xref: /openbsd-src/sys/kern/kern_resource.c (revision 62a742911104f98b9185b2c6b6007d9b1c36396c)
1 /*	$OpenBSD: kern_resource.c,v 1.8 1999/02/26 05:10:40 art 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. All advertising materials mentioning features or use of this software
22  *    must display the following acknowledgement:
23  *	This product includes software developed by the University of
24  *	California, Berkeley and its contributors.
25  * 4. Neither the name of the University nor the names of its contributors
26  *    may be used to endorse or promote products derived from this software
27  *    without specific prior written permission.
28  *
29  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
30  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
31  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
32  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
33  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
34  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
35  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
36  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
37  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
38  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
39  * SUCH DAMAGE.
40  *
41  *	@(#)kern_resource.c	8.5 (Berkeley) 1/21/94
42  */
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/kernel.h>
47 #include <sys/file.h>
48 #include <sys/resourcevar.h>
49 #include <sys/malloc.h>
50 #include <sys/proc.h>
51 
52 #include <sys/mount.h>
53 #include <sys/syscallargs.h>
54 
55 #include <vm/vm.h>
56 
57 #if defined(UVM)
58 #include <uvm/uvm_extern.h>
59 #endif
60 
61 void limfree __P((struct plimit *));
62 /*
63  * Resource controls and accounting.
64  */
65 
66 int
67 sys_getpriority(curp, v, retval)
68 	struct proc *curp;
69 	void *v;
70 	register_t *retval;
71 {
72 	register struct sys_getpriority_args /* {
73 		syscallarg(int) which;
74 		syscallarg(int) who;
75 	} */ *uap = v;
76 	register struct proc *p;
77 	register int low = NZERO + PRIO_MAX + 1;
78 
79 	switch (SCARG(uap, which)) {
80 
81 	case PRIO_PROCESS:
82 		if (SCARG(uap, who) == 0)
83 			p = curp;
84 		else
85 			p = pfind(SCARG(uap, who));
86 		if (p == 0)
87 			break;
88 		low = p->p_nice;
89 		break;
90 
91 	case PRIO_PGRP: {
92 		register struct pgrp *pg;
93 
94 		if (SCARG(uap, who) == 0)
95 			pg = curp->p_pgrp;
96 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
97 			break;
98 		for (p = pg->pg_members.lh_first; p != 0; p = p->p_pglist.le_next) {
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 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
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(curp, v, retval)
126 	struct proc *curp;
127 	void *v;
128 	register_t *retval;
129 {
130 	register struct sys_setpriority_args /* {
131 		syscallarg(int) which;
132 		syscallarg(int) who;
133 		syscallarg(int) prio;
134 	} */ *uap = v;
135 	register struct proc *p;
136 	int found = 0, error = 0;
137 
138 	switch (SCARG(uap, which)) {
139 
140 	case PRIO_PROCESS:
141 		if (SCARG(uap, who) == 0)
142 			p = curp;
143 		else
144 			p = pfind(SCARG(uap, who));
145 		if (p == 0)
146 			break;
147 		error = donice(curp, p, SCARG(uap, prio));
148 		found++;
149 		break;
150 
151 	case PRIO_PGRP: {
152 		register struct pgrp *pg;
153 
154 		if (SCARG(uap, who) == 0)
155 			pg = curp->p_pgrp;
156 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
157 			break;
158 		for (p = pg->pg_members.lh_first; p != 0;
159 		    p = p->p_pglist.le_next) {
160 			error = donice(curp, p, SCARG(uap, prio));
161 			found++;
162 		}
163 		break;
164 	}
165 
166 	case PRIO_USER:
167 		if (SCARG(uap, who) == 0)
168 			SCARG(uap, who) = curp->p_ucred->cr_uid;
169 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
170 			if (p->p_ucred->cr_uid == SCARG(uap, who)) {
171 				error = donice(curp, p, SCARG(uap, prio));
172 				found++;
173 			}
174 		break;
175 
176 	default:
177 		return (EINVAL);
178 	}
179 	if (found == 0)
