xref: /netbsd-src/sys/kern/kern_resource.c (revision 21e37cc72a480a47828990a439cde7ac9ffaf0c6)
1 /*	$NetBSD: kern_resource.c,v 1.85 2004/05/13 17:56:14 kleink 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 1.85 2004/05/13 17:56:14 kleink 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/pool.h>
49 #include <sys/proc.h>
50 #include <sys/sysctl.h>
51 
52 #include <sys/mount.h>
53 #include <sys/sa.h>
54 #include <sys/syscallargs.h>
55 
56 #include <uvm/uvm_extern.h>
57 
58 /*
59  * Maximum process data and stack limits.
60  * They are variables so they are patchable.
61  */
62 rlim_t maxdmap = MAXDSIZ;
63 rlim_t maxsmap = MAXSSIZ;
64 
65 struct uihashhead *uihashtbl;
66 u_long uihash;		/* size of hash table - 1 */
67 
68 static struct uidinfo *getuidinfo(uid_t);
69 static void freeuidinfo(struct uidinfo *);
70 static struct uidinfo *allocuidinfo(uid_t);
71 
72 /*
73  * Resource controls and accounting.
74  */
75 
76 int
77 sys_getpriority(l, v, retval)
78 	struct lwp *l;
79 	void *v;
80 	register_t *retval;
81 {
82 	struct sys_getpriority_args /* {
83 		syscallarg(int) which;
84 		syscallarg(id_t) who;
85 	} */ *uap = v;
86 	struct proc *curp = l->l_proc, *p;
87 	int low = NZERO + PRIO_MAX + 1;
88 
89 	switch (SCARG(uap, which)) {
90 
91 	case PRIO_PROCESS:
92 		if (SCARG(uap, who) == 0)
93 			p = curp;
94 		else
95 			p = pfind(SCARG(uap, who));
96 		if (p == 0)
97 			break;
98 		low = p->p_nice;
99 		break;
100 
101 	case PRIO_PGRP: {
102 		struct pgrp *pg;
103 
104 		if (SCARG(uap, who) == 0)
105 			pg = curp->p_pgrp;
106 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
107 			break;
108 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
109 			if (p->p_nice < low)
110 				low = p->p_nice;
111 		}
112 		break;
113 	}
114 
115 	case PRIO_USER:
116 		if (SCARG(uap, who) == 0)
117 			SCARG(uap, who) = curp->p_ucred->cr_uid;
118 		proclist_lock_read();
119 		LIST_FOREACH(p, &allproc, p_list) {
120 			if (p->p_ucred->cr_uid == (uid_t) SCARG(uap, who) &&
121 			    p->p_nice < low)
122 				low = p->p_nice;
123 		}
124 		proclist_unlock_read();
125 		break;
126 
127 	default:
128 		return (EINVAL);
129 	}
130 	if (low == NZERO + PRIO_MAX + 1)
131 		return (ESRCH);
132 	*retval = low - NZERO;
133 	return (0);
134 }
135 
136 /* ARGSUSED */
137 int
138 sys_setpriority(l, v, retval)
139 	struct lwp *l;
140 	void *v;
141 	register_t *retval;
142 {
143 	struct sys_setpriority_args /* {
144 		syscallarg(int) which;
145 		syscallarg(id_t) who;
146 		syscallarg(int) prio;
147 	} */ *uap = v;
148 	struct proc *curp = l->l_proc, *p;
149 	int found = 0, error = 0;
150 
151 	switch (SCARG(uap, which)) {
152 
153 	case PRIO_PROCESS:
154 		if (SCARG(uap, who) == 0)
155 			p = curp;
156 		else
157 			p = pfind(SCARG(uap, who));
158 		if (p == 0)
159 			break;
160 		error = donice(curp, p, SCARG(uap, prio));
161 		found++;
162 		break;
163 
164 	case PRIO_PGRP: {
165 		struct pgrp *pg;
166 
167 		if (SCARG(uap, who) == 0)
168 			pg = curp->p_pgrp;
169 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
170 			break;
171 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
172 			error = donice(curp, p, SCARG(uap, prio));
173 			found++;
174 		}
175 		break;
176 	}
177 
178 	case PRIO_USER:
179 		if (SCARG(uap, who) == 0)
180 			SCARG(uap, who) = curp->p_ucred->cr_uid;
