1/*
2 * Copyright (c) 2000-2020 Apple Inc. All rights reserved.
3 *
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5 *
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
14 *
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
17 *
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
25 *
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27 */
28/* Copyright (c) 1995, 1997 Apple Computer, Inc. All Rights Reserved */
29/*-
30 * Copyright (c) 1982, 1986, 1991, 1993
31 * The Regents of the University of California. All rights reserved.
32 * (c) UNIX System Laboratories, Inc.
33 * All or some portions of this file are derived from material licensed
34 * to the University of California by American Telephone and Telegraph
35 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
36 * the permission of UNIX System Laboratories, Inc.
37 *
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
40 * are met:
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 3. All advertising materials mentioning features or use of this software
47 * must display the following acknowledgement:
48 * This product includes software developed by the University of
49 * California, Berkeley and its contributors.
50 * 4. Neither the name of the University nor the names of its contributors
51 * may be used to endorse or promote products derived from this software
52 * without specific prior written permission.
53 *
54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64 * SUCH DAMAGE.
65 *
66 * @(#)kern_resource.c 8.5 (Berkeley) 1/21/94
67 */
68/*
69 * NOTICE: This file was modified by SPARTA, Inc. in 2005 to introduce
70 * support for mandatory and extensible security protections. This notice
71 * is included in support of clause 2.2 (b) of the Apple Public License,
72 * Version 2.0.
73 */
74
75#include <sys/param.h>
76#include <sys/systm.h>
77#include <sys/sysctl.h>
78#include <sys/kernel.h>
79#include <sys/file_internal.h>
80#include <sys/resourcevar.h>
81#include <sys/malloc.h>
82#include <sys/proc_internal.h>
83#include <sys/kauth.h>
84#include <sys/mount_internal.h>
85#include <sys/sysproto.h>
86
87#include <security/audit/audit.h>
88
89#include <machine/vmparam.h>
90
91#include <mach/mach_types.h>
92#include <mach/time_value.h>
93#include <mach/task.h>
94#include <mach/task_info.h>
95#include <mach/vm_map.h>
96#include <mach/mach_vm.h>
97#include <mach/thread_act.h> /* for thread_policy_set( ) */
98#include <kern/thread.h>
99#include <kern/policy_internal.h>
100
101#include <kern/task.h>
102#include <kern/clock.h> /* for absolutetime_to_microtime() */
103#include <netinet/in.h> /* for TRAFFIC_MGT_SO_* */
104#if CONFIG_FREEZE
105#include <sys/kern_memorystatus_freeze.h> /* for memorystatus_freeze_mark_ui_transition */
106#endif /* CONFIG_FREEZE */
107#include <sys/socketvar.h> /* for struct socket */
108#if NECP
109#include <net/necp.h>
110#endif /* NECP */
111
112#include <vm/vm_map.h>
113
114#include <kern/assert.h>
115#include <sys/resource.h>
116#include <sys/resource_private.h>
117#include <sys/priv.h>
118#include <IOKit/IOBSD.h>
119
120#if CONFIG_MACF
121#include <security/mac_framework.h>
122#endif
123
124static void proc_limitblock(proc_t p);
125static void proc_limitunblock(proc_t p);
126static void proc_limitupdate(proc_t p, bool unblock,
127 void (^update)(struct plimit *plim));
128
129static int donice(struct proc *curp, struct proc *chgp, int n);
130static int dosetrlimit(struct proc *p, u_int which, struct rlimit *limp);
131static void do_background_socket(struct proc *p, thread_t thread);
132static int do_background_thread(thread_t thread, int priority);
133static int do_background_proc(struct proc *curp, struct proc *targetp, int priority);
134static int set_gpudeny_proc(struct proc *curp, struct proc *targetp, int priority);
135static int proc_set_darwin_role(proc_t curp, proc_t targetp, int priority);
136static int proc_get_darwin_role(proc_t curp, proc_t targetp, int *priority);
137static int proc_set_game_mode(proc_t targetp, int priority);
138static int proc_get_game_mode(proc_t targetp, int *priority);
139static int get_background_proc(struct proc *curp, struct proc *targetp, int *priority);
140
141int fill_task_rusage(task_t task, rusage_info_current *ri);
142void fill_task_billed_usage(task_t task, rusage_info_current *ri);
143int fill_task_io_rusage(task_t task, rusage_info_current *ri);
144int fill_task_qos_rusage(task_t task, rusage_info_current *ri);
145uint64_t get_task_logical_writes(task_t task, bool external);
146
147rlim_t maxdmap = MAXDSIZ; /* XXX */
148rlim_t maxsmap = MAXSSIZ - PAGE_MAX_SIZE; /* XXX */
149
150/* For plimit reference count */
151os_refgrp_decl(, rlimit_refgrp, "plimit_refcnt", NULL);
152
153static KALLOC_TYPE_DEFINE(plimit_zone, struct plimit, KT_DEFAULT);
154
155/*
156 * Limits on the number of open files per process, and the number
157 * of child processes per process.
158 *
159 * Note: would be in kern/subr_param.c in FreeBSD.
160 */
161__private_extern__ int maxfilesperproc = OPEN_MAX; /* per-proc open files limit */
162
163SYSCTL_INT(_kern, KERN_MAXPROCPERUID, maxprocperuid, CTLFLAG_RW | CTLFLAG_LOCKED,
164 &maxprocperuid, 0, "Maximum processes allowed per userid" );
165
166SYSCTL_INT(_kern, KERN_MAXFILESPERPROC, maxfilesperproc, CTLFLAG_RW | CTLFLAG_LOCKED,
167 &maxfilesperproc, 0, "Maximum files allowed open per process" );
168
169/* Args and fn for proc_iteration callback used in setpriority */
170struct puser_nice_args {
171 proc_t curp;
172 int prio;
173 id_t who;
174 int * foundp;
175 int * errorp;
176};
177static int puser_donice_callback(proc_t p, void * arg);
178
179
180/* Args and fn for proc_iteration callback used in setpriority */
181struct ppgrp_nice_args {
182 proc_t curp;
183 int prio;
184 int * foundp;
185 int * errorp;
186};
187static int ppgrp_donice_callback(proc_t p, void * arg);
188
189/*
190 * Resource controls and accounting.
191 */
192int
193getpriority(struct proc *curp, struct getpriority_args *uap, int32_t *retval)
194{
195 struct proc *p;
196 int low = PRIO_MAX + 1;
197 kauth_cred_t my_cred;
198 int refheld = 0;
199 int error = 0;
200
201 /* would also test (uap->who < 0), but id_t is unsigned */
202 if (uap->who > 0x7fffffff) {
203 return EINVAL;
204 }
205
206 switch (uap->which) {
207 case PRIO_PROCESS:
208 if (uap->who == 0) {
209 p = curp;
210 low = p->p_nice;
211 } else {
212 p = proc_find(pid: uap->who);
213 if (p == 0) {
214 break;
215 }
216 low = p->p_nice;
217 proc_rele(p);
218 }
219 break;
220
221 case PRIO_PGRP: {
222 struct pgrp *pg = PGRP_NULL;
223
224 if (uap->who == 0) {
225 /* returns the pgrp to ref */
226 pg = proc_pgrp(curp, NULL);
227 } else if ((pg = pgrp_find(uap->who)) == PGRP_NULL) {
228 break;
229 }
230 /* No need for iteration as it is a simple scan */
231 pgrp_lock(pgrp: pg);
232 PGMEMBERS_FOREACH(pg, p) {
233 if (p->p_nice < low) {
234 low = p->p_nice;
235 }
236 }
237 pgrp_unlock(pgrp: pg);
238 pgrp_rele(pgrp: pg);
239 break;
240 }
241
242 case PRIO_USER:
243 if (uap->who == 0) {
244 uap->who = kauth_cred_getuid(cred: kauth_cred_get());
245 }
246
247 proc_list_lock();
248
249 for (p = allproc.lh_first; p != 0; p = p->p_list.le_next) {
250 my_cred = kauth_cred_proc_ref(procp: p);
251 if (kauth_cred_getuid(cred: my_cred) == uap->who &&
252 p->p_nice < low) {
253 low = p->p_nice;
254 }
255 kauth_cred_unref(&my_cred);
256 }
257
258 proc_list_unlock();
259
260 break;
261
262 case PRIO_DARWIN_THREAD:
263 /* we currently only support the current thread */
264 if (uap->who != 0) {
265 return EINVAL;
266 }
267
268 low = proc_get_thread_policy(thread: current_thread(), TASK_POLICY_INTERNAL, TASK_POLICY_DARWIN_BG);
269
270 break;
271
272 case PRIO_DARWIN_PROCESS:
273 if (uap->who == 0) {
274 p = curp;
275 } else {
276 p = proc_find(pid: uap->who);
277 if (p == PROC_NULL) {
278 break;
279 }
280 refheld = 1;
281 }
282
283 error = get_background_proc(curp, targetp: p, priority: &low);
284
285 if (refheld) {
286 proc_rele(p);
287 }
288 if (error) {
289 return error;
290 }
291 break;
292
293 case PRIO_DARWIN_ROLE:
294 if (uap->who == 0) {
295 p = curp;
296 } else {
297 p = proc_find(pid: uap->who);
298 if (p == PROC_NULL) {
299 break;
300 }
301 refheld = 1;
302 }
303
304 error = proc_get_darwin_role(curp, targetp: p, priority: &low);
305
306 if (refheld) {
307 proc_rele(p);
308 }
309 if (error) {
310 return error;
311 }
312 break;
313
314 case PRIO_DARWIN_GAME_MODE:
315 if (uap->who == 0) {
316 p = curp;
317 } else {
318 p = proc_find(pid: uap->who);
319 if (p == PROC_NULL) {
320 break;
321 }
322 refheld = 1;
323 }
324
325
326 error = proc_get_game_mode(targetp: p, priority: &low);
327
328 if (refheld) {
329 proc_rele(p);
330 }
331 if (error) {
332 return error;
333 }
334 break;
335
336 default:
337 return EINVAL;
338 }
339 if (low == PRIO_MAX + 1) {
340 return ESRCH;
341 }
342 *retval = low;
343 return 0;
344}
345
346/* call back function used for proc iteration in PRIO_USER */
347static int
348puser_donice_callback(proc_t p, void * arg)
349{
350 int error, n;
351 struct puser_nice_args * pun = (struct puser_nice_args *)arg;
352 kauth_cred_t my_cred;
353
354 my_cred = kauth_cred_proc_ref(procp: p);
355 if (kauth_cred_getuid(cred: my_cred) == pun->who) {
356 error = donice(curp: pun->curp, chgp: p, n: pun->prio);
357 if (pun->errorp != NULL) {
358 *pun->errorp = error;
359 }
360 if (pun->foundp != NULL) {
361 n = *pun->foundp;
362 *pun->foundp = n + 1;
363 }
364 }
365 kauth_cred_unref(&my_cred);
366
367 return PROC_RETURNED;
368}
369
370/* call back function used for proc iteration in PRIO_PGRP */
371static int
372ppgrp_donice_callback(proc_t p, void * arg)
373{
374 int error;
375 struct ppgrp_nice_args * pun = (struct ppgrp_nice_args *)arg;
376 int n;
377
378 error = donice(curp: pun->curp, chgp: p, n: pun->prio);
379 if (pun->errorp != NULL) {
380 *pun->errorp = error;
381 }
382 if (pun->foundp != NULL) {
383 n = *pun->foundp;
384 *pun->foundp = n + 1;
385 }
386
387 return PROC_RETURNED;
388}
389
390/*
391 * Returns: 0 Success
392 * EINVAL
393 * ESRCH
394 * donice:EPERM
395 * donice:EACCES
396 */
397/* ARGSUSED */
398int
399setpriority(struct proc *curp, struct setpriority_args *uap, int32_t *retval)
400{
401 struct proc *p;
402 int found = 0, error = 0;
403 int refheld = 0;
404
405 AUDIT_ARG(cmd, uap->which);
406 AUDIT_ARG(owner, uap->who, 0);
407 AUDIT_ARG(value32, uap->prio);
408
409 /* would also test (uap->who < 0), but id_t is unsigned */
410 if (uap->who > 0x7fffffff) {
411 return EINVAL;
412 }
413
414 switch (uap->which) {
415 case PRIO_PROCESS:
416 if (uap->who == 0) {
417 p = curp;
418 } else {
419 p = proc_find(pid: uap->who);
420 if (p == 0) {
421 break;
422 }
423 refheld = 1;
424 }
425 error = donice(curp, chgp: p, n: uap->prio);
426 found++;
427 if (refheld != 0) {
428 proc_rele(p);
429 }
430 break;
431
432 case PRIO_PGRP: {
433 struct pgrp *pg = PGRP_NULL;
434 struct ppgrp_nice_args ppgrp;
435
436 if (uap->who == 0) {
437 pg = proc_pgrp(curp, NULL);
438 } else if ((pg = pgrp_find(uap->who)) == PGRP_NULL) {
439 break;
440 }
441
442 ppgrp.curp = curp;
443 ppgrp.prio = uap->prio;
444 ppgrp.foundp = &found;
445 ppgrp.errorp = &error;
446
447 pgrp_iterate(pgrp: pg, callout: ppgrp_donice_callback, arg: (void *)&ppgrp, NULL);
448 pgrp_rele(pgrp: pg);
449
450 break;
451 }
452
453 case PRIO_USER: {
454 struct puser_nice_args punice;
455
456 if (uap->who == 0) {
457 uap->who = kauth_cred_getuid(cred: kauth_cred_get());
458 }
459
460 punice.curp = curp;
461 punice.prio = uap->prio;
462 punice.who = uap->who;
463 punice.foundp = &found;
464 error = 0;
465 punice.errorp = &error;
466 proc_iterate(PROC_ALLPROCLIST, callout: puser_donice_callback, arg: (void *)&punice, NULL, NULL);
467
468 break;
469 }
470
471 case PRIO_DARWIN_THREAD: {
472 /* we currently only support the current thread */
473 if (uap->who != 0) {
474 return EINVAL;
475 }
476
477 error = do_background_thread(thread: current_thread(), priority: uap->prio);
478 found++;
479 break;
480 }
481
482 case PRIO_DARWIN_PROCESS: {
483 if (uap->who == 0) {
484 p = curp;
485 } else {
486 p = proc_find(pid: uap->who);
487 if (p == 0) {
488 break;
489 }
490 refheld = 1;
491 }
492
493 error = do_background_proc(curp, targetp: p, priority: uap->prio);
494
495 found++;
496 if (refheld != 0) {
497 proc_rele(p);
498 }
499 break;
500 }
501
502 case PRIO_DARWIN_GPU: {
503 if (uap->who == 0) {
504 return EINVAL;
505 }
506
507 p = proc_find(pid: uap->who);
508 if (p == PROC_NULL) {
509 break;
510 }
511
512 error = set_gpudeny_proc(curp, targetp: p, priority: uap->prio);
513
514 found++;
515 proc_rele(p);
516 break;
517 }
518
519 case PRIO_DARWIN_ROLE: {
520 if (uap->who == 0) {
521 p = curp;
522 } else {
523 p = proc_find(pid: uap->who);
524 if (p == PROC_NULL) {
525 break;
526 }
527 refheld = 1;
528 }
529
530 error = proc_set_darwin_role(curp, targetp: p, priority: uap->prio);
531
532 found++;
533 if (refheld != 0) {
534 proc_rele(p);
535 }
536 break;
537 }
538
539 case PRIO_DARWIN_GAME_MODE: {
540 if (uap->who == 0) {
541 p = curp;
542 } else {
543 p = proc_find(pid: uap->who);
544 if (p == PROC_NULL) {
545 break;
546 }
547 refheld = 1;
548 }
549
550
551 error = proc_set_game_mode(targetp: p, priority: uap->prio);
552
553 found++;
554 if (refheld != 0) {
555 proc_rele(p);
556 }
557 break;
558 }
559
560 default:
561 return EINVAL;
562 }
563 if (found == 0) {
564 return ESRCH;
565 }
566 if (error == EIDRM) {
567 *retval = -2;
568 error = 0;
569 }
570 return error;
571}
572
573
574/*
575 * Returns: 0 Success
576 * EPERM
577 * EACCES
578 * mac_check_proc_sched:???
