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
| 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: head/sys/kern/kern_resource.c 103367 2002-09-15 23:52:25Z julian $
| 39 * $FreeBSD: head/sys/kern/kern_resource.c 103767 2002-09-21 22:07:17Z jake $
|
40 */
41
42#include "opt_compat.h"
43
44#include <sys/param.h>
45#include <sys/systm.h>
46#include <sys/sysproto.h>
47#include <sys/file.h>
48#include <sys/kernel.h>
49#include <sys/lock.h>
50#include <sys/malloc.h>
51#include <sys/mutex.h>
52#include <sys/proc.h>
53#include <sys/resourcevar.h>
54#include <sys/sx.h>
| 40 */
41
42#include "opt_compat.h"
43
44#include <sys/param.h>
45#include <sys/systm.h>
46#include <sys/sysproto.h>
47#include <sys/file.h>
48#include <sys/kernel.h>
49#include <sys/lock.h>
50#include <sys/malloc.h>
51#include <sys/mutex.h>
52#include <sys/proc.h>
53#include <sys/resourcevar.h>
54#include <sys/sx.h>
|
| 55#include <sys/sysent.h>
|
55#include <sys/time.h>
56
57#include <vm/vm.h>
58#include <vm/vm_param.h>
59#include <vm/pmap.h>
60#include <vm/vm_map.h>
61
62static int donice(struct thread *td, struct proc *chgp, int n);
63
64static MALLOC_DEFINE(M_UIDINFO, "uidinfo", "uidinfo structures");
65#define UIHASH(uid) (&uihashtbl[(uid) & uihash])
66static struct mtx uihashtbl_mtx;
67static LIST_HEAD(uihashhead, uidinfo) *uihashtbl;
68static u_long uihash; /* size of hash table - 1 */
69
70static struct uidinfo *uilookup(uid_t uid);
71
72/*
73 * Resource controls and accounting.
74 */
75
76#ifndef _SYS_SYSPROTO_H_
77struct getpriority_args {
78 int which;
79 int who;
80};
81#endif
82/*
83 * MPSAFE
84 */
85int
86getpriority(td, uap)
87 struct thread *td;
88 register struct getpriority_args *uap;
89{
90 struct proc *p;
91 int low = PRIO_MAX + 1;
92 int error = 0;
93 struct ksegrp *kg;
94
95 mtx_lock(&Giant);
96
97 switch (uap->which) {
98 case PRIO_PROCESS:
99 if (uap->who == 0)
100 low = td->td_ksegrp->kg_nice;
101 else {
102 p = pfind(uap->who);
103 if (p == NULL)
104 break;
105 if (p_cansee(td, p) == 0) {
106 FOREACH_KSEGRP_IN_PROC(p, kg) {
107 if (kg->kg_nice < low)
108 low = kg->kg_nice;
109 }
110 }
111 PROC_UNLOCK(p);
112 }
113 break;
114
115 case PRIO_PGRP: {
116 register struct pgrp *pg;
117
118 sx_slock(&proctree_lock);
119 if (uap->who == 0) {
120 pg = td->td_proc->p_pgrp;
121 PGRP_LOCK(pg);
122 } else {
123 pg = pgfind(uap->who);
124 if (pg == NULL) {
125 sx_sunlock(&proctree_lock);
126 break;
127 }
128 }
129 sx_sunlock(&proctree_lock);
130 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
131 PROC_LOCK(p);
132 if (!p_cansee(td, p)) {
133 FOREACH_KSEGRP_IN_PROC(p, kg) {
134 if (kg->kg_nice < low)
135 low = kg->kg_nice;
136 }
137 }
138 PROC_UNLOCK(p);
139 }
140 PGRP_UNLOCK(pg);
141 break;
142 }
143
144 case PRIO_USER:
145 if (uap->who == 0)
146 uap->who = td->td_ucred->cr_uid;
147 sx_slock(&allproc_lock);
148 LIST_FOREACH(p, &allproc, p_list) {
149 PROC_LOCK(p);
150 if (!p_cansee(td, p) &&
151 p->p_ucred->cr_uid == uap->who) {
152 FOREACH_KSEGRP_IN_PROC(p, kg) {
153 if (kg->kg_nice < low)
154 low = kg->kg_nice;
155 }
156 }
157 PROC_UNLOCK(p);
158 }
159 sx_sunlock(&allproc_lock);
160 break;
161
162 default:
163 error = EINVAL;
164 break;
165 }
166 if (low == PRIO_MAX + 1 && error == 0)
167 error = ESRCH;
168 td->td_retval[0] = low;
169 mtx_unlock(&Giant);
170 return (error);
171}
172
173#ifndef _SYS_SYSPROTO_H_
174struct setpriority_args {
175 int which;
176 int who;
177 int prio;
178};
179#endif
180/*
181 * MPSAFE
182 */
183/* ARGSUSED */
184int
185setpriority(td, uap)
186 struct thread *td;
187 register struct setpriority_args *uap;
188{
189 struct proc *curp = td->td_proc;
190 register struct proc *p;
191 int found = 0, error = 0;
192
193 mtx_lock(&Giant);
194
195 switch (uap->which) {
196 case PRIO_PROCESS:
197 if (uap->who == 0) {
198 PROC_LOCK(curp);
199 error = donice(td, curp, uap->prio);
200 PROC_UNLOCK(curp);
201 } else {
202 p = pfind(uap->who);
203 if (p == 0)
204 break;
205 if (p_cansee(td, p) == 0)
206 error = donice(td, p, uap->prio);
207 PROC_UNLOCK(p);
208 }
209 found++;
210 break;
211
212 case PRIO_PGRP: {
213 register struct pgrp *pg;
214
215 sx_slock(&proctree_lock);
216 if (uap->who == 0) {
217 pg = curp->p_pgrp;
218 PGRP_LOCK(pg);
219 } else {
220 pg = pgfind(uap->who);
221 if (pg == NULL) {
222 sx_sunlock(&proctree_lock);
223 break;
224 }
225 }
226 sx_sunlock(&proctree_lock);
227 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
228 PROC_LOCK(p);
229 if (!p_cansee(td, p)) {
230 error = donice(td, p, uap->prio);
231 found++;
232 }
233 PROC_UNLOCK(p);
234 }
235 PGRP_UNLOCK(pg);
236 break;
237 }
238
239 case PRIO_USER:
240 if (uap->who == 0)
241 uap->who = td->td_ucred->cr_uid;
242 sx_slock(&allproc_lock);
243 FOREACH_PROC_IN_SYSTEM(p) {
244 PROC_LOCK(p);
245 if (p->p_ucred->cr_uid == uap->who &&
246 !p_cansee(td, p)) {
247 error = donice(td, p, uap->prio);
248 found++;
249 }
250 PROC_UNLOCK(p);
251 }
252 sx_sunlock(&allproc_lock);
253 break;
254
255 default:
256 error = EINVAL;
257 break;
258 }
259 if (found == 0 && error == 0)
260 error = ESRCH;
261 mtx_unlock(&Giant);
262 return (error);
263}
264
265/*
266 * Set "nice" for a process. Doesn't really understand threaded processes well
267 * but does try. Has the unfortunate side effect of making all the NICE
268 * values for a process's ksegrps the same.. This suggests that
269 * NICE valuse should be stored as a process nice and deltas for the ksegrps.
270 * (but not yet).
271 */
272static int
273donice(struct thread *td, struct proc *p, int n)
274{
275 int error;
276 int low = PRIO_MAX + 1;
277 struct ksegrp *kg;
278
279 PROC_LOCK_ASSERT(p, MA_OWNED);
280 if ((error = p_cansched(td, p)))
281 return (error);
282 if (n > PRIO_MAX)
283 n = PRIO_MAX;
284 if (n < PRIO_MIN)
285 n = PRIO_MIN;
286 /*
287 * Only allow nicing if to more than the lowest nice.
