xref: /netbsd-src/sys/kern/kern_resource.c (revision 503611ba29d4c920cb1878a9ece7ebd1e0ac2e16)
1 /*	$NetBSD: kern_resource.c,v 1.104 2006/09/08 20:58:57 elad 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 1.104 2006/09/08 20:58:57 elad 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/sa.h>
56 #include <sys/syscallargs.h>
57 
58 #include <uvm/uvm_extern.h>
59 
60 /*
61  * Maximum process data and stack limits.
62  * They are variables so they are patchable.
63  */
64 rlim_t maxdmap = MAXDSIZ;
65 rlim_t maxsmap = MAXSSIZ;
66 
67 struct uihashhead *uihashtbl;
68 u_long uihash;		/* size of hash table - 1 */
69 struct simplelock uihashtbl_slock = SIMPLELOCK_INITIALIZER;
70 
71 
72 /*
73  * Resource controls and accounting.
74  */
75 
76 int
77 sys_getpriority(struct lwp *l, void *v, register_t *retval)
78 {
79 	struct sys_getpriority_args /* {
80 		syscallarg(int) which;
81 		syscallarg(id_t) who;
82 	} */ *uap = v;
83 	struct proc *curp = l->l_proc, *p;
84 	int low = NZERO + PRIO_MAX + 1;
85 
86 	switch (SCARG(uap, which)) {
87 
88 	case PRIO_PROCESS:
89 		if (SCARG(uap, who) == 0)
90 			p = curp;
91 		else
92 			p = pfind(SCARG(uap, who));
93 		if (p == 0)
94 			break;
95 		low = p->p_nice;
96 		break;
97 
98 	case PRIO_PGRP: {
99 		struct pgrp *pg;
100 
101 		if (SCARG(uap, who) == 0)
102 			pg = curp->p_pgrp;
103 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
104 			break;
105 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
106 			if (p->p_nice < low)
107 				low = p->p_nice;
108 		}
109 		break;
110 	}
111 
112 	case PRIO_USER:
113 		if (SCARG(uap, who) == 0)
114 			SCARG(uap, who) = kauth_cred_geteuid(l->l_cred);
115 		proclist_lock_read();
116 		PROCLIST_FOREACH(p, &allproc) {
117 			if (kauth_cred_geteuid(p->p_cred) ==
118 			    (uid_t) SCARG(uap, who) && p->p_nice < low)
119 				low = p->p_nice;
120 		}
121 		proclist_unlock_read();
122 		break;
123 
124 	default:
125 		return (EINVAL);
126 	}
127 	if (low == NZERO + PRIO_MAX + 1)
128 		return (ESRCH);
129 	*retval = low - NZERO;
130 	return (0);
131 }
132 
133 /* ARGSUSED */
134 int
135 sys_setpriority(struct lwp *l, void *v, register_t *retval)
136 {
137 	struct sys_setpriority_args /* {
138 		syscallarg(int) which;
139 		syscallarg(id_t) who;
140 		syscallarg(int) prio;
141 	} */ *uap = v;
142 	struct proc *curp = l->l_proc, *p;
143 	int found = 0, error = 0;
144 
145 	switch (SCARG(uap, which)) {
146 
147 	case PRIO_PROCESS:
148 		if (SCARG(uap, who) == 0)
149 			p = curp;
150 		else
151 			p = pfind(SCARG(uap, who));
152 		if (p == 0)
153 			break;
154 		error = donice(l, p, SCARG(uap, prio));
155 		found++;
156 		break;
157 
158 	case PRIO_PGRP: {
159 		struct pgrp *pg;
160 
161 		if (SCARG(uap, who) == 0)
162 			pg = curp->p_pgrp;
163 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
164 			break;
165 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
166 			error = donice(l, p, SCARG(uap, prio));
167 			found++;
168 		}
169 		break;
170 	}
171 
172 	case PRIO_USER:
173 		if (SCARG(uap, who) == 0)
174 			SCARG(uap, who) = kauth_cred_geteuid(l->l_cred);
175 		proclist_lock_read();
176 		PROCLIST_FOREACH(p, &allproc) {
177 			if (kauth_cred_geteuid(p->p_cred) ==
178 			    (uid_t)SCARG(uap, who)) {
179 				error = donice(l, p, SCARG(uap, prio));
