xref: /openbsd-src/sys/kern/kern_resource.c (revision b33a8d55775a3013157d6b0e3602c1fe178b48a6) 
1 /*	$OpenBSD: kern_resource.c,v 1.54 2015/02/09 09:39:09 miod 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 
57 /*
58  * Patchable maximum data and stack limits.
59  */
60 rlim_t maxdmap = MAXDSIZ;
61 rlim_t maxsmap = MAXSSIZ;
62 
63 /*
64  * Resource controls and accounting.
65  */
66 
67 int
68 sys_getpriority(struct proc *curp, void *v, register_t *retval)
69 {
70 	struct sys_getpriority_args /* {
71 		syscallarg(int) which;
72 		syscallarg(id_t) who;
73 	} */ *uap = v;
74 	struct process *pr;
75 	int low = NZERO + PRIO_MAX + 1;
76 
77 	switch (SCARG(uap, which)) {
78 
79 	case PRIO_PROCESS:
80 		if (SCARG(uap, who) == 0)
81 			pr = curp->p_p;
82 		else
83 			pr = prfind(SCARG(uap, who));
84 		if (pr == NULL)
85 			break;
86 		if (pr->ps_nice < low)
87 			low = pr->ps_nice;
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 			if (pr->ps_nice < low)
99 				low = pr->ps_nice;
100 		break;
101 	}
102 
103 	case PRIO_USER:
104 		if (SCARG(uap, who) == 0)
105 			SCARG(uap, who) = curp->p_ucred->cr_uid;
106 		LIST_FOREACH(pr, &allprocess, ps_list)
107 			if (pr->ps_ucred->cr_uid == SCARG(uap, who) &&
108 			    pr->ps_nice < low)
109 				low = pr->ps_nice;
110 		break;
111 
112 	default:
113 		return (EINVAL);
114 	}
115 	if (low == NZERO + PRIO_MAX + 1)
116 		return (ESRCH);
117 	*retval = low - NZERO;
118 	return (0);
119 }
120 
121 /* ARGSUSED */
122 int
123 sys_setpriority(struct proc *curp, void *v, register_t *retval)
124 {
125 	struct sys_setpriority_args /* {
126 		syscallarg(int) which;
127 		syscallarg(id_t) who;
128 		syscallarg(int) prio;
129 	} */ *uap = v;
130 	struct process *pr;
131 	int found = 0, error = 0;
132 
133 	switch (SCARG(uap, which)) {
134 
135 	case PRIO_PROCESS:
136 		if (SCARG(uap, who) == 0)
137 			pr = curp->p_p;
138 		else
139 			pr = prfind(SCARG(uap, who));
140 		if (pr == NULL)
141 			break;
142 		error = donice(curp, pr, SCARG(uap, prio));
143 		found++;
144 		break;
145 
146 	case PRIO_PGRP: {
147 		struct pgrp *pg;
148 
149 		if (SCARG(uap, who) == 0)
150 			pg = curp->p_p->ps_pgrp;
151 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
152 			break;
153 		LIST_FOREACH(pr, &pg->pg_members, ps_pglist) {
154 			error = donice(curp, pr, SCARG(uap, prio));
155 			found++;
156 		}
157 		break;
158 	}
159 
160 	case PRIO_USER:
161 		if (SCARG(uap, who) == 0)
162 			SCARG(uap, who) = curp->p_ucred->cr_uid;
163 		LIST_FOREACH(pr, &allprocess, ps_list)
164 			if (pr->ps_ucred->cr_uid == SCARG(uap, who)) {
165 				error = donice(curp, pr, SCARG(uap, prio));
166 				found++;
167 			}
168 		break;
169 
170 	default:
171 		return (EINVAL);