181 		proclist_lock_read();
182 		LIST_FOREACH(p, &allproc, p_list) {
183 			if (p->p_ucred->cr_uid == (uid_t) SCARG(uap, who)) {
184 				error = donice(curp, p, SCARG(uap, prio));
185 				found++;
186 			}
187 		}
188 		proclist_unlock_read();
189 		break;
190 
191 	default:
192 		return (EINVAL);
193 	}
194 	if (found == 0)
195 		return (ESRCH);
196 	return (error);
197 }
198 
199 int
200 donice(curp, chgp, n)
201 	struct proc *curp, *chgp;
202 	int n;
203 {
204 	struct pcred *pcred = curp->p_cred;
205 	int s;
206 
207 	if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
208 	    pcred->pc_ucred->cr_uid != chgp->p_ucred->cr_uid &&
209 	    pcred->p_ruid != chgp->p_ucred->cr_uid)
210 		return (EPERM);
211 	if (n > PRIO_MAX)
212 		n = PRIO_MAX;
213 	if (n < PRIO_MIN)
214 		n = PRIO_MIN;
215 	n += NZERO;
216 	if (n < chgp->p_nice && suser(pcred->pc_ucred, &curp->p_acflag))
217 		return (EACCES);
218 	chgp->p_nice = n;
219 	SCHED_LOCK(s);
220 	(void)resetprocpriority(chgp);
221 	SCHED_UNLOCK(s);
222 	return (0);
223 }
224 
225 /* ARGSUSED */
226 int
227 sys_setrlimit(l, v, retval)
228 	struct lwp *l;
229 	void *v;
230 	register_t *retval;
231 {
232 	struct sys_setrlimit_args /* {
233 		syscallarg(int) which;
234 		syscallarg(const struct rlimit *) rlp;
235 	} */ *uap = v;
236 	struct proc *p = l->l_proc;
237 	int which = SCARG(uap, which);
238 	struct rlimit alim;
239 	int error;
240 
241 	error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit));
242 	if (error)
243 		return (error);
244 	return (dosetrlimit(p, p->p_cred, which, &alim));
245 }
246 
247 int
248 dosetrlimit(p, cred, which, limp)
249 	struct proc *p;
250 	struct  pcred *cred;
251 	int which;
252 	struct rlimit *limp;
253 {
254 	struct rlimit *alimp;
255 	struct plimit *oldplim;
256 	int error;
257 
258 	if ((u_int)which >= RLIM_NLIMITS)
259 		return (EINVAL);
260 
261 	if (limp->rlim_cur < 0 || limp->rlim_max < 0)
262 		return (EINVAL);
263 
264 	alimp = &p->p_rlimit[which];
265 	/* if we don't change the value, no need to limcopy() */
266 	if (limp->rlim_cur == alimp->rlim_cur &&
267 	    limp->rlim_max == alimp->rlim_max)
268 		return 0;
269 
270 	if (limp->rlim_cur > limp->rlim_max) {
271 		/*
272 		 * This is programming error. According to SUSv2, we should
273 		 * return error in this case.
274 		 */
275 		return (EINVAL);
276 	}
277 	if (limp->rlim_max > alimp->rlim_max
278 	    && (error = suser(cred->pc_ucred, &p->p_acflag)) != 0)
279 			return (error);
280 
281 	if (p->p_limit->p_refcnt > 1 &&
282 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
283 		p->p_limit = limcopy(oldplim = p->p_limit);
284 		limfree(oldplim);
285 		alimp = &p->p_rlimit[which];
286 	}
287 
288 	switch (which) {
289 
290 	case RLIMIT_DATA:
291 		if (limp->rlim_cur > maxdmap)
292 			limp->rlim_cur = maxdmap;
293 		if (limp->rlim_max > maxdmap)
294 			limp->rlim_max = maxdmap;
295 		break;
296 
297 	case RLIMIT_STACK:
298 		if (limp->rlim_cur > maxsmap)
299 			limp->rlim_cur = maxsmap;
300 		if (limp->rlim_max > maxsmap)
301 			limp->rlim_max = maxsmap;
302 
303 		/*
304 		 * Return EINVAL if the new stack size limit is lower than
305 		 * current usage. Otherwise, the process would get SIGSEGV the
306 		 * moment it would try to access anything on it's current stack.
