xref: /dflybsd-src/sys/kern/kern_resource.c (revision b370aff7747b2e03ce9b829fbf2877dffdadfb64)
1 /*-
2  * Copyright (c) 1982, 1986, 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * (c) UNIX System Laboratories, Inc.
5  * All or some portions of this file are derived from material licensed
6  * to the University of California by American Telephone and Telegraph
7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8  * the permission of UNIX System Laboratories, Inc.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. All advertising materials mentioning features or use of this software
19  *    must display the following acknowledgement:
20  *	This product includes software developed by the University of
21  *	California, Berkeley and its contributors.
22  * 4. Neither the name of the University nor the names of its contributors
23  *    may be used to endorse or promote products derived from this software
24  *    without specific prior written permission.
25  *
26  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36  * SUCH DAMAGE.
37  *
38  *	@(#)kern_resource.c	8.5 (Berkeley) 1/21/94
39  * $FreeBSD: src/sys/kern/kern_resource.c,v 1.55.2.5 2001/11/03 01:41:08 ps Exp $
40  * $DragonFly: src/sys/kern/kern_resource.c,v 1.35 2008/05/27 05:25:34 dillon Exp $
41  */
42 
43 #include "opt_compat.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/sysproto.h>
48 #include <sys/file.h>
49 #include <sys/kern_syscall.h>
50 #include <sys/kernel.h>
51 #include <sys/resourcevar.h>
52 #include <sys/malloc.h>
53 #include <sys/proc.h>
54 #include <sys/priv.h>
55 #include <sys/time.h>
56 #include <sys/lockf.h>
57 
58 #include <vm/vm.h>
59 #include <vm/vm_param.h>
60 #include <sys/lock.h>
61 #include <vm/pmap.h>
62 #include <vm/vm_map.h>
63 
64 #include <sys/thread2.h>
65 #include <sys/spinlock2.h>
66 
67 static int donice (struct proc *chgp, int n);
68 static int doionice (struct proc *chgp, int n);
69 
70 static MALLOC_DEFINE(M_UIDINFO, "uidinfo", "uidinfo structures");
71 #define	UIHASH(uid)	(&uihashtbl[(uid) & uihash])
72 static struct spinlock uihash_lock;
73 static LIST_HEAD(uihashhead, uidinfo) *uihashtbl;
74 static u_long uihash;		/* size of hash table - 1 */
75 
76 static struct uidinfo	*uicreate (uid_t uid);
77 static struct uidinfo	*uilookup (uid_t uid);
78 
79 /*
80  * Resource controls and accounting.
81  */
82 
83 struct getpriority_info {
84 	int low;
85 	int who;
86 };
87 
88 static int getpriority_callback(struct proc *p, void *data);
89 
90 /*
91  * MPALMOSTSAFE
92  */
93 int
94 sys_getpriority(struct getpriority_args *uap)
95 {
96 	struct getpriority_info info;
97 	struct proc *curp = curproc;
98 	struct proc *p;
99 	int low = PRIO_MAX + 1;
100 	int error;
101 
102 	switch (uap->which) {
103 	case PRIO_PROCESS:
104 		if (uap->who == 0) {
105 			p = curp;
106 			PHOLD(p);
107 		} else {
108 			p = pfind(uap->who);
109 		}
110 		if (p) {
111 			if (PRISON_CHECK(curp->p_ucred, p->p_ucred)) {
112 				low = p->p_nice;
113 			}
114 			PRELE(p);
115 		}
116 		break;
117 
118 	case PRIO_PGRP:
119 	{
120 		struct pgrp *pg;
121 
122 		if (uap->who == 0) {
123 			pg = curp->p_pgrp;
124 			pgref(pg);
125 		} else if ((pg = pgfind(uap->who)) == NULL) {
126 			break;
127 		} /* else ref held from pgfind */
128 
129 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
130 			if (PRISON_CHECK(curp->p_ucred, p->p_ucred) &&
131 			    p->p_nice < low) {
132 				low = p->p_nice;
133 			}
134 		}
135 		pgrel(pg);
136 		break;
137 	}
138 	case PRIO_USER:
139 		if (uap->who == 0)
140 			uap->who = curp->p_ucred->cr_uid;
141 		info.low = low;
142 		info.who = uap->who;
143 		allproc_scan(getpriority_callback, &info);
144 		low = info.low;
145 		break;
146 
147 	default:
148 		error = EINVAL;
149 		goto done;
150 	}
151 	if (low == PRIO_MAX + 1) {
152 		error = ESRCH;
153 		goto done;
154 	}
155 	uap->sysmsg_result = low;
156 	error = 0;
157 done:
158 	return (error);
159 }
160 
161 /*
162  * Figure out the current lowest nice priority for processes owned
163  * by the specified user.
