xref: /freebsd/sys/kern/kern_resource.c (revision e627b39baccd1ec9129690167cf5e6d860509655)
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 1.19 1996/03/11 06:04:20 hsu Exp $
40  */
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/sysproto.h>
45 #include <sys/kernel.h>
46 #include <sys/file.h>
47 #include <sys/resourcevar.h>
48 #include <sys/malloc.h>
49 #include <sys/proc.h>
50 
51 #include <vm/vm.h>
52 #include <vm/vm_param.h>
53 #include <vm/vm_prot.h>
54 #include <vm/lock.h>
55 #include <vm/pmap.h>
56 #include <vm/vm_map.h>
57 
58 int	donice __P((struct proc *curp, struct proc *chgp, int n));
59 int	dosetrlimit __P((struct proc *p, u_int which, struct rlimit *limp));
60 
61 /*
62  * Resource controls and accounting.
63  */
64 
65 #ifndef _SYS_SYSPROTO_H_
66 struct getpriority_args {
67 	int	which;
68 	int	who;
69 };
70 #endif
71 int
72 getpriority(curp, uap, retval)
73 	struct proc *curp;
74 	register struct getpriority_args *uap;
75 	int *retval;
76 {
77 	register struct proc *p;
78 	register int low = PRIO_MAX + 1;
79 
80 	switch (uap->which) {
81 
82 	case PRIO_PROCESS:
83 		if (uap->who == 0)
84 			p = curp;
85 		else
86 			p = pfind(uap->who);
87 		if (p == 0)
88 			break;
89 		low = p->p_nice;
90 		break;
91 
92 	case PRIO_PGRP: {
93 		register struct pgrp *pg;
94 
95 		if (uap->who == 0)
96 			pg = curp->p_pgrp;
97 		else if ((pg = pgfind(uap->who)) == NULL)
98 			break;
99 		for (p = pg->pg_members.lh_first; p != 0;
100 		     p = p->p_pglist.le_next) {
101 			if (p->p_nice < low)
102 				low = p->p_nice;
103 		}
104 		break;
105 	}
106 
107 	case PRIO_USER:
108 		if (uap->who == 0)
109 			uap->who = curp->p_ucred->cr_uid;
110 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
111 			if (p->p_ucred->cr_uid == uap->who &&
112 			    p->p_nice < low)
113 				low = p->p_nice;
114 		break;
115 
116 	default:
117 		return (EINVAL);
118 	}
119 	if (low == PRIO_MAX + 1)
120 		return (ESRCH);
121 	*retval = low;
122 	return (0);
123 }
124 
125 #ifndef _SYS_SYSPROTO_H_
126 struct setpriority_args {
127 	int	which;
128 	int	who;
129 	int	prio;
130 };
131 #endif
132 /* ARGSUSED */
133 int
134 setpriority(curp, uap, retval)
135 	struct proc *curp;
136 	register struct setpriority_args *uap;
137 	int *retval;
138 {
139 	register struct proc *p;
140 	int found = 0, error = 0;
141 
142 	switch (uap->which) {
143 
144 	case PRIO_PROCESS:
145 		if (uap->who == 0)
146 			p = curp;
147 		else
148 			p = pfind(uap->who);
149 		if (p == 0)
150 			break;
151 		error = donice(curp, p, uap->prio);
152 		found++;
153 		break;
154 
155 	case PRIO_PGRP: {
156 		register struct pgrp *pg;
157 
158 		if (uap->who == 0)
159 			pg = curp->p_pgrp;
160 		else if ((pg = pgfind(uap->who)) == NULL)
161 			break;
162 		for (p = pg->pg_members.lh_first; p != 0;
163 		    p = p->p_pglist.le_next) {
164 			error = donice(curp, p, uap->prio);
165 			found++;
166 		}
167 		break;
168 	}
169 
170 	case PRIO_USER:
171 		if (uap->who == 0)
172 			uap->who = curp->p_ucred->cr_uid;
173 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
174 			if (p->p_ucred->cr_uid == uap->who) {
175 				error = donice(curp, p, uap->prio);
176 				found++;
177 			}
178 		break;
179 
180 	default:
181 		return (EINVAL);
182 	}
183 	if (found == 0)
184 		return (ESRCH);
185 	return (error);
186 }
187 
188 int
189 donice(curp, chgp, n)
190 	register struct proc *curp, *chgp;
191 	register int n;
192 {
193 	register struct pcred *pcred = curp->p_cred;
194 
195 	if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
196 	    pcred->pc_ucred->cr_uid != chgp->p_ucred->cr_uid &&
197 	    pcred->p_ruid != chgp->p_ucred->cr_uid)
198 		return (EPERM);
199 	if (n > PRIO_MAX)
200 		n = PRIO_MAX;
201 	if (n < PRIO_MIN)
202 		n = PRIO_MIN;
203 	if (n < chgp->p_nice && suser(pcred->pc_ucred, &curp->p_acflag))
204 		return (EACCES);
205 	chgp->p_nice = n;
206 	(void)resetpriority(chgp);
