MFC r1.3 r1.13 r1.8 (HEAD):
[dragonfly.git] / sys / kern / kern_resource.c
blob2e074294da6af2f2c4030c0cba4a7598512f0c71
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.
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.
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.
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 $
43 #include "opt_compat.h"
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/time.h>
55 #include <sys/lockf.h>
57 #include <vm/vm.h>
58 #include <vm/vm_param.h>
59 #include <sys/lock.h>
60 #include <vm/pmap.h>
61 #include <vm/vm_map.h>
63 #include <sys/thread2.h>
65 static int donice (struct proc *chgp, int n);
67 static MALLOC_DEFINE(M_UIDINFO, "uidinfo", "uidinfo structures");
68 #define UIHASH(uid) (&uihashtbl[(uid) & uihash])
69 static LIST_HEAD(uihashhead, uidinfo) *uihashtbl;
70 static u_long uihash; /* size of hash table - 1 */
72 static struct uidinfo *uicreate (uid_t uid);
73 static struct uidinfo *uilookup (uid_t uid);
76 * Resource controls and accounting.
79 struct getpriority_info {
80 int low;
81 int who;
84 static int getpriority_callback(struct proc *p, void *data);
86 int
87 sys_getpriority(struct getpriority_args *uap)
89 struct getpriority_info info;
90 struct proc *curp = curproc;
91 struct proc *p;
92 int low = PRIO_MAX + 1;
94 switch (uap->which) {
95 case PRIO_PROCESS:
96 if (uap->who == 0)
97 p = curp;
98 else
99 p = pfind(uap->who);
100 if (p == 0)
101 break;
102 if (!PRISON_CHECK(curp->p_ucred, p->p_ucred))
103 break;
104 low = p->p_nice;
105 break;
107 case PRIO_PGRP:
109 struct pgrp *pg;
111 if (uap->who == 0)
112 pg = curp->p_pgrp;
113 else if ((pg = pgfind(uap->who)) == NULL)
114 break;
115 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
116 if ((PRISON_CHECK(curp->p_ucred, p->p_ucred) && p->p_nice < low))
117 low = p->p_nice;
119 break;
121 case PRIO_USER:
122 if (uap->who == 0)
123 uap->who = curp->p_ucred->cr_uid;
124 info.low = low;
125 info.who = uap->who;
126 allproc_scan(getpriority_callback, &info);
127 low = info.low;
128 break;
130 default:
131 return (EINVAL);
133 if (low == PRIO_MAX + 1)
134 return (ESRCH);
135 uap->sysmsg_result = low;
136 return (0);
140 * Figure out the current lowest nice priority for processes owned
141 * by the specified user.
143 static
145 getpriority_callback(struct proc *p, void *data)
147 struct getpriority_info *info = data;
149 if (PRISON_CHECK(curproc->p_ucred, p->p_ucred) &&
150 p->p_ucred->cr_uid == info->who &&
151 p->p_nice < info->low) {
152 info->low = p->p_nice;
154 return(0);
157 struct setpriority_info {
158 int prio;
159 int who;
160 int error;
161 int found;
164 static int setpriority_callback(struct proc *p, void *data);
167 sys_setpriority(struct setpriority_args *uap)
169 struct setpriority_info info;
170 struct proc *curp = curproc;
171 struct proc *p;
172 int found = 0, error = 0;
174 switch (uap->which) {
175 case PRIO_PROCESS:
176 if (uap->who == 0)
177 p = curp;
178 else
179 p = pfind(uap->who);
180 if (p == 0)
181 break;
182 if (!PRISON_CHECK(curp->p_ucred, p->p_ucred))
183 break;
184 error = donice(p, uap->prio);
185 found++;
186 break;
188 case PRIO_PGRP:
190 struct pgrp *pg;
192 if (uap->who == 0)
193 pg = curp->p_pgrp;
194 else if ((pg = pgfind(uap->who)) == NULL)
195 break;
196 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
197 if (PRISON_CHECK(curp->p_ucred, p->p_ucred)) {
198 error = donice(p, uap->prio);
199 found++;
202 break;
204 case PRIO_USER:
205 if (uap->who == 0)
206 uap->who = curp->p_ucred->cr_uid;
207 info.prio = uap->prio;
208 info.who = uap->who;
209 info.error = 0;
210 info.found = 0;
211 allproc_scan(setpriority_callback, &info);
212 error = info.error;
213 found = info.found;
214 break;
216 default:
217 return (EINVAL);
219 if (found == 0)
220 return (ESRCH);
221 return (error);
224 static
226 setpriority_callback(struct proc *p, void *data)
228 struct setpriority_info *info = data;
