xref: /netbsd-src/sys/kern/kern_resource.c (revision d20841bb642898112fe68f0ad3f7b26dddf56f07)
1 /*	$NetBSD: kern_resource.c,v 1.75 2003/12/06 04:25:57 atatat 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 1.75 2003/12/06 04:25:57 atatat 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  * XXXX Do we really need them to be patchable?
63  */
64 rlim_t maxdmap = MAXDSIZ;
65 rlim_t maxsmap = MAXSSIZ;
66 
67 /*
68  * Resource controls and accounting.
69  */
70 
71 int
72 sys_getpriority(l, v, retval)
73 	struct lwp *l;
74 	void *v;
75 	register_t *retval;
76 {
77 	struct sys_getpriority_args /* {
78 		syscallarg(int) which;
79 		syscallarg(int) 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 		LIST_FOREACH(p, &allproc, p_list) {
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(l, v, retval)
134 	struct lwp *l;
135 	void *v;
136 	register_t *retval;
137 {
138 	struct sys_setpriority_args /* {
139 		syscallarg(int) which;
140 		syscallarg(int) who;
141 		syscallarg(int) prio;
142 	} */ *uap = v;
143 	struct proc *curp = l->l_proc, *p;
144 	int found = 0, error = 0;
145 
146 	switch (SCARG(uap, which)) {
147 
148 	case PRIO_PROCESS:
149 		if (SCARG(uap, who) == 0)
150 			p = curp;
151 		else
152 			p = pfind(SCARG(uap, who));
153 		if (p == 0)
154 			break;
155 		error = donice(curp, p, SCARG(uap, prio));
156 		found++;
157 		break;
158 
159 	case PRIO_PGRP: {
160 		struct pgrp *pg;
161 
162 		if (SCARG(uap, who) == 0)
163 			pg = curp->p_pgrp;
164 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
165 			break;
166 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
167 			error = donice(curp, p, SCARG(uap, prio));
168 			found++;
169 		}
170 		break;
171 	}
172 
173 	case PRIO_USER:
174 		if (SCARG(uap, who) == 0)
175 			SCARG(uap, who) = curp->p_ucred->cr_uid;
176 		proclist_lock_read();
177 		LIST_FOREACH(p, &allproc, p_list) {
178 			if (p->p_ucred->cr_uid == (uid_t) SCARG(uap, who)) {
179 				error = donice(curp, 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(curp, chgp, n)
196 	struct proc *curp, *chgp;
197 	int n;
198 {
199 	struct pcred *pcred = curp->p_cred;
200 	int s;
201 
202 	if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
203 	    pcred->pc_ucred->cr_uid != chgp->p_ucred->cr_uid &&
204 	    pcred->p_ruid != chgp->p_ucred->cr_uid)
205 		return (EPERM);
206 	if (n > PRIO_MAX)
207 		n = PRIO_MAX;
208 	if (n < PRIO_MIN)
209 		n = PRIO_MIN;
210 	n += NZERO;
211 	if (n < chgp->p_nice && suser(pcred->pc_ucred, &curp->p_acflag))
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(l, v, retval)
223 	struct lwp *l;
224 	void *v;
225 	register_t *retval;
226 {
227 	struct sys_setrlimit_args /* {
228 		syscallarg(int) which;
229 		syscallarg(const struct rlimit *) rlp;
230 	} */ *uap = v;
231 	struct proc *p = l->l_proc;
232 	int which = SCARG(uap, which);
233 	struct rlimit alim;
234 	int error;
235 
236 	error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit));
237 	if (error)
238 		return (error);
239 	return (dosetrlimit(p, p->p_cred, which, &alim));
240 }
241 
242 int
243 dosetrlimit(p, cred, which, limp)
244 	struct proc *p;
245 	struct  pcred *cred;
246 	int which;
247 	struct rlimit *limp;
248 {
249 	struct rlimit *alimp;
250 	struct plimit *newplim;
251 	int error;
252 
253 	if ((u_int)which >= RLIM_NLIMITS)
254 		return (EINVAL);
255 
256 	if (limp->rlim_cur < 0 || limp->rlim_max < 0)
257 		return (EINVAL);
258 
259 	alimp = &p->p_rlimit[which];
260 	/* if we don't change the value, no need to limcopy() */
261 	if (limp->rlim_cur == alimp->rlim_cur &&
262 	    limp->rlim_max == alimp->rlim_max)
263 		return 0;
264 
265 	if (limp->rlim_cur > limp->rlim_max) {
266 		/*
267 		 * This is programming error. According to SUSv2, we should
268 		 * return error in this case.
