kern_resource.c revision 1.116.2.1 1 /* $NetBSD: kern_resource.c,v 1.116.2.1 2007/03/21 20:16:31 ad 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 1.116.2.1 2007/03/21 20:16:31 ad 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 mutex_enter(&proclist_lock);
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 mutex_exit(&proclist_lock);
125 return (EINVAL);
126 }
127 mutex_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 mutex_enter(&proclist_lock);
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 mutex_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 KASSERT(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 KASSERT(mutex_owned(&p->p_mutex));
539
540 mutex_exit(&p->p_mutex);
541 newlim = pool_get(&plimit_pool, PR_WAITOK);
542 mutex_init(&newlim->p_lock, MUTEX_DEFAULT, IPL_NONE);
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 mutex_enter(&lim->p_lock);
550 if (lim->pl_corename != defcorename) {
551 l = strlen(lim->pl_corename) + 1;
552
553 mutex_exit(&lim->p_lock);
554 mutex_exit(&p->p_mutex);
555 corename = malloc(l, M_TEMP, M_WAITOK);
556 mutex_enter(&p->p_mutex);
557 mutex_enter(&lim->p_lock);
558
559 if (l != strlen(lim->pl_corename) + 1) {
560 mutex_exit(&lim->p_lock);
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 mutex_exit(&lim->p_lock);
576 break;
577 }
578
579 return (newlim);
580 }
581
582 void
583 limfree(struct plimit *lim)
584 {
585 int n;
586
587 mutex_enter(&lim->p_lock);
588 n = --lim->p_refcnt;
589 mutex_exit(&lim->p_lock);
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 mutex_destroy(&lim->p_lock);
599 pool_put(&plimit_pool, lim);
600 }
601
602 struct pstats *
603 pstatscopy(struct pstats *ps)
604 {
605
606 struct pstats *newps;
607
608 newps = pool_get(&pstats_pool, PR_WAITOK);
609
610 memset(&newps->pstat_startzero, 0,
611 (unsigned) ((char *)&newps->pstat_endzero -
612 (char *)&newps->pstat_startzero));
613 memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy,
614 ((char *)&newps->pstat_endcopy -
615 (char *)&newps->pstat_startcopy));
616
617 return (newps);
618
619 }
620
621 void
622 pstatsfree(struct pstats *ps)
623 {
624
625 pool_put(&pstats_pool, ps);
626 }
627
628 /*
629 * sysctl interface in five parts
630 */
631
632 /*
633 * a routine for sysctl proc subtree helpers that need to pick a valid
634 * process by pid.
635 */
636 static int
637 sysctl_proc_findproc(struct lwp *l, struct proc **p2, pid_t pid)
638 {
639 struct proc *ptmp;
640 int error = 0;
641
642 if (pid == PROC_CURPROC)
643 ptmp = l->l_proc;
644 else if ((ptmp = pfind(pid)) == NULL)
645 error = ESRCH;
646
647 *p2 = ptmp;
648 return (error);
649 }
650
651 /*
652 * sysctl helper routine for setting a process's specific corefile
653 * name. picks the process based on the given pid and checks the
654 * correctness of the new value.
655 */
656 static int
657 sysctl_proc_corename(SYSCTLFN_ARGS)
658 {
659 struct proc *ptmp;
660 struct plimit *lim;
661 int error = 0, len;
662 char *cname;
663 char *tmp;
664 struct sysctlnode node;
665
666 /*
667 * is this all correct?
668 */
669 if (namelen != 0)
670 return (EINVAL);
671 if (name[-1] != PROC_PID_CORENAME)
672 return (EINVAL);
673
674 /*
675 * whom are we tweaking?