164  */
165 static
166 int
167 getpriority_callback(struct proc *p, void *data)
168 {
169 	struct getpriority_info *info = data;
170 
171 	if (PRISON_CHECK(curproc->p_ucred, p->p_ucred) &&
172 	    p->p_ucred->cr_uid == info->who &&
173 	    p->p_nice < info->low) {
174 		info->low = p->p_nice;
175 	}
176 	return(0);
177 }
178 
179 struct setpriority_info {
180 	int prio;
181 	int who;
182 	int error;
183 	int found;
184 };
185 
186 static int setpriority_callback(struct proc *p, void *data);
187 
188 /*
189  * MPALMOSTSAFE
190  */
191 int
192 sys_setpriority(struct setpriority_args *uap)
193 {
194 	struct setpriority_info info;
195 	struct proc *curp = curproc;
196 	struct proc *p;
197 	int found = 0, error = 0;
198 
199 	lwkt_gettoken(&proc_token);
200 
201 	switch (uap->which) {
202 	case PRIO_PROCESS:
203 		if (uap->who == 0) {
204 			p = curp;
205 			PHOLD(p);
206 		} else {
207 			p = pfind(uap->who);
208 		}
209 		if (p) {
210 			if (PRISON_CHECK(curp->p_ucred, p->p_ucred)) {
211 				error = donice(p, uap->prio);
212 				found++;
213 			}
214 			PRELE(p);
215 		}
216 		break;
217 
218 	case PRIO_PGRP:
219 	{
220 		struct pgrp *pg;
221 
222 		if (uap->who == 0) {
223 			pg = curp->p_pgrp;
224 			pgref(pg);
225 		} else if ((pg = pgfind(uap->who)) == NULL) {
226 			break;
227 		} /* else ref held from pgfind */
228 
229 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
230 			if (PRISON_CHECK(curp->p_ucred, p->p_ucred)) {
231 				error = donice(p, uap->prio);
232 				found++;
233 			}
234 		}
235 		pgrel(pg);
236 		break;
237 	}
238 	case PRIO_USER:
239 		if (uap->who == 0)
240 			uap->who = curp->p_ucred->cr_uid;
241 		info.prio = uap->prio;
242 		info.who = uap->who;
243 		info.error = 0;
244 		info.found = 0;
245 		allproc_scan(setpriority_callback, &info);
246 		error = info.error;
247 		found = info.found;
248 		break;
249 
250 	default:
251 		error = EINVAL;
252 		found = 1;
253 		break;
254 	}
255 
256 	lwkt_reltoken(&proc_token);
257 
258 	if (found == 0)
259 		error = ESRCH;
260 	return (error);
261 }
262 
263 static
264 int
265 setpriority_callback(struct proc *p, void *data)
266 {
267 	struct setpriority_info *info = data;
268 	int error;
269 
270 	if (p->p_ucred->cr_uid == info->who &&
271 	    PRISON_CHECK(curproc->p_ucred, p->p_ucred)) {
272 		error = donice(p, info->prio);
273 		if (error)
274 			info->error = error;
275 		++info->found;
276 	}
277 	return(0);
278 }
279 
280 static int
281 donice(struct proc *chgp, int n)
282 {
283 	struct proc *curp = curproc;
284 	struct ucred *cr = curp->p_ucred;
285 	struct lwp *lp;
286 
287 	if (cr->cr_uid && cr->cr_ruid &&
288 	    cr->cr_uid != chgp->p_ucred->cr_uid &&
289 	    cr->cr_ruid != chgp->p_ucred->cr_uid)
290 		return (EPERM);
291 	if (n > PRIO_MAX)
292 		n = PRIO_MAX;
293 	if (n < PRIO_MIN)
294 		n = PRIO_MIN;
295 	if (n < chgp->p_nice && priv_check_cred(cr, PRIV_SCHED_SETPRIORITY, 0))
296 		return (EACCES);
297 	chgp->p_nice = n;
298 	FOREACH_LWP_IN_PROC(lp, chgp) {
299 		LWPHOLD(lp);
300 		chgp->p_usched->resetpriority(lp);
301 		LWPRELE(lp);
302 	}
303 	return (0);
304 }
305 
306 
307 struct ioprio_get_info {
308 	int high;
309 	int who;
310 };
311 
312 static int ioprio_get_callback(struct proc *p, void *data);
313 
314 /*
315  * MPALMOSTSAFE
316  */
317 int
318 sys_ioprio_get(struct ioprio_get_args *uap)
319 {
320 	struct ioprio_get_info info;