288 * e.g. nices of 4,3,2 allow nice to 3 but not 1
289 */
290 FOREACH_KSEGRP_IN_PROC(p, kg) {
291 if (kg->kg_nice < low)
292 low = kg->kg_nice;
293 }
294 if (n < low && suser(td))
295 return (EACCES);
296 FOREACH_KSEGRP_IN_PROC(p, kg) {
297 kg->kg_nice = n;
298 (void)resetpriority(kg);
299 }
300 return (0);
301}
302
303/* rtprio system call */
304#ifndef _SYS_SYSPROTO_H_
305struct rtprio_args {
306 int function;
307 pid_t pid;
308 struct rtprio *rtp;
309};
310#endif
311
312/*
313 * Set realtime priority
314 */
315
316/*
317 * MPSAFE
318 */
319/* ARGSUSED */
320int
321rtprio(td, uap)
322 struct thread *td;
323 register struct rtprio_args *uap;
324{
325 struct proc *curp = td->td_proc;
326 register struct proc *p;
327 struct rtprio rtp;
328 int error, cierror = 0;
329
330 /* Perform copyin before acquiring locks if needed. */
331 if (uap->function == RTP_SET)
332 cierror = copyin(uap->rtp, &rtp, sizeof(struct rtprio));
333
334 if (uap->pid == 0) {
335 p = curp;
336 PROC_LOCK(p);
337 } else {
338 p = pfind(uap->pid);
339 if (p == NULL)
340 return (ESRCH);
341 }
342
343 switch (uap->function) {
344 case RTP_LOOKUP:
345 if ((error = p_cansee(td, p)))
346 break;
347 mtx_lock_spin(&sched_lock);
348 pri_to_rtp(FIRST_KSEGRP_IN_PROC(p), &rtp);
349 mtx_unlock_spin(&sched_lock);
350 PROC_UNLOCK(p);
351 return (copyout(&rtp, uap->rtp, sizeof(struct rtprio)));
352 case RTP_SET:
353 if ((error = p_cansched(td, p)) || (error = cierror))
354 break;
355 /* disallow setting rtprio in most cases if not superuser */
356 if (suser(td) != 0) {
357 /* can't set someone else's */
358 if (uap->pid) {
359 error = EPERM;
360 break;
361 }
362 /* can't set realtime priority */
363/*
364 * Realtime priority has to be restricted for reasons which should be
365 * obvious. However, for idle priority, there is a potential for
366 * system deadlock if an idleprio process gains a lock on a resource
367 * that other processes need (and the idleprio process can't run
368 * due to a CPU-bound normal process). Fix me! XXX
369 */
370#if 0
371 if (RTP_PRIO_IS_REALTIME(rtp.type))
372#endif
373 if (rtp.type != RTP_PRIO_NORMAL) {
374 error = EPERM;
375 break;
376 }
377 }
378 mtx_lock_spin(&sched_lock);
379 error = rtp_to_pri(&rtp, FIRST_KSEGRP_IN_PROC(p));
380 mtx_unlock_spin(&sched_lock);
381 break;
382 default:
383 error = EINVAL;
384 break;
385 }
386 PROC_UNLOCK(p);
387 return (error);
388}
389
390int
391rtp_to_pri(struct rtprio *rtp, struct ksegrp *kg)
392{
393
394 if (rtp->prio > RTP_PRIO_MAX)
395 return (EINVAL);
396 switch (RTP_PRIO_BASE(rtp->type)) {
397 case RTP_PRIO_REALTIME:
398 kg->kg_user_pri = PRI_MIN_REALTIME + rtp->prio;
399 break;
400 case RTP_PRIO_NORMAL:
401 kg->kg_user_pri = PRI_MIN_TIMESHARE + rtp->prio;
402 break;
403 case RTP_PRIO_IDLE:
404 kg->kg_user_pri = PRI_MIN_IDLE + rtp->prio;
405 break;
406 default:
407 return (EINVAL);
408 }
409 kg->kg_pri_class = rtp->type;
410 if (curthread->td_ksegrp == kg) {
411 curthread->td_base_pri = kg->kg_user_pri;
412 curthread->td_priority = kg->kg_user_pri; /* XXX dubious */
413 }
414 return (0);
415}
416
417void
418pri_to_rtp(struct ksegrp *kg, struct rtprio *rtp)
419{
420
421 switch (PRI_BASE(kg->kg_pri_class)) {
422 case PRI_REALTIME:
423 rtp->prio = kg->kg_user_pri - PRI_MIN_REALTIME;
424 break;
425 case PRI_TIMESHARE:
426 rtp->prio = kg->kg_user_pri - PRI_MIN_TIMESHARE;
427 break;
428 case PRI_IDLE:
429 rtp->prio = kg->kg_user_pri - PRI_MIN_IDLE;
430 break;
431 default:
432 break;
433 }
434 rtp->type = kg->kg_pri_class;
435}
436
437#if defined(COMPAT_43) || defined(COMPAT_SUNOS)
438#ifndef _SYS_SYSPROTO_H_
439struct osetrlimit_args {
440 u_int which;
441 struct orlimit *rlp;
442};
443#endif
444/*
445 * MPSAFE
446 */
447/* ARGSUSED */
448int
449osetrlimit(td, uap)
450 struct thread *td;
451 register struct osetrlimit_args *uap;
452{
453 struct orlimit olim;
454 struct rlimit lim;
455 int error;
456
457 if ((error = copyin(uap->rlp, &olim, sizeof(struct orlimit))))
458 return (error);
459 lim.rlim_cur = olim.rlim_cur;
460 lim.rlim_max = olim.rlim_max;
461 mtx_lock(&Giant);
462 error = dosetrlimit(td, uap->which, &lim);
463 mtx_unlock(&Giant);
464 return (error);
465}
466
467#ifndef _SYS_SYSPROTO_H_
468struct ogetrlimit_args {
469 u_int which;
470 struct orlimit *rlp;
471};
472#endif
473/*
474 * MPSAFE
475 */
476/* ARGSUSED */
477int
478ogetrlimit(td, uap)
479 struct thread *td;
480 register struct ogetrlimit_args *uap;
481{
482 struct proc *p = td->td_proc;
483 struct orlimit olim;
484 int error;
485
486 if (uap->which >= RLIM_NLIMITS)
487 return (EINVAL);
488 mtx_lock(&Giant);
489 olim.rlim_cur = p->p_rlimit[uap->which].rlim_cur;
490 if (olim.rlim_cur == -1)
491 olim.rlim_cur = 0x7fffffff;
492 olim.rlim_max = p->p_rlimit[uap->which].rlim_max;
493 if (olim.rlim_max == -1)
494 olim.rlim_max = 0x7fffffff;
495 error = copyout(&olim, uap->rlp, sizeof(olim));
496 mtx_unlock(&Giant);
497 return (error);
498}
499#endif /* COMPAT_43 || COMPAT_SUNOS */
500
501#ifndef _SYS_SYSPROTO_H_
502struct __setrlimit_args {
503 u_int which;
504 struct rlimit *rlp;
505};
506#endif
507/*
508 * MPSAFE
509 */
510/* ARGSUSED */
511int
512setrlimit(td, uap)
513 struct thread *td;
514 register struct __setrlimit_args *uap;
515{
516 struct rlimit alim;
517 int error;
518
519 if ((error = copyin(uap->rlp, &alim, sizeof (struct rlimit))))
520 return (error);
521 mtx_lock(&Giant);
522 error = dosetrlimit(td, uap->which, &alim);
523 mtx_unlock(&Giant);
524 return (error);
525}
526
527int
528dosetrlimit(td, which, limp)
529 struct thread *td;
530 u_int which;
531 struct rlimit *limp;
532{
533 struct proc *p = td->td_proc;
534 register struct rlimit *alimp;
535 int error;
536
537 GIANT_REQUIRED;
538
539 if (which >= RLIM_NLIMITS)
540 return (EINVAL);
541 alimp = &p->p_rlimit[which];
542
543 /*
544 * Preserve historical bugs by treating negative limits as unsigned.