676 */
677 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]);
678 if (error)
679 return (error);
680
681 /* XXX this should be in p_find() */
682 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
683 ptmp, NULL, NULL, NULL);
684 if (error)
685 return (error);
686
687 cname = PNBUF_GET();
688 /*
689 * let them modify a temporary copy of the core name
690 */
691 node = *rnode;
692 strlcpy(cname, ptmp->p_limit->pl_corename, MAXPATHLEN);
693 node.sysctl_data = cname;
694 error = sysctl_lookup(SYSCTLFN_CALL(&node));
695
696 /*
697 * if that failed, or they have nothing new to say, or we've
698 * heard it before...
699 */
700 if (error || newp == NULL ||
701 strcmp(cname, ptmp->p_limit->pl_corename) == 0) {
702 goto done;
703 }
704
705 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CORENAME,
706 ptmp, cname, NULL, NULL);
707 if (error)
708 return (error);
709
710 /*
711 * no error yet and cname now has the new core name in it.
712 * let's see if it looks acceptable. it must be either "core"
713 * or end in ".core" or "/core".
714 */
715 len = strlen(cname);
716 if (len < 4) {
717 error = EINVAL;
718 } else if (strcmp(cname + len - 4, "core") != 0) {
719 error = EINVAL;
720 } else if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.') {
721 error = EINVAL;
722 }
723 if (error != 0) {
724 goto done;
725 }
726
727 /*
728 * hmm...looks good. now...where do we put it?
729 */
730 tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL);
731 if (tmp == NULL) {
732 error = ENOMEM;
733 goto done;
734 }
735 strlcpy(tmp, cname, len + 1);
736
737 mutex_enter(&ptmp->p_mutex);
738 lim = ptmp->p_limit;
739 if (lim->p_refcnt > 1 && (lim->p_lflags & PL_SHAREMOD) == 0) {
740 ptmp->p_limit = limcopy(ptmp);
741 limfree(lim);
742 lim = ptmp->p_limit;
743 }
744 if (lim->pl_corename != defcorename)
745 free(lim->pl_corename, M_TEMP);
746 lim->pl_corename = tmp;
747 mutex_exit(&ptmp->p_mutex);
748 done:
749 PNBUF_PUT(cname);
750 return error;
751 }
752
753 /*
754 * sysctl helper routine for checking/setting a process's stop flags,
755 * one for fork and one for exec.
756 */
757 static int
758 sysctl_proc_stop(SYSCTLFN_ARGS)
759 {
760 struct proc *ptmp;
761 int i, f, error = 0;
762 struct sysctlnode node;
763
764 if (namelen != 0)
765 return (EINVAL);
766
767 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]);
768 if (error)
769 return (error);
770
771 /* XXX this should be in p_find() */
772 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
773 ptmp, NULL, NULL, NULL);
774 if (error)
775 return (error);
776
777 switch (rnode->sysctl_num) {
778 case PROC_PID_STOPFORK:
779 f = PS_STOPFORK;
780 break;
781 case PROC_PID_STOPEXEC:
782 f = PS_STOPEXEC;
783 break;
784 case PROC_PID_STOPEXIT:
785 f = PS_STOPEXIT;
786 break;
787 default:
788 return (EINVAL);
789 }
790
791 i = (ptmp->p_flag & f) ? 1 : 0;
792 node = *rnode;
793 node.sysctl_data = &i;
794 error = sysctl_lookup(SYSCTLFN_CALL(&node));
795 if (error || newp == NULL)
796 return (error);
797
798 mutex_enter(&ptmp->p_smutex);
799 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_STOPFLAG,
800 ptmp, KAUTH_ARG(f), NULL, NULL);
801 if (error)
802 return (error);
803 if (i)
804 ptmp->p_sflag |= f;
805 else
806 ptmp->p_sflag &= ~f;
807 mutex_exit(&ptmp->p_smutex);
808
809 return (0);
810 }
811
812 /*
813 * sysctl helper routine for a process's rlimits as exposed by sysctl.