321 	struct proc *curp = curproc;
322 	struct proc *p;
323 	int high = IOPRIO_MIN-2;
324 	int error;
325 
326 	lwkt_gettoken(&proc_token);
327 
328 	switch (uap->which) {
329 	case PRIO_PROCESS:
330 		if (uap->who == 0) {
331 			p = curp;
332 			PHOLD(p);
333 		} else {
334 			p = pfind(uap->who);
335 		}
336 		if (p) {
337 			if (PRISON_CHECK(curp->p_ucred, p->p_ucred))
338 				high = p->p_ionice;
339 			PRELE(p);
340 		}
341 		break;
342 
343 	case PRIO_PGRP:
344 	{
345 		struct pgrp *pg;
346 
347 		if (uap->who == 0) {
348 			pg = curp->p_pgrp;
349 			pgref(pg);
350 		} else if ((pg = pgfind(uap->who)) == NULL) {
351 			break;
352 		} /* else ref held from pgfind */
353 
354 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
355 			if (PRISON_CHECK(curp->p_ucred, p->p_ucred) &&
356 			    p->p_nice > high)
357 				high = p->p_ionice;
358 		}
359 		pgrel(pg);
360 		break;
361 	}
362 	case PRIO_USER:
363 		if (uap->who == 0)
364 			uap->who = curp->p_ucred->cr_uid;
365 		info.high = high;
366 		info.who = uap->who;
367 		allproc_scan(ioprio_get_callback, &info);
368 		high = info.high;
369 		break;
370 
371 	default:
372 		error = EINVAL;
373 		goto done;
374 	}
375 	if (high == IOPRIO_MIN-2) {
376 		error = ESRCH;
377 		goto done;
378 	}
379 	uap->sysmsg_result = high;
380 	error = 0;
381 done:
382 	lwkt_reltoken(&proc_token);
383 
384 	return (error);
385 }
386 
387 /*
388  * Figure out the current lowest nice priority for processes owned
389  * by the specified user.
390  */
391 static
392 int
393 ioprio_get_callback(struct proc *p, void *data)
394 {
395 	struct ioprio_get_info *info = data;
396 
397 	if (PRISON_CHECK(curproc->p_ucred, p->p_ucred) &&
398 	    p->p_ucred->cr_uid == info->who &&
399 	    p->p_ionice > info->high) {
400 		info->high = p->p_ionice;
401 	}
402 	return(0);
403 }
404 
405 
406 struct ioprio_set_info {
407 	int prio;
408 	int who;
409 	int error;
410 	int found;
411 };
412 
413 static int ioprio_set_callback(struct proc *p, void *data);
414 
415 /*
416  * MPALMOSTSAFE
417  */
418 int
419 sys_ioprio_set(struct ioprio_set_args *uap)
420 {
421 	struct ioprio_set_info info;
422 	struct proc *curp = curproc;
423 	struct proc *p;
424 	int found = 0, error = 0;
425 
426 	lwkt_gettoken(&proc_token);
427 
428 	switch (uap->which) {
429 	case PRIO_PROCESS:
430 		if (uap->who == 0) {
431 			p = curp;
432 			PHOLD(p);
433 		} else {
434 			p = pfind(uap->who);
435 		}
436 		if (p) {
437 			if (PRISON_CHECK(curp->p_ucred, p->p_ucred)) {
438 				error = doionice(p, uap->prio);
439 				found++;
440 			}
441 			PRELE(p);
442 		}
443 		break;
444 
445 	case PRIO_PGRP:
446 	{
447 		struct pgrp *pg;
448 
449 		if (uap->who == 0) {
450 			pg = curp->p_pgrp;
451 			pgref(pg);
452 		} else if ((pg = pgfind(uap->who)) == NULL) {
453 			break;
454 		} /* else ref held from pgfind */
455 
456 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
457 			if (PRISON_CHECK(curp->p_ucred, p->p_ucred)) {
458 				error = doionice(p, uap->prio);
459 				found++;
460 			}
461 		}
462 		pgrel(pg);
463 		break;
464 	}
465 	case PRIO_USER:
466 		if (uap->who == 0)
467 			uap->who = curp->p_ucred->cr_uid;
468 		info.prio = uap->prio;
469 		info.who = uap->who;
470 		info.error = 0;
471 		info.found = 0;
472 		allproc_scan(ioprio_set_callback, &info);
473 		error = info.error;
474 		found = info.found;
475 		break;
