kern_resource.c revision 1.115 1 /* $NetBSD: kern_resource.c,v 1.115 2007/03/04 06:03:06 christos Exp $ */
2
3 /*-
4 * Copyright (c) 1982, 1986, 1991, 1993
5 * The Regents of the University of California. All rights reserved.
6 * (c) UNIX System Laboratories, Inc.
7 * All or some portions of this file are derived from material licensed
8 * to the University of California by American Telephone and Telegraph
9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10 * the permission of UNIX System Laboratories, Inc.
11 *
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
14 * are met:
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 *
36 * @(#)kern_resource.c 8.8 (Berkeley) 2/14/95
37 */
38
39 #include <sys/cdefs.h>
40 __KERNEL_RCSID(0, "$NetBSD: kern_resource.c,v 1.115 2007/03/04 06:03:06 christos Exp $");
41
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/kernel.h>
45 #include <sys/file.h>
46 #include <sys/resourcevar.h>
47 #include <sys/malloc.h>
48 #include <sys/namei.h>
49 #include <sys/pool.h>
50 #include <sys/proc.h>
51 #include <sys/sysctl.h>
52 #include <sys/kauth.h>
53
54 #include <sys/mount.h>
55 #include <sys/syscallargs.h>
56
57 #include <uvm/uvm_extern.h>
58
59 /*
60 * Maximum process data and stack limits.
61 * They are variables so they are patchable.
62 */
63 rlim_t maxdmap = MAXDSIZ;
64 rlim_t maxsmap = MAXSSIZ;
65
66 struct uihashhead *uihashtbl;
67 u_long uihash; /* size of hash table - 1 */
68 struct simplelock uihashtbl_slock = SIMPLELOCK_INITIALIZER;
69
70
71 /*
72 * Resource controls and accounting.
73 */
74
75 int
76 sys_getpriority(struct lwp *l, void *v, register_t *retval)
77 {
78 struct sys_getpriority_args /* {
79 syscallarg(int) which;
80 syscallarg(id_t) who;
81 } */ *uap = v;
82 struct proc *curp = l->l_proc, *p;
83 int low = NZERO + PRIO_MAX + 1;
84 int who = SCARG(uap, who);
85
86 rw_enter(&proclist_lock, RW_READER);
87 switch (SCARG(uap, which)) {
88 case PRIO_PROCESS:
89 if (who == 0)
90 p = curp;
91 else
92 p = p_find(who, PFIND_LOCKED);
93 if (p != NULL)
94 low = p->p_nice;
95 break;
96
97 case PRIO_PGRP: {
98 struct pgrp *pg;
99
100 if (who == 0)
101 pg = curp->p_pgrp;
102 else if ((pg = pg_find(who, PFIND_LOCKED)) == NULL)
103 break;
104 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
105 if (p->p_nice < low)
106 low = p->p_nice;
107 }
108 break;
109 }
110
111 case PRIO_USER:
112 if (who == 0)
113 who = (int)kauth_cred_geteuid(l->l_cred);
114 PROCLIST_FOREACH(p, &allproc) {
115 mutex_enter(&p->p_mutex);
116 if (kauth_cred_geteuid(p->p_cred) ==
117 (uid_t)who && p->p_nice < low)
118 low = p->p_nice;
119 mutex_exit(&p->p_mutex);
120 }
121 break;
122
123 default:
124 rw_exit(&proclist_lock);
125 return (EINVAL);
126 }
127 rw_exit(&proclist_lock);
128
129 if (low == NZERO + PRIO_MAX + 1)
130 return (ESRCH);
131 *retval = low - NZERO;
132 return (0);
133 }
134
135 /* ARGSUSED */
136 int
137 sys_setpriority(struct lwp *l, void *v, register_t *retval)
138 {
139 struct sys_setpriority_args /* {
140 syscallarg(int) which;
141 syscallarg(id_t) who;
142 syscallarg(int) prio;
143 } */ *uap = v;
144 struct proc *curp = l->l_proc, *p;