814 */
815 static int
816 sysctl_proc_plimit(SYSCTLFN_ARGS)
817 {
818 struct proc *ptmp;
819 u_int limitno;
820 int which, error = 0;
821 struct rlimit alim;
822 struct sysctlnode node;
823
824 if (namelen != 0)
825 return (EINVAL);
826
827 which = name[-1];
828 if (which != PROC_PID_LIMIT_TYPE_SOFT &&
829 which != PROC_PID_LIMIT_TYPE_HARD)
830 return (EINVAL);
831
832 limitno = name[-2] - 1;
833 if (limitno >= RLIM_NLIMITS)
834 return (EINVAL);
835
836 if (name[-3] != PROC_PID_LIMIT)
837 return (EINVAL);
838
839 error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-4]);
840 if (error)
841 return (error);
842
843 /* XXX this should be in p_find() */
844 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
845 ptmp, NULL, NULL, NULL);
846 if (error)
847 return (error);
848
849 node = *rnode;
850 memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim));
851 if (which == PROC_PID_LIMIT_TYPE_HARD)
852 node.sysctl_data = &alim.rlim_max;
853 else
854 node.sysctl_data = &alim.rlim_cur;
855
856 error = sysctl_lookup(SYSCTLFN_CALL(&node));
857 if (error || newp == NULL)
858 return (error);
859
860 return (dosetrlimit(l, ptmp, limitno, &alim));
861 }
862
863 /*
864 * and finally, the actually glue that sticks it to the tree
865 */
866 SYSCTL_SETUP(sysctl_proc_setup, "sysctl proc subtree setup")
867 {
868
869 sysctl_createv(clog, 0, NULL, NULL,
870 CTLFLAG_PERMANENT,
871 CTLTYPE_NODE, "proc", NULL,
872 NULL, 0, NULL, 0,
873 CTL_PROC, CTL_EOL);
874 sysctl_createv(clog, 0, NULL, NULL,
875 CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER,
876 CTLTYPE_NODE, "curproc",
877 SYSCTL_DESCR("Per-process settings"),
878 NULL, 0, NULL, 0,
879 CTL_PROC, PROC_CURPROC, CTL_EOL);
880
881 sysctl_createv(clog, 0, NULL, NULL,
882 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
883 CTLTYPE_STRING, "corename",
884 SYSCTL_DESCR("Core file name"),
885 sysctl_proc_corename, 0, NULL, MAXPATHLEN,
886 CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL);
887 sysctl_createv(clog, 0, NULL, NULL,
888 CTLFLAG_PERMANENT,
889 CTLTYPE_NODE, "rlimit",
890 SYSCTL_DESCR("Process limits"),
891 NULL, 0, NULL, 0,
892 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL);
893
894 #define create_proc_plimit(s, n) do { \
895 sysctl_createv(clog, 0, NULL, NULL, \
896 CTLFLAG_PERMANENT, \
897 CTLTYPE_NODE, s, \
898 SYSCTL_DESCR("Process " s " limits"), \
899 NULL, 0, NULL, 0, \
900 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \
901 CTL_EOL); \
902 sysctl_createv(clog, 0, NULL, NULL, \
903 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \
904 CTLTYPE_QUAD, "soft", \
905 SYSCTL_DESCR("Process soft " s " limit"), \
906 sysctl_proc_plimit, 0, NULL, 0, \
907 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \
908 PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL); \
909 sysctl_createv(clog, 0, NULL, NULL, \
910 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \
911 CTLTYPE_QUAD, "hard", \
912 SYSCTL_DESCR("Process hard " s " limit"), \
913 sysctl_proc_plimit, 0, NULL, 0, \
914 CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n, \