476 
477 	default:
478 		error = EINVAL;
479 		found = 1;
480 		break;
481 	}
482 
483 	lwkt_reltoken(&proc_token);
484 
485 	if (found == 0)
486 		error = ESRCH;
487 	return (error);
488 }
489 
490 static
491 int
492 ioprio_set_callback(struct proc *p, void *data)
493 {
494 	struct ioprio_set_info *info = data;
495 	int error;
496 
497 	if (p->p_ucred->cr_uid == info->who &&
498 	    PRISON_CHECK(curproc->p_ucred, p->p_ucred)) {
499 		error = doionice(p, info->prio);
500 		if (error)
501 			info->error = error;
502 		++info->found;
503 	}
504 	return(0);
505 }
506 
507 int
508 doionice(struct proc *chgp, int n)
509 {
510 	struct proc *curp = curproc;
511 	struct ucred *cr = curp->p_ucred;
512 
513 	if (cr->cr_uid && cr->cr_ruid &&
514 	    cr->cr_uid != chgp->p_ucred->cr_uid &&
515 	    cr->cr_ruid != chgp->p_ucred->cr_uid)
516 		return (EPERM);
517 	if (n > IOPRIO_MAX)
518 		n = IOPRIO_MAX;
519 	if (n < IOPRIO_MIN)
520 		n = IOPRIO_MIN;
521 	if (n < chgp->p_ionice && priv_check_cred(cr, PRIV_SCHED_SETPRIORITY, 0))
522 		return (EACCES);
523 	chgp->p_ionice = n;
524 
525 	return (0);
526 
527 }
528 
529 /*
530  * MPALMOSTSAFE
531  */
532 int
533 sys_lwp_rtprio(struct lwp_rtprio_args *uap)
534 {
535 	struct proc *p;
536 	struct lwp *lp;
537 	struct rtprio rtp;
538 	struct ucred *cr = curthread->td_ucred;
539 	int error;
540 
541 	error = copyin(uap->rtp, &rtp, sizeof(struct rtprio));
542 	if (error)
543 		return error;
544 	if (uap->pid < 0)
545 		return EINVAL;
546 
547 	lwkt_gettoken(&proc_token);
548 
549 	if (uap->pid == 0) {
550 		p = curproc;
551 		PHOLD(p);
552 	} else {
553 		p = pfind(uap->pid);
554 	}
555 
556 	if (p == NULL) {
557 		error = ESRCH;
558 		goto done;
559 	}
560 
561 	if (uap->tid < -1) {
562 		error = EINVAL;
563 		goto done;
564 	}
565 	if (uap->tid == -1) {
566 		/*
567 		 * sadly, tid can be 0 so we can't use 0 here
568 		 * like sys_rtprio()
569 		 */
570 		lp = curthread->td_lwp;
571 	} else {
572 		lp = lwp_rb_tree_RB_LOOKUP(&p->p_lwp_tree, uap->tid);
573 		if (lp == NULL) {
574 			error = ESRCH;
575 			goto done;
576 		}
577 	}
578 
579 	switch (uap->function) {
580 	case RTP_LOOKUP:
581 		error = copyout(&lp->lwp_rtprio, uap->rtp,
582 				sizeof(struct rtprio));
583 		break;
584 	case RTP_SET:
585 		if (cr->cr_uid && cr->cr_ruid &&
586 		    cr->cr_uid != p->p_ucred->cr_uid &&
587 		    cr->cr_ruid != p->p_ucred->cr_uid) {
588 			error = EPERM;
589 			break;
590 		}
591 		/* disallow setting rtprio in most cases if not superuser */
592 		if (priv_check_cred(cr, PRIV_SCHED_RTPRIO, 0)) {
593 			/* can't set someone else's */
594 			if (uap->pid) { /* XXX */
595 				error = EPERM;
596 				break;
597 			}
598 			/* can't set realtime priority */
599 /*
600  * Realtime priority has to be restricted for reasons which should be
601  * obvious. However, for idle priority, there is a potential for
602  * system deadlock if an idleprio process gains a lock on a resource
603  * that other processes need (and the idleprio process can't run
604  * due to a CPU-bound normal process). Fix me! XXX
605  */
606  			if (RTP_PRIO_IS_REALTIME(rtp.type)) {
607 				error = EPERM;
608 				break;
609 			}
610 		}
611 		switch (rtp.type) {
612 #ifdef RTP_PRIO_FIFO
613 		case RTP_PRIO_FIFO:
614 #endif
615 		case RTP_PRIO_REALTIME:
616 		case RTP_PRIO_NORMAL:
617 		case RTP_PRIO_IDLE:
618 			if (rtp.prio > RTP_PRIO_MAX) {
