xref: /freebsd/sys/kern/kern_resource.c (revision 0ea3482342b4d7d6e71f3007ce4dafe445c639fd)
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 1.15 1995/11/11 01:48:17 bde 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 
53 int	donice __P((struct proc *, struct proc *, int));
54 int	dosetrlimit __P((struct proc *, u_int, struct rlimit *));
55 
56 /*
57  * Resource controls and accounting.
58  */
59 
60 #ifndef _SYS_SYSPROTO_H_
61 struct getpriority_args {
62 	int	which;
63 	int	who;
64 };
65 #endif
66 int
67 getpriority(curp, uap, retval)
68 	struct proc *curp;
69 	register struct getpriority_args *uap;
70 	int *retval;
71 {
72 	register struct proc *p;
73 	register int low = PRIO_MAX + 1;
74 
75 	switch (uap->which) {
76 
77 	case PRIO_PROCESS:
78 		if (uap->who == 0)
79 			p = curp;
80 		else
81 			p = pfind(uap->who);
82 		if (p == 0)
83 			break;
84 		low = p->p_nice;
85 		break;
86 
87 	case PRIO_PGRP: {
88 		register struct pgrp *pg;
89 
90 		if (uap->who == 0)
91 			pg = curp->p_pgrp;
92 		else if ((pg = pgfind(uap->who)) == NULL)
93 			break;
94 		for (p = pg->pg_mem; p != NULL; p = p->p_pgrpnxt) {
95 			if (p->p_nice < low)
96 				low = p->p_nice;
97 		}
98 		break;
99 	}
100 
101 	case PRIO_USER:
102 		if (uap->who == 0)
103 			uap->who = curp->p_ucred->cr_uid;
104 		for (p = (struct proc *)allproc; p != NULL; p = p->p_next) {
105 			if (p->p_ucred->cr_uid == uap->who &&
106 			    p->p_nice < low)
107 				low = p->p_nice;
108 		}
109 		break;
110 
111 	default:
112 		return (EINVAL);
113 	}
114 	if (low == PRIO_MAX + 1)
115 		return (ESRCH);
116 	*retval = low;
117 	return (0);
118 }
119 
120 #ifndef _SYS_SYSPROTO_H_
121 struct setpriority_args {
122 	int	which;
123 	int	who;
124 	int	prio;
125 };
126 #endif
127 /* ARGSUSED */
128 int
129 setpriority(curp, uap, retval)
130 	struct proc *curp;
131 	register struct setpriority_args *uap;
132 	int *retval;
133 {
134 	register struct proc *p;
135 	int found = 0, error = 0;
136 
137 	switch (uap->which) {
138 
139 	case PRIO_PROCESS:
140 		if (uap->who == 0)
141 			p = curp;
142 		else
143 			p = pfind(uap->who);
144 		if (p == 0)
145 			break;
146 		error = donice(curp, p, uap->prio);
147 		found++;
148 		break;
149 
150 	case PRIO_PGRP: {
151 		register struct pgrp *pg;
152 
153 		if (uap->who == 0)
154 			pg = curp->p_pgrp;
155 		else if ((pg = pgfind(uap->who)) == NULL)
156 			break;
157 		for (p = pg->pg_mem; p != NULL; p = p->p_pgrpnxt) {
158 			error = donice(curp, p, uap->prio);
159 			found++;
160 		}
161 		break;
162 	}
163 
164 	case PRIO_USER:
165 		if (uap->who == 0)
166 			uap->who = curp->p_ucred->cr_uid;
167 		for (p = (struct proc *)allproc; p != NULL; p = p->p_next)
168 			if (p->p_ucred->cr_uid == uap->who) {
169 				error = donice(curp, p, uap->prio);
170 				found++;
171 			}
172 		break;
173 
174 	default:
175 		return (EINVAL);
176 	}
177 	if (found == 0)
178 		return (ESRCH);
179 	return (error);
180 }
181 
182 int
183 donice(curp, chgp, n)
184 	register struct proc *curp, *chgp;
185 	register int n;
186 {
187 	register struct pcred *pcred = curp->p_cred;
188 
189 	if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
190 	    pcred->pc_ucred->cr_uid != chgp->p_ucred->cr_uid &&
191 	    pcred->p_ruid != chgp->p_ucred->cr_uid)
192 		return (EPERM);
193 	if (n > PRIO_MAX)
194 		n = PRIO_MAX;
195 	if (n < PRIO_MIN)
196 		n = PRIO_MIN;
197 	if (n < chgp->p_nice && suser(pcred->pc_ucred, &curp->p_acflag))
198 		return (EACCES);
199 	chgp->p_nice = n;
200 	(void)resetpriority(chgp);
201 	return (0);
202 }
203 
204 /* rtprio system call */
205 #ifndef _SYS_SYSPROTO_H_
206 struct rtprio_args {
207 	int		function;
208 	pid_t		pid;
209 	struct rtprio	*rtp;
210 };
211 #endif