180 				found++;
181 			}
182 		}
183 		proclist_unlock_read();
184 		break;
185 
186 	default:
187 		return (EINVAL);
188 	}
189 	if (found == 0)
190 		return (ESRCH);
191 	return (error);
192 }
193 
194 int
195 donice(struct lwp *l, struct proc *chgp, int n)
196 {
197 	kauth_cred_t cred = l->l_cred;
198 	int s;
199 
200 	if (kauth_cred_geteuid(cred) && kauth_cred_getuid(cred) &&
201 	    kauth_cred_geteuid(cred) != kauth_cred_geteuid(chgp->p_cred) &&
202 	    kauth_cred_getuid(cred) != kauth_cred_geteuid(chgp->p_cred))
203 		return (EPERM);
204 	if (n > PRIO_MAX)
205 		n = PRIO_MAX;
206 	if (n < PRIO_MIN)
207 		n = PRIO_MIN;
208 	n += NZERO;
209 	if (n < chgp->p_nice && kauth_authorize_process(cred,
210 	    KAUTH_PROCESS_RESOURCE, chgp, (void *)KAUTH_REQ_PROCESS_RESOURCE_NICE,
211 	    (void *)(u_long)n, NULL))
212 		return (EACCES);
213 	chgp->p_nice = n;
214 	SCHED_LOCK(s);
215 	(void)resetprocpriority(chgp);
216 	SCHED_UNLOCK(s);
217 	return (0);
218 }
219 
220 /* ARGSUSED */
221 int
222 sys_setrlimit(struct lwp *l, void *v, register_t *retval)
223 {
224 	struct sys_setrlimit_args /* {
225 		syscallarg(int) which;
226 		syscallarg(const struct rlimit *) rlp;
227 	} */ *uap = v;
228 	int which = SCARG(uap, which);
229 	struct rlimit alim;
230 	int error;
231 
232 	error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit));
233 	if (error)
234 		return (error);
235 	return (dosetrlimit(l, l->l_proc, which, &alim));
236 }
237 
238 int
239 dosetrlimit(struct lwp *l, struct proc *p, int which, struct rlimit *limp)
240 {
241 	struct rlimit *alimp;
242 	struct plimit *oldplim;
243 	int error;
244 
245 	if ((u_int)which >= RLIM_NLIMITS)
246 		return (EINVAL);
247 
248 	if (limp->rlim_cur < 0 || limp->rlim_max < 0)
249 		return (EINVAL);
250 
251 	alimp = &p->p_rlimit[which];
252 	/* if we don't change the value, no need to limcopy() */
253 	if (limp->rlim_cur == alimp->rlim_cur &&
254 	    limp->rlim_max == alimp->rlim_max)
255 		return 0;
256 
257 	if (limp->rlim_cur > limp->rlim_max) {
258 		/*
259 		 * This is programming error. According to SUSv2, we should
260 		 * return error in this case.
261 		 */
262 		return (EINVAL);
263 	}
264 	if (limp->rlim_max > alimp->rlim_max && (error =
265 	    kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RESOURCE,
266 	    p, (void *)KAUTH_REQ_PROCESS_RESOURCE_RLIMIT, limp,
267 	    (void *)(u_long)which)) != KAUTH_RESULT_ALLOW)
268 			return (error);
269 
270 	if (p->p_limit->p_refcnt > 1 &&
271 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
272 		p->p_limit = limcopy(oldplim = p->p_limit);
273 		limfree(oldplim);
274 		alimp = &p->p_rlimit[which];
275 	}
276 
277 	switch (which) {
278 
279 	case RLIMIT_DATA:
280 		if (limp->rlim_cur > maxdmap)
281 			limp->rlim_cur = maxdmap;
282 		if (limp->rlim_max > maxdmap)
283 			limp->rlim_max = maxdmap;
284 		break;
285 
286 	case RLIMIT_STACK:
287 		if (limp->rlim_cur > maxsmap)
288 			limp->rlim_cur = maxsmap;
289 		if (limp->rlim_max > maxsmap)
290 			limp->rlim_max = maxsmap;
291 
292 		/*
293 		 * Return EINVAL if the new stack size limit is lower than
294 		 * current usage. Otherwise, the process would get SIGSEGV the
295 		 * moment it would try to access anything on it's current stack.