619 				error = EINVAL;
620 			} else {
621 				lp->lwp_rtprio = rtp;
622 				error = 0;
623 			}
624 			break;
625 		default:
626 			error = EINVAL;
627 			break;
628 		}
629 		break;
630 	default:
631 		error = EINVAL;
632 		break;
633 	}
634 
635 done:
636 	if (p)
637 		PRELE(p);
638 	lwkt_reltoken(&proc_token);
639 
640 	return (error);
641 }
642 
643 /*
644  * Set realtime priority
645  *
646  * MPALMOSTSAFE
647  */
648 int
649 sys_rtprio(struct rtprio_args *uap)
650 {
651 	struct proc *p;
652 	struct lwp *lp;
653 	struct ucred *cr = curthread->td_ucred;
654 	struct rtprio rtp;
655 	int error;
656 
657 	error = copyin(uap->rtp, &rtp, sizeof(struct rtprio));
658 	if (error)
659 		return (error);
660 
661 	lwkt_gettoken(&proc_token);
662 
663 	if (uap->pid == 0) {
664 		p = curproc;
665 		PHOLD(p);
666 	} else {
667 		p = pfind(uap->pid);
668 	}
669 
670 	if (p == NULL) {
671 		error = ESRCH;
672 		goto done;
673 	}
674 
675 	/* XXX lwp */
676 	lp = FIRST_LWP_IN_PROC(p);
677 	switch (uap->function) {
678 	case RTP_LOOKUP:
679 		error = copyout(&lp->lwp_rtprio, uap->rtp,
680 				sizeof(struct rtprio));
681 		break;
682 	case RTP_SET:
683 		if (cr->cr_uid && cr->cr_ruid &&
684 		    cr->cr_uid != p->p_ucred->cr_uid &&
685 		    cr->cr_ruid != p->p_ucred->cr_uid) {
686 			error = EPERM;
687 			break;
688 		}
689 		/* disallow setting rtprio in most cases if not superuser */
690 		if (priv_check_cred(cr, PRIV_SCHED_RTPRIO, 0)) {
691 			/* can't set someone else's */
692 			if (uap->pid) {
693 				error = EPERM;
694 				break;
695 			}
696 			/* can't set realtime priority */
697 /*
698  * Realtime priority has to be restricted for reasons which should be
699  * obvious. However, for idle priority, there is a potential for
700  * system deadlock if an idleprio process gains a lock on a resource
701  * that other processes need (and the idleprio process can't run
702  * due to a CPU-bound normal process). Fix me! XXX
703  */
704 			if (RTP_PRIO_IS_REALTIME(rtp.type)) {
705 				error = EPERM;
706 				break;
707 			}
708 		}
709 		switch (rtp.type) {
710 #ifdef RTP_PRIO_FIFO
711 		case RTP_PRIO_FIFO:
712 #endif
713 		case RTP_PRIO_REALTIME:
714 		case RTP_PRIO_NORMAL:
715 		case RTP_PRIO_IDLE:
716 			if (rtp.prio > RTP_PRIO_MAX) {
717 				error = EINVAL;
718 				break;
719 			}
720 			lp->lwp_rtprio = rtp;
721 			error = 0;
722 			break;
723 		default:
724 			error = EINVAL;
725 			break;
726 		}
727 		break;
728 	default:
729 		error = EINVAL;
730 		break;
731 	}
732 done:
733 	if (p)
734 		PRELE(p);
735 	lwkt_reltoken(&proc_token);
736 
737 	return (error);
738 }
739 
740 /*
741  * MPSAFE
742  */
743 int
744 sys_setrlimit(struct __setrlimit_args *uap)
745 {
746 	struct rlimit alim;
747 	int error;
748 
749 	error = copyin(uap->rlp, &alim, sizeof(alim));
750 	if (error)
751 		return (error);
752 
753 	error = kern_setrlimit(uap->which, &alim);
754 
755 	return (error);
756 }
757 
758 /*
759  * MPSAFE
760  */
761 int
762 sys_getrlimit(struct __getrlimit_args *uap)
763 {
764 	struct rlimit lim;
765 	int error;
766 
767 	error = kern_getrlimit(uap->which, &lim);
768 
769 	if (error == 0)
770 		error = copyout(&lim, uap->rlp, sizeof(*uap->rlp));
771 	return error;
772 }
773 
774 /*
775  * Transform the running time and tick information in lwp lp's thread into user,
776  * system, and interrupt time usage.
