xref: /netbsd-src/sys/kern/kern_resource.c (revision 001c68bd94f75ce9270b69227c4199fbf34ee396)
1 /*	$NetBSD: kern_resource.c,v 1.71 2003/05/16 14:25:03 itojun 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. All advertising materials mentioning features or use of this software
21  *    must display the following acknowledgement:
22  *	This product includes software developed by the University of
23  *	California, Berkeley and its contributors.
24  * 4. Neither the name of the University nor the names of its contributors
25  *    may be used to endorse or promote products derived from this software
26  *    without specific prior written permission.
27  *
28  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
29  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
30  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
31  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
32  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
33  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
34  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
35  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
36  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
37  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38  * SUCH DAMAGE.
39  *
40  *	@(#)kern_resource.c	8.8 (Berkeley) 2/14/95
41  */
42 
43 #include <sys/cdefs.h>
44 __KERNEL_RCSID(0, "$NetBSD: kern_resource.c,v 1.71 2003/05/16 14:25:03 itojun Exp $");
45 
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/kernel.h>
49 #include <sys/file.h>
50 #include <sys/resourcevar.h>
51 #include <sys/malloc.h>
52 #include <sys/pool.h>
53 #include <sys/proc.h>
54 
55 #include <sys/mount.h>
56 #include <sys/sa.h>
57 #include <sys/syscallargs.h>
58 
59 #include <uvm/uvm_extern.h>
60 
61 /*
62  * Maximum process data and stack limits.
63  * They are variables so they are patchable.
64  *
65  * XXXX Do we really need them to be patchable?
66  */
67 rlim_t maxdmap = MAXDSIZ;
68 rlim_t maxsmap = MAXSSIZ;
69 
70 /*
71  * Resource controls and accounting.
72  */
73 
74 int
75 sys_getpriority(l, v, retval)
76 	struct lwp *l;
77 	void *v;
78 	register_t *retval;
79 {
80 	struct sys_getpriority_args /* {
81 		syscallarg(int) which;
82 		syscallarg(int) who;
83 	} */ *uap = v;
84 	struct proc *curp = l->l_proc, *p;
85 	int low = NZERO + PRIO_MAX + 1;
86 
87 	switch (SCARG(uap, which)) {
88 
89 	case PRIO_PROCESS:
90 		if (SCARG(uap, who) == 0)
91 			p = curp;
92 		else
93 			p = pfind(SCARG(uap, who));
94 		if (p == 0)
95 			break;
96 		low = p->p_nice;
97 		break;
98 
99 	case PRIO_PGRP: {
100 		struct pgrp *pg;
101 
102 		if (SCARG(uap, who) == 0)
103 			pg = curp->p_pgrp;
104 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
105 			break;
106 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
107 			if (p->p_nice < low)
108 				low = p->p_nice;
109 		}
110 		break;
111 	}
112 
113 	case PRIO_USER:
114 		if (SCARG(uap, who) == 0)
115 			SCARG(uap, who) = curp->p_ucred->cr_uid;
116 		proclist_lock_read();
117 		LIST_FOREACH(p, &allproc, p_list) {
118 			if (p->p_ucred->cr_uid == (uid_t) SCARG(uap, who) &&
119 			    p->p_nice < low)
120 				low = p->p_nice;
121 		}
122 		proclist_unlock_read();
123 		break;
124 
125 	default:
126 		return (EINVAL);
127 	}
128 	if (low == NZERO + PRIO_MAX + 1)
129 		return (ESRCH);
130 	*retval = low - NZERO;
131 	return (0);
132 }
133 
134 /* ARGSUSED */
135 int
136 sys_setpriority(l, v, retval)
137 	struct lwp *l;
138 	void *v;
139 	register_t *retval;
140 {
141 	struct sys_setpriority_args /* {
142 		syscallarg(int) which;
143 		syscallarg(int) who;
144 		syscallarg(int) prio;
145 	} */ *uap = v;
146 	struct proc *curp = l->l_proc, *p;
147 	int found = 0, error = 0;
148 
149 	switch (SCARG(uap, which)) {
150 
151 	case PRIO_PROCESS:
152 		if (SCARG(uap, who) == 0)
153 			p = curp;
154 		else