269 		 */
270 		return (EINVAL);
271 	}
272 	if (limp->rlim_max > alimp->rlim_max
273 	    && (error = suser(cred->pc_ucred, &p->p_acflag)) != 0)
274 			return (error);
275 
276 	if (p->p_limit->p_refcnt > 1 &&
277 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
278 		newplim = limcopy(p->p_limit);
279 		limfree(p->p_limit);
280 		p->p_limit = newplim;
281 		alimp = &p->p_rlimit[which];
282 	}
283 
284 	switch (which) {
285 
286 	case RLIMIT_DATA:
287 		if (limp->rlim_cur > maxdmap)
288 			limp->rlim_cur = maxdmap;
289 		if (limp->rlim_max > maxdmap)
290 			limp->rlim_max = maxdmap;
291 		break;
292 
293 	case RLIMIT_STACK:
294 		if (limp->rlim_cur > maxsmap)
295 			limp->rlim_cur = maxsmap;
296 		if (limp->rlim_max > maxsmap)
297 			limp->rlim_max = maxsmap;
298 
299 		/*
300 		 * Return EINVAL if the new stack size limit is lower than
301 		 * current usage. Otherwise, the process would get SIGSEGV the
302 		 * moment it would try to access anything on it's current stack.
303 		 * This conforms to SUSv2.
304 		 */
305 		if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE
306 		    || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE)
307 			return (EINVAL);
308 
309 		/*
310 		 * Stack is allocated to the max at exec time with
311 		 * only "rlim_cur" bytes accessible (In other words,
312 		 * allocates stack dividing two contiguous regions at
313 		 * "rlim_cur" bytes boundary).
314 		 *
315 		 * Since allocation is done in terms of page, roundup
316 		 * "rlim_cur" (otherwise, contiguous regions
317 		 * overlap).  If stack limit is going up make more
318 		 * accessible, if going down make inaccessible.
319 		 */
320 		limp->rlim_cur = round_page(limp->rlim_cur);
321 		if (limp->rlim_cur != alimp->rlim_cur) {
322 			vaddr_t addr;
323 			vsize_t size;
324 			vm_prot_t prot;
325 
326 			if (limp->rlim_cur > alimp->rlim_cur) {
327 				prot = VM_PROT_READ | VM_PROT_WRITE;
328 				size = limp->rlim_cur - alimp->rlim_cur;
329 				addr = USRSTACK - limp->rlim_cur;
330 			} else {
331 				prot = VM_PROT_NONE;
332 				size = alimp->rlim_cur - limp->rlim_cur;
333 				addr = USRSTACK - alimp->rlim_cur;
334 			}
335 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
336 					      addr, addr+size, prot, FALSE);
337 		}
338 		break;
339 
340 	case RLIMIT_NOFILE:
341 		if (limp->rlim_cur > maxfiles)
342 			limp->rlim_cur = maxfiles;
343 		if (limp->rlim_max > maxfiles)
344 			limp->rlim_max = maxfiles;
345 		break;
346 
347 	case RLIMIT_NPROC:
348 		if (limp->rlim_cur > maxproc)
349 			limp->rlim_cur = maxproc;
350 		if (limp->rlim_max > maxproc)
351 			limp->rlim_max = maxproc;
352 		break;
353 	}
354 	*alimp = *limp;
355 	return (0);
356 }
357 
358 /* ARGSUSED */
359 int
360 sys_getrlimit(l, v, retval)
361 	struct lwp *l;
362 	void *v;
363 	register_t *retval;
364 {
365 	struct sys_getrlimit_args /* {
366 		syscallarg(int) which;
367 		syscallarg(struct rlimit *) rlp;
368 	} */ *uap = v;
369 	struct proc *p = l->l_proc;
370 	int which = SCARG(uap, which);
371 
372 	if ((u_int)which >= RLIM_NLIMITS)
373 		return (EINVAL);
374 	return (copyout(&p->p_rlimit[which], SCARG(uap, rlp),
375 	    sizeof(struct rlimit)));
376 }
377 
378 /*
379  * Transform the running time and tick information in proc p into user,
380  * system, and interrupt time usage.
