kern_racct.c revision 220372
1/*-
2 * Copyright (c) 2010 The FreeBSD Foundation
3 * All rights reserved.
4 *
5 * This software was developed by Edward Tomasz Napierala under sponsorship
6 * from the FreeBSD Foundation.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 *    notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 *    notice, this list of conditions and the following disclaimer in the
15 *    documentation and/or other materials provided with the distribution.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * $FreeBSD: head/sys/kern/kern_racct.c 220372 2011-04-05 19:50:34Z trasz $
30 */
31
32#include <sys/cdefs.h>
33__FBSDID("$FreeBSD: head/sys/kern/kern_racct.c 220372 2011-04-05 19:50:34Z trasz $");
34
35#include "opt_kdtrace.h"
36
37#include <sys/param.h>
38#include <sys/eventhandler.h>
39#include <sys/param.h>
40#include <sys/jail.h>
41#include <sys/kernel.h>
42#include <sys/kthread.h>
43#include <sys/lock.h>
44#include <sys/loginclass.h>
45#include <sys/malloc.h>
46#include <sys/mutex.h>
47#include <sys/proc.h>
48#include <sys/racct.h>
49#include <sys/resourcevar.h>
50#include <sys/sbuf.h>
51#include <sys/sched.h>
52#include <sys/sdt.h>
53#include <sys/sx.h>
54#include <sys/sysent.h>
55#include <sys/sysproto.h>
56#include <sys/systm.h>
57#include <sys/umtx.h>
58
59#ifdef RCTL
60#include <sys/rctl.h>
61#endif
62
63#ifdef RACCT
64
65FEATURE(racct, "Resource Accounting");
66
67static struct mtx racct_lock;
68MTX_SYSINIT(racct_lock, &racct_lock, "racct lock", MTX_DEF);
69
70static uma_zone_t racct_zone;
71
72static void racct_sub_racct(struct racct *dest, const struct racct *src);
73static void racct_sub_cred_locked(struct ucred *cred, int resource,
74		uint64_t amount);
75static void racct_add_cred_locked(struct ucred *cred, int resource,
76		uint64_t amount);
77
78SDT_PROVIDER_DEFINE(racct);
79SDT_PROBE_DEFINE3(racct, kernel, rusage, add, add, "struct proc *", "int",
80    "uint64_t");
81SDT_PROBE_DEFINE3(racct, kernel, rusage, add_failure, add-failure,
82    "struct proc *", "int", "uint64_t");
83SDT_PROBE_DEFINE3(racct, kernel, rusage, add_cred, add-cred, "struct ucred *",
84    "int", "uint64_t");
85SDT_PROBE_DEFINE3(racct, kernel, rusage, add_force, add-force, "struct proc *",
86    "int", "uint64_t");
87SDT_PROBE_DEFINE3(racct, kernel, rusage, set, set, "struct proc *", "int",
88    "uint64_t");
89SDT_PROBE_DEFINE3(racct, kernel, rusage, set_failure, set-failure,
90    "struct proc *", "int", "uint64_t");
91SDT_PROBE_DEFINE3(racct, kernel, rusage, sub, sub, "struct proc *", "int",
92    "uint64_t");
93SDT_PROBE_DEFINE3(racct, kernel, rusage, sub_cred, sub-cred, "struct ucred *",
94    "int", "uint64_t");
