xref: /freebsd/sys/kern/kern_resource.c (revision ef5d438ed4bc17ad7ece3e40fe4d1f9baf3aadf7)
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 1.17 1995/12/07 12:46:48 davidg 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 *, struct proc *, int));
59 int	dosetrlimit __P((struct proc *, u_int, struct rlimit *));
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_mem; p != NULL; p = p->p_pgrpnxt) {
100 			if (p->p_nice < low)
101 				low = p->p_nice;
102 		}
103 		break;
104 	}
105 
106 	case PRIO_USER:
107 		if (uap->who == 0)
108 			uap->who = curp->p_ucred->cr_uid;
109 		for (p = (struct proc *)allproc; p != NULL; p = p->p_next) {
110 			if (p->p_ucred->cr_uid == uap->who &&
111 			    p->p_nice < low)
112 				low = p->p_nice;
113 		}
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_mem; p != NULL; p = p->p_pgrpnxt) {
163 			error = donice(curp, p, uap->prio);
164 			found++;
165 		}
166 		break;
167 	}
168 
169 	case PRIO_USER:
170 		if (uap->who == 0)
171 			uap->who = curp->p_ucred->cr_uid;
172 		for (p = (struct proc *)allproc; p != NULL; p = p->p_next)
173 			if (p->p_ucred->cr_uid == uap->who) {
174 				error = donice(curp, p, uap->prio);
175 				found++;
176 			}
177 		break;
178 
179 	default:
180 		return (EINVAL);
181 	}
182 	if (found == 0)
183 		return (ESRCH);
184 	return (error);
185 }
186 
187 int
188 donice(curp, chgp, n)
189 	register struct proc *curp, *chgp;
190 	register int n;
191 {
192 	register struct pcred *pcred = curp->p_cred;
193 
194 	if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
195 	    pcred->pc_ucred->cr_uid != chgp->p_ucred->cr_uid &&
196 	    pcred->p_ruid != chgp->p_ucred->cr_uid)
197 		return (EPERM);
198 	if (n > PRIO_MAX)
199 		n = PRIO_MAX;
200 	if (n < PRIO_MIN)
201 		n = PRIO_MIN;
202 	if (n < chgp->p_nice && suser(pcred->pc_ucred, &curp->p_acflag))
203 		return (EACCES);
204 	chgp->p_nice = n;
205 	(void)resetpriority(chgp);
206 	return (0);
207 }
208 
209 /* rtprio system call */
210 #ifndef _SYS_SYSPROTO_H_
211 struct rtprio_args {
212 	int		function;
213 	pid_t		pid;
214 	struct rtprio	*rtp;
215 };
216 #endif
217 
218 /*
219  * Set realtime priority
220  */
221 
222 /* ARGSUSED */
223 int
224 rtprio(curp, uap, retval)
225 	struct proc *curp;
226 	register struct rtprio_args *uap;
227 	int *retval;
228 {
229 	register struct proc *p;
230 	register struct pcred *pcred = curp->p_cred;
231 	struct rtprio rtp;
232 	int error;
233 
234 	error = copyin(uap->rtp, &rtp, sizeof(struct rtprio));
235 	if (error)
236 		return (error);
237 
238 	if (uap->pid == 0)
239 		p = curp;
240 	else
241 		p = pfind(uap->pid);
242 
243 	if (p == 0)
244 		return (ESRCH);
245 
246 	switch (uap->function) {
247 	case RTP_LOOKUP:
248 		return (copyout(&p->p_rtprio, uap->rtp, sizeof(struct rtprio)));
249 	case RTP_SET:
250 		if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
251 		    pcred->pc_ucred->cr_uid != p->p_ucred->cr_uid &&
252 		    pcred->p_ruid != p->p_ucred->cr_uid)
253 		        return (EPERM);
254 		/* disallow setting rtprio in most cases if not superuser */
255 		if (suser(pcred->pc_ucred, &curp->p_acflag)) {
256 			/* can't set someone else's */
257 			if (uap->pid)
258 				return (EPERM);
259 			/* can't set realtime priority */
260 			if (rtp.type == RTP_PRIO_REALTIME)
261 				return (EPERM);
262 		}
263 		switch (rtp.type) {
264 		case RTP_PRIO_REALTIME:
265 		case RTP_PRIO_NORMAL:
266 		case RTP_PRIO_IDLE:
267 			if (rtp.prio > RTP_PRIO_MAX)
268 				return (EINVAL);
