_CHILDREN:
1771 		utime = p->signal->cutime;
1772 		stime = p->signal->cstime;
1773 		r->ru_nvcsw = p->signal->cnvcsw;
1774 		r->ru_nivcsw = p->signal->cnivcsw;
1775 		r->ru_minflt = p->signal->cmin_flt;
1776 		r->ru_majflt = p->signal->cmaj_flt;
1777 		r->ru_inblock = p->signal->cinblock;
1778 		r->ru_oublock = p->signal->coublock;
1779 		maxrss = p->signal->cmaxrss;
1780 
1781 		if (who == RUSAGE_CHILDREN)
1782 			break;
1783 		fallthrough;
1784 
1785 	case RUSAGE_SELF:
1786 		thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1787 		utime += tgutime;
1788 		stime += tgstime;
1789 		r->ru_nvcsw += p->signal->nvcsw;
1790 		r->ru_nivcsw += p->signal->nivcsw;
1791 		r->ru_minflt += p->signal->min_flt;
1792 		r->ru_majflt += p->signal->maj_flt;
1793 		r->ru_inblock += p->signal->inblock;
1794 		r->ru_oublock += p->signal->oublock;
1795 		if (maxrss < p->signal->maxrss)
1796 			maxrss = p->signal->maxrss;
1797 		t = p;
1798 		do {
1799 			accumulate_thread_rusage(t, r);
1800 		} while_each_thread(p, t);
1801 		break;
1802 
1803 	default:
1804 		BUG();
1805 	}
1806 	unlock_task_sighand(p, &flags);
1807 
1808 out:
1809 	r->ru_utime = ns_to_kernel_old_timeval(utime);
1810 	r->ru_stime = ns_to_kernel_old_timeval(stime);
1811 
1812 	if (who != RUSAGE_CHILDREN) {
1813 		struct mm_struct *mm = get_task_mm(p);
1814 
1815 		if (mm) {
1816 			setmax_mm_hiwater_rss(&maxrss, mm);
1817 			mmput(mm);
1818 		}
1819 	}
1820 	r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1821 }
1822 
1823 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1824 {
1825 	struct rusage r;
1826 
1827 	if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1828 	    who != RUSAGE_THREAD)
1829 		return -EINVAL;
1830 
1831 	getrusage(current, who, &r);
1832 	return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1833 }
1834 
1835 #ifdef CONFIG_COMPAT
1836 COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru)
1837 {
1838 	struct rusage r;
1839 
1840 	if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1841 	    who != RUSAGE_THREAD)
1842 		return -EINVAL;
1843 
1844 	getrusage(current, who, &r);
1845 	return put_compat_rusage(&r, ru);
1846 }
1847 #endif
1848 
1849 SYSCALL_DEFINE1(umask, int, mask)
1850 {
1851 	mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1852 	return mask;
1853 }
1854 
1855 static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1856 {
1857 	struct fd exe;
1858 	struct inode *inode;
1859 	int err;
1860 
1861 	exe = fdget(fd);
1862 	if (!exe.file)
1863 		return -EBADF;
1864 
1865 	inode = file_inode(exe.file);
1866 
1867 	/*
1868 	 * Because the original mm->exe_file points to executable file, make
1869 	 * sure that this one is executable as well, to avoid breaking an
1870 	 * overall picture.
1871 	 */
1872 	err = -EACCES;
1873 	if (!S_ISREG(inode->i_mode) || path_noexec(&exe.file->f_path))
1874 		goto exit;
1875 
1876 	err = file_permission(exe.file, MAY_EXEC);
1877 	if (err)
1878 		goto exit;
1879 
1880 	err = replace_mm_exe_file(mm, exe.file);
1881 exit:
1882 	fdput(exe);
1883 	return err;
1884 }
1885 
1886 /*
1887  * Check arithmetic relations of passed addresses.
1888  *
1889  * WARNING: we don't require any capability here so be very careful
1890  * in what is allowed for modification from userspace.
1891  */
1892 static int validate_prctl_map_addr(struct prctl_mm_map *prctl_map)
1893 {
1894 	unsigned long mmap_max_addr = TASK_SIZE;
1895 	int error = -EINVAL, i;
1896 
1897 	static const unsigned char offsets[] = {
1898 		offsetof(struct prctl_mm_map, start_code),
1899 		offsetof(struct prctl_mm_map, end_code),
1900 		offsetof(struct prctl_mm_map, start_data),
1901 		offsetof(struct prctl_mm_map, end_data),
1902 		offsetof(struct prctl_mm_map, start_brk),
1903 		offsetof(struct prctl_mm_map, brk),
1904 		offsetof(struct prctl_mm_map, start_stack),
1905 		offsetof(struct prctl_mm_map, arg_start),
1906 		offsetof(struct prctl_mm_map, arg_end),
1907 		offsetof(struct prctl_mm_map, env_start),
1908 		offsetof(struct prctl_mm_map, env_end),
1909 	};
1910 
1911 	/*
1912 	 * Make sure the members are not somewhere outside
1913 	 * of allowed address space.