145 int found = 0, error = 0;
146 int who = SCARG(uap, who);
147
148 rw_enter(&proclist_lock, RW_READER);
149 switch (SCARG(uap, which)) {
150 case PRIO_PROCESS:
151 if (who == 0)
152 p = curp;
153 else
154 p = p_find(who, PFIND_LOCKED);
155 if (p != 0) {
156 mutex_enter(&p->p_mutex);
157 error = donice(l, p, SCARG(uap, prio));
158 mutex_exit(&p->p_mutex);
159 }
160 found++;
161 break;
162
163 case PRIO_PGRP: {
164 struct pgrp *pg;
165
166 if (who == 0)
167 pg = curp->p_pgrp;
168 else if ((pg = pg_find(who, PFIND_LOCKED)) == NULL)
169 break;
170 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
171 mutex_enter(&p->p_mutex);
172 error = donice(l, p, SCARG(uap, prio));
173 mutex_exit(&p->p_mutex);
174 found++;
175 }
176 break;
177 }
178
179 case PRIO_USER:
180 if (who == 0)
181 who = (int)kauth_cred_geteuid(l->l_cred);
182 PROCLIST_FOREACH(p, &allproc) {
183 mutex_enter(&p->p_mutex);
184 if (kauth_cred_geteuid(p->p_cred) ==
185 (uid_t)SCARG(uap, who)) {
186 error = donice(l, p, SCARG(uap, prio));
187 found++;
188 }
189 mutex_exit(&p->p_mutex);
190 }
191 break;
192
193 default:
194 error = EINVAL;
195 break;
196 }
197 rw_exit(&proclist_lock);
198 if (found == 0)
199 return (ESRCH);
200 return (error);
201 }
202
203 /*
204 * Renice a process.
205 *
206 * Call with the target process' credentials locked.
207 */
208 int
209 donice(struct lwp *l, struct proc *chgp, int n)
210 {
211 kauth_cred_t cred = l->l_cred;
212 int onice;
213
214 LOCK_ASSERT(mutex_owned(&chgp->p_mutex));
215
216 if (n > PRIO_MAX)
217 n = PRIO_MAX;
218 if (n < PRIO_MIN)
219 n = PRIO_MIN;
220 n += NZERO;
221 onice = chgp->p_nice;
222 onice = chgp->p_nice;
223
224 again:
225 if (kauth_authorize_process(cred, KAUTH_PROCESS_NICE, chgp,
226 KAUTH_ARG(n), NULL, NULL))
227 return (EACCES);
228 mutex_spin_enter(&chgp->p_stmutex);
229 if (onice != chgp->p_nice) {
230 mutex_spin_exit(&chgp->p_stmutex);
231 goto again;
232 }
233 chgp->p_nice = n;
234 (void)resetprocpriority(chgp);
235 mutex_spin_exit(&chgp->p_stmutex);
236 return (0);
237 }
238
239 /* ARGSUSED */
240 int
241 sys_setrlimit(struct lwp *l, void *v, register_t *retval)
242 {
243 struct sys_setrlimit_args /* {
244 syscallarg(int) which;
245 syscallarg(const struct rlimit *) rlp;
246 } */ *uap = v;
247 int which = SCARG(uap, which);
248 struct rlimit alim;
249 int error;
250
251 error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit));
252 if (error)
253 return (error);
254 return (dosetrlimit(l, l->l_proc, which, &alim));
255 }
256
257 int
258 dosetrlimit(struct lwp *l, struct proc *p, int which, struct rlimit *limp)
259 {
260 struct rlimit *alimp;
261 struct plimit *oldplim;
262 int error;
263
264 if ((u_int)which >= RLIM_NLIMITS)
265 return (EINVAL);
266
267 if (limp->rlim_cur < 0 || limp->rlim_max < 0)
268 return (EINVAL);
269
270 alimp = &p->p_rlimit[which];
271 /* if we don't change the value, no need to limcopy() */
272 if (limp->rlim_cur == alimp->rlim_cur &&
273 limp->rlim_max == alimp->rlim_max)
274 return 0;
275
276 if (limp->rlim_cur > limp->rlim_max) {
277 /*
278 * This is programming error. According to SUSv2, we should
279 * return error in this case.