579 */
580static int
581donice(struct proc *curp, struct proc *chgp, int n)
582{
583 int error = 0;
584 kauth_cred_t ucred;
585 kauth_cred_t my_cred;
586
587 ucred = kauth_cred_proc_ref(procp: curp);
588 my_cred = kauth_cred_proc_ref(procp: chgp);
589
590 if (suser(cred: ucred, NULL) && kauth_cred_getruid(cred: ucred) &&
591 kauth_cred_getuid(cred: ucred) != kauth_cred_getuid(cred: my_cred) &&
592 kauth_cred_getruid(cred: ucred) != kauth_cred_getuid(cred: my_cred)) {
593 error = EPERM;
594 goto out;
595 }
596 if (n > PRIO_MAX) {
597 n = PRIO_MAX;
598 }
599 if (n < PRIO_MIN) {
600 n = PRIO_MIN;
601 }
602 if (n < chgp->p_nice && suser(cred: ucred, acflag: &curp->p_acflag)) {
603 error = EACCES;
604 goto out;
605 }
606#if CONFIG_MACF
607 error = mac_proc_check_sched(proc: curp, proc2: chgp);
608 if (error) {
609 goto out;
610 }
611#endif
612 proc_lock(chgp);
613 chgp->p_nice = (char)n;
614 proc_unlock(chgp);
615 (void)resetpriority(chgp);
616out:
617 kauth_cred_unref(&ucred);
618 kauth_cred_unref(&my_cred);
619 return error;
620}
621
622static int
623set_gpudeny_proc(struct proc *curp, struct proc *targetp, int priority)
624{
625 int error = 0;
626 kauth_cred_t ucred;
627 kauth_cred_t target_cred;
628
629 ucred = kauth_cred_get();
630 target_cred = kauth_cred_proc_ref(procp: targetp);
631
632 /* TODO: Entitlement instead of uid check */
633
634 if (!kauth_cred_issuser(cred: ucred) && kauth_cred_getruid(cred: ucred) &&
635 kauth_cred_getuid(cred: ucred) != kauth_cred_getuid(cred: target_cred) &&
636 kauth_cred_getruid(cred: ucred) != kauth_cred_getuid(cred: target_cred)) {
637 error = EPERM;
638 goto out;
639 }
640
641 if (curp == targetp) {
642 error = EPERM;
643 goto out;
644 }
645
646#if CONFIG_MACF
647 error = mac_proc_check_sched(proc: curp, proc2: targetp);
648 if (error) {
649 goto out;
650 }
651#endif
652
653 switch (priority) {
654 case PRIO_DARWIN_GPU_DENY:
655 task_set_gpu_denied(task: proc_task(targetp), TRUE);
656 break;
657 case PRIO_DARWIN_GPU_ALLOW:
658 task_set_gpu_denied(task: proc_task(targetp), FALSE);
659 break;
660 default:
661 error = EINVAL;
662 goto out;
663 }
664
665out:
666 kauth_cred_unref(&target_cred);
667 return error;
668}
669
670static int
671proc_set_darwin_role(proc_t curp, proc_t targetp, int priority)
672{
673 int error = 0;
674 uint32_t flagsp = 0;
675
676 kauth_cred_t ucred, target_cred;
677
678 ucred = kauth_cred_get();
679 target_cred = kauth_cred_proc_ref(procp: targetp);
680
681 if (!kauth_cred_issuser(cred: ucred) && kauth_cred_getruid(cred: ucred) &&
682 kauth_cred_getuid(cred: ucred) != kauth_cred_getuid(cred: target_cred) &&
683 kauth_cred_getruid(cred: ucred) != kauth_cred_getuid(cred: target_cred)) {
684 if (priv_check_cred(cred: ucred, PRIV_SETPRIORITY_DARWIN_ROLE, flags: 0) != 0) {
685 error = EPERM;
686 goto out;
687 }
688 }
689
690 if (curp != targetp) {
691#if CONFIG_MACF
692 if ((error = mac_proc_check_sched(proc: curp, proc2: targetp))) {
693 goto out;
694 }
695#endif
696 }
697
698 proc_get_darwinbgstate(task: proc_task(targetp), flagsp: &flagsp);
699 if ((flagsp & PROC_FLAG_APPLICATION) != PROC_FLAG_APPLICATION) {
700 error = ENOTSUP;
701 goto out;
702 }
703
704 task_role_t role = TASK_UNSPECIFIED;
705
706 if ((error = proc_darwin_role_to_task_role(darwin_role: priority, task_role: &role))) {
707 goto out;
708 }
709
710 proc_set_task_policy(task: proc_task(targetp), TASK_POLICY_ATTRIBUTE,
711 TASK_POLICY_ROLE, value: role);
712
713#if CONFIG_FREEZE
714 if (priority == PRIO_DARWIN_ROLE_UI_FOCAL || priority == PRIO_DARWIN_ROLE_UI || priority == PRIO_DARWIN_ROLE_UI_NON_FOCAL) {
715 memorystatus_freezer_mark_ui_transition(targetp);
716 }
717#endif /* CONFIG_FREEZE */
718
719out:
720 kauth_cred_unref(&target_cred);
721 return error;
722}
723
724static int
725proc_get_darwin_role(proc_t curp, proc_t targetp, int *priority)
726{
727 int error = 0;
728 int role = 0;
729
730 kauth_cred_t ucred, target_cred;
731
732 ucred = kauth_cred_get();
733 target_cred = kauth_cred_proc_ref(procp: targetp);
734
735 if (!kauth_cred_issuser(cred: ucred) && kauth_cred_getruid(cred: ucred) &&
736 kauth_cred_getuid(cred: ucred) != kauth_cred_getuid(cred: target_cred) &&
737 kauth_cred_getruid(cred: ucred) != kauth_cred_getuid(cred: target_cred)) {
738 error = EPERM;
739 goto out;
740 }
741
742 if (curp != targetp) {
743#if CONFIG_MACF
744 if ((error = mac_proc_check_sched(proc: curp, proc2: targetp))) {
745 goto out;
746 }
747#endif
748 }
749
750 role = proc_get_task_policy(task: proc_task(targetp), TASK_POLICY_ATTRIBUTE, TASK_POLICY_ROLE);
751
752 *priority = proc_task_role_to_darwin_role(task_role: role);
753
754out:
755 kauth_cred_unref(&target_cred);
756 return error;
757}
758
759#define SET_GAME_MODE_ENTITLEMENT "com.apple.private.set-game-mode"
760
761static int
762proc_set_game_mode(proc_t targetp, int priority)
763{
764 int error = 0;
765
766 kauth_cred_t ucred, target_cred;
767
768 ucred = kauth_cred_get();
769 target_cred = kauth_cred_proc_ref(procp: targetp);
770
771 boolean_t entitled = FALSE;
772 entitled = IOCurrentTaskHasEntitlement(SET_GAME_MODE_ENTITLEMENT);
773 if (!entitled) {
774 error = EPERM;
775 goto out;
776 }
777
778 /* Even with entitlement, non-root is only alllowed to set same-user */
779 if (!kauth_cred_issuser(cred: ucred) &&
780 kauth_cred_getuid(cred: ucred) != kauth_cred_getuid(cred: target_cred)) {
781 error = EPERM;
782 goto out;
783 }
784
785 switch (priority) {
786 case PRIO_DARWIN_GAME_MODE_OFF:
787 task_set_game_mode(task: proc_task(targetp), false);
788 break;
789 case PRIO_DARWIN_GAME_MODE_ON:
790 task_set_game_mode(task: proc_task(targetp), true);
791 break;
792 default:
793 error = EINVAL;
794 goto out;
795 }
796
797out:
798 kauth_cred_unref(&target_cred);
799 return error;
800}
801
802static int
803proc_get_game_mode(proc_t targetp, int *priority)
804{
805 int error = 0;
806
807 kauth_cred_t ucred, target_cred;
808
809 ucred = kauth_cred_get();
810 target_cred = kauth_cred_proc_ref(procp: targetp);
811
812 boolean_t entitled = FALSE;
813 entitled = IOCurrentTaskHasEntitlement(SET_GAME_MODE_ENTITLEMENT);
814
815 /* Root is allowed to get without entitlement */
816 if (!kauth_cred_issuser(cred: ucred) && !entitled) {
817 error = EPERM;
818 goto out;
819 }
820
821 /* Even with entitlement, non-root is only alllowed to see same-user */
822 if (!kauth_cred_issuser(cred: ucred) &&
823 kauth_cred_getuid(cred: ucred) != kauth_cred_getuid(cred: target_cred)) {
824 error = EPERM;
825 goto out;
826 }
827
828 if (task_get_game_mode(task: proc_task(targetp))) {
829 *priority = PRIO_DARWIN_GAME_MODE_ON;
830 } else {
831 *priority = PRIO_DARWIN_GAME_MODE_OFF;
832 }
833
834out:
835 kauth_cred_unref(&target_cred);
836 return error;
837}
838
839
840
841static int
842get_background_proc(struct proc *curp, struct proc *targetp, int *priority)
843{
844 int external = 0;
845 int error = 0;
846 kauth_cred_t ucred, target_cred;
847
848 ucred = kauth_cred_get();
849 target_cred = kauth_cred_proc_ref(procp: targetp);
850
851 if (!kauth_cred_issuser(cred: ucred) && kauth_cred_getruid(cred: ucred) &&
852 kauth_cred_getuid(cred: ucred) != kauth_cred_getuid(cred: target_cred) &&
853 kauth_cred_getruid(cred: ucred) != kauth_cred_getuid(cred: target_cred)) {
854 error = EPERM;
855 goto out;
856 }
857
858 external = (curp == targetp) ? TASK_POLICY_INTERNAL : TASK_POLICY_EXTERNAL;
859
860 *priority = proc_get_task_policy(task: current_task(), category: external, TASK_POLICY_DARWIN_BG);
861
862out:
863 kauth_cred_unref(&target_cred);
864 return error;
865}
866
867static int
868do_background_proc(struct proc *curp, struct proc *targetp, int priority)
869{
870#if !CONFIG_MACF
871#pragma unused(curp)
872#endif
873 int error = 0;
874 kauth_cred_t ucred;
875 kauth_cred_t target_cred;
876 int external;
877 int enable;
878
879 ucred = kauth_cred_get();
880 target_cred = kauth_cred_proc_ref(procp: targetp);
881
882 if (!kauth_cred_issuser(cred: ucred) && kauth_cred_getruid(cred: ucred) &&
883 kauth_cred_getuid(cred: ucred) != kauth_cred_getuid(cred: target_cred) &&
884 kauth_cred_getruid(cred: ucred) != kauth_cred_getuid(cred: target_cred)) {
885 error = EPERM;
886 goto out;
887 }
888
889#if CONFIG_MACF
890 error = mac_proc_check_sched(proc: curp, proc2: targetp);
891 if (error) {
892 goto out;
893 }
894#endif
895
896 external = (curp == targetp) ? TASK_POLICY_INTERNAL : TASK_POLICY_EXTERNAL;
897
898 switch (priority) {
899 case PRIO_DARWIN_BG:
900 enable = TASK_POLICY_ENABLE;
901 break;
902 case PRIO_DARWIN_NONUI:
903 /* ignored for compatibility */
904 goto out;
905 default:
906 /* TODO: EINVAL if priority != 0 */
907 enable = TASK_POLICY_DISABLE;
908 break;
909 }
910
911 proc_set_task_policy(task: proc_task(targetp), category: external, TASK_POLICY_DARWIN_BG, value: enable);
912
913out:
914 kauth_cred_unref(&target_cred);
915 return error;
916}
917
918static void
919do_background_socket(struct proc *p, thread_t thread)
920{
921#if SOCKETS
922 struct fileproc *fp;
923 int background = false;
924#if NECP
925 int update_necp = false;
926#endif /* NECP */
927
928 if (thread != THREAD_NULL &&
929 get_threadtask(thread) != proc_task(p)) {
930 return;
931 }
932
933 proc_fdlock(p);
934
935 if (thread != THREAD_NULL) {
936 background = proc_get_effective_thread_policy(thread, TASK_POLICY_ALL_SOCKETS_BG);
937 } else {
938 background = proc_get_effective_task_policy(task: proc_task(p), TASK_POLICY_ALL_SOCKETS_BG);
939 }
940
941 if (background) {
942 /*
943 * For PRIO_DARWIN_PROCESS (thread is NULL), simply mark
944 * the sockets with the background flag. There's nothing
945 * to do here for the PRIO_DARWIN_THREAD case.