1914 	 */
1915 	for (i = 0; i < ARRAY_SIZE(offsets); i++) {
1916 		u64 val = *(u64 *)((char *)prctl_map + offsets[i]);
1917 
1918 		if ((unsigned long)val >= mmap_max_addr ||
1919 		    (unsigned long)val < mmap_min_addr)
1920 			goto out;
1921 	}
1922 
1923 	/*
1924 	 * Make sure the pairs are ordered.
1925 	 */
1926 #define __prctl_check_order(__m1, __op, __m2)				\
1927 	((unsigned long)prctl_map->__m1 __op				\
1928 	 (unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1929 	error  = __prctl_check_order(start_code, <, end_code);
1930 	error |= __prctl_check_order(start_data,<=, end_data);
1931 	error |= __prctl_check_order(start_brk, <=, brk);
1932 	error |= __prctl_check_order(arg_start, <=, arg_end);
1933 	error |= __prctl_check_order(env_start, <=, env_end);
1934 	if (error)
1935 		goto out;
1936 #undef __prctl_check_order
1937 
1938 	error = -EINVAL;
1939 
1940 	/*
1941 	 * Neither we should allow to override limits if they set.
1942 	 */
1943 	if (check_data_rlimit(rlimit(RLIMIT_DATA), prctl_map->brk,
1944 			      prctl_map->start_brk, prctl_map->end_data,
1945 			      prctl_map->start_data))
1946 			goto out;
1947 
1948 	error = 0;
1949 out:
1950 	return error;
1951 }
1952 
1953 #ifdef CONFIG_CHECKPOINT_RESTORE
1954 static int prctl_set_mm_map(int opt, const void __user *addr, unsigned long data_size)
1955 {
1956 	struct prctl_mm_map prctl_map = { .exe_fd = (u32)-1, };
1957 	unsigned long user_auxv[AT_VECTOR_SIZE];
1958 	struct mm_struct *mm = current->mm;
1959 	int error;
1960 
1961 	BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
1962 	BUILD_BUG_ON(sizeof(struct prctl_mm_map) > 256);
1963 
1964 	if (opt == PR_SET_MM_MAP_SIZE)
1965 		return put_user((unsigned int)sizeof(prctl_map),
1966 				(unsigned int __user *)addr);
1967 
1968 	if (data_size != sizeof(prctl_map))
1969 		return -EINVAL;
1970 
1971 	if (copy_from_user(&prctl_map, addr, sizeof(prctl_map)))
1972 		return -EFAULT;
1973 
1974 	error = validate_prctl_map_addr(&prctl_map);
1975 	if (error)
1976 		return error;
1977 
1978 	if (prctl_map.auxv_size) {
1979 		/*
1980 		 * Someone is trying to cheat the auxv vector.
1981 		 */
1982 		if (!prctl_map.auxv ||
1983 				prctl_map.auxv_size > sizeof(mm->saved_auxv))
1984 			return -EINVAL;
1985 
1986 		memset(user_auxv, 0, sizeof(user_auxv));
1987 		if (copy_from_user(user_auxv,
1988 				   (const void __user *)prctl_map.auxv,
1989 				   prctl_map.auxv_size))
1990 			return -EFAULT;
1991 
1992 		/* Last entry must be AT_NULL as specification requires */
1993 		user_auxv[AT_VECTOR_SIZE - 2] = AT_NULL;
1994 		user_auxv[AT_VECTOR_SIZE - 1] = AT_NULL;
1995 	}
1996 
1997 	if (prctl_map.exe_fd != (u32)-1) {
1998 		/*
1999 		 * Check if the current user is checkpoint/restore capable.
2000 		 * At the time of this writing, it checks for CAP_SYS_ADMIN
2001 		 * or CAP_CHECKPOINT_RESTORE.
2002 		 * Note that a user with access to ptrace can masquerade an
2003 		 * arbitrary program as any executable, even setuid ones.
2004 		 * This may have implications in the tomoyo subsystem.
2005 		 */
2006 		if (!checkpoint_restore_ns_capable(current_user_ns()))
2007 			return -EPERM;
2008 
2009 		error = prctl_set_mm_exe_file(mm, prctl_map.exe_fd);
2010 		if (error)
2011 			return error;
2012 	}
2013 
2014 	/*
2015 	 * arg_lock protects concurrent updates but we still need mmap_lock for
2016 	 * read to exclude races with sys_brk.