269 			p->p_rtprio = rtp;
270 			return (0);
271 		default:
272 			return (EINVAL);
273 		}
274 
275 	default:
276 		return (EINVAL);
277 	}
278 }
279 
280 #if defined(COMPAT_43) || defined(COMPAT_SUNOS)
281 #ifndef _SYS_SYSPROTO_H_
282 struct osetrlimit_args {
283 	u_int	which;
284 	struct	orlimit *rlp;
285 };
286 #endif
287 /* ARGSUSED */
288 int
289 osetrlimit(p, uap, retval)
290 	struct proc *p;
291 	register struct osetrlimit_args *uap;
292 	int *retval;
293 {
294 	struct orlimit olim;
295 	struct rlimit lim;
296 	int error;
297 
298 	if ((error =
299 	    copyin((caddr_t)uap->rlp, (caddr_t)&olim, sizeof(struct orlimit))))
300 		return (error);
301 	lim.rlim_cur = olim.rlim_cur;
302 	lim.rlim_max = olim.rlim_max;
303 	return (dosetrlimit(p, uap->which, &lim));
304 }
305 
306 #ifndef _SYS_SYSPROTO_H_
307 struct ogetrlimit_args {
308 	u_int	which;
309 	struct	orlimit *rlp;
310 };
311 #endif
312 /* ARGSUSED */
313 int
314 ogetrlimit(p, uap, retval)
315 	struct proc *p;
316 	register struct ogetrlimit_args *uap;
317 	int *retval;
318 {
319 	struct orlimit olim;
320 
321 	if (uap->which >= RLIM_NLIMITS)
322 		return (EINVAL);
323 	olim.rlim_cur = p->p_rlimit[uap->which].rlim_cur;
324 	if (olim.rlim_cur == -1)
325 		olim.rlim_cur = 0x7fffffff;
326 	olim.rlim_max = p->p_rlimit[uap->which].rlim_max;
327 	if (olim.rlim_max == -1)
328 		olim.rlim_max = 0x7fffffff;
329 	return (copyout((caddr_t)&olim, (caddr_t)uap->rlp, sizeof(olim)));
330 }
331 #endif /* COMPAT_43 || COMPAT_SUNOS */
332 
333 #ifndef _SYS_SYSPROTO_H_
334 struct __setrlimit_args {
335 	u_int	which;
336 	struct	rlimit *rlp;
337 };
338 #endif
339 /* ARGSUSED */
340 int
341 setrlimit(p, uap, retval)
342 	struct proc *p;
343 	register struct __setrlimit_args *uap;
344 	int *retval;
345 {
346 	struct rlimit alim;
347 	int error;
348 
349 	if ((error =
350 	    copyin((caddr_t)uap->rlp, (caddr_t)&alim, sizeof (struct rlimit))))
351 		return (error);
352 	return (dosetrlimit(p, uap->which, &alim));
353 }
354 
355 int
356 dosetrlimit(p, which, limp)
357 	struct proc *p;
358 	u_int which;
359 	struct rlimit *limp;
360 {
361 	register struct rlimit *alimp;
362 	int error;
363 
364 	if (which >= RLIM_NLIMITS)
365 		return (EINVAL);
366 	alimp = &p->p_rlimit[which];
367 
368 	/*
369 	 * Preserve historical bugs by treating negative limits as unsigned.
370 	 */
371 	if (limp->rlim_cur < 0)
372 		limp->rlim_cur = RLIM_INFINITY;
373 	if (limp->rlim_max < 0)
374 		limp->rlim_max = RLIM_INFINITY;
375 
376 	if (limp->rlim_cur > alimp->rlim_max ||
377 	    limp->rlim_max > alimp->rlim_max)
378 		if ((error = suser(p->p_ucred, &p->p_acflag)))
379 			return (error);
380 	if (limp->rlim_cur > limp->rlim_max)
381 		limp->rlim_cur = limp->rlim_max;
382 	if (p->p_limit->p_refcnt > 1 &&
383 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
384 		p->p_limit->p_refcnt--;
385 		p->p_limit = limcopy(p->p_limit);
386 		alimp = &p->p_rlimit[which];
387 	}
388 
389 	switch (which) {
390 
391 	case RLIMIT_DATA:
392 		if (limp->rlim_cur > MAXDSIZ)
393 			limp->rlim_cur = MAXDSIZ;
394 		if (limp->rlim_max > MAXDSIZ)
395 			limp->rlim_max = MAXDSIZ;
396 		break;
397 
398 	case RLIMIT_STACK:
399 		if (limp->rlim_cur > MAXSSIZ)
400 			limp->rlim_cur = MAXSSIZ;
401 		if (limp->rlim_max > MAXSSIZ)
402 			limp->rlim_max = MAXSSIZ;
403 		/*
404 		 * Stack is allocated to the max at exec time with only
405 		 * "rlim_cur" bytes accessible.  If stack limit is going
406 		 * up make more accessible, if going down make inaccessible.
