xref: /netbsd-src/sys/kern/kern_resource.c (revision df0caa2637da0538ecdf6b878c4d08e684b43d8f)
1 /*	$NetBSD: kern_resource.c,v 1.98 2005/06/23 23:15:12 thorpej 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 1.98 2005/06/23 23:15:12 thorpej 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(struct lwp *l, void *v, register_t *retval)
76 {
77 	struct sys_getpriority_args /* {
78 		syscallarg(int) which;
79 		syscallarg(id_t) who;
80 	} */ *uap = v;
81 	struct proc *curp = l->l_proc, *p;
82 	int low = NZERO + PRIO_MAX + 1;
83 
84 	switch (SCARG(uap, which)) {
85 
86 	case PRIO_PROCESS:
87 		if (SCARG(uap, who) == 0)
88 			p = curp;
89 		else
90 			p = pfind(SCARG(uap, who));
91 		if (p == 0)
92 			break;
93 		low = p->p_nice;
94 		break;
95 
96 	case PRIO_PGRP: {
97 		struct pgrp *pg;
98 
99 		if (SCARG(uap, who) == 0)
100 			pg = curp->p_pgrp;
101 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
102 			break;
103 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
104 			if (p->p_nice < low)
105 				low = p->p_nice;
106 		}
107 		break;
108 	}
109 
110 	case PRIO_USER:
111 		if (SCARG(uap, who) == 0)
112 			SCARG(uap, who) = curp->p_ucred->cr_uid;
113 		proclist_lock_read();
114 		PROCLIST_FOREACH(p, &allproc) {
115 			if (p->p_ucred->cr_uid == (uid_t) SCARG(uap, who) &&
116 			    p->p_nice < low)
117 				low = p->p_nice;
118 		}
119 		proclist_unlock_read();
120 		break;
121 
122 	default:
123 		return (EINVAL);
124 	}
125 	if (low == NZERO + PRIO_MAX + 1)
126 		return (ESRCH);
127 	*retval = low - NZERO;
128 	return (0);
129 }
130 
131 /* ARGSUSED */
132 int
133 sys_setpriority(struct lwp *l, void *v, register_t *retval)
134 {
135 	struct sys_setpriority_args /* {
136 		syscallarg(int) which;
137 		syscallarg(id_t) who;
138 		syscallarg(int) prio;
139 	} */ *uap = v;
140 	struct proc *curp = l->l_proc, *p;
141 	int found = 0, error = 0;
142 
143 	switch (SCARG(uap, which)) {
144 
145 	case PRIO_PROCESS:
146 		if (SCARG(uap, who) == 0)
147 			p = curp;
148 		else
149 			p = pfind(SCARG(uap, who));
150 		if (p == 0)
151 			break;
152 		error = donice(curp, p, SCARG(uap, prio));
153 		found++;
154 		break;
155 
156 	case PRIO_PGRP: {
157 		struct pgrp *pg;
158 
159 		if (SCARG(uap, who) == 0)
160 			pg = curp->p_pgrp;
161 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
162 			break;
163 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
164 			error = donice(curp, p, SCARG(uap, prio));
165 			found++;
166 		}
167 		break;
168 	}
169 
170 	case PRIO_USER:
171 		if (SCARG(uap, who) == 0)
172 			SCARG(uap, who) = curp->p_ucred->cr_uid;
173 		proclist_lock_read();
174 		PROCLIST_FOREACH(p, &allproc) {
175 			if (p->p_ucred->cr_uid == (uid_t) SCARG(uap, who)) {
176 				error = donice(curp, p, SCARG(uap, prio));
177 				found++;
178 			}
179 		}
180 		proclist_unlock_read();
181 		break;
182 
183 	default:
184 		return (EINVAL);
185 	}
186 	if (found == 0)
187 		return (ESRCH);
188 	return (error);
189 }
190 
191 int
192 donice(struct proc *curp, struct proc *chgp, int n)
193 {
194 	struct pcred *pcred = curp->p_cred;
195 	int s;
196 
197 	if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
198 	    pcred->pc_ucred->cr_uid != chgp->p_ucred->cr_uid &&