2017 	 */
2018 	mmap_read_lock(mm);
2019 
2020 	/*
2021 	 * We don't validate if these members are pointing to
2022 	 * real present VMAs because application may have correspond
2023 	 * VMAs already unmapped and kernel uses these members for statistics
2024 	 * output in procfs mostly, except
2025 	 *
2026 	 *  - @start_brk/@brk which are used in do_brk_flags but kernel lookups
2027 	 *    for VMAs when updating these members so anything wrong written
2028 	 *    here cause kernel to swear at userspace program but won't lead
2029 	 *    to any problem in kernel itself
2030 	 */
2031 
2032 	spin_lock(&mm->arg_lock);
2033 	mm->start_code	= prctl_map.start_code;
2034 	mm->end_code	= prctl_map.end_code;
2035 	mm->start_data	= prctl_map.start_data;
2036 	mm->end_data	= prctl_map.end_data;
2037 	mm->start_brk	= prctl_map.start_brk;
2038 	mm->brk		= prctl_map.brk;
2039 	mm->start_stack	= prctl_map.start_stack;
2040 	mm->arg_start	= prctl_map.arg_start;
2041 	mm->arg_end	= prctl_map.arg_end;
2042 	mm->env_start	= prctl_map.env_start;
2043 	mm->env_end	= prctl_map.env_end;
2044 	spin_unlock(&mm->arg_lock);
2045 
2046 	/*
2047 	 * Note this update of @saved_auxv is lockless thus
2048 	 * if someone reads this member in procfs while we're
2049 	 * updating -- it may get partly updated results. It's
2050 	 * known and acceptable trade off: we leave it as is to
2051 	 * not introduce additional locks here making the kernel
2052 	 * more complex.
2053 	 */
2054 	if (prctl_map.auxv_size)
2055 		memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv));
2056 
2057 	mmap_read_unlock(mm);
2058 	return 0;
2059 }
2060 #endif /* CONFIG_CHECKPOINT_RESTORE */
2061 
2062 static int prctl_set_auxv(struct mm_struct *mm, unsigned long addr,
2063 			  unsigned long len)
2064 {
2065 	/*
2066 	 * This doesn't move the auxiliary vector itself since it's pinned to
2067 	 * mm_struct, but it permits filling the vector with new values.  It's
2068 	 * up to the caller to provide sane values here, otherwise userspace
2069 	 * tools which use this vector might be unhappy.
2070 	 */
2071 	unsigned long user_auxv[AT_VECTOR_SIZE] = {};
2072 
2073 	if (len > sizeof(user_auxv))
2074 		return -EINVAL;
2075 
2076 	if (copy_from_user(user_auxv, (const void __user *)addr, len))
2077 		return -EFAULT;
2078 
2079 	/* Make sure the last entry is always AT_NULL */
2080 	user_auxv[AT_VECTOR_SIZE - 2] = 0;
2081 	user_auxv[AT_VECTOR_SIZE - 1] = 0;
2082 
2083 	BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
2084 
2085 	task_lock(current);
2086 	memcpy(mm->saved_auxv, user_auxv, len);
2087 	task_unlock(current);
2088 
2089 	return 0;
2090 }
2091 
2092 static int prctl_set_mm(int opt, unsigned long addr,
2093 			unsigned long arg4, unsigned long arg5)
2094 {
2095 	struct mm_struct *mm = current->mm;
2096 	struct prctl_mm_map prctl_map = {
2097 		.auxv = NULL,
2098 		.auxv_size = 0,
2099 		.exe_fd = -1,
2100 	};
2101 	struct vm_area_struct *vma;
2102 	int error;
2103 
2104 	if (arg5 || (arg4 && (opt != PR_SET_MM_AUXV &&
2105 			      opt != PR_SET_MM_MAP &&
2106 			      opt != PR_SET_MM_MAP_SIZE)))
2107 		return -EINVAL;
2108 
2109 #ifdef CONFIG_CHECKPOINT_RESTORE
2110 	if (opt == PR_SET_MM_MAP || opt == PR_SET_MM_MAP_SIZE)
2111 		return prctl_set_mm_map(opt, (const void __user *)addr, arg4);
2112 #endif
2113 
2114 	if (!capable(CAP_SYS_RESOURCE))
2115 		return -EPERM;
2116 
2117 	if (opt == PR_SET_MM_EXE_FILE)
2118 		return prctl_set_mm_exe_file(mm, (unsigned int)addr);
2119 
2120 	if (opt == PR_SET_MM_AUXV)
2121 		return prctl_set_auxv(mm, addr, arg4);
2122 
2123 	if (addr >= TASK_SIZE || addr < mmap_min_addr)
2124 		return -EINVAL;
2125 
2126 	error = -EINVAL;
2127 
2128 	/*
2129 	 * arg_lock protects concurrent updates of arg boundaries, we need
2130 	 * mmap_lock for a) concurrent sys_brk, b) finding VMA for addr
2131 	 * validation.