296 		 * This conforms to SUSv2.
297 		 */
298 		if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE
299 		    || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE)
300 			return (EINVAL);
301 
302 		/*
303 		 * Stack is allocated to the max at exec time with
304 		 * only "rlim_cur" bytes accessible (In other words,
305 		 * allocates stack dividing two contiguous regions at
306 		 * "rlim_cur" bytes boundary).
307 		 *
308 		 * Since allocation is done in terms of page, roundup
309 		 * "rlim_cur" (otherwise, contiguous regions
310 		 * overlap).  If stack limit is going up make more
311 		 * accessible, if going down make inaccessible.
312 		 */
313 		limp->rlim_cur = round_page(limp->rlim_cur);
314 		if (limp->rlim_cur != alimp->rlim_cur) {
315 			vaddr_t addr;
316 			vsize_t size;
317 			vm_prot_t prot;
318 
319 			if (limp->rlim_cur > alimp->rlim_cur) {
320 				prot = VM_PROT_READ | VM_PROT_WRITE;
321 				size = limp->rlim_cur - alimp->rlim_cur;
322 				addr = (vaddr_t)p->p_vmspace->vm_minsaddr -
323 				    limp->rlim_cur;
324 			} else {
325 				prot = VM_PROT_NONE;
326 				size = alimp->rlim_cur - limp->rlim_cur;
327 				addr = (vaddr_t)p->p_vmspace->vm_minsaddr -
328 				     alimp->rlim_cur;
329 			}
330 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
331 			    addr, addr+size, prot, FALSE);
332 		}
333 		break;
334 
335 	case RLIMIT_NOFILE:
336 		if (limp->rlim_cur > maxfiles)
337 			limp->rlim_cur = maxfiles;
338 		if (limp->rlim_max > maxfiles)
339 			limp->rlim_max = maxfiles;
340 		break;
341 
342 	case RLIMIT_NPROC:
343 		if (limp->rlim_cur > maxproc)
344 			limp->rlim_cur = maxproc;
345 		if (limp->rlim_max > maxproc)
346 			limp->rlim_max = maxproc;
347 		break;
348 	}
349 	*alimp = *limp;
350 	return (0);
351 }
352 
353 /* ARGSUSED */
354 int
355 sys_getrlimit(struct lwp *l, void *v, register_t *retval)
356 {
357 	struct sys_getrlimit_args /* {
358 		syscallarg(int) which;
359 		syscallarg(struct rlimit *) rlp;
360 	} */ *uap = v;
361 	struct proc *p = l->l_proc;
362 	int which = SCARG(uap, which);
363 
364 	if ((u_int)which >= RLIM_NLIMITS)
365 		return (EINVAL);
366 	return (copyout(&p->p_rlimit[which], SCARG(uap, rlp),
367 	    sizeof(struct rlimit)));
368 }
369 
370 /*
371  * Transform the running time and tick information in proc p into user,
372  * system, and interrupt time usage.