381  */
382 void
383 calcru(p, up, sp, ip)
384 	struct proc *p;
385 	struct timeval *up;
386 	struct timeval *sp;
387 	struct timeval *ip;
388 {
389 	u_quad_t u, st, ut, it, tot;
390 	unsigned long sec;
391 	long usec;
392 	int s;
393 	struct timeval tv;
394 	struct lwp *l;
395 
396 	s = splstatclock();
397 	st = p->p_sticks;
398 	ut = p->p_uticks;
399 	it = p->p_iticks;
400 	splx(s);
401 
402 	sec = p->p_rtime.tv_sec;
403 	usec = p->p_rtime.tv_usec;
404 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
405 		if (l->l_stat == LSONPROC) {
406 			struct schedstate_percpu *spc;
407 
408 			KDASSERT(l->l_cpu != NULL);
409 			spc = &l->l_cpu->ci_schedstate;
410 
411 			/*
412 			 * Adjust for the current time slice.  This is
413 			 * actually fairly important since the error
414 			 * here is on the order of a time quantum,
415 			 * which is much greater than the sampling
416 			 * error.
417 			 */
418 			microtime(&tv);
419 			sec += tv.tv_sec - spc->spc_runtime.tv_sec;
420 			usec += tv.tv_usec - spc->spc_runtime.tv_usec;
421 		}
422 	}
423 
424 	tot = st + ut + it;
425 	u = sec * 1000000ull + usec;
426 
427 	if (tot == 0) {
428 		/* No ticks, so can't use to share time out, split 50-50 */
429 		st = ut = u / 2;
430 	} else {
431 		st = (u * st) / tot;
432 		ut = (u * ut) / tot;
433 	}
434 	sp->tv_sec = st / 1000000;
435 	sp->tv_usec = st % 1000000;
436 	up->tv_sec = ut / 1000000;
437 	up->tv_usec = ut % 1000000;
438 	if (ip != NULL) {
439 		if (it != 0)
440 			it = (u * it) / tot;
441 		ip->tv_sec = it / 1000000;
442 		ip->tv_usec = it % 1000000;
443 	}
444 }
445 
446 /* ARGSUSED */
447 int
448 sys_getrusage(l, v, retval)
449 	struct lwp *l;
450 	void *v;
451 	register_t *retval;
452 {
453 	struct sys_getrusage_args /* {
454 		syscallarg(int) who;
455 		syscallarg(struct rusage *) rusage;
456 	} */ *uap = v;
457 	struct rusage *rup;
458 	struct proc *p = l->l_proc;
459 
460 	switch (SCARG(uap, who)) {
461 
462 	case RUSAGE_SELF:
463 		rup = &p->p_stats->p_ru;
464 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
465 		break;
466 
467 	case RUSAGE_CHILDREN:
468 		rup = &p->p_stats->p_cru;
469 		break;
470 
471 	default:
472 		return (EINVAL);
473 	}
474 	return (copyout(rup, SCARG(uap, rusage), sizeof(struct rusage)));
475 }
476 
477 void
478 ruadd(ru, ru2)
479 	struct rusage *ru, *ru2;
480 {
481 	long *ip, *ip2;
482 	int i;
483 
484 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
485 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
486 	if (ru->ru_maxrss < ru2->ru_maxrss)
487 		ru->ru_maxrss = ru2->ru_maxrss;
488 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
489 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
490 		*ip++ += *ip2++;
491 }
492 
493 /*
494  * Make a copy of the plimit structure.
495  * We share these structures copy-on-write after fork,
496  * and copy when a limit is changed.
