xref: /openbsd-src/sys/kern/kern_resource.c (revision 08be1c182fbd154e4a9fe23d2f7954d0b00a3b1d) 
1 /*	$OpenBSD: kern_resource.c,v 1.42 2013/06/03 16:55:22 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/ktrace.h>
48 #include <sys/sched.h>
49 
50 #include <sys/mount.h>
51 #include <sys/syscallargs.h>
52 
53 #include <uvm/uvm_extern.h>
54 
55 void	tuagg_sub(struct tusage *, struct proc *);
56 void	tuagg(struct process *, struct proc *);
57 
58 /*
59  * Patchable maximum data and stack limits.
60  */
61 rlim_t maxdmap = MAXDSIZ;
62 rlim_t maxsmap = MAXSSIZ;
63 
64 /*
65  * Resource controls and accounting.
66  */
67 
68 int
69 sys_getpriority(struct proc *curp, void *v, register_t *retval)
70 {
71 	struct sys_getpriority_args /* {
72 		syscallarg(int) which;
73 		syscallarg(id_t) who;
74 	} */ *uap = v;
75 	struct process *pr;
76 	struct proc *p;
77 	int low = NZERO + PRIO_MAX + 1;
78 
79 	switch (SCARG(uap, which)) {
80 
81 	case PRIO_PROCESS:
82 		if (SCARG(uap, who) == 0)
83 			pr = curp->p_p;
84 		else
85 			pr = prfind(SCARG(uap, who));
86 		if (pr == NULL)
87 			break;
88 		if (pr->ps_nice < low)
89 			low = pr->ps_nice;
90 		break;
91 
92 	case PRIO_PGRP: {
93 		struct pgrp *pg;
94 
95 		if (SCARG(uap, who) == 0)
96 			pg = curp->p_p->ps_pgrp;
97 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
98 			break;
99 		LIST_FOREACH(pr, &pg->pg_members, ps_pglist)
100 			if (pr->ps_nice < low)
101 				low = pr->ps_nice;
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_flag & P_THREAD) == 0 &&
110 			    p->p_ucred->cr_uid == SCARG(uap, who) &&
111 			    p->p_p->ps_nice < low)
112 				low = p->p_p->ps_nice;
113 		break;
114 
115 	default:
116 		return (EINVAL);
117 	}
118 	if (low == NZERO + PRIO_MAX + 1)
119 		return (ESRCH);
120 	*retval = low - NZERO;
121 	return (0);
122 }
123 
124 /* ARGSUSED */
125 int
126 sys_setpriority(struct proc *curp, void *v, register_t *retval)
127 {
128 	struct sys_setpriority_args /* {
129 		syscallarg(int) which;
130 		syscallarg(id_t) who;
131 		syscallarg(int) prio;
132 	} */ *uap = v;
133 	struct process *pr;
134 	int found = 0, error = 0;
135 
136 	switch (SCARG(uap, which)) {
137 
138 	case PRIO_PROCESS:
139 		if (SCARG(uap, who) == 0)
140 			pr = curp->p_p;
141 		else
142 			pr = prfind(SCARG(uap, who));
143 		if (pr == NULL)
144 			break;
145 		error = donice(curp, pr, SCARG(uap, prio));
146 		found++;
147 		break;
148 
149 	case PRIO_PGRP: {
150 		struct pgrp *pg;
151 
152 		if (SCARG(uap, who) == 0)
153 			pg = curp->p_p->ps_pgrp;
154 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
155 			break;
156 		LIST_FOREACH(pr, &pg->pg_members, ps_pglist) {
157 			error = donice(curp, pr, SCARG(uap, prio));
158 			found++;
159 		}
160 		break;
161 	}
162 
163 	case PRIO_USER: {
164 		struct proc *p;
165 		if (SCARG(uap, who) == 0)
166 			SCARG(uap, who) = curp->p_ucred->cr_uid;
167 		LIST_FOREACH(p, &allproc, p_list)
168 			if ((p->p_flag & P_THREAD) == 0 &&
169 			    p->p_ucred->cr_uid == SCARG(uap, who)) {
170 				error = donice(curp, p->p_p, SCARG(uap, prio));
171 				found++;
172 			}
173 		break;
174 	}
175 
176 	default:
177 		return (EINVAL);
178 	}
179 	if (found == 0)
180 		return (ESRCH);
181 	return (error);
182 }
183 
184 int
185 donice(struct proc *curp, struct process *chgpr, int n)
186 {
187 	struct pcred *pcred = curp->p_cred;
188 	struct proc *p;
189 	int s;
190 
191 	if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
192 	    pcred->pc_ucred->cr_uid != chgpr->ps_cred->pc_ucred->cr_uid &&
193 	    pcred->p_ruid != chgpr->ps_cred->pc_ucred->cr_uid)
194 		return (EPERM);
195 	if (n > PRIO_MAX)
196 		n = PRIO_MAX;
197 	if (n < PRIO_MIN)
198 		n = PRIO_MIN;
199 	n += NZERO;
200 	if (n < chgpr->ps_nice && suser(curp, 0))
201 		return (EACCES);
202 	chgpr->ps_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 #ifdef KTRACE
226 	if (KTRPOINT(p, KTR_STRUCT))
