xref: /openbsd-src/sys/kern/kern_resource.c (revision 207e4b38a785bf665de26471f4fc5fde67fa99f0) 
1 /*	$OpenBSD: kern_resource.c,v 1.36 2010/07/26 01:56:27 guenther Exp $	*/
2 /*	$NetBSD: kern_resource.c,v 1.38 1996/10/23 07:19:38 matthias Exp $	*/
3 
4 /*-
5  * Copyright (c) 1982, 1986, 1991, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  * (c) UNIX System Laboratories, Inc.
8  * All or some portions of this file are derived from material licensed
9  * to the University of California by American Telephone and Telegraph
10  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
11  * the permission of UNIX System Laboratories, Inc.
12  *
13  * Redistribution and use in source and binary forms, with or without
14  * modification, are permitted provided that the following conditions
15  * are met:
16  * 1. Redistributions of source code must retain the above copyright
17  *    notice, this list of conditions and the following disclaimer.
18  * 2. Redistributions in binary form must reproduce the above copyright
19  *    notice, this list of conditions and the following disclaimer in the
20  *    documentation and/or other materials provided with the distribution.
21  * 3. Neither the name of the University nor the names of its contributors
22  *    may be used to endorse or promote products derived from this software
23  *    without specific prior written permission.
24  *
25  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35  * SUCH DAMAGE.
36  *
37  *	@(#)kern_resource.c	8.5 (Berkeley) 1/21/94
38  */
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/kernel.h>
43 #include <sys/file.h>
44 #include <sys/resourcevar.h>
45 #include <sys/pool.h>
46 #include <sys/proc.h>
47 #include <sys/sched.h>
48 
49 #include <sys/mount.h>
50 #include <sys/syscallargs.h>
51 
52 #include <uvm/uvm_extern.h>
53 
54 /*
55  * Patchable maximum data and stack limits.
56  */
57 rlim_t maxdmap = MAXDSIZ;
58 rlim_t maxsmap = MAXSSIZ;
59 
60 /*
61  * Resource controls and accounting.
62  */
63 
64 int
65 sys_getpriority(struct proc *curp, void *v, register_t *retval)
66 {
67 	struct sys_getpriority_args /* {
68 		syscallarg(int) which;
69 		syscallarg(id_t) who;
70 	} */ *uap = v;
71 	struct process *pr;
72 	struct proc *p;
73 	int low = NZERO + PRIO_MAX + 1;
74 
75 	switch (SCARG(uap, which)) {
76 
77 	case PRIO_PROCESS:
78 		if (SCARG(uap, who) == 0)
79 			pr = curp->p_p;
80 		else
81 			pr = prfind(SCARG(uap, who));
82 		if (pr == NULL)
83 			break;
84 		TAILQ_FOREACH(p, &pr->ps_threads, p_thr_link) {
85 			if (p->p_nice < low)
86 				low = p->p_nice;
87 		}
88 		break;
89 
90 	case PRIO_PGRP: {
91 		struct pgrp *pg;
92 
93 		if (SCARG(uap, who) == 0)
94 			pg = curp->p_p->ps_pgrp;
95 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
96 			break;
97 		LIST_FOREACH(pr, &pg->pg_members, ps_pglist)
98 			TAILQ_FOREACH(p, &pr->ps_threads, p_thr_link) {
99 				if (p->p_nice < low)
100 					low = p->p_nice;
101 			}
102 		break;
103 	}