915 PROC_PID_LIMIT_TYPE_HARD, CTL_EOL); \
916 } while (0/*CONSTCOND*/)
917
918 create_proc_plimit("cputime", PROC_PID_LIMIT_CPU);
919 create_proc_plimit("filesize", PROC_PID_LIMIT_FSIZE);
920 create_proc_plimit("datasize", PROC_PID_LIMIT_DATA);
921 create_proc_plimit("stacksize", PROC_PID_LIMIT_STACK);
922 create_proc_plimit("coredumpsize", PROC_PID_LIMIT_CORE);
923 create_proc_plimit("memoryuse", PROC_PID_LIMIT_RSS);
924 create_proc_plimit("memorylocked", PROC_PID_LIMIT_MEMLOCK);
925 create_proc_plimit("maxproc", PROC_PID_LIMIT_NPROC);
926 create_proc_plimit("descriptors", PROC_PID_LIMIT_NOFILE);
927 create_proc_plimit("sbsize", PROC_PID_LIMIT_SBSIZE);
928
929 #undef create_proc_plimit
930
931 sysctl_createv(clog, 0, NULL, NULL,
932 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
933 CTLTYPE_INT, "stopfork",
934 SYSCTL_DESCR("Stop process at fork(2)"),
935 sysctl_proc_stop, 0, NULL, 0,
936 CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL);
937 sysctl_createv(clog, 0, NULL, NULL,
938 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
939 CTLTYPE_INT, "stopexec",
940 SYSCTL_DESCR("Stop process at execve(2)"),
941 sysctl_proc_stop, 0, NULL, 0,
942 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL);
943 sysctl_createv(clog, 0, NULL, NULL,
944 CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
945 CTLTYPE_INT, "stopexit",
946 SYSCTL_DESCR("Stop process before completing exit"),
947 sysctl_proc_stop, 0, NULL, 0,
948 CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL);
949 }
950
951 struct uidinfo *
952 uid_find(uid_t uid)
953 {
954 struct uidinfo *uip;
955 struct uidinfo *newuip = NULL;
956 struct uihashhead *uipp;
957
958 uipp = UIHASH(uid);
959
960 again:
961 simple_lock(&uihashtbl_slock);
962 LIST_FOREACH(uip, uipp, ui_hash)
963 if (uip->ui_uid == uid) {
964 simple_unlock(&uihashtbl_slock);
965 if (newuip)
966 free(newuip, M_PROC);
967 return uip;
968 }
969
970 if (newuip == NULL) {
971 simple_unlock(&uihashtbl_slock);
972 newuip = malloc(sizeof(*uip), M_PROC, M_WAITOK | M_ZERO);
973 goto again;
974 }
975 uip = newuip;
976
977 LIST_INSERT_HEAD(uipp, uip, ui_hash);
978 uip->ui_uid = uid;
979 simple_lock_init(&uip->ui_slock);
980 simple_unlock(&uihashtbl_slock);
981
982 return uip;
983 }
984
985 /*
986 * Change the count associated with number of processes
987 * a given user is using.
988 */
989 int
990 chgproccnt(uid_t uid, int diff)
991 {
992 struct uidinfo *uip;
993 int s;
994
995 if (diff == 0)
996 return 0;
997
998 uip = uid_find(uid);
999 UILOCK(uip, s);
1000 uip->ui_proccnt += diff;
1001 KASSERT(uip->ui_proccnt >= 0);
1002 UIUNLOCK(uip, s);
1003 return uip->ui_proccnt;
1004 }
1005
1006 int
1007 chgsbsize(struct uidinfo *uip, u_long *hiwat, u_long to, rlim_t xmax)
1008 {
1009 rlim_t nsb;
1010 int s;
1011
1012 UILOCK(uip, s);
1013 nsb = uip->ui_sbsize + to - *hiwat;
1014 if (to > *hiwat && nsb > xmax) {
1015 UIUNLOCK(uip, s);
1016 splx(s);
1017 return 0;
1018 }
1019 *hiwat = to;
1020 uip->ui_sbsize = nsb;
1021 KASSERT(uip->ui_sbsize >= 0);
1022 UIUNLOCK(uip, s);
1023 return 1;
1024 }
1025