227 		ktrrlimit(p, &alim);
228 #endif
229 	return (dosetrlimit(p, SCARG(uap, which), &alim));
230 }
231 
232 int
233 dosetrlimit(struct proc *p, u_int which, struct rlimit *limp)
234 {
235 	struct rlimit *alimp;
236 	rlim_t maxlim;
237 	int error;
238 
239 	if (which >= RLIM_NLIMITS || limp->rlim_cur > limp->rlim_max)
240 		return (EINVAL);
241 
242 	alimp = &p->p_rlimit[which];
243 	if (limp->rlim_max > alimp->rlim_max)
244 		if ((error = suser(p, 0)) != 0)
245 			return (error);
246 	if (p->p_p->ps_limit->p_refcnt > 1) {
247 		struct plimit *l = p->p_p->ps_limit;
248 
249 		/* limcopy() can sleep, so copy before decrementing refcnt */
250 		p->p_p->ps_limit = limcopy(l);
251 		limfree(l);
252 		alimp = &p->p_rlimit[which];
253 	}
254 
255 	switch (which) {
256 	case RLIMIT_DATA:
257 		maxlim = maxdmap;
258 		break;
259 	case RLIMIT_STACK:
260 		maxlim = maxsmap;
261 		break;
262 	case RLIMIT_NOFILE:
263 		maxlim = maxfiles;
264 		break;
265 	case RLIMIT_NPROC:
266 		maxlim = maxprocess;
267 		break;
268 	default:
269 		maxlim = RLIM_INFINITY;
270 		break;
271 	}
272 
273 	if (limp->rlim_max > maxlim)
274 		limp->rlim_max = maxlim;
275 	if (limp->rlim_cur > limp->rlim_max)
276 		limp->rlim_cur = limp->rlim_max;
277 
278 	if (which == RLIMIT_STACK) {
279 		/*
280 		 * Stack is allocated to the max at exec time with only
281 		 * "rlim_cur" bytes accessible.  If stack limit is going
282 		 * up make more accessible, if going down make inaccessible.
283 		 */
284 		if (limp->rlim_cur != alimp->rlim_cur) {
285 			vaddr_t addr;
286 			vsize_t size;
287 			vm_prot_t prot;
288 
289 			if (limp->rlim_cur > alimp->rlim_cur) {
290 				prot = VM_PROT_READ|VM_PROT_WRITE;
291 				size = limp->rlim_cur - alimp->rlim_cur;
292 #ifdef MACHINE_STACK_GROWS_UP
293 				addr = USRSTACK + alimp->rlim_cur;
294 #else
295 				addr = USRSTACK - limp->rlim_cur;
296 #endif
297 			} else {
298 				prot = VM_PROT_NONE;
299 				size = alimp->rlim_cur - limp->rlim_cur;
300 #ifdef MACHINE_STACK_GROWS_UP
301 				addr = USRSTACK + limp->rlim_cur;
302 #else
303 				addr = USRSTACK - alimp->rlim_cur;
304 #endif
305 			}
306 			addr = trunc_page(addr);
307 			size = round_page(size);
308 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
309 					      addr, addr+size, prot, FALSE);
310 		}
311 	}
312 
313 	*alimp = *limp;
314 	return (0);
315 }
316 
317 /* ARGSUSED */
318 int
319 sys_getrlimit(struct proc *p, void *v, register_t *retval)
320 {
321 	struct sys_getrlimit_args /* {
322 		syscallarg(int) which;
323 		syscallarg(struct rlimit *) rlp;
324 	} */ *uap = v;
325 	struct rlimit *alimp;
326 	int error;
327 
328 	if (SCARG(uap, which) < 0 || SCARG(uap, which) >= RLIM_NLIMITS)
329 		return (EINVAL);
330 	alimp = &p->p_rlimit[SCARG(uap, which)];
331 	error = copyout(alimp, SCARG(uap, rlp), sizeof(struct rlimit));
332 #ifdef KTRACE
333 	if (error == 0 && KTRPOINT(p, KTR_STRUCT))
334 		ktrrlimit(p, alimp);
335 #endif
336 	return (error);
337 }
338 
339 void
340 tuagg_sub(struct tusage *tup, struct proc *p)
341 {
342 	timespecadd(&tup->tu_runtime, &p->p_rtime, &tup->tu_runtime);
343 	tup->tu_uticks += p->p_uticks;
344 	tup->tu_sticks += p->p_sticks;
345 	tup->tu_iticks += p->p_iticks;
346 }
347 
348 /*
349  * Aggregate a single thread's immediate time counts into the running
350  * totals for the thread and process
351  */
352 void
353 tuagg_unlocked(struct process *pr, struct proc *p)
354 {
355 	tuagg_sub(&pr->ps_tu, p);
356 	tuagg_sub(&p->p_tu, p);
357 	timespecclear(&p->p_rtime);
358 	p->p_uticks = 0;
359 	p->p_sticks = 0;
360 	p->p_iticks = 0;
361 }
362 
363 void
364 tuagg(struct process *pr, struct proc *p)
365 {
366 	int s;
367 
368 	SCHED_LOCK(s);
369 	tuagg_unlocked(pr, p);
370 	SCHED_UNLOCK(s);
371 }
372 
373 /*
374  * Transform the running time and tick information in a struct tusage
375  * into user, system, and interrupt time usage.