155 			p = pfind(SCARG(uap, who));
156 		if (p == 0)
157 			break;
158 		error = donice(curp, p, SCARG(uap, prio));
159 		found++;
160 		break;
161 
162 	case PRIO_PGRP: {
163 		struct pgrp *pg;
164 
165 		if (SCARG(uap, who) == 0)
166 			pg = curp->p_pgrp;
167 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
168 			break;
169 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
170 			error = donice(curp, p, SCARG(uap, prio));
171 			found++;
172 		}
173 		break;
174 	}
175 
176 	case PRIO_USER:
177 		if (SCARG(uap, who) == 0)
178 			SCARG(uap, who) = curp->p_ucred->cr_uid;
179 		proclist_lock_read();
180 		LIST_FOREACH(p, &allproc, p_list) {
181 			if (p->p_ucred->cr_uid == (uid_t) SCARG(uap, who)) {
182 				error = donice(curp, p, SCARG(uap, prio));
183 				found++;
184 			}
185 		}
186 		proclist_unlock_read();
187 		break;
188 
189 	default:
190 		return (EINVAL);
191 	}
192 	if (found == 0)
193 		return (ESRCH);
194 	return (error);
195 }
196 
197 int
198 donice(curp, chgp, n)
199 	struct proc *curp, *chgp;
200 	int n;
201 {
202 	struct pcred *pcred = curp->p_cred;
203 	int s;
204 
205 	if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
206 	    pcred->pc_ucred->cr_uid != chgp->p_ucred->cr_uid &&
207 	    pcred->p_ruid != chgp->p_ucred->cr_uid)
208 		return (EPERM);
209 	if (n > PRIO_MAX)
210 		n = PRIO_MAX;
211 	if (n < PRIO_MIN)
212 		n = PRIO_MIN;
213 	n += NZERO;
214 	if (n < chgp->p_nice && suser(pcred->pc_ucred, &curp->p_acflag))
215 		return (EACCES);
216 	chgp->p_nice = n;
217 	SCHED_LOCK(s);
218 	(void)resetprocpriority(chgp);
219 	SCHED_UNLOCK(s);
220 	return (0);
221 }
222 
223 /* ARGSUSED */
224 int
225 sys_setrlimit(l, v, retval)
226 	struct lwp *l;
227 	void *v;
228 	register_t *retval;
229 {
230 	struct sys_setrlimit_args /* {
231 		syscallarg(int) which;
232 		syscallarg(const struct rlimit *) rlp;
233 	} */ *uap = v;
234 	struct proc *p = l->l_proc;
235 	int which = SCARG(uap, which);
236 	struct rlimit alim;
237 	int error;
238 
239 	error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit));
240 	if (error)
241 		return (error);
242 	return (dosetrlimit(p, p->p_cred, which, &alim));
243 }
244 
245 int
246 dosetrlimit(p, cred, which, limp)
247 	struct proc *p;
248 	struct  pcred *cred;
249 	int which;
250 	struct rlimit *limp;
251 {
252 	struct rlimit *alimp;
253 	struct plimit *newplim;
254 	int error;
255 
256 	if ((u_int)which >= RLIM_NLIMITS)
257 		return (EINVAL);
258 
259 	if (limp->rlim_cur < 0 || limp->rlim_max < 0)
260 		return (EINVAL);
261 
262 	alimp = &p->p_rlimit[which];
263 	/* if we don't change the value, no need to limcopy() */
264 	if (limp->rlim_cur == alimp->rlim_cur &&
265 	    limp->rlim_max == alimp->rlim_max)
266 		return 0;
267 
268 	if (limp->rlim_cur > limp->rlim_max) {
269 		/*
270 		 * This is programming error. According to SUSv2, we should
271 		 * return error in this case.
272 		 */
273 		return (EINVAL);
274 	}
275 	if (limp->rlim_max > alimp->rlim_max
276 	    && (error = suser(cred->pc_ucred, &p->p_acflag)) != 0)
277 			return (error);
278 
279 	if (p->p_limit->p_refcnt > 1 &&
280 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
281 		newplim = limcopy(p->p_limit);
282 		limfree(p->p_limit);
283 		p->p_limit = newplim;
284 		alimp = &p->p_rlimit[which];
285 	}
286 
287 	switch (which) {
288 
289 	case RLIMIT_DATA:
290 		if (limp->rlim_cur > maxdmap)
291 			limp->rlim_cur = maxdmap;
292 		if (limp->rlim_max > maxdmap)
293 			limp->rlim_max = maxdmap;
294 		break;
295 
296 	case RLIMIT_STACK:
297 		if (limp->rlim_cur > maxsmap)
298 			limp->rlim_cur = maxsmap;
299 		if (limp->rlim_max > maxsmap)
300 			limp->rlim_max = maxsmap;
301 
302 		/*
303 		 * Return EINVAL if the new stack size limit is lower than
304 		 * current usage. Otherwise, the process would get SIGSEGV the
305 		 * moment it would try to access anything on it's current stack.