95SDT_PROBE_DEFINE1(racct, kernel, racct, create, create, "struct racct *");
96SDT_PROBE_DEFINE1(racct, kernel, racct, destroy, destroy, "struct racct *");
97SDT_PROBE_DEFINE2(racct, kernel, racct, join, join, "struct racct *",
98    "struct racct *");
99SDT_PROBE_DEFINE2(racct, kernel, racct, join_failure, join-failure,
100    "struct racct *", "struct racct *");
101SDT_PROBE_DEFINE2(racct, kernel, racct, leave, leave, "struct racct *",
102    "struct racct *");
103
104int racct_types[] = {
105	[RACCT_CPU] =
106		RACCT_IN_THOUSANDS,
107	[RACCT_FSIZE] =
108		RACCT_RECLAIMABLE | RACCT_INHERITABLE | RACCT_DENIABLE,
109	[RACCT_DATA] =
110		RACCT_RECLAIMABLE | RACCT_INHERITABLE | RACCT_DENIABLE,
111	[RACCT_STACK] =
112		RACCT_RECLAIMABLE | RACCT_INHERITABLE | RACCT_DENIABLE,
113	[RACCT_CORE] =
114		RACCT_DENIABLE,
115	[RACCT_RSS] =
116		RACCT_RECLAIMABLE,
117	[RACCT_MEMLOCK] =
118		RACCT_RECLAIMABLE | RACCT_DENIABLE,
119	[RACCT_NPROC] =
120		RACCT_RECLAIMABLE | RACCT_DENIABLE,
121	[RACCT_NOFILE] =
122		RACCT_RECLAIMABLE | RACCT_INHERITABLE | RACCT_DENIABLE,
123	[RACCT_SBSIZE] =
124		RACCT_RECLAIMABLE | RACCT_DENIABLE | RACCT_SLOPPY,
125	[RACCT_VMEM] =
126		RACCT_RECLAIMABLE | RACCT_INHERITABLE | RACCT_DENIABLE,
127	[RACCT_NPTS] =
128		RACCT_RECLAIMABLE | RACCT_DENIABLE | RACCT_SLOPPY,
129	[RACCT_SWAP] =
130		RACCT_RECLAIMABLE | RACCT_DENIABLE | RACCT_SLOPPY,
131	[RACCT_NTHR] =
132		RACCT_RECLAIMABLE | RACCT_DENIABLE,
133	[RACCT_MSGQQUEUED] =
134		RACCT_RECLAIMABLE | RACCT_DENIABLE | RACCT_SLOPPY,
135	[RACCT_MSGQSIZE] =
136		RACCT_RECLAIMABLE | RACCT_DENIABLE | RACCT_SLOPPY,
137	[RACCT_NMSGQ] =
138		RACCT_RECLAIMABLE | RACCT_DENIABLE | RACCT_SLOPPY,
139	[RACCT_NSEM] =
140		RACCT_RECLAIMABLE | RACCT_DENIABLE | RACCT_SLOPPY,
141	[RACCT_NSEMOP] =
142		RACCT_RECLAIMABLE | RACCT_INHERITABLE | RACCT_DENIABLE,
143	[RACCT_NSHM] =
144		RACCT_RECLAIMABLE | RACCT_DENIABLE | RACCT_SLOPPY,
145	[RACCT_SHMSIZE] =
146		RACCT_RECLAIMABLE | RACCT_DENIABLE | RACCT_SLOPPY,
147	[RACCT_WALLCLOCK] =
148		RACCT_IN_THOUSANDS };
149
150static void
151racct_add_racct(struct racct *dest, const struct racct *src)
152{
153	int i;
154
155	mtx_assert(&racct_lock, MA_OWNED);
156
157	/*
158	 * Update resource usage in dest.
159	 */
160	for (i = 0; i <= RACCT_MAX; i++) {
161		KASSERT(dest->r_resources[i] >= 0,
162		    ("racct propagation meltdown: dest < 0"));
163		KASSERT(src->r_resources[i] >= 0,
164		    ("racct propagation meltdown: src < 0"));
165		dest->r_resources[i] += src->r_resources[i];
166	}
167}
168
169static void
170racct_sub_racct(struct racct *dest, const struct racct *src)
171{
172	int i;
173
174	mtx_assert(&racct_lock, MA_OWNED);
175
176	/*
177	 * Update resource usage in dest.