497  */
498 struct plimit *
499 limcopy(lim)
500 	struct plimit *lim;
501 {
502 	struct plimit *newlim;
503 	size_t l;
504 
505 	newlim = pool_get(&plimit_pool, PR_WAITOK);
506 	memcpy(newlim->pl_rlimit, lim->pl_rlimit,
507 	    sizeof(struct rlimit) * RLIM_NLIMITS);
508 	if (lim->pl_corename == defcorename) {
509 		newlim->pl_corename = defcorename;
510 	} else {
511 		l = strlen(lim->pl_corename) + 1;
512 		newlim->pl_corename = malloc(l, M_TEMP, M_WAITOK);
513 		strlcpy(newlim->pl_corename, lim->pl_corename, l);
514 	}
515 	newlim->p_lflags = 0;
516 	newlim->p_refcnt = 1;
517 	return (newlim);
518 }
519 
520 void
521 limfree(lim)
522 	struct plimit *lim;
523 {
524 
525 	if (--lim->p_refcnt > 0)
526 		return;
527 #ifdef DIAGNOSTIC
528 	if (lim->p_refcnt < 0)
529 		panic("limfree");
530 #endif
531 	if (lim->pl_corename != defcorename)
532 		free(lim->pl_corename, M_TEMP);
533 	pool_put(&plimit_pool, lim);
534 }
535 
536 struct pstats *
537 pstatscopy(ps)
538 	struct pstats *ps;
539 {
540 
541 	struct pstats *newps;
542 
543 	newps = pool_get(&pstats_pool, PR_WAITOK);
544 
545 	memset(&newps->pstat_startzero, 0,
546 	(unsigned) ((caddr_t)&newps->pstat_endzero -
547 		    (caddr_t)&newps->pstat_startzero));
548 	memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy,
549 	((caddr_t)&newps->pstat_endcopy -
550 	 (caddr_t)&newps->pstat_startcopy));
551 
552 	return (newps);
553 
554 }
555 
556 void
557 pstatsfree(ps)
558 	struct pstats *ps;
559 {
560 
561 	pool_put(&pstats_pool, ps);
562 }
563 
564 /*
565  * sysctl interface in five parts
566  */
567 
568 /*
569  * a routine for sysctl proc subtree helpers that need to pick a valid
570  * process by pid.
571  */
572 static int
573 sysctl_proc_findproc(struct proc *p, struct proc **p2, pid_t pid)
574 {
575 	struct proc *ptmp;
576 	int i, error = 0;
577 
578 	if (pid == PROC_CURPROC)
579 		ptmp = p;
580 	else if ((ptmp = pfind(pid)) == NULL)
581 		error = ESRCH;
582 	else {
583 		/*
584 		 * suid proc of ours or proc not ours
585 		 */
586 		if (p->p_cred->p_ruid != ptmp->p_cred->p_ruid ||
587 		    p->p_cred->p_ruid != ptmp->p_cred->p_svuid)
588 			error = suser(p->p_ucred, &p->p_acflag);
589 
590 		/*
591 		 * sgid proc has sgid back to us temporarily
592 		 */
593 		else if (ptmp->p_cred->p_rgid != ptmp->p_cred->p_svgid)
594 			error = suser(p->p_ucred, &p->p_acflag);
595 
596 		/*
597 		 * our rgid must be in target's group list (ie,
598 		 * sub-processes started by a sgid process)
599 		 */
600 		else {
601 			for (i = 0; i < p->p_ucred->cr_ngroups; i++) {
602 				if (p->p_ucred->cr_groups[i] ==
603 				    ptmp->p_cred->p_rgid)
604 					break;
605 			}
606 			if (i == p->p_ucred->cr_ngroups)
607 				error = suser(p->p_ucred, &p->p_acflag);
608 		}
609 	}
610 
611 	*p2 = ptmp;
612 	return (error);
613 }
614 
615 /*
616  * sysctl helper routine for setting a process's specific corefile
617  * name.  picks the process based on the given pid and checks the
618  * correctness of the new value.