212 
213 /*
214  * Set realtime priority
215  */
216 
217 /* ARGSUSED */
218 int
219 rtprio(curp, uap, retval)
220 	struct proc *curp;
221 	register struct rtprio_args *uap;
222 	int *retval;
223 {
224 	register struct proc *p;
225 	register struct pcred *pcred = curp->p_cred;
226 	struct rtprio rtp;
227 	int error;
228 
229 	error = copyin(uap->rtp, &rtp, sizeof(struct rtprio));
230 	if (error)
231 		return (error);
232 
233 	if (uap->pid == 0)
234 		p = curp;
235 	else
236 		p = pfind(uap->pid);
237 
238 	if (p == 0)
239 		return (ESRCH);
240 
241 	switch (uap->function) {
242 	case RTP_LOOKUP:
243 		return (copyout(&p->p_rtprio, uap->rtp, sizeof(struct rtprio)));
244 	case RTP_SET:
245 		if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
246 		    pcred->pc_ucred->cr_uid != p->p_ucred->cr_uid &&
247 		    pcred->p_ruid != p->p_ucred->cr_uid)
248 		        return (EPERM);
249 		/* disallow setting rtprio in most cases if not superuser */
250 		if (suser(pcred->pc_ucred, &curp->p_acflag)) {
251 			/* can't set someone else's */
252 			if (uap->pid)
253 				return (EPERM);
254 			/* can't set realtime priority */
255 			if (rtp.type == RTP_PRIO_REALTIME)
256 				return (EPERM);
257 		}
258 		switch (rtp.type) {
259 		case RTP_PRIO_REALTIME:
260 		case RTP_PRIO_NORMAL:
261 		case RTP_PRIO_IDLE:
262 			if (rtp.prio > RTP_PRIO_MAX)
263 				return (EINVAL);
264 			p->p_rtprio = rtp;
265 			return (0);
266 		default:
267 			return (EINVAL);
268 		}
269 
270 	default:
271 		return (EINVAL);
272 	}
273 }
274 
275 #if defined(COMPAT_43) || defined(COMPAT_SUNOS)
276 #ifndef _SYS_SYSPROTO_H_
277 struct osetrlimit_args {
278 	u_int	which;
279 	struct	orlimit *rlp;
280 };
281 #endif
282 /* ARGSUSED */
283 int
284 osetrlimit(p, uap, retval)
285 	struct proc *p;
286 	register struct osetrlimit_args *uap;
287 	int *retval;
288 {
289 	struct orlimit olim;
290 	struct rlimit lim;
291 	int error;
292 
293 	if ((error =
294 	    copyin((caddr_t)uap->rlp, (caddr_t)&olim, sizeof(struct orlimit))))
295 		return (error);
296 	lim.rlim_cur = olim.rlim_cur;
297 	lim.rlim_max = olim.rlim_max;
298 	return (dosetrlimit(p, uap->which, &lim));
299 }
300 
301 #ifndef _SYS_SYSPROTO_H_
302 struct ogetrlimit_args {
303 	u_int	which;
304 	struct	orlimit *rlp;
305 };
306 #endif
307 /* ARGSUSED */
308 int
309 ogetrlimit(p, uap, retval)
310 	struct proc *p;
311 	register struct ogetrlimit_args *uap;
312 	int *retval;
313 {
314 	struct orlimit olim;
315 
316 	if (uap->which >= RLIM_NLIMITS)
317 		return (EINVAL);
318 	olim.rlim_cur = p->p_rlimit[uap->which].rlim_cur;
319 	if (olim.rlim_cur == -1)
320 		olim.rlim_cur = 0x7fffffff;
321 	olim.rlim_max = p->p_rlimit[uap->which].rlim_max;
322 	if (olim.rlim_max == -1)
323 		olim.rlim_max = 0x7fffffff;
324 	return (copyout((caddr_t)&olim, (caddr_t)uap->rlp, sizeof(olim)));
325 }
326 #endif /* COMPAT_43 || COMPAT_SUNOS */
327 
328 #ifndef _SYS_SYSPROTO_H_
329 struct __setrlimit_args {
330 	u_int	which;
331 	struct	rlimit *rlp;
332 };
333 #endif
334 /* ARGSUSED */
335 int
336 setrlimit(p, uap, retval)
337 	struct proc *p;
338 	register struct __setrlimit_args *uap;
339 	int *retval;
340 {
341 	struct rlimit alim;
342 	int error;
343 
344 	if ((error =
345 	    copyin((caddr_t)uap->rlp, (caddr_t)&alim, sizeof (struct rlimit))))
346 		return (error);
347 	return (dosetrlimit(p, uap->which, &alim));
348 }
349 
350 int
351 dosetrlimit(p, which, limp)
352 	struct proc *p;
353 	u_int which;
354 	struct rlimit *limp;
355 {
356 	register struct rlimit *alimp;
357 	int error;
358 
359 	if (which >= RLIM_NLIMITS)
360 		return (EINVAL);
361 	alimp = &p->p_rlimit[which];
362 
363 	/*
364 	 * Preserve historical bugs by treating negative limits as unsigned.