307 		 * This conforms to SUSv2.
308 		 */
309 		if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE
310 		    || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE)
311 			return (EINVAL);
312 
313 		/*
314 		 * Stack is allocated to the max at exec time with
315 		 * only "rlim_cur" bytes accessible (In other words,
316 		 * allocates stack dividing two contiguous regions at
317 		 * "rlim_cur" bytes boundary).
318 		 *
319 		 * Since allocation is done in terms of page, roundup
320 		 * "rlim_cur" (otherwise, contiguous regions
321 		 * overlap).  If stack limit is going up make more
322 		 * accessible, if going down make inaccessible.
323 		 */
324 		limp->rlim_cur = round_page(limp->rlim_cur);
325 		if (limp->rlim_cur != alimp->rlim_cur) {
326 			vaddr_t addr;
327 			vsize_t size;
328 			vm_prot_t prot;
329 
330 			if (limp->rlim_cur > alimp->rlim_cur) {
331 				prot = VM_PROT_READ | VM_PROT_WRITE;
332 				size = limp->rlim_cur - alimp->rlim_cur;
333 				addr = USRSTACK - limp->rlim_cur;
334 			} else {
335 				prot = VM_PROT_NONE;
336 				size = alimp->rlim_cur - limp->rlim_cur;
337 				addr = USRSTACK - alimp->rlim_cur;
338 			}
339 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
340 					      addr, addr+size, prot, FALSE);
341 		}
342 		break;
343 
344 	case RLIMIT_NOFILE:
345 		if (limp->rlim_cur > maxfiles)
346 			limp->rlim_cur = maxfiles;
347 		if (limp->rlim_max > maxfiles)
348 			limp->rlim_max = maxfiles;
349 		break;
350 
351 	case RLIMIT_NPROC:
352 		if (limp->rlim_cur > maxproc)
353 			limp->rlim_cur = maxproc;
354 		if (limp->rlim_max > maxproc)
355 			limp->rlim_max = maxproc;
356 		break;
357 	}
358 	*alimp = *limp;
359 	return (0);
360 }
361 
362 /* ARGSUSED */
363 int
364 sys_getrlimit(l, v, retval)
365 	struct lwp *l;
366 	void *v;
367 	register_t *retval;
368 {
369 	struct sys_getrlimit_args /* {
370 		syscallarg(int) which;
371 		syscallarg(struct rlimit *) rlp;
372 	} */ *uap = v;
373 	struct proc *p = l->l_proc;
374 	int which = SCARG(uap, which);
375 
376 	if ((u_int)which >= RLIM_NLIMITS)
377 		return (EINVAL);
378 	return (copyout(&p->p_rlimit[which], SCARG(uap, rlp),
379 	    sizeof(struct rlimit)));
380 }
381 
382 /*
383  * Transform the running time and tick information in proc p into user,
384  * system, and interrupt time usage.
385  */
386 void
387 calcru(p, up, sp, ip)
388 	struct proc *p;
389 	struct timeval *up;
390 	struct timeval *sp;
391 	struct timeval *ip;
392 {
393 	u_quad_t u, st, ut, it, tot;
394 	unsigned long sec;
395 	long usec;
396 	int s;
397 	struct timeval tv;
398 	struct lwp *l;
399 
400 	s = splstatclock();
401 	st = p->p_sticks;
402 	ut = p->p_uticks;
403 	it = p->p_iticks;
404 	splx(s);
405 
406 	sec = p->p_rtime.tv_sec;
407 	usec = p->p_rtime.tv_usec;
408 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
409 		if (l->l_stat == LSONPROC) {
410 			struct schedstate_percpu *spc;
411 
412 			KDASSERT(l->l_cpu != NULL);
413 			spc = &l->l_cpu->ci_schedstate;
414 
415 			/*
416 			 * Adjust for the current time slice.  This is
417 			 * actually fairly important since the error
418 			 * here is on the order of a time quantum,
419 			 * which is much greater than the sampling
420 			 * error.