2132 	 */
2133 	mmap_read_lock(mm);
2134 	vma = find_vma(mm, addr);
2135 
2136 	spin_lock(&mm->arg_lock);
2137 	prctl_map.start_code	= mm->start_code;
2138 	prctl_map.end_code	= mm->end_code;
2139 	prctl_map.start_data	= mm->start_data;
2140 	prctl_map.end_data	= mm->end_data;
2141 	prctl_map.start_brk	= mm->start_brk;
2142 	prctl_map.brk		= mm->brk;
2143 	prctl_map.start_stack	= mm->start_stack;
2144 	prctl_map.arg_start	= mm->arg_start;
2145 	prctl_map.arg_end	= mm->arg_end;
2146 	prctl_map.env_start	= mm->env_start;
2147 	prctl_map.env_end	= mm->env_end;
2148 
2149 	switch (opt) {
2150 	case PR_SET_MM_START_CODE:
2151 		prctl_map.start_code = addr;
2152 		break;
2153 	case PR_SET_MM_END_CODE:
2154 		prctl_map.end_code = addr;
2155 		break;
2156 	case PR_SET_MM_START_DATA:
2157 		prctl_map.start_data = addr;
2158 		break;
2159 	case PR_SET_MM_END_DATA:
2160 		prctl_map.end_data = addr;
2161 		break;
2162 	case PR_SET_MM_START_STACK:
2163 		prctl_map.start_stack = addr;
2164 		break;
2165 	case PR_SET_MM_START_BRK:
2166 		prctl_map.start_brk = addr;
2167 		break;
2168 	case PR_SET_MM_BRK:
2169 		prctl_map.brk = addr;
2170 		break;
2171 	case PR_SET_MM_ARG_START:
2172 		prctl_map.arg_start = addr;
2173 		break;
2174 	case PR_SET_MM_ARG_END:
2175 		prctl_map.arg_end = addr;
2176 		break;
2177 	case PR_SET_MM_ENV_START:
2178 		prctl_map.env_start = addr;
2179 		break;
2180 	case PR_SET_MM_ENV_END:
2181 		prctl_map.env_end = addr;
2182 		break;
2183 	default:
2184 		goto out;
2185 	}
2186 
2187 	error = validate_prctl_map_addr(&prctl_map);
2188 	if (error)
2189 		goto out;
2190 
2191 	switch (opt) {
2192 	/*
2193 	 * If command line arguments and environment
2194 	 * are placed somewhere else on stack, we can
2195 	 * set them up here, ARG_START/END to setup
2196 	 * command line arguments and ENV_START/END
2197 	 * for environment.
2198 	 */
2199 	case PR_SET_MM_START_STACK:
2200 	case PR_SET_MM_ARG_START:
2201 	case PR_SET_MM_ARG_END:
2202 	case PR_SET_MM_ENV_START:
2203 	case PR_SET_MM_ENV_END:
2204 		if (!vma) {
2205 			error = -EFAULT;
2206 			goto out;
2207 		}
2208 	}
2209 
2210 	mm->start_code	= prctl_map.start_code;
2211 	mm->end_code	= prctl_map.end_code;
2212 	mm->start_data	= prctl_map.start_data;
2213 	mm->end_data	= prctl_map.end_data;
2214 	mm->start_brk	= prctl_map.start_brk;
2215 	mm->brk		= prctl_map.brk;
2216 	mm->start_stack	= prctl_map.start_stack;
2217 	mm->arg_start	= prctl_map.arg_start;
2218 	mm->arg_end	= prctl_map.arg_end;
2219 	mm->env_start	= prctl_map.env_start;
2220 	mm->env_end	= prctl_map.env_end;
2221 
2222 	error = 0;
2223 out:
2224 	spin_unlock(&mm->arg_lock);
2225 	mmap_read_unlock(mm);
2226 	return error;
2227 }
2228 
2229 #ifdef CONFIG_CHECKPOINT_RESTORE
2230 static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr)
2231 {
2232 	return put_user(me->clear_child_tid, tid_addr);
2233 }
2234 #else
2235 static int prctl_get_tid_address(struct task_struct *me, int __user * __user *tid_addr)
2236 {
2237 	return -EINVAL;
2238 }
2239 #endif
2240 
2241 static int propagate_has_child_subreaper(struct task_struct *p, void *data)
2242 {
2243 	/*
2244 	 * If task has has_child_subreaper - all its descendants
2245 	 * already have these flag too and new descendants will
2246 	 * inherit it on fork, skip them.
2247 	 *
2248 	 * If we've found child_reaper - skip descendants in
2249 	 * it's subtree as they will never get out pidns.