178	 */
179	for (i = 0; i <= RACCT_MAX; i++) {
180		if (!racct_is_sloppy(i)) {
181			KASSERT(dest->r_resources[i] >= 0,
182			    ("racct propagation meltdown: dest < 0"));
183			KASSERT(src->r_resources[i] >= 0,
184			    ("racct propagation meltdown: src < 0"));
185			KASSERT(src->r_resources[i] <= dest->r_resources[i],
186			    ("racct propagation meltdown: src > dest"));
187		}
188		if (racct_is_reclaimable(i)) {
189			dest->r_resources[i] -= src->r_resources[i];
190			if (dest->r_resources[i] < 0) {
191				KASSERT(racct_is_sloppy(i),
192				    ("racct_sub_racct: usage < 0"));
193				dest->r_resources[i] = 0;
194			}
195		}
196	}
197}
198
199void
200racct_create(struct racct **racctp)
201{
202
203	SDT_PROBE(racct, kernel, racct, create, racctp, 0, 0, 0, 0);
204
205	KASSERT(*racctp == NULL, ("racct already allocated"));
206
207	*racctp = uma_zalloc(racct_zone, M_WAITOK | M_ZERO);
208}
209
210static void
211racct_destroy_locked(struct racct **racctp)
212{
213	int i;
214	struct racct *racct;
215
216	SDT_PROBE(racct, kernel, racct, destroy, racctp, 0, 0, 0, 0);
217
218	mtx_assert(&racct_lock, MA_OWNED);
219	KASSERT(racctp != NULL, ("NULL racctp"));
220	KASSERT(*racctp != NULL, ("NULL racct"));
221
222	racct = *racctp;
223
224	for (i = 0; i <= RACCT_MAX; i++) {
225		if (racct_is_sloppy(i))
226			continue;
227		if (!racct_is_reclaimable(i))
228			continue;
229		KASSERT(racct->r_resources[i] == 0,
230		    ("destroying non-empty racct: "
231		    "%ju allocated for resource %d\n",
232		    racct->r_resources[i], i));
233	}
234	uma_zfree(racct_zone, racct);
235	*racctp = NULL;
236}
237
238void
239racct_destroy(struct racct **racct)
240{
241
242	mtx_lock(&racct_lock);
243	racct_destroy_locked(racct);
244	mtx_unlock(&racct_lock);
245}
246
247/*
248 * Increase consumption of 'resource' by 'amount' for 'racct'
249 * and all its parents.  Differently from other cases, 'amount' here
250 * may be less than zero.
251 */
252static void
253racct_alloc_resource(struct racct *racct, int resource,
254    uint64_t amount)
255{
256
257	mtx_assert(&racct_lock, MA_OWNED);
258	KASSERT(racct != NULL, ("NULL racct"));
259
260	racct->r_resources[resource] += amount;
261	if (racct->r_resources[resource] < 0) {
262		KASSERT(racct_is_sloppy(resource),
263		    ("racct_alloc_resource: usage < 0"));
264		racct->r_resources[resource] = 0;
265	}
266}
267
268/*
269 * Increase allocation of 'resource' by 'amount' for process 'p'.
270 * Return 0 if it's below limits, or errno, if it's not.
271 */
272int
273racct_add(struct proc *p, int resource, uint64_t amount)
274{
275#ifdef RCTL
276	int error;
277#endif
278
279	if (p->p_flag & P_SYSTEM)
280		return (0);
281
282	SDT_PROBE(racct, kernel, rusage, add, p, resource, amount, 0, 0);
283
284	/*
285	 * We need proc lock to dereference p->p_ucred.
286	 */
287	PROC_LOCK_ASSERT(p, MA_OWNED);
288
289	mtx_lock(&racct_lock);
290#ifdef RCTL
291	error = rctl_enforce(p, resource, amount);
292	if (error && racct_is_deniable(resource)) {
293		SDT_PROBE(racct, kernel, rusage, add_failure, p, resource,
294		    amount, 0, 0);
295		mtx_unlock(&racct_lock);
296		return (error);
297	}
298#endif
299	racct_alloc_resource(p->p_racct, resource, amount);
300	racct_add_cred_locked(p->p_ucred, resource, amount);
301	mtx_unlock(&racct_lock);
302
303	return (0);
304}
305
306static void
307racct_add_cred_locked(struct ucred *cred, int resource, uint64_t amount)
308{
309	struct prison *pr;
310
311	SDT_PROBE(racct, kernel, rusage, add_cred, cred, resource, amount,
312	    0, 0);
313
314	racct_alloc_resource(cred->cr_ruidinfo->ui_racct, resource, amount);
315	for (pr = cred->cr_prison; pr != NULL; pr = pr->pr_parent)
316		racct_alloc_resource(pr->pr_racct, resource, amount);
317	racct_alloc_resource(cred->cr_loginclass->lc_racct, resource, amount);
318}
319
320/*
321 * Increase allocation of 'resource' by 'amount' for credential 'cred'.