421 			 */
422 			microtime(&tv);
423 			sec += tv.tv_sec - spc->spc_runtime.tv_sec;
424 			usec += tv.tv_usec - spc->spc_runtime.tv_usec;
425 		}
426 	}
427 
428 	tot = st + ut + it;
429 	u = sec * 1000000ull + usec;
430 
431 	if (tot == 0) {
432 		/* No ticks, so can't use to share time out, split 50-50 */
433 		st = ut = u / 2;
434 	} else {
435 		st = (u * st) / tot;
436 		ut = (u * ut) / tot;
437 	}
438 	sp->tv_sec = st / 1000000;
439 	sp->tv_usec = st % 1000000;
440 	up->tv_sec = ut / 1000000;
441 	up->tv_usec = ut % 1000000;
442 	if (ip != NULL) {
443 		if (it != 0)
444 			it = (u * it) / tot;
445 		ip->tv_sec = it / 1000000;
446 		ip->tv_usec = it % 1000000;
447 	}
448 }
449 
450 /* ARGSUSED */
451 int
452 sys_getrusage(l, v, retval)
453 	struct lwp *l;
454 	void *v;
455 	register_t *retval;
456 {
457 	struct sys_getrusage_args /* {
458 		syscallarg(int) who;
459 		syscallarg(struct rusage *) rusage;
460 	} */ *uap = v;
461 	struct rusage *rup;
462 	struct proc *p = l->l_proc;
463 
464 	switch (SCARG(uap, who)) {
465 
466 	case RUSAGE_SELF:
467 		rup = &p->p_stats->p_ru;
468 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
469 		break;
470 
471 	case RUSAGE_CHILDREN:
472 		rup = &p->p_stats->p_cru;
473 		break;
474 
475 	default:
476 		return (EINVAL);
477 	}
478 	return (copyout(rup, SCARG(uap, rusage), sizeof(struct rusage)));
479 }
480 
481 void
482 ruadd(ru, ru2)
483 	struct rusage *ru, *ru2;
484 {
485 	long *ip, *ip2;
486 	int i;
487 
488 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
489 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
490 	if (ru->ru_maxrss < ru2->ru_maxrss)
491 		ru->ru_maxrss = ru2->ru_maxrss;
492 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
493 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
494 		*ip++ += *ip2++;
495 }
496 
497 /*
498  * Make a copy of the plimit structure.
499  * We share these structures copy-on-write after fork,
500  * and copy when a limit is changed.