946 */
947 if (thread == THREAD_NULL) {
948 fdt_foreach(fp, p) {
949 if (FILEGLOB_DTYPE(fp->fp_glob) == DTYPE_SOCKET) {
950 struct socket *sockp = (struct socket *)fp_get_data(fp);
951 socket_set_traffic_mgt_flags(so: sockp, TRAFFIC_MGT_SO_BACKGROUND);
952 sockp->so_background_thread = NULL;
953 }
954#if NECP
955 else if (FILEGLOB_DTYPE(fp->fp_glob) == DTYPE_NETPOLICY) {
956 if (necp_set_client_as_background(proc: p, fp, background)) {
957 update_necp = true;
958 }
959 }
960#endif /* NECP */
961 }
962 }
963 } else {
964 /* disable networking IO throttle.
965 * NOTE - It is a known limitation of the current design that we
966 * could potentially clear TRAFFIC_MGT_SO_BACKGROUND bit for
967 * sockets created by other threads within this process.
968 */
969 fdt_foreach(fp, p) {
970 struct socket *sockp;
971
972 if (FILEGLOB_DTYPE(fp->fp_glob) == DTYPE_SOCKET) {
973 sockp = (struct socket *)fp_get_data(fp);
974 /* skip if only clearing this thread's sockets */
975 if ((thread) && (sockp->so_background_thread != thread)) {
976 continue;
977 }
978 socket_clear_traffic_mgt_flags(so: sockp, TRAFFIC_MGT_SO_BACKGROUND);
979 sockp->so_background_thread = NULL;
980 }
981#if NECP
982 else if (FILEGLOB_DTYPE(fp->fp_glob) == DTYPE_NETPOLICY) {
983 if (necp_set_client_as_background(proc: p, fp, background)) {
984 update_necp = true;
985 }
986 }
987#endif /* NECP */
988 }
989 }
990
991 proc_fdunlock(p);
992
993#if NECP
994 if (update_necp) {
995 necp_update_all_clients();
996 }
997#endif /* NECP */
998#else
999#pragma unused(p, thread)
1000#endif
1001}
1002
1003
1004/*
1005 * do_background_thread
1006 *
1007 * Requires: thread reference
1008 *
1009 * Returns: 0 Success
1010 * EPERM Tried to background while in vfork
1011 * XXX - todo - does this need a MACF hook?
1012 */
1013static int
1014do_background_thread(thread_t thread, int priority)
1015{
1016 int enable, external;
1017 int rv = 0;
1018
1019 /* Backgrounding is unsupported for workq threads */
1020 if (thread_is_static_param(thread)) {
1021 return EPERM;
1022 }
1023
1024 /* Not allowed to combine QoS and DARWIN_BG, doing so strips the QoS */
1025 if (thread_has_qos_policy(thread)) {
1026 thread_remove_qos_policy(thread);
1027 rv = EIDRM;
1028 }
1029
1030 /* TODO: Fail if someone passes something besides 0 or PRIO_DARWIN_BG */
1031 enable = (priority == PRIO_DARWIN_BG) ? TASK_POLICY_ENABLE : TASK_POLICY_DISABLE;
1032 external = (current_thread() == thread) ? TASK_POLICY_INTERNAL : TASK_POLICY_EXTERNAL;
1033
1034 proc_set_thread_policy(thread, category: external, TASK_POLICY_DARWIN_BG, value: enable);
1035
1036 return rv;
1037}
1038
1039
1040/*
1041 * Returns: 0 Success
1042 * copyin:EFAULT
1043 * dosetrlimit:
1044 */
1045/* ARGSUSED */
1046int
1047setrlimit(struct proc *p, struct setrlimit_args *uap, __unused int32_t *retval)
1048{
1049 struct rlimit alim;
1050 int error;
1051
1052 if ((error = copyin(uap->rlp, (caddr_t)&alim,
1053 sizeof(struct rlimit)))) {
1054 return error;
1055 }
1056
1057 return dosetrlimit(p, which: uap->which, limp: &alim);
1058}
1059
1060/*
1061 * Returns: 0 Success
1062 * EINVAL
1063 * suser:EPERM
1064 *
1065 * Notes: EINVAL is returned both for invalid arguments, and in the
1066 * case that the current usage (e.g. RLIMIT_STACK) is already
1067 * in excess of the requested limit.
1068 */
1069static int
1070dosetrlimit(struct proc *p, u_int which, struct rlimit *newrlim)
1071{
1072 struct rlimit rlim, stack_rlim = {.rlim_cur = 0, .rlim_max = 0};
1073 int error;
1074 kern_return_t kr;
1075
1076 /* Mask out POSIX flag, saved above */
1077 which &= ~_RLIMIT_POSIX_FLAG;
1078
1079 /* Unknown resource */
1080 if (which >= RLIM_NLIMITS) {
1081 return EINVAL;
1082 }
1083
1084 proc_lock(p);
1085
1086 /* Only one thread is able to change the current process's rlimit values */
1087 proc_limitblock(p);
1088
1089 /*
1090 * Take a snapshot of the current rlimit values and read this throughout
1091 * this routine. This minimizes the critical sections and allow other
1092 * processes in the system to access the plimit while we are in the
1093 * middle of this setrlimit call.
1094 */
1095 rlim = smr_serialized_load(&p->p_limit)->pl_rlimit[which];
1096
1097 proc_unlock(p);
1098
1099 error = 0;
1100 /* Sanity check: new soft limit cannot exceed new hard limit */
1101 if (newrlim->rlim_cur > newrlim->rlim_max) {
1102 error = EINVAL;
1103 }
1104 /*
1105 * Sanity check: only super-user may raise the hard limit.
1106 * newrlim->rlim_cur > rlim.rlim_max implies that the call
1107 * is increasing the hard limit as well.
1108 */
1109 else if (newrlim->rlim_cur > rlim.rlim_max || newrlim->rlim_max > rlim.rlim_max) {
1110 /* suser() returns 0 if the calling thread is super user. */
1111 error = suser(cred: kauth_cred_get(), acflag: &p->p_acflag);
1112 }
1113
1114 if (error) {
1115 /* Invalid setrlimit request: EINVAL or EPERM */
1116 goto out;
1117 }
1118
1119 /* We have the reader lock of the process's plimit so it's safe to read the rlimit values */
1120 switch (which) {
1121 case RLIMIT_CPU:
1122 if (newrlim->rlim_cur == RLIM_INFINITY) {
1123 task_vtimer_clear(task: proc_task(p), TASK_VTIMER_RLIM);
1124 timerclear(&p->p_rlim_cpu);
1125 } else {
1126 task_absolutetime_info_data_t tinfo;
1127 mach_msg_type_number_t count;
1128 struct timeval ttv, tv;
1129 clock_sec_t tv_sec;
1130 clock_usec_t tv_usec;
1131
1132 count = TASK_ABSOLUTETIME_INFO_COUNT;
1133 task_info(task: proc_task(p), TASK_ABSOLUTETIME_INFO, task_info_out: (task_info_t)&tinfo, task_info_count: &count);
1134 absolutetime_to_microtime(abstime: tinfo.total_user + tinfo.total_system, secs: &tv_sec, microsecs: &tv_usec);
1135 ttv.tv_sec = tv_sec;
1136 ttv.tv_usec = tv_usec;
1137
1138 tv.tv_sec = (newrlim->rlim_cur > __INT_MAX__ ? __INT_MAX__ : (__darwin_time_t)newrlim->rlim_cur);
1139 tv.tv_usec = 0;
1140 timersub(&tv, &ttv, &p->p_rlim_cpu);
1141
1142 timerclear(&tv);
1143 if (timercmp(&p->p_rlim_cpu, &tv, >)) {
1144 task_vtimer_set(task: proc_task(p), TASK_VTIMER_RLIM);
1145 } else {
1146 task_vtimer_clear(task: proc_task(p), TASK_VTIMER_RLIM);
1147
1148 timerclear(&p->p_rlim_cpu);
1149
1150 psignal(p, SIGXCPU);
1151 }
1152 }
1153 break;
1154
1155 case RLIMIT_DATA:
1156#if 00
1157 if (newrlim->rlim_cur > maxdmap) {
1158 newrlim->rlim_cur = maxdmap;
1159 }
1160 if (newrlim->rlim_max > maxdmap) {
1161 newrlim->rlim_max = maxdmap;
1162 }
1163#endif
1164
1165 /* Over to Mach VM to validate the new data limit */
1166 if (vm_map_set_data_limit(map: current_map(), limit: newrlim->rlim_cur) != KERN_SUCCESS) {
1167 /* The limit specified cannot be lowered because current usage is already higher than the limit. */
1168 error = EINVAL;
1169 goto out;
1170 }
1171 break;
1172
1173 case RLIMIT_STACK:
1174 if (p->p_lflag & P_LCUSTOM_STACK) {
1175 /* Process has a custom stack set - rlimit cannot be used to change it */
1176 error = EINVAL;
1177 goto out;
1178 }
1179
1180 /*
1181 * Note: the real stack size limit is enforced by maxsmap, not a process's RLIMIT_STACK.
1182 *
1183 * The kernel uses maxsmap to control the actual stack size limit. While we allow
1184 * processes to set RLIMIT_STACK to RLIM_INFINITY (UNIX 03), accessing memory
1185 * beyond the maxsmap will still trigger an exception.
1186 *
1187 * stack_rlim is used to store the user-defined RLIMIT_STACK values while we adjust
1188 * the stack size using kernel limit (i.e. maxsmap).
1189 */
1190 if (newrlim->rlim_cur > maxsmap ||
1191 newrlim->rlim_max > maxsmap) {
1192 if (newrlim->rlim_cur > maxsmap) {
1193 stack_rlim.rlim_cur = newrlim->rlim_cur;
1194 newrlim->rlim_cur = maxsmap;
1195 }
1196 if (newrlim->rlim_max > maxsmap) {
1197 stack_rlim.rlim_max = newrlim->rlim_max;
1198 newrlim->rlim_max = maxsmap;
1199 }
1200 }
1201
1202 /*
1203 * Stack is allocated to the max at exec time with only
1204 * "rlim_cur" bytes accessible. If stack limit is going
1205 * up make more accessible, if going down make inaccessible.
1206 */
1207 if (newrlim->rlim_cur > rlim.rlim_cur) {
1208 mach_vm_offset_t addr;
1209 mach_vm_size_t size;
1210
1211 /* grow stack */
1212 size = newrlim->rlim_cur;
1213 if (round_page_overflow(size, &size)) {
1214 error = EINVAL;
1215 goto out;
1216 }
1217 size -= round_page_64(x: rlim.rlim_cur);
1218
1219 addr = (mach_vm_offset_t)(p->user_stack - round_page_64(x: newrlim->rlim_cur));
1220 kr = mach_vm_protect(target_task: current_map(), address: addr, size, FALSE, VM_PROT_DEFAULT);
1221 if (kr != KERN_SUCCESS) {
1222 error = EINVAL;
1223 goto out;
1224 }
1225 } else if (newrlim->rlim_cur < rlim.rlim_cur) {
1226 mach_vm_offset_t addr;
1227 mach_vm_size_t size;
1228 uint64_t cur_sp;
1229
1230 /* shrink stack */
1231
1232 /*
1233 * First check if new stack limit would agree
1234 * with current stack usage.
1235 * Get the current thread's stack pointer...
1236 */
1237 cur_sp = thread_adjuserstack(thread: current_thread(), adjust: 0);
1238 if (cur_sp <= p->user_stack &&
1239 cur_sp > (p->user_stack - round_page_64(x: rlim.rlim_cur))) {
1240 /* stack pointer is in main stack */
1241 if (cur_sp <= (p->user_stack - round_page_64(x: newrlim->rlim_cur))) {
1242 /*
1243 * New limit would cause current usage to be invalid:
1244 * reject new limit.
1245 */
1246 error = EINVAL;
1247 goto out;
1248 }
1249 } else {
1250 /* not on the main stack: reject */
1251 error = EINVAL;
1252 goto out;
1253 }
1254
1255 size = round_page_64(x: rlim.rlim_cur);
1256 size -= round_page_64(x: rlim.rlim_cur);
1257
1258 addr = (mach_vm_offset_t)(p->user_stack - round_page_64(x: rlim.rlim_cur));
1259
1260 kr = mach_vm_protect(target_task: current_map(), address: addr, size, FALSE, VM_PROT_NONE);
1261 if (kr != KERN_SUCCESS) {
1262 error = EINVAL;
1263 goto out;
1264 }
1265 } else {
1266 /* no change ... */
1267 }
1268
1269 /*
1270 * We've adjusted the process's stack region. If the user-defined limit is greater
1271 * than maxsmap, we need to reflect this change in rlimit interface.