365 	 */
366 	if (limp->rlim_cur < 0)
367 		limp->rlim_cur = RLIM_INFINITY;
368 	if (limp->rlim_max < 0)
369 		limp->rlim_max = RLIM_INFINITY;
370 
371 	if (limp->rlim_cur > alimp->rlim_max ||
372 	    limp->rlim_max > alimp->rlim_max)
373 		if ((error = suser(p->p_ucred, &p->p_acflag)))
374 			return (error);
375 	if (limp->rlim_cur > limp->rlim_max)
376 		limp->rlim_cur = limp->rlim_max;
377 	if (p->p_limit->p_refcnt > 1 &&
378 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
379 		p->p_limit->p_refcnt--;
380 		p->p_limit = limcopy(p->p_limit);
381 		alimp = &p->p_rlimit[which];
382 	}
383 
384 	switch (which) {
385 
386 	case RLIMIT_DATA:
387 		if (limp->rlim_cur > MAXDSIZ)
388 			limp->rlim_cur = MAXDSIZ;
389 		if (limp->rlim_max > MAXDSIZ)
390 			limp->rlim_max = MAXDSIZ;
391 		break;
392 
393 	case RLIMIT_STACK:
394 		if (limp->rlim_cur > MAXSSIZ)
395 			limp->rlim_cur = MAXSSIZ;
396 		if (limp->rlim_max > MAXSSIZ)
397 			limp->rlim_max = MAXSSIZ;
398 		/*
399 		 * Stack is allocated to the max at exec time with only
400 		 * "rlim_cur" bytes accessible.  If stack limit is going
401 		 * up make more accessible, if going down make inaccessible.
402 		 */
403 		if (limp->rlim_cur != alimp->rlim_cur) {
404 			vm_offset_t addr;
405 			vm_size_t size;
406 			vm_prot_t prot;
407 
408 			if (limp->rlim_cur > alimp->rlim_cur) {
409 				prot = VM_PROT_ALL;
410 				size = limp->rlim_cur - alimp->rlim_cur;
411 				addr = USRSTACK - limp->rlim_cur;
412 			} else {
413 				prot = VM_PROT_NONE;
414 				size = alimp->rlim_cur - limp->rlim_cur;
415 				addr = USRSTACK - alimp->rlim_cur;
416 			}
417 			addr = trunc_page(addr);
418 			size = round_page(size);
419 			(void) vm_map_protect(&p->p_vmspace->vm_map,
420 					      addr, addr+size, prot, FALSE);
421 		}
422 		break;
423 
424 	case RLIMIT_NOFILE:
425 		if (limp->rlim_cur > maxfilesperproc)
426 			limp->rlim_cur = maxfilesperproc;
427 		if (limp->rlim_max > maxfilesperproc)
428 			limp->rlim_max = maxfilesperproc;
429 		break;
430 
431 	case RLIMIT_NPROC:
432 		if (limp->rlim_cur > maxprocperuid)
433 			limp->rlim_cur = maxprocperuid;
434 		if (limp->rlim_max > maxprocperuid)
435 			limp->rlim_max = maxprocperuid;
436 		break;
437 	}
438 	*alimp = *limp;
439 	return (0);
440 }
441 
442 #ifndef _SYS_SYSPROTO_H_
443 struct __getrlimit_args {
444 	u_int	which;
445 	struct	rlimit *rlp;
446 };
447 #endif
448 /* ARGSUSED */
449 int
450 getrlimit(p, uap, retval)
451 	struct proc *p;
452 	register struct __getrlimit_args *uap;
453 	int *retval;
454 {
455 
456 	if (uap->which >= RLIM_NLIMITS)
457 		return (EINVAL);
458 	return (copyout((caddr_t)&p->p_rlimit[uap->which], (caddr_t)uap->rlp,
459 	    sizeof (struct rlimit)));
460 }
461 
462 /*
463  * Transform the running time and tick information in proc p into user,
464  * system, and interrupt time usage.