501  */
502 struct plimit *
503 limcopy(lim)
504 	struct plimit *lim;
505 {
506 	struct plimit *newlim;
507 	size_t l = 0;
508 
509 	simple_lock(&lim->p_slock);
510 	if (lim->pl_corename != defcorename)
511 		l = strlen(lim->pl_corename) + 1;
512 	simple_unlock(&lim->p_slock);
513 
514 	newlim = pool_get(&plimit_pool, PR_WAITOK);
515 	simple_lock_init(&newlim->p_slock);
516 	newlim->p_lflags = 0;
517 	newlim->p_refcnt = 1;
518 	newlim->pl_corename = (l != 0)
519 		? malloc(l, M_TEMP, M_WAITOK)
520 		: defcorename;
521 
522 	simple_lock(&lim->p_slock);
523 	memcpy(newlim->pl_rlimit, lim->pl_rlimit,
524 	    sizeof(struct rlimit) * RLIM_NLIMITS);
525 
526 	if (l != 0)
527 		strlcpy(newlim->pl_corename, lim->pl_corename, l);
528 	simple_unlock(&lim->p_slock);
529 
530 	return (newlim);
531 }
532 
533 void
534 limfree(lim)
535 	struct plimit *lim;
536 {
537 	int n;
538 
539 	simple_lock(&lim->p_slock);
540 	n = --lim->p_refcnt;
541 	simple_unlock(&lim->p_slock);
542 	if (n > 0)
543 		return;
544 #ifdef DIAGNOSTIC
545 	if (n < 0)
546 		panic("limfree");
547 #endif
548 	if (lim->pl_corename != defcorename)
549 		free(lim->pl_corename, M_TEMP);
550 	pool_put(&plimit_pool, lim);
551 }
552 
553 struct pstats *
554 pstatscopy(ps)
555 	struct pstats *ps;
556 {
557 
558 	struct pstats *newps;
559 
560 	newps = pool_get(&pstats_pool, PR_WAITOK);
561 
562 	memset(&newps->pstat_startzero, 0,
563 	(unsigned) ((caddr_t)&newps->pstat_endzero -
564 		    (caddr_t)&newps->pstat_startzero));
565 	memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy,
566 	((caddr_t)&newps->pstat_endcopy -
567 	 (caddr_t)&newps->pstat_startcopy));
568 
569 	return (newps);
570 
571 }
572 
573 void
574 pstatsfree(ps)
575 	struct pstats *ps;
576 {
577 
578 	pool_put(&pstats_pool, ps);
579 }
580 
581 /*
582  * sysctl interface in five parts
583  */
584 
585 /*
586  * a routine for sysctl proc subtree helpers that need to pick a valid
587  * process by pid.
588  */
589 static int
590 sysctl_proc_findproc(struct proc *p, struct proc **p2, pid_t pid)
591 {
592 	struct proc *ptmp;
593 	int i, error = 0;
594 
595 	if (pid == PROC_CURPROC)
596 		ptmp = p;
597 	else if ((ptmp = pfind(pid)) == NULL)
598 		error = ESRCH;
599 	else {
600 		/*
601 		 * suid proc of ours or proc not ours
602 		 */
603 		if (p->p_cred->p_ruid != ptmp->p_cred->p_ruid ||
604 		    p->p_cred->p_ruid != ptmp->p_cred->p_svuid)
605 			error = suser(p->p_ucred, &p->p_acflag);
606 
607 		/*
608 		 * sgid proc has sgid back to us temporarily
609 		 */
610 		else if (ptmp->p_cred->p_rgid != ptmp->p_cred->p_svgid)
611 			error = suser(p->p_ucred, &p->p_acflag);
612 
613 		/*
614 		 * our rgid must be in target's group list (ie,
615 		 * sub-processes started by a sgid process)
616 		 */
617 		else {
618 			for (i = 0; i < p->p_ucred->cr_ngroups; i++) {
619 				if (p->p_ucred->cr_groups[i] ==
620 				    ptmp->p_cred->p_rgid)
621 					break;
622 			}
623 			if (i == p->p_ucred->cr_ngroups)
624 				error = suser(p->p_ucred, &p->p_acflag);
625 		}
626 	}
627 
628 	*p2 = ptmp;
629 	return (error);
630 }
631 
632 /*
633  * sysctl helper routine for setting a process's specific corefile
634  * name.  picks the process based on the given pid and checks the
635  * correctness of the new value.
636  */
637 static int
638 sysctl_proc_corename(SYSCTLFN_ARGS)
639 {
640 	struct proc *ptmp, *p;
641 	struct plimit *lim;
642 	int error = 0, len;
643 	char cname[MAXPATHLEN], *tmp;
644 	struct sysctlnode node;
645 
646 	/*
647 	 * is this all correct?
648 	 */
649 	if (namelen != 0)
650 		return (EINVAL);
651 	if (name[-1] != PROC_PID_CORENAME)
652 		return (EINVAL);
653 
654 	/*
655 	 * whom are we tweaking?