373  */
374 void
375 calcru(struct proc *p, struct timeval *up, struct timeval *sp,
376     struct timeval *ip)
377 {
378 	u_quad_t u, st, ut, it, tot;
379 	unsigned long sec;
380 	long usec;
381 	int s;
382 	struct timeval tv;
383 	struct lwp *l;
384 
385 	s = splstatclock();
386 	st = p->p_sticks;
387 	ut = p->p_uticks;
388 	it = p->p_iticks;
389 	splx(s);
390 
391 	sec = p->p_rtime.tv_sec;
392 	usec = p->p_rtime.tv_usec;
393 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
394 		if (l->l_stat == LSONPROC) {
395 			struct schedstate_percpu *spc;
396 
397 			KDASSERT(l->l_cpu != NULL);
398 			spc = &l->l_cpu->ci_schedstate;
399 
400 			/*
401 			 * Adjust for the current time slice.  This is
402 			 * actually fairly important since the error
403 			 * here is on the order of a time quantum,
404 			 * which is much greater than the sampling
405 			 * error.
406 			 */
407 			microtime(&tv);
408 			sec += tv.tv_sec - spc->spc_runtime.tv_sec;
409 			usec += tv.tv_usec - spc->spc_runtime.tv_usec;
410 		}
411 	}
412 
413 	tot = st + ut + it;
414 	u = sec * 1000000ull + usec;
415 
416 	if (tot == 0) {
417 		/* No ticks, so can't use to share time out, split 50-50 */
418 		st = ut = u / 2;
419 	} else {
420 		st = (u * st) / tot;
421 		ut = (u * ut) / tot;
422 	}
423 	sp->tv_sec = st / 1000000;
424 	sp->tv_usec = st % 1000000;
425 	up->tv_sec = ut / 1000000;
426 	up->tv_usec = ut % 1000000;
427 	if (ip != NULL) {
428 		if (it != 0)
429 			it = (u * it) / tot;
430 		ip->tv_sec = it / 1000000;
431 		ip->tv_usec = it % 1000000;
432 	}
433 }
434 
435 /* ARGSUSED */
436 int
437 sys_getrusage(struct lwp *l, void *v, register_t *retval)
438 {
439 	struct sys_getrusage_args /* {
440 		syscallarg(int) who;
441 		syscallarg(struct rusage *) rusage;
442 	} */ *uap = v;
443 	struct rusage *rup;
444 	struct proc *p = l->l_proc;
445 
446 	switch (SCARG(uap, who)) {
447 
448 	case RUSAGE_SELF:
449 		rup = &p->p_stats->p_ru;
450 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
451 		break;
452 
453 	case RUSAGE_CHILDREN:
454 		rup = &p->p_stats->p_cru;
455 		break;
456 
457 	default:
458 		return (EINVAL);
459 	}
460 	return (copyout(rup, SCARG(uap, rusage), sizeof(struct rusage)));
461 }
462 
463 void
464 ruadd(struct rusage *ru, struct rusage *ru2)
465 {
466 	long *ip, *ip2;
467 	int i;
468 
469 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
470 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
471 	if (ru->ru_maxrss < ru2->ru_maxrss)
472 		ru->ru_maxrss = ru2->ru_maxrss;
473 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
474 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
475 		*ip++ += *ip2++;
476 }
477 
478 /*
479  * Make a copy of the plimit structure.
480  * We share these structures copy-on-write after fork,
481  * and copy when a limit is changed.