322 * Doesn't check for limits and never fails.
323 *
324 * XXX: Shouldn't this ever return an error?
325 */
326void
327racct_add_cred(struct ucred *cred, int resource, uint64_t amount)
328{
329
330	mtx_lock(&racct_lock);
331	racct_add_cred_locked(cred, resource, amount);
332	mtx_unlock(&racct_lock);
333}
334
335/*
336 * Increase allocation of 'resource' by 'amount' for process 'p'.
337 * Doesn't check for limits and never fails.
338 */
339void
340racct_add_force(struct proc *p, int resource, uint64_t amount)
341{
342
343	if (p->p_flag & P_SYSTEM)
344		return;
345
346	SDT_PROBE(racct, kernel, rusage, add_force, p, resource, amount, 0, 0);
347
348	/*
349	 * We need proc lock to dereference p->p_ucred.
350	 */
351	PROC_LOCK_ASSERT(p, MA_OWNED);
352
353	mtx_lock(&racct_lock);
354	racct_alloc_resource(p->p_racct, resource, amount);
355	mtx_unlock(&racct_lock);
356	racct_add_cred(p->p_ucred, resource, amount);
357}
358
359static int
360racct_set_locked(struct proc *p, int resource, uint64_t amount)
361{
362	int64_t diff;
363#ifdef RCTL
364	int error;
365#endif
366
367	if (p->p_flag & P_SYSTEM)
368		return (0);
369
370	SDT_PROBE(racct, kernel, rusage, set, p, resource, amount, 0, 0);
371
372	/*
373	 * We need proc lock to dereference p->p_ucred.
374	 */
375	PROC_LOCK_ASSERT(p, MA_OWNED);
376
377	diff = amount - p->p_racct->r_resources[resource];
378#ifdef notyet
379	KASSERT(diff >= 0 || racct_is_reclaimable(resource),
380	    ("racct_set: usage of non-reclaimable resource %d dropping",
381	     resource));
382#endif
383#ifdef RCTL
384	if (diff > 0) {
385		error = rctl_enforce(p, resource, diff);
386		if (error && racct_is_deniable(resource)) {
387			SDT_PROBE(racct, kernel, rusage, set_failure, p,
388			    resource, amount, 0, 0);
389			return (error);
390		}
391	}
392#endif
393	racct_alloc_resource(p->p_racct, resource, diff);
394	if (diff > 0)
395		racct_add_cred_locked(p->p_ucred, resource, diff);
396	else if (diff < 0)
397		racct_sub_cred_locked(p->p_ucred, resource, -diff);
398
399	return (0);
400}
401
402/*
403 * Set allocation of 'resource' to 'amount' for process 'p'.
404 * Return 0 if it's below limits, or errno, if it's not.
405 *
406 * Note that decreasing the allocation always returns 0,
407 * even if it's above the limit.
408 */
409int
410racct_set(struct proc *p, int resource, uint64_t amount)
411{
412	int error;
413
414	mtx_lock(&racct_lock);
415	error = racct_set_locked(p, resource, amount);
416	mtx_unlock(&racct_lock);
417	return (error);
418}
419
420void
421racct_set_force(struct proc *p, int resource, uint64_t amount)
422{
423	int64_t diff;
424
425	if (p->p_flag & P_SYSTEM)
426		return;
427
428	SDT_PROBE(racct, kernel, rusage, set, p, resource, amount, 0, 0);
429
430	/*
431	 * We need proc lock to dereference p->p_ucred.