619  */
620 static int
621 sysctl_proc_corename(SYSCTLFN_ARGS)
622 {
623 	struct proc *ptmp, *p;
624 	struct plimit *newplim;
625 	int error = 0, len;
626 	char cname[MAXPATHLEN], *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 	p = l->l_proc;
641 	error = sysctl_proc_findproc(p, &ptmp, (pid_t)name[-2]);
642 	if (error)
643 		return (error);
644 
645 	/*
646 	 * let them modify a temporary copy of the core name
647 	 */
648 	node = *rnode;
649 	strlcpy(cname, ptmp->p_limit->pl_corename, sizeof(cname));
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 		return (error);
660 
661 	/*
662 	 * no error yet and cname now has the new core name in it.
663 	 * let's see if it looks acceptable.  it must be either "core"
664 	 * or end in ".core" or "/core".
665 	 */
666 	len = strlen(cname);
667 	if (len < 4)
668 		return (EINVAL);
669 	if (strcmp(cname + len - 4, "core") != 0)
670 		return (EINVAL);
671 	if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.')
672 		return (EINVAL);
673 
674 	/*
675 	 * hmm...looks good.  now...where do we put it?
676 	 */
677 	tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL);
678 	if (tmp == NULL)
679 		return (ENOMEM);
680 	strlcpy(tmp, cname, len + 1);
681 
682 	if (ptmp->p_limit->p_refcnt > 1 &&
683 	    (ptmp->p_limit->p_lflags & PL_SHAREMOD) == 0) {
684 		newplim = limcopy(ptmp->p_limit);
685 		limfree(ptmp->p_limit);
686 		ptmp->p_limit = newplim;
687 	}
688 	if (ptmp->p_limit->pl_corename != defcorename)
689 		FREE(ptmp->p_limit->pl_corename, M_SYSCTLDATA);
690 	ptmp->p_limit->pl_corename = tmp;
691 
692 	return (error);
693 }
694 
695 /*
696  * sysctl helper routine for checking/setting a process's stop flags,
697  * one for fork and one for exec.
698  */
699 static int
700 sysctl_proc_stop(SYSCTLFN_ARGS)
701 {
702 	struct proc *p, *ptmp;
703 	int i, f, error = 0;
704 	struct sysctlnode node;
705 
706 	if (namelen != 0)
707 		return (EINVAL);
708 
709 	p = l->l_proc;
710 	error = sysctl_proc_findproc(p, &ptmp, (pid_t)name[-2]);
711 	if (error)
712 		return (error);
713 
714 	switch (rnode->sysctl_num) {
715 	case PROC_PID_STOPFORK:
716 		f = P_STOPFORK;
717 		break;
718 	case PROC_PID_STOPEXEC:
719 		f = P_STOPEXEC;
720 		break;
721 	case PROC_PID_STOPEXIT:
722 		f = P_STOPEXIT;
723 		break;
724 	default:
725 		return (EINVAL);
726 	}
727 
728 	i = (ptmp->p_flag & f) ? 1 : 0;
729 	node = *rnode;
730 	node.sysctl_data = &i;
731 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
732 	if (error || newp == NULL)
733 		return (error);
734 
735 	if (i)
736 		ptmp->p_flag |= f;
737 	else
738 		ptmp->p_flag &= ~f;
739 
740 	return (0);
741 }
742 
743 /*
744  * sysctl helper routine for a process's rlimits as exposed by sysctl.