482  */
483 struct plimit *
484 limcopy(struct plimit *lim)
485 {
486 	struct plimit *newlim;
487 	size_t l = 0;
488 
489 	simple_lock(&lim->p_slock);
490 	if (lim->pl_corename != defcorename)
491 		l = strlen(lim->pl_corename) + 1;
492 	simple_unlock(&lim->p_slock);
493 
494 	newlim = pool_get(&plimit_pool, PR_WAITOK);
495 	simple_lock_init(&newlim->p_slock);
496 	newlim->p_lflags = 0;
497 	newlim->p_refcnt = 1;
498 	newlim->pl_corename = (l != 0)
499 		? malloc(l, M_TEMP, M_WAITOK)
500 		: defcorename;
501 
502 	simple_lock(&lim->p_slock);
503 	memcpy(newlim->pl_rlimit, lim->pl_rlimit,
504 	    sizeof(struct rlimit) * RLIM_NLIMITS);
505 
506 	if (l != 0)
507 		strlcpy(newlim->pl_corename, lim->pl_corename, l);
508 	simple_unlock(&lim->p_slock);
509 
510 	return (newlim);
511 }
512 
513 void
514 limfree(struct plimit *lim)
515 {
516 	int n;
517 
518 	simple_lock(&lim->p_slock);
519 	n = --lim->p_refcnt;
520 	simple_unlock(&lim->p_slock);
521 	if (n > 0)
522 		return;
523 #ifdef DIAGNOSTIC
524 	if (n < 0)
525 		panic("limfree");
526 #endif
527 	if (lim->pl_corename != defcorename)
528 		free(lim->pl_corename, M_TEMP);
529 	pool_put(&plimit_pool, lim);
530 }
531 
532 struct pstats *
533 pstatscopy(struct pstats *ps)
534 {
535 
536 	struct pstats *newps;
537 
538 	newps = pool_get(&pstats_pool, PR_WAITOK);
539 
540 	memset(&newps->pstat_startzero, 0,
541 	(unsigned) ((caddr_t)&newps->pstat_endzero -
542 		    (caddr_t)&newps->pstat_startzero));
543 	memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy,
544 	((caddr_t)&newps->pstat_endcopy -
545 	 (caddr_t)&newps->pstat_startcopy));
546 
547 	return (newps);
548 
549 }
550 
551 void
552 pstatsfree(struct pstats *ps)
553 {
554 
555 	pool_put(&pstats_pool, ps);
556 }
557 
558 /*
559  * sysctl interface in five parts
560  */
561 
562 /*
563  * a routine for sysctl proc subtree helpers that need to pick a valid
564  * process by pid.
565  */
566 static int
567 sysctl_proc_findproc(struct lwp *l, struct proc **p2, pid_t pid)
568 {
569 	struct proc *ptmp;
570 	int error = 0;
571 
572 	if (pid == PROC_CURPROC)
573 		ptmp = l->l_proc;
574 	else if ((ptmp = pfind(pid)) == NULL)
575 		error = ESRCH;
576 	else {
577 		boolean_t isroot = kauth_authorize_generic(l->l_cred,
578 		    KAUTH_GENERIC_ISSUSER, NULL);
579 		/*
580 		 * suid proc of ours or proc not ours
581 		 */
582 		if (kauth_cred_getuid(l->l_cred) !=
583 		    kauth_cred_getuid(ptmp->p_cred) ||
584 		    kauth_cred_getuid(l->l_cred) !=
585 		    kauth_cred_getsvuid(ptmp->p_cred))
586 			error = isroot ? 0 : EPERM;
587 
588 		/*
589 		 * sgid proc has sgid back to us temporarily
590 		 */
591 		else if (kauth_cred_getgid(ptmp->p_cred) !=
592 		    kauth_cred_getsvgid(ptmp->p_cred))
593 			error = isroot ? 0 : EPERM;
594 
595 		/*
596 		 * our rgid must be in target's group list (ie,
597 		 * sub-processes started by a sgid process)
598 		 */
599 		else {
600 			int ismember = 0;
601 
602 			if (kauth_cred_ismember_gid(l->l_cred,
603 			    kauth_cred_getgid(ptmp->p_cred), &ismember) != 0 ||
604 			    !ismember) {
605 				error = isroot ? 0 : EPERM;
606 			}
607 		}
608 	}
609 
610 	*p2 = ptmp;
611 	return (error);
612 }
613 
614 /*
615  * sysctl helper routine for setting a process's specific corefile
616  * name.  picks the process based on the given pid and checks the
617  * correctness of the new value.
618  */
619 static int
620 sysctl_proc_corename(SYSCTLFN_ARGS)
621 {
622 	struct proc *ptmp;
623 	struct plimit *lim;
624 	int error = 0, len;
625 	char *cname;
626 	char *tmp;
627 	struct sysctlnode node;
628 
629 	/*
630 	 * is this all correct?