2250 	 */
2251 	if (p->signal->has_child_subreaper ||
2252 	    is_child_reaper(task_pid(p)))
2253 		return 0;
2254 
2255 	p->signal->has_child_subreaper = 1;
2256 	return 1;
2257 }
2258 
2259 int __weak arch_prctl_spec_ctrl_get(struct task_struct *t, unsigned long which)
2260 {
2261 	return -EINVAL;
2262 }
2263 
2264 int __weak arch_prctl_spec_ctrl_set(struct task_struct *t, unsigned long which,
2265 				    unsigned long ctrl)
2266 {
2267 	return -EINVAL;
2268 }
2269 
2270 #define PR_IO_FLUSHER (PF_MEMALLOC_NOIO | PF_LOCAL_THROTTLE)
2271 
2272 #ifdef CONFIG_ANON_VMA_NAME
2273 
2274 #define ANON_VMA_NAME_MAX_LEN		80
2275 #define ANON_VMA_NAME_INVALID_CHARS	"\\`$[]"
2276 
2277 static inline bool is_valid_name_char(char ch)
2278 {
2279 	/* printable ascii characters, excluding ANON_VMA_NAME_INVALID_CHARS */
2280 	return ch > 0x1f && ch < 0x7f &&
2281 		!strchr(ANON_VMA_NAME_INVALID_CHARS, ch);
2282 }
2283 
2284 static int prctl_set_vma(unsigned long opt, unsigned long addr,
2285 			 unsigned long size, unsigned long arg)
2286 {
2287 	struct mm_struct *mm = current->mm;
2288 	const char __user *uname;
2289 	char *name, *pch;
2290 	int error;
2291 
2292 	switch (opt) {
2293 	case PR_SET_VMA_ANON_NAME:
2294 		uname = (const char __user *)arg;
2295 		if (uname) {
2296 			name = strndup_user(uname, ANON_VMA_NAME_MAX_LEN);
2297 
2298 			if (IS_ERR(name))
2299 				return PTR_ERR(name);
2300 
2301 			for (pch = name; *pch != '\0'; pch++) {
2302 				if (!is_valid_name_char(*pch)) {
2303 					kfree(name);
2304 					return -EINVAL;
2305 				}
2306 			}
2307 		} else {
2308 			/* Reset the name */
2309 			name = NULL;
2310 		}
2311 
2312 		mmap_write_lock(mm);
2313 		error = madvise_set_anon_name(mm, addr, size, name);
2314 		mmap_write_unlock(mm);
2315 		kfree(name);
2316 		break;
2317 	default:
2318 		error = -EINVAL;
2319 	}
2320 
2321 	return error;
2322 }
2323 
2324 #else /* CONFIG_ANON_VMA_NAME */
2325 static int prctl_set_vma(unsigned long opt, unsigned long start,
2326 			 unsigned long size, unsigned long arg)
2327 {
2328 	return -EINVAL;
2329 }
2330 #endif /* CONFIG_ANON_VMA_NAME */
2331 
2332 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2333 		unsigned long, arg4, unsigned long, arg5)
2334 {
2335 	struct task_struct *me = current;
2336 	unsigned char comm[sizeof(me->comm)];
2337 	long error;
2338 
2339 	error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2340 	if (error != -ENOSYS)
2341 		return error;
2342 
2343 	error = 0;
2344 	switch (option) {
2345 	case PR_SET_PDEATHSIG:
2346 		if (!valid_signal(arg2)) {
2347 			error = -EINVAL;
2348 			break;
2349 		}
2350 		me->pdeath_signal = arg2;
2351 		break;
2352 	case PR_GET_PDEATHSIG:
2353 		error = put_user(me->pdeath_signal, (int __user *)arg2);
2354 		break;
2355 	case PR_GET_DUMPABLE:
2356 		error = get_dumpable(me->mm);
2357 		break;
2358 	case PR_SET_DUMPABLE:
2359 		if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) {
2360 			error = -EINVAL;
2361 			break;
2362 		}
2363 		set_dumpable(me->mm, arg2);
2364 		break;
2365 
2366 	case PR_SET_UNALIGN:
2367 		error = SET_UNALIGN_CTL(me, arg2);
2368 		break;
2369 	case PR_GET_UNALIGN:
2370 		error = GET_UNALIGN_CTL(me, arg2);
2371 		break;
2372 	case PR_SET_FPEMU:
2373 		error = SET_FPEMU_CTL(me, arg2);
2374 		break;
2375 	case PR_GET_FPEMU:
2376 		error = GET_FPEMU_CTL(me, arg2);
2377 		break;
2378 	case PR_SET_FPEXC:
2379 		error = SET_FPEXC_CTL(me, arg2);
2380 		break;
2381 	case PR_GET_FPEXC:
2382 		error = GET_FPEXC_CTL(me, arg2);
2383 		break;
2384 	case PR_GET_TIMING:
2385 		error = PR_TIMING_STATISTICAL;
2386 		break;
2387 	case PR_SET_TIMING:
2388 		if (arg2 != PR_TIMING_STATISTICAL)
2389 			error = -EINVAL;
2390 		break;
2391 	case PR_SET_NAME:
2392 		comm[sizeof(me->comm) - 1] = 0;