432	 */
433	PROC_LOCK_ASSERT(p, MA_OWNED);
434
435	mtx_lock(&racct_lock);
436	diff = amount - p->p_racct->r_resources[resource];
437	racct_alloc_resource(p->p_racct, resource, diff);
438	if (diff > 0)
439		racct_add_cred_locked(p->p_ucred, resource, diff);
440	else if (diff < 0)
441		racct_sub_cred_locked(p->p_ucred, resource, -diff);
442	mtx_unlock(&racct_lock);
443}
444
445/*
446 * Returns amount of 'resource' the process 'p' can keep allocated.
447 * Allocating more than that would be denied, unless the resource
448 * is marked undeniable.  Amount of already allocated resource does
449 * not matter.
450 */
451uint64_t
452racct_get_limit(struct proc *p, int resource)
453{
454
455#ifdef RCTL
456	return (rctl_get_limit(p, resource));
457#else
458	return (UINT64_MAX);
459#endif
460}
461
462/*
463 * Returns amount of 'resource' the process 'p' can keep allocated.
464 * Allocating more than that would be denied, unless the resource
465 * is marked undeniable.  Amount of already allocated resource does
466 * matter.
467 */
468uint64_t
469racct_get_available(struct proc *p, int resource)
470{
471
472#ifdef RCTL
473	return (rctl_get_available(p, resource));
474#else
475	return (UINT64_MAX);
476#endif
477}
478
479/*
480 * Decrease allocation of 'resource' by 'amount' for process 'p'.
481 */
482void
483racct_sub(struct proc *p, int resource, uint64_t amount)
484{
485
486	if (p->p_flag & P_SYSTEM)
487		return;
488
489	SDT_PROBE(racct, kernel, rusage, sub, p, resource, amount, 0, 0);
490
491	/*
492	 * We need proc lock to dereference p->p_ucred.
493	 */
494	PROC_LOCK_ASSERT(p, MA_OWNED);
495	KASSERT(racct_is_reclaimable(resource),
496	    ("racct_sub: called for non-reclaimable resource %d", resource));
497
498	mtx_lock(&racct_lock);
499	KASSERT(amount <= p->p_racct->r_resources[resource],
500	    ("racct_sub: freeing %ju of resource %d, which is more "
501	     "than allocated %jd for %s (pid %d)", amount, resource,
502	    (intmax_t)p->p_racct->r_resources[resource], p->p_comm, p->p_pid));
503
504	racct_alloc_resource(p->p_racct, resource, -amount);
505	racct_sub_cred_locked(p->p_ucred, resource, amount);
506	mtx_unlock(&racct_lock);
507}
508
509static void
510racct_sub_cred_locked(struct ucred *cred, int resource, uint64_t amount)
511{
512	struct prison *pr;
513
514	SDT_PROBE(racct, kernel, rusage, sub_cred, cred, resource, amount,
515	    0, 0);
516
517#ifdef notyet
518	KASSERT(racct_is_reclaimable(resource),
519	    ("racct_sub_cred: called for non-reclaimable resource %d",
520	     resource));
521#endif
522
523	racct_alloc_resource(cred->cr_ruidinfo->ui_racct, resource, -amount);
524	for (pr = cred->cr_prison; pr != NULL; pr = pr->pr_parent)
525		racct_alloc_resource(pr->pr_racct, resource, -amount);
526	racct_alloc_resource(cred->cr_loginclass->lc_racct, resource, -amount);
527}
528
529/*
530 * Decrease allocation of 'resource' by 'amount' for credential 'cred'.
531 */
532void
533racct_sub_cred(struct ucred *cred, int resource, uint64_t amount)
534{
535
536	mtx_lock(&racct_lock);
537	racct_sub_cred_locked(cred, resource, amount);
538	mtx_unlock(&racct_lock);
539}
540
541/*
542 * Inherit resource usage information from the parent process.
543 */
544int
545racct_proc_fork(struct proc *parent, struct proc *child)
546{
547	int i, error = 0;
548
549	/*
550	 * Create racct for the child process.
