xref: /freebsd/sys/kern/kern_resource.c (revision ce834215a70ff69e7e222827437116eee2f9ac6f)
1 /*-
2  * Copyright (c) 1982, 1986, 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * (c) UNIX System Laboratories, Inc.
5  * All or some portions of this file are derived from material licensed
6  * to the University of California by American Telephone and Telegraph
7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8  * the permission of UNIX System Laboratories, Inc.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. All advertising materials mentioning features or use of this software
19  *    must display the following acknowledgement:
20  *	This product includes software developed by the University of
21  *	California, Berkeley and its contributors.
22  * 4. Neither the name of the University nor the names of its contributors
23  *    may be used to endorse or promote products derived from this software
24  *    without specific prior written permission.
25  *
26  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36  * SUCH DAMAGE.
37  *
38  *	@(#)kern_resource.c	8.5 (Berkeley) 1/21/94
39  * $Id: kern_resource.c,v 1.24 1997/02/22 09:39:09 peter Exp $
40  */
41 
42 #include "opt_rlimit.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/sysproto.h>
47 #include <sys/kernel.h>
48 #include <sys/file.h>
49 #include <sys/resourcevar.h>
50 #include <sys/malloc.h>
51 #include <sys/proc.h>
52 
53 #include <vm/vm.h>
54 #include <vm/vm_param.h>
55 #include <vm/vm_prot.h>
56 #include <sys/lock.h>
57 #include <vm/pmap.h>
58 #include <vm/vm_map.h>
59 
60 int	donice __P((struct proc *curp, struct proc *chgp, int n));
61 int	dosetrlimit __P((struct proc *p, u_int which, struct rlimit *limp));
62 
63 /*
64  * Resource controls and accounting.
65  */
66 
67 #ifndef _SYS_SYSPROTO_H_
68 struct getpriority_args {
69 	int	which;
70 	int	who;
71 };
72 #endif
73 int
74 getpriority(curp, uap, retval)
75 	struct proc *curp;
76 	register struct getpriority_args *uap;
77 	int *retval;
78 {
79 	register struct proc *p;
80 	register int low = PRIO_MAX + 1;
81 
82 	switch (uap->which) {
83 
84 	case PRIO_PROCESS:
85 		if (uap->who == 0)
86 			p = curp;
87 		else
88 			p = pfind(uap->who);
89 		if (p == 0)
90 			break;
91 		low = p->p_nice;
92 		break;
93 
94 	case PRIO_PGRP: {
95 		register struct pgrp *pg;
96 
97 		if (uap->who == 0)
98 			pg = curp->p_pgrp;
99 		else if ((pg = pgfind(uap->who)) == NULL)
100 			break;
101 		for (p = pg->pg_members.lh_first; p != 0;
102 		     p = p->p_pglist.le_next) {
103 			if (p->p_nice < low)
104 				low = p->p_nice;
105 		}
106 		break;
107 	}
108 
109 	case PRIO_USER:
110 		if (uap->who == 0)
111 			uap->who = curp->p_ucred->cr_uid;
112 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
113 			if (p->p_ucred->cr_uid == uap->who &&
114 			    p->p_nice < low)
115 				low = p->p_nice;
116 		break;
117 
118 	default:
119 		return (EINVAL);
120 	}
121 	if (low == PRIO_MAX + 1)
122 		return (ESRCH);
123 	*retval = low;
124 	return (0);
125 }
126 
127 #ifndef _SYS_SYSPROTO_H_
128 struct setpriority_args {
129 	int	which;
130 	int	who;
131 	int	prio;
132 };
133 #endif
134 /* ARGSUSED */
135 int
136 setpriority(curp, uap, retval)
137 	struct proc *curp;
138 	register struct setpriority_args *uap;
139 	int *retval;
140 {
141 	register struct proc *p;
142 	int found = 0, error = 0;
143 
144 	switch (uap->which) {
145 
146 	case PRIO_PROCESS:
147 		if (uap->who == 0)
148 			p = curp;
149 		else
150 			p = pfind(uap->who);
151 		if (p == 0)
152 			break;
153 		error = donice(curp, p, uap->prio);
154 		found++;
155 		break;
156 
157 	case PRIO_PGRP: {
158 		register struct pgrp *pg;
159 
160 		if (uap->who == 0)
161 			pg = curp->p_pgrp;
162 		else if ((pg = pgfind(uap->who)) == NULL)
163 			break;
164 		for (p = pg->pg_members.lh_first; p != 0;
165 		    p = p->p_pglist.le_next) {
166 			error = donice(curp, p, uap->prio);
167 			found++;
168 		}
169 		break;
170 	}
171 
172 	case PRIO_USER:
173 		if (uap->who == 0)
174 			uap->who = curp->p_ucred->cr_uid;
175 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
176 			if (p->p_ucred->cr_uid == uap->who) {
177 				error = donice(curp, p, uap->prio);
178 				found++;
179 			}
180 		break;
181 
182 	default:
183 		return (EINVAL);
184 	}
185 	if (found == 0)