1272 */
1273 if (stack_rlim.rlim_cur != 0) {
1274 newrlim->rlim_cur = stack_rlim.rlim_cur;
1275 }
1276 if (stack_rlim.rlim_max != 0) {
1277 newrlim->rlim_max = stack_rlim.rlim_max;
1278 }
1279 break;
1280
1281 case RLIMIT_NOFILE:
1282 /*
1283 * Nothing to be done here as we already performed the sanity checks before entering the switch code block.
1284 * The real NOFILE limits enforced by the kernel is capped at MIN(RLIMIT_NOFILE, maxfilesperproc)
1285 */
1286 break;
1287
1288 case RLIMIT_AS:
1289 /* Over to Mach VM to validate the new address space limit */
1290 if (vm_map_set_size_limit(map: current_map(), limit: newrlim->rlim_cur) != KERN_SUCCESS) {
1291 /* The limit specified cannot be lowered because current usage is already higher than the limit. */
1292 error = EINVAL;
1293 goto out;
1294 }
1295 break;
1296
1297 case RLIMIT_NPROC:
1298 /*
1299 * Only root can set to the maxproc limits, as it is
1300 * systemwide resource; all others are limited to
1301 * maxprocperuid (presumably less than maxproc).
1302 */
1303 if (kauth_cred_issuser(cred: kauth_cred_get())) {
1304 if (newrlim->rlim_cur > (rlim_t)maxproc) {
1305 newrlim->rlim_cur = maxproc;
1306 }
1307 if (newrlim->rlim_max > (rlim_t)maxproc) {
1308 newrlim->rlim_max = maxproc;
1309 }
1310 } else {
1311 if (newrlim->rlim_cur > (rlim_t)maxprocperuid) {
1312 newrlim->rlim_cur = maxprocperuid;
1313 }
1314 if (newrlim->rlim_max > (rlim_t)maxprocperuid) {
1315 newrlim->rlim_max = maxprocperuid;
1316 }
1317 }
1318 break;
1319
1320 case RLIMIT_MEMLOCK:
1321 /*
1322 * Tell the Mach VM layer about the new limit value.
1323 */
1324 newrlim->rlim_cur = (vm_size_t)newrlim->rlim_cur;
1325 vm_map_set_user_wire_limit(map: current_map(), limit: (vm_size_t)newrlim->rlim_cur);
1326 break;
1327 } /* switch... */
1328
1329 /* Everything checks out and we are now ready to update the rlimit */
1330 error = 0;
1331
1332out:
1333
1334 if (error == 0) {
1335 /*
1336 * COW the current plimit if it's shared, otherwise update it in place.
1337 * Finally unblock other threads wishing to change plimit.
1338 */
1339 proc_limitupdate(p, true, update: ^(struct plimit *plim) {
1340 plim->pl_rlimit[which] = *newrlim;
1341 });
1342 } else {
1343 /*
1344 * This setrlimit has failed, just leave the plimit as is and unblock other
1345 * threads wishing to change plimit.
1346 */
1347 proc_lock(p);
1348 proc_limitunblock(p);
1349 proc_unlock(p);
1350 }
1351
1352 return error;
1353}
1354
1355/* ARGSUSED */
1356int
1357getrlimit(struct proc *p, struct getrlimit_args *uap, __unused int32_t *retval)
1358{
1359 struct rlimit lim = {};
1360
1361 /*
1362 * Take out flag now in case we need to use it to trigger variant
1363 * behaviour later.
1364 */
1365 uap->which &= ~_RLIMIT_POSIX_FLAG;
1366
1367 if (uap->which >= RLIM_NLIMITS) {
1368 return EINVAL;
1369 }
1370 lim = proc_limitget(p, which: uap->which);
1371 return copyout((caddr_t)&lim,
1372 uap->rlp, sizeof(struct rlimit));
1373}
1374
1375/*
1376 * Transform the running time and tick information in proc p into user,
1377 * system, and interrupt time usage.
1378 */
1379/* No lock on proc is held for this.. */
1380void
1381calcru(struct proc *p, struct timeval *up, struct timeval *sp, struct timeval *ip)
1382{
1383 task_t task;
1384
1385 timerclear(up);
1386 timerclear(sp);
1387 if (ip != NULL) {
1388 timerclear(ip);
1389 }
1390
1391 task = proc_task(p);
1392 if (task) {
1393 mach_task_basic_info_data_t tinfo;
1394 task_thread_times_info_data_t ttimesinfo;
1395 task_events_info_data_t teventsinfo;
1396 mach_msg_type_number_t task_info_count, task_ttimes_count;
1397 mach_msg_type_number_t task_events_count;
1398 struct timeval ut, st;
1399
1400 task_info_count = MACH_TASK_BASIC_INFO_COUNT;
1401 task_info(task, MACH_TASK_BASIC_INFO,
1402 task_info_out: (task_info_t)&tinfo, task_info_count: &task_info_count);
1403 ut.tv_sec = tinfo.user_time.seconds;
1404 ut.tv_usec = tinfo.user_time.microseconds;
1405 st.tv_sec = tinfo.system_time.seconds;
1406 st.tv_usec = tinfo.system_time.microseconds;
1407 timeradd(&ut, up, up);
1408 timeradd(&st, sp, sp);
1409
1410 task_ttimes_count = TASK_THREAD_TIMES_INFO_COUNT;
1411 task_info(task, TASK_THREAD_TIMES_INFO,
1412 task_info_out: (task_info_t)&ttimesinfo, task_info_count: &task_ttimes_count);
1413
1414 ut.tv_sec = ttimesinfo.user_time.seconds;
1415 ut.tv_usec = ttimesinfo.user_time.microseconds;
1416 st.tv_sec = ttimesinfo.system_time.seconds;
1417 st.tv_usec = ttimesinfo.system_time.microseconds;
1418 timeradd(&ut, up, up);
1419 timeradd(&st, sp, sp);
1420
1421 task_events_count = TASK_EVENTS_INFO_COUNT;
1422 task_info(task, TASK_EVENTS_INFO,
1423 task_info_out: (task_info_t)&teventsinfo, task_info_count: &task_events_count);
1424
1425 /*
1426 * No need to lock "p": this does not need to be
1427 * completely consistent, right ?
1428 */
1429 p->p_stats->p_ru.ru_minflt = (teventsinfo.faults -
1430 teventsinfo.pageins);
1431 p->p_stats->p_ru.ru_majflt = teventsinfo.pageins;
1432 p->p_stats->p_ru.ru_nivcsw = (teventsinfo.csw -
1433 p->p_stats->p_ru.ru_nvcsw);
1434 if (p->p_stats->p_ru.ru_nivcsw < 0) {
1435 p->p_stats->p_ru.ru_nivcsw = 0;
1436 }
1437
1438 p->p_stats->p_ru.ru_maxrss = (long)tinfo.resident_size_max;
1439 }
1440}
1441
1442__private_extern__ void munge_user64_rusage(struct rusage *a_rusage_p, struct user64_rusage *a_user_rusage_p);
1443__private_extern__ void munge_user32_rusage(struct rusage *a_rusage_p, struct user32_rusage *a_user_rusage_p);
1444
1445/* ARGSUSED */
1446int
1447getrusage(struct proc *p, struct getrusage_args *uap, __unused int32_t *retval)
1448{
1449 struct rusage *rup, rubuf;
1450 struct user64_rusage rubuf64 = {};
1451 struct user32_rusage rubuf32 = {};
1452 size_t retsize = sizeof(rubuf); /* default: 32 bits */
1453 caddr_t retbuf = (caddr_t)&rubuf; /* default: 32 bits */
1454 struct timeval utime;
1455 struct timeval stime;
1456
1457
1458 switch (uap->who) {
1459 case RUSAGE_SELF:
1460 calcru(p, up: &utime, sp: &stime, NULL);
1461 proc_lock(p);
1462 rup = &p->p_stats->p_ru;
1463 rup->ru_utime = utime;
1464 rup->ru_stime = stime;
1465
1466 rubuf = *rup;
1467 proc_unlock(p);
1468
1469 break;
1470
1471 case RUSAGE_CHILDREN:
1472 proc_lock(p);
1473 rup = &p->p_stats->p_cru;
1474 rubuf = *rup;
1475 proc_unlock(p);
1476 break;
1477
1478 default:
1479 return EINVAL;
1480 }
1481 if (IS_64BIT_PROCESS(p)) {
1482 retsize = sizeof(rubuf64);
1483 retbuf = (caddr_t)&rubuf64;
1484 munge_user64_rusage(a_rusage_p: &rubuf, a_user_rusage_p: &rubuf64);
1485 } else {
1486 retsize = sizeof(rubuf32);
1487 retbuf = (caddr_t)&rubuf32;
1488 munge_user32_rusage(a_rusage_p: &rubuf, a_user_rusage_p: &rubuf32);
1489 }
1490
1491 return copyout(retbuf, uap->rusage, retsize);
1492}
1493
1494void
1495ruadd(struct rusage *ru, struct rusage *ru2)
1496{
1497 long *ip, *ip2;
1498 long i;
1499
1500 timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
1501 timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
1502 if (ru->ru_maxrss < ru2->ru_maxrss) {
1503 ru->ru_maxrss = ru2->ru_maxrss;
1504 }
1505 ip = &ru->ru_first; ip2 = &ru2->ru_first;
1506 for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--) {
1507 *ip++ += *ip2++;
1508 }
1509}
1510
1511/*
1512 * Add the rusage stats of child in parent.
1513 *
1514 * It adds rusage statistics of child process and statistics of all its
1515 * children to its parent.
1516 *
1517 * Note: proc lock of parent should be held while calling this function.
1518 */
1519void
1520update_rusage_info_child(struct rusage_info_child *ri, rusage_info_current *ri_current)
1521{
1522 ri->ri_child_user_time += (ri_current->ri_user_time +
1523 ri_current->ri_child_user_time);
1524 ri->ri_child_system_time += (ri_current->ri_system_time +
1525 ri_current->ri_child_system_time);
1526 ri->ri_child_pkg_idle_wkups += (ri_current->ri_pkg_idle_wkups +
1527 ri_current->ri_child_pkg_idle_wkups);
1528 ri->ri_child_interrupt_wkups += (ri_current->ri_interrupt_wkups +
1529 ri_current->ri_child_interrupt_wkups);
1530 ri->ri_child_pageins += (ri_current->ri_pageins +
1531 ri_current->ri_child_pageins);
1532 ri->ri_child_elapsed_abstime += ((ri_current->ri_proc_exit_abstime -
1533 ri_current->ri_proc_start_abstime) + ri_current->ri_child_elapsed_abstime);
1534}
1535
1536static void
1537proc_limit_free(smr_node_t node)
1538{
1539 struct plimit *plimit = __container_of(node, struct plimit, pl_node);
1540
1541 zfree(plimit_zone, plimit);
1542}
1543
1544static void
1545proc_limit_release(struct plimit *plimit)
1546{
1547 if (os_ref_release(rc: &plimit->pl_refcnt) == 0) {
1548 smr_proc_task_call(&plimit->pl_node, sizeof(*plimit), proc_limit_free);
1549 }
1550}
1551
1552/*
1553 * Reading soft limit from specified resource.
1554 */
1555rlim_t
1556proc_limitgetcur(proc_t p, int which)
1557{
1558 rlim_t rlim_cur;
1559
1560 assert(p);
1561 assert(which < RLIM_NLIMITS);
1562
1563 smr_proc_task_enter();
1564 rlim_cur = smr_entered_load(&p->p_limit)->pl_rlimit[which].rlim_cur;
1565 smr_proc_task_leave();
1566
1567 return rlim_cur;
1568}
1569
1570/*
1571 * Handle commonly asked limit that needs to be clamped with maxfilesperproc.
1572 */
1573int
1574proc_limitgetcur_nofile(struct proc *p)
1575{
1576 rlim_t lim = proc_limitgetcur(p, RLIMIT_NOFILE);
1577
1578 return (int)MIN(lim, maxfilesperproc);
1579}
1580
1581/*
1582 * Writing soft limit to specified resource. This is an internal function
1583 * used only by proc_exit to update RLIMIT_FSIZE in
1584 * place without invoking setrlimit.
1585 */
1586void
1587proc_limitsetcur_fsize(proc_t p, rlim_t value)
1588{
1589 proc_limitupdate(p, false, update: ^(struct plimit *plimit) {
1590 plimit->pl_rlimit[RLIMIT_FSIZE].rlim_cur = value;
1591 });
1592}
1593
1594struct rlimit
1595proc_limitget(proc_t p, int which)
1596{
1597 struct rlimit lim;
1598
1599 assert(which < RLIM_NLIMITS);
1600
1601 smr_proc_task_enter();
1602 lim = smr_entered_load(&p->p_limit)->pl_rlimit[which];
1603 smr_proc_task_leave();
1604
1605 return lim;
1606}
1607
1608void
1609proc_limitfork(proc_t parent, proc_t child)
1610{
1611 struct plimit *plim;
1612
1613 proc_lock(parent);
1614 plim = smr_serialized_load(&parent->p_limit);
1615 os_ref_retain(rc: &plim->pl_refcnt);
1616 proc_unlock(parent);
1617
1618 smr_init_store(&child->p_limit, plim);
1619}
1620
1621void
1622proc_limitdrop(proc_t p)
1623{
1624 struct plimit *plimit = NULL;
1625
1626 proc_lock(p);
1627 plimit = smr_serialized_load(&p->p_limit);
1628 smr_clear_store(&p->p_limit);
1629 proc_unlock(p);
1630
1631 proc_limit_release(plimit);
1632}
1633
1634/*
1635 * proc_limitblock/unblock are used to serialize access to plimit
1636 * from concurrent threads within the same process.
