xref: /netbsd-src/sys/kern/kern_resource.c (revision 2a399c6883d870daece976daec6ffa7bb7f934ce)
1 /*	$NetBSD: kern_resource.c,v 1.42 1997/10/15 17:04:02 mycroft 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	8.5 (Berkeley) 1/21/94
41  */
42 
43 #include <sys/param.h>
44 #include <sys/systm.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 <sys/mount.h>
52 #include <sys/syscallargs.h>
53 
54 #include <vm/vm.h>
55 
56 void limfree __P((struct plimit *));
57 /*
58  * Resource controls and accounting.
59  */
60 
61 int
62 sys_getpriority(curp, v, retval)
63 	struct proc *curp;
64 	void *v;
65 	register_t *retval;
66 {
67 	register struct sys_getpriority_args /* {
68 		syscallarg(int) which;
69 		syscallarg(int) who;
70 	} */ *uap = v;
71 	register struct proc *p;
72 	register int low = NZERO + PRIO_MAX + 1;
73 
74 	switch (SCARG(uap, which)) {
75 
76 	case PRIO_PROCESS:
77 		if (SCARG(uap, who) == 0)
78 			p = curp;
79 		else
80 			p = pfind(SCARG(uap, who));
81 		if (p == 0)
82 			break;
83 		low = p->p_nice;
84 		break;
85 
86 	case PRIO_PGRP: {
87 		register struct pgrp *pg;
88 
89 		if (SCARG(uap, who) == 0)
90 			pg = curp->p_pgrp;
91 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
92 			break;
93 		for (p = pg->pg_members.lh_first; p != 0; p = p->p_pglist.le_next) {
94 			if (p->p_nice < low)
95 				low = p->p_nice;
96 		}
97 		break;
98 	}
99 
100 	case PRIO_USER:
101 		if (SCARG(uap, who) == 0)
102 			SCARG(uap, who) = curp->p_ucred->cr_uid;
103 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
104 			if (p->p_ucred->cr_uid == SCARG(uap, who) &&
105 			    p->p_nice < low)
106 				low = p->p_nice;
107 		break;
108 
109 	default:
110 		return (EINVAL);
111 	}
112 	if (low == NZERO + PRIO_MAX + 1)
113 		return (ESRCH);
114 	*retval = low - NZERO;
115 	return (0);
116 }
117 
118 /* ARGSUSED */
119 int
120 sys_setpriority(curp, v, retval)
121 	struct proc *curp;
122 	void *v;
123 	register_t *retval;
124 {
125 	register struct sys_setpriority_args /* {
126 		syscallarg(int) which;
127 		syscallarg(int) who;
128 		syscallarg(int) prio;
129 	} */ *uap = v;
130 	register struct proc *p;
131 	int found = 0, error = 0;
132 
133 	switch (SCARG(uap, which)) {
134 
135 	case PRIO_PROCESS:
136 		if (SCARG(uap, who) == 0)
137 			p = curp;
138 		else
139 			p = pfind(SCARG(uap, who));
140 		if (p == 0)
141 			break;
142 		error = donice(curp, p, SCARG(uap, prio));
143 		found++;
144 		break;
145 
146 	case PRIO_PGRP: {
147 		register struct pgrp *pg;
148 
149 		if (SCARG(uap, who) == 0)
150 			pg = curp->p_pgrp;
151 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
152 			break;
153 		for (p = pg->pg_members.lh_first; p != 0;
154 		    p = p->p_pglist.le_next) {
155 			error = donice(curp, p, SCARG(uap, prio));
156 			found++;
157 		}
158 		break;
159 	}
160 
161 	case PRIO_USER:
162 		if (SCARG(uap, who) == 0)
163 			SCARG(uap, who) = curp->p_ucred->cr_uid;
164 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
165 			if (p->p_ucred->cr_uid == SCARG(uap, who)) {
166 				error = donice(curp, p, SCARG(uap, prio));
167 				found++;
168 			}
169 		break;
170 
171 	default:
172 		return (EINVAL);
173 	}
174 	if (found == 0)
175 		return (ESRCH);
176 	return (error);
177 }
178 
179 int
180 donice(curp, chgp, n)
181 	register struct proc *curp, *chgp;
182 	register int n;
183 {
184 	register struct pcred *pcred = curp->p_cred;
185 
186 	if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
187 	    pcred->pc_ucred->cr_uid != chgp->p_ucred->cr_uid &&
188 	    pcred->p_ruid != chgp->p_ucred->cr_uid)
189 		return (EPERM);
190 	if (n > PRIO_MAX)
191 		n = PRIO_MAX;
192 	if (n < PRIO_MIN)
193 		n = PRIO_MIN;
194 	n += NZERO;
195 	if (n < chgp->p_nice && suser(pcred->pc_ucred, &curp->p_acflag))
196 		return (EACCES);
197 	chgp->p_nice = n;
198 	(void)resetpriority(chgp);