631 	 */
632 	if (namelen != 0)
633 		return (EINVAL);
634 	if (name[-1] != PROC_PID_CORENAME)
635 		return (EINVAL);
636 
637 	/*
638 	 * whom are we tweaking?
639 	 */
640 	error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]);
641 	if (error)
642 		return (error);
643 
644 	cname = PNBUF_GET();
645 	/*
646 	 * let them modify a temporary copy of the core name
647 	 */
648 	node = *rnode;
649 	strlcpy(cname, ptmp->p_limit->pl_corename, MAXPATHLEN);
650 	node.sysctl_data = cname;
651 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
652 
653 	/*
654 	 * if that failed, or they have nothing new to say, or we've
655 	 * heard it before...
656 	 */
657 	if (error || newp == NULL ||
658 	    strcmp(cname, ptmp->p_limit->pl_corename) == 0) {
659 		goto done;
660 	}
661 
662 	if (securelevel > 1)
663 		return (EPERM);
664 
665 	/*
666 	 * no error yet and cname now has the new core name in it.
667 	 * let's see if it looks acceptable.  it must be either "core"
668 	 * or end in ".core" or "/core".
669 	 */
670 	len = strlen(cname);
671 	if (len < 4) {
672 		error = EINVAL;
673 	} else if (strcmp(cname + len - 4, "core") != 0) {
674 		error = EINVAL;
675 	} else if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.') {
676 		error = EINVAL;
677 	}
678 	if (error != 0) {
679 		goto done;
680 	}
681 
682 	/*
683 	 * hmm...looks good.  now...where do we put it?
684 	 */
685 	tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL);
686 	if (tmp == NULL) {
687 		error = ENOMEM;
688 		goto done;
689 	}
690 	strlcpy(tmp, cname, len + 1);
691 
692 	lim = ptmp->p_limit;
693 	if (lim->p_refcnt > 1 && (lim->p_lflags & PL_SHAREMOD) == 0) {
694 		ptmp->p_limit = limcopy(lim);
695 		limfree(lim);
696 		lim = ptmp->p_limit;
697 	}
698 	if (lim->pl_corename != defcorename)
699 		free(lim->pl_corename, M_TEMP);
700 	lim->pl_corename = tmp;
701 done:
702 	PNBUF_PUT(cname);
703 	return error;
704 }
705 
706 /*
707  * sysctl helper routine for checking/setting a process's stop flags,
708  * one for fork and one for exec.
709  */
710 static int
711 sysctl_proc_stop(SYSCTLFN_ARGS)
712 {
713 	struct proc *ptmp;
714 	int i, f, error = 0;
715 	struct sysctlnode node;
716 
717 	if (namelen != 0)
718 		return (EINVAL);
719 
720 	error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]);
721 	if (error)
722 		return (error);
723 
724 	switch (rnode->sysctl_num) {
725 	case PROC_PID_STOPFORK:
726 		f = P_STOPFORK;
727 		break;
728 	case PROC_PID_STOPEXEC:
729 		f = P_STOPEXEC;
730 		break;
731 	case PROC_PID_STOPEXIT:
732 		f = P_STOPEXIT;
733 		break;
734 	default:
735 		return (EINVAL);
736 	}
737 
738 	i = (ptmp->p_flag & f) ? 1 : 0;
739 	node = *rnode;
740 	node.sysctl_data = &i;
741 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
742 	if (error || newp == NULL)
743 		return (error);
744 
745 	if (i)
746 		ptmp->p_flag |= f;
747 	else
748 		ptmp->p_flag &= ~f;
749 
750 	return (0);
751 }
752 
753 /*
754  * sysctl helper routine for a process's rlimits as exposed by sysctl.