1637 * Callers must be holding the proc lock to enter, return with
1638 * the proc lock locked
1639 */
1640static void
1641proc_limitblock(proc_t p)
1642{
1643 lck_mtx_assert(lck: &p->p_mlock, LCK_MTX_ASSERT_OWNED);
1644
1645 while (p->p_lflag & P_LLIMCHANGE) {
1646 p->p_lflag |= P_LLIMWAIT;
1647 msleep(chan: &p->p_limit, mtx: &p->p_mlock, pri: 0, wmesg: "proc_limitblock", NULL);
1648 }
1649 p->p_lflag |= P_LLIMCHANGE;
1650}
1651
1652/*
1653 * Callers must be holding the proc lock to enter, return with
1654 * the proc lock locked
1655 */
1656static void
1657proc_limitunblock(proc_t p)
1658{
1659 lck_mtx_assert(lck: &p->p_mlock, LCK_MTX_ASSERT_OWNED);
1660
1661 p->p_lflag &= ~P_LLIMCHANGE;
1662 if (p->p_lflag & P_LLIMWAIT) {
1663 p->p_lflag &= ~P_LLIMWAIT;
1664 wakeup(chan: &p->p_limit);
1665 }
1666}
1667
1668/*
1669 * Perform an rlimit update (as defined by the arbitrary `update` function).
1670 *
1671 * Because plimits are accessed without holding any locks,
1672 * with only a hazard reference, the struct plimit is always
1673 * copied, updated, and replaced, to implement a const value type.
1674 */
1675static void
1676proc_limitupdate(proc_t p, bool unblock, void (^update)(struct plimit *))
1677{
1678 struct plimit *cur_plim;
1679 struct plimit *copy_plim;
1680
1681 copy_plim = zalloc_flags(plimit_zone, Z_WAITOK | Z_ZERO | Z_NOFAIL);
1682
1683 proc_lock(p);
1684
1685 cur_plim = smr_serialized_load(&p->p_limit);
1686
1687 os_ref_init_count(&copy_plim->pl_refcnt, &rlimit_refgrp, 1);
1688 bcopy(src: cur_plim->pl_rlimit, dst: copy_plim->pl_rlimit,
1689 n: sizeof(struct rlimit) * RLIM_NLIMITS);
1690
1691 update(copy_plim);
1692
1693 smr_serialized_store(&p->p_limit, copy_plim);
1694
1695 if (unblock) {
1696 proc_limitunblock(p);
1697 }
1698 proc_unlock(p);
1699
1700 proc_limit_release(plimit: cur_plim);
1701}
1702
1703static int
1704iopolicysys_disk(struct proc *p, int cmd, int scope, int policy, struct _iopol_param_t *iop_param);
1705static int
1706iopolicysys_vfs_hfs_case_sensitivity(struct proc *p, int cmd, int scope, int policy, struct _iopol_param_t *iop_param);
1707static int
1708iopolicysys_vfs_atime_updates(struct proc *p, int cmd, int scope, int policy, struct _iopol_param_t *iop_param);
1709static int
1710iopolicysys_vfs_statfs_no_data_volume(struct proc *p, int cmd, int scope, int policy, struct _iopol_param_t *iop_param);
1711static int
1712iopolicysys_vfs_trigger_resolve(struct proc *p, int cmd, int scope, int policy, struct _iopol_param_t *iop_param);
1713static int
1714iopolicysys_vfs_ignore_content_protection(struct proc *p, int cmd, int scope, int policy, struct _iopol_param_t *iop_param);
1715static int
1716iopolicysys_vfs_ignore_node_permissions(struct proc *p, int cmd, int scope, int policy, struct _iopol_param_t *ipo_param);
1717static int
1718iopolicysys_vfs_skip_mtime_update(struct proc *p, int cmd, int scope, int policy, struct _iopol_param_t *iop_param);
1719static int
1720iopolicysys_vfs_allow_lowspace_writes(struct proc *p, int cmd, int scope, int policy, struct _iopol_param_t *iop_param);
1721static int
1722iopolicysys_vfs_disallow_rw_for_o_evtonly(struct proc *p, int cmd, int scope, int policy, struct _iopol_param_t *iop_param);
1723static int iopolicysys_vfs_altlink(struct proc *p, int cmd, int scope, int policy, struct _iopol_param_t *iop_param);
1724static int iopolicysys_vfs_nocache_write_fs_blksize(struct proc *p, int cmd, int scope, int policy, struct _iopol_param_t *iop_param);
1725
1726/*
1727 * iopolicysys
1728 *
1729 * Description: System call MUX for use in manipulating I/O policy attributes of the current process or thread
1730 *
1731 * Parameters: cmd Policy command
1732 * arg Pointer to policy arguments
1733 *
1734 * Returns: 0 Success
1735 * EINVAL Invalid command or invalid policy arguments
1736 *
1737 */
1738int
1739iopolicysys(struct proc *p, struct iopolicysys_args *uap, int32_t *retval)
1740{
1741 int error = 0;
1742 struct _iopol_param_t iop_param;
1743
1744 if ((error = copyin(uap->arg, &iop_param, sizeof(iop_param))) != 0) {
1745 goto out;
1746 }
1747
1748 switch (iop_param.iop_iotype) {
1749 case IOPOL_TYPE_DISK:
1750 error = iopolicysys_disk(p, cmd: uap->cmd, scope: iop_param.iop_scope, policy: iop_param.iop_policy, iop_param: &iop_param);
1751 if (error == EIDRM) {
1752 *retval = -2;
1753 error = 0;
1754 }
1755 if (error) {
1756 goto out;
1757 }
1758 break;
1759 case IOPOL_TYPE_VFS_HFS_CASE_SENSITIVITY:
1760 error = iopolicysys_vfs_hfs_case_sensitivity(p, cmd: uap->cmd, scope: iop_param.iop_scope, policy: iop_param.iop_policy, iop_param: &iop_param);
1761 if (error) {
1762 goto out;
1763 }
1764 break;
1765 case IOPOL_TYPE_VFS_ATIME_UPDATES:
1766 error = iopolicysys_vfs_atime_updates(p, cmd: uap->cmd, scope: iop_param.iop_scope, policy: iop_param.iop_policy, iop_param: &iop_param);
1767 if (error) {
1768 goto out;
1769 }
1770 break;
1771 case IOPOL_TYPE_VFS_MATERIALIZE_DATALESS_FILES:
1772 error = iopolicysys_vfs_materialize_dataless_files(p, cmd: uap->cmd, scope: iop_param.iop_scope, policy: iop_param.iop_policy, iop_param: &iop_param);
1773 if (error) {
1774 goto out;
1775 }
1776 break;
1777 case IOPOL_TYPE_VFS_STATFS_NO_DATA_VOLUME:
1778 error = iopolicysys_vfs_statfs_no_data_volume(p, cmd: uap->cmd, scope: iop_param.iop_scope, policy: iop_param.iop_policy, iop_param: &iop_param);
1779 if (error) {
1780 goto out;
1781 }
1782 break;
1783 case IOPOL_TYPE_VFS_TRIGGER_RESOLVE:
1784 error = iopolicysys_vfs_trigger_resolve(p, cmd: uap->cmd, scope: iop_param.iop_scope, policy: iop_param.iop_policy, iop_param: &iop_param);
1785 if (error) {
1786 goto out;
1787 }
1788 break;
1789 case IOPOL_TYPE_VFS_IGNORE_CONTENT_PROTECTION:
1790 error = iopolicysys_vfs_ignore_content_protection(p, cmd: uap->cmd, scope: iop_param.iop_scope, policy: iop_param.iop_policy, iop_param: &iop_param);
1791 if (error) {
1792 goto out;
1793 }
1794 break;
1795 case IOPOL_TYPE_VFS_IGNORE_PERMISSIONS:
1796 error = iopolicysys_vfs_ignore_node_permissions(p, cmd: uap->cmd, scope: iop_param.iop_scope, policy: iop_param.iop_policy, ipo_param: &iop_param);
1797 if (error) {
1798 goto out;
1799 }
1800 break;
1801 case IOPOL_TYPE_VFS_SKIP_MTIME_UPDATE:
1802 error = iopolicysys_vfs_skip_mtime_update(p, cmd: uap->cmd, scope: iop_param.iop_scope, policy: iop_param.iop_policy, iop_param: &iop_param);
1803 if (error) {
1804 goto out;
1805 }
1806 break;
1807 case IOPOL_TYPE_VFS_ALLOW_LOW_SPACE_WRITES:
1808 error = iopolicysys_vfs_allow_lowspace_writes(p, cmd: uap->cmd, scope: iop_param.iop_scope, policy: iop_param.iop_policy, iop_param: &iop_param);
1809 if (error) {
1810 goto out;
1811 }
1812 break;
1813 case IOPOL_TYPE_VFS_DISALLOW_RW_FOR_O_EVTONLY:
1814 error = iopolicysys_vfs_disallow_rw_for_o_evtonly(p, cmd: uap->cmd, scope: iop_param.iop_scope, policy: iop_param.iop_policy, iop_param: &iop_param);
1815 if (error) {
1816 goto out;
1817 }
1818 break;
1819 case IOPOL_TYPE_VFS_ALTLINK:
1820 error = iopolicysys_vfs_altlink(p, cmd: uap->cmd, scope: iop_param.iop_scope, policy: iop_param.iop_policy, iop_param: &iop_param);
1821 if (error) {
1822 goto out;
1823 }
1824 break;
1825 case IOPOL_TYPE_VFS_NOCACHE_WRITE_FS_BLKSIZE:
1826 error = iopolicysys_vfs_nocache_write_fs_blksize(p, cmd: uap->cmd, scope: iop_param.iop_scope, policy: iop_param.iop_policy, iop_param: &iop_param);
1827 if (error) {
1828 goto out;
1829 }
1830 break;
1831
1832 default:
1833 error = EINVAL;
1834 goto out;
1835 }
1836
1837 /* Individual iotype handlers are expected to update iop_param, if requested with a GET command */
1838 if (uap->cmd == IOPOL_CMD_GET) {
1839 error = copyout((caddr_t)&iop_param, uap->arg, sizeof(iop_param));
1840 if (error) {
1841 goto out;
1842 }
1843 }
1844
1845out:
1846 return error;
1847}
1848
1849static int
1850iopolicysys_disk(struct proc *p __unused, int cmd, int scope, int policy, struct _iopol_param_t *iop_param)
1851{
1852 int error = 0;
1853 thread_t thread;
1854 int policy_flavor;
1855
1856 /* Validate scope */
1857 switch (scope) {
1858 case IOPOL_SCOPE_PROCESS:
1859 thread = THREAD_NULL;
1860 policy_flavor = TASK_POLICY_IOPOL;
1861 break;
1862
1863 case IOPOL_SCOPE_THREAD:
1864 thread = current_thread();
1865 policy_flavor = TASK_POLICY_IOPOL;
1866
1867 /* Not allowed to combine QoS and (non-PASSIVE) IO policy, doing so strips the QoS */
1868 if (cmd == IOPOL_CMD_SET && thread_has_qos_policy(thread)) {
1869 switch (policy) {
1870 case IOPOL_DEFAULT:
1871 case IOPOL_PASSIVE:
1872 break;
1873 case IOPOL_UTILITY:
1874 case IOPOL_THROTTLE:
1875 case IOPOL_IMPORTANT:
1876 case IOPOL_STANDARD:
1877 if (!thread_is_static_param(thread)) {
1878 thread_remove_qos_policy(thread);
1879 /*
1880 * This is not an error case, this is to return a marker to user-space that
1881 * we stripped the thread of its QoS class.