2393 		if (strncpy_from_user(comm, (char __user *)arg2,
2394 				      sizeof(me->comm) - 1) < 0)
2395 			return -EFAULT;
2396 		set_task_comm(me, comm);
2397 		proc_comm_connector(me);
2398 		break;
2399 	case PR_GET_NAME:
2400 		get_task_comm(comm, me);
2401 		if (copy_to_user((char __user *)arg2, comm, sizeof(comm)))
2402 			return -EFAULT;
2403 		break;
2404 	case PR_GET_ENDIAN:
2405 		error = GET_ENDIAN(me, arg2);
2406 		break;
2407 	case PR_SET_ENDIAN:
2408 		error = SET_ENDIAN(me, arg2);
2409 		break;
2410 	case PR_GET_SECCOMP:
2411 		error = prctl_get_seccomp();
2412 		break;
2413 	case PR_SET_SECCOMP:
2414 		error = prctl_set_seccomp(arg2, (char __user *)arg3);
2415 		break;
2416 	case PR_GET_TSC:
2417 		error = GET_TSC_CTL(arg2);
2418 		break;
2419 	case PR_SET_TSC:
2420 		error = SET_TSC_CTL(arg2);
2421 		break;
2422 	case PR_TASK_PERF_EVENTS_DISABLE:
2423 		error = perf_event_task_disable();
2424 		break;
2425 	case PR_TASK_PERF_EVENTS_ENABLE:
2426 		error = perf_event_task_enable();
2427 		break;
2428 	case PR_GET_TIMERSLACK:
2429 		if (current->timer_slack_ns > ULONG_MAX)
2430 			error = ULONG_MAX;
2431 		else
2432 			error = current->timer_slack_ns;
2433 		break;
2434 	case PR_SET_TIMERSLACK:
2435 		if (arg2 <= 0)
2436 			current->timer_slack_ns =
2437 					current->default_timer_slack_ns;
2438 		else
2439 			current->timer_slack_ns = arg2;
2440 		break;
2441 	case PR_MCE_KILL:
2442 		if (arg4 | arg5)
2443 			return -EINVAL;
2444 		switch (arg2) {
2445 		case PR_MCE_KILL_CLEAR:
2446 			if (arg3 != 0)
2447 				return -EINVAL;
2448 			current->flags &= ~PF_MCE_PROCESS;
2449 			break;
2450 		case PR_MCE_KILL_SET:
2451 			current->flags |= PF_MCE_PROCESS;
2452 			if (arg3 == PR_MCE_KILL_EARLY)
2453 				current->flags |= PF_MCE_EARLY;
2454 			else if (arg3 == PR_MCE_KILL_LATE)
2455 				current->flags &= ~PF_MCE_EARLY;
2456 			else if (arg3 == PR_MCE_KILL_DEFAULT)
2457 				current->flags &=
2458 						~(PF_MCE_EARLY|PF_MCE_PROCESS);
2459 			else
2460 				return -EINVAL;
2461 			break;
2462 		default:
2463 			return -EINVAL;
2464 		}
2465 		break;
2466 	case PR_MCE_KILL_GET:
2467 		if (arg2 | arg3 | arg4 | arg5)
2468 			return -EINVAL;
2469 		if (current->flags & PF_MCE_PROCESS)
2470 			error = (current->flags & PF_MCE_EARLY) ?
2471 				PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2472 		else
2473 			error = PR_MCE_KILL_DEFAULT;
2474 		break;
2475 	case PR_SET_MM:
2476 		error = prctl_set_mm(arg2, arg3, arg4, arg5);
2477 		break;
2478 	case PR_GET_TID_ADDRESS:
2479 		error = prctl_get_tid_address(me, (int __user * __user *)arg2);
2480 		break;
2481 	case PR_SET_CHILD_SUBREAPER:
2482 		me->signal->is_child_subreaper = !!arg2;
2483 		if (!arg2)
2484 			break;
2485 
2486 		walk_process_tree(me, propagate_has_child_subreaper, NULL);
2487 		break;
2488 	case PR_GET_CHILD_SUBREAPER:
2489 		error = put_user(me->signal->is_child_subreaper,
2490 				 (int __user *)arg2);
2491 		break;
2492 	case PR_SET_NO_NEW_PRIVS:
2493 		if (arg2 != 1 || arg3 || arg4 || arg5)
2494 			return -EINVAL;
2495 
2496 		task_set_no_new_privs(current);
2497 		break;
2498 	case PR_GET_NO_NEW_PRIVS:
2499 		if (arg2 || arg3 || arg4 || arg5)
2500 			return -EINVAL;
2501 		return task_no_new_privs(current) ? 1 : 0;
2502 	case PR_GET_THP_DISABLE:
2503 		if (arg2 || arg3 || arg4 || arg5)
2504 			return -EINVAL;
2505 		error = !!test_bit(MMF_DISABLE_THP, &me->mm->flags);
2506 		break;
2507 	case PR_SET_THP_DISABLE:
2508 		if (arg3 || arg4 || arg5)
2509 			return -EINVAL;
2510 		if (mmap_write_lock_killable(me->mm))
2511 			return -EINTR;
2512 		if (arg2)
2513 			set_bit(MMF_DISABLE_THP, &me->mm->flags);
2514 		else
2515 			clear_bit(MMF_DISABLE_THP, &me->mm->flags);
2516 		mmap_write_unlock(me->mm);
2517 		break;