656 	 */
657 	p = l->l_proc;
658 	error = sysctl_proc_findproc(p, &ptmp, (pid_t)name[-2]);
659 	if (error)
660 		return (error);
661 
662 	/*
663 	 * let them modify a temporary copy of the core name
664 	 */
665 	node = *rnode;
666 	strlcpy(cname, ptmp->p_limit->pl_corename, sizeof(cname));
667 	node.sysctl_data = cname;
668 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
669 
670 	/*
671 	 * if that failed, or they have nothing new to say, or we've
672 	 * heard it before...
673 	 */
674 	if (error || newp == NULL ||
675 	    strcmp(cname, ptmp->p_limit->pl_corename) == 0)
676 		return (error);
677 
678 	/*
679 	 * no error yet and cname now has the new core name in it.
680 	 * let's see if it looks acceptable.  it must be either "core"
681 	 * or end in ".core" or "/core".
682 	 */
683 	len = strlen(cname);
684 	if (len < 4)
685 		return (EINVAL);
686 	if (strcmp(cname + len - 4, "core") != 0)
687 		return (EINVAL);
688 	if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.')
689 		return (EINVAL);
690 
691 	/*
692 	 * hmm...looks good.  now...where do we put it?
693 	 */
694 	tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL);
695 	if (tmp == NULL)
696 		return (ENOMEM);
697 	strlcpy(tmp, cname, len + 1);
698 
699 	lim = ptmp->p_limit;
700 	if (lim->p_refcnt > 1 && (lim->p_lflags & PL_SHAREMOD) == 0) {
701 		ptmp->p_limit = limcopy(lim);
702 		limfree(lim);
703 		lim = ptmp->p_limit;
704 	}
705 	if (lim->pl_corename != defcorename)
706 		free(lim->pl_corename, M_TEMP);
707 	lim->pl_corename = tmp;
708 
709 	return (error);
710 }
711 
712 /*
713  * sysctl helper routine for checking/setting a process's stop flags,
714  * one for fork and one for exec.
715  */
716 static int
717 sysctl_proc_stop(SYSCTLFN_ARGS)
718 {
719 	struct proc *p, *ptmp;
720 	int i, f, error = 0;
721 	struct sysctlnode node;
722 
723 	if (namelen != 0)
724 		return (EINVAL);
725 
726 	p = l->l_proc;
727 	error = sysctl_proc_findproc(p, &ptmp, (pid_t)name[-2]);
728 	if (error)
729 		return (error);
730 
731 	switch (rnode->sysctl_num) {
732 	case PROC_PID_STOPFORK:
733 		f = P_STOPFORK;
734 		break;
735 	case PROC_PID_STOPEXEC:
736 		f = P_STOPEXEC;
737 		break;
738 	case PROC_PID_STOPEXIT:
739 		f = P_STOPEXIT;
740 		break;
741 	default:
742 		return (EINVAL);
743 	}
744 
745 	i = (ptmp->p_flag & f) ? 1 : 0;
746 	node = *rnode;
747 	node.sysctl_data = &i;
748 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
749 	if (error || newp == NULL)
750 		return (error);
751 
752 	if (i)
753 		ptmp->p_flag |= f;
754 	else
755 		ptmp->p_flag &= ~f;
756 
757 	return (0);
758 }
759 
760 /*
761  * sysctl helper routine for a process's rlimits as exposed by sysctl.