745  */
746 static int
747 sysctl_proc_plimit(SYSCTLFN_ARGS)
748 {
749 	struct proc *ptmp, *p;
750 	u_int limitno;
751 	int which, error = 0;
752         struct rlimit alim;
753 	struct sysctlnode node;
754 
755 	if (namelen != 0)
756 		return (EINVAL);
757 
758 	which = name[-1];
759 	if (which != PROC_PID_LIMIT_TYPE_SOFT &&
760 	    which != PROC_PID_LIMIT_TYPE_HARD)
761 		return (EINVAL);
762 
763 	limitno = name[-2] - 1;
764 	if (limitno >= RLIM_NLIMITS)
765 		return (EINVAL);
766 
767 	if (name[-3] != PROC_PID_LIMIT)
768 		return (EINVAL);
769 
770 	p = l->l_proc;
771 	error = sysctl_proc_findproc(p, &ptmp, (pid_t)name[-4]);
772 	if (error)
773 		return (error);
774 
775 	node = *rnode;
776 	memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim));
777 	if (which == PROC_PID_LIMIT_TYPE_HARD)
778 		node.sysctl_data = &alim.rlim_max;
779 	else
780 		node.sysctl_data = &alim.rlim_cur;
781 
782 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
783 	if (error || newp == NULL)
784 		return (error);
785 
786 	return (dosetrlimit(ptmp, p->p_cred, limitno, &alim));
787 }
788 
789 /*
790  * and finally, the actually glue that sticks it to the tree
791  */
792 SYSCTL_SETUP(sysctl_proc_setup, "sysctl proc subtree setup")
793 {
794 
795 	sysctl_createv(SYSCTL_PERMANENT,
796 		       CTLTYPE_NODE, "proc", NULL,
797 		       NULL, 0, NULL, 0,
798 		       CTL_PROC, CTL_EOL);
799 	sysctl_createv(SYSCTL_PERMANENT|SYSCTL_ANYNUMBER,
800 		       CTLTYPE_NODE, "curproc", NULL,
801 		       NULL, 0, NULL, 0,
802 		       CTL_PROC, PROC_CURPROC, CTL_EOL);
803 
804 	sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READONLY2|SYSCTL_ANYWRITE,
805 		       CTLTYPE_STRING, "corename", NULL,
806 		       sysctl_proc_corename, 0, NULL, MAXPATHLEN,
807 		       CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL);
808 	sysctl_createv(SYSCTL_PERMANENT,
809 		       CTLTYPE_NODE, "rlimit", NULL,
810 		       NULL, 0, NULL, 0,
811 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL);
812 
813 #define create_proc_plimit(s, n) do {					\
814 	sysctl_createv(SYSCTL_PERMANENT,				\
815 		       CTLTYPE_NODE, s, NULL,				\
816 		       NULL, 0, NULL, 0,				\
817 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
818 		       CTL_EOL);					\
819 	sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE|SYSCTL_ANYWRITE, \
820 		       CTLTYPE_QUAD, "soft", NULL,			\
821 		       sysctl_proc_plimit, 0, NULL, 0,			\
822 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
823 		       PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL);		\
824 	sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE|SYSCTL_ANYWRITE, \
825 		       CTLTYPE_QUAD, "hard", NULL,			\
826 		       sysctl_proc_plimit, 0, NULL, 0,			\
827 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
828 		       PROC_PID_LIMIT_TYPE_HARD, CTL_EOL);		\
829 	} while (0/*CONSTCOND*/)
830 
831 	create_proc_plimit("cputime",		PROC_PID_LIMIT_CPU);
832 	create_proc_plimit("filesize",		PROC_PID_LIMIT_FSIZE);
833 	create_proc_plimit("datasize",		PROC_PID_LIMIT_DATA);
834 	create_proc_plimit("stacksize",		PROC_PID_LIMIT_STACK);
835 	create_proc_plimit("coredumpsize",	PROC_PID_LIMIT_CORE);
836 	create_proc_plimit("memoryuse",		PROC_PID_LIMIT_RSS);
837 	create_proc_plimit("memorylocked",	PROC_PID_LIMIT_MEMLOCK);
838 	create_proc_plimit("maxproc",		PROC_PID_LIMIT_NPROC);
839 	create_proc_plimit("descriptors",	PROC_PID_LIMIT_NOFILE);
840 
841 #undef create_proc_plimit
842 
843 	sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE|SYSCTL_ANYWRITE,
844 		       CTLTYPE_INT, "stopfork", NULL,
845 		       sysctl_proc_stop, 0, NULL, 0,
846 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL);
847 	sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE|SYSCTL_ANYWRITE,
848 		       CTLTYPE_INT, "stopexec", NULL,
849 		       sysctl_proc_stop, 0, NULL, 0,
850 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL);
851 	sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE|SYSCTL_ANYWRITE,
852 		       CTLTYPE_INT, "stopexit", NULL,
853 		       sysctl_proc_stop, 0, NULL, 0,
854 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL);
855 }
856