104 
105 	case PRIO_USER:
106 		if (SCARG(uap, who) == 0)
107 			SCARG(uap, who) = curp->p_ucred->cr_uid;
108 		LIST_FOREACH(p, &allproc, p_list)
109 			if (p->p_ucred->cr_uid == SCARG(uap, who) &&
110 			    p->p_nice < low)
111 				low = p->p_nice;
112 		break;
113 
114 	default:
115 		return (EINVAL);
116 	}
117 	if (low == NZERO + PRIO_MAX + 1)
118 		return (ESRCH);
119 	*retval = low - NZERO;
120 	return (0);
121 }
122 
123 /* ARGSUSED */
124 int
125 sys_setpriority(struct proc *curp, void *v, register_t *retval)
126 {
127 	struct sys_setpriority_args /* {
128 		syscallarg(int) which;
129 		syscallarg(id_t) who;
130 		syscallarg(int) prio;
131 	} */ *uap = v;
132 	struct process *pr;
133 	int found = 0, error = 0;
134 
135 	switch (SCARG(uap, which)) {
136 
137 	case PRIO_PROCESS:
138 		if (SCARG(uap, who) == 0)
139 			pr = curp->p_p;
140 		else
141 			pr = prfind(SCARG(uap, who));
142 		if (pr == NULL)
143 			break;
144 		error = donice(curp, pr, SCARG(uap, prio));
145 		found++;
146 		break;
147 
148 	case PRIO_PGRP: {
149 		struct pgrp *pg;
150 
151 		if (SCARG(uap, who) == 0)
152 			pg = curp->p_p->ps_pgrp;
153 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
154 			break;
155 		LIST_FOREACH(pr, &pg->pg_members, ps_pglist) {
156 			error = donice(curp, pr, SCARG(uap, prio));
157 			found++;
158 		}
159 		break;
160 	}
161 
162 	case PRIO_USER: {
163 		struct proc *p;
164 		if (SCARG(uap, who) == 0)
165 			SCARG(uap, who) = curp->p_ucred->cr_uid;
166 		LIST_FOREACH(p, &allproc, p_list)
167 			if ((p->p_flag & P_THREAD) == 0 &&
168 			    p->p_ucred->cr_uid == SCARG(uap, who)) {
169 				error = donice(curp, p->p_p, SCARG(uap, prio));
170 				found++;
171 			}
172 		break;
173 	}
174 
175 	default:
176 		return (EINVAL);
177 	}
178 	if (found == 0)
179 		return (ESRCH);
180 	return (error);
181 }
182 
183 int
184 donice(struct proc *curp, struct process *chgpr, int n)
185 {
186 	struct pcred *pcred = curp->p_cred;
187 	struct proc *p;
188 	int s;
189 
190 	if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
191 	    pcred->pc_ucred->cr_uid != chgpr->ps_cred->pc_ucred->cr_uid &&
192 	    pcred->p_ruid != chgpr->ps_cred->pc_ucred->cr_uid)
193 		return (EPERM);
194 	if (n > PRIO_MAX)
195 		n = PRIO_MAX;
196 	if (n < PRIO_MIN)
197 		n = PRIO_MIN;
198 	n += NZERO;
199 	/* XXX wrong: p_nice should be in process */
200 	if (n < chgpr->ps_mainproc->p_nice && suser(curp, 0))
201 		return (EACCES);
202 	chgpr->ps_mainproc->p_nice = n;
203 	SCHED_LOCK(s);
204 	TAILQ_FOREACH(p, &chgpr->ps_threads, p_thr_link)
205 		(void)resetpriority(p);
206 	SCHED_UNLOCK(s);
207 	return (0);
208 }
209 
210 /* ARGSUSED */
211 int
212 sys_setrlimit(struct proc *p, void *v, register_t *retval)
213 {
214 	struct sys_setrlimit_args /* {
215 		syscallarg(int) which;
216 		syscallarg(const struct rlimit *) rlp;
217 	} */ *uap = v;
218 	struct rlimit alim;
219 	int error;
220 
221 	error = copyin((caddr_t)SCARG(uap, rlp), (caddr_t)&alim,