2518 	case PR_MPX_ENABLE_MANAGEMENT:
2519 	case PR_MPX_DISABLE_MANAGEMENT:
2520 		/* No longer implemented: */
2521 		return -EINVAL;
2522 	case PR_SET_FP_MODE:
2523 		error = SET_FP_MODE(me, arg2);
2524 		break;
2525 	case PR_GET_FP_MODE:
2526 		error = GET_FP_MODE(me);
2527 		break;
2528 	case PR_SVE_SET_VL:
2529 		error = SVE_SET_VL(arg2);
2530 		break;
2531 	case PR_SVE_GET_VL:
2532 		error = SVE_GET_VL();
2533 		break;
2534 	case PR_GET_SPECULATION_CTRL:
2535 		if (arg3 || arg4 || arg5)
2536 			return -EINVAL;
2537 		error = arch_prctl_spec_ctrl_get(me, arg2);
2538 		break;
2539 	case PR_SET_SPECULATION_CTRL:
2540 		if (arg4 || arg5)
2541 			return -EINVAL;
2542 		error = arch_prctl_spec_ctrl_set(me, arg2, arg3);
2543 		break;
2544 	case PR_PAC_RESET_KEYS:
2545 		if (arg3 || arg4 || arg5)
2546 			return -EINVAL;
2547 		error = PAC_RESET_KEYS(me, arg2);
2548 		break;
2549 	case PR_PAC_SET_ENABLED_KEYS:
2550 		if (arg4 || arg5)
2551 			return -EINVAL;
2552 		error = PAC_SET_ENABLED_KEYS(me, arg2, arg3);
2553 		break;
2554 	case PR_PAC_GET_ENABLED_KEYS:
2555 		if (arg2 || arg3 || arg4 || arg5)
2556 			return -EINVAL;
2557 		error = PAC_GET_ENABLED_KEYS(me);
2558 		break;
2559 	case PR_SET_TAGGED_ADDR_CTRL:
2560 		if (arg3 || arg4 || arg5)
2561 			return -EINVAL;
2562 		error = SET_TAGGED_ADDR_CTRL(arg2);
2563 		break;
2564 	case PR_GET_TAGGED_ADDR_CTRL:
2565 		if (arg2 || arg3 || arg4 || arg5)
2566 			return -EINVAL;
2567 		error = GET_TAGGED_ADDR_CTRL();
2568 		break;
2569 	case PR_SET_IO_FLUSHER:
2570 		if (!capable(CAP_SYS_RESOURCE))
2571 			return -EPERM;
2572 
2573 		if (arg3 || arg4 || arg5)
2574 			return -EINVAL;
2575 
2576 		if (arg2 == 1)
2577 			current->flags |= PR_IO_FLUSHER;
2578 		else if (!arg2)
2579 			current->flags &= ~PR_IO_FLUSHER;
2580 		else
2581 			return -EINVAL;
2582 		break;
2583 	case PR_GET_IO_FLUSHER:
2584 		if (!capable(CAP_SYS_RESOURCE))
2585 			return -EPERM;
2586 
2587 		if (arg2 || arg3 || arg4 || arg5)
2588 			return -EINVAL;
2589 
2590 		error = (current->flags & PR_IO_FLUSHER) == PR_IO_FLUSHER;
2591 		break;
2592 	case PR_SET_SYSCALL_USER_DISPATCH:
2593 		error = set_syscall_user_dispatch(arg2, arg3, arg4,
2594 						  (char __user *) arg5);
2595 		break;
2596 #ifdef CONFIG_SCHED_CORE
2597 	case PR_SCHED_CORE:
2598 		error = sched_core_share_pid(arg2, arg3, arg4, arg5);
2599 		break;
2600 #endif
2601 	case PR_SET_VMA:
2602 		error = prctl_set_vma(arg2, arg3, arg4, arg5);
2603 		break;
2604 	default:
2605 		error = -EINVAL;
2606 		break;
2607 	}
2608 	return error;
2609 }
2610 
2611 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2612 		struct getcpu_cache __user *, unused)
2613 {
2614 	int err = 0;
2615 	int cpu = raw_smp_processor_id();
2616 
2617 	if (cpup)
2618 		err |= put_user(cpu, cpup);
2619 	if (nodep)
2620 		err |= put_user(cpu_to_node(cpu), nodep);
2621 	return err ? -EFAULT : 0;
2622 }
2623 
2624 /**
2625  * do_sysinfo - fill in sysinfo struct
2626  * @info: pointer to buffer to fill
2627  */
2628 static int do_sysinfo(struct sysinfo *info)
2629 {
2630 	unsigned long mem_total, sav_total;
2631 	unsigned int mem_unit, bitcount;
2632 	struct timespec64 tp;
2633 
2634 	memset(info, 0, sizeof(struct sysinfo));
2635 
2636 	ktime_get_boottime_ts64(&tp);
2637 	timens_add_boottime(&tp);
2638 	info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
2639 
2640 	get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
2641 
2642 	info->procs = nr_threads;
2643 
2644 	si_meminfo(info);
2645 	si_swapinfo(info);
2646 
2647 	/*
2648 	 * If the sum of all the available memory (i.e. ram + swap)
2649 	 * is less than can be stored in a 32 bit unsigned long then
2650 	 * we can be binary compatible with 2.2.x kernels.  If not,
2651 	 * well, in that case 2.2.x was broken anyways...