207 	return (0);
208 }
209 
210 /* rtprio system call */
211 #ifndef _SYS_SYSPROTO_H_
212 struct rtprio_args {
213 	int		function;
214 	pid_t		pid;
215 	struct rtprio	*rtp;
216 };
217 #endif
218 
219 /*
220  * Set realtime priority
221  */
222 
223 /* ARGSUSED */
224 int
225 rtprio(curp, uap, retval)
226 	struct proc *curp;
227 	register struct rtprio_args *uap;
228 	int *retval;
229 {
230 	register struct proc *p;
231 	register struct pcred *pcred = curp->p_cred;
232 	struct rtprio rtp;
233 	int error;
234 
235 	error = copyin(uap->rtp, &rtp, sizeof(struct rtprio));
236 	if (error)
237 		return (error);
238 
239 	if (uap->pid == 0)
240 		p = curp;
241 	else
242 		p = pfind(uap->pid);
243 
244 	if (p == 0)
245 		return (ESRCH);
246 
247 	switch (uap->function) {
248 	case RTP_LOOKUP:
249 		return (copyout(&p->p_rtprio, uap->rtp, sizeof(struct rtprio)));
250 	case RTP_SET:
251 		if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
252 		    pcred->pc_ucred->cr_uid != p->p_ucred->cr_uid &&
253 		    pcred->p_ruid != p->p_ucred->cr_uid)
254 		        return (EPERM);
255 		/* disallow setting rtprio in most cases if not superuser */
256 		if (suser(pcred->pc_ucred, &curp->p_acflag)) {
257 			/* can't set someone else's */
258 			if (uap->pid)
259 				return (EPERM);
260 			/* can't set realtime priority */
261 			if (rtp.type == RTP_PRIO_REALTIME)
262 				return (EPERM);
263 		}
264 		switch (rtp.type) {
265 		case RTP_PRIO_REALTIME:
266 		case RTP_PRIO_NORMAL:
267 		case RTP_PRIO_IDLE:
268 			if (rtp.prio > RTP_PRIO_MAX)
269 				return (EINVAL);
270 			p->p_rtprio = rtp;
271 			return (0);
272 		default:
273 			return (EINVAL);
274 		}
275 
276 	default:
277 		return (EINVAL);
278 	}
279 }
280 
281 #if defined(COMPAT_43) || defined(COMPAT_SUNOS)
282 #ifndef _SYS_SYSPROTO_H_
283 struct osetrlimit_args {
284 	u_int	which;
285 	struct	orlimit *rlp;
286 };
287 #endif
288 /* ARGSUSED */
289 int
290 osetrlimit(p, uap, retval)
291 	struct proc *p;
292 	register struct osetrlimit_args *uap;
293 	int *retval;
294 {
295 	struct orlimit olim;
296 	struct rlimit lim;
297 	int error;
298 
299 	if ((error =
300 	    copyin((caddr_t)uap->rlp, (caddr_t)&olim, sizeof(struct orlimit))))
301 		return (error);
302 	lim.rlim_cur = olim.rlim_cur;
303 	lim.rlim_max = olim.rlim_max;
304 	return (dosetrlimit(p, uap->which, &lim));
305 }
306 
307 #ifndef _SYS_SYSPROTO_H_
308 struct ogetrlimit_args {
309 	u_int	which;
310 	struct	orlimit *rlp;
311 };
312 #endif
313 /* ARGSUSED */
314 int
315 ogetrlimit(p, uap, retval)
316 	struct proc *p;
317 	register struct ogetrlimit_args *uap;
318 	int *retval;
319 {
320 	struct orlimit olim;
321 
322 	if (uap->which >= RLIM_NLIMITS)
323 		return (EINVAL);
324 	olim.rlim_cur = p->p_rlimit[uap->which].rlim_cur;
325 	if (olim.rlim_cur == -1)
326 		olim.rlim_cur = 0x7fffffff;
327 	olim.rlim_max = p->p_rlimit[uap->which].rlim_max;
328 	if (olim.rlim_max == -1)
329 		olim.rlim_max = 0x7fffffff;
330 	return (copyout((caddr_t)&olim, (caddr_t)uap->rlp, sizeof(olim)));
331 }
332 #endif /* COMPAT_43 || COMPAT_SUNOS */
333 
334 #ifndef _SYS_SYSPROTO_H_
335 struct __setrlimit_args {
336 	u_int	which;
337 	struct	rlimit *rlp;
338 };
339 #endif
340 /* ARGSUSED */
341 int
342 setrlimit(p, uap, retval)
343 	struct proc *p;
344 	register struct __setrlimit_args *uap;
345 	int *retval;
346 {
347 	struct rlimit alim;
348 	int error;
349 
350 	if ((error =
351 	    copyin((caddr_t)uap->rlp, (caddr_t)&alim, sizeof (struct rlimit))))
352 		return (error);
353 	return (dosetrlimit(p, uap->which, &alim));
354 }
355 
356 int
357 dosetrlimit(p, which, limp)
358 	struct proc *p;
359 	u_int which;
360 	struct rlimit *limp;
361 {
362 	register struct rlimit *alimp;
363 	int error;
364 
365 	if (which >= RLIM_NLIMITS)
366 		return (EINVAL);
367 	alimp = &p->p_rlimit[which];
368 
369 	/*
370 	 * Preserve historical bugs by treating negative limits as unsigned.