777  *
778  * Since we are limited to statclock tick granularity this is a statisical
779  * calculation which will be correct over the long haul, but should not be
780  * expected to measure fine grained deltas.
781  *
782  * It is possible to catch a lwp in the midst of being created, so
783  * check whether lwp_thread is NULL or not.
784  */
785 void
786 calcru(struct lwp *lp, struct timeval *up, struct timeval *sp)
787 {
788 	struct thread *td;
789 
790 	/*
791 	 * Calculate at the statclock level.  YYY if the thread is owned by
792 	 * another cpu we need to forward the request to the other cpu, or
793 	 * have a token to interlock the information in order to avoid racing
794 	 * thread destruction.
795 	 */
796 	if ((td = lp->lwp_thread) != NULL) {
797 		crit_enter();
798 		up->tv_sec = td->td_uticks / 1000000;
799 		up->tv_usec = td->td_uticks % 1000000;
800 		sp->tv_sec = td->td_sticks / 1000000;
801 		sp->tv_usec = td->td_sticks % 1000000;
802 		crit_exit();
803 	}
804 }
805 
806 /*
807  * Aggregate resource statistics of all lwps of a process.
808  *
809  * proc.p_ru keeps track of all statistics directly related to a proc.  This
810  * consists of RSS usage and nswap information and aggregate numbers for all
811  * former lwps of this proc.
812  *
813  * proc.p_cru is the sum of all stats of reaped children.
814  *
815  * lwp.lwp_ru contains the stats directly related to one specific lwp, meaning
816  * packet, scheduler switch or page fault counts, etc.  This information gets
817  * added to lwp.lwp_proc.p_ru when the lwp exits.
818  */
819 void
820 calcru_proc(struct proc *p, struct rusage *ru)
821 {
822 	struct timeval upt, spt;
823 	long *rip1, *rip2;
824 	struct lwp *lp;
825 
826 	*ru = p->p_ru;
827 
828 	FOREACH_LWP_IN_PROC(lp, p) {
829 		calcru(lp, &upt, &spt);
830 		timevaladd(&ru->ru_utime, &upt);
831 		timevaladd(&ru->ru_stime, &spt);
832 		for (rip1 = &ru->ru_first, rip2 = &lp->lwp_ru.ru_first;
833 		     rip1 <= &ru->ru_last;
834 		     rip1++, rip2++)
835 			*rip1 += *rip2;
836 	}
837 }
838 
839 
840 /*
841  * MPALMOSTSAFE
842  */
843 int
844 sys_getrusage(struct getrusage_args *uap)
845 {
846 	struct rusage ru;
847 	struct rusage *rup;
848 	int error;
849 
850 	lwkt_gettoken(&proc_token);
851 
852 	switch (uap->who) {
853 	case RUSAGE_SELF:
854 		rup = &ru;
855 		calcru_proc(curproc, rup);
856 		error = 0;
857 		break;
858 	case RUSAGE_CHILDREN:
859 		rup = &curproc->p_cru;
860 		error = 0;
861 		break;
862 	default:
863 		error = EINVAL;
864 		break;
865 	}
866 	if (error == 0)
867 		error = copyout(rup, uap->rusage, sizeof(struct rusage));
868 	lwkt_reltoken(&proc_token);
869 	return (error);
870 }
871 
872 void
873 ruadd(struct rusage *ru, struct rusage *ru2)
874 {
875 	long *ip, *ip2;
876 	int i;
877 
878 	timevaladd(&ru->ru_utime, &ru2->ru_utime);
879 	timevaladd(&ru->ru_stime, &ru2->ru_stime);
880 	if (ru->ru_maxrss < ru2->ru_maxrss)
881 		ru->ru_maxrss = ru2->ru_maxrss;
882 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
883 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
884 		*ip++ += *ip2++;
885 }
886 
887 /*
888  * Find the uidinfo structure for a uid.  This structure is used to
889  * track the total resource consumption (process count, socket buffer
890  * size, etc.) for the uid and impose limits.