199 	    pcred->p_ruid != chgp->p_ucred->cr_uid)
200 		return (EPERM);
201 	if (n > PRIO_MAX)
202 		n = PRIO_MAX;
203 	if (n < PRIO_MIN)
204 		n = PRIO_MIN;
205 	n += NZERO;
206 	if (n < chgp->p_nice && suser(pcred->pc_ucred, &curp->p_acflag))
207 		return (EACCES);
208 	chgp->p_nice = n;
209 	SCHED_LOCK(s);
210 	(void)resetprocpriority(chgp);
211 	SCHED_UNLOCK(s);
212 	return (0);
213 }
214 
215 /* ARGSUSED */
216 int
217 sys_setrlimit(struct lwp *l, void *v, register_t *retval)
218 {
219 	struct sys_setrlimit_args /* {
220 		syscallarg(int) which;
221 		syscallarg(const struct rlimit *) rlp;
222 	} */ *uap = v;
223 	struct proc *p = l->l_proc;
224 	int which = SCARG(uap, which);
225 	struct rlimit alim;
226 	int error;
227 
228 	error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit));
229 	if (error)
230 		return (error);
231 	return (dosetrlimit(p, p->p_cred, which, &alim));
232 }
233 
234 int
235 dosetrlimit(struct proc *p, struct pcred *cred, int which, struct rlimit *limp)
236 {
237 	struct rlimit *alimp;
238 	struct plimit *oldplim;
239 	int error;
240 
241 	if ((u_int)which >= RLIM_NLIMITS)
242 		return (EINVAL);
243 
244 	if (limp->rlim_cur < 0 || limp->rlim_max < 0)
245 		return (EINVAL);
246 
247 	alimp = &p->p_rlimit[which];
248 	/* if we don't change the value, no need to limcopy() */
249 	if (limp->rlim_cur == alimp->rlim_cur &&
250 	    limp->rlim_max == alimp->rlim_max)
251 		return 0;
252 
253 	if (limp->rlim_cur > limp->rlim_max) {
254 		/*
255 		 * This is programming error. According to SUSv2, we should
256 		 * return error in this case.
257 		 */
258 		return (EINVAL);
259 	}
260 	if (limp->rlim_max > alimp->rlim_max
261 	    && (error = suser(cred->pc_ucred, &p->p_acflag)) != 0)
262 			return (error);
263 
264 	if (p->p_limit->p_refcnt > 1 &&
265 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
266 		p->p_limit = limcopy(oldplim = p->p_limit);
267 		limfree(oldplim);
268 		alimp = &p->p_rlimit[which];
269 	}
270 
271 	switch (which) {
272 
273 	case RLIMIT_DATA:
274 		if (limp->rlim_cur > maxdmap)
275 			limp->rlim_cur = maxdmap;
276 		if (limp->rlim_max > maxdmap)
277 			limp->rlim_max = maxdmap;
278 		break;
279 
280 	case RLIMIT_STACK:
281 		if (limp->rlim_cur > maxsmap)
282 			limp->rlim_cur = maxsmap;
283 		if (limp->rlim_max > maxsmap)
284 			limp->rlim_max = maxsmap;
285 
286 		/*
287 		 * Return EINVAL if the new stack size limit is lower than
288 		 * current usage. Otherwise, the process would get SIGSEGV the
289 		 * moment it would try to access anything on it's current stack.
290 		 * This conforms to SUSv2.
291 		 */
292 		if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE
293 		    || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE)
294 			return (EINVAL);
295 
296 		/*
297 		 * Stack is allocated to the max at exec time with
298 		 * only "rlim_cur" bytes accessible (In other words,
299 		 * allocates stack dividing two contiguous regions at
300 		 * "rlim_cur" bytes boundary).
301 		 *
302 		 * Since allocation is done in terms of page, roundup
303 		 * "rlim_cur" (otherwise, contiguous regions
304 		 * overlap).  If stack limit is going up make more
305 		 * accessible, if going down make inaccessible.
