xref: /netbsd-src/sys/kern/kern_resource.c (revision 20e85ad185ab16980f1219a557c42e057edb42ea)
1 /*	$NetBSD: kern_resource.c,v 1.93 2005/03/29 18:18:06 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 1.93 2005/03/29 18:18:06 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/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 struct simplelock uihashtbl_slock = SIMPLELOCK_INITIALIZER;
68 
69 
70 /*
71  * Resource controls and accounting.
72  */
73 
74 int
75 sys_getpriority(l, v, retval)
76 	struct lwp *l;
77 	void *v;
78 	register_t *retval;
79 {
80 	struct sys_getpriority_args /* {
81 		syscallarg(int) which;
82 		syscallarg(id_t) who;
83 	} */ *uap = v;
84 	struct proc *curp = l->l_proc, *p;
85 	int low = NZERO + PRIO_MAX + 1;
86 
87 	switch (SCARG(uap, which)) {
88 
89 	case PRIO_PROCESS:
90 		if (SCARG(uap, who) == 0)
91 			p = curp;
92 		else
93 			p = pfind(SCARG(uap, who));
94 		if (p == 0)
95 			break;
96 		low = p->p_nice;
97 		break;
98 
99 	case PRIO_PGRP: {
100 		struct pgrp *pg;
101 
102 		if (SCARG(uap, who) == 0)
103 			pg = curp->p_pgrp;
104 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
105 			break;
106 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
107 			if (p->p_nice < low)
108 				low = p->p_nice;
109 		}
110 		break;
111 	}
112 
113 	case PRIO_USER:
114 		if (SCARG(uap, who) == 0)
115 			SCARG(uap, who) = curp->p_ucred->cr_uid;
116 		proclist_lock_read();
117 		PROCLIST_FOREACH(p, &allproc) {
118 			if (p->p_ucred->cr_uid == (uid_t) SCARG(uap, who) &&
119 			    p->p_nice < low)
120 				low = p->p_nice;
121 		}
122 		proclist_unlock_read();
123 		break;
124 
125 	default:
126 		return (EINVAL);
127 	}
128 	if (low == NZERO + PRIO_MAX + 1)
129 		return (ESRCH);
130 	*retval = low - NZERO;
131 	return (0);
132 }
133 
134 /* ARGSUSED */
135 int
136 sys_setpriority(l, v, retval)
137 	struct lwp *l;
138 	void *v;
139 	register_t *retval;
140 {
141 	struct sys_setpriority_args /* {
142 		syscallarg(int) which;
143 		syscallarg(id_t) who;
144 		syscallarg(int) prio;
145 	} */ *uap = v;
146 	struct proc *curp = l->l_proc, *p;
147 	int found = 0, error = 0;
148 
149 	switch (SCARG(uap, which)) {
150 
151 	case PRIO_PROCESS:
152 		if (SCARG(uap, who) == 0)
153 			p = curp;
154 		else
155 			p = pfind(SCARG(uap, who));
156 		if (p == 0)
157 			break;
158 		error = donice(curp, p, SCARG(uap, prio));
159 		found++;
160 		break;
161 
162 	case PRIO_PGRP: {
163 		struct pgrp *pg;
164 
165 		if (SCARG(uap, who) == 0)
166 			pg = curp->p_pgrp;
167 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
168 			break;
169 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
170 			error = donice(curp, p, SCARG(uap, prio));
171 			found++;
172 		}
173 		break;
174 	}
175 
176 	case PRIO_USER:
177 		if (SCARG(uap, who) == 0)
178 			SCARG(uap, who) = curp->p_ucred->cr_uid;
179 		proclist_lock_read();
180 		PROCLIST_FOREACH(p, &allproc) {
181 			if (p->p_ucred->cr_uid == (uid_t) SCARG(uap, who)) {
182 				error = donice(curp, p, SCARG(uap, prio));
183 				found++;
184 			}
185 		}
186 		proclist_unlock_read();
187 		break;
188 
189 	default:
190 		return (EINVAL);
191 	}
192 	if (found == 0)
193 		return (ESRCH);
194 	return (error);
195 }
196 
197 int
198 donice(curp, chgp, n)
199 	struct proc *curp, *chgp;
200 	int n;
201 {
202 	struct pcred *pcred = curp->p_cred;
203 	int s;
204 
205 	if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
206 	    pcred->pc_ucred->cr_uid != chgp->p_ucred->cr_uid &&
207 	    pcred->p_ruid != chgp->p_ucred->cr_uid)
208 		return (EPERM);
209 	if (n > PRIO_MAX)
210 		n = PRIO_MAX;
211 	if (n < PRIO_MIN)
212 		n = PRIO_MIN;
213 	n += NZERO;
214 	if (n < chgp->p_nice && suser(pcred->pc_ucred, &curp->p_acflag))
215 		return (EACCES);
216 	chgp->p_nice = n;
217 	SCHED_LOCK(s);
218 	(void)resetprocpriority(chgp);
219 	SCHED_UNLOCK(s);
220 	return (0);
221 }
222 
223 /* ARGSUSED */
224 int
225 sys_setrlimit(l, v, retval)
226 	struct lwp *l;
227 	void *v;
228 	register_t *retval;
229 {
230 	struct sys_setrlimit_args /* {
231 		syscallarg(int) which;
232 		syscallarg(const struct rlimit *) rlp;
233 	} */ *uap = v;
234 	struct proc *p = l->l_proc;
235 	int which = SCARG(uap, which);
236 	struct rlimit alim;
237 	int error;
238 
239 	error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit));
240 	if (error)
241 		return (error);
242 	return (dosetrlimit(p, p->p_cred, which, &alim));
243 }
244 
245 int
246 dosetrlimit(p, cred, which, limp)
247 	struct proc *p;
248 	struct  pcred *cred;
249 	int which;
250 	struct rlimit *limp;
251 {
252 	struct rlimit *alimp;
253 	struct plimit *oldplim;
254 	int error;
255 
256 	if ((u_int)which >= RLIM_NLIMITS)
257 		return (EINVAL);
258 
259 	if (limp->rlim_cur < 0 || limp->rlim_max < 0)
260 		return (EINVAL);
261 
262 	alimp = &p->p_rlimit[which];
263 	/* if we don't change the value, no need to limcopy() */
264 	if (limp->rlim_cur == alimp->rlim_cur &&
265 	    limp->rlim_max == alimp->rlim_max)
266 		return 0;
267 
268 	if (limp->rlim_cur > limp->rlim_max) {
269 		/*
270 		 * This is programming error. According to SUSv2, we should
271 		 * return error in this case.