222 		       sizeof (struct rlimit));
223 	if (error)
224 		return (error);
225 	return (dosetrlimit(p, SCARG(uap, which), &alim));
226 }
227 
228 int
229 dosetrlimit(struct proc *p, u_int which, struct rlimit *limp)
230 {
231 	struct rlimit *alimp;
232 	rlim_t maxlim;
233 	int error;
234 
235 	if (which >= RLIM_NLIMITS)
236 		return (EINVAL);
237 
238 	alimp = &p->p_rlimit[which];
239 	if (limp->rlim_cur > alimp->rlim_max ||
240 	    limp->rlim_max > alimp->rlim_max)
241 		if ((error = suser(p, 0)) != 0)
242 			return (error);
243 	if (p->p_p->ps_limit->p_refcnt > 1) {
244 		struct plimit *l = p->p_p->ps_limit;
245 
246 		/* limcopy() can sleep, so copy before decrementing refcnt */
247 		p->p_p->ps_limit = limcopy(l);
248 		l->p_refcnt--;
249 		alimp = &p->p_rlimit[which];
250 	}
251 
252 	switch (which) {
253 	case RLIMIT_DATA:
254 		maxlim = maxdmap;
255 		break;
256 	case RLIMIT_STACK:
257 		maxlim = maxsmap;
258 		break;
259 	case RLIMIT_NOFILE:
260 		maxlim = maxfiles;
261 		break;
262 	case RLIMIT_NPROC:
263 		maxlim = maxproc;
264 		break;
265 	default:
266 		maxlim = RLIM_INFINITY;
267 		break;
268 	}
269 
270 	if (limp->rlim_max > maxlim)
271 		limp->rlim_max = maxlim;
272 	if (limp->rlim_cur > limp->rlim_max)
273 		limp->rlim_cur = limp->rlim_max;
274 
275 	if (which == RLIMIT_STACK) {
276 		/*
277 		 * Stack is allocated to the max at exec time with only
278 		 * "rlim_cur" bytes accessible.  If stack limit is going
279 		 * up make more accessible, if going down make inaccessible.
280 		 */
281 		if (limp->rlim_cur != alimp->rlim_cur) {
282 			vaddr_t addr;
283 			vsize_t size;
284 			vm_prot_t prot;
285 
286 			if (limp->rlim_cur > alimp->rlim_cur) {
287 				prot = VM_PROT_READ|VM_PROT_WRITE;
288 				size = limp->rlim_cur - alimp->rlim_cur;
289 #ifdef MACHINE_STACK_GROWS_UP
290 				addr = USRSTACK + alimp->rlim_cur;
291 #else
292 				addr = USRSTACK - limp->rlim_cur;
293 #endif
294 			} else {
295 				prot = VM_PROT_NONE;
296 				size = alimp->rlim_cur - limp->rlim_cur;
297 #ifdef MACHINE_STACK_GROWS_UP
298 				addr = USRSTACK + limp->rlim_cur;
299 #else
300 				addr = USRSTACK - alimp->rlim_cur;
301 #endif
302 			}
303 			addr = trunc_page(addr);
304 			size = round_page(size);
305 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
306 					      addr, addr+size, prot, FALSE);
307 		}
308 	}
309 
310 	*alimp = *limp;
311 	return (0);
312 }
313 
314 /* ARGSUSED */
315 int
316 sys_getrlimit(struct proc *p, void *v, register_t *retval)
317 {
318 	struct sys_getrlimit_args /* {
319 		syscallarg(int) which;
320 		syscallarg(struct rlimit *) rlp;
321 	} */ *uap = v;
322 
323 	if (SCARG(uap, which) < 0 || SCARG(uap, which) >= RLIM_NLIMITS)
324 		return (EINVAL);
325 	return (copyout((caddr_t)&p->p_rlimit[SCARG(uap, which)],
326 	    (caddr_t)SCARG(uap, rlp), sizeof (struct rlimit)));
327 }
328 
329 /*
330  * Transform the running time and tick information in proc p into user,
331  * system, and interrupt time usage.