755  */
756 static int
757 sysctl_proc_plimit(SYSCTLFN_ARGS)
758 {
759 	struct proc *ptmp;
760 	u_int limitno;
761 	int which, error = 0;
762         struct rlimit alim;
763 	struct sysctlnode node;
764 
765 	if (namelen != 0)
766 		return (EINVAL);
767 
768 	which = name[-1];
769 	if (which != PROC_PID_LIMIT_TYPE_SOFT &&
770 	    which != PROC_PID_LIMIT_TYPE_HARD)
771 		return (EINVAL);
772 
773 	limitno = name[-2] - 1;
774 	if (limitno >= RLIM_NLIMITS)
775 		return (EINVAL);
776 
777 	if (name[-3] != PROC_PID_LIMIT)
778 		return (EINVAL);
779 
780 	error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-4]);
781 	if (error)
782 		return (error);
783 
784 	node = *rnode;
785 	memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim));
786 	if (which == PROC_PID_LIMIT_TYPE_HARD)
787 		node.sysctl_data = &alim.rlim_max;
788 	else
789 		node.sysctl_data = &alim.rlim_cur;
790 
791 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
792 	if (error || newp == NULL)
793 		return (error);
794 
795 	return (dosetrlimit(l, ptmp, limitno, &alim));
796 }
797 
798 /*
799  * and finally, the actually glue that sticks it to the tree
800  */
801 SYSCTL_SETUP(sysctl_proc_setup, "sysctl proc subtree setup")
802 {
803 
804 	sysctl_createv(clog, 0, NULL, NULL,
805 		       CTLFLAG_PERMANENT,
806 		       CTLTYPE_NODE, "proc", NULL,
807 		       NULL, 0, NULL, 0,
808 		       CTL_PROC, CTL_EOL);
809 	sysctl_createv(clog, 0, NULL, NULL,
810 		       CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER,
811 		       CTLTYPE_NODE, "curproc",
812 		       SYSCTL_DESCR("Per-process settings"),
813 		       NULL, 0, NULL, 0,
814 		       CTL_PROC, PROC_CURPROC, CTL_EOL);
815 
816 	sysctl_createv(clog, 0, NULL, NULL,
817 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
818 		       CTLTYPE_STRING, "corename",
819 		       SYSCTL_DESCR("Core file name"),
820 		       sysctl_proc_corename, 0, NULL, MAXPATHLEN,
821 		       CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL);
822 	sysctl_createv(clog, 0, NULL, NULL,
823 		       CTLFLAG_PERMANENT,
824 		       CTLTYPE_NODE, "rlimit",
825 		       SYSCTL_DESCR("Process limits"),
826 		       NULL, 0, NULL, 0,
827 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL);
828 
829 #define create_proc_plimit(s, n) do {					\
830 	sysctl_createv(clog, 0, NULL, NULL,				\
831 		       CTLFLAG_PERMANENT,				\
832 		       CTLTYPE_NODE, s,					\
833 		       SYSCTL_DESCR("Process " s " limits"),		\
834 		       NULL, 0, NULL, 0,				\
835 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
836 		       CTL_EOL);					\
837 	sysctl_createv(clog, 0, NULL, NULL,				\
838 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \
839 		       CTLTYPE_QUAD, "soft",				\
840 		       SYSCTL_DESCR("Process soft " s " limit"),	\
841 		       sysctl_proc_plimit, 0, NULL, 0,			\
842 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
843 		       PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL);		\
844 	sysctl_createv(clog, 0, NULL, NULL,				\
845 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \
846 		       CTLTYPE_QUAD, "hard",				\
847 		       SYSCTL_DESCR("Process hard " s " limit"),	\
848 		       sysctl_proc_plimit, 0, NULL, 0,			\
849 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