186 		return (ESRCH);
187 	return (error);
188 }
189 
190 int
191 donice(curp, chgp, n)
192 	register struct proc *curp, *chgp;
193 	register int n;
194 {
195 	register struct pcred *pcred = curp->p_cred;
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 	if (n < chgp->p_nice && suser(pcred->pc_ucred, &curp->p_acflag))
206 		return (EACCES);
207 	chgp->p_nice = n;
208 	(void)resetpriority(chgp);
209 	return (0);
210 }
211 
212 /* rtprio system call */
213 #ifndef _SYS_SYSPROTO_H_
214 struct rtprio_args {
215 	int		function;
216 	pid_t		pid;
217 	struct rtprio	*rtp;
218 };
219 #endif
220 
221 /*
222  * Set realtime priority
223  */
224 
225 /* ARGSUSED */
226 int
227 rtprio(curp, uap, retval)
228 	struct proc *curp;
229 	register struct rtprio_args *uap;
230 	int *retval;
231 {
232 	register struct proc *p;
233 	register struct pcred *pcred = curp->p_cred;
234 	struct rtprio rtp;
235 	int error;
236 
237 	error = copyin(uap->rtp, &rtp, sizeof(struct rtprio));
238 	if (error)
239 		return (error);
240 
241 	if (uap->pid == 0)
242 		p = curp;
243 	else
244 		p = pfind(uap->pid);
245 
246 	if (p == 0)
247 		return (ESRCH);
248 
249 	switch (uap->function) {
250 	case RTP_LOOKUP:
251 		return (copyout(&p->p_rtprio, uap->rtp, sizeof(struct rtprio)));
252 	case RTP_SET:
253 		if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
254 		    pcred->pc_ucred->cr_uid != p->p_ucred->cr_uid &&
255 		    pcred->p_ruid != p->p_ucred->cr_uid)
256 		        return (EPERM);
257 		/* disallow setting rtprio in most cases if not superuser */
258 		if (suser(pcred->pc_ucred, &curp->p_acflag)) {
259 			/* can't set someone else's */
260 			if (uap->pid)
261 				return (EPERM);
262 			/* can't set realtime priority */
263 			if (rtp.type == RTP_PRIO_REALTIME)
264 				return (EPERM);
265 		}
266 		switch (rtp.type) {
267 		case RTP_PRIO_REALTIME:
268 		case RTP_PRIO_NORMAL:
269 		case RTP_PRIO_IDLE:
270 			if (rtp.prio > RTP_PRIO_MAX)
271 				return (EINVAL);
272 			p->p_rtprio = rtp;
273 			return (0);
274 		default:
275 			return (EINVAL);
276 		}
277 
278 	default:
279 		return (EINVAL);
280 	}
281 }
282 
283 #if defined(COMPAT_43) || defined(COMPAT_SUNOS)
284 #ifndef _SYS_SYSPROTO_H_
285 struct osetrlimit_args {
286 	u_int	which;
287 	struct	orlimit *rlp;
288 };
289 #endif
290 /* ARGSUSED */
291 int
292 osetrlimit(p, uap, retval)
293 	struct proc *p;
294 	register struct osetrlimit_args *uap;
295 	int *retval;
296 {
297 	struct orlimit olim;
298 	struct rlimit lim;
299 	int error;
300 
301 	if ((error =
302 	    copyin((caddr_t)uap->rlp, (caddr_t)&olim, sizeof(struct orlimit))))
303 		return (error);
304 	lim.rlim_cur = olim.rlim_cur;
305 	lim.rlim_max = olim.rlim_max;
306 	return (dosetrlimit(p, uap->which, &lim));
307 }
308 
309 #ifndef _SYS_SYSPROTO_H_
310 struct ogetrlimit_args {
311 	u_int	which;
312 	struct	orlimit *rlp;
313 };
314 #endif
315 /* ARGSUSED */
316 int
317 ogetrlimit(p, uap, retval)
318 	struct proc *p;
319 	register struct ogetrlimit_args *uap;
320 	int *retval;
321 {
322 	struct orlimit olim;
323 
324 	if (uap->which >= RLIM_NLIMITS)
325 		return (EINVAL);
326 	olim.rlim_cur = p->p_rlimit[uap->which].rlim_cur;
327 	if (olim.rlim_cur == -1)
328 		olim.rlim_cur = 0x7fffffff;
329 	olim.rlim_max = p->p_rlimit[uap->which].rlim_max;
330 	if (olim.rlim_max == -1)
331 		olim.rlim_max = 0x7fffffff;
332 	return (copyout((caddr_t)&olim, (caddr_t)uap->rlp, sizeof(olim)));
333 }
334 #endif /* COMPAT_43 || COMPAT_SUNOS */
335 
336 #ifndef _SYS_SYSPROTO_H_
337 struct __setrlimit_args {
338 	u_int	which;
339 	struct	rlimit *rlp;
340 };
341 #endif
342 /* ARGSUSED */
343 int
344 setrlimit(p, uap, retval)
345 	struct proc *p;
346 	register struct __setrlimit_args *uap;
347 	int *retval;
348 {
349 	struct rlimit alim;
350 	int error;
351 
352 	if ((error =
353 	    copyin((caddr_t)uap->rlp, (caddr_t)&alim, sizeof (struct rlimit))))
354 		return (error);
355 	return (dosetrlimit(p, uap->which, &alim));
356 }
357 
358 int
359 dosetrlimit(p, which, limp)
360 	struct proc *p;
361 	u_int which;
362 	struct rlimit *limp;
363 {
364 	register struct rlimit *alimp;
365 	int error;
366 
367 	if (which >= RLIM_NLIMITS)
368 		return (EINVAL);
369 	alimp = &p->p_rlimit[which];
370 
371 	/*
372 	 * Preserve historical bugs by treating negative limits as unsigned.