172 	}
173 	if (found == 0)
174 		return (ESRCH);
175 	return (error);
176 }
177 
178 int
179 donice(struct proc *curp, struct process *chgpr, int n)
180 {
181 	struct ucred *ucred = curp->p_ucred;
182 	struct proc *p;
183 	int s;
184 
185 	if (ucred->cr_uid != 0 && ucred->cr_ruid != 0 &&
186 	    ucred->cr_uid != chgpr->ps_ucred->cr_uid &&
187 	    ucred->cr_ruid != chgpr->ps_ucred->cr_uid)
188 		return (EPERM);
189 	if (n > PRIO_MAX)
190 		n = PRIO_MAX;
191 	if (n < PRIO_MIN)
192 		n = PRIO_MIN;
193 	n += NZERO;
194 	if (n < chgpr->ps_nice && suser(curp, 0))
195 		return (EACCES);
196 	chgpr->ps_nice = n;
197 	SCHED_LOCK(s);
198 	TAILQ_FOREACH(p, &chgpr->ps_threads, p_thr_link)
199 		(void)resetpriority(p);
200 	SCHED_UNLOCK(s);
201 	return (0);
202 }
203 
204 /* ARGSUSED */
205 int
206 sys_setrlimit(struct proc *p, void *v, register_t *retval)
207 {
208 	struct sys_setrlimit_args /* {
209 		syscallarg(int) which;
210 		syscallarg(const struct rlimit *) rlp;
211 	} */ *uap = v;
212 	struct rlimit alim;
213 	int error;
214 
215 	error = copyin((caddr_t)SCARG(uap, rlp), (caddr_t)&alim,
216 		       sizeof (struct rlimit));
217 	if (error)
218 		return (error);
219 #ifdef KTRACE
220 	if (KTRPOINT(p, KTR_STRUCT))
221 		ktrrlimit(p, &alim);
222 #endif
223 	return (dosetrlimit(p, SCARG(uap, which), &alim));
224 }
225 
226 int
227 dosetrlimit(struct proc *p, u_int which, struct rlimit *limp)
228 {
229 	struct rlimit *alimp;
230 	rlim_t maxlim;
231 	int error;
232 
233 	if (which >= RLIM_NLIMITS || limp->rlim_cur > limp->rlim_max)
234 		return (EINVAL);
235 
236 	alimp = &p->p_rlimit[which];
237 	if (limp->rlim_max > alimp->rlim_max)
238 		if ((error = suser(p, 0)) != 0)
239 			return (error);
240 	if (p->p_p->ps_limit->p_refcnt > 1) {
241 		struct plimit *l = p->p_p->ps_limit;
242 
243 		/* limcopy() can sleep, so copy before decrementing refcnt */
244 		p->p_p->ps_limit = limcopy(l);
245 		limfree(l);
246 		alimp = &p->p_rlimit[which];
247 	}
248 
249 	switch (which) {
250 	case RLIMIT_DATA:
251 		maxlim = maxdmap;
252 		break;
253 	case RLIMIT_STACK:
254 		maxlim = maxsmap;
255 		break;
256 	case RLIMIT_NOFILE:
257 		maxlim = maxfiles;
258 		break;
259 	case RLIMIT_NPROC:
260 		maxlim = maxprocess;
261 		break;
262 	default:
263 		maxlim = RLIM_INFINITY;
264 		break;
265 	}
266 
267 	if (limp->rlim_max > maxlim)
268 		limp->rlim_max = maxlim;
269 	if (limp->rlim_cur > limp->rlim_max)
270 		limp->rlim_cur = limp->rlim_max;
271 
272 	if (which == RLIMIT_STACK) {
273 		/*
274 		 * Stack is allocated to the max at exec time with only
275 		 * "rlim_cur" bytes accessible.  If stack limit is going
276 		 * up make more accessible, if going down make inaccessible.