229 int error;
231 if (p->p_ucred->cr_uid == info->who &&
232 PRISON_CHECK(curproc->p_ucred, p->p_ucred)) {
233 error = donice(p, info->prio);
234 if (error)
235 info->error = error;
236 ++info->found;
238 return(0);
241 static int
242 donice(struct proc *chgp, int n)
244 struct proc *curp = curproc;
245 struct ucred *cr = curp->p_ucred;
246 struct lwp *lp;
248 if (cr->cr_uid && cr->cr_ruid &&
249 cr->cr_uid != chgp->p_ucred->cr_uid &&
250 cr->cr_ruid != chgp->p_ucred->cr_uid)
251 return (EPERM);
252 if (n > PRIO_MAX)
253 n = PRIO_MAX;
254 if (n < PRIO_MIN)
255 n = PRIO_MIN;
256 if (n < chgp->p_nice && suser_cred(cr, 0))
257 return (EACCES);
258 chgp->p_nice = n;
259 FOREACH_LWP_IN_PROC(lp, chgp)
260 chgp->p_usched->resetpriority(lp);
261 return (0);
265 sys_lwp_rtprio(struct lwp_rtprio_args *uap)
267 struct proc *p = curproc;
268 struct lwp *lp;
269 struct rtprio rtp;
270 struct ucred *cr = p->p_ucred;
271 int error;
273 error = copyin(uap->rtp, &rtp, sizeof(struct rtprio));
274 if (error)
275 return error;
277 if (uap->pid < 0) {
278 return EINVAL;
279 } else if (uap->pid == 0) {
280 /* curproc already loaded on p */
281 } else {
282 p = pfind(uap->pid);
285 if (p == 0) {
286 return ESRCH;
289 if (uap->tid < -1) {
290 return EINVAL;
291 } else if (uap->tid == -1) {
293 * sadly, tid can be 0 so we can't use 0 here
294 * like sys_rtprio()
296 lp = curthread->td_lwp;
297 } else {
298 lp = lwp_rb_tree_RB_LOOKUP(&p->p_lwp_tree, uap->tid);
299 if (lp == NULL)
300 return ESRCH;
303 switch (uap->function) {
304 case RTP_LOOKUP:
305 return (copyout(&lp->lwp_rtprio, uap->rtp,
306 sizeof(struct rtprio)));
307 case RTP_SET:
308 if (cr->cr_uid && cr->cr_ruid &&
309 cr->cr_uid != p->p_ucred->cr_uid &&
310 cr->cr_ruid != p->p_ucred->cr_uid) {
311 return EPERM;
313 /* disallow setting rtprio in most cases if not superuser */
314 if (suser_cred(cr, 0)) {
315 /* can't set someone else's */
316 if (uap->pid) { /* XXX */
317 return EPERM;
319 /* can't set realtime priority */
321 * Realtime priority has to be restricted for reasons which should be
322 * obvious. However, for idle priority, there is a potential for
323 * system deadlock if an idleprio process gains a lock on a resource
324 * that other processes need (and the idleprio process can't run
325 * due to a CPU-bound normal process). Fix me! XXX
327 if (RTP_PRIO_IS_REALTIME(rtp.type)) {
328 return EPERM;
331 switch (rtp.type) {
332 #ifdef RTP_PRIO_FIFO
333 case RTP_PRIO_FIFO:
334 #endif
335 case RTP_PRIO_REALTIME:
336 case RTP_PRIO_NORMAL:
337 case RTP_PRIO_IDLE:
338 if (rtp.prio > RTP_PRIO_MAX)
339 return EINVAL;
340 lp->lwp_rtprio = rtp;
341 return 0;
342 default:
343 return EINVAL;
345 default:
346 return EINVAL;
348 panic("can't get here");
352 * Set realtime priority
354 /* ARGSUSED */
356 sys_rtprio(struct rtprio_args *uap)
358 struct proc *curp = curproc;
359 struct proc *p;
360 struct lwp *lp;
361 struct ucred *cr = curp->p_ucred;
362 struct rtprio rtp;
363 int error;
365 error = copyin(uap->rtp, &rtp, sizeof(struct rtprio));
366 if (error)
367 return (error);
369 if (uap->pid == 0)
370 p = curp;
371 else
372 p = pfind(uap->pid);
374 if (p == 0)
375 return (ESRCH);
377 /* XXX lwp */
378 lp = FIRST_LWP_IN_PROC(p);
379 switch (uap->function) {
380 case RTP_LOOKUP:
381 return (copyout(&lp->lwp_rtprio, uap->rtp, sizeof(struct rtprio)));
382 case RTP_SET:
383 if (cr->cr_uid && cr->cr_ruid &&
384 cr->cr_uid != p->p_ucred->cr_uid &&
385 cr->cr_ruid != p->p_ucred->cr_uid)
386 return (EPERM);
387 /* disallow setting rtprio in most cases if not superuser */
388 if (suser_cred(cr, 0)) {
389 /* can't set someone else's */
390 if (uap->pid)
391 return (EPERM);
392 /* can't set realtime priority */
394 * Realtime priority has to be restricted for reasons which should be
395 * obvious. However, for idle priority, there is a potential for
396 * system deadlock if an idleprio process gains a lock on a resource