373 	 */
374 	if (limp->rlim_cur < 0)
375 		limp->rlim_cur = RLIM_INFINITY;
376 	if (limp->rlim_max < 0)
377 		limp->rlim_max = RLIM_INFINITY;
378 
379 	if (limp->rlim_cur > alimp->rlim_max ||
380 	    limp->rlim_max > alimp->rlim_max)
381 		if ((error = suser(p->p_ucred, &p->p_acflag)))
382 			return (error);
383 	if (limp->rlim_cur > limp->rlim_max)
384 		limp->rlim_cur = limp->rlim_max;
385 	if (p->p_limit->p_refcnt > 1 &&
386 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
387 		p->p_limit->p_refcnt--;
388 		p->p_limit = limcopy(p->p_limit);
389 		alimp = &p->p_rlimit[which];
390 	}
391 
392 	switch (which) {
393 
394 	case RLIMIT_DATA:
395 		if (limp->rlim_cur > MAXDSIZ)
396 			limp->rlim_cur = MAXDSIZ;
397 		if (limp->rlim_max > MAXDSIZ)
398 			limp->rlim_max = MAXDSIZ;
399 		break;
400 
401 	case RLIMIT_STACK:
402 		if (limp->rlim_cur > MAXSSIZ)
403 			limp->rlim_cur = MAXSSIZ;
404 		if (limp->rlim_max > MAXSSIZ)
405 			limp->rlim_max = MAXSSIZ;
406 		/*
407 		 * Stack is allocated to the max at exec time with only
408 		 * "rlim_cur" bytes accessible.  If stack limit is going
409 		 * up make more accessible, if going down make inaccessible.
410 		 */
411 		if (limp->rlim_cur != alimp->rlim_cur) {
412 			vm_offset_t addr;
413 			vm_size_t size;
414 			vm_prot_t prot;
415 
416 			if (limp->rlim_cur > alimp->rlim_cur) {
417 				prot = VM_PROT_ALL;
418 				size = limp->rlim_cur - alimp->rlim_cur;
419 				addr = USRSTACK - limp->rlim_cur;
420 			} else {
421 				prot = VM_PROT_NONE;
422 				size = alimp->rlim_cur - limp->rlim_cur;
423 				addr = USRSTACK - alimp->rlim_cur;
424 			}
425 			addr = trunc_page(addr);
426 			size = round_page(size);
427 			(void) vm_map_protect(&p->p_vmspace->vm_map,
428 					      addr, addr+size, prot, FALSE);
429 		}
430 		break;
431 
432 	case RLIMIT_NOFILE:
433 		if (limp->rlim_cur > maxfilesperproc)
434 			limp->rlim_cur = maxfilesperproc;
435 		if (limp->rlim_max > maxfilesperproc)
436 			limp->rlim_max = maxfilesperproc;
437 		break;
438 
439 	case RLIMIT_NPROC:
440 		if (limp->rlim_cur > maxprocperuid)
441 			limp->rlim_cur = maxprocperuid;
442 		if (limp->rlim_max > maxprocperuid)
443 			limp->rlim_max = maxprocperuid;
444 		break;
445 	}
446 	*alimp = *limp;
447 	return (0);
448 }
449 
450 #ifndef _SYS_SYSPROTO_H_
451 struct __getrlimit_args {
452 	u_int	which;
453 	struct	rlimit *rlp;
454 };
455 #endif
456 /* ARGSUSED */
457 int
458 getrlimit(p, uap, retval)
459 	struct proc *p;
460 	register struct __getrlimit_args *uap;
461 	int *retval;
462 {
463 
464 	if (uap->which >= RLIM_NLIMITS)
465 		return (EINVAL);
466 	return (copyout((caddr_t)&p->p_rlimit[uap->which], (caddr_t)uap->rlp,
467 	    sizeof (struct rlimit)));
468 }
469 
470 /*
471  * Transform the running time and tick information in proc p into user,
472  * system, and interrupt time usage.