407 		 */
408 		if (limp->rlim_cur != alimp->rlim_cur) {
409 			vm_offset_t addr;
410 			vm_size_t size;
411 			vm_prot_t prot;
412 
413 			if (limp->rlim_cur > alimp->rlim_cur) {
414 				prot = VM_PROT_ALL;
415 				size = limp->rlim_cur - alimp->rlim_cur;
416 				addr = USRSTACK - limp->rlim_cur;
417 			} else {
418 				prot = VM_PROT_NONE;
419 				size = alimp->rlim_cur - limp->rlim_cur;
420 				addr = USRSTACK - alimp->rlim_cur;
421 			}
422 			addr = trunc_page(addr);
423 			size = round_page(size);
424 			(void) vm_map_protect(&p->p_vmspace->vm_map,
425 					      addr, addr+size, prot, FALSE);
426 		}
427 		break;
428 
429 	case RLIMIT_NOFILE:
430 		if (limp->rlim_cur > maxfilesperproc)
431 			limp->rlim_cur = maxfilesperproc;
432 		if (limp->rlim_max > maxfilesperproc)
433 			limp->rlim_max = maxfilesperproc;
434 		break;
435 
436 	case RLIMIT_NPROC:
437 		if (limp->rlim_cur > maxprocperuid)
438 			limp->rlim_cur = maxprocperuid;
439 		if (limp->rlim_max > maxprocperuid)
440 			limp->rlim_max = maxprocperuid;
441 		break;
442 	}
443 	*alimp = *limp;
444 	return (0);
445 }
446 
447 #ifndef _SYS_SYSPROTO_H_
448 struct __getrlimit_args {
449 	u_int	which;
450 	struct	rlimit *rlp;
451 };
452 #endif
453 /* ARGSUSED */
454 int
455 getrlimit(p, uap, retval)
456 	struct proc *p;
457 	register struct __getrlimit_args *uap;
458 	int *retval;
459 {
460 
461 	if (uap->which >= RLIM_NLIMITS)
462 		return (EINVAL);
463 	return (copyout((caddr_t)&p->p_rlimit[uap->which], (caddr_t)uap->rlp,
464 	    sizeof (struct rlimit)));
465 }
466 
467 /*
468  * Transform the running time and tick information in proc p into user,
469  * system, and interrupt time usage.
470  */
471 void
472 calcru(p, up, sp, ip)
473 	struct proc *p;
474 	struct timeval *up;
475 	struct timeval *sp;
476 	struct timeval *ip;
477 {
478 	quad_t totusec;
479 	u_quad_t u, st, ut, it, tot;
480 	long sec, usec;
481 	int s;
482 	struct timeval tv;
483 
484 	s = splstatclock();
485 	st = p->p_sticks;
486 	ut = p->p_uticks;
487 	it = p->p_iticks;
488 	splx(s);
489 
490 	tot = st + ut + it;
491 	if (tot == 0) {
492 		up->tv_sec = up->tv_usec = 0;
493 		sp->tv_sec = sp->tv_usec = 0;
494 		if (ip != NULL)
495 			ip->tv_sec = ip->tv_usec = 0;
496 		return;
497 	}
498 
499 	sec = p->p_rtime.tv_sec;
500 	usec = p->p_rtime.tv_usec;
501 	if (p == curproc) {
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 		printf("calcru: negative time: %d usec\n", totusec);
514 		totusec = 0;
515 	}
516 	u = totusec;
517 	st = (u * st) / tot;
518 	sp->tv_sec = st / 1000000;
519 	sp->tv_usec = st % 1000000;
520 	ut = (u * ut) / tot;
521 	up->tv_sec = ut / 1000000;
522 	up->tv_usec = ut % 1000000;
523 	if (ip != NULL) {
524 		it = (u * it) / tot;
525 		ip->tv_sec = it / 1000000;
526 		ip->tv_usec = it % 1000000;
527 	}
528 }
529 
530 #ifndef _SYS_SYSPROTO_H_
531 struct getrusage_args {
532 	int	who;
533 	struct	rusage *rusage;
534 };
535 #endif
536 /* ARGSUSED */
537 int
538 getrusage(p, uap, retval)
539 	register struct proc *p;
540 	register struct getrusage_args *uap;
541 	int *retval;
542 {
543 	register struct rusage *rup;
544 
545 	switch (uap->who) {
546 
547 	case RUSAGE_SELF:
548 		rup = &p->p_stats->p_ru;
549 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
550 		break;
551 
552 	case RUSAGE_CHILDREN:
553 		rup = &p->p_stats->p_cru;
554 		break;
555 
556 	default:
557 		return (EINVAL);
558 	}
559 	return (copyout((caddr_t)rup, (caddr_t)uap->rusage,
560 	    sizeof (struct rusage)));
561 }
562 
563 void
564 ruadd(ru, ru2)
565 	register struct rusage *ru, *ru2;
566 {
567 	register long *ip, *ip2;
568 	register int i;
569 
570 	timevaladd(&ru->ru_utime, &ru2->ru_utime);
571 	timevaladd(&ru->ru_stime, &ru2->ru_stime);
572 	if (ru->ru_maxrss < ru2->ru_maxrss)
573 		ru->ru_maxrss = ru2->ru_maxrss;
574 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
575 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
576 		*ip++ += *ip2++;
577 }
578 
579 /*
580  * Make a copy of the plimit structure.
581  * We share these structures copy-on-write after fork,
582  * and copy when a limit is changed.
583  */
584 struct plimit *
585 limcopy(lim)
586 	struct plimit *lim;
587 {
588 	register struct plimit *copy;
589 
590 	MALLOC(copy, struct plimit *, sizeof(struct plimit),
591 	    M_SUBPROC, M_WAITOK);
592 	bcopy(lim->pl_rlimit, copy->pl_rlimit,
593 	    sizeof(struct rlimit) * RLIM_NLIMITS);
594 	copy->p_lflags = 0;
595 	copy->p_refcnt = 1;
596 	return (copy);
597 }
598