306 		 */
307 		limp->rlim_cur = round_page(limp->rlim_cur);
308 		if (limp->rlim_cur != alimp->rlim_cur) {
309 			vaddr_t addr;
310 			vsize_t size;
311 			vm_prot_t prot;
312 
313 			if (limp->rlim_cur > alimp->rlim_cur) {
314 				prot = VM_PROT_READ | VM_PROT_WRITE;
315 				size = limp->rlim_cur - alimp->rlim_cur;
316 				addr = (vaddr_t)p->p_vmspace->vm_minsaddr -
317 				    limp->rlim_cur;
318 			} else {
319 				prot = VM_PROT_NONE;
320 				size = alimp->rlim_cur - limp->rlim_cur;
321 				addr = (vaddr_t)p->p_vmspace->vm_minsaddr -
322 				     alimp->rlim_cur;
323 			}
324 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
325 					      addr, addr+size, prot, FALSE);
326 		}
327 		break;
328 
329 	case RLIMIT_NOFILE:
330 		if (limp->rlim_cur > maxfiles)
331 			limp->rlim_cur = maxfiles;
332 		if (limp->rlim_max > maxfiles)
333 			limp->rlim_max = maxfiles;
334 		break;
335 
336 	case RLIMIT_NPROC:
337 		if (limp->rlim_cur > maxproc)
338 			limp->rlim_cur = maxproc;
339 		if (limp->rlim_max > maxproc)
340 			limp->rlim_max = maxproc;
341 		break;
342 	}
343 	*alimp = *limp;
344 	return (0);
345 }
346 
347 /* ARGSUSED */
348 int
349 sys_getrlimit(struct lwp *l, void *v, register_t *retval)
350 {
351 	struct sys_getrlimit_args /* {
352 		syscallarg(int) which;
353 		syscallarg(struct rlimit *) rlp;
354 	} */ *uap = v;
355 	struct proc *p = l->l_proc;
356 	int which = SCARG(uap, which);
357 
358 	if ((u_int)which >= RLIM_NLIMITS)
359 		return (EINVAL);
360 	return (copyout(&p->p_rlimit[which], SCARG(uap, rlp),
361 	    sizeof(struct rlimit)));
362 }
363 
364 /*
365  * Transform the running time and tick information in proc p into user,
366  * system, and interrupt time usage.
367  */
368 void
369 calcru(struct proc *p, struct timeval *up, struct timeval *sp,
370     struct timeval *ip)
371 {
372 	u_quad_t u, st, ut, it, tot;
373 	unsigned long sec;
374 	long usec;
375 	int s;
376 	struct timeval tv;
377 	struct lwp *l;
378 
379 	s = splstatclock();
380 	st = p->p_sticks;
381 	ut = p->p_uticks;
382 	it = p->p_iticks;
383 	splx(s);
384 
385 	sec = p->p_rtime.tv_sec;
386 	usec = p->p_rtime.tv_usec;
387 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
388 		if (l->l_stat == LSONPROC) {
389 			struct schedstate_percpu *spc;
390 
391 			KDASSERT(l->l_cpu != NULL);
392 			spc = &l->l_cpu->ci_schedstate;
393 
394 			/*
395 			 * Adjust for the current time slice.  This is
396 			 * actually fairly important since the error
397 			 * here is on the order of a time quantum,
398 			 * which is much greater than the sampling
399 			 * error.
400 			 */
401 			microtime(&tv);
402 			sec += tv.tv_sec - spc->spc_runtime.tv_sec;
403 			usec += tv.tv_usec - spc->spc_runtime.tv_usec;
404 		}
405 	}
406 
407 	tot = st + ut + it;
408 	u = sec * 1000000ull + usec;
409 
410 	if (tot == 0) {
411 		/* No ticks, so can't use to share time out, split 50-50 */
412 		st = ut = u / 2;
413 	} else {
414 		st = (u * st) / tot;
415 		ut = (u * ut) / tot;
416 	}
417 	sp->tv_sec = st / 1000000;
418 	sp->tv_usec = st % 1000000;
419 	up->tv_sec = ut / 1000000;
420 	up->tv_usec = ut % 1000000;
421 	if (ip != NULL) {
422 		if (it != 0)
423 			it = (u * it) / tot;
424 		ip->tv_sec = it / 1000000;
425 		ip->tv_usec = it % 1000000;
426 	}
427 }
428 
429 /* ARGSUSED */
430 int
431 sys_getrusage(struct lwp *l, void *v, register_t *retval)
432 {
433 	struct sys_getrusage_args /* {
434 		syscallarg(int) who;
435 		syscallarg(struct rusage *) rusage;
436 	} */ *uap = v;
437 	struct rusage *rup;
438 	struct proc *p = l->l_proc;
439 
440 	switch (SCARG(uap, who)) {
441 
442 	case RUSAGE_SELF:
443 		rup = &p->p_stats->p_ru;
444 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
445 		break;
446 
447 	case RUSAGE_CHILDREN:
448 		rup = &p->p_stats->p_cru;
449 		break;
450 
451 	default:
452 		return (EINVAL);
453 	}
454 	return (copyout(rup, SCARG(uap, rusage), sizeof(struct rusage)));
455 }
456 
457 void
458 ruadd(struct rusage *ru, struct rusage *ru2)
459 {
460 	long *ip, *ip2;
461 	int i;
462 
463 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
464 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
465 	if (ru->ru_maxrss < ru2->ru_maxrss)
466 		ru->ru_maxrss = ru2->ru_maxrss;
467 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
468 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
469 		*ip++ += *ip2++;
470 }
471 
472 /*
473  * Make a copy of the plimit structure.