473  */
474 void
475 calcru(p, up, sp, ip)
476 	struct proc *p;
477 	struct timeval *up;
478 	struct timeval *sp;
479 	struct timeval *ip;
480 {
481 	quad_t totusec;
482 	u_quad_t u, st, ut, it, tot;
483 	long sec, usec;
484 	int s;
485 	struct timeval tv;
486 
487 	s = splstatclock();
488 	st = p->p_sticks;
489 	ut = p->p_uticks;
490 	it = p->p_iticks;
491 	splx(s);
492 
493 	tot = st + ut + it;
494 	if (tot == 0) {
495 		st = 1;
496 		tot = 1;
497 	}
498 
499 	sec = p->p_rtime.tv_sec;
500 	usec = p->p_rtime.tv_usec;
501 	if (p == curproc) { /* XXX what if it's running on another cpu?? */
502 		/*
503 		 * Adjust for the current time slice.  This is actually fairly
504 		 * important since the error here is on the order of a time
505 		 * quantum, which is much greater than the sampling error.
506 		 */
507 		microtime(&tv);
508 		sec += tv.tv_sec - runtime.tv_sec;
509 		usec += tv.tv_usec - runtime.tv_usec;
510 	}
511 	totusec = (quad_t)sec * 1000000 + usec;
512 	if (totusec < 0) {
513 #ifndef SMP	/* sigh, microtime and fork/exit madness here */
514 		/* XXX no %qd in kernel.  Truncate. */
515 		printf("calcru: negative time: %ld usec\n", (long)totusec);
516 #endif
517 		totusec = 0;
518 	}
519 	u = totusec;
520 	st = (u * st) / tot;
521 	sp->tv_sec = st / 1000000;
522 	sp->tv_usec = st % 1000000;
523 	ut = (u * ut) / tot;
524 	up->tv_sec = ut / 1000000;
525 	up->tv_usec = ut % 1000000;
526 	if (ip != NULL) {
527 		it = (u * it) / tot;
528 		ip->tv_sec = it / 1000000;
529 		ip->tv_usec = it % 1000000;
530 	}
531 }
532 
533 #ifndef _SYS_SYSPROTO_H_
534 struct getrusage_args {
535 	int	who;
536 	struct	rusage *rusage;
537 };
538 #endif
539 /* ARGSUSED */
540 int
541 getrusage(p, uap, retval)
542 	register struct proc *p;
543 	register struct getrusage_args *uap;
544 	int *retval;
545 {
546 	register struct rusage *rup;
547 
548 	switch (uap->who) {
549 
550 	case RUSAGE_SELF:
551 		rup = &p->p_stats->p_ru;
552 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
553 		break;
554 
555 	case RUSAGE_CHILDREN:
556 		rup = &p->p_stats->p_cru;
557 		break;
558 
559 	default:
560 		return (EINVAL);
561 	}
562 	return (copyout((caddr_t)rup, (caddr_t)uap->rusage,
563 	    sizeof (struct rusage)));
564 }
565 
566 void
567 ruadd(ru, ru2)
568 	register struct rusage *ru, *ru2;
569 {
570 	register long *ip, *ip2;
571 	register int i;
572 
573 	timevaladd(&ru->ru_utime, &ru2->ru_utime);
574 	timevaladd(&ru->ru_stime, &ru2->ru_stime);
575 	if (ru->ru_maxrss < ru2->ru_maxrss)
576 		ru->ru_maxrss = ru2->ru_maxrss;
577 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
578 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
579 		*ip++ += *ip2++;
580 }
581 
582 /*
583  * Make a copy of the plimit structure.
584  * We share these structures copy-on-write after fork,
585  * and copy when a limit is changed.
586  */
587 struct plimit *
588 limcopy(lim)
589 	struct plimit *lim;
590 {
591 	register struct plimit *copy;
592 
593 	MALLOC(copy, struct plimit *, sizeof(struct plimit),
594 	    M_SUBPROC, M_WAITOK);
595 	bcopy(lim->pl_rlimit, copy->pl_rlimit,
596 	    sizeof(struct rlimit) * RLIM_NLIMITS);
597 	copy->p_lflags = 0;
598 	copy->p_refcnt = 1;
599 	return (copy);
600 }
601