397 * that other processes need (and the idleprio process can't run
398 * due to a CPU-bound normal process). Fix me! XXX
400 if (RTP_PRIO_IS_REALTIME(rtp.type))
401 return (EPERM);
403 switch (rtp.type) {
404 #ifdef RTP_PRIO_FIFO
405 case RTP_PRIO_FIFO:
406 #endif
407 case RTP_PRIO_REALTIME:
408 case RTP_PRIO_NORMAL:
409 case RTP_PRIO_IDLE:
410 if (rtp.prio > RTP_PRIO_MAX)
411 return (EINVAL);
412 lp->lwp_rtprio = rtp;
413 return (0);
414 default:
415 return (EINVAL);
418 default:
419 return (EINVAL);
424 sys_setrlimit(struct __setrlimit_args *uap)
426 struct rlimit alim;
427 int error;
429 error = copyin(uap->rlp, &alim, sizeof(alim));
430 if (error)
431 return (error);
433 error = kern_setrlimit(uap->which, &alim);
435 return (error);
439 sys_getrlimit(struct __getrlimit_args *uap)
441 struct rlimit lim;
442 int error;
444 error = kern_getrlimit(uap->which, &lim);
446 if (error == 0)
447 error = copyout(&lim, uap->rlp, sizeof(*uap->rlp));
448 return error;
452 * Transform the running time and tick information in lwp lp's thread into user,
453 * system, and interrupt time usage.
455 * Since we are limited to statclock tick granularity this is a statisical
456 * calculation which will be correct over the long haul, but should not be
457 * expected to measure fine grained deltas.
459 * It is possible to catch a lwp in the midst of being created, so
460 * check whether lwp_thread is NULL or not.
462 void
463 calcru(struct lwp *lp, struct timeval *up, struct timeval *sp)
465 struct thread *td;
468 * Calculate at the statclock level. YYY if the thread is owned by
469 * another cpu we need to forward the request to the other cpu, or
470 * have a token to interlock the information in order to avoid racing
471 * thread destruction.
473 if ((td = lp->lwp_thread) != NULL) {
474 crit_enter();
475 up->tv_sec = td->td_uticks / 1000000;
476 up->tv_usec = td->td_uticks % 1000000;
477 sp->tv_sec = td->td_sticks / 1000000;
478 sp->tv_usec = td->td_sticks % 1000000;
479 crit_exit();
484 * Aggregate resource statistics of all lwps of a process.
486 * proc.p_ru keeps track of all statistics directly related to a proc. This
487 * consists of RSS usage and nswap information and aggregate numbers for all
488 * former lwps of this proc.
490 * proc.p_cru is the sum of all stats of reaped children.
492 * lwp.lwp_ru contains the stats directly related to one specific lwp, meaning
493 * packet, scheduler switch or page fault counts, etc. This information gets
494 * added to lwp.lwp_proc.p_ru when the lwp exits.
496 void
497 calcru_proc(struct proc *p, struct rusage *ru)
499 struct timeval upt, spt;
500 long *rip1, *rip2;
501 struct lwp *lp;
503 *ru = p->p_ru;
505 FOREACH_LWP_IN_PROC(lp, p) {
506 calcru(lp, &upt, &spt);
507 timevaladd(&ru->ru_utime, &upt);
508 timevaladd(&ru->ru_stime, &spt);
509 for (rip1 = &ru->ru_first, rip2 = &lp->lwp_ru.ru_first;
510 rip1 <= &ru->ru_last;
511 rip1++, rip2++)
512 *rip1 += *rip2;
517 /* ARGSUSED */
519 sys_getrusage(struct getrusage_args *uap)
521 struct rusage ru;
522 struct rusage *rup;
524 switch (uap->who) {
526 case RUSAGE_SELF:
527 rup = &ru;
528 calcru_proc(curproc, rup);
529 break;
531 case RUSAGE_CHILDREN:
532 rup = &curproc->p_cru;
533 break;
535 default:
536 return (EINVAL);
538 return (copyout((caddr_t)rup, (caddr_t)uap->rusage,
539 sizeof (struct rusage)));
542 void
543 ruadd(struct rusage *ru, struct rusage *ru2)
545 long *ip, *ip2;
546 int i;
548 timevaladd(&ru->ru_utime, &ru2->ru_utime);
549 timevaladd(&ru->ru_stime, &ru2->ru_stime);
550 if (ru->ru_maxrss < ru2->ru_maxrss)
551 ru->ru_maxrss = ru2->ru_maxrss;
552 ip = &ru->ru_first; ip2 = &ru2->ru_first;
553 for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
554 *ip++ += *ip2++;
558 * Find the uidinfo structure for a uid. This structure is used to
559 * track the total resource consumption (process count, socket buffer
560 * size, etc.) for the uid and impose limits.