474  * We share these structures copy-on-write after fork,
475  * and copy when a limit is changed.
476  */
477 struct plimit *
478 limcopy(struct plimit *lim)
479 {
480 	struct plimit *newlim;
481 	size_t l = 0;
482 
483 	simple_lock(&lim->p_slock);
484 	if (lim->pl_corename != defcorename)
485 		l = strlen(lim->pl_corename) + 1;
486 	simple_unlock(&lim->p_slock);
487 
488 	newlim = pool_get(&plimit_pool, PR_WAITOK);
489 	simple_lock_init(&newlim->p_slock);
490 	newlim->p_lflags = 0;
491 	newlim->p_refcnt = 1;
492 	newlim->pl_corename = (l != 0)
493 		? malloc(l, M_TEMP, M_WAITOK)
494 		: defcorename;
495 
496 	simple_lock(&lim->p_slock);
497 	memcpy(newlim->pl_rlimit, lim->pl_rlimit,
498 	    sizeof(struct rlimit) * RLIM_NLIMITS);
499 
500 	if (l != 0)
501 		strlcpy(newlim->pl_corename, lim->pl_corename, l);
502 	simple_unlock(&lim->p_slock);
503 
504 	return (newlim);
505 }
506 
507 void
508 limfree(struct plimit *lim)
509 {
510 	int n;
511 
512 	simple_lock(&lim->p_slock);
513 	n = --lim->p_refcnt;
514 	simple_unlock(&lim->p_slock);
515 	if (n > 0)
516 		return;
517 #ifdef DIAGNOSTIC
518 	if (n < 0)
519 		panic("limfree");
520 #endif
521 	if (lim->pl_corename != defcorename)
522 		free(lim->pl_corename, M_TEMP);
523 	pool_put(&plimit_pool, lim);
524 }
525 
526 struct pstats *
527 pstatscopy(struct pstats *ps)
528 {
529 
530 	struct pstats *newps;
531 
532 	newps = pool_get(&pstats_pool, PR_WAITOK);
533 
534 	memset(&newps->pstat_startzero, 0,
535 	(unsigned) ((caddr_t)&newps->pstat_endzero -
536 		    (caddr_t)&newps->pstat_startzero));
537 	memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy,
538 	((caddr_t)&newps->pstat_endcopy -
539 	 (caddr_t)&newps->pstat_startcopy));
540 
541 	return (newps);
542 
543 }
544 
545 void
546 pstatsfree(struct pstats *ps)
547 {
548 
549 	pool_put(&pstats_pool, ps);
550 }
551 
552 /*
553  * sysctl interface in five parts
554  */
555 
556 /*
557  * a routine for sysctl proc subtree helpers that need to pick a valid
558  * process by pid.
559  */
560 static int
561 sysctl_proc_findproc(struct proc *p, struct proc **p2, pid_t pid)
562 {
563 	struct proc *ptmp;
564 	int i, error = 0;
565 
566 	if (pid == PROC_CURPROC)
567 		ptmp = p;
568 	else if ((ptmp = pfind(pid)) == NULL)
569 		error = ESRCH;
570 	else {
571 		/*
572 		 * suid proc of ours or proc not ours
573 		 */
574 		if (p->p_cred->p_ruid != ptmp->p_cred->p_ruid ||
575 		    p->p_cred->p_ruid != ptmp->p_cred->p_svuid)
576 			error = suser(p->p_ucred, &p->p_acflag);
577 
578 		/*
579 		 * sgid proc has sgid back to us temporarily
580 		 */
581 		else if (ptmp->p_cred->p_rgid != ptmp->p_cred->p_svgid)
582 			error = suser(p->p_ucred, &p->p_acflag);
583 
584 		/*
585 		 * our rgid must be in target's group list (ie,
586 		 * sub-processes started by a sgid process)
587 		 */
588 		else {
589 			for (i = 0; i < p->p_ucred->cr_ngroups; i++) {
590 				if (p->p_ucred->cr_groups[i] ==
591 				    ptmp->p_cred->p_rgid)
592 					break;
593 			}
594 			if (i == p->p_ucred->cr_ngroups)
595 				error = suser(p->p_ucred, &p->p_acflag);
596 		}
597 	}
598 
599 	*p2 = ptmp;
600 	return (error);
601 }
602 
603 /*
604  * sysctl helper routine for setting a process's specific corefile
605  * name.  picks the process based on the given pid and checks the
606  * correctness of the new value.