277 		 */
278 		if (limp->rlim_cur != alimp->rlim_cur) {
279 			vaddr_t addr;
280 			vsize_t size;
281 			vm_prot_t prot;
282 			struct vmspace *vm = p->p_vmspace;
283 
284 			if (limp->rlim_cur > alimp->rlim_cur) {
285 				prot = PROT_READ | PROT_WRITE;
286 				size = limp->rlim_cur - alimp->rlim_cur;
287 #ifdef MACHINE_STACK_GROWS_UP
288 				addr = (vaddr_t)vm->vm_maxsaddr +
289 				    alimp->rlim_cur;
290 #else
291 				addr = (vaddr_t)vm->vm_minsaddr -
292 				    limp->rlim_cur;
293 #endif
294 			} else {
295 				prot = PROT_NONE;
296 				size = alimp->rlim_cur - limp->rlim_cur;
297 #ifdef MACHINE_STACK_GROWS_UP
298 				addr = (vaddr_t)vm->vm_maxsaddr +
299 				    limp->rlim_cur;
300 #else
301 				addr = (vaddr_t)vm->vm_minsaddr -
302 				    alimp->rlim_cur;
303 #endif
304 			}
305 			addr = trunc_page(addr);
306 			size = round_page(size);
307 			(void) uvm_map_protect(&vm->vm_map,
308 					      addr, addr+size, prot, FALSE);
309 		}
310 	}
311 
312 	*alimp = *limp;
313 	return (0);
314 }
315 
316 /* ARGSUSED */
317 int
318 sys_getrlimit(struct proc *p, void *v, register_t *retval)
319 {
320 	struct sys_getrlimit_args /* {
321 		syscallarg(int) which;
322 		syscallarg(struct rlimit *) rlp;
323 	} */ *uap = v;
324 	struct rlimit *alimp;
325 	int error;
326 
327 	if (SCARG(uap, which) < 0 || SCARG(uap, which) >= RLIM_NLIMITS)
328 		return (EINVAL);
329 	alimp = &p->p_rlimit[SCARG(uap, which)];
330 	error = copyout(alimp, SCARG(uap, rlp), sizeof(struct rlimit));
331 #ifdef KTRACE
332 	if (error == 0 && KTRPOINT(p, KTR_STRUCT))
333 		ktrrlimit(p, alimp);
334 #endif
335 	return (error);
336 }
337 
338 void
339 tuagg_sub(struct tusage *tup, struct proc *p)
340 {
341 	timespecadd(&tup->tu_runtime, &p->p_rtime, &tup->tu_runtime);
342 	tup->tu_uticks += p->p_uticks;
343 	tup->tu_sticks += p->p_sticks;
344 	tup->tu_iticks += p->p_iticks;
345 }
346 
347 /*
348  * Aggregate a single thread's immediate time counts into the running
349  * totals for the thread and process
350  */
351 void
352 tuagg_unlocked(struct process *pr, struct proc *p)
353 {
354 	tuagg_sub(&pr->ps_tu, p);
355 	tuagg_sub(&p->p_tu, p);
356 	timespecclear(&p->p_rtime);
357 	p->p_uticks = 0;
358 	p->p_sticks = 0;
359 	p->p_iticks = 0;
360 }
361 
362 void
363 tuagg(struct process *pr, struct proc *p)
364 {
365 	int s;
366 
367 	SCHED_LOCK(s);
368 	tuagg_unlocked(pr, p);
369 	SCHED_UNLOCK(s);
370 }
371 
372 /*
373  * Transform the running time and tick information in a struct tusage
374  * into user, system, and interrupt time usage.