180 		return (ESRCH);
181 	return (error);
182 }
183 
184 int
185 donice(curp, chgp, n)
186 	register struct proc *curp, *chgp;
187 	register int n;
188 {
189 	register struct pcred *pcred = curp->p_cred;
190 
191 	if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
192 	    pcred->pc_ucred->cr_uid != chgp->p_ucred->cr_uid &&
193 	    pcred->p_ruid != chgp->p_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 < chgp->p_nice && suser(pcred->pc_ucred, &curp->p_acflag))
201 		return (EACCES);
202 	chgp->p_nice = n;
203 	(void)resetpriority(chgp);
204 	return (0);
205 }
206 
207 /* ARGSUSED */
208 int
209 sys_setrlimit(p, v, retval)
210 	struct proc *p;
211 	void *v;
212 	register_t *retval;
213 {
214 	register struct sys_setrlimit_args /* {
215 		syscallarg(u_int) which;
216 		syscallarg(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(p, which, limp)
230 	struct proc *p;
231 	u_int which;
232 	struct rlimit *limp;
233 {
234 	register struct rlimit *alimp;
235 	extern unsigned maxdmap, maxsmap;
236 	int error;
237 
238 	if (which >= RLIM_NLIMITS)
239 		return (EINVAL);
240 
241 	if (limp->rlim_cur < 0 || limp->rlim_max < 0)
242 		return (EINVAL);
243 
244 	alimp = &p->p_rlimit[which];
245 	if (limp->rlim_cur > alimp->rlim_max ||
246 	    limp->rlim_max > alimp->rlim_max)
247 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
248 			return (error);
249 	if (limp->rlim_cur > limp->rlim_max)
250 		limp->rlim_cur = limp->rlim_max;
251 	if (p->p_limit->p_refcnt > 1 &&
252 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
253 		p->p_limit->p_refcnt--;
254 		p->p_limit = limcopy(p->p_limit);
255 		alimp = &p->p_rlimit[which];
256 	}
257 
258 	switch (which) {
259 
260 	case RLIMIT_DATA:
261 		if (limp->rlim_cur > maxdmap)
262 			limp->rlim_cur = maxdmap;
263 		if (limp->rlim_max > maxdmap)
264 			limp->rlim_max = maxdmap;
265 		break;
266 
267 	case RLIMIT_STACK:
268 		if (limp->rlim_cur > maxsmap)
269 			limp->rlim_cur = maxsmap;
270 		if (limp->rlim_max > maxsmap)
271 			limp->rlim_max = maxsmap;
272 		/*
273 		 * Stack is allocated to the max at exec time with only
274 		 * "rlim_cur" bytes accessible.  If stack limit is going
275 		 * up make more accessible, if going down make inaccessible.
276 		 */
277 		if (limp->rlim_cur != alimp->rlim_cur) {
278 			vm_offset_t addr;
279 			vm_size_t size;
280 			vm_prot_t prot;
281 
282 			if (limp->rlim_cur > alimp->rlim_cur) {
283 				prot = VM_PROT_ALL;
284 				size = limp->rlim_cur - alimp->rlim_cur;
285 				addr = USRSTACK - limp->rlim_cur;
286 			} else {
287 				prot = VM_PROT_NONE;
288 				size = alimp->rlim_cur - limp->rlim_cur;
289 				addr = USRSTACK - alimp->rlim_cur;
290 			}
291 			addr = trunc_page(addr);
292 			size = round_page(size);
293 #if defined(UVM)
294 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
295 					      addr, addr+size, prot, FALSE);
296 #else
297 			(void) vm_map_protect(&p->p_vmspace->vm_map,
298 					      addr, addr+size, prot, FALSE);
299 #endif
300 		}
301 		break;
302 
303 	case RLIMIT_NOFILE:
304 		if (limp->rlim_cur > maxfiles)
305 			limp->rlim_cur = maxfiles;
306 		if (limp->rlim_max > maxfiles)
307 			limp->rlim_max = maxfiles;
308 		break;
309 
310 	case RLIMIT_NPROC:
311 		if (limp->rlim_cur > maxproc)
312 			limp->rlim_cur = maxproc;
313 		if (limp->rlim_max > maxproc)
314 			limp->rlim_max = maxproc;
315 		break;
316 	}
317 	*alimp = *limp;
318 	return (0);
319 }
320 
321 /* ARGSUSED */
322 int
323 sys_getrlimit(p, v, retval)
324 	struct proc *p;
325 	void *v;
326 	register_t *retval;
327 {
328 	register struct sys_getrlimit_args /* {
329 		syscallarg(u_int) which;
330 		syscallarg(struct rlimit *) rlp;
331 	} */ *uap = v;
332 
333 	if (SCARG(uap, which) >= RLIM_NLIMITS)
334 		return (EINVAL);
335 	return (copyout((caddr_t)&p->p_rlimit[SCARG(uap, which)],
336 	    (caddr_t)SCARG(uap, rlp), sizeof (struct rlimit)));
337 }
338 
339 /*
340  * Transform the running time and tick information in proc p into user,
341  * system, and interrupt time usage.
