xref: /netbsd-src/sys/kern/kern_resource.c (revision 06be8101a16cc95f40783b3cb7afd12112103a9a)
1 /*	$NetBSD: kern_resource.c,v 1.61 2001/11/12 15:25:14 lukem 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 1.61 2001/11/12 15:25:14 lukem 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/syscallargs.h>
57 
58 #include <uvm/uvm_extern.h>
59 
60 /*
61  * Maximum process data and stack limits.
62  * They are variables so they are patchable.
63  *
64  * XXXX Do we really need them to be patchable?
65  */
66 rlim_t maxdmap = MAXDSIZ;
67 rlim_t maxsmap = MAXSSIZ;
68 
69 /*
70  * Resource controls and accounting.
71  */
72 
73 int
74 sys_getpriority(curp, v, retval)
75 	struct proc *curp;
76 	void *v;
77 	register_t *retval;
78 {
79 	struct sys_getpriority_args /* {
80 		syscallarg(int) which;
81 		syscallarg(int) who;
82 	} */ *uap = v;
83 	struct proc *p;
84 	int low = NZERO + PRIO_MAX + 1;
85 
86 	switch (SCARG(uap, which)) {
87 
88 	case PRIO_PROCESS:
89 		if (SCARG(uap, who) == 0)
90 			p = curp;
91 		else
92 			p = pfind(SCARG(uap, who));
93 		if (p == 0)
94 			break;
95 		low = p->p_nice;
96 		break;
97 
98 	case PRIO_PGRP: {
99 		struct pgrp *pg;
100 
101 		if (SCARG(uap, who) == 0)
102 			pg = curp->p_pgrp;
103 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
104 			break;
105 		for (p = pg->pg_members.lh_first; p != 0;
106 		     p = p->p_pglist.le_next) {
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 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
118 			if (p->p_ucred->cr_uid == SCARG(uap, who) &&
119 			    p->p_nice < low)
120 				low = p->p_nice;
121 		proclist_unlock_read();
122 		break;
123 
124 	default:
125 		return (EINVAL);
126 	}
127 	if (low == NZERO + PRIO_MAX + 1)
128 		return (ESRCH);
129 	*retval = low - NZERO;
130 	return (0);
131 }
132 
133 /* ARGSUSED */
134 int
135 sys_setpriority(curp, v, retval)
136 	struct proc *curp;
137 	void *v;
138 	register_t *retval;
139 {
140 	struct sys_setpriority_args /* {
141 		syscallarg(int) which;
142 		syscallarg(int) who;
143 		syscallarg(int) prio;
144 	} */ *uap = v;
145 	struct proc *p;
146 	int found = 0, error = 0;
147 
148 	switch (SCARG(uap, which)) {
149 
150 	case PRIO_PROCESS:
151 		if (SCARG(uap, who) == 0)
152 			p = curp;
153 		else
154 			p = pfind(SCARG(uap, who));
155 		if (p == 0)
156 			break;
157 		error = donice(curp, p, SCARG(uap, prio));
158 		found++;
159 		break;
160 
161 	case PRIO_PGRP: {
162 		struct pgrp *pg;
163 
164 		if (SCARG(uap, who) == 0)
165 			pg = curp->p_pgrp;
166 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
167 			break;
168 		for (p = pg->pg_members.lh_first; p != 0;
169 		    p = p->p_pglist.le_next) {
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 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
181 			if (p->p_ucred->cr_uid == SCARG(uap, who)) {
182 				error = donice(curp, p, SCARG(uap, prio));
183 				found++;
184 			}
185 		proclist_unlock_read();
186 		break;
187 
188 	default:
189 		return (EINVAL);
190 	}
191 	if (found == 0)
192 		return (ESRCH);
193 	return (error);
194 }
195 
196 int
197 donice(curp, chgp, n)
198 	struct proc *curp, *chgp;
199 	int n;
200 {
201 	struct pcred *pcred = curp->p_cred;
202 	int s;
203 
204 	if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
205 	    pcred->pc_ucred->cr_uid != chgp->p_ucred->cr_uid &&
206 	    pcred->p_ruid != chgp->p_ucred->cr_uid)
207 		return (EPERM);
208 	if (n > PRIO_MAX)
209 		n = PRIO_MAX;
210 	if (n < PRIO_MIN)
211 		n = PRIO_MIN;
212 	n += NZERO;
213 	if (n < chgp->p_nice && suser(pcred->pc_ucred, &curp->p_acflag))
214 		return (EACCES);
215 	chgp->p_nice = n;
216 	SCHED_LOCK(s);
217 	(void)resetpriority(chgp);
218 	SCHED_UNLOCK(s);
219 	return (0);
220 }
221 
222 /* ARGSUSED */
223 int
224 sys_setrlimit(p, v, retval)
225 	struct proc *p;
226 	void *v;
227 	register_t *retval;
228 {