545 */
546 if (limp->rlim_cur < 0)
547 limp->rlim_cur = RLIM_INFINITY;
548 if (limp->rlim_max < 0)
549 limp->rlim_max = RLIM_INFINITY;
550
551 if (limp->rlim_cur > alimp->rlim_max ||
552 limp->rlim_max > alimp->rlim_max)
553 if ((error = suser_cred(td->td_ucred, PRISON_ROOT)))
554 return (error);
555 if (limp->rlim_cur > limp->rlim_max)
556 limp->rlim_cur = limp->rlim_max;
557 if (p->p_limit->p_refcnt > 1 &&
558 (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
559 p->p_limit->p_refcnt--;
560 p->p_limit = limcopy(p->p_limit);
561 alimp = &p->p_rlimit[which];
562 }
563
564 switch (which) {
565
566 case RLIMIT_CPU:
567 if (limp->rlim_cur > RLIM_INFINITY / (rlim_t)1000000)
568 p->p_limit->p_cpulimit = RLIM_INFINITY;
569 else
570 p->p_limit->p_cpulimit =
571 (rlim_t)1000000 * limp->rlim_cur;
572 break;
573 case RLIMIT_DATA:
574 if (limp->rlim_cur > maxdsiz)
575 limp->rlim_cur = maxdsiz;
576 if (limp->rlim_max > maxdsiz)
577 limp->rlim_max = maxdsiz;
578 break;
579
580 case RLIMIT_STACK:
581 if (limp->rlim_cur > maxssiz)
582 limp->rlim_cur = maxssiz;
583 if (limp->rlim_max > maxssiz)
584 limp->rlim_max = maxssiz;
585 /*
586 * Stack is allocated to the max at exec time with only
587 * "rlim_cur" bytes accessible. If stack limit is going
588 * up make more accessible, if going down make inaccessible.
589 */
590 if (limp->rlim_cur != alimp->rlim_cur) {
591 vm_offset_t addr;
592 vm_size_t size;
593 vm_prot_t prot;
594
595 if (limp->rlim_cur > alimp->rlim_cur) {
| 56#include <sys/time.h>
57
58#include <vm/vm.h>
59#include <vm/vm_param.h>
60#include <vm/pmap.h>
61#include <vm/vm_map.h>
62
63static int donice(struct thread *td, struct proc *chgp, int n);
64
65static MALLOC_DEFINE(M_UIDINFO, "uidinfo", "uidinfo structures");
66#define UIHASH(uid) (&uihashtbl[(uid) & uihash])
67static struct mtx uihashtbl_mtx;
68static LIST_HEAD(uihashhead, uidinfo) *uihashtbl;
69static u_long uihash; /* size of hash table - 1 */
70
71static struct uidinfo *uilookup(uid_t uid);
72
73/*
74 * Resource controls and accounting.
75 */
76
77#ifndef _SYS_SYSPROTO_H_
78struct getpriority_args {
79 int which;
80 int who;
81};
82#endif
83/*
84 * MPSAFE
85 */
86int
87getpriority(td, uap)
88 struct thread *td;
89 register struct getpriority_args *uap;
90{
91 struct proc *p;
92 int low = PRIO_MAX + 1;
93 int error = 0;
94 struct ksegrp *kg;
95
96 mtx_lock(&Giant);
97
98 switch (uap->which) {
99 case PRIO_PROCESS:
100 if (uap->who == 0)
101 low = td->td_ksegrp->kg_nice;
102 else {
103 p = pfind(uap->who);
104 if (p == NULL)
105 break;
106 if (p_cansee(td, p) == 0) {
107 FOREACH_KSEGRP_IN_PROC(p, kg) {
108 if (kg->kg_nice < low)
109 low = kg->kg_nice;
110 }
111 }
112 PROC_UNLOCK(p);
113 }
114 break;
115
116 case PRIO_PGRP: {
117 register struct pgrp *pg;
118
119 sx_slock(&proctree_lock);
120 if (uap->who == 0) {
121 pg = td->td_proc->p_pgrp;
122 PGRP_LOCK(pg);
123 } else {
124 pg = pgfind(uap->who);
125 if (pg == NULL) {
126 sx_sunlock(&proctree_lock);
127 break;
128 }
129 }
130 sx_sunlock(&proctree_lock);
131 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
132 PROC_LOCK(p);
133 if (!p_cansee(td, p)) {
134 FOREACH_KSEGRP_IN_PROC(p, kg) {
135 if (kg->kg_nice < low)
136 low = kg->kg_nice;
137 }
138 }
139 PROC_UNLOCK(p);
140 }
141 PGRP_UNLOCK(pg);
142 break;
143 }
144
145 case PRIO_USER:
146 if (uap->who == 0)
147 uap->who = td->td_ucred->cr_uid;
148 sx_slock(&allproc_lock);
149 LIST_FOREACH(p, &allproc, p_list) {
150 PROC_LOCK(p);
151 if (!p_cansee(td, p) &&
152 p->p_ucred->cr_uid == uap->who) {
153 FOREACH_KSEGRP_IN_PROC(p, kg) {
154 if (kg->kg_nice < low)
155 low = kg->kg_nice;
156 }
157 }
158 PROC_UNLOCK(p);
159 }
160 sx_sunlock(&allproc_lock);
161 break;
162
163 default:
164 error = EINVAL;
165 break;
166 }
167 if (low == PRIO_MAX + 1 && error == 0)
168 error = ESRCH;
169 td->td_retval[0] = low;
170 mtx_unlock(&Giant);
171 return (error);
172}
173
174#ifndef _SYS_SYSPROTO_H_
175struct setpriority_args {
176 int which;
177 int who;
178 int prio;
179};
180#endif
181/*
182 * MPSAFE
183 */
184/* ARGSUSED */
185int
186setpriority(td, uap)
187 struct thread *td;
188 register struct setpriority_args *uap;
189{
190 struct proc *curp = td->td_proc;
191 register struct proc *p;
192 int found = 0, error = 0;
193
194 mtx_lock(&Giant);
195
196 switch (uap->which) {
197 case PRIO_PROCESS:
198 if (uap->who == 0) {
199 PROC_LOCK(curp);
200 error = donice(td, curp, uap->prio);
201 PROC_UNLOCK(curp);
202 } else {
203 p = pfind(uap->who);
204 if (p == 0)
205 break;
206 if (p_cansee(td, p) == 0)
207 error = donice(td, p, uap->prio);
208 PROC_UNLOCK(p);
209 }
210 found++;
211 break;
212
213 case PRIO_PGRP: {
214 register struct pgrp *pg;
215
216 sx_slock(&proctree_lock);
217 if (uap->who == 0) {
218 pg = curp->p_pgrp;
219 PGRP_LOCK(pg);
220 } else {
221 pg = pgfind(uap->who);
222 if (pg == NULL) {
223 sx_sunlock(&proctree_lock);
224 break;
225 }
226 }
227 sx_sunlock(&proctree_lock);
228 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
229 PROC_LOCK(p);
230 if (!p_cansee(td, p)) {
231 error = donice(td, p, uap->prio);
232 found++;
233 }
234 PROC_UNLOCK(p);
235 }
236 PGRP_UNLOCK(pg);
237 break;
238 }
239
240 case PRIO_USER:
241 if (uap->who == 0)
242 uap->who = td->td_ucred->cr_uid;
243 sx_slock(&allproc_lock);
244 FOREACH_PROC_IN_SYSTEM(p) {
245 PROC_LOCK(p);
246 if (p->p_ucred->cr_uid == uap->who &&
247 !p_cansee(td, p)) {
248 error = donice(td, p, uap->prio);
249 found++;
250 }
251 PROC_UNLOCK(p);
252 }
253 sx_sunlock(&allproc_lock);
254 break;
255
256 default:
257 error = EINVAL;
258 break;
259 }
260 if (found == 0 && error == 0)
261 error = ESRCH;
262 mtx_unlock(&Giant);
263 return (error);
264}
265
266/*
267 * Set "nice" for a process. Doesn't really understand threaded processes well
268 * but does try. Has the unfortunate side effect of making all the NICE
269 * values for a process's ksegrps the same.. This suggests that
270 * NICE valuse should be stored as a process nice and deltas for the ksegrps.
271 * (but not yet).