1882 */
1883 error = EIDRM;
1884 break;
1885 }
1886 OS_FALLTHROUGH;
1887 default:
1888 error = EINVAL;
1889 goto out;
1890 }
1891 }
1892 break;
1893
1894 case IOPOL_SCOPE_DARWIN_BG:
1895#if !defined(XNU_TARGET_OS_OSX)
1896 /* We don't want this on platforms outside of macOS as BG is always IOPOL_THROTTLE */
1897 error = ENOTSUP;
1898 goto out;
1899#else /* !defined(XNU_TARGET_OS_OSX) */
1900 thread = THREAD_NULL;
1901 policy_flavor = TASK_POLICY_DARWIN_BG_IOPOL;
1902 break;
1903#endif /* !defined(XNU_TARGET_OS_OSX) */
1904
1905 default:
1906 error = EINVAL;
1907 goto out;
1908 }
1909
1910 /* Validate policy */
1911 if (cmd == IOPOL_CMD_SET) {
1912 switch (policy) {
1913 case IOPOL_DEFAULT:
1914 if (scope == IOPOL_SCOPE_DARWIN_BG) {
1915 /* the current default BG throttle level is UTILITY */
1916 policy = IOPOL_UTILITY;
1917 } else {
1918 policy = IOPOL_IMPORTANT;
1919 }
1920 break;
1921 case IOPOL_UTILITY:
1922 /* fall-through */
1923 case IOPOL_THROTTLE:
1924 /* These levels are OK */
1925 break;
1926 case IOPOL_IMPORTANT:
1927 /* fall-through */
1928 case IOPOL_STANDARD:
1929 /* fall-through */
1930 case IOPOL_PASSIVE:
1931 if (scope == IOPOL_SCOPE_DARWIN_BG) {
1932 /* These levels are invalid for BG */
1933 error = EINVAL;
1934 goto out;
1935 } else {
1936 /* OK for other scopes */
1937 }
1938 break;
1939 default:
1940 error = EINVAL;
1941 goto out;
1942 }
1943 }
1944
1945 /* Perform command */
1946 switch (cmd) {
1947 case IOPOL_CMD_SET:
1948 if (thread != THREAD_NULL) {
1949 proc_set_thread_policy(thread, TASK_POLICY_INTERNAL, flavor: policy_flavor, value: policy);
1950 } else {
1951 proc_set_task_policy(task: current_task(), TASK_POLICY_INTERNAL, flavor: policy_flavor, value: policy);
1952 }
1953 break;
1954 case IOPOL_CMD_GET:
1955 if (thread != THREAD_NULL) {
1956 policy = proc_get_thread_policy(thread, TASK_POLICY_INTERNAL, flavor: policy_flavor);
1957 } else {
1958 policy = proc_get_task_policy(task: current_task(), TASK_POLICY_INTERNAL, flavor: policy_flavor);
1959 }
1960 iop_param->iop_policy = policy;
1961 break;
1962 default:
1963 error = EINVAL; /* unknown command */
1964 break;
1965 }
1966
1967out:
1968 return error;
1969}
1970
1971static int
1972iopolicysys_vfs_hfs_case_sensitivity(struct proc *p, int cmd, int scope, int policy, struct _iopol_param_t *iop_param)
1973{
1974 int error = 0;
1975
1976 /* Validate scope */
1977 switch (scope) {
1978 case IOPOL_SCOPE_PROCESS:
1979 /* Only process OK */
1980 break;
1981 default:
1982 error = EINVAL;
1983 goto out;
1984 }
1985
1986 /* Validate policy */
1987 if (cmd == IOPOL_CMD_SET) {
1988 switch (policy) {
1989 case IOPOL_VFS_HFS_CASE_SENSITIVITY_DEFAULT:
1990 /* fall-through */
1991 case IOPOL_VFS_HFS_CASE_SENSITIVITY_FORCE_CASE_SENSITIVE:
1992 /* These policies are OK */
1993 break;
1994 default:
1995 error = EINVAL;
1996 goto out;
1997 }
1998 }
1999
2000 /* Perform command */
2001 switch (cmd) {
2002 case IOPOL_CMD_SET:
2003 if (0 == kauth_cred_issuser(cred: kauth_cred_get())) {
2004 /* If it's a non-root process, it needs to have the entitlement to set the policy */
2005 boolean_t entitled = FALSE;
2006 entitled = IOCurrentTaskHasEntitlement(entitlement: "com.apple.private.iopol.case_sensitivity");
2007 if (!entitled) {
2008 error = EPERM;
2009 goto out;
2010 }
2011 }
2012
2013 switch (policy) {
2014 case IOPOL_VFS_HFS_CASE_SENSITIVITY_DEFAULT:
2015 OSBitAndAtomic16(mask: ~((uint32_t)P_VFS_IOPOLICY_FORCE_HFS_CASE_SENSITIVITY), address: &p->p_vfs_iopolicy);
2016 break;
2017 case IOPOL_VFS_HFS_CASE_SENSITIVITY_FORCE_CASE_SENSITIVE:
2018 OSBitOrAtomic16(mask: (uint32_t)P_VFS_IOPOLICY_FORCE_HFS_CASE_SENSITIVITY, address: &p->p_vfs_iopolicy);
2019 break;
2020 default:
2021 error = EINVAL;
2022 goto out;
2023 }
2024
2025 break;
2026 case IOPOL_CMD_GET:
2027 iop_param->iop_policy = (p->p_vfs_iopolicy & P_VFS_IOPOLICY_FORCE_HFS_CASE_SENSITIVITY)
2028 ? IOPOL_VFS_HFS_CASE_SENSITIVITY_FORCE_CASE_SENSITIVE
2029 : IOPOL_VFS_HFS_CASE_SENSITIVITY_DEFAULT;
2030 break;
2031 default:
2032 error = EINVAL; /* unknown command */
2033 break;
2034 }
2035
2036out:
2037 return error;
2038}
2039
2040static inline int
2041get_thread_atime_policy(struct uthread *ut)
2042{
2043 return (ut->uu_flag & UT_ATIME_UPDATE) ? IOPOL_ATIME_UPDATES_OFF : IOPOL_ATIME_UPDATES_DEFAULT;
2044}
2045
2046static inline void
2047set_thread_atime_policy(struct uthread *ut, int policy)
2048{
2049 if (policy == IOPOL_ATIME_UPDATES_OFF) {
2050 ut->uu_flag |= UT_ATIME_UPDATE;
2051 } else {
2052 ut->uu_flag &= ~UT_ATIME_UPDATE;
2053 }
2054}
2055
2056static inline void
2057set_task_atime_policy(struct proc *p, int policy)
2058{
2059 if (policy == IOPOL_ATIME_UPDATES_OFF) {
2060 OSBitOrAtomic16(mask: (uint16_t)P_VFS_IOPOLICY_ATIME_UPDATES, address: &p->p_vfs_iopolicy);
2061 } else {
2062 OSBitAndAtomic16(mask: ~((uint16_t)P_VFS_IOPOLICY_ATIME_UPDATES), address: &p->p_vfs_iopolicy);
2063 }
2064}
2065
2066static inline int
2067get_task_atime_policy(struct proc *p)
2068{
2069 return (p->p_vfs_iopolicy & P_VFS_IOPOLICY_ATIME_UPDATES) ? IOPOL_ATIME_UPDATES_OFF : IOPOL_ATIME_UPDATES_DEFAULT;
2070}
2071
2072static int
2073iopolicysys_vfs_atime_updates(struct proc *p __unused, int cmd, int scope, int policy, struct _iopol_param_t *iop_param)
2074{
2075 int error = 0;
2076 thread_t thread;
2077
2078 /* Validate scope */
2079 switch (scope) {
2080 case IOPOL_SCOPE_THREAD:
2081 thread = current_thread();
2082 break;
2083 case IOPOL_SCOPE_PROCESS:
2084 thread = THREAD_NULL;
2085 break;
2086 default:
2087 error = EINVAL;
2088 goto out;
2089 }
2090
2091 /* Validate policy */
2092 if (cmd == IOPOL_CMD_SET) {
2093 switch (policy) {
2094 case IOPOL_ATIME_UPDATES_DEFAULT:
2095 case IOPOL_ATIME_UPDATES_OFF:
2096 break;
2097 default:
2098 error = EINVAL;
2099 goto out;
2100 }
2101 }
2102
2103 /* Perform command */
2104 switch (cmd) {
2105 case IOPOL_CMD_SET:
2106 if (thread != THREAD_NULL) {
2107 set_thread_atime_policy(ut: get_bsdthread_info(thread), policy);
2108 } else {
2109 set_task_atime_policy(p, policy);
2110 }
2111 break;
2112 case IOPOL_CMD_GET:
2113 if (thread != THREAD_NULL) {
2114 policy = get_thread_atime_policy(ut: get_bsdthread_info(thread));
2115 } else {
2116 policy = get_task_atime_policy(p);
2117 }
2118 iop_param->iop_policy = policy;
2119 break;
2120 default:
2121 error = EINVAL; /* unknown command */
2122 break;
2123 }
2124
2125out:
2126 return error;
2127}
2128
2129static inline int
2130get_thread_materialize_policy(struct uthread *ut)
2131{
2132 if (ut->uu_flag & UT_NSPACE_NODATALESSFAULTS) {
2133 return IOPOL_MATERIALIZE_DATALESS_FILES_OFF;
2134 } else if (ut->uu_flag & UT_NSPACE_FORCEDATALESSFAULTS) {
2135 return IOPOL_MATERIALIZE_DATALESS_FILES_ON;
2136 }
2137 /* Default thread behavior is "inherit process behavior". */
2138 return IOPOL_MATERIALIZE_DATALESS_FILES_DEFAULT;
2139}
2140
2141static inline void
2142set_thread_materialize_policy(struct uthread *ut, int policy)
2143{
2144 if (policy == IOPOL_MATERIALIZE_DATALESS_FILES_OFF) {
2145 ut->uu_flag &= ~UT_NSPACE_FORCEDATALESSFAULTS;
2146 ut->uu_flag |= UT_NSPACE_NODATALESSFAULTS;
2147 } else if (policy == IOPOL_MATERIALIZE_DATALESS_FILES_ON) {
2148 ut->uu_flag &= ~UT_NSPACE_NODATALESSFAULTS;
2149 ut->uu_flag |= UT_NSPACE_FORCEDATALESSFAULTS;
2150 } else {
2151 ut->uu_flag &= ~(UT_NSPACE_NODATALESSFAULTS | UT_NSPACE_FORCEDATALESSFAULTS);
2152 }
2153}
2154
2155static inline void
2156set_proc_materialize_policy(struct proc *p, int policy)
2157{
2158 if (policy == IOPOL_MATERIALIZE_DATALESS_FILES_DEFAULT) {
2159 /*
2160 * Caller has specified "use the default policy".
2161 * The default policy is to NOT materialize dataless
2162 * files.
2163 */
2164 policy = IOPOL_MATERIALIZE_DATALESS_FILES_OFF;
2165 }
2166 if (policy == IOPOL_MATERIALIZE_DATALESS_FILES_ON) {
2167 OSBitOrAtomic16(mask: (uint16_t)P_VFS_IOPOLICY_MATERIALIZE_DATALESS_FILES, address: &p->p_vfs_iopolicy);
2168 } else {
2169 OSBitAndAtomic16(mask: ~((uint16_t)P_VFS_IOPOLICY_MATERIALIZE_DATALESS_FILES), address: &p->p_vfs_iopolicy);
2170 }
2171}
2172
2173static int
2174get_proc_materialize_policy(struct proc *p)
2175{
2176 return (p->p_vfs_iopolicy & P_VFS_IOPOLICY_MATERIALIZE_DATALESS_FILES) ? IOPOL_MATERIALIZE_DATALESS_FILES_ON : IOPOL_MATERIALIZE_DATALESS_FILES_OFF;
2177}
2178
2179int
2180iopolicysys_vfs_materialize_dataless_files(struct proc *p __unused, int cmd, int scope, int policy, struct _iopol_param_t *iop_param)
2181{
2182 int error = 0;
2183 thread_t thread;
2184
2185 /* Validate scope */
2186 switch (scope) {
2187 case IOPOL_SCOPE_THREAD:
2188 thread = current_thread();
2189 break;
2190 case IOPOL_SCOPE_PROCESS:
2191 thread = THREAD_NULL;
2192 break;
2193 default:
2194 error = EINVAL;
2195 goto out;
2196 }
2197
2198 /* Validate policy */
2199 if (cmd == IOPOL_CMD_SET) {
2200 switch (policy) {
2201 case IOPOL_MATERIALIZE_DATALESS_FILES_DEFAULT:
2202 case IOPOL_MATERIALIZE_DATALESS_FILES_OFF:
2203 case IOPOL_MATERIALIZE_DATALESS_FILES_ON:
2204 break;
2205 default:
2206 error = EINVAL;
2207 goto out;
2208 }
2209 }
2210
2211 /* Perform command */
2212 switch (cmd) {
2213 case IOPOL_CMD_SET:
2214 if (thread != THREAD_NULL) {
2215 set_thread_materialize_policy(ut: get_bsdthread_info(thread), policy);
2216 } else {
2217 set_proc_materialize_policy(p, policy);
2218 }
2219 break;
2220 case IOPOL_CMD_GET:
2221 if (thread != THREAD_NULL) {
2222 policy = get_thread_materialize_policy(ut: get_bsdthread_info(thread));
2223 } else {
2224 policy = get_proc_materialize_policy(p);
2225 }
2226 iop_param->iop_policy = policy;
2227 break;
2228 default:
2229 error = EINVAL; /* unknown command */
2230 break;
2231 }
2232
2233out:
2234 return error;
2235}
2236
2237static int
2238iopolicysys_vfs_statfs_no_data_volume(struct proc *p __unused, int cmd,
2239 int scope, int policy, struct _iopol_param_t *iop_param)
2240{
2241 int error = 0;
2242
2243 /* Validate scope */
2244 switch (scope) {
2245 case IOPOL_SCOPE_PROCESS:
2246 /* Only process OK */
2247 break;
2248 default:
2249 error = EINVAL;
2250 goto out;
2251 }
2252
2253 /* Validate policy */