371 	 */
372 	if (limp->rlim_cur < 0)
373 		limp->rlim_cur = RLIM_INFINITY;
374 	if (limp->rlim_max < 0)
375 		limp->rlim_max = RLIM_INFINITY;
376 
377 	if (limp->rlim_cur > alimp->rlim_max ||
378 	    limp->rlim_max > alimp->rlim_max)
379 		if ((error = suser(p->p_ucred, &p->p_acflag)))
380 			return (error);
381 	if (limp->rlim_cur > limp->rlim_max)
382 		limp->rlim_cur = limp->rlim_max;
383 	if (p->p_limit->p_refcnt > 1 &&
384 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
385 		p->p_limit->p_refcnt--;
386 		p->p_limit = limcopy(p->p_limit);
387 		alimp = &p->p_rlimit[which];
388 	}
389 
390 	switch (which) {
391 
392 	case RLIMIT_DATA:
393 		if (limp->rlim_cur > MAXDSIZ)
394 			limp->rlim_cur = MAXDSIZ;
395 		if (limp->rlim_max > MAXDSIZ)
396 			limp->rlim_max = MAXDSIZ;
397 		break;
398 
399 	case RLIMIT_STACK:
400 		if (limp->rlim_cur > MAXSSIZ)
401 			limp->rlim_cur = MAXSSIZ;
402 		if (limp->rlim_max > MAXSSIZ)
403 			limp->rlim_max = MAXSSIZ;
404 		/*
405 		 * Stack is allocated to the max at exec time with only
406 		 * "rlim_cur" bytes accessible.  If stack limit is going
407 		 * up make more accessible, if going down make inaccessible.
408 		 */
409 		if (limp->rlim_cur != alimp->rlim_cur) {
410 			vm_offset_t addr;
411 			vm_size_t size;
412 			vm_prot_t prot;
413 
414 			if (limp->rlim_cur > alimp->rlim_cur) {
415 				prot = VM_PROT_ALL;
416 				size = limp->rlim_cur - alimp->rlim_cur;
417 				addr = USRSTACK - limp->rlim_cur;
418 			} else {
419 				prot = VM_PROT_NONE;
420 				size = alimp->rlim_cur - limp->rlim_cur;
421 				addr = USRSTACK - alimp->rlim_cur;
422 			}
423 			addr = trunc_page(addr);
424 			size = round_page(size);
425 			(void) vm_map_protect(&p->p_vmspace->vm_map,
426 					      addr, addr+size, prot, FALSE);
427 		}
428 		break;
429 
430 	case RLIMIT_NOFILE:
431 		if (limp->rlim_cur > maxfilesperproc)
432 			limp->rlim_cur = maxfilesperproc;
433 		if (limp->rlim_max > maxfilesperproc)
434 			limp->rlim_max = maxfilesperproc;
435 		break;
436 
437 	case RLIMIT_NPROC:
438 		if (limp->rlim_cur > maxprocperuid)
439 			limp->rlim_cur = maxprocperuid;
440 		if (limp->rlim_max > maxprocperuid)
441 			limp->rlim_max = maxprocperuid;
442 		break;
443 	}
444 	*alimp = *limp;
445 	return (0);
446 }
447 
448 #ifndef _SYS_SYSPROTO_H_
449 struct __getrlimit_args {
450 	u_int	which;
451 	struct	rlimit *rlp;
452 };
453 #endif
454 /* ARGSUSED */
455 int
456 getrlimit(p, uap, retval)
457 	struct proc *p;
458 	register struct __getrlimit_args *uap;
459 	int *retval;
460 {
461 
462 	if (uap->which >= RLIM_NLIMITS)
463 		return (EINVAL);
464 	return (copyout((caddr_t)&p->p_rlimit[uap->which], (caddr_t)uap->rlp,
465 	    sizeof (struct rlimit)));
466 }
467 
468 /*
469  * Transform the running time and tick information in proc p into user,
470  * system, and interrupt time usage.