272 */
273static int
274donice(struct thread *td, struct proc *p, int n)
275{
276 int error;
277 int low = PRIO_MAX + 1;
278 struct ksegrp *kg;
279
280 PROC_LOCK_ASSERT(p, MA_OWNED);
281 if ((error = p_cansched(td, p)))
282 return (error);
283 if (n > PRIO_MAX)
284 n = PRIO_MAX;
285 if (n < PRIO_MIN)
286 n = PRIO_MIN;
287 /*
288 * Only allow nicing if to more than the lowest nice.
289 * e.g. nices of 4,3,2 allow nice to 3 but not 1
290 */
291 FOREACH_KSEGRP_IN_PROC(p, kg) {
292 if (kg->kg_nice < low)
293 low = kg->kg_nice;
294 }
295 if (n < low && suser(td))
296 return (EACCES);
297 FOREACH_KSEGRP_IN_PROC(p, kg) {
298 kg->kg_nice = n;
299 (void)resetpriority(kg);
300 }
301 return (0);
302}
303
304/* rtprio system call */
305#ifndef _SYS_SYSPROTO_H_
306struct rtprio_args {
307 int function;
308 pid_t pid;
309 struct rtprio *rtp;
310};
311#endif
312
313/*
314 * Set realtime priority
315 */
316
317/*
318 * MPSAFE
319 */
320/* ARGSUSED */
321int
322rtprio(td, uap)
323 struct thread *td;
324 register struct rtprio_args *uap;
325{
326 struct proc *curp = td->td_proc;
327 register struct proc *p;
328 struct rtprio rtp;
329 int error, cierror = 0;
330
331 /* Perform copyin before acquiring locks if needed. */
332 if (uap->function == RTP_SET)
333 cierror = copyin(uap->rtp, &rtp, sizeof(struct rtprio));
334
335 if (uap->pid == 0) {
336 p = curp;
337 PROC_LOCK(p);
338 } else {
339 p = pfind(uap->pid);
340 if (p == NULL)
341 return (ESRCH);
342 }
343
344 switch (uap->function) {
345 case RTP_LOOKUP:
346 if ((error = p_cansee(td, p)))
347 break;
348 mtx_lock_spin(&sched_lock);
349 pri_to_rtp(FIRST_KSEGRP_IN_PROC(p), &rtp);
350 mtx_unlock_spin(&sched_lock);
351 PROC_UNLOCK(p);
352 return (copyout(&rtp, uap->rtp, sizeof(struct rtprio)));
353 case RTP_SET:
354 if ((error = p_cansched(td, p)) || (error = cierror))
355 break;
356 /* disallow setting rtprio in most cases if not superuser */
357 if (suser(td) != 0) {
358 /* can't set someone else's */
359 if (uap->pid) {
360 error = EPERM;
361 break;
362 }
363 /* can't set realtime priority */
364/*
365 * Realtime priority has to be restricted for reasons which should be
366 * obvious. However, for idle priority, there is a potential for
367 * system deadlock if an idleprio process gains a lock on a resource
368 * that other processes need (and the idleprio process can't run
369 * due to a CPU-bound normal process). Fix me! XXX
370 */
371#if 0
372 if (RTP_PRIO_IS_REALTIME(rtp.type))
373#endif
374 if (rtp.type != RTP_PRIO_NORMAL) {
375 error = EPERM;
376 break;
377 }
378 }
379 mtx_lock_spin(&sched_lock);
380 error = rtp_to_pri(&rtp, FIRST_KSEGRP_IN_PROC(p));
381 mtx_unlock_spin(&sched_lock);
382 break;
383 default:
384 error = EINVAL;
385 break;
386 }
387 PROC_UNLOCK(p);
388 return (error);
389}
390
391int
392rtp_to_pri(struct rtprio *rtp, struct ksegrp *kg)
393{
394
395 if (rtp->prio > RTP_PRIO_MAX)
396 return (EINVAL);
397 switch (RTP_PRIO_BASE(rtp->type)) {
398 case RTP_PRIO_REALTIME:
399 kg->kg_user_pri = PRI_MIN_REALTIME + rtp->prio;
400 break;
401 case RTP_PRIO_NORMAL:
402 kg->kg_user_pri = PRI_MIN_TIMESHARE + rtp->prio;
403 break;
404 case RTP_PRIO_IDLE:
405 kg->kg_user_pri = PRI_MIN_IDLE + rtp->prio;
406 break;
407 default:
408 return (EINVAL);
409 }
410 kg->kg_pri_class = rtp->type;
411 if (curthread->td_ksegrp == kg) {
412 curthread->td_base_pri = kg->kg_user_pri;
413 curthread->td_priority = kg->kg_user_pri; /* XXX dubious */
414 }
415 return (0);
416}
417
418void
419pri_to_rtp(struct ksegrp *kg, struct rtprio *rtp)
420{
421
422 switch (PRI_BASE(kg->kg_pri_class)) {
423 case PRI_REALTIME:
424 rtp->prio = kg->kg_user_pri - PRI_MIN_REALTIME;
425 break;
426 case PRI_TIMESHARE:
427 rtp->prio = kg->kg_user_pri - PRI_MIN_TIMESHARE;
428 break;
429 case PRI_IDLE:
430 rtp->prio = kg->kg_user_pri - PRI_MIN_IDLE;
431 break;
432 default:
433 break;
434 }
435 rtp->type = kg->kg_pri_class;
436}
437
438#if defined(COMPAT_43) || defined(COMPAT_SUNOS)
439#ifndef _SYS_SYSPROTO_H_
440struct osetrlimit_args {
441 u_int which;
442 struct orlimit *rlp;
443};
444#endif
445/*
446 * MPSAFE
447 */
448/* ARGSUSED */
449int
450osetrlimit(td, uap)
451 struct thread *td;
452 register struct osetrlimit_args *uap;
453{
454 struct orlimit olim;
455 struct rlimit lim;
456 int error;
457
458 if ((error = copyin(uap->rlp, &olim, sizeof(struct orlimit))))
459 return (error);
460 lim.rlim_cur = olim.rlim_cur;
461 lim.rlim_max = olim.rlim_max;
462 mtx_lock(&Giant);
463 error = dosetrlimit(td, uap->which, &lim);
464 mtx_unlock(&Giant);
465 return (error);
466}
467
468#ifndef _SYS_SYSPROTO_H_
469struct ogetrlimit_args {
470 u_int which;
471 struct orlimit *rlp;
472};
473#endif
474/*
475 * MPSAFE
476 */
477/* ARGSUSED */
478int
479ogetrlimit(td, uap)
480 struct thread *td;
481 register struct ogetrlimit_args *uap;
482{
483 struct proc *p = td->td_proc;
484 struct orlimit olim;
485 int error;
486
487 if (uap->which >= RLIM_NLIMITS)
488 return (EINVAL);
489 mtx_lock(&Giant);
490 olim.rlim_cur = p->p_rlimit[uap->which].rlim_cur;
491 if (olim.rlim_cur == -1)
492 olim.rlim_cur = 0x7fffffff;
493 olim.rlim_max = p->p_rlimit[uap->which].rlim_max;
494 if (olim.rlim_max == -1)
495 olim.rlim_max = 0x7fffffff;
496 error = copyout(&olim, uap->rlp, sizeof(olim));
497 mtx_unlock(&Giant);
498 return (error);
499}
500#endif /* COMPAT_43 || COMPAT_SUNOS */
501
502#ifndef _SYS_SYSPROTO_H_
503struct __setrlimit_args {
504 u_int which;
505 struct rlimit *rlp;
506};
507#endif
508/*
509 * MPSAFE
510 */
511/* ARGSUSED */
512int
513setrlimit(td, uap)
514 struct thread *td;
515 register struct __setrlimit_args *uap;
516{
517 struct rlimit alim;
518 int error;
519
520 if ((error = copyin(uap->rlp, &alim, sizeof (struct rlimit))))
521 return (error);
522 mtx_lock(&Giant);
523 error = dosetrlimit(td, uap->which, &alim);
524 mtx_unlock(&Giant);
525 return (error);
526}
527
528int
529dosetrlimit(td, which, limp)
530 struct thread *td;
531 u_int which;
532 struct rlimit *limp;
533{
534 struct proc *p = td->td_proc;
535 register struct rlimit *alimp;
536 int error;
537
538 GIANT_REQUIRED;
539
540 if (which >= RLIM_NLIMITS)
541 return (EINVAL);
542 alimp = &p->p_rlimit[which];
543
544 /*
545 * Preserve historical bugs by treating negative limits as unsigned.