280 */
281 return (EINVAL);
282 }
283 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT,
284 p, limp, KAUTH_ARG(which), NULL);
285 if (error)
286 return (error);
287
288 mutex_enter(&p->p_mutex);
289 if (p->p_limit->p_refcnt > 1 &&
290 (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
291 oldplim = p->p_limit;
292 p->p_limit = limcopy(p);
293 limfree(oldplim);
294 alimp = &p->p_rlimit[which];
295 }
296
297 switch (which) {
298
299 case RLIMIT_DATA:
300 if (limp->rlim_cur > maxdmap)
301 limp->rlim_cur = maxdmap;
302 if (limp->rlim_max > maxdmap)
303 limp->rlim_max = maxdmap;
304 break;
305
306 case RLIMIT_STACK:
307 if (limp->rlim_cur > maxsmap)
308 limp->rlim_cur = maxsmap;
309 if (limp->rlim_max > maxsmap)
310 limp->rlim_max = maxsmap;
311
312 /*
313 * Return EINVAL if the new stack size limit is lower than
314 * current usage. Otherwise, the process would get SIGSEGV the
315 * moment it would try to access anything on it's current stack.
316 * This conforms to SUSv2.
317 */
318 if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE
319 || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE) {
320 mutex_exit(&p->p_mutex);
321 return (EINVAL);
322 }
323
324 /*
325 * Stack is allocated to the max at exec time with
326 * only "rlim_cur" bytes accessible (In other words,
327 * allocates stack dividing two contiguous regions at
328 * "rlim_cur" bytes boundary).
329 *
330 * Since allocation is done in terms of page, roundup
331 * "rlim_cur" (otherwise, contiguous regions
332 * overlap). If stack limit is going up make more
333 * accessible, if going down make inaccessible.
334 */
335 limp->rlim_cur = round_page(limp->rlim_cur);
336 if (limp->rlim_cur != alimp->rlim_cur) {
337 vaddr_t addr;
338 vsize_t size;
339 vm_prot_t prot;
340
341 if (limp->rlim_cur > alimp->rlim_cur) {
342 prot = VM_PROT_READ | VM_PROT_WRITE;
343 size = limp->rlim_cur - alimp->rlim_cur;
344 addr = (vaddr_t)p->p_vmspace->vm_minsaddr -
345 limp->rlim_cur;
346 } else {
347 prot = VM_PROT_NONE;
348 size = alimp->rlim_cur - limp->rlim_cur;
349 addr = (vaddr_t)p->p_vmspace->vm_minsaddr -
350 alimp->rlim_cur;
351 }
352 (void) uvm_map_protect(&p->p_vmspace->vm_map,
353 addr, addr+size, prot, false);
354 }
355 break;
356
357 case RLIMIT_NOFILE:
358 if (limp->rlim_cur > maxfiles)
359 limp->rlim_cur = maxfiles;
360 if (limp->rlim_max > maxfiles)
361 limp->rlim_max = maxfiles;
362 break;
363
364 case RLIMIT_NPROC:
365 if (limp->rlim_cur > maxproc)
366 limp->rlim_cur = maxproc;
367 if (limp->rlim_max > maxproc)
368 limp->rlim_max = maxproc;
369 break;
370 }
371 *alimp = *limp;
372 mutex_exit(&p->p_mutex);
373 return (0);
374 }
375
376 /* ARGSUSED */
377 int
378 sys_getrlimit(struct lwp *l, void *v, register_t *retval)
379 {
380 struct sys_getrlimit_args /* {
381 syscallarg(int) which;
382 syscallarg(struct rlimit *) rlp;
383 } */ *uap = v;
384 struct proc *p = l->l_proc;
385 int which = SCARG(uap, which);
386
387 if ((u_int)which >= RLIM_NLIMITS)
388 return (EINVAL);
389 return (copyout(&p->p_rlimit[which], SCARG(uap, rlp),
390 sizeof(struct rlimit)));
391 }
392
393 /*
394 * Transform the running time and tick information in proc p into user,
395 * system, and interrupt time usage.