2254 if (cmd == IOPOL_CMD_SET) {
2255 switch (policy) {
2256 case IOPOL_VFS_STATFS_NO_DATA_VOLUME_DEFAULT:
2257 /* fall-through */
2258 case IOPOL_VFS_STATFS_FORCE_NO_DATA_VOLUME:
2259 /* These policies are OK */
2260 break;
2261 default:
2262 error = EINVAL;
2263 goto out;
2264 }
2265 }
2266
2267 /* Perform command */
2268 switch (cmd) {
2269 case IOPOL_CMD_SET:
2270 if (0 == kauth_cred_issuser(cred: kauth_cred_get())) {
2271 /* If it's a non-root process, it needs to have the entitlement to set the policy */
2272 boolean_t entitled = FALSE;
2273 entitled = IOCurrentTaskHasEntitlement(entitlement: "com.apple.private.iopol.case_sensitivity");
2274 if (!entitled) {
2275 error = EPERM;
2276 goto out;
2277 }
2278 }
2279
2280 switch (policy) {
2281 case IOPOL_VFS_STATFS_NO_DATA_VOLUME_DEFAULT:
2282 OSBitAndAtomic16(mask: ~((uint32_t)P_VFS_IOPOLICY_STATFS_NO_DATA_VOLUME), address: &p->p_vfs_iopolicy);
2283 break;
2284 case IOPOL_VFS_STATFS_FORCE_NO_DATA_VOLUME:
2285 OSBitOrAtomic16(mask: (uint32_t)P_VFS_IOPOLICY_STATFS_NO_DATA_VOLUME, address: &p->p_vfs_iopolicy);
2286 break;
2287 default:
2288 error = EINVAL;
2289 goto out;
2290 }
2291
2292 break;
2293 case IOPOL_CMD_GET:
2294 iop_param->iop_policy = (p->p_vfs_iopolicy & P_VFS_IOPOLICY_STATFS_NO_DATA_VOLUME)
2295 ? IOPOL_VFS_STATFS_FORCE_NO_DATA_VOLUME
2296 : IOPOL_VFS_STATFS_NO_DATA_VOLUME_DEFAULT;
2297 break;
2298 default:
2299 error = EINVAL; /* unknown command */
2300 break;
2301 }
2302
2303out:
2304 return error;
2305}
2306
2307static int
2308iopolicysys_vfs_trigger_resolve(struct proc *p __unused, int cmd,
2309 int scope, int policy, struct _iopol_param_t *iop_param)
2310{
2311 int error = 0;
2312
2313 /* Validate scope */
2314 switch (scope) {
2315 case IOPOL_SCOPE_PROCESS:
2316 /* Only process OK */
2317 break;
2318 default:
2319 error = EINVAL;
2320 goto out;
2321 }
2322
2323 /* Validate policy */
2324 if (cmd == IOPOL_CMD_SET) {
2325 switch (policy) {
2326 case IOPOL_VFS_TRIGGER_RESOLVE_DEFAULT:
2327 /* fall-through */
2328 case IOPOL_VFS_TRIGGER_RESOLVE_OFF:
2329 /* These policies are OK */
2330 break;
2331 default:
2332 error = EINVAL;
2333 goto out;
2334 }
2335 }
2336
2337 /* Perform command */
2338 switch (cmd) {
2339 case IOPOL_CMD_SET:
2340 switch (policy) {
2341 case IOPOL_VFS_TRIGGER_RESOLVE_DEFAULT:
2342 OSBitAndAtomic16(mask: ~((uint32_t)P_VFS_IOPOLICY_TRIGGER_RESOLVE_DISABLE), address: &p->p_vfs_iopolicy);
2343 break;
2344 case IOPOL_VFS_TRIGGER_RESOLVE_OFF:
2345 OSBitOrAtomic16(mask: (uint32_t)P_VFS_IOPOLICY_TRIGGER_RESOLVE_DISABLE, address: &p->p_vfs_iopolicy);
2346 break;
2347 default:
2348 error = EINVAL;
2349 goto out;
2350 }
2351
2352 break;
2353 case IOPOL_CMD_GET:
2354 iop_param->iop_policy = (p->p_vfs_iopolicy & P_VFS_IOPOLICY_TRIGGER_RESOLVE_DISABLE)
2355 ? IOPOL_VFS_TRIGGER_RESOLVE_OFF
2356 : IOPOL_VFS_TRIGGER_RESOLVE_DEFAULT;
2357 break;
2358 default:
2359 error = EINVAL; /* unknown command */
2360 break;
2361 }
2362
2363out:
2364 return error;
2365}
2366
2367static int
2368iopolicysys_vfs_ignore_content_protection(struct proc *p, int cmd, int scope,
2369 int policy, struct _iopol_param_t *iop_param)
2370{
2371 int error = 0;
2372
2373 /* Validate scope */
2374 switch (scope) {
2375 case IOPOL_SCOPE_PROCESS:
2376 /* Only process OK */
2377 break;
2378 default:
2379 error = EINVAL;
2380 goto out;
2381 }
2382
2383 /* Validate policy */
2384 if (cmd == IOPOL_CMD_SET) {
2385 switch (policy) {
2386 case IOPOL_VFS_CONTENT_PROTECTION_DEFAULT:
2387 OS_FALLTHROUGH;
2388 case IOPOL_VFS_CONTENT_PROTECTION_IGNORE:
2389 /* These policies are OK */
2390 break;
2391 default:
2392 error = EINVAL;
2393 goto out;
2394 }
2395 }
2396
2397 /* Perform command */
2398 switch (cmd) {
2399 case IOPOL_CMD_SET:
2400 if (0 == kauth_cred_issuser(cred: kauth_cred_get())) {
2401 /* If it's a non-root process, it needs to have the entitlement to set the policy */
2402 boolean_t entitled = FALSE;
2403 entitled = IOCurrentTaskHasEntitlement(entitlement: "com.apple.private.iopol.case_sensitivity");
2404 if (!entitled) {
2405 error = EPERM;
2406 goto out;
2407 }
2408 }
2409
2410 switch (policy) {
2411 case IOPOL_VFS_CONTENT_PROTECTION_DEFAULT:
2412 os_atomic_andnot(&p->p_vfs_iopolicy, P_VFS_IOPOLICY_IGNORE_CONTENT_PROTECTION, relaxed);
2413 break;
2414 case IOPOL_VFS_CONTENT_PROTECTION_IGNORE:
2415 os_atomic_or(&p->p_vfs_iopolicy, P_VFS_IOPOLICY_IGNORE_CONTENT_PROTECTION, relaxed);
2416 break;
2417 default:
2418 error = EINVAL;
2419 goto out;
2420 }
2421
2422 break;
2423 case IOPOL_CMD_GET:
2424 iop_param->iop_policy = (os_atomic_load(&p->p_vfs_iopolicy, relaxed) & P_VFS_IOPOLICY_IGNORE_CONTENT_PROTECTION)
2425 ? IOPOL_VFS_CONTENT_PROTECTION_IGNORE
2426 : IOPOL_VFS_CONTENT_PROTECTION_DEFAULT;
2427 break;
2428 default:
2429 error = EINVAL; /* unknown command */
2430 break;
2431 }
2432
2433out:
2434 return error;
2435}
2436
2437#define AUTHORIZED_ACCESS_ENTITLEMENT \
2438 "com.apple.private.vfs.authorized-access"
2439int
2440iopolicysys_vfs_ignore_node_permissions(struct proc *p, int cmd, int scope,
2441 int policy, __unused struct _iopol_param_t *iop_param)
2442{
2443 int error = EINVAL;
2444
2445 switch (scope) {
2446 case IOPOL_SCOPE_PROCESS:
2447 break;
2448 default:
2449 goto out;
2450 }
2451
2452 switch (cmd) {
2453 case IOPOL_CMD_GET:
2454 policy = os_atomic_load(&p->p_vfs_iopolicy, relaxed) & P_VFS_IOPOLICY_IGNORE_NODE_PERMISSIONS ?
2455 IOPOL_VFS_IGNORE_PERMISSIONS_ON : IOPOL_VFS_IGNORE_PERMISSIONS_OFF;
2456 iop_param->iop_policy = policy;
2457 goto out_ok;
2458 case IOPOL_CMD_SET:
2459 /* SET is handled after the switch */
2460 break;
2461 default:
2462 goto out;
2463 }
2464
2465 if (!IOCurrentTaskHasEntitlement(AUTHORIZED_ACCESS_ENTITLEMENT)) {
2466 error = EPERM;
2467 goto out;
2468 }
2469
2470 switch (policy) {
2471 case IOPOL_VFS_IGNORE_PERMISSIONS_OFF:
2472 os_atomic_andnot(&p->p_vfs_iopolicy, P_VFS_IOPOLICY_IGNORE_NODE_PERMISSIONS, relaxed);
2473 break;
2474 case IOPOL_VFS_IGNORE_PERMISSIONS_ON:
2475 os_atomic_or(&p->p_vfs_iopolicy, P_VFS_IOPOLICY_IGNORE_NODE_PERMISSIONS, relaxed);
2476 break;
2477 default:
2478 break;
2479 }
2480
2481out_ok:
2482 error = 0;
2483out:
2484 return error;
2485}
2486
2487#define SKIP_MTIME_UPDATE_ENTITLEMENT \
2488 "com.apple.private.vfs.skip-mtime-updates"
2489int
2490iopolicysys_vfs_skip_mtime_update(struct proc *p, int cmd, int scope,
2491 int policy, __unused struct _iopol_param_t *iop_param)
2492{
2493 int error = EINVAL;
2494
2495 switch (scope) {
2496 case IOPOL_SCOPE_PROCESS:
2497 break;
2498 default:
2499 goto out;
2500 }
2501
2502 switch (cmd) {
2503 case IOPOL_CMD_GET:
2504 policy = os_atomic_load(&p->p_vfs_iopolicy, relaxed) & P_VFS_IOPOLICY_SKIP_MTIME_UPDATE ?
2505 IOPOL_VFS_SKIP_MTIME_UPDATE_ON : IOPOL_VFS_SKIP_MTIME_UPDATE_OFF;
2506 iop_param->iop_policy = policy;
2507 goto out_ok;
2508 case IOPOL_CMD_SET:
2509 break;
2510 default:
2511 break;
2512 }
2513
2514 if (!IOCurrentTaskHasEntitlement(SKIP_MTIME_UPDATE_ENTITLEMENT)) {
2515 error = EPERM;
2516 goto out;
2517 }
2518
2519 switch (policy) {
2520 case IOPOL_VFS_SKIP_MTIME_UPDATE_OFF:
2521 os_atomic_andnot(&p->p_vfs_iopolicy, P_VFS_IOPOLICY_SKIP_MTIME_UPDATE, relaxed);
2522 break;
2523 case IOPOL_VFS_SKIP_MTIME_UPDATE_ON:
2524 os_atomic_or(&p->p_vfs_iopolicy, P_VFS_IOPOLICY_SKIP_MTIME_UPDATE, relaxed);
2525 break;
2526 default:
2527 break;
2528 }
2529
2530out_ok:
2531 error = 0;
2532out:
2533 return error;
2534}
2535
2536#define ALLOW_LOW_SPACE_WRITES_ENTITLEMENT \
2537 "com.apple.private.vfs.allow-low-space-writes"
2538static int
2539iopolicysys_vfs_allow_lowspace_writes(struct proc *p, int cmd, int scope,
2540 int policy, __unused struct _iopol_param_t *iop_param)
2541{
2542 int error = EINVAL;
2543
2544 switch (scope) {
2545 case IOPOL_SCOPE_PROCESS:
2546 break;
2547 default:
2548 goto out;
2549 }
2550
2551 switch (cmd) {
2552 case IOPOL_CMD_GET:
2553 policy = os_atomic_load(&p->p_vfs_iopolicy, relaxed) & P_VFS_IOPOLICY_ALLOW_LOW_SPACE_WRITES ?
2554 IOPOL_VFS_ALLOW_LOW_SPACE_WRITES_ON : IOPOL_VFS_ALLOW_LOW_SPACE_WRITES_OFF;
2555 iop_param->iop_policy = policy;
2556 goto out_ok;
2557 case IOPOL_CMD_SET:
2558 break;
2559 default:
2560 break;
2561 }
2562
2563 if (!IOCurrentTaskHasEntitlement(ALLOW_LOW_SPACE_WRITES_ENTITLEMENT)) {
2564 error = EPERM;
2565 goto out;
2566 }
2567
2568 switch (policy) {
2569 case IOPOL_VFS_ALLOW_LOW_SPACE_WRITES_OFF:
2570 os_atomic_andnot(&p->p_vfs_iopolicy, P_VFS_IOPOLICY_ALLOW_LOW_SPACE_WRITES, relaxed);
2571 break;
2572 case IOPOL_VFS_ALLOW_LOW_SPACE_WRITES_ON:
2573 os_atomic_or(&p->p_vfs_iopolicy, P_VFS_IOPOLICY_ALLOW_LOW_SPACE_WRITES, relaxed);
2574 break;
2575 default:
2576 break;
2577 }
2578
2579out_ok:
2580 error = 0;
2581out:
2582 return error;
2583}
2584
2585static int
2586iopolicysys_vfs_disallow_rw_for_o_evtonly(struct proc *p, int cmd, int scope,
2587 int policy, __unused struct _iopol_param_t *iop_param)
2588{
2589 int error = EINVAL;
2590
2591 switch (scope) {
2592 case IOPOL_SCOPE_PROCESS:
2593 break;
2594 default:
2595 goto out;
2596 }
2597
2598 switch (cmd) {
2599 case IOPOL_CMD_GET:
2600 policy = (os_atomic_load(&p->p_vfs_iopolicy, relaxed) &
2601 P_VFS_IOPOLICY_DISALLOW_RW_FOR_O_EVTONLY) ?
2602 IOPOL_VFS_DISALLOW_RW_FOR_O_EVTONLY_ON :
2603 IOPOL_VFS_DISALLOW_RW_FOR_O_EVTONLY_DEFAULT;
2604 iop_param->iop_policy = policy;
2605 goto out_ok;
2606 case IOPOL_CMD_SET:
2607 break;
2608 default:
2609 goto out;
2610 }
2611
2612 /* Once set, we don't allow the process to clear it. */
2613 switch (policy) {
2614 case IOPOL_VFS_DISALLOW_RW_FOR_O_EVTONLY_ON:
2615 os_atomic_or(&p->p_vfs_iopolicy,
2616 P_VFS_IOPOLICY_DISALLOW_RW_FOR_O_EVTONLY, relaxed);
2617 break;
2618 default:
2619 goto out;
2620 }
2621
2622out_ok:
2623 error = 0;
2624out:
2625 return error;
2626}
2627
2628static int
2629iopolicysys_vfs_altlink(struct proc *p, int cmd, int scope, int policy,
2630 struct _iopol_param_t *iop_param)
2631{
2632 if (scope != IOPOL_SCOPE_PROCESS) {
2633 return EINVAL;
2634 }
2635
2636 if (cmd == IOPOL_CMD_GET) {
2637 policy = (os_atomic_load(&p->p_vfs_iopolicy, relaxed) & P_VFS_IOPOLICY_ALTLINK) ?