546 */
547 if (limp->rlim_cur < 0)
548 limp->rlim_cur = RLIM_INFINITY;
549 if (limp->rlim_max < 0)
550 limp->rlim_max = RLIM_INFINITY;
551
552 if (limp->rlim_cur > alimp->rlim_max ||
553 limp->rlim_max > alimp->rlim_max)
554 if ((error = suser_cred(td->td_ucred, PRISON_ROOT)))
555 return (error);
556 if (limp->rlim_cur > limp->rlim_max)
557 limp->rlim_cur = limp->rlim_max;
558 if (p->p_limit->p_refcnt > 1 &&
559 (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
560 p->p_limit->p_refcnt--;
561 p->p_limit = limcopy(p->p_limit);
562 alimp = &p->p_rlimit[which];
563 }
564
565 switch (which) {
566
567 case RLIMIT_CPU:
568 if (limp->rlim_cur > RLIM_INFINITY / (rlim_t)1000000)
569 p->p_limit->p_cpulimit = RLIM_INFINITY;
570 else
571 p->p_limit->p_cpulimit =
572 (rlim_t)1000000 * limp->rlim_cur;
573 break;
574 case RLIMIT_DATA:
575 if (limp->rlim_cur > maxdsiz)
576 limp->rlim_cur = maxdsiz;
577 if (limp->rlim_max > maxdsiz)
578 limp->rlim_max = maxdsiz;
579 break;
580
581 case RLIMIT_STACK:
582 if (limp->rlim_cur > maxssiz)
583 limp->rlim_cur = maxssiz;
584 if (limp->rlim_max > maxssiz)
585 limp->rlim_max = maxssiz;
586 /*
587 * Stack is allocated to the max at exec time with only
588 * "rlim_cur" bytes accessible. If stack limit is going
589 * up make more accessible, if going down make inaccessible.
590 */
591 if (limp->rlim_cur != alimp->rlim_cur) {
592 vm_offset_t addr;
593 vm_size_t size;
594 vm_prot_t prot;
595
596 if (limp->rlim_cur > alimp->rlim_cur) {
|
596 prot = VM_PROT_ALL;
| 597 prot = p->p_sysent->sv_stackprot;
|
597 size = limp->rlim_cur - alimp->rlim_cur;
| 598 size = limp->rlim_cur - alimp->rlim_cur;
|
598 addr = USRSTACK - limp->rlim_cur;
| 599 addr = p->p_sysent->sv_usrstack -
600 limp->rlim_cur;
|
599 } else {
600 prot = VM_PROT_NONE;
601 size = alimp->rlim_cur - limp->rlim_cur;
| 601 } else {
602 prot = VM_PROT_NONE;
603 size = alimp->rlim_cur - limp->rlim_cur;
|
602 addr = USRSTACK - alimp->rlim_cur;
| 604 addr = p->p_sysent->sv_usrstack -
605 alimp->rlim_cur;
|
603 }
604 addr = trunc_page(addr);
605 size = round_page(size);
606 (void) vm_map_protect(&p->p_vmspace->vm_map,
607 addr, addr+size, prot, FALSE);
608 }
609 break;
610
611 case RLIMIT_NOFILE:
612 if (limp->rlim_cur > maxfilesperproc)
613 limp->rlim_cur = maxfilesperproc;
614 if (limp->rlim_max > maxfilesperproc)
615 limp->rlim_max = maxfilesperproc;
616 break;
617
618 case RLIMIT_NPROC:
619 if (limp->rlim_cur > maxprocperuid)
620 limp->rlim_cur = maxprocperuid;
621 if (limp->rlim_max > maxprocperuid)
622 limp->rlim_max = maxprocperuid;
623 if (limp->rlim_cur < 1)
624 limp->rlim_cur = 1;
625 if (limp->rlim_max < 1)
626 limp->rlim_max = 1;
627 break;
628 }
629 *alimp = *limp;
630 return (0);
631}
632
633#ifndef _SYS_SYSPROTO_H_
634struct __getrlimit_args {
635 u_int which;
636 struct rlimit *rlp;
637};
638#endif
639/*
640 * MPSAFE
641 */
642/* ARGSUSED */
643int
644getrlimit(td, uap)
645 struct thread *td;
646 register struct __getrlimit_args *uap;
647{
648 int error;
649 struct proc *p = td->td_proc;
650
651 if (uap->which >= RLIM_NLIMITS)
652 return (EINVAL);
653 mtx_lock(&Giant);
654 error = copyout(&p->p_rlimit[uap->which], uap->rlp,
655 sizeof (struct rlimit));
656 mtx_unlock(&Giant);
657 return(error);
658}
659
660/*
661 * Transform the running time and tick information in proc p into user,
662 * system, and interrupt time usage.
663 */
664void
665calcru(p, up, sp, ip)
666 struct proc *p;
667 struct timeval *up;
668 struct timeval *sp;
669 struct timeval *ip;
670{
671 /* {user, system, interrupt, total} {ticks, usec}; previous tu: */
672 u_int64_t ut, uu, st, su, it, iu, tt, tu, ptu;
673 u_int64_t uut = 0, sut = 0, iut = 0;
674 int s;
675 struct timeval tv;
676 struct bintime bt;
677 struct kse *ke;
678 struct ksegrp *kg;
679
680 mtx_assert(&sched_lock, MA_OWNED);
681 /* XXX: why spl-protect ? worst case is an off-by-one report */
682
683 FOREACH_KSEGRP_IN_PROC(p, kg) {
684 /* we could accumulate per ksegrp and per process here*/
685 FOREACH_KSE_IN_GROUP(kg, ke) {
686 s = splstatclock();
687 ut = ke->ke_uticks;
688 st = ke->ke_sticks;
689 it = ke->ke_iticks;
690 splx(s);
691
692 tt = ut + st + it;
693 if (tt == 0) {
694 st = 1;
695 tt = 1;
696 }
697
698 if (ke == curthread->td_kse) {
699 /*
700 * Adjust for the current time slice. This is actually fairly
701 * important since the error here is on the order of a time
702 * quantum, which is much greater than the sampling error.
703 * XXXKSE use a different test due to threads on other
704 * processors also being 'current'.