891  */
892 void
893 uihashinit(void)
894 {
895 	spin_init(&uihash_lock);
896 	uihashtbl = hashinit(maxproc / 16, M_UIDINFO, &uihash);
897 }
898 
899 /*
900  * NOTE: Must be called with uihash_lock held
901  *
902  * MPSAFE
903  */
904 static struct uidinfo *
905 uilookup(uid_t uid)
906 {
907 	struct	uihashhead *uipp;
908 	struct	uidinfo *uip;
909 
910 	uipp = UIHASH(uid);
911 	LIST_FOREACH(uip, uipp, ui_hash) {
912 		if (uip->ui_uid == uid)
913 			break;
914 	}
915 	return (uip);
916 }
917 
918 /*
919  * Helper function to creat ea uid that could not be found.
920  * This function will properly deal with races.
921  *
922  * MPSAFE
923  */
924 static struct uidinfo *
925 uicreate(uid_t uid)
926 {
927 	struct	uidinfo *uip, *tmp;
928 
929 	/*
930 	 * Allocate space and check for a race
931 	 */
932 	uip = kmalloc(sizeof(*uip), M_UIDINFO, M_WAITOK|M_ZERO);
933 
934 	/*
935 	 * Initialize structure and enter it into the hash table
936 	 */
937 	spin_init(&uip->ui_lock);
938 	uip->ui_uid = uid;
939 	uip->ui_ref = 1;	/* we're returning a ref */
940 	varsymset_init(&uip->ui_varsymset, NULL);
941 
942 	/*
943 	 * Somebody may have already created the uidinfo for this
944 	 * uid. If so, return that instead.
945 	 */
946 	spin_lock(&uihash_lock);
947 	tmp = uilookup(uid);
948 	if (tmp != NULL) {
949 		uihold(tmp);
950 		spin_unlock(&uihash_lock);
951 
952 		spin_uninit(&uip->ui_lock);
953 		varsymset_clean(&uip->ui_varsymset);
954 		FREE(uip, M_UIDINFO);
955 		uip = tmp;
956 	} else {
957 		LIST_INSERT_HEAD(UIHASH(uid), uip, ui_hash);
958 		spin_unlock(&uihash_lock);
959 	}
960 	return (uip);
961 }
962 
963 /*
964  *
965  *
966  * MPSAFE
967  */
968 struct uidinfo *
969 uifind(uid_t uid)
970 {
971 	struct	uidinfo *uip;
972 
973 	spin_lock(&uihash_lock);
974 	uip = uilookup(uid);
975 	if (uip == NULL) {
976 		spin_unlock(&uihash_lock);
977 		uip = uicreate(uid);
978 	} else {
979 		uihold(uip);
980 		spin_unlock(&uihash_lock);
981 	}
982 	return (uip);
983 }
984 
985 /*
986  * Helper funtion to remove a uidinfo whos reference count is
987  * transitioning from 1->0.  The reference count is 1 on call.
988  *
989  * Zero is returned on success, otherwise non-zero and the
990  * uiphas not been removed.
991  *
992  * MPSAFE
993  */
994 static __inline int
995 uifree(struct uidinfo *uip)
996 {
997 	/*
998 	 * If we are still the only holder after acquiring the uihash_lock
999 	 * we can safely unlink the uip and destroy it.  Otherwise we lost
1000 	 * a race and must fail.