551	 */
552	racct_create(&child->p_racct);
553
554	/*
555	 * No resource accounting for kernel processes.
556	 */
557	if (child->p_flag & P_SYSTEM)
558		return (0);
559
560	PROC_LOCK(parent);
561	PROC_LOCK(child);
562	mtx_lock(&racct_lock);
563
564	/*
565	 * Inherit resource usage.
566	 */
567	for (i = 0; i <= RACCT_MAX; i++) {
568		if (parent->p_racct->r_resources[i] == 0 ||
569		    !racct_is_inheritable(i))
570			continue;
571
572		error = racct_set_locked(child, i,
573		    parent->p_racct->r_resources[i]);
574		if (error != 0) {
575			/*
576			 * XXX: The only purpose of these two lines is
577			 * to prevent from tripping checks in racct_destroy().
578			 */
579			for (i = 0; i <= RACCT_MAX; i++)
580				racct_set_locked(child, i, 0);
581			goto out;
582		}
583	}
584
585#ifdef RCTL
586	error = rctl_proc_fork(parent, child);
587	if (error != 0) {
588		/*
589		 * XXX: The only purpose of these two lines is to prevent from
590		 * tripping checks in racct_destroy().
591		 */
592		for (i = 0; i <= RACCT_MAX; i++)
593			racct_set_locked(child, i, 0);
594	}
595#endif
596
597out:
598	if (error != 0)
599		racct_destroy_locked(&child->p_racct);
600	mtx_unlock(&racct_lock);
601	PROC_UNLOCK(child);
602	PROC_UNLOCK(parent);
603
604	return (error);
605}
606
607void
608racct_proc_exit(struct proc *p)
609{
610	uint64_t runtime;
611
612	PROC_LOCK(p);
613	/*
614	 * We don't need to calculate rux, proc_reap() has already done this.
615	 */
616	runtime = cputick2usec(p->p_rux.rux_runtime);
617#ifdef notyet
618	KASSERT(runtime >= p->p_prev_runtime, ("runtime < p_prev_runtime"));
619#else
620	if (runtime < p->p_prev_runtime)
621		runtime = p->p_prev_runtime;
622#endif
623	racct_set(p, RACCT_CPU, runtime);
624
625	/*
626	 * XXX: Free this some other way.
627	 */
628	racct_set(p, RACCT_FSIZE, 0);
629	racct_set(p, RACCT_NPTS, 0);
630	racct_set(p, RACCT_NTHR, 0);
631	racct_set(p, RACCT_RSS, 0);
632	PROC_UNLOCK(p);
633
634#ifdef RCTL
635	rctl_racct_release(p->p_racct);
636#endif
637	racct_destroy(&p->p_racct);
638}
639
640/*
641 * Called after credentials change, to move resource utilisation
642 * between raccts.