705 */
706
707 binuptime(&bt);
708 bintime_sub(&bt, PCPU_PTR(switchtime));
709 bintime_add(&bt, &p->p_runtime);
710 } else {
711 bt = p->p_runtime;
712 }
713 bintime2timeval(&bt, &tv);
714 tu = (u_int64_t)tv.tv_sec * 1000000 + tv.tv_usec;
715 ptu = ke->ke_uu + ke->ke_su + ke->ke_iu;
716 if (tu < ptu || (int64_t)tu < 0) {
717 /* XXX no %qd in kernel. Truncate. */
718 printf("calcru: negative time of %ld usec for pid %d (%s)\n",
719 (long)tu, p->p_pid, p->p_comm);
720 tu = ptu;
721 }
722
723 /* Subdivide tu. */
724 uu = (tu * ut) / tt;
725 su = (tu * st) / tt;
726 iu = tu - uu - su;
727
728 /* Enforce monotonicity. */
729 if (uu < ke->ke_uu || su < ke->ke_su || iu < ke->ke_iu) {
730 if (uu < ke->ke_uu)
731 uu = ke->ke_uu;
732 else if (uu + ke->ke_su + ke->ke_iu > tu)
733 uu = tu - ke->ke_su - ke->ke_iu;
734 if (st == 0)
735 su = ke->ke_su;
736 else {
737 su = ((tu - uu) * st) / (st + it);
738 if (su < ke->ke_su)
739 su = ke->ke_su;
740 else if (uu + su + ke->ke_iu > tu)
741 su = tu - uu - ke->ke_iu;
742 }
743 KASSERT(uu + su + ke->ke_iu <= tu,
744 ("calcru: monotonisation botch 1"));
745 iu = tu - uu - su;
746 KASSERT(iu >= ke->ke_iu,
747 ("calcru: monotonisation botch 2"));
748 }
749 ke->ke_uu = uu;
750 ke->ke_su = su;
751 ke->ke_iu = iu;
752 uut += uu;
753 sut += su;
754 iut += iu;
755
756 } /* end kse loop */
757 } /* end kseg loop */
758 up->tv_sec = uut / 1000000;
759 up->tv_usec = uut % 1000000;
760 sp->tv_sec = sut / 1000000;
761 sp->tv_usec = sut % 1000000;
762 if (ip != NULL) {
763 ip->tv_sec = iut / 1000000;
764 ip->tv_usec = iut % 1000000;
765 }
766}
767
768#ifndef _SYS_SYSPROTO_H_
769struct getrusage_args {
770 int who;
771 struct rusage *rusage;
772};
773#endif
774/*
775 * MPSAFE
776 */
777/* ARGSUSED */
778int
779getrusage(td, uap)
780 register struct thread *td;
781 register struct getrusage_args *uap;
782{
783 struct proc *p = td->td_proc;
784 register struct rusage *rup;
785 int error = 0;
786
787 mtx_lock(&Giant);
788
789 switch (uap->who) {
790 case RUSAGE_SELF:
791 rup = &p->p_stats->p_ru;
792 mtx_lock_spin(&sched_lock);
793 calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
794 mtx_unlock_spin(&sched_lock);
795 break;
796
797 case RUSAGE_CHILDREN:
798 rup = &p->p_stats->p_cru;
799 break;
800
801 default:
802 rup = NULL;
803 error = EINVAL;
804 break;
805 }
806 mtx_unlock(&Giant);
807 if (error == 0) {
808 error = copyout(rup, uap->rusage, sizeof (struct rusage));
809 }
810 return(error);
811}
812
813void
814ruadd(ru, ru2)
815 register struct rusage *ru, *ru2;
816{
817 register long *ip, *ip2;
818 register int i;
819
820 timevaladd(&ru->ru_utime, &ru2->ru_utime);
821 timevaladd(&ru->ru_stime, &ru2->ru_stime);
822 if (ru->ru_maxrss < ru2->ru_maxrss)
823 ru->ru_maxrss = ru2->ru_maxrss;
824 ip = &ru->ru_first; ip2 = &ru2->ru_first;
825 for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
826 *ip++ += *ip2++;
827}
828
829/*
830 * Make a copy of the plimit structure.
831 * We share these structures copy-on-write after fork,
832 * and copy when a limit is changed.
833 */
834struct plimit *
835limcopy(lim)
836 struct plimit *lim;
837{
838 register struct plimit *copy;
839
840 MALLOC(copy, struct plimit *, sizeof(struct plimit),
841 M_SUBPROC, M_WAITOK);
842 bcopy(lim->pl_rlimit, copy->pl_rlimit, sizeof(struct plimit));
843 copy->p_lflags = 0;
844 copy->p_refcnt = 1;
845 return (copy);
846}
847
848/*
849 * Find the uidinfo structure for a uid. This structure is used to
850 * track the total resource consumption (process count, socket buffer
851 * size, etc.) for the uid and impose limits.
852 */
853void
854uihashinit()
855{
856
857 uihashtbl = hashinit(maxproc / 16, M_UIDINFO, &uihash);
858 mtx_init(&uihashtbl_mtx, "uidinfo hash", NULL, MTX_DEF);
859}
860
861/*
862 * lookup a uidinfo struct for the parameter uid.
863 * uihashtbl_mtx must be locked.
864 */
865static struct uidinfo *
866uilookup(uid)
867 uid_t uid;
868{
869 struct uihashhead *uipp;
870 struct uidinfo *uip;
871
872 mtx_assert(&uihashtbl_mtx, MA_OWNED);
873 uipp = UIHASH(uid);
874 LIST_FOREACH(uip, uipp, ui_hash)
875 if (uip->ui_uid == uid)
876 break;
877
878 return (uip);
879}
880
881/*
882 * Find or allocate a struct uidinfo for a particular uid.
883 * Increase refcount on uidinfo struct returned.
884 * uifree() should be called on a struct uidinfo when released.
885 */
886struct uidinfo *
887uifind(uid)
888 uid_t uid;
889{
890 struct uidinfo *uip;
891
892 mtx_lock(&uihashtbl_mtx);
893 uip = uilookup(uid);
894 if (uip == NULL) {
895 struct uidinfo *old_uip;
896
897 mtx_unlock(&uihashtbl_mtx);
898 uip = malloc(sizeof(*uip), M_UIDINFO, M_WAITOK | M_ZERO);
899 mtx_lock(&uihashtbl_mtx);
900 /*
901 * There's a chance someone created our uidinfo while we
902 * were in malloc and not holding the lock, so we have to
903 * make sure we don't insert a duplicate uidinfo
904 */
905 if ((old_uip = uilookup(uid)) != NULL) {
906 /* someone else beat us to it */
907 free(uip, M_UIDINFO);
908 uip = old_uip;
909 } else {
910 uip->ui_mtxp = mtx_pool_alloc();
911 uip->ui_uid = uid;
912 LIST_INSERT_HEAD(UIHASH(uid), uip, ui_hash);
913 }
914 }
915 uihold(uip);
916 mtx_unlock(&uihashtbl_mtx);
917 return (uip);
918}
919
920/*
921 * Place another refcount on a uidinfo struct.
922 */
923void
924uihold(uip)
925 struct uidinfo *uip;
926{
927
928 UIDINFO_LOCK(uip);
929 uip->ui_ref++;
930 UIDINFO_UNLOCK(uip);
931}
932
933/*-
934 * Since uidinfo structs have a long lifetime, we use an
935 * opportunistic refcounting scheme to avoid locking the lookup hash
936 * for each release.
937 *
938 * If the refcount hits 0, we need to free the structure,
939 * which means we need to lock the hash.
940 * Optimal case:
941 * After locking the struct and lowering the refcount, if we find
942 * that we don't need to free, simply unlock and return.
943 * Suboptimal case:
944 * If refcount lowering results in need to free, bump the count
945 * back up, loose the lock and aquire the locks in the proper
946 * order to try again.
947 */
948void
949uifree(uip)
950 struct uidinfo *uip;
951{
952
953 /* Prepare for optimal case. */
954 UIDINFO_LOCK(uip);
955
956 if (--uip->ui_ref != 0) {
957 UIDINFO_UNLOCK(uip);
958 return;
959 }
960
961 /* Prepare for suboptimal case. */
962 uip->ui_ref++;
963 UIDINFO_UNLOCK(uip);
964 mtx_lock(&uihashtbl_mtx);
965 UIDINFO_LOCK(uip);
966
967 /*
968 * We must subtract one from the count again because we backed out
969 * our initial subtraction before dropping the lock.
970 * Since another thread may have added a reference after we dropped the
971 * initial lock we have to test for zero again.
972 */
973 if (--uip->ui_ref == 0) {
974 LIST_REMOVE(uip, ui_hash);
975 mtx_unlock(&uihashtbl_mtx);
976 if (uip->ui_sbsize != 0)
977 /* XXX no %qd in kernel. Truncate. */
978 printf("freeing uidinfo: uid = %d, sbsize = %ld\n",
979 uip->ui_uid, (long)uip->ui_sbsize);
980 if (uip->ui_proccnt != 0)
981 printf("freeing uidinfo: uid = %d, proccnt = %ld\n",
982 uip->ui_uid, uip->ui_proccnt);
983 UIDINFO_UNLOCK(uip);
984 FREE(uip, M_UIDINFO);
985 return;
986 }
987
988 mtx_unlock(&uihashtbl_mtx);
989 UIDINFO_UNLOCK(uip);
990}
991
992/*
993 * Change the count associated with number of processes
994 * a given user is using. When 'max' is 0, don't enforce a limit
995 */
996int
997chgproccnt(uip, diff, max)
998 struct uidinfo *uip;
999 int diff;
1000 int max;
1001{
1002
1003 UIDINFO_LOCK(uip);
1004 /* don't allow them to exceed max, but allow subtraction */
1005 if (diff > 0 && uip->ui_proccnt + diff > max && max != 0) {
1006 UIDINFO_UNLOCK(uip);
1007 return (0);
1008 }
1009 uip->ui_proccnt += diff;
1010 if (uip->ui_proccnt < 0)
1011 printf("negative proccnt for uid = %d\n", uip->ui_uid);
1012 UIDINFO_UNLOCK(uip);
1013 return (1);
1014}
1015
1016/*
1017 * Change the total socket buffer size a user has used.