607  */
608 static int
609 sysctl_proc_corename(SYSCTLFN_ARGS)
610 {
611 	struct proc *ptmp, *p;
612 	struct plimit *lim;
613 	int error = 0, len;
614 	char cname[MAXPATHLEN], *tmp;
615 	struct sysctlnode node;
616 
617 	/*
618 	 * is this all correct?
619 	 */
620 	if (namelen != 0)
621 		return (EINVAL);
622 	if (name[-1] != PROC_PID_CORENAME)
623 		return (EINVAL);
624 
625 	/*
626 	 * whom are we tweaking?
627 	 */
628 	p = l->l_proc;
629 	error = sysctl_proc_findproc(p, &ptmp, (pid_t)name[-2]);
630 	if (error)
631 		return (error);
632 
633 	/*
634 	 * let them modify a temporary copy of the core name
635 	 */
636 	node = *rnode;
637 	strlcpy(cname, ptmp->p_limit->pl_corename, sizeof(cname));
638 	node.sysctl_data = cname;
639 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
640 
641 	/*
642 	 * if that failed, or they have nothing new to say, or we've
643 	 * heard it before...
644 	 */
645 	if (error || newp == NULL ||
646 	    strcmp(cname, ptmp->p_limit->pl_corename) == 0)
647 		return (error);
648 
649 	/*
650 	 * no error yet and cname now has the new core name in it.
651 	 * let's see if it looks acceptable.  it must be either "core"
652 	 * or end in ".core" or "/core".
653 	 */
654 	len = strlen(cname);
655 	if (len < 4)
656 		return (EINVAL);
657 	if (strcmp(cname + len - 4, "core") != 0)
658 		return (EINVAL);
659 	if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.')
660 		return (EINVAL);
661 
662 	/*
663 	 * hmm...looks good.  now...where do we put it?
664 	 */
665 	tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL);
666 	if (tmp == NULL)
667 		return (ENOMEM);
668 	strlcpy(tmp, cname, len + 1);
669 
670 	lim = ptmp->p_limit;
671 	if (lim->p_refcnt > 1 && (lim->p_lflags & PL_SHAREMOD) == 0) {
672 		ptmp->p_limit = limcopy(lim);
673 		limfree(lim);
674 		lim = ptmp->p_limit;
675 	}
676 	if (lim->pl_corename != defcorename)
677 		free(lim->pl_corename, M_TEMP);
678 	lim->pl_corename = tmp;
679 
680 	return (error);
681 }
682 
683 /*
684  * sysctl helper routine for checking/setting a process's stop flags,
685  * one for fork and one for exec.
686  */
687 static int
688 sysctl_proc_stop(SYSCTLFN_ARGS)
689 {
690 	struct proc *p, *ptmp;
691 	int i, f, error = 0;
692 	struct sysctlnode node;
693 
694 	if (namelen != 0)
695 		return (EINVAL);
696 
697 	p = l->l_proc;
698 	error = sysctl_proc_findproc(p, &ptmp, (pid_t)name[-2]);
699 	if (error)
700 		return (error);
701 
702 	switch (rnode->sysctl_num) {
703 	case PROC_PID_STOPFORK:
704 		f = P_STOPFORK;
705 		break;
706 	case PROC_PID_STOPEXEC:
707 		f = P_STOPEXEC;
708 		break;
709 	case PROC_PID_STOPEXIT:
710 		f = P_STOPEXIT;
711 		break;
712 	default:
713 		return (EINVAL);
714 	}
715 
716 	i = (ptmp->p_flag & f) ? 1 : 0;
717 	node = *rnode;
718 	node.sysctl_data = &i;
719 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
720 	if (error || newp == NULL)
721 		return (error);
722 
723 	if (i)
724 		ptmp->p_flag |= f;
725 	else
726 		ptmp->p_flag &= ~f;
727 
728 	return (0);
729 }
730 
731 /*
732  * sysctl helper routine for a process's rlimits as exposed by sysctl.