2652 	 *
2653 	 *  -Erik Andersen <andersee@debian.org>
2654 	 */
2655 
2656 	mem_total = info->totalram + info->totalswap;
2657 	if (mem_total < info->totalram || mem_total < info->totalswap)
2658 		goto out;
2659 	bitcount = 0;
2660 	mem_unit = info->mem_unit;
2661 	while (mem_unit > 1) {
2662 		bitcount++;
2663 		mem_unit >>= 1;
2664 		sav_total = mem_total;
2665 		mem_total <<= 1;
2666 		if (mem_total < sav_total)
2667 			goto out;
2668 	}
2669 
2670 	/*
2671 	 * If mem_total did not overflow, multiply all memory values by
2672 	 * info->mem_unit and set it to 1.  This leaves things compatible
2673 	 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2674 	 * kernels...
2675 	 */
2676 
2677 	info->mem_unit = 1;
2678 	info->totalram <<= bitcount;
2679 	info->freeram <<= bitcount;
2680 	info->sharedram <<= bitcount;
2681 	info->bufferram <<= bitcount;
2682 	info->totalswap <<= bitcount;
2683 	info->freeswap <<= bitcount;
2684 	info->totalhigh <<= bitcount;
2685 	info->freehigh <<= bitcount;
2686 
2687 out:
2688 	return 0;
2689 }
2690 
2691 SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
2692 {
2693 	struct sysinfo val;
2694 
2695 	do_sysinfo(&val);
2696 
2697 	if (copy_to_user(info, &val, sizeof(struct sysinfo)))
2698 		return -EFAULT;
2699 
2700 	return 0;
2701 }
2702 
2703 #ifdef CONFIG_COMPAT
2704 struct compat_sysinfo {
2705 	s32 uptime;
2706 	u32 loads[3];
2707 	u32 totalram;
2708 	u32 freeram;
2709 	u32 sharedram;
2710 	u32 bufferram;
2711 	u32 totalswap;
2712 	u32 freeswap;
2713 	u16 procs;
2714 	u16 pad;
2715 	u32 totalhigh;
2716 	u32 freehigh;
2717 	u32 mem_unit;
2718 	char _f[20-2*sizeof(u32)-sizeof(int)];
2719 };
2720 
2721 COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2722 {
2723 	struct sysinfo s;
2724 	struct compat_sysinfo s_32;
2725 
2726 	do_sysinfo(&s);
2727 
2728 	/* Check to see if any memory value is too large for 32-bit and scale
2729 	 *  down if needed
2730 	 */
2731 	if (upper_32_bits(s.totalram) || upper_32_bits(s.totalswap)) {
2732 		int bitcount = 0;
2733 
2734 		while (s.mem_unit < PAGE_SIZE) {
2735 			s.mem_unit <<= 1;
2736 			bitcount++;
2737 		}
2738 
2739 		s.totalram >>= bitcount;
2740 		s.freeram >>= bitcount;
2741 		s.sharedram >>= bitcount;
2742 		s.bufferram >>= bitcount;
2743 		s.totalswap >>= bitcount;
2744 		s.freeswap >>= bitcount;
2745 		s.totalhigh >>= bitcount;
2746 		s.freehigh >>= bitcount;
2747 	}
2748 
2749 	memset(&s_32, 0, sizeof(s_32));
2750 	s_32.uptime = s.uptime;
2751 	s_32.loads[0] = s.loads[0];
2752 	s_32.loads[1] = s.loads[1];
2753 	s_32.loads[2] = s.loads[2];
2754 	s_32.totalram = s.totalram;
2755 	s_32.freeram = s.freeram;
2756 	s_32.sharedram = s.sharedram;
2757 	s_32.bufferram = s.bufferram;
2758 	s_32.totalswap = s.totalswap;
2759 	s_32.freeswap = s.freeswap;
2760 	s_32.procs = s.procs;
2761 	s_32.totalhigh = s.totalhigh;
2762 	s_32.freehigh = s.freehigh;
2763 	s_32.mem_unit = s.mem_unit;
2764 	if (copy_to_user(info, &s_32, sizeof(s_32)))
2765 		return -EFAULT;
2766 	return 0;
2767 }
2768 #endif /* CONFIG_COMPAT */
2769