306 		 * This conforms to SUSv2.
307 		 */
308 		if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE
309 		    || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE)
310 			return (EINVAL);
311 
312 		/*
313 		 * Stack is allocated to the max at exec time with
314 		 * only "rlim_cur" bytes accessible (In other words,
315 		 * allocates stack dividing two contiguous regions at
316 		 * "rlim_cur" bytes boundary).
317 		 *
318 		 * Since allocation is done in terms of page, roundup
319 		 * "rlim_cur" (otherwise, contiguous regions
320 		 * overlap).  If stack limit is going up make more
321 		 * accessible, if going down make inaccessible.
322 		 */
323 		limp->rlim_cur = round_page(limp->rlim_cur);
324 		if (limp->rlim_cur != alimp->rlim_cur) {
325 			vaddr_t addr;
326 			vsize_t size;
327 			vm_prot_t prot;
328 
329 			if (limp->rlim_cur > alimp->rlim_cur) {
330 				prot = VM_PROT_ALL;
331 				size = limp->rlim_cur - alimp->rlim_cur;
332 				addr = USRSTACK - limp->rlim_cur;
333 			} else {
334 				prot = VM_PROT_NONE;
335 				size = alimp->rlim_cur - limp->rlim_cur;
336 				addr = USRSTACK - alimp->rlim_cur;
337 			}
338 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
339 					      addr, addr+size, prot, FALSE);
340 		}
341 		break;
342 
343 	case RLIMIT_NOFILE:
344 		if (limp->rlim_cur > maxfiles)
345 			limp->rlim_cur = maxfiles;
346 		if (limp->rlim_max > maxfiles)
347 			limp->rlim_max = maxfiles;
348 		break;
349 
350 	case RLIMIT_NPROC:
351 		if (limp->rlim_cur > maxproc)
352 			limp->rlim_cur = maxproc;
353 		if (limp->rlim_max > maxproc)
354 			limp->rlim_max = maxproc;
355 		break;
356 	}
357 	*alimp = *limp;
358 	return (0);
359 }
360 
361 /* ARGSUSED */
362 int
363 sys_getrlimit(l, v, retval)
364 	struct lwp *l;
365 	void *v;
366 	register_t *retval;
367 {
368 	struct sys_getrlimit_args /* {
369 		syscallarg(int) which;
370 		syscallarg(struct rlimit *) rlp;
371 	} */ *uap = v;
372 	struct proc *p = l->l_proc;
373 	int which = SCARG(uap, which);
374 
375 	if ((u_int)which >= RLIM_NLIMITS)
376 		return (EINVAL);
377 	return (copyout(&p->p_rlimit[which], SCARG(uap, rlp),
378 	    sizeof(struct rlimit)));
379 }
380 
381 /*
382  * Transform the running time and tick information in proc p into user,
383  * system, and interrupt time usage.
384  */
385 void
386 calcru(p, up, sp, ip)
387 	struct proc *p;
388 	struct timeval *up;
389 	struct timeval *sp;
390 	struct timeval *ip;
391 {
392 	u_quad_t u, st, ut, it, tot;
393 	unsigned long sec;
394 	long usec;
395 	int s;
396 	struct timeval tv;
397 	struct lwp *l;
398 
399 	s = splstatclock();
400 	st = p->p_sticks;
401 	ut = p->p_uticks;
402 	it = p->p_iticks;
403 	splx(s);
404 
405 	sec = p->p_rtime.tv_sec;
406 	usec = p->p_rtime.tv_usec;
407 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
408 		if (l->l_stat == LSONPROC) {
409 			struct schedstate_percpu *spc;
410 
411 			KDASSERT(l->l_cpu != NULL);
412 			spc = &l->l_cpu->ci_schedstate;
413 
414 			/*
415 			 * Adjust for the current time slice.  This is
416 			 * actually fairly important since the error
417 			 * here is on the order of a time quantum,
418 			 * which is much greater than the sampling
419 			 * error.