733  */
734 static int
735 sysctl_proc_plimit(SYSCTLFN_ARGS)
736 {
737 	struct proc *ptmp, *p;
738 	u_int limitno;
739 	int which, error = 0;
740         struct rlimit alim;
741 	struct sysctlnode node;
742 
743 	if (namelen != 0)
744 		return (EINVAL);
745 
746 	which = name[-1];
747 	if (which != PROC_PID_LIMIT_TYPE_SOFT &&
748 	    which != PROC_PID_LIMIT_TYPE_HARD)
749 		return (EINVAL);
750 
751 	limitno = name[-2] - 1;
752 	if (limitno >= RLIM_NLIMITS)
753 		return (EINVAL);
754 
755 	if (name[-3] != PROC_PID_LIMIT)
756 		return (EINVAL);
757 
758 	p = l->l_proc;
759 	error = sysctl_proc_findproc(p, &ptmp, (pid_t)name[-4]);
760 	if (error)
761 		return (error);
762 
763 	node = *rnode;
764 	memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim));
765 	if (which == PROC_PID_LIMIT_TYPE_HARD)
766 		node.sysctl_data = &alim.rlim_max;
767 	else
768 		node.sysctl_data = &alim.rlim_cur;
769 
770 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
771 	if (error || newp == NULL)
772 		return (error);
773 
774 	return (dosetrlimit(ptmp, p->p_cred, limitno, &alim));
775 }
776 
777 /*
778  * and finally, the actually glue that sticks it to the tree
779  */
780 SYSCTL_SETUP(sysctl_proc_setup, "sysctl proc subtree setup")
781 {
782 
783 	sysctl_createv(clog, 0, NULL, NULL,
784 		       CTLFLAG_PERMANENT,
785 		       CTLTYPE_NODE, "proc", NULL,
786 		       NULL, 0, NULL, 0,
787 		       CTL_PROC, CTL_EOL);
788 	sysctl_createv(clog, 0, NULL, NULL,
789 		       CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER,
790 		       CTLTYPE_NODE, "curproc",
791 		       SYSCTL_DESCR("Per-process settings"),
792 		       NULL, 0, NULL, 0,
793 		       CTL_PROC, PROC_CURPROC, CTL_EOL);
794 
795 	sysctl_createv(clog, 0, NULL, NULL,
796 		       CTLFLAG_PERMANENT|CTLFLAG_READONLY2|CTLFLAG_ANYWRITE,
797 		       CTLTYPE_STRING, "corename",
798 		       SYSCTL_DESCR("Core file name"),
799 		       sysctl_proc_corename, 0, NULL, MAXPATHLEN,
800 		       CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL);
801 	sysctl_createv(clog, 0, NULL, NULL,
802 		       CTLFLAG_PERMANENT,
803 		       CTLTYPE_NODE, "rlimit",
804 		       SYSCTL_DESCR("Process limits"),
805 		       NULL, 0, NULL, 0,
806 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL);
807 
808 #define create_proc_plimit(s, n) do {					\
809 	sysctl_createv(clog, 0, NULL, NULL,				\
810 		       CTLFLAG_PERMANENT,				\
811 		       CTLTYPE_NODE, s,					\
812 		       SYSCTL_DESCR("Process " s " limits"),		\
813 		       NULL, 0, NULL, 0,				\
814 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
815 		       CTL_EOL);					\
816 	sysctl_createv(clog, 0, NULL, NULL,				\
817 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \
818 		       CTLTYPE_QUAD, "soft",				\
819 		       SYSCTL_DESCR("Process soft " s " limit"),	\
820 		       sysctl_proc_plimit, 0, NULL, 0,			\
821 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
822 		       PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL);		\
823 	sysctl_createv(clog, 0, NULL, NULL,				\
824 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \
825 		       CTLTYPE_QUAD, "hard",				\
826 		       SYSCTL_DESCR("Process hard " s " limit"),	\
827 		       sysctl_proc_plimit, 0, NULL, 0,			\