199 	return (0);
200 }
201 
202 /* ARGSUSED */
203 int
204 sys_setrlimit(p, v, retval)
205 	struct proc *p;
206 	void *v;
207 	register_t *retval;
208 {
209 	register struct sys_setrlimit_args /* {
210 		syscallarg(int) which;
211 		syscallarg(const struct rlimit *) rlp;
212 	} */ *uap = v;
213 	int which = SCARG(uap, which);
214 	struct rlimit alim;
215 	int error;
216 
217 	error = copyin(SCARG(uap, rlp), &alim, sizeof (struct rlimit));
218 	if (error)
219 		return (error);
220 	return (dosetrlimit(p, which, &alim));
221 }
222 
223 int
224 dosetrlimit(p, which, limp)
225 	struct proc *p;
226 	int which;
227 	struct rlimit *limp;
228 {
229 	register struct rlimit *alimp;
230 	extern unsigned maxdmap, maxsmap;
231 	int error;
232 
233 	if ((u_int)which >= RLIM_NLIMITS)
234 		return (EINVAL);
235 
236 	if (limp->rlim_cur < 0 || limp->rlim_max < 0)
237 		return (EINVAL);
238 
239 	alimp = &p->p_rlimit[which];
240 	if (limp->rlim_cur > alimp->rlim_max ||
241 	    limp->rlim_max > alimp->rlim_max)
242 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
243 			return (error);
244 	if (limp->rlim_cur > limp->rlim_max)
245 		limp->rlim_cur = limp->rlim_max;
246 	if (p->p_limit->p_refcnt > 1 &&
247 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
248 		p->p_limit->p_refcnt--;
249 		p->p_limit = limcopy(p->p_limit);
250 		alimp = &p->p_rlimit[which];
251 	}
252 
253 	switch (which) {
254 
255 	case RLIMIT_DATA:
256 		if (limp->rlim_cur > maxdmap)
257 			limp->rlim_cur = maxdmap;
258 		if (limp->rlim_max > maxdmap)
259 			limp->rlim_max = maxdmap;
260 		break;
261 
262 	case RLIMIT_STACK:
263 		if (limp->rlim_cur > maxsmap)
264 			limp->rlim_cur = maxsmap;
265 		if (limp->rlim_max > maxsmap)
266 			limp->rlim_max = maxsmap;
267 
268 		/*
269 		 * Stack is allocated to the max at exec time with
270 		 * only "rlim_cur" bytes accessible (In other words,
271 		 * allocates stack dividing two contiguous regions at
272 		 * "rlim_cur" bytes boundary).
273 		 *
274 		 * Since allocation is done in terms of page, roundup
275 		 * "rlim_cur" (otherwise, contiguous regions
276 		 * overlap).  If stack limit is going up make more
277 		 * accessible, if going down make inaccessible.
278 		 */
279 		limp->rlim_cur = round_page(limp->rlim_cur);
280 		if (limp->rlim_cur != alimp->rlim_cur) {
281 			vm_offset_t addr;
282 			vm_size_t size;
283 			vm_prot_t prot;
284 
285 			if (limp->rlim_cur > alimp->rlim_cur) {
286 				prot = VM_PROT_ALL;
287 				size = limp->rlim_cur - alimp->rlim_cur;
288 				addr = USRSTACK - limp->rlim_cur;
289 			} else {
290 				prot = VM_PROT_NONE;
291 				size = alimp->rlim_cur - limp->rlim_cur;
292 				addr = USRSTACK - alimp->rlim_cur;
293 			}
294 			(void)vm_map_protect(&p->p_vmspace->vm_map,
295 			    addr, addr+size, prot, FALSE);
296 		}
297 		break;
298 
299 	case RLIMIT_NOFILE:
300 		if (limp->rlim_cur > maxfiles)
301 			limp->rlim_cur = maxfiles;
302 		if (limp->rlim_max > maxfiles)
303 			limp->rlim_max = maxfiles;
304 		break;
305 
306 	case RLIMIT_NPROC:
307 		if (limp->rlim_cur > maxproc)
308 			limp->rlim_cur = maxproc;
309 		if (limp->rlim_max > maxproc)
310 			limp->rlim_max = maxproc;
311 		break;
312 	}
313 	*alimp = *limp;
314 	return (0);
315 }
316 
317 /* ARGSUSED */
318 int
319 sys_getrlimit(p, v, retval)
320 	struct proc *p;
321 	void *v;
322 	register_t *retval;
323 {
324 	register struct sys_getrlimit_args /* {
325 		syscallarg(int) which;
326 		syscallarg(struct rlimit *) rlp;
327 	} */ *uap = v;
328 	int which = SCARG(uap, which);
329 
330 	if ((u_int)which >= RLIM_NLIMITS)
331 		return (EINVAL);
332 	return (copyout(&p->p_rlimit[which], SCARG(uap, rlp),
333 	    sizeof (struct rlimit)));
334 }
335 
336 /*
337  * Transform the running time and tick information in proc p into user,
338  * system, and interrupt time usage.