562 void
563 uihashinit(void)
565 uihashtbl = hashinit(maxproc / 16, M_UIDINFO, &uihash);
568 static struct uidinfo *
569 uilookup(uid_t uid)
571 struct uihashhead *uipp;
572 struct uidinfo *uip;
574 uipp = UIHASH(uid);
575 LIST_FOREACH(uip, uipp, ui_hash) {
576 if (uip->ui_uid == uid)
577 break;
579 return (uip);
582 static struct uidinfo *
583 uicreate(uid_t uid)
585 struct uidinfo *uip, *norace;
588 * Allocate space and check for a race
590 MALLOC(uip, struct uidinfo *, sizeof(*uip), M_UIDINFO, M_WAITOK);
591 norace = uilookup(uid);
592 if (norace != NULL) {
593 FREE(uip, M_UIDINFO);
594 return (norace);
598 * Initialize structure and enter it into the hash table
600 LIST_INSERT_HEAD(UIHASH(uid), uip, ui_hash);
601 uip->ui_uid = uid;
602 uip->ui_proccnt = 0;
603 uip->ui_sbsize = 0;
604 uip->ui_ref = 0;
605 uip->ui_posixlocks = 0;
606 varsymset_init(&uip->ui_varsymset, NULL);
607 return (uip);
610 struct uidinfo *
611 uifind(uid_t uid)
613 struct uidinfo *uip;
615 uip = uilookup(uid);
616 if (uip == NULL)
617 uip = uicreate(uid);
618 uip->ui_ref++;
619 return (uip);
622 static __inline void
623 uifree(struct uidinfo *uip)
625 if (uip->ui_sbsize != 0)
626 /* XXX no %qd in kernel. Truncate. */
627 kprintf("freeing uidinfo: uid = %d, sbsize = %ld\n",
628 uip->ui_uid, (long)uip->ui_sbsize);
629 if (uip->ui_proccnt != 0)
630 kprintf("freeing uidinfo: uid = %d, proccnt = %ld\n",
631 uip->ui_uid, uip->ui_proccnt);
632 LIST_REMOVE(uip, ui_hash);
633 varsymset_clean(&uip->ui_varsymset);
634 FREE(uip, M_UIDINFO);
637 void
638 uihold(struct uidinfo *uip)
640 ++uip->ui_ref;
641 KKASSERT(uip->ui_ref > 0);
644 void
645 uidrop(struct uidinfo *uip)
647 KKASSERT(uip->ui_ref > 0);
648 if (--uip->ui_ref == 0)
649 uifree(uip);
652 void
653 uireplace(struct uidinfo **puip, struct uidinfo *nuip)
655 uidrop(*puip);
656 *puip = nuip;
660 * Change the count associated with number of processes
661 * a given user is using. When 'max' is 0, don't enforce a limit
664 chgproccnt(struct uidinfo *uip, int diff, int max)
666 /* don't allow them to exceed max, but allow subtraction */
667 if (diff > 0 && uip->ui_proccnt + diff > max && max != 0)
668 return (0);
669 uip->ui_proccnt += diff;
670 if (uip->ui_proccnt < 0)
671 kprintf("negative proccnt for uid = %d\n", uip->ui_uid);
672 return (1);
676 * Change the total socket buffer size a user has used.
679 chgsbsize(struct uidinfo *uip, u_long *hiwat, u_long to, rlim_t max)
681 rlim_t new;
683 crit_enter();
684 new = uip->ui_sbsize + to - *hiwat;
687 * If we are trying to increase the socket buffer size
688 * Scale down the hi water mark when we exceed the user's
689 * allowed socket buffer space.
691 * We can't scale down too much or we will blow up atomic packet
692 * operations.
694 if (to > *hiwat && to > MCLBYTES && new > max) {
695 to = to * max / new;
696 if (to < MCLBYTES)
697 to = MCLBYTES;
699 uip->ui_sbsize = new;
700 *hiwat = to;
701 if (uip->ui_sbsize < 0)
702 kprintf("negative sbsize for uid = %d\n", uip->ui_uid);
703 crit_exit();
704 return (1);