2638 IOPOL_VFS_ALTLINK_ENABLED : IOPOL_VFS_ALTLINK_DISABLED;
2639 iop_param->iop_policy = policy;
2640 return 0;
2641 }
2642
2643 /* Once set, we don't allow the process to clear it. */
2644 if (policy == IOPOL_VFS_ALTLINK_ENABLED) {
2645 os_atomic_or(&p->p_vfs_iopolicy, P_VFS_IOPOLICY_ALTLINK, relaxed);
2646 return 0;
2647 }
2648
2649 return EINVAL;
2650}
2651
2652static int
2653iopolicysys_vfs_nocache_write_fs_blksize(struct proc *p, int cmd, int scope, int policy,
2654 struct _iopol_param_t *iop_param)
2655{
2656 thread_t thread;
2657
2658 switch (scope) {
2659 case IOPOL_SCOPE_THREAD:
2660 thread = current_thread();
2661 break;
2662 case IOPOL_SCOPE_PROCESS:
2663 thread = THREAD_NULL;
2664 break;
2665 default:
2666 return EINVAL;
2667 }
2668
2669 if (cmd == IOPOL_CMD_GET) {
2670 if (thread != THREAD_NULL) {
2671 struct uthread *ut = get_bsdthread_info(thread);
2672 policy = ut->uu_flag & UT_FS_BLKSIZE_NOCACHE_WRITES ?
2673 IOPOL_VFS_NOCACHE_WRITE_FS_BLKSIZE_ON : IOPOL_VFS_NOCACHE_WRITE_FS_BLKSIZE_DEFAULT;
2674 } else {
2675 policy = (os_atomic_load(&p->p_vfs_iopolicy, relaxed) & P_VFS_IOPOLICY_NOCACHE_WRITE_FS_BLKSIZE) ?
2676 IOPOL_VFS_NOCACHE_WRITE_FS_BLKSIZE_ON : IOPOL_VFS_NOCACHE_WRITE_FS_BLKSIZE_DEFAULT;
2677 }
2678 iop_param->iop_policy = policy;
2679 return 0;
2680 }
2681
2682 /* Once set, we don't allow the process or thread to clear it. */
2683 if ((cmd == IOPOL_CMD_SET) && (policy == IOPOL_VFS_NOCACHE_WRITE_FS_BLKSIZE_ON)) {
2684#if 0
2685 if (thread != THREAD_NULL) {
2686 struct uthread *ut = get_bsdthread_info(thread);
2687 ut->uu_flag |= UT_FS_BLKSIZE_NOCACHE_WRITES;
2688 } else {
2689 os_atomic_or(&p->p_vfs_iopolicy, P_VFS_IOPOLICY_NOCACHE_WRITE_FS_BLKSIZE, relaxed);
2690 }
2691#endif
2692 return 0;
2693 }
2694
2695 return EINVAL;
2696}
2697
2698void
2699proc_apply_task_networkbg(int pid, thread_t thread)
2700{
2701 proc_t p = proc_find(pid);
2702
2703 if (p != PROC_NULL) {
2704 do_background_socket(p, thread);
2705 proc_rele(p);
2706 }
2707}
2708
2709void
2710gather_rusage_info(proc_t p, rusage_info_current *ru, int flavor)
2711{
2712 struct rusage_info_child *ri_child;
2713
2714 assert(p->p_stats != NULL);
2715 memset(s: ru, c: 0, n: sizeof(*ru));
2716 switch (flavor) {
2717 case RUSAGE_INFO_V6:
2718 /* Any P-specific resource counters are captured in fill_task_rusage. */
2719 OS_FALLTHROUGH;
2720
2721 case RUSAGE_INFO_V5:
2722#if __has_feature(ptrauth_calls)
2723 if (vm_shared_region_is_reslide(proc_task(p))) {
2724 ru->ri_flags |= RU_PROC_RUNS_RESLIDE;
2725 }
2726#endif /* __has_feature(ptrauth_calls) */
2727 OS_FALLTHROUGH;
2728
2729 case RUSAGE_INFO_V4:
2730 ru->ri_logical_writes = get_task_logical_writes(task: proc_task(p), false);
2731 ru->ri_lifetime_max_phys_footprint = get_task_phys_footprint_lifetime_max(proc_task(p));
2732#if CONFIG_LEDGER_INTERVAL_MAX
2733 ru->ri_interval_max_phys_footprint = get_task_phys_footprint_interval_max(proc_task(p), FALSE);
2734#endif
2735 OS_FALLTHROUGH;
2736
2737 case RUSAGE_INFO_V3:
2738 fill_task_qos_rusage(task: proc_task(p), ri: ru);
2739 fill_task_billed_usage(task: proc_task(p), ri: ru);
2740 OS_FALLTHROUGH;
2741
2742 case RUSAGE_INFO_V2:
2743 fill_task_io_rusage(task: proc_task(p), ri: ru);
2744 OS_FALLTHROUGH;
2745
2746 case RUSAGE_INFO_V1:
2747 /*
2748 * p->p_stats->ri_child statistics are protected under proc lock.
2749 */
2750 proc_lock(p);
2751
2752 ri_child = &(p->p_stats->ri_child);
2753 ru->ri_child_user_time = ri_child->ri_child_user_time;
2754 ru->ri_child_system_time = ri_child->ri_child_system_time;
2755 ru->ri_child_pkg_idle_wkups = ri_child->ri_child_pkg_idle_wkups;
2756 ru->ri_child_interrupt_wkups = ri_child->ri_child_interrupt_wkups;
2757 ru->ri_child_pageins = ri_child->ri_child_pageins;
2758 ru->ri_child_elapsed_abstime = ri_child->ri_child_elapsed_abstime;
2759
2760 proc_unlock(p);
2761 OS_FALLTHROUGH;
2762
2763 case RUSAGE_INFO_V0:
2764 proc_getexecutableuuid(p, (unsigned char *)&ru->ri_uuid, sizeof(ru->ri_uuid));
2765 fill_task_rusage(task: proc_task(p), ri: ru);
2766 ru->ri_proc_start_abstime = p->p_stats->ps_start;
2767 }
2768}
2769
2770int
2771proc_get_rusage(proc_t p, int flavor, user_addr_t buffer, __unused int is_zombie)
2772{
2773 rusage_info_current ri_current = {};
2774
2775 size_t size = 0;
2776
2777 switch (flavor) {
2778 case RUSAGE_INFO_V0:
2779 size = sizeof(struct rusage_info_v0);
2780 break;
2781
2782 case RUSAGE_INFO_V1:
2783 size = sizeof(struct rusage_info_v1);
2784 break;
2785
2786 case RUSAGE_INFO_V2:
2787 size = sizeof(struct rusage_info_v2);
2788 break;
2789
2790 case RUSAGE_INFO_V3:
2791 size = sizeof(struct rusage_info_v3);
2792 break;
2793
2794 case RUSAGE_INFO_V4:
2795 size = sizeof(struct rusage_info_v4);
2796 break;
2797
2798 case RUSAGE_INFO_V5:
2799 size = sizeof(struct rusage_info_v5);
2800 break;
2801
2802 case RUSAGE_INFO_V6:
2803 size = sizeof(struct rusage_info_v6);
2804 break;
2805 default:
2806 return EINVAL;
2807 }
2808
2809 if (size == 0) {
2810 return EINVAL;
2811 }
2812
2813 /*
2814 * If task is still alive, collect info from the live task itself.
2815 * Otherwise, look to the cached info in the zombie proc.
2816 */
2817 if (p->p_ru) {
2818 return copyout(&p->p_ru->ri, buffer, size);
2819 } else {
2820 gather_rusage_info(p, ru: &ri_current, flavor);
2821 ri_current.ri_proc_exit_abstime = 0;
2822 return copyout(&ri_current, buffer, size);
2823 }
2824}
2825
2826static int
2827mach_to_bsd_rv(int mach_rv)
2828{
2829 int bsd_rv = 0;
2830
2831 switch (mach_rv) {
2832 case KERN_SUCCESS:
2833 bsd_rv = 0;
2834 break;
2835 case KERN_INVALID_ARGUMENT:
2836 bsd_rv = EINVAL;
2837 break;
2838 default:
2839 panic("unknown error %#x", mach_rv);
2840 }
2841
2842 return bsd_rv;
2843}
2844
2845/*
2846 * Resource limit controls
2847 *
2848 * uap->flavor available flavors:
2849 *
2850 * RLIMIT_WAKEUPS_MONITOR
2851 * RLIMIT_CPU_USAGE_MONITOR
2852 * RLIMIT_THREAD_CPULIMITS
2853 * RLIMIT_FOOTPRINT_INTERVAL
2854 */
2855int
2856proc_rlimit_control(__unused struct proc *p, struct proc_rlimit_control_args *uap, __unused int32_t *retval)
2857{
2858 proc_t targetp;
2859 int error = 0;
2860 struct proc_rlimit_control_wakeupmon wakeupmon_args;
2861 uint32_t cpumon_flags;
2862 uint32_t cpulimits_flags;
2863 kauth_cred_t my_cred, target_cred;
2864#if CONFIG_LEDGER_INTERVAL_MAX
2865 uint32_t footprint_interval_flags;
2866 uint64_t interval_max_footprint;
2867#endif /* CONFIG_LEDGER_INTERVAL_MAX */
2868
2869 /* -1 implicitly means our own process (perhaps even the current thread for per-thread attributes) */
2870 if (uap->pid == -1) {
2871 targetp = proc_self();
2872 } else {
2873 targetp = proc_find(pid: uap->pid);
2874 }
2875
2876 /* proc_self() can return NULL for an exiting process */
2877 if (targetp == PROC_NULL) {
2878 return ESRCH;
2879 }
2880
2881 my_cred = kauth_cred_get();
2882 target_cred = kauth_cred_proc_ref(procp: targetp);
2883
2884 if (!kauth_cred_issuser(cred: my_cred) && kauth_cred_getruid(cred: my_cred) &&
2885 kauth_cred_getuid(cred: my_cred) != kauth_cred_getuid(cred: target_cred) &&
2886 kauth_cred_getruid(cred: my_cred) != kauth_cred_getuid(cred: target_cred)) {
2887 proc_rele(p: targetp);
2888 kauth_cred_unref(&target_cred);
2889 return EACCES;
2890 }
2891
2892 switch (uap->flavor) {
2893 case RLIMIT_WAKEUPS_MONITOR:
2894 if ((error = copyin(uap->arg, &wakeupmon_args, sizeof(wakeupmon_args))) != 0) {
2895 break;
2896 }
2897 if ((error = mach_to_bsd_rv(mach_rv: task_wakeups_monitor_ctl(task: proc_task(targetp), rate_hz: &wakeupmon_args.wm_flags,
2898 flags: &wakeupmon_args.wm_rate))) != 0) {
2899 break;
2900 }
2901 error = copyout(&wakeupmon_args, uap->arg, sizeof(wakeupmon_args));
2902 break;
2903 case RLIMIT_CPU_USAGE_MONITOR:
2904 cpumon_flags = (uint32_t)uap->arg; // XXX temporarily stashing flags in argp (12592127)
2905 error = mach_to_bsd_rv(mach_rv: task_cpu_usage_monitor_ctl(task: proc_task(targetp), flags: &cpumon_flags));
2906 break;
2907 case RLIMIT_THREAD_CPULIMITS:
2908 cpulimits_flags = (uint32_t)uap->arg; // only need a limited set of bits, pass in void * argument
2909
2910 if (uap->pid != -1) {
2911 error = EINVAL;
2912 break;
2913 }
2914
2915 uint8_t percent = 0;
2916 uint32_t ms_refill = 0;
2917 uint64_t ns_refill;
2918
2919 percent = (uint8_t)(cpulimits_flags & 0xffU); /* low 8 bits for percent */
2920 ms_refill = (cpulimits_flags >> 8) & 0xffffff; /* next 24 bits represent ms refill value */
2921 if (percent >= 100 || percent == 0) {
2922 error = EINVAL;
2923 break;
2924 }
2925
2926 ns_refill = ((uint64_t)ms_refill) * NSEC_PER_MSEC;
2927
2928 error = mach_to_bsd_rv(mach_rv: thread_set_cpulimit(THREAD_CPULIMIT_BLOCK, percentage: percent, interval_ns: ns_refill));
2929 break;
2930
2931#if CONFIG_LEDGER_INTERVAL_MAX
2932 case RLIMIT_FOOTPRINT_INTERVAL:
2933 footprint_interval_flags = (uint32_t)uap->arg; // XXX temporarily stashing flags in argp (12592127)
2934 /*
2935 * There is currently only one option for this flavor.
2936 */
2937 if ((footprint_interval_flags & FOOTPRINT_INTERVAL_RESET) == 0) {
2938 error = EINVAL;
2939 break;
2940 }
2941 interval_max_footprint = get_task_phys_footprint_interval_max(proc_task(targetp), TRUE);
2942 break;
2943#endif /* CONFIG_LEDGER_INTERVAL_MAX */
2944 default:
2945 error = EINVAL;
2946 break;
2947 }
2948
2949 proc_rele(p: targetp);
2950 kauth_cred_unref(&target_cred);
2951
2952 /*
2953 * Return value from this function becomes errno to userland caller.
2954 */
2955 return error;
2956}
2957
2958/*
2959 * Return the current amount of CPU consumed by this thread (in either user or kernel mode)
2960 */
2961int
2962thread_selfusage(struct proc *p __unused, struct thread_selfusage_args *uap __unused, uint64_t *retval)
2963{
2964 uint64_t runtime;
2965
2966 runtime = thread_get_runtime_self();
2967 *retval = runtime;
2968
2969 return 0;
2970}
2971