339  */
340 void
341 calcru(p, up, sp, ip)
342 	register struct proc *p;
343 	register struct timeval *up;
344 	register struct timeval *sp;
345 	register struct timeval *ip;
346 {
347 	register u_quad_t u, st, ut, it, tot;
348 	register long sec, usec;
349 	register int s;
350 	struct timeval tv;
351 
352 	s = splstatclock();
353 	st = p->p_sticks;
354 	ut = p->p_uticks;
355 	it = p->p_iticks;
356 	splx(s);
357 
358 	tot = st + ut + it;
359 	if (tot == 0) {
360 		up->tv_sec = up->tv_usec = 0;
361 		sp->tv_sec = sp->tv_usec = 0;
362 		if (ip != NULL)
363 			ip->tv_sec = ip->tv_usec = 0;
364 		return;
365 	}
366 
367 	sec = p->p_rtime.tv_sec;
368 	usec = p->p_rtime.tv_usec;
369 	if (p == curproc) {
370 		/*
371 		 * Adjust for the current time slice.  This is actually fairly
372 		 * important since the error here is on the order of a time
373 		 * quantum, which is much greater than the sampling error.
374 		 */
375 		microtime(&tv);
376 		sec += tv.tv_sec - runtime.tv_sec;
377 		usec += tv.tv_usec - runtime.tv_usec;
378 	}
379 	u = (u_quad_t) sec * 1000000 + usec;
380 	st = (u * st) / tot;
381 	sp->tv_sec = st / 1000000;
382 	sp->tv_usec = st % 1000000;
383 	ut = (u * ut) / tot;
384 	up->tv_sec = ut / 1000000;
385 	up->tv_usec = ut % 1000000;
386 	if (ip != NULL) {
387 		it = (u * it) / tot;
388 		ip->tv_sec = it / 1000000;
389 		ip->tv_usec = it % 1000000;
390 	}
391 }
392 
393 /* ARGSUSED */
394 int
395 sys_getrusage(p, v, retval)
396 	register struct proc *p;
397 	void *v;
398 	register_t *retval;
399 {
400 	register struct sys_getrusage_args /* {
401 		syscallarg(int) who;
402 		syscallarg(struct rusage *) rusage;
403 	} */ *uap = v;
404 	register struct rusage *rup;
405 
406 	switch (SCARG(uap, who)) {
407 
408 	case RUSAGE_SELF:
409 		rup = &p->p_stats->p_ru;
410 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
411 		break;
412 
413 	case RUSAGE_CHILDREN:
414 		rup = &p->p_stats->p_cru;
415 		break;
416 
417 	default:
418 		return (EINVAL);
419 	}
420 	return (copyout(rup, SCARG(uap, rusage), sizeof (struct rusage)));
421 }
422 
423 void
424 ruadd(ru, ru2)
425 	register struct rusage *ru, *ru2;
426 {
427 	register long *ip, *ip2;
428 	register int i;
429 
430 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
431 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
432 	if (ru->ru_maxrss < ru2->ru_maxrss)
433 		ru->ru_maxrss = ru2->ru_maxrss;
434 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
435 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
436 		*ip++ += *ip2++;
437 }
438 
439 /*
440  * Make a copy of the plimit structure.
441  * We share these structures copy-on-write after fork,
442  * and copy when a limit is changed.
443  */
444 struct plimit *
445 limcopy(lim)
446 	struct plimit *lim;
447 {
448 	register struct plimit *newlim;
449 
450 	MALLOC(newlim, struct plimit *, sizeof(struct plimit),
451 	    M_SUBPROC, M_WAITOK);
452 	bcopy(lim->pl_rlimit, newlim->pl_rlimit,
453 	    sizeof(struct rlimit) * RLIM_NLIMITS);
454 	newlim->p_lflags = 0;
455 	newlim->p_refcnt = 1;
456 	return (newlim);
457 }
458 
459 void
460 limfree(lim)
461 	struct plimit *lim;
462 {
463 
464 	if (--lim->p_refcnt > 0)
465 		return;
466 	FREE(lim, M_SUBPROC);
467 }
468