396 *
397 * Should be called with p->p_smutex held unless called from exit1().
398 */
399 void
400 calcru(struct proc *p, struct timeval *up, struct timeval *sp,
401 struct timeval *ip, struct timeval *rp)
402 {
403 u_quad_t u, st, ut, it, tot;
404 unsigned long sec;
405 long usec;
406 struct timeval tv;
407 struct lwp *l;
408
409 mutex_spin_enter(&p->p_stmutex);
410 st = p->p_sticks;
411 ut = p->p_uticks;
412 it = p->p_iticks;
413 mutex_spin_exit(&p->p_stmutex);
414
415 sec = p->p_rtime.tv_sec;
416 usec = p->p_rtime.tv_usec;
417
418 LIST_FOREACH(l, &p->p_lwps, l_sibling) {
419 lwp_lock(l);
420 sec += l->l_rtime.tv_sec;
421 if ((usec += l->l_rtime.tv_usec) >= 1000000) {
422 sec++;
423 usec -= 1000000;
424 }
425 if (l->l_cpu == curcpu()) {
426 struct schedstate_percpu *spc;
427
428 KDASSERT(l->l_cpu != NULL);
429 spc = &l->l_cpu->ci_schedstate;
430
431 /*
432 * Adjust for the current time slice. This is
433 * actually fairly important since the error
434 * here is on the order of a time quantum,
435 * which is much greater than the sampling
436 * error.
437 */
438 microtime(&tv);
439 sec += tv.tv_sec - spc->spc_runtime.tv_sec;
440 usec += tv.tv_usec - spc->spc_runtime.tv_usec;
441 if (usec >= 1000000) {
442 sec++;
443 usec -= 1000000;
444 }
445 }
446 lwp_unlock(l);
447 }
448
449 tot = st + ut + it;
450 u = sec * 1000000ull + usec;
451
452 if (tot == 0) {
453 /* No ticks, so can't use to share time out, split 50-50 */
454 st = ut = u / 2;
455 } else {
456 st = (u * st) / tot;
457 ut = (u * ut) / tot;
458 }
459 if (sp != NULL) {
460 sp->tv_sec = st / 1000000;
461 sp->tv_usec = st % 1000000;
462 }
463 if (up != NULL) {
464 up->tv_sec = ut / 1000000;
465 up->tv_usec = ut % 1000000;
466 }
467 if (ip != NULL) {
468 if (it != 0)
469 it = (u * it) / tot;
470 ip->tv_sec = it / 1000000;
471 ip->tv_usec = it % 1000000;
472 }
473 if (rp != NULL) {
474 rp->tv_sec = sec;
475 rp->tv_usec = usec;
476 }
477 }
478
479 /* ARGSUSED */
480 int
481 sys_getrusage(struct lwp *l, void *v, register_t *retval)
482 {
483 struct sys_getrusage_args /* {
484 syscallarg(int) who;
485 syscallarg(struct rusage *) rusage;
486 } */ *uap = v;
487 struct rusage *rup;
488 struct proc *p = l->l_proc;
489
490 switch (SCARG(uap, who)) {
491
492 case RUSAGE_SELF:
493 rup = &p->p_stats->p_ru;
494 mutex_enter(&p->p_smutex);
495 calcru(p, &rup->ru_utime, &rup->ru_stime, NULL, NULL);
496 mutex_exit(&p->p_smutex);
497 break;
498
499 case RUSAGE_CHILDREN:
500 rup = &p->p_stats->p_cru;
501 break;
502
503 default:
504 return (EINVAL);
505 }
506 return (copyout(rup, SCARG(uap, rusage), sizeof(struct rusage)));
507 }
508
509 void
510 ruadd(struct rusage *ru, struct rusage *ru2)
511 {
512 long *ip, *ip2;
513 int i;
514
515 timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
516 timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
517 if (ru->ru_maxrss < ru2->ru_maxrss)
518 ru->ru_maxrss = ru2->ru_maxrss;
519 ip = &ru->ru_first; ip2 = &ru2->ru_first;
520 for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
521 *ip++ += *ip2++;
522 }
523
524 /*
525 * Make a copy of the plimit structure.