420 			 */
421 			microtime(&tv);
422 			sec += tv.tv_sec - spc->spc_runtime.tv_sec;
423 			usec += tv.tv_usec - spc->spc_runtime.tv_usec;
424 		}
425 	}
426 
427 	tot = st + ut + it;
428 	u = sec * 1000000ull + usec;
429 
430 	if (tot == 0) {
431 		/* No ticks, so can't use to share time out, split 50-50 */
432 		st = ut = u / 2;
433 	} else {
434 		st = (u * st) / tot;
435 		ut = (u * ut) / tot;
436 	}
437 	sp->tv_sec = st / 1000000;
438 	sp->tv_usec = st % 1000000;
439 	up->tv_sec = ut / 1000000;
440 	up->tv_usec = ut % 1000000;
441 	if (ip != NULL) {
442 		if (it != 0)
443 			it = (u * it) / tot;
444 		ip->tv_sec = it / 1000000;
445 		ip->tv_usec = it % 1000000;
446 	}
447 }
448 
449 /* ARGSUSED */
450 int
451 sys_getrusage(l, v, retval)
452 	struct lwp *l;
453 	void *v;
454 	register_t *retval;
455 {
456 	struct sys_getrusage_args /* {
457 		syscallarg(int) who;
458 		syscallarg(struct rusage *) rusage;
459 	} */ *uap = v;
460 	struct rusage *rup;
461 	struct proc *p = l->l_proc;
462 
463 	switch (SCARG(uap, who)) {
464 
465 	case RUSAGE_SELF:
466 		rup = &p->p_stats->p_ru;
467 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
468 		break;
469 
470 	case RUSAGE_CHILDREN:
471 		rup = &p->p_stats->p_cru;
472 		break;
473 
474 	default:
475 		return (EINVAL);
476 	}
477 	return (copyout(rup, SCARG(uap, rusage), sizeof(struct rusage)));
478 }
479 
480 void
481 ruadd(ru, ru2)
482 	struct rusage *ru, *ru2;
483 {
484 	long *ip, *ip2;
485 	int i;
486 
487 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
488 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
489 	if (ru->ru_maxrss < ru2->ru_maxrss)
490 		ru->ru_maxrss = ru2->ru_maxrss;
491 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
492 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
493 		*ip++ += *ip2++;
494 }
495 
496 /*
497  * Make a copy of the plimit structure.
498  * We share these structures copy-on-write after fork,
499  * and copy when a limit is changed.
500  */
501 struct plimit *
502 limcopy(lim)
503 	struct plimit *lim;
504 {
505 	struct plimit *newlim;
506 	size_t l;
507 
508 	newlim = pool_get(&plimit_pool, PR_WAITOK);
509 	memcpy(newlim->pl_rlimit, lim->pl_rlimit,
510 	    sizeof(struct rlimit) * RLIM_NLIMITS);
511 	if (lim->pl_corename == defcorename) {
512 		newlim->pl_corename = defcorename;
513 	} else {
514 		l = strlen(lim->pl_corename) + 1;
515 		newlim->pl_corename = malloc(l, M_TEMP, M_WAITOK);
516 		strlcpy(newlim->pl_corename, lim->pl_corename, l);
517 	}
518 	newlim->p_lflags = 0;
519 	newlim->p_refcnt = 1;
520 	return (newlim);
521 }
522 
523 void
524 limfree(lim)
525 	struct plimit *lim;
526 {
527 
528 	if (--lim->p_refcnt > 0)
529 		return;
530 #ifdef DIAGNOSTIC
531 	if (lim->p_refcnt < 0)
532 		panic("limfree");
533 #endif
534 	if (lim->pl_corename != defcorename)
535 		free(lim->pl_corename, M_TEMP);
536 	pool_put(&plimit_pool, lim);
537 }
538 
539 struct pstats *
540 pstatscopy(ps)
541 	struct pstats *ps;
542 {
543 
544 	struct pstats *newps;
545 
546 	newps = pool_get(&pstats_pool, PR_WAITOK);
547 
548 	memset(&newps->pstat_startzero, 0,
549 	(unsigned) ((caddr_t)&newps->pstat_endzero -
550 		    (caddr_t)&newps->pstat_startzero));
551 	memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy,
552 	((caddr_t)&newps->pstat_endcopy -
553 	 (caddr_t)&newps->pstat_startcopy));
554 
555 	return (newps);
556 
557 }
558 
559 void
560 pstatsfree(ps)
561 	struct pstats *ps;
562 {
563 
564 	pool_put(&pstats_pool, ps);
565 }
566