375  */
376 void
377 calctsru(struct tusage *tup, struct timespec *up, struct timespec *sp,
378     struct timespec *ip)
379 {
380 	u_quad_t st, ut, it;
381 	int freq;
382 
383 	st = tup->tu_sticks;
384 	ut = tup->tu_uticks;
385 	it = tup->tu_iticks;
386 
387 	if (st + ut + it == 0) {
388 		timespecclear(up);
389 		timespecclear(sp);
390 		if (ip != NULL)
391 			timespecclear(ip);
392 		return;
393 	}
394 
395 	freq = stathz ? stathz : hz;
396 
397 	st = st * 1000000000 / freq;
398 	sp->tv_sec = st / 1000000000;
399 	sp->tv_nsec = st % 1000000000;
400 	ut = ut * 1000000000 / freq;
401 	up->tv_sec = ut / 1000000000;
402 	up->tv_nsec = ut % 1000000000;
403 	if (ip != NULL) {
404 		it = it * 1000000000 / freq;
405 		ip->tv_sec = it / 1000000000;
406 		ip->tv_nsec = it % 1000000000;
407 	}
408 }
409 
410 void
411 calcru(struct tusage *tup, struct timeval *up, struct timeval *sp,
412     struct timeval *ip)
413 {
414 	struct timespec u, s, i;
415 
416 	calctsru(tup, &u, &s, ip != NULL ? &i : NULL);
417 	TIMESPEC_TO_TIMEVAL(up, &u);
418 	TIMESPEC_TO_TIMEVAL(sp, &s);
419 	if (ip != NULL)
420 		TIMESPEC_TO_TIMEVAL(ip, &i);
421 }
422 
423 /* ARGSUSED */
424 int
425 sys_getrusage(struct proc *p, void *v, register_t *retval)
426 {
427 	struct sys_getrusage_args /* {
428 		syscallarg(int) who;
429 		syscallarg(struct rusage *) rusage;
430 	} */ *uap = v;
431 	struct rusage ru;
432 	int error;
433 
434 	error = dogetrusage(p, SCARG(uap, who), &ru);
435 	if (error == 0) {
436 		error = copyout(&ru, SCARG(uap, rusage), sizeof(ru));
437 #ifdef KTRACE
438 		if (error == 0 && KTRPOINT(p, KTR_STRUCT))
439 			ktrrusage(p, &ru);
440 #endif
441 	}
442 	return (error);
443 }
444 
445 int
446 dogetrusage(struct proc *p, int who, struct rusage *rup)
447 {
448 	struct process *pr = p->p_p;
449 	struct proc *q;
450 
451 	switch (who) {
452 
453 	case RUSAGE_SELF:
454 		/* start with the sum of dead threads, if any */
455 		if (pr->ps_ru != NULL)
456 			*rup = *pr->ps_ru;
457 		else
458 			memset(rup, 0, sizeof(*rup));
459 
460 		/* add on all living threads */
461 		TAILQ_FOREACH(q, &pr->ps_threads, p_thr_link) {
462 			ruadd(rup, &q->p_ru);
463 			tuagg(pr, q);
464 		}
465 
466 		calcru(&pr->ps_tu, &rup->ru_utime, &rup->ru_stime, NULL);
467 		break;
468 
469 	case RUSAGE_THREAD:
470 		*rup = p->p_ru;
471 		calcru(&p->p_tu, &rup->ru_utime, &rup->ru_stime, NULL);
472 		break;
473 
474 	case RUSAGE_CHILDREN:
475 		*rup = pr->ps_cru;
476 		break;
477 
478 	default:
479 		return (EINVAL);
480 	}
481 	return (0);
482 }
483 
484 void
485 ruadd(struct rusage *ru, struct rusage *ru2)
486 {
487 	long *ip, *ip2;
488 	int i;
489 
490 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
491 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
492 	if (ru->ru_maxrss < ru2->ru_maxrss)
493 		ru->ru_maxrss = ru2->ru_maxrss;
494 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
495 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
496 		*ip++ += *ip2++;
497 }
498 
499 struct pool plimit_pool;
500 
501 /*
502  * Make a copy of the plimit structure.
503  * We share these structures copy-on-write after fork,
504  * and copy when a limit is changed.
505  */
506 struct plimit *
507 limcopy(struct plimit *lim)
508 {
509 	struct plimit *newlim;
510 	static int initialized;
511 
512 	if (!initialized) {
513 		pool_init(&plimit_pool, sizeof(struct plimit), 0, 0, PR_WAITOK,
514 		    "plimitpl", NULL);
515 		initialized = 1;
516 	}
517 
518 	newlim = pool_get(&plimit_pool, PR_WAITOK);
519 	memcpy(newlim->pl_rlimit, lim->pl_rlimit,
520 	    sizeof(struct rlimit) * RLIM_NLIMITS);
521 	newlim->p_refcnt = 1;
522 	return (newlim);
523 }
524 
525 void
526 limfree(struct plimit *lim)
527 {
528 	if (--lim->p_refcnt > 0)
529 		return;
530 	pool_put(&plimit_pool, lim);
531 }
532