526 * We share these structures copy-on-write after fork,
527 * and copy when a limit is changed.
528 *
529 * XXXSMP This is atrocious, need to simplify.
530 */
531 struct plimit *
532 limcopy(struct proc *p)
533 {
534 struct plimit *lim, *newlim;
535 char *corename;
536 size_t l;
537
538 LOCK_ASSERT(mutex_owned(&p->p_mutex));
539
540 mutex_exit(&p->p_mutex);
541 newlim = pool_get(&plimit_pool, PR_WAITOK);
542 simple_lock_init(&newlim->p_slock);
543 newlim->p_lflags = 0;
544 newlim->p_refcnt = 1;
545 mutex_enter(&p->p_mutex);
546
547 for (;;) {
548 lim = p->p_limit;
549 simple_lock(&lim->p_slock);
550 if (lim->pl_corename != defcorename) {
551 l = strlen(lim->pl_corename) + 1;
552
553 simple_unlock(&lim->p_slock);
554 mutex_exit(&p->p_mutex);
555 corename = malloc(l, M_TEMP, M_WAITOK);
556 mutex_enter(&p->p_mutex);
557 simple_lock(&lim->p_slock);
558
559 if (l != strlen(lim->pl_corename) + 1) {
560 simple_unlock(&lim->p_slock);
561 mutex_exit(&p->p_mutex);
562 free(corename, M_TEMP);
563 mutex_enter(&p->p_mutex);
564 continue;
565 }
566 } else
567 l = 0;
568
569 memcpy(newlim->pl_rlimit, lim->pl_rlimit,
570 sizeof(struct rlimit) * RLIM_NLIMITS);
571 if (l != 0)
572 strlcpy(newlim->pl_corename, lim->pl_corename, l);
573 else
574 newlim->pl_corename = defcorename;
575 simple_unlock(&lim->p_slock);
576 break;
577 }
578
579 return (newlim);
580 }
581
582 void
583 limfree(struct plimit *lim)
584 {
585 int n;
586
587 simple_lock(&lim->p_slock);
588 n = --lim->p_refcnt;
589 simple_unlock(&lim->p_slock);
590 if (n > 0)
591 return;
592 #ifdef DIAGNOSTIC
593 if (n < 0)
594 panic("limfree");
595 #endif
596 if (lim->pl_corename != defcorename)
597 free(lim->pl_corename, M_TEMP);
598 pool_put(&plimit_pool, lim);
599 }
600
601 struct pstats *
602 pstatscopy(struct pstats *ps)
603 {
604
605 struct pstats *newps;
606
607 newps = pool_get(&pstats_pool, PR_WAITOK);
608
609 memset(&newps->pstat_startzero, 0,
610 (unsigned) ((char *)&newps->pstat_endzero -
611 (char *)&newps->pstat_startzero));
612 memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy,
613 ((char *)&newps->pstat_endcopy -
614 (char *)&newps->pstat_startcopy));
615
616 return (newps);
617
618 }
619
620 void
621 pstatsfree(struct pstats *ps)
622 {
623
624 pool_put(&pstats_pool, ps);
625 }
626
627 /*
628 * sysctl interface in five parts
629 */
630
631 /*
632 * a routine for sysctl proc subtree helpers that need to pick a valid
633 * process by pid.
634 */
635 static int
636 sysctl_proc_findproc(struct lwp *l, struct proc **p2, pid_t pid)
637 {
638 struct proc *ptmp;
639 int error = 0;
640
641 if (pid == PROC_CURPROC)
642 ptmp = l->l_proc;
643 else if ((ptmp = pfind(pid)) == NULL)
644 error = ESRCH;
645
646 *p2 = ptmp;
647 return (error);
648 }
649
650 /*
651 * sysctl helper routine for setting a process's specific corefile
652 * name. picks the process based on the given pid and checks the
653 * correctness of the new value.