828 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
829 		       PROC_PID_LIMIT_TYPE_HARD, CTL_EOL);		\
830 	} while (0/*CONSTCOND*/)
831 
832 	create_proc_plimit("cputime",		PROC_PID_LIMIT_CPU);
833 	create_proc_plimit("filesize",		PROC_PID_LIMIT_FSIZE);
834 	create_proc_plimit("datasize",		PROC_PID_LIMIT_DATA);
835 	create_proc_plimit("stacksize",		PROC_PID_LIMIT_STACK);
836 	create_proc_plimit("coredumpsize",	PROC_PID_LIMIT_CORE);
837 	create_proc_plimit("memoryuse",		PROC_PID_LIMIT_RSS);
838 	create_proc_plimit("memorylocked",	PROC_PID_LIMIT_MEMLOCK);
839 	create_proc_plimit("maxproc",		PROC_PID_LIMIT_NPROC);
840 	create_proc_plimit("descriptors",	PROC_PID_LIMIT_NOFILE);
841 	create_proc_plimit("sbsize",		PROC_PID_LIMIT_SBSIZE);
842 
843 #undef create_proc_plimit
844 
845 	sysctl_createv(clog, 0, NULL, NULL,
846 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
847 		       CTLTYPE_INT, "stopfork",
848 		       SYSCTL_DESCR("Stop process at fork(2)"),
849 		       sysctl_proc_stop, 0, NULL, 0,
850 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL);
851 	sysctl_createv(clog, 0, NULL, NULL,
852 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
853 		       CTLTYPE_INT, "stopexec",
854 		       SYSCTL_DESCR("Stop process at execve(2)"),
855 		       sysctl_proc_stop, 0, NULL, 0,
856 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL);
857 	sysctl_createv(clog, 0, NULL, NULL,
858 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
859 		       CTLTYPE_INT, "stopexit",
860 		       SYSCTL_DESCR("Stop process before completing exit"),
861 		       sysctl_proc_stop, 0, NULL, 0,
862 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL);
863 }
864 
865 struct uidinfo *
866 uid_find(uid_t uid)
867 {
868 	struct uidinfo *uip;
869 	struct uidinfo *newuip = NULL;
870 	struct uihashhead *uipp;
871 
872 	uipp = UIHASH(uid);
873 
874 again:
875 	simple_lock(&uihashtbl_slock);
876 	LIST_FOREACH(uip, uipp, ui_hash)
877 		if (uip->ui_uid == uid) {
878 			simple_unlock(&uihashtbl_slock);
879 			if (newuip)
880 				free(newuip, M_PROC);
881 			return uip;
882 		}
883 
884 	if (newuip == NULL) {
885 		simple_unlock(&uihashtbl_slock);
886 		newuip = malloc(sizeof(*uip), M_PROC, M_WAITOK | M_ZERO);
887 		goto again;
888 	}
889 	uip = newuip;
890 
891 	LIST_INSERT_HEAD(uipp, uip, ui_hash);
892 	uip->ui_uid = uid;
893 	simple_lock_init(&uip->ui_slock);
894 	simple_unlock(&uihashtbl_slock);
895 
896 	return uip;
897 }
898 
899 /*
900  * Change the count associated with number of processes
901  * a given user is using.
902  */
903 int
904 chgproccnt(uid_t uid, int diff)
905 {
906 	struct uidinfo *uip;
907 	int s;
908 
909 	if (diff == 0)
910 		return 0;
911 
912 	uip = uid_find(uid);
913 	UILOCK(uip, s);
914 	uip->ui_proccnt += diff;
915 	KASSERT(uip->ui_proccnt >= 0);
916 	UIUNLOCK(uip, s);
917 	return uip->ui_proccnt;
918 }
919 
920 int
921 chgsbsize(struct uidinfo *uip, u_long *hiwat, u_long to, rlim_t xmax)
922 {
923 	rlim_t nsb;
924 	int s;
925 
926 	UILOCK(uip, s);
927 	nsb = uip->ui_sbsize + to - *hiwat;
928 	if (to > *hiwat && nsb > xmax) {
929 		UIUNLOCK(uip, s);
930 		splx(s);
931 		return 0;
932 	}
933 	*hiwat = to;
934 	uip->ui_sbsize = nsb;
935 	KASSERT(uip->ui_sbsize >= 0);
936 	UIUNLOCK(uip, s);
937 	return 1;
938 }
939