229 	struct sys_setrlimit_args /* {
230 		syscallarg(int) which;
231 		syscallarg(const struct rlimit *) rlp;
232 	} */ *uap = v;
233 	int which = SCARG(uap, which);
234 	struct rlimit alim;
235 	int error;
236 
237 	error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit));
238 	if (error)
239 		return (error);
240 	return (dosetrlimit(p, p->p_cred, which, &alim));
241 }
242 
243 int
244 dosetrlimit(p, cred, which, limp)
245 	struct proc *p;
246 	struct  pcred *cred;
247 	int which;
248 	struct rlimit *limp;
249 {
250 	struct rlimit *alimp;
251 	struct plimit *newplim;
252 	int error;
253 
254 	if ((u_int)which >= RLIM_NLIMITS)
255 		return (EINVAL);
256 
257 	if (limp->rlim_cur < 0 || limp->rlim_max < 0)
258 		return (EINVAL);
259 
260 	alimp = &p->p_rlimit[which];
261 	/* if we don't change the value, no need to limcopy() */
262 	if (limp->rlim_cur == alimp->rlim_cur &&
263 	    limp->rlim_max == alimp->rlim_max)
264 		return 0;
265 
266 	if (limp->rlim_cur > alimp->rlim_max ||
267 	    limp->rlim_max > alimp->rlim_max)
268 		if ((error = suser(cred->pc_ucred, &p->p_acflag)) != 0)
269 			return (error);
270 	if (limp->rlim_cur > limp->rlim_max)
271 		limp->rlim_cur = limp->rlim_max;
272 	if (p->p_limit->p_refcnt > 1 &&
273 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
274 		newplim = limcopy(p->p_limit);
275 		limfree(p->p_limit);
276 		p->p_limit = newplim;
277 		alimp = &p->p_rlimit[which];
278 	}
279 
280 	switch (which) {
281 
282 	case RLIMIT_DATA:
283 		if (limp->rlim_cur > maxdmap)
284 			limp->rlim_cur = maxdmap;
285 		if (limp->rlim_max > maxdmap)
286 			limp->rlim_max = maxdmap;
287 		break;
288 
289 	case RLIMIT_STACK:
290 		if (limp->rlim_cur > maxsmap)
291 			limp->rlim_cur = maxsmap;
292 		if (limp->rlim_max > maxsmap)
293 			limp->rlim_max = maxsmap;
294 
295 		/*
296 		 * Stack is allocated to the max at exec time with
297 		 * only "rlim_cur" bytes accessible (In other words,
298 		 * allocates stack dividing two contiguous regions at
299 		 * "rlim_cur" bytes boundary).
300 		 *
301 		 * Since allocation is done in terms of page, roundup
302 		 * "rlim_cur" (otherwise, contiguous regions
303 		 * overlap).  If stack limit is going up make more
304 		 * accessible, if going down make inaccessible.
305 		 */
306 		limp->rlim_cur = round_page(limp->rlim_cur);
307 		if (limp->rlim_cur != alimp->rlim_cur) {
308 			vaddr_t addr;
309 			vsize_t size;
310 			vm_prot_t prot;
311 
312 			if (limp->rlim_cur > alimp->rlim_cur) {
313 				prot = VM_PROT_ALL;
314 				size = limp->rlim_cur - alimp->rlim_cur;
315 				addr = USRSTACK - limp->rlim_cur;
316 			} else {
317 				prot = VM_PROT_NONE;
318 				size = alimp->rlim_cur - limp->rlim_cur;
319 				addr = USRSTACK - alimp->rlim_cur;
320 			}
321 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
322 					      addr, addr+size, prot, FALSE);
323 		}
324 		break;
325 
326 	case RLIMIT_NOFILE:
327 		if (limp->rlim_cur > maxfiles)
328 			limp->rlim_cur = maxfiles;
329 		if (limp->rlim_max > maxfiles)
330 			limp->rlim_max = maxfiles;
331 		break;
332 
333 	case RLIMIT_NPROC:
334 		if (limp->rlim_cur > maxproc)
335 			limp->rlim_cur = maxproc;
336 		if (limp->rlim_max > maxproc)
337 			limp->rlim_max = maxproc;
338 		break;
339 	}
340 	*alimp = *limp;
341 	return (0);
342 }
343 
344 /* ARGSUSED */
345 int
346 sys_getrlimit(p, v, retval)
347 	struct proc *p;
348 	void *v;
349 	register_t *retval;
350 {
351 	struct sys_getrlimit_args /* {
352 		syscallarg(int) which;
353 		syscallarg(struct rlimit *) rlp;
354 	} */ *uap = v;
355 	int which = SCARG(uap, which);
356 
357 	if ((u_int)which >= RLIM_NLIMITS)
358 		return (EINVAL);
359 	return (copyout(&p->p_rlimit[which], SCARG(uap, rlp),
360 	    sizeof(struct rlimit)));
361 }
362 
363 /*
364  * Transform the running time and tick information in proc p into user,
365  * system, and interrupt time usage.