654 */
655 static int
656 sysctl_proc_corename(SYSCTLFN_ARGS)
657 {
658 struct proc *ptmp;
659 struct plimit *lim;
660 int error = 0, len;
661 char *cname;
662 char *tmp;
663 struct sysctlnode node;
664
665 /*
666 * is this all correct?
667 */
668 if (namelen != 0)
669 return (EINVAL);
670 if (name[-1] != PROC_PID_CORENAME)
671 return (EINVAL);
672
673 /*
674 * whom are we tweaking?
675 */
676 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]);
677 if (error)
678 return (error);
679
680 /* XXX this should be in p_find() */
681 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
682 ptmp, NULL, NULL, NULL);
683 if (error)
684 return (error);
685
686 cname = PNBUF_GET();
687 /*
688 * let them modify a temporary copy of the core name
689 */
690 node = *rnode;
691 strlcpy(cname, ptmp->p_limit->pl_corename, MAXPATHLEN);
692 node.sysctl_data = cname;
693 error = sysctl_lookup(SYSCTLFN_CALL(&node));
694
695 /*
696 * if that failed, or they have nothing new to say, or we've
697 * heard it before...
698 */
699 if (error || newp == NULL ||
700 strcmp(cname, ptmp->p_limit->pl_corename) == 0) {
701 goto done;
702 }
703
704 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CORENAME,
705 ptmp, cname, NULL, NULL);
706 if (error)
707 return (error);
708
709 /*
710 * no error yet and cname now has the new core name in it.
711 * let's see if it looks acceptable. it must be either "core"
712 * or end in ".core" or "/core".
713 */
714 len = strlen(cname);
715 if (len < 4) {
716 error = EINVAL;
717 } else if (strcmp(cname + len - 4, "core") != 0) {
718 error = EINVAL;
719 } else if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.') {
720 error = EINVAL;
721 }
722 if (error != 0) {
723 goto done;
724 }
725
726 /*
727 * hmm...looks good. now...where do we put it?
728 */
729 tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL);
730 if (tmp == NULL) {
731 error = ENOMEM;
732 goto done;
733 }
734 strlcpy(tmp, cname, len + 1);
735
736 mutex_enter(&ptmp->p_mutex);
737 lim = ptmp->p_limit;
738 if (lim->p_refcnt > 1 && (lim->p_lflags & PL_SHAREMOD) == 0) {
739 ptmp->p_limit = limcopy(ptmp);
740 limfree(lim);
741 lim = ptmp->p_limit;
742 }
743 if (lim->pl_corename != defcorename)
744 free(lim->pl_corename, M_TEMP);
745 lim->pl_corename = tmp;
746 mutex_exit(&ptmp->p_mutex);
747 done:
748 PNBUF_PUT(cname);
749 return error;
750 }
751
752 /*
753 * sysctl helper routine for checking/setting a process's stop flags,
754 * one for fork and one for exec.
755 */
756 static int
757 sysctl_proc_stop(SYSCTLFN_ARGS)
758 {
759 struct proc *ptmp;
760 int i, f, error = 0;
761 struct sysctlnode node;
762
763 if (namelen != 0)
764 return (EINVAL);
765
766 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]);
767 if (error)
768 return (error);
769
770 /* XXX this should be in p_find() */
771 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
772 ptmp, NULL, NULL, NULL);
773 if (error)
774 return (error);
775
776 switch (rnode->sysctl_num) {
777 case PROC_PID_STOPFORK:
778 f = PS_STOPFORK;
779 break;
780 case PROC_PID_STOPEXEC:
781 f = PS_STOPEXEC;
782 break;
783 case PROC_PID_STOPEXIT:
784 f = PS_STOPEXIT;
785 break;
786 default:
787 return (EINVAL);
788 }
789
790 i = (ptmp->p_flag & f) ? 1 : 0;
791 node = *rnode;
792 node.sysctl_data = &i;
793 error = sysctl_lookup(SYSCTLFN_CALL(&node));
794 if (error || newp == NULL)
795 return (error);
796
797 mutex_enter(&ptmp->p_smutex);
798 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_STOPFLAG,
799 ptmp, KAUTH_ARG(f), NULL, NULL);
800 if (error)
801 return (error);
802 if (i)
803 ptmp->p_sflag |= f;
804 else
805 ptmp->p_sflag &= ~f;
806 mutex_exit(&ptmp->p_smutex);
807
808 return (0);
809 }
810
811 /*
812 * sysctl helper routine for a process's rlimits as exposed by sysctl.