850 		       PROC_PID_LIMIT_TYPE_HARD, CTL_EOL);		\
851 	} while (0/*CONSTCOND*/)
852 
853 	create_proc_plimit("cputime",		PROC_PID_LIMIT_CPU);
854 	create_proc_plimit("filesize",		PROC_PID_LIMIT_FSIZE);
855 	create_proc_plimit("datasize",		PROC_PID_LIMIT_DATA);
856 	create_proc_plimit("stacksize",		PROC_PID_LIMIT_STACK);
857 	create_proc_plimit("coredumpsize",	PROC_PID_LIMIT_CORE);
858 	create_proc_plimit("memoryuse",		PROC_PID_LIMIT_RSS);
859 	create_proc_plimit("memorylocked",	PROC_PID_LIMIT_MEMLOCK);
860 	create_proc_plimit("maxproc",		PROC_PID_LIMIT_NPROC);
861 	create_proc_plimit("descriptors",	PROC_PID_LIMIT_NOFILE);
862 	create_proc_plimit("sbsize",		PROC_PID_LIMIT_SBSIZE);
863 
864 #undef create_proc_plimit
865 
866 	sysctl_createv(clog, 0, NULL, NULL,
867 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
868 		       CTLTYPE_INT, "stopfork",
869 		       SYSCTL_DESCR("Stop process at fork(2)"),
870 		       sysctl_proc_stop, 0, NULL, 0,
871 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL);
872 	sysctl_createv(clog, 0, NULL, NULL,
873 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
874 		       CTLTYPE_INT, "stopexec",
875 		       SYSCTL_DESCR("Stop process at execve(2)"),
876 		       sysctl_proc_stop, 0, NULL, 0,
877 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL);
878 	sysctl_createv(clog, 0, NULL, NULL,
879 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
880 		       CTLTYPE_INT, "stopexit",
881 		       SYSCTL_DESCR("Stop process before completing exit"),
882 		       sysctl_proc_stop, 0, NULL, 0,
883 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL);
884 }
885 
886 struct uidinfo *
887 uid_find(uid_t uid)
888 {
889 	struct uidinfo *uip;
890 	struct uidinfo *newuip = NULL;
891 	struct uihashhead *uipp;
892 
893 	uipp = UIHASH(uid);
894 
895 again:
896 	simple_lock(&uihashtbl_slock);
897 	LIST_FOREACH(uip, uipp, ui_hash)
898 		if (uip->ui_uid == uid) {
899 			simple_unlock(&uihashtbl_slock);
900 			if (newuip)
901 				free(newuip, M_PROC);
902 			return uip;
903 		}
904 
905 	if (newuip == NULL) {
906 		simple_unlock(&uihashtbl_slock);
907 		newuip = malloc(sizeof(*uip), M_PROC, M_WAITOK | M_ZERO);
908 		goto again;
909 	}
910 	uip = newuip;
911 
912 	LIST_INSERT_HEAD(uipp, uip, ui_hash);
913 	uip->ui_uid = uid;
914 	simple_lock_init(&uip->ui_slock);
915 	simple_unlock(&uihashtbl_slock);
916 
917 	return uip;
918 }
919 
920 /*
921  * Change the count associated with number of processes
922  * a given user is using.
923  */
924 int
925 chgproccnt(uid_t uid, int diff)
926 {
927 	struct uidinfo *uip;
928 	int s;
929 
930 	if (diff == 0)
931 		return 0;
932 
933 	uip = uid_find(uid);
934 	UILOCK(uip, s);
935 	uip->ui_proccnt += diff;
936 	KASSERT(uip->ui_proccnt >= 0);
937 	UIUNLOCK(uip, s);
938 	return uip->ui_proccnt;
939 }
940 
941 int
942 chgsbsize(struct uidinfo *uip, u_long *hiwat, u_long to, rlim_t xmax)
943 {
944 	rlim_t nsb;
945 	int s;
946 
947 	UILOCK(uip, s);
948 	nsb = uip->ui_sbsize + to - *hiwat;
949 	if (to > *hiwat && nsb > xmax) {
950 		UIUNLOCK(uip, s);
951 		splx(s);
952 		return 0;
953 	}
954 	*hiwat = to;
955 	uip->ui_sbsize = nsb;
956 	KASSERT(uip->ui_sbsize >= 0);
957 	UIUNLOCK(uip, s);
958 	return 1;
959 }
960