465  */
466 void
467 calcru(p, up, sp, ip)
468 	register struct proc *p;
469 	register struct timeval *up;
470 	register struct timeval *sp;
471 	register struct timeval *ip;
472 {
473 	register quad_t totusec;
474 	register u_quad_t u, st, ut, it, tot;
475 	register long sec, usec;
476 	register int s;
477 	struct timeval tv;
478 
479 	s = splstatclock();
480 	st = p->p_sticks;
481 	ut = p->p_uticks;
482 	it = p->p_iticks;
483 	splx(s);
484 
485 	tot = st + ut + it;
486 	if (tot == 0) {
487 		up->tv_sec = up->tv_usec = 0;
488 		sp->tv_sec = sp->tv_usec = 0;
489 		if (ip != NULL)
490 			ip->tv_sec = ip->tv_usec = 0;
491 		return;
492 	}
493 
494 	sec = p->p_rtime.tv_sec;
495 	usec = p->p_rtime.tv_usec;
496 	if (p == curproc) {
497 		/*
498 		 * Adjust for the current time slice.  This is actually fairly
499 		 * important since the error here is on the order of a time
500 		 * quantum, which is much greater than the sampling error.
501 		 */
502 		microtime(&tv);
503 		sec += tv.tv_sec - runtime.tv_sec;
504 		usec += tv.tv_usec - runtime.tv_usec;
505 	}
506 	totusec = (quad_t)sec * 1000000 + usec;
507 	if (totusec < 0) {
508 		printf("calcru: negative time: %qd usec\n", totusec);
509 		totusec = 0;
510 	}
511 	u = totusec;
512 	st = (u * st) / tot;
513 	sp->tv_sec = st / 1000000;
514 	sp->tv_usec = st % 1000000;
515 	ut = (u * ut) / tot;
516 	up->tv_sec = ut / 1000000;
517 	up->tv_usec = ut % 1000000;
518 	if (ip != NULL) {
519 		it = (u * it) / tot;
520 		ip->tv_sec = it / 1000000;
521 		ip->tv_usec = it % 1000000;
522 	}
523 }
524 
525 #ifndef _SYS_SYSPROTO_H_
526 struct getrusage_args {
527 	int	who;
528 	struct	rusage *rusage;
529 };
530 #endif
531 /* ARGSUSED */
532 int
533 getrusage(p, uap, retval)
534 	register struct proc *p;
535 	register struct getrusage_args *uap;
536 	int *retval;
537 {
538 	register struct rusage *rup;
539 
540 	switch (uap->who) {
541 
542 	case RUSAGE_SELF:
543 		rup = &p->p_stats->p_ru;
544 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
545 		break;
546 
547 	case RUSAGE_CHILDREN:
548 		rup = &p->p_stats->p_cru;
549 		break;
550 
551 	default:
552 		return (EINVAL);
553 	}
554 	return (copyout((caddr_t)rup, (caddr_t)uap->rusage,
555 	    sizeof (struct rusage)));
556 }
557 
558 void
559 ruadd(ru, ru2)
560 	register struct rusage *ru, *ru2;
561 {
562 	register long *ip, *ip2;
563 	register int i;
564 
565 	timevaladd(&ru->ru_utime, &ru2->ru_utime);
566 	timevaladd(&ru->ru_stime, &ru2->ru_stime);
567 	if (ru->ru_maxrss < ru2->ru_maxrss)
568 		ru->ru_maxrss = ru2->ru_maxrss;
569 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
570 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
571 		*ip++ += *ip2++;
572 }
573 
574 /*
575  * Make a copy of the plimit structure.
576  * We share these structures copy-on-write after fork,
577  * and copy when a limit is changed.
578  */
579 struct plimit *
580 limcopy(lim)
581 	struct plimit *lim;
582 {
583 	register struct plimit *copy;
584 
585 	MALLOC(copy, struct plimit *, sizeof(struct plimit),
586 	    M_SUBPROC, M_WAITOK);
587 	bcopy(lim->pl_rlimit, copy->pl_rlimit,
588 	    sizeof(struct rlimit) * RLIM_NLIMITS);
589 	copy->p_lflags = 0;
590 	copy->p_refcnt = 1;
591 	return (copy);
592 }
593