332  */
333 void
334 calcru(struct proc *p, struct timeval *up, struct timeval *sp,
335     struct timeval *ip)
336 {
337 	u_quad_t st, ut, it;
338 	int freq;
339 	int s;
340 
341 	s = splstatclock();
342 	st = p->p_sticks;
343 	ut = p->p_uticks;
344 	it = p->p_iticks;
345 	splx(s);
346 
347 	if (st + ut + it == 0) {
348 		timerclear(up);
349 		timerclear(sp);
350 		if (ip != NULL)
351 			timerclear(ip);
352 		return;
353 	}
354 
355 	freq = stathz ? stathz : hz;
356 
357 	st = st * 1000000 / freq;
358 	sp->tv_sec = st / 1000000;
359 	sp->tv_usec = st % 1000000;
360 	ut = ut * 1000000 / freq;
361 	up->tv_sec = ut / 1000000;
362 	up->tv_usec = ut % 1000000;
363 	if (ip != NULL) {
364 		it = it * 1000000 / freq;
365 		ip->tv_sec = it / 1000000;
366 		ip->tv_usec = it % 1000000;
367 	}
368 }
369 
370 /* ARGSUSED */
371 int
372 sys_getrusage(struct proc *p, void *v, register_t *retval)
373 {
374 	struct sys_getrusage_args /* {
375 		syscallarg(int) who;
376 		syscallarg(struct rusage *) rusage;
377 	} */ *uap = v;
378 	struct process *pr = p->p_p;
379 	struct rusage ru;
380 	struct rusage *rup;
381 
382 	switch (SCARG(uap, who)) {
383 
384 	case RUSAGE_SELF:
385 		calcru(p, &p->p_stats->p_ru.ru_utime,
386 		    &p->p_stats->p_ru.ru_stime, NULL);
387 		ru = p->p_stats->p_ru;
388 		rup = &ru;
389 
390 		/* XXX add on already dead threads */
391 
392 		/* add on other living threads */
393 		{
394 			struct proc *q;
395 
396 			TAILQ_FOREACH(q, &pr->ps_threads, p_thr_link) {
397 				if (q == p || P_ZOMBIE(q))
398 					continue;
399 				/*
400 				 * XXX this is approximate: no call
401 				 * to calcru in other running threads
402 				 */
403 				ruadd(rup, &q->p_stats->p_ru);
404 			}
405 		}
406 		break;
407 
408 	case RUSAGE_THREAD:
409 		rup = &p->p_stats->p_ru;
410 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
411 		ru = *rup;
412 		break;
413 
414 	case RUSAGE_CHILDREN:
415 		rup = &p->p_stats->p_cru;
416 		break;
417 
418 	default:
419 		return (EINVAL);
420 	}
421 	return (copyout((caddr_t)rup, (caddr_t)SCARG(uap, rusage),
422 	    sizeof (struct rusage)));
423 }
424 
425 void
426 ruadd(struct rusage *ru, struct rusage *ru2)
427 {
428 	long *ip, *ip2;
429 	int i;
430 
431 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
432 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
433 	if (ru->ru_maxrss < ru2->ru_maxrss)
434 		ru->ru_maxrss = ru2->ru_maxrss;
435 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
436 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
437 		*ip++ += *ip2++;
438 }
439 
440 struct pool plimit_pool;
441 
442 /*
443  * Make a copy of the plimit structure.
444  * We share these structures copy-on-write after fork,
445  * and copy when a limit is changed.
446  */
447 struct plimit *
448 limcopy(struct plimit *lim)
449 {
450 	struct plimit *newlim;
451 	static int initialized;
452 
453 	if (!initialized) {
454 		pool_init(&plimit_pool, sizeof(struct plimit), 0, 0, 0,
455 		    "plimitpl", &pool_allocator_nointr);
456 		initialized = 1;
457 	}
458 
459 	newlim = pool_get(&plimit_pool, PR_WAITOK);
460 	bcopy(lim->pl_rlimit, newlim->pl_rlimit,
461 	    sizeof(struct rlimit) * RLIM_NLIMITS);
462 	newlim->p_refcnt = 1;
463 	return (newlim);
464 }
465 
466 void
467 limfree(struct plimit *lim)
468 {
469 	if (--lim->p_refcnt > 0)
470 		return;
471 	pool_put(&plimit_pool, lim);
472 }
473