1018 */
1019int
1020chgsbsize(uip, hiwat, to, max)
1021 struct uidinfo *uip;
1022 u_int *hiwat;
1023 u_int to;
1024 rlim_t max;
1025{
1026 rlim_t new;
1027 int s;
1028
1029 s = splnet();
1030 UIDINFO_LOCK(uip);
1031 new = uip->ui_sbsize + to - *hiwat;
1032 /* don't allow them to exceed max, but allow subtraction */
1033 if (to > *hiwat && new > max) {
1034 splx(s);
1035 UIDINFO_UNLOCK(uip);
1036 return (0);
1037 }
1038 uip->ui_sbsize = new;
1039 *hiwat = to;
1040 if (uip->ui_sbsize < 0)
1041 printf("negative sbsize for uid = %d\n", uip->ui_uid);
1042 splx(s);
1043 UIDINFO_UNLOCK(uip);
1044 return (1);
1045}
| 606 }
607 addr = trunc_page(addr);
608 size = round_page(size);
609 (void) vm_map_protect(&p->p_vmspace->vm_map,
610 addr, addr+size, prot, FALSE);
611 }
612 break;
613
614 case RLIMIT_NOFILE:
615 if (limp->rlim_cur > maxfilesperproc)
616 limp->rlim_cur = maxfilesperproc;
617 if (limp->rlim_max > maxfilesperproc)
618 limp->rlim_max = maxfilesperproc;
619 break;
620
621 case RLIMIT_NPROC:
622 if (limp->rlim_cur > maxprocperuid)
623 limp->rlim_cur = maxprocperuid;
624 if (limp->rlim_max > maxprocperuid)
625 limp->rlim_max = maxprocperuid;
626 if (limp->rlim_cur < 1)
627 limp->rlim_cur = 1;
628 if (limp->rlim_max < 1)
629 limp->rlim_max = 1;
630 break;
631 }
632 *alimp = *limp;
633 return (0);
634}
635
636#ifndef _SYS_SYSPROTO_H_
637struct __getrlimit_args {
638 u_int which;
639 struct rlimit *rlp;
640};
641#endif
642/*
643 * MPSAFE
644 */
645/* ARGSUSED */
646int
647getrlimit(td, uap)
648 struct thread *td;
649 register struct __getrlimit_args *uap;
650{
651 int error;
652 struct proc *p = td->td_proc;
653
654 if (uap->which >= RLIM_NLIMITS)
655 return (EINVAL);
656 mtx_lock(&Giant);
657 error = copyout(&p->p_rlimit[uap->which], uap->rlp,
658 sizeof (struct rlimit));
659 mtx_unlock(&Giant);
660 return(error);
661}
662
663/*
664 * Transform the running time and tick information in proc p into user,
665 * system, and interrupt time usage.
666 */
667void
668calcru(p, up, sp, ip)
669 struct proc *p;
670 struct timeval *up;
671 struct timeval *sp;
672 struct timeval *ip;
673{
674 /* {user, system, interrupt, total} {ticks, usec}; previous tu: */
675 u_int64_t ut, uu, st, su, it, iu, tt, tu, ptu;
676 u_int64_t uut = 0, sut = 0, iut = 0;
677 int s;
678 struct timeval tv;
679 struct bintime bt;
680 struct kse *ke;
681 struct ksegrp *kg;
682
683 mtx_assert(&sched_lock, MA_OWNED);
684 /* XXX: why spl-protect ? worst case is an off-by-one report */
685
686 FOREACH_KSEGRP_IN_PROC(p, kg) {
687 /* we could accumulate per ksegrp and per process here*/
688 FOREACH_KSE_IN_GROUP(kg, ke) {
689 s = splstatclock();
690 ut = ke->ke_uticks;
691 st = ke->ke_sticks;
692 it = ke->ke_iticks;
693 splx(s);
694
695 tt = ut + st + it;
696 if (tt == 0) {
697 st = 1;
698 tt = 1;
699 }
700
701 if (ke == curthread->td_kse) {
702 /*
703 * Adjust for the current time slice. This is actually fairly
704 * important since the error here is on the order of a time
705 * quantum, which is much greater than the sampling error.
706 * XXXKSE use a different test due to threads on other
707 * processors also being 'current'.
708 */
709
710 binuptime(&bt);
711 bintime_sub(&bt, PCPU_PTR(switchtime));
712 bintime_add(&bt, &p->p_runtime);
713 } else {
714 bt = p->p_runtime;
715 }
716 bintime2timeval(&bt, &tv);
717 tu = (u_int64_t)tv.tv_sec * 1000000 + tv.tv_usec;
718 ptu = ke->ke_uu + ke->ke_su + ke->ke_iu;
719 if (tu < ptu || (int64_t)tu < 0) {
720 /* XXX no %qd in kernel. Truncate. */
721 printf("calcru: negative time of %ld usec for pid %d (%s)\n",
722 (long)tu, p->p_pid, p->p_comm);
723 tu = ptu;
724 }
725
726 /* Subdivide tu. */
727 uu = (tu * ut) / tt;
728 su = (tu * st) / tt;
729 iu = tu - uu - su;
730
731 /* Enforce monotonicity. */
732 if (uu < ke->ke_uu || su < ke->ke_su || iu < ke->ke_iu) {
733 if (uu < ke->ke_uu)
734 uu = ke->ke_uu;
735 else if (uu + ke->ke_su + ke->ke_iu > tu)
736 uu = tu - ke->ke_su - ke->ke_iu;
737 if (st == 0)
738 su = ke->ke_su;
739 else {
740 su = ((tu - uu) * st) / (st + it);
741 if (su < ke->ke_su)
742 su = ke->ke_su;
743 else if (uu + su + ke->ke_iu > tu)
744 su = tu - uu - ke->ke_iu;
745 }
746 KASSERT(uu + su + ke->ke_iu <= tu,
747 ("calcru: monotonisation botch 1"));
748 iu = tu - uu - su;
749 KASSERT(iu >= ke->ke_iu,
750 ("calcru: monotonisation botch 2"));
751 }
752 ke->ke_uu = uu;
753 ke->ke_su = su;
754 ke->ke_iu = iu;
755 uut += uu;
756 sut += su;
757 iut += iu;
758
759 } /* end kse loop */
760 } /* end kseg loop */
761 up->tv_sec = uut / 1000000;
762 up->tv_usec = uut % 1000000;
763 sp->tv_sec = sut / 1000000;
764 sp->tv_usec = sut % 1000000;
765 if (ip != NULL) {
766 ip->tv_sec = iut / 1000000;
767 ip->tv_usec = iut % 1000000;
768 }
769}
770
771#ifndef _SYS_SYSPROTO_H_
772struct getrusage_args {
773 int who;
774 struct rusage *rusage;
775};
776#endif
777/*
778 * MPSAFE
779 */
780/* ARGSUSED */
781int
782getrusage(td, uap)
783 register struct thread *td;
784 register struct getrusage_args *uap;
785{
786 struct proc *p = td->td_proc;
787 register struct rusage *rup;
788 int error = 0;
789
790 mtx_lock(&Giant);
791
792 switch (uap->who) {
793 case RUSAGE_SELF:
794 rup = &p->p_stats->p_ru;
795 mtx_lock_spin(&sched_lock);
796 calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
797 mtx_unlock_spin(&sched_lock);
798 break;
799
800 case RUSAGE_CHILDREN:
801 rup = &p->p_stats->p_cru;
802 break;
803
804 default:
805 rup = NULL;
806 error = EINVAL;
807 break;
808 }
809 mtx_unlock(&Giant);
810 if (error == 0) {
811 error = copyout(rup, uap->rusage, sizeof (struct rusage));
812 }
813 return(error);
814}
815
816void
817ruadd(ru, ru2)
818 register struct rusage *ru, *ru2;
819{
820 register long *ip, *ip2;
821 register int i;
822
823 timevaladd(&ru->ru_utime, &ru2->ru_utime);
824 timevaladd(&ru->ru_stime, &ru2->ru_stime);
825 if (ru->ru_maxrss < ru2->ru_maxrss)
826 ru->ru_maxrss = ru2->ru_maxrss;
827 ip = &ru->ru_first; ip2 = &ru2->ru_first;
828 for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
829 *ip++ += *ip2++;
830}
831
832/*
833 * Make a copy of the plimit structure.