366  */
367 void
368 calcru(p, up, sp, ip)
369 	struct proc *p;
370 	struct timeval *up;
371 	struct timeval *sp;
372 	struct timeval *ip;
373 {
374 	u_quad_t u, st, ut, it, tot;
375 	long sec, usec;
376 	int s;
377 	struct timeval tv;
378 
379 	s = splstatclock();
380 	st = p->p_sticks;
381 	ut = p->p_uticks;
382 	it = p->p_iticks;
383 	splx(s);
384 
385 	tot = st + ut + it;
386 	if (tot == 0) {
387 		up->tv_sec = up->tv_usec = 0;
388 		sp->tv_sec = sp->tv_usec = 0;
389 		if (ip != NULL)
390 			ip->tv_sec = ip->tv_usec = 0;
391 		return;
392 	}
393 
394 	sec = p->p_rtime.tv_sec;
395 	usec = p->p_rtime.tv_usec;
396 	if (p->p_stat == SONPROC) {
397 		struct schedstate_percpu *spc;
398 
399 		KDASSERT(p->p_cpu != NULL);
400 		spc = &p->p_cpu->ci_schedstate;
401 
402 		/*
403 		 * Adjust for the current time slice.  This is actually fairly
404 		 * important since the error here is on the order of a time
405 		 * quantum, which is much greater than the sampling error.
406 		 */
407 		microtime(&tv);
408 		sec += tv.tv_sec - spc->spc_runtime.tv_sec;
409 		usec += tv.tv_usec - spc->spc_runtime.tv_usec;
410 	}
411 	u = (u_quad_t) sec * 1000000 + usec;
412 	st = (u * st) / tot;
413 	sp->tv_sec = st / 1000000;
414 	sp->tv_usec = st % 1000000;
415 	ut = (u * ut) / tot;
416 	up->tv_sec = ut / 1000000;
417 	up->tv_usec = ut % 1000000;
418 	if (ip != NULL) {
419 		it = (u * it) / tot;
420 		ip->tv_sec = it / 1000000;
421 		ip->tv_usec = it % 1000000;
422 	}
423 }
424 
425 /* ARGSUSED */
426 int
427 sys_getrusage(p, v, retval)
428 	struct proc *p;
429 	void *v;
430 	register_t *retval;
431 {
432 	struct sys_getrusage_args /* {
433 		syscallarg(int) who;
434 		syscallarg(struct rusage *) rusage;
435 	} */ *uap = v;
436 	struct rusage *rup;
437 
438 	switch (SCARG(uap, who)) {
439 
440 	case RUSAGE_SELF:
441 		rup = &p->p_stats->p_ru;
442 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
443 		break;
444 
445 	case RUSAGE_CHILDREN:
446 		rup = &p->p_stats->p_cru;
447 		break;
448 
449 	default:
450 		return (EINVAL);
451 	}
452 	return (copyout(rup, SCARG(uap, rusage), sizeof(struct rusage)));
453 }
454 
455 void
456 ruadd(ru, ru2)
457 	struct rusage *ru, *ru2;
458 {
459 	long *ip, *ip2;
460 	int i;
461 
462 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
463 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
464 	if (ru->ru_maxrss < ru2->ru_maxrss)
465 		ru->ru_maxrss = ru2->ru_maxrss;
466 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
467 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
468 		*ip++ += *ip2++;
469 }
470 
471 /*
472  * Make a copy of the plimit structure.
473  * We share these structures copy-on-write after fork,
474  * and copy when a limit is changed.
475  */
476 struct plimit *
477 limcopy(lim)
478 	struct plimit *lim;
479 {
480 	struct plimit *newlim;
481 
482 	newlim = pool_get(&plimit_pool, PR_WAITOK);
483 	memcpy(newlim->pl_rlimit, lim->pl_rlimit,
484 	    sizeof(struct rlimit) * RLIM_NLIMITS);
485 	if (lim->pl_corename == defcorename) {
486 		newlim->pl_corename = defcorename;
487 	} else {
488 		newlim->pl_corename = malloc(strlen(lim->pl_corename)+1,
489 		    M_TEMP, M_WAITOK);
490 		strcpy(newlim->pl_corename, lim->pl_corename);
491 	}
492 	newlim->p_lflags = 0;
493 	newlim->p_refcnt = 1;
494 	return (newlim);
495 }
496 
497 void
498 limfree(lim)
499 	struct plimit *lim;
500 {
501 
502 	if (--lim->p_refcnt > 0)
503 		return;
504 #ifdef DIAGNOSTIC
505 	if (lim->p_refcnt < 0)
506 		panic("limfree");
507 #endif
508 	if (lim->pl_corename != defcorename)
509 		free(lim->pl_corename, M_TEMP);
510 	pool_put(&plimit_pool, lim);
511 }
512