1001 	 */
1002 	spin_lock(&uihash_lock);
1003 	if (uip->ui_ref != 1) {
1004 		spin_unlock(&uihash_lock);
1005 		return(-1);
1006 	}
1007 	LIST_REMOVE(uip, ui_hash);
1008 	spin_unlock(&uihash_lock);
1009 
1010 	/*
1011 	 * The uip is now orphaned and we can destroy it at our
1012 	 * leisure.
1013 	 */
1014 	if (uip->ui_sbsize != 0)
1015 		kprintf("freeing uidinfo: uid = %d, sbsize = %jd\n",
1016 		    uip->ui_uid, (intmax_t)uip->ui_sbsize);
1017 	if (uip->ui_proccnt != 0)
1018 		kprintf("freeing uidinfo: uid = %d, proccnt = %ld\n",
1019 		    uip->ui_uid, uip->ui_proccnt);
1020 
1021 	varsymset_clean(&uip->ui_varsymset);
1022 	lockuninit(&uip->ui_varsymset.vx_lock);
1023 	spin_uninit(&uip->ui_lock);
1024 	FREE(uip, M_UIDINFO);
1025 	return(0);
1026 }
1027 
1028 /*
1029  * MPSAFE
1030  */
1031 void
1032 uihold(struct uidinfo *uip)
1033 {
1034 	atomic_add_int(&uip->ui_ref, 1);
1035 	KKASSERT(uip->ui_ref >= 0);
1036 }
1037 
1038 /*
1039  * NOTE: It is important for us to not drop the ref count to 0
1040  *	 because this can cause a 2->0/2->0 race with another
1041  *	 concurrent dropper.  Losing the race in that situation
1042  *	 can cause uip to become stale for one of the other
1043  *	 threads.
1044  *
1045  * MPSAFE
1046  */
1047 void
1048 uidrop(struct uidinfo *uip)
1049 {
1050 	int ref;
1051 
1052 	KKASSERT(uip->ui_ref > 0);
1053 
1054 	for (;;) {
1055 		ref = uip->ui_ref;
1056 		cpu_ccfence();
1057 		if (ref == 1) {
1058 			if (uifree(uip) == 0)
1059 				break;
1060 		} else if (atomic_cmpset_int(&uip->ui_ref, ref, ref - 1)) {
1061 			break;
1062 		}
1063 		/* else retry */
1064 	}
1065 }
1066 
1067 void
1068 uireplace(struct uidinfo **puip, struct uidinfo *nuip)
1069 {
1070 	uidrop(*puip);
1071 	*puip = nuip;
1072 }
1073 
1074 /*
1075  * Change the count associated with number of processes
1076  * a given user is using.  When 'max' is 0, don't enforce a limit
1077  */
1078 int
1079 chgproccnt(struct uidinfo *uip, int diff, int max)
1080 {
1081 	int ret;
1082 	spin_lock(&uip->ui_lock);
1083 	/* don't allow them to exceed max, but allow subtraction */
1084 	if (diff > 0 && uip->ui_proccnt + diff > max && max != 0) {
1085 		ret = 0;
1086 	} else {
1087 		uip->ui_proccnt += diff;
1088 		if (uip->ui_proccnt < 0)
1089 			kprintf("negative proccnt for uid = %d\n", uip->ui_uid);
1090 		ret = 1;
1091 	}
1092 	spin_unlock(&uip->ui_lock);
1093 	return ret;
1094 }
1095 
1096 /*
1097  * Change the total socket buffer size a user has used.
1098  */
1099 int
1100 chgsbsize(struct uidinfo *uip, u_long *hiwat, u_long to, rlim_t max)
1101 {
1102 	rlim_t new;
1103 
1104 	spin_lock(&uip->ui_lock);
1105 	new = uip->ui_sbsize + to - *hiwat;
1106 	KKASSERT(new >= 0);
1107 
1108 	/*
1109 	 * If we are trying to increase the socket buffer size
1110 	 * Scale down the hi water mark when we exceed the user's
1111 	 * allowed socket buffer space.
1112 	 *
1113 	 * We can't scale down too much or we will blow up atomic packet
1114 	 * operations.
1115 	 */
1116 	if (to > *hiwat && to > MCLBYTES && new > max) {
1117 		to = to * max / new;
1118 		if (to < MCLBYTES)
1119 			to = MCLBYTES;
1120 	}
1121 	uip->ui_sbsize = new;
1122 	*hiwat = to;
1123 	spin_unlock(&uip->ui_lock);
1124 	return (1);
1125 }
1126 
1127