834 * We share these structures copy-on-write after fork,
835 * and copy when a limit is changed.
836 */
837struct plimit *
838limcopy(lim)
839 struct plimit *lim;
840{
841 register struct plimit *copy;
842
843 MALLOC(copy, struct plimit *, sizeof(struct plimit),
844 M_SUBPROC, M_WAITOK);
845 bcopy(lim->pl_rlimit, copy->pl_rlimit, sizeof(struct plimit));
846 copy->p_lflags = 0;
847 copy->p_refcnt = 1;
848 return (copy);
849}
850
851/*
852 * Find the uidinfo structure for a uid. This structure is used to
853 * track the total resource consumption (process count, socket buffer
854 * size, etc.) for the uid and impose limits.
855 */
856void
857uihashinit()
858{
859
860 uihashtbl = hashinit(maxproc / 16, M_UIDINFO, &uihash);
861 mtx_init(&uihashtbl_mtx, "uidinfo hash", NULL, MTX_DEF);
862}
863
864/*
865 * lookup a uidinfo struct for the parameter uid.
866 * uihashtbl_mtx must be locked.
867 */
868static struct uidinfo *
869uilookup(uid)
870 uid_t uid;
871{
872 struct uihashhead *uipp;
873 struct uidinfo *uip;
874
875 mtx_assert(&uihashtbl_mtx, MA_OWNED);
876 uipp = UIHASH(uid);
877 LIST_FOREACH(uip, uipp, ui_hash)
878 if (uip->ui_uid == uid)
879 break;
880
881 return (uip);
882}
883
884/*
885 * Find or allocate a struct uidinfo for a particular uid.
886 * Increase refcount on uidinfo struct returned.
887 * uifree() should be called on a struct uidinfo when released.
888 */
889struct uidinfo *
890uifind(uid)
891 uid_t uid;
892{
893 struct uidinfo *uip;
894
895 mtx_lock(&uihashtbl_mtx);
896 uip = uilookup(uid);
897 if (uip == NULL) {
898 struct uidinfo *old_uip;
899
900 mtx_unlock(&uihashtbl_mtx);
901 uip = malloc(sizeof(*uip), M_UIDINFO, M_WAITOK | M_ZERO);
902 mtx_lock(&uihashtbl_mtx);
903 /*
904 * There's a chance someone created our uidinfo while we
905 * were in malloc and not holding the lock, so we have to
906 * make sure we don't insert a duplicate uidinfo
907 */
908 if ((old_uip = uilookup(uid)) != NULL) {
909 /* someone else beat us to it */
910 free(uip, M_UIDINFO);
911 uip = old_uip;
912 } else {
913 uip->ui_mtxp = mtx_pool_alloc();
914 uip->ui_uid = uid;
915 LIST_INSERT_HEAD(UIHASH(uid), uip, ui_hash);
916 }
917 }
918 uihold(uip);
919 mtx_unlock(&uihashtbl_mtx);
920 return (uip);
921}
922
923/*
924 * Place another refcount on a uidinfo struct.
925 */
926void
927uihold(uip)
928 struct uidinfo *uip;
929{
930
931 UIDINFO_LOCK(uip);
932 uip->ui_ref++;
933 UIDINFO_UNLOCK(uip);
934}
935
936/*-
937 * Since uidinfo structs have a long lifetime, we use an
938 * opportunistic refcounting scheme to avoid locking the lookup hash
939 * for each release.
940 *
941 * If the refcount hits 0, we need to free the structure,
942 * which means we need to lock the hash.
943 * Optimal case:
944 * After locking the struct and lowering the refcount, if we find
945 * that we don't need to free, simply unlock and return.
946 * Suboptimal case:
947 * If refcount lowering results in need to free, bump the count
948 * back up, loose the lock and aquire the locks in the proper
949 * order to try again.
950 */
951void
952uifree(uip)
953 struct uidinfo *uip;
954{
955
956 /* Prepare for optimal case. */
957 UIDINFO_LOCK(uip);
958
959 if (--uip->ui_ref != 0) {
960 UIDINFO_UNLOCK(uip);
961 return;
962 }
963
964 /* Prepare for suboptimal case. */
965 uip->ui_ref++;
966 UIDINFO_UNLOCK(uip);
967 mtx_lock(&uihashtbl_mtx);
968 UIDINFO_LOCK(uip);
969
970 /*
971 * We must subtract one from the count again because we backed out
972 * our initial subtraction before dropping the lock.
973 * Since another thread may have added a reference after we dropped the
974 * initial lock we have to test for zero again.
975 */
976 if (--uip->ui_ref == 0) {
977 LIST_REMOVE(uip, ui_hash);
978 mtx_unlock(&uihashtbl_mtx);
979 if (uip->ui_sbsize != 0)
980 /* XXX no %qd in kernel. Truncate. */
981 printf("freeing uidinfo: uid = %d, sbsize = %ld\n",
982 uip->ui_uid, (long)uip->ui_sbsize);
983 if (uip->ui_proccnt != 0)
984 printf("freeing uidinfo: uid = %d, proccnt = %ld\n",
985 uip->ui_uid, uip->ui_proccnt);
986 UIDINFO_UNLOCK(uip);
987 FREE(uip, M_UIDINFO);
988 return;
989 }
990
991 mtx_unlock(&uihashtbl_mtx);
992 UIDINFO_UNLOCK(uip);
993}
994
995/*
996 * Change the count associated with number of processes
997 * a given user is using. When 'max' is 0, don't enforce a limit
998 */
999int
1000chgproccnt(uip, diff, max)
1001 struct uidinfo *uip;
1002 int diff;
1003 int max;
1004{
1005
1006 UIDINFO_LOCK(uip);
1007 /* don't allow them to exceed max, but allow subtraction */
1008 if (diff > 0 && uip->ui_proccnt + diff > max && max != 0) {
1009 UIDINFO_UNLOCK(uip);
1010 return (0);
1011 }
1012 uip->ui_proccnt += diff;
1013 if (uip->ui_proccnt < 0)
1014 printf("negative proccnt for uid = %d\n", uip->ui_uid);
1015 UIDINFO_UNLOCK(uip);
1016 return (1);
1017}
1018
1019/*
1020 * Change the total socket buffer size a user has used.
1021 */
1022int
1023chgsbsize(uip, hiwat, to, max)
1024 struct uidinfo *uip;
1025 u_int *hiwat;
1026 u_int to;
1027 rlim_t max;
1028{
1029 rlim_t new;
1030 int s;
1031
1032 s = splnet();
1033 UIDINFO_LOCK(uip);
1034 new = uip->ui_sbsize + to - *hiwat;
1035 /* don't allow them to exceed max, but allow subtraction */
1036 if (to > *hiwat && new > max) {
1037 splx(s);
1038 UIDINFO_UNLOCK(uip);
1039 return (0);
1040 }
1041 uip->ui_sbsize = new;
1042 *hiwat = to;
1043 if (uip->ui_sbsize < 0)
1044 printf("negative sbsize for uid = %d\n", uip->ui_uid);
1045 splx(s);
1046 UIDINFO_UNLOCK(uip);
1047 return (1);
1048}
|