762  */
763 static int
764 sysctl_proc_plimit(SYSCTLFN_ARGS)
765 {
766 	struct proc *ptmp, *p;
767 	u_int limitno;
768 	int which, error = 0;
769         struct rlimit alim;
770 	struct sysctlnode node;
771 
772 	if (namelen != 0)
773 		return (EINVAL);
774 
775 	which = name[-1];
776 	if (which != PROC_PID_LIMIT_TYPE_SOFT &&
777 	    which != PROC_PID_LIMIT_TYPE_HARD)
778 		return (EINVAL);
779 
780 	limitno = name[-2] - 1;
781 	if (limitno >= RLIM_NLIMITS)
782 		return (EINVAL);
783 
784 	if (name[-3] != PROC_PID_LIMIT)
785 		return (EINVAL);
786 
787 	p = l->l_proc;
788 	error = sysctl_proc_findproc(p, &ptmp, (pid_t)name[-4]);
789 	if (error)
790 		return (error);
791 
792 	node = *rnode;
793 	memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim));
794 	if (which == PROC_PID_LIMIT_TYPE_HARD)
795 		node.sysctl_data = &alim.rlim_max;
796 	else
797 		node.sysctl_data = &alim.rlim_cur;
798 
799 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
800 	if (error || newp == NULL)
801 		return (error);
802 
803 	return (dosetrlimit(ptmp, p->p_cred, limitno, &alim));
804 }
805 
806 /*
807  * and finally, the actually glue that sticks it to the tree
808  */
809 SYSCTL_SETUP(sysctl_proc_setup, "sysctl proc subtree setup")
810 {
811 
812 	sysctl_createv(clog, 0, NULL, NULL,
813 		       CTLFLAG_PERMANENT,
814 		       CTLTYPE_NODE, "proc", NULL,
815 		       NULL, 0, NULL, 0,
816 		       CTL_PROC, CTL_EOL);
817 	sysctl_createv(clog, 0, NULL, NULL,
818 		       CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER,
819 		       CTLTYPE_NODE, "curproc",
820 		       SYSCTL_DESCR("Per-process settings"),
821 		       NULL, 0, NULL, 0,
822 		       CTL_PROC, PROC_CURPROC, CTL_EOL);
823 
824 	sysctl_createv(clog, 0, NULL, NULL,
825 		       CTLFLAG_PERMANENT|CTLFLAG_READONLY2|CTLFLAG_ANYWRITE,
826 		       CTLTYPE_STRING, "corename",
827 		       SYSCTL_DESCR("Core file name"),
828 		       sysctl_proc_corename, 0, NULL, MAXPATHLEN,
829 		       CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL);
830 	sysctl_createv(clog, 0, NULL, NULL,
831 		       CTLFLAG_PERMANENT,
832 		       CTLTYPE_NODE, "rlimit",
833 		       SYSCTL_DESCR("Process limits"),
834 		       NULL, 0, NULL, 0,
835 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL);
836 
837 #define create_proc_plimit(s, n) do {					\
838 	sysctl_createv(clog, 0, NULL, NULL,				\
839 		       CTLFLAG_PERMANENT,				\
840 		       CTLTYPE_NODE, s,					\
841 		       SYSCTL_DESCR("Process " s " limits"),		\
842 		       NULL, 0, NULL, 0,				\
843 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
844 		       CTL_EOL);					\
845 	sysctl_createv(clog, 0, NULL, NULL,				\
846 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \
847 		       CTLTYPE_QUAD, "soft",				\
848 		       SYSCTL_DESCR("Process soft " s " limit"),	\
849 		       sysctl_proc_plimit, 0, NULL, 0,			\
850 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
851 		       PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL);		\
852 	sysctl_createv(clog, 0, NULL, NULL,				\
853 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \
854 		       CTLTYPE_QUAD, "hard",				\
855 		       SYSCTL_DESCR("Process hard " s " limit"),	\
856 		       sysctl_proc_plimit, 0, NULL, 0,			\
857 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
858 		       PROC_PID_LIMIT_TYPE_HARD, CTL_EOL);		\
859 	} while (0/*CONSTCOND*/)
860 
861 	create_proc_plimit("cputime",		PROC_PID_LIMIT_CPU);
862 	create_proc_plimit("filesize",		PROC_PID_LIMIT_FSIZE);