342  */
343 void
344 calcru(p, up, sp, ip)
345 	register struct proc *p;
346 	register struct timeval *up;
347 	register struct timeval *sp;
348 	register struct timeval *ip;
349 {
350 	register u_quad_t u, st, ut, it, tot;
351 	register long sec, usec;
352 	register int s;
353 	struct timeval tv;
354 
355 	s = splstatclock();
356 	st = p->p_sticks;
357 	ut = p->p_uticks;
358 	it = p->p_iticks;
359 	splx(s);
360 
361 	tot = st + ut + it;
362 	if (tot == 0) {
363 		up->tv_sec = up->tv_usec = 0;
364 		sp->tv_sec = sp->tv_usec = 0;
365 		if (ip != NULL)
366 			ip->tv_sec = ip->tv_usec = 0;
367 		return;
368 	}
369 
370 	sec = p->p_rtime.tv_sec;
371 	usec = p->p_rtime.tv_usec;
372 	if (p == curproc) {
373 		/*
374 		 * Adjust for the current time slice.  This is actually fairly
375 		 * important since the error here is on the order of a time
376 		 * quantum, which is much greater than the sampling error.
377 		 */
378 		microtime(&tv);
379 		sec += tv.tv_sec - runtime.tv_sec;
380 		usec += tv.tv_usec - runtime.tv_usec;
381 	}
382 	u = (u_quad_t) sec * 1000000 + usec;
383 	st = (u * st) / tot;
384 	sp->tv_sec = st / 1000000;
385 	sp->tv_usec = st % 1000000;
386 	ut = (u * ut) / tot;
387 	up->tv_sec = ut / 1000000;
388 	up->tv_usec = ut % 1000000;
389 	if (ip != NULL) {
390 		it = (u * it) / tot;
391 		ip->tv_sec = it / 1000000;
392 		ip->tv_usec = it % 1000000;
393 	}
394 }
395 
396 /* ARGSUSED */
397 int
398 sys_getrusage(p, v, retval)
399 	register struct proc *p;
400 	void *v;
401 	register_t *retval;
402 {
403 	register struct sys_getrusage_args /* {
404 		syscallarg(int) who;
405 		syscallarg(struct rusage *) rusage;
406 	} */ *uap = v;
407 	register struct rusage *rup;
408 
409 	switch (SCARG(uap, who)) {
410 
411 	case RUSAGE_SELF:
412 		rup = &p->p_stats->p_ru;
413 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
414 		break;
415 
416 	case RUSAGE_CHILDREN:
417 		rup = &p->p_stats->p_cru;
418 		break;
419 
420 	default:
421 		return (EINVAL);
422 	}
423 	return (copyout((caddr_t)rup, (caddr_t)SCARG(uap, rusage),
424 	    sizeof (struct rusage)));
425 }
426 
427 void
428 ruadd(ru, ru2)
429 	register struct rusage *ru, *ru2;
430 {
431 	register long *ip, *ip2;
432 	register int i;
433 
434 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
435 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
436 	if (ru->ru_maxrss < ru2->ru_maxrss)
437 		ru->ru_maxrss = ru2->ru_maxrss;
438 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
439 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
440 		*ip++ += *ip2++;
441 }
442 
443 /*
444  * Make a copy of the plimit structure.
445  * We share these structures copy-on-write after fork,
446  * and copy when a limit is changed.
447  */
448 struct plimit *
449 limcopy(lim)
450 	struct plimit *lim;
451 {
452 	register struct plimit *newlim;
453 
454 	MALLOC(newlim, struct plimit *, sizeof(struct plimit),
455 	    M_SUBPROC, M_WAITOK);
456 	bcopy(lim->pl_rlimit, newlim->pl_rlimit,
457 	    sizeof(struct rlimit) * RLIM_NLIMITS);
458 	newlim->p_lflags = 0;
459 	newlim->p_refcnt = 1;
460 	return (newlim);
461 }
462 
463 void
464 limfree(lim)
465 	struct plimit *lim;
466 {
467 
468 	if (--lim->p_refcnt > 0)
469 		return;
470 	FREE(lim, M_SUBPROC);
471 }
472