471  */
472 void
473 calcru(p, up, sp, ip)
474 	struct proc *p;
475 	struct timeval *up;
476 	struct timeval *sp;
477 	struct timeval *ip;
478 {
479 	quad_t totusec;
480 	u_quad_t u, st, ut, it, tot;
481 	long sec, usec;
482 	int s;
483 	struct timeval tv;
484 
485 	s = splstatclock();
486 	st = p->p_sticks;
487 	ut = p->p_uticks;
488 	it = p->p_iticks;
489 	splx(s);
490 
491 	tot = st + ut + it;
492 	if (tot == 0) {
493 		st = 1;
494 		tot = 1;
495 	}
496 
497 	sec = p->p_rtime.tv_sec;
498 	usec = p->p_rtime.tv_usec;
499 	if (p == curproc) {
500 		/*
501 		 * Adjust for the current time slice.  This is actually fairly
502 		 * important since the error here is on the order of a time
503 		 * quantum, which is much greater than the sampling error.
504 		 */
505 		microtime(&tv);
506 		sec += tv.tv_sec - runtime.tv_sec;
507 		usec += tv.tv_usec - runtime.tv_usec;
508 	}
509 	totusec = (quad_t)sec * 1000000 + usec;
510 	if (totusec < 0) {
511 		/* XXX no %qd in kernel.  Truncate. */
512 		printf("calcru: negative time: %ld usec\n", (long)totusec);
513 		totusec = 0;
514 	}
515 	u = totusec;
516 	st = (u * st) / tot;
517 	sp->tv_sec = st / 1000000;
518 	sp->tv_usec = st % 1000000;
519 	ut = (u * ut) / tot;
520 	up->tv_sec = ut / 1000000;
521 	up->tv_usec = ut % 1000000;
522 	if (ip != NULL) {
523 		it = (u * it) / tot;
524 		ip->tv_sec = it / 1000000;
525 		ip->tv_usec = it % 1000000;
526 	}
527 }
528 
529 #ifndef _SYS_SYSPROTO_H_
530 struct getrusage_args {
531 	int	who;
532 	struct	rusage *rusage;
533 };
534 #endif
535 /* ARGSUSED */
536 int
537 getrusage(p, uap, retval)
538 	register struct proc *p;
539 	register struct getrusage_args *uap;
540 	int *retval;
541 {
542 	register struct rusage *rup;
543 
544 	switch (uap->who) {
545 
546 	case RUSAGE_SELF:
547 		rup = &p->p_stats->p_ru;
548 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
549 		break;
550 
551 	case RUSAGE_CHILDREN:
552 		rup = &p->p_stats->p_cru;
553 		break;
554 
555 	default:
556 		return (EINVAL);
557 	}
558 	return (copyout((caddr_t)rup, (caddr_t)uap->rusage,
559 	    sizeof (struct rusage)));
560 }
561 
562 void
563 ruadd(ru, ru2)
564 	register struct rusage *ru, *ru2;
565 {
566 	register long *ip, *ip2;
567 	register int i;
568 
569 	timevaladd(&ru->ru_utime, &ru2->ru_utime);
570 	timevaladd(&ru->ru_stime, &ru2->ru_stime);
571 	if (ru->ru_maxrss < ru2->ru_maxrss)
572 		ru->ru_maxrss = ru2->ru_maxrss;
573 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
574 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
575 		*ip++ += *ip2++;
576 }
577 
578 /*
579  * Make a copy of the plimit structure.
580  * We share these structures copy-on-write after fork,
581  * and copy when a limit is changed.
582  */
583 struct plimit *
584 limcopy(lim)
585 	struct plimit *lim;
586 {
587 	register struct plimit *copy;
588 
589 	MALLOC(copy, struct plimit *, sizeof(struct plimit),
590 	    M_SUBPROC, M_WAITOK);
591 	bcopy(lim->pl_rlimit, copy->pl_rlimit,
592 	    sizeof(struct rlimit) * RLIM_NLIMITS);
593 	copy->p_lflags = 0;
594 	copy->p_refcnt = 1;
595 	return (copy);
596 }
597