863 	create_proc_plimit("datasize",		PROC_PID_LIMIT_DATA);
864 	create_proc_plimit("stacksize",		PROC_PID_LIMIT_STACK);
865 	create_proc_plimit("coredumpsize",	PROC_PID_LIMIT_CORE);
866 	create_proc_plimit("memoryuse",		PROC_PID_LIMIT_RSS);
867 	create_proc_plimit("memorylocked",	PROC_PID_LIMIT_MEMLOCK);
868 	create_proc_plimit("maxproc",		PROC_PID_LIMIT_NPROC);
869 	create_proc_plimit("descriptors",	PROC_PID_LIMIT_NOFILE);
870 	create_proc_plimit("sbsize",		PROC_PID_LIMIT_SBSIZE);
871 
872 #undef create_proc_plimit
873 
874 	sysctl_createv(clog, 0, NULL, NULL,
875 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
876 		       CTLTYPE_INT, "stopfork",
877 		       SYSCTL_DESCR("Stop process at fork(2)"),
878 		       sysctl_proc_stop, 0, NULL, 0,
879 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL);
880 	sysctl_createv(clog, 0, NULL, NULL,
881 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
882 		       CTLTYPE_INT, "stopexec",
883 		       SYSCTL_DESCR("Stop process at execve(2)"),
884 		       sysctl_proc_stop, 0, NULL, 0,
885 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL);
886 	sysctl_createv(clog, 0, NULL, NULL,
887 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
888 		       CTLTYPE_INT, "stopexit",
889 		       SYSCTL_DESCR("Stop process before completing exit"),
890 		       sysctl_proc_stop, 0, NULL, 0,
891 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL);
892 }
893 
894 static struct uidinfo *
895 getuidinfo(uid_t uid)
896 {
897 	struct uidinfo *uip;
898 	struct uihashhead *uipp;
899 
900 	uipp = UIHASH(uid);
901 
902 	LIST_FOREACH(uip, uipp, ui_hash)
903 		if (uip->ui_uid == uid)
904 			return uip;
905 	return NULL;
906 }
907 
908 static void
909 freeuidinfo(struct uidinfo *uip)
910 {
911 	LIST_REMOVE(uip, ui_hash);
912 	FREE(uip, M_PROC);
913 }
914 
915 static struct uidinfo *
916 allocuidinfo(uid_t uid)
917 {
918 	struct uidinfo *uip;
919 	struct uihashhead *uipp;
920 
921 	uipp = UIHASH(uid);
922 	MALLOC(uip, struct uidinfo *, sizeof(*uip), M_PROC, M_WAITOK);
923 	LIST_INSERT_HEAD(uipp, uip, ui_hash);
924 	uip->ui_uid = uid;
925 	uip->ui_proccnt = 0;
926 	uip->ui_sbsize = 0;
927 	return uip;
928 }
929 
930 /*
931  * Change the count associated with number of processes
932  * a given user is using.
933  */
934 int
935 chgproccnt(uid_t uid, int diff)
936 {
937 	struct uidinfo *uip;
938 
939 	if (diff == 0)
940 		return 0;
941 
942 	if ((uip = getuidinfo(uid)) != NULL) {
943 		uip->ui_proccnt += diff;
944 		KASSERT(uip->ui_proccnt >= 0);
945 		if (uip->ui_proccnt > 0)
946 			return uip->ui_proccnt;
947 		else {
948 			if (uip->ui_sbsize == 0)
949 				freeuidinfo(uip);
950 			return 0;
951 		}
952 	} else {
953 		if (diff < 0)
954 			panic("chgproccnt: lost user %lu", (unsigned long)uid);
955 		uip = allocuidinfo(uid);
956 		uip->ui_proccnt = diff;
957 		return uip->ui_proccnt;
958 	}
959 }
960 
961 int
962 chgsbsize(uid_t uid, u_long *hiwat, u_long to, rlim_t max)
963 {
964 	*hiwat = to;
965 	return 1;
966 #ifdef notyet
967 	struct uidinfo *uip;
968 	rlim_t nsb;
969 	int rv = 0;
970 
971 	if ((uip = getuidinfo(uid)) == NULL)
972 		uip = allocuidinfo(uid);
973 	nsb = uip->ui_sbsize + to - *hiwat;
974 	if (to > *hiwat && nsb > max)
975 		goto done;
976 	*hiwat = to;
977 	uip->ui_sbsize = nsb;
978 	rv = 1;
979 	KASSERT(uip->ui_sbsize >= 0);
980 done:
981 	if (uip->ui_sbsize == 0 && uip->ui_proccnt == 0)
982 		freeuidinfo(uip);
983 	return rv;
984 #endif
985 }
986