813 */
814 static int
815 sysctl_proc_plimit(SYSCTLFN_ARGS)
816 {
817 struct proc *ptmp;
818 u_int limitno;
819 int which, error = 0;
820 struct rlimit alim;
821 struct sysctlnode node;
822
823 if (namelen != 0)
824 return (EINVAL);
825
826 which = name[-1];
827 if (which != PROC_PID_LIMIT_TYPE_SOFT &&
828 which != PROC_PID_LIMIT_TYPE_HARD)
829 return (EINVAL);
830
831 limitno = name[-2] - 1;
832 if (limitno >= RLIM_NLIMITS)
833 return (EINVAL);
834
835 if (name[-3] != PROC_PID_LIMIT)
836 return (EINVAL);
837
838 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-4]);
839 if (error)
840 return (error);
841
842 /* XXX this should be in p_find() */
843 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
844 ptmp, NULL, NULL, NULL);
845 if (error)
846 return (error);
847
848 node = *rnode;
849 memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim));
850 if (which == PROC_PID_LIMIT_TYPE_HARD)
851 node.sysctl_data = &alim.rlim_max;
852 else
853 node.sysctl_data = &alim.rlim_cur;
854
855 error = sysctl_lookup(SYSCTLFN_CALL(&node));
856 if (error || newp == NULL)
857 return (error);
858
859 return (dosetrlimit(l, ptmp, limitno, &alim));
860 }
861
862 /*
863 * and finally, the actually glue that sticks it to the tree
864 */
865 SYSCTL_SETUP(sysctl_proc_setup, "sysctl proc subtree setup")
866 {
867
868 sysctl_createv(clog, 0, NULL, NULL,
869 CTLFLAG_PERMANENT,
870 CTLTYPE_NODE, "proc", NULL,
871 NULL, 0, NULL, 0,
872 CTL_PROC, CTL_EOL);
873 sysctl_createv(clog, 0, NULL, NULL,
874 CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER,
875 CTLTYPE_NODE, "curproc",
876 SYSCTL_DESCR("Per-process settings"),
877 NULL, 0, NULL, 0,
878 CTL_PROC, PROC_CURPROC, CTL_EOL);
879
880 sysctl_createv(clog, 0, NULL, NULL,
881 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
882 CTLTYPE_STRING, "corename",
883 SYSCTL_DESCR("Core file name"),
884 sysctl_proc_corename, 0, NULL, MAXPATHLEN,
885 CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL);
886 sysctl_createv(clog, 0, NULL, NULL,
887 CTLFLAG_PERMANENT,
888 CTLTYPE_NODE, "rlimit",
889 SYSCTL_DESCR("Process limits"),
890 NULL, 0, NULL, 0,
891 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL);
892
893 #define create_proc_plimit(s, n) do { \
894 sysctl_createv(clog, 0, NULL, NULL, \
895 CTLFLAG_PERMANENT, \
896 CTLTYPE_NODE, s, \
897 SYSCTL_DESCR("Process " s " limits"), \
898 NULL, 0, NULL, 0, \
899 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \
900 CTL_EOL); \
901 sysctl_createv(clog, 0, NULL, NULL, \
902 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \
903 CTLTYPE_QUAD, "soft", \
904 SYSCTL_DESCR("Process soft " s " limit"), \
905 sysctl_proc_plimit, 0, NULL, 0, \
906 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \
907 PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL); \
908 sysctl_createv(clog, 0, NULL, NULL, \
909 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \
910 CTLTYPE_QUAD, "hard", \
911 SYSCTL_DESCR("Process hard " s " limit"), \
912 sysctl_proc_plimit, 0, NULL, 0, \
913 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \
914 PROC_PID_LIMIT_TYPE_HARD, CTL_EOL); \
915 } while (0/*CONSTCOND*/)
916
917 create_proc_plimit("cputime", PROC_PID_LIMIT_CPU);
918 create_proc_plimit("filesize", PROC_PID_LIMIT_FSIZE);
919 create_proc_plimit("datasize", PROC_PID_LIMIT_DATA);
920 create_proc_plimit("stacksize", PROC_PID_LIMIT_STACK);
921 create_proc_plimit("coredumpsize", PROC_PID_LIMIT_CORE);
922 create_proc_plimit("memoryuse", PROC_PID_LIMIT_RSS);
923 create_proc_plimit("memorylocked", PROC_PID_LIMIT_MEMLOCK);
924 create_proc_plimit("maxproc", PROC_PID_LIMIT_NPROC);
925 create_proc_plimit("descriptors", PROC_PID_LIMIT_NOFILE);
926 create_proc_plimit("sbsize", PROC_PID_LIMIT_SBSIZE);
927
928 #undef create_proc_plimit
929
930 sysctl_createv(clog, 0, NULL, NULL,
931 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
932 CTLTYPE_INT, "stopfork",
933 SYSCTL_DESCR("Stop process at fork(2)"),
934 sysctl_proc_stop, 0, NULL, 0,
935 CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL);
936 sysctl_createv(clog, 0, NULL, NULL,
937 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
938 CTLTYPE_INT, "stopexec",
939 SYSCTL_DESCR("Stop process at execve(2)"),
940 sysctl_proc_stop, 0, NULL, 0,
941 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL);
942 sysctl_createv(clog, 0, NULL, NULL,
943 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
944 CTLTYPE_INT, "stopexit",
945 SYSCTL_DESCR("Stop process before completing exit"),
946 sysctl_proc_stop, 0, NULL, 0,
947 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL);
948 }
949
950 struct uidinfo *
951 uid_find(uid_t uid)
952 {
953 struct uidinfo *uip;
954 struct uidinfo *newuip = NULL;
955 struct uihashhead *uipp;
956
957 uipp = UIHASH(uid);
958
959 again:
960 simple_lock(&uihashtbl_slock);
961 LIST_FOREACH(uip, uipp, ui_hash)
962 if (uip->ui_uid == uid) {
963 simple_unlock(&uihashtbl_slock);
964 if (newuip)
965 free(newuip, M_PROC);
966 return uip;
967 }
968
969 if (newuip == NULL) {
970 simple_unlock(&uihashtbl_slock);
971 newuip = malloc(sizeof(*uip), M_PROC, M_WAITOK | M_ZERO);
972 goto again;
973 }
974 uip = newuip;
975
976 LIST_INSERT_HEAD(uipp, uip, ui_hash);
977 uip->ui_uid = uid;
978 simple_lock_init(&uip->ui_slock);
979 simple_unlock(&uihashtbl_slock);
980
981 return uip;
982 }
983
984 /*
985 * Change the count associated with number of processes
986 * a given user is using.
987 */
988 int
989 chgproccnt(uid_t uid, int diff)
990 {
991 struct uidinfo *uip;
992 int s;
993
994 if (diff == 0)
995 return 0;
996
997 uip = uid_find(uid);
998 UILOCK(uip, s);
999 uip->ui_proccnt += diff;
1000 KASSERT(uip->ui_proccnt >= 0);
1001 UIUNLOCK(uip, s);
1002 return uip->ui_proccnt;
1003 }
1004
1005 int
1006 chgsbsize(struct uidinfo *uip, u_long *hiwat, u_long to, rlim_t xmax)
1007 {
1008 rlim_t nsb;
1009 int s;
1010
1011 UILOCK(uip, s);
1012 nsb = uip->ui_sbsize + to - *hiwat;
1013 if (to > *hiwat && nsb > xmax) {
1014 UIUNLOCK(uip, s);
1015 splx(s);
1016 return 0;
1017 }
1018 *hiwat = to;
1019 uip->ui_sbsize = nsb;
1020 KASSERT(uip->ui_sbsize >= 0);
1021 UIUNLOCK(uip, s);
1022 return 1;
1023 }
1024