643 */
644void
645racct_proc_ucred_changed(struct proc *p, struct ucred *oldcred,
646    struct ucred *newcred)
647{
648	struct uidinfo *olduip, *newuip;
649	struct loginclass *oldlc, *newlc;
650	struct prison *oldpr, *newpr, *pr;
651
652	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
653
654	newuip = newcred->cr_ruidinfo;
655	olduip = oldcred->cr_ruidinfo;
656	newlc = newcred->cr_loginclass;
657	oldlc = oldcred->cr_loginclass;
658	newpr = newcred->cr_prison;
659	oldpr = oldcred->cr_prison;
660
661	mtx_lock(&racct_lock);
662	if (newuip != olduip) {
663		racct_sub_racct(olduip->ui_racct, p->p_racct);
664		racct_add_racct(newuip->ui_racct, p->p_racct);
665	}
666	if (newlc != oldlc) {
667		racct_sub_racct(oldlc->lc_racct, p->p_racct);
668		racct_add_racct(newlc->lc_racct, p->p_racct);
669	}
670	if (newpr != oldpr) {
671		for (pr = oldpr; pr != NULL; pr = pr->pr_parent)
672			racct_sub_racct(pr->pr_racct, p->p_racct);
673		for (pr = newpr; pr != NULL; pr = pr->pr_parent)
674			racct_add_racct(pr->pr_racct, p->p_racct);
675	}
676	mtx_unlock(&racct_lock);
677
678#ifdef RCTL
679	rctl_proc_ucred_changed(p, newcred);
680#endif
681}
682
683static void
684racctd(void)
685{
686	struct thread *td;
687	struct proc *p;
688	struct timeval wallclock;
689	uint64_t runtime;
690
691	for (;;) {
692		sx_slock(&allproc_lock);
693
694		FOREACH_PROC_IN_SYSTEM(p) {
695			if (p->p_state != PRS_NORMAL)
696				continue;
697			if (p->p_flag & P_SYSTEM)
698				continue;
699
700			microuptime(&wallclock);
701			timevalsub(&wallclock, &p->p_stats->p_start);
702			PROC_LOCK(p);
703			PROC_SLOCK(p);
704			FOREACH_THREAD_IN_PROC(p, td) {
705				ruxagg(p, td);
706				thread_lock(td);
707				thread_unlock(td);
708			}
709			runtime = cputick2usec(p->p_rux.rux_runtime);
710			PROC_SUNLOCK(p);
711#ifdef notyet
712			KASSERT(runtime >= p->p_prev_runtime,
713			    ("runtime < p_prev_runtime"));
714#else
715			if (runtime < p->p_prev_runtime)
716				runtime = p->p_prev_runtime;
717#endif
718			p->p_prev_runtime = runtime;
719			mtx_lock(&racct_lock);
720			racct_set_locked(p, RACCT_CPU, runtime);
721			racct_set_locked(p, RACCT_WALLCLOCK,
722			    wallclock.tv_sec * 1000000 + wallclock.tv_usec);
723			mtx_unlock(&racct_lock);
724			PROC_UNLOCK(p);
725		}
726		sx_sunlock(&allproc_lock);
727		pause("-", hz);
728	}
729}
730
731static struct kproc_desc racctd_kp = {
732	"racctd",
733	racctd,
734	NULL
735};
736SYSINIT(racctd, SI_SUB_RACCTD, SI_ORDER_FIRST, kproc_start, &racctd_kp);
737
738static void
739racct_init(void)
740{
741
742	racct_zone = uma_zcreate("racct", sizeof(struct racct),
743	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
744	/*
745	 * XXX: Move this somewhere.
746	 */
747	racct_create(&prison0.pr_racct);
748}
749SYSINIT(racct, SI_SUB_RACCT, SI_ORDER_FIRST, racct_init, NULL);
750
751#else /* !RACCT */
752
753int
754racct_add(struct proc *p, int resource, uint64_t amount)
755{
756
757	return (0);
758}
759
760void
761racct_add_cred(struct ucred *cred, int resource, uint64_t amount)
762{
763}
764
765void
766racct_add_force(struct proc *p, int resource, uint64_t amount)
767{
768
769	return;
770}
771
772int
773racct_set(struct proc *p, int resource, uint64_t amount)
774{
775
776	return (0);
777}
778
779void
780racct_set_force(struct proc *p, int resource, uint64_t amount)
781{
782}
783
784void
785racct_sub(struct proc *p, int resource, uint64_t amount)
786{
787}
788
789void
790racct_sub_cred(struct ucred *cred, int resource, uint64_t amount)
791{
792}
793
794uint64_t
795racct_get_limit(struct proc *p, int resource)
796{
797
798	return (UINT64_MAX);
799}
800
801uint64_t
802racct_get_available(struct proc *p, int resource)
803{
804
805	return (UINT64_MAX);
806}
807
808void
809racct_create(struct racct **racctp)
810{
811}
812
813void
814racct_destroy(struct racct **racctp)
815{
816}
817
818int
819racct_proc_fork(struct proc *parent, struct proc *child)
820{
821
822	return (0);
823}
824
825void
826racct_proc_exit(struct proc *p)
827{
828}
829
830#endif /* !RACCT */
831