272 		 */
273 		return (EINVAL);
274 	}
275 	if (limp->rlim_max > alimp->rlim_max
276 	    && (error = suser(cred->pc_ucred, &p->p_acflag)) != 0)
277 			return (error);
278 
279 	if (p->p_limit->p_refcnt > 1 &&
280 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
281 		p->p_limit = limcopy(oldplim = p->p_limit);
282 		limfree(oldplim);
283 		alimp = &p->p_rlimit[which];
284 	}
285 
286 	switch (which) {
287 
288 	case RLIMIT_DATA:
289 		if (limp->rlim_cur > maxdmap)
290 			limp->rlim_cur = maxdmap;
291 		if (limp->rlim_max > maxdmap)
292 			limp->rlim_max = maxdmap;
293 		break;
294 
295 	case RLIMIT_STACK:
296 		if (limp->rlim_cur > maxsmap)
297 			limp->rlim_cur = maxsmap;
298 		if (limp->rlim_max > maxsmap)
299 			limp->rlim_max = maxsmap;
300 
301 		/*
302 		 * Return EINVAL if the new stack size limit is lower than
303 		 * current usage. Otherwise, the process would get SIGSEGV the
304 		 * moment it would try to access anything on it's current stack.
305 		 * This conforms to SUSv2.
306 		 */
307 		if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE
308 		    || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE)
309 			return (EINVAL);
310 
311 		/*
312 		 * Stack is allocated to the max at exec time with
313 		 * only "rlim_cur" bytes accessible (In other words,
314 		 * allocates stack dividing two contiguous regions at
315 		 * "rlim_cur" bytes boundary).
316 		 *
317 		 * Since allocation is done in terms of page, roundup
318 		 * "rlim_cur" (otherwise, contiguous regions
319 		 * overlap).  If stack limit is going up make more
320 		 * accessible, if going down make inaccessible.
321 		 */
322 		limp->rlim_cur = round_page(limp->rlim_cur);
323 		if (limp->rlim_cur != alimp->rlim_cur) {
324 			vaddr_t addr;
325 			vsize_t size;
326 			vm_prot_t prot;
327 
328 			if (limp->rlim_cur > alimp->rlim_cur) {
329 				prot = VM_PROT_READ | VM_PROT_WRITE;
330 				size = limp->rlim_cur - alimp->rlim_cur;
331 				addr = (vaddr_t)p->p_vmspace->vm_minsaddr -
332 				    limp->rlim_cur;
333 			} else {
334 				prot = VM_PROT_NONE;
335 				size = alimp->rlim_cur - limp->rlim_cur;
336 				addr = (vaddr_t)p->p_vmspace->vm_minsaddr -
337 				     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 struct uidinfo *
895 uid_find(uid_t uid)
896 {
897 	struct uidinfo *uip;
898 	struct uidinfo *newuip = NULL;
899 	struct uihashhead *uipp;
900 
901 	uipp = UIHASH(uid);
902 
903 again:
904 	simple_lock(&uihashtbl_slock);
905 	LIST_FOREACH(uip, uipp, ui_hash)
906 		if (uip->ui_uid == uid) {
907 			simple_unlock(&uihashtbl_slock);
908 			if (newuip)
909 				free(newuip, M_PROC);
910 			return uip;
911 		}
912 
913 	if (newuip == NULL) {
914 		simple_unlock(&uihashtbl_slock);
915 		newuip = malloc(sizeof(*uip), M_PROC, M_WAITOK | M_ZERO);
916 		goto again;
917 	}
918 	uip = newuip;
919 
920 	LIST_INSERT_HEAD(uipp, uip, ui_hash);
921 	uip->ui_uid = uid;
922 	simple_unlock(&uihashtbl_slock);
923 
924 	return uip;
925 }
926 
927 /*
928  * Change the count associated with number of processes
929  * a given user is using.
930  */
931 int
932 chgproccnt(uid_t uid, int diff)
933 {
934 	struct uidinfo *uip;
935 
936 	if (diff == 0)
937 		return 0;
938 
939 	uip = uid_find(uid);
940 	uip->ui_proccnt += diff;
941 	KASSERT(uip->ui_proccnt >= 0);
942 	return uip->ui_proccnt;
943 }
944 
945 int
946 chgsbsize(uid_t uid, u_long *hiwat, u_long to, rlim_t max)
947 {
948 	struct uidinfo *uip;
949 	rlim_t nsb;
950 
951 	uip = uid_find(uid);
952 	nsb = uip->ui_sbsize + to - *hiwat;
953 	if (to > *hiwat && nsb > max)
954 		return 0;
955 	*hiwat = to;
956 	uip->ui_sbsize = nsb;
957 	KASSERT(uip->ui_sbsize >= 0);
958 	return 1;
959 }
960