376  */
377 void
378 calctsru(struct tusage *tup, struct timespec *up, struct timespec *sp,
379     struct timespec *ip)
380 {
381 	u_quad_t st, ut, it;
382 	int freq;
383 
384 	st = tup->tu_sticks;
385 	ut = tup->tu_uticks;
386 	it = tup->tu_iticks;
387 
388 	if (st + ut + it == 0) {
389 		timespecclear(up);
390 		timespecclear(sp);
391 		if (ip != NULL)
392 			timespecclear(ip);
393 		return;
394 	}
395 
396 	freq = stathz ? stathz : hz;
397 
398 	st = st * 1000000000 / freq;
399 	sp->tv_sec = st / 1000000000;
400 	sp->tv_nsec = st % 1000000000;
401 	ut = ut * 1000000000 / freq;
402 	up->tv_sec = ut / 1000000000;
403 	up->tv_nsec = ut % 1000000000;
404 	if (ip != NULL) {
405 		it = it * 1000000000 / freq;
406 		ip->tv_sec = it / 1000000000;
407 		ip->tv_nsec = it % 1000000000;
408 	}
409 }
410 
411 void
412 calcru(struct tusage *tup, struct timeval *up, struct timeval *sp,
413     struct timeval *ip)
414 {
415 	struct timespec u, s, i;
416 
417 	calctsru(tup, &u, &s, ip != NULL ? &i : NULL);
418 	TIMESPEC_TO_TIMEVAL(up, &u);
419 	TIMESPEC_TO_TIMEVAL(sp, &s);
420 	if (ip != NULL)
421 		TIMESPEC_TO_TIMEVAL(ip, &i);
422 }
423 
424 
425 /* ARGSUSED */
426 int
427 sys_getrusage(struct proc *p, void *v, register_t *retval)
428 {
429 	struct sys_getrusage_args /* {
430 		syscallarg(int) who;
431 		syscallarg(struct rusage *) rusage;
432 	} */ *uap = v;
433 	struct process *pr = p->p_p;
434 	struct proc *q;
435 	struct rusage ru;
436 	struct rusage *rup;
437 
438 	switch (SCARG(uap, who)) {
439 
440 	case RUSAGE_SELF:
441 		/* start with the sum of dead threads, if any */
442 		if (pr->ps_ru != NULL)
443 			ru = *pr->ps_ru;
444 		else
445 			bzero(&ru, sizeof(ru));
446 		rup = &ru;
447 
448 		/* add on all living threads */
449 		TAILQ_FOREACH(q, &pr->ps_threads, p_thr_link) {
450 			ruadd(rup, &q->p_ru);
451 			tuagg(pr, q);
452 		}
453 
454 		calcru(&pr->ps_tu, &rup->ru_utime, &rup->ru_stime, NULL);
455 		break;
456 
457 	case RUSAGE_THREAD:
458 		rup = &p->p_ru;
459 		calcru(&p->p_tu, &rup->ru_utime, &rup->ru_stime, NULL);
460 		break;
461 
462 	case RUSAGE_CHILDREN:
463 		rup = &pr->ps_cru;
464 		break;
465 
466 	default:
467 		return (EINVAL);
468 	}
469 	return (copyout((caddr_t)rup, (caddr_t)SCARG(uap, rusage),
470 	    sizeof (struct rusage)));
471 }
472 
473 void
474 ruadd(struct rusage *ru, struct rusage *ru2)
475 {
476 	long *ip, *ip2;
477 	int i;
478 
479 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
480 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
481 	if (ru->ru_maxrss < ru2->ru_maxrss)
482 		ru->ru_maxrss = ru2->ru_maxrss;
483 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
484 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
485 		*ip++ += *ip2++;
486 }
487 
488 struct pool plimit_pool;
489 
490 /*
491  * Make a copy of the plimit structure.
492  * We share these structures copy-on-write after fork,
493  * and copy when a limit is changed.
494  */
495 struct plimit *
496 limcopy(struct plimit *lim)
497 {
498 	struct plimit *newlim;
499 	static int initialized;
500 
501 	if (!initialized) {
502 		pool_init(&plimit_pool, sizeof(struct plimit), 0, 0, 0,
503 		    "plimitpl", &pool_allocator_nointr);
504 		initialized = 1;
505 	}
506 
507 	newlim = pool_get(&plimit_pool, PR_WAITOK);
508 	bcopy(lim->pl_rlimit, newlim->pl_rlimit,
509 	    sizeof(struct rlimit) * RLIM_NLIMITS);
510 	newlim->p_refcnt = 1;
511 	return (newlim);
512 }
513 
514 void
515 limfree(struct plimit *lim)
516 {
517 	if (--lim->p_refcnt > 0)
518 		return;
519 	pool_put(&plimit_pool, lim);
520 }
521