Home | History | Annotate | Line # | Download | only in kern
kern_resource.c revision 1.52
      1 /*	$NetBSD: kern_resource.c,v 1.52 1999/07/25 06:30:34 thorpej 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	8.8 (Berkeley) 2/14/95
     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/pool.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 #include <uvm/uvm_extern.h>
     58 
     59 void limfree __P((struct plimit *));
     60 /*
     61  * Resource controls and accounting.
     62  */
     63 
     64 int
     65 sys_getpriority(curp, v, retval)
     66 	struct proc *curp;
     67 	void *v;
     68 	register_t *retval;
     69 {
     70 	register struct sys_getpriority_args /* {
     71 		syscallarg(int) which;
     72 		syscallarg(int) who;
     73 	} */ *uap = v;
     74 	register struct proc *p;
     75 	register int low = NZERO + PRIO_MAX + 1;
     76 
     77 	switch (SCARG(uap, which)) {
     78 
     79 	case PRIO_PROCESS:
     80 		if (SCARG(uap, who) == 0)
     81 			p = curp;
     82 		else
     83 			p = pfind(SCARG(uap, who));
     84 		if (p == 0)
     85 			break;
     86 		low = p->p_nice;
     87 		break;
     88 
     89 	case PRIO_PGRP: {
     90 		register struct pgrp *pg;
     91 
     92 		if (SCARG(uap, who) == 0)
     93 			pg = curp->p_pgrp;
     94 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
     95 			break;
     96 		for (p = pg->pg_members.lh_first; p != 0;
     97 		     p = p->p_pglist.le_next) {
     98 			if (p->p_nice < low)
     99 				low = p->p_nice;
    100 		}
    101 		break;
    102 	}
    103 
    104 	case PRIO_USER:
    105 		if (SCARG(uap, who) == 0)
    106 			SCARG(uap, who) = curp->p_ucred->cr_uid;
    107 		proclist_lock_read();
    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 		proclist_unlock_read();
    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(curp, v, retval)
    127 	struct proc *curp;
    128 	void *v;
    129 	register_t *retval;
    130 {
    131 	register struct sys_setpriority_args /* {
    132 		syscallarg(int) which;
    133 		syscallarg(int) who;
    134 		syscallarg(int) prio;
    135 	} */ *uap = v;
    136 	register struct proc *p;
    137 	int found = 0, error = 0;
    138 
    139 	switch (SCARG(uap, which)) {
    140 
    141 	case PRIO_PROCESS:
    142 		if (SCARG(uap, who) == 0)
    143 			p = curp;
    144 		else
    145 			p = pfind(SCARG(uap, who));
    146 		if (p == 0)
    147 			break;
    148 		error = donice(curp, p, SCARG(uap, prio));
    149 		found++;
    150 		break;
    151 
    152 	case PRIO_PGRP: {
    153 		register struct pgrp *pg;
    154 
    155 		if (SCARG(uap, who) == 0)
    156 			pg = curp->p_pgrp;
    157 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
    158 			break;
    159 		for (p = pg->pg_members.lh_first; p != 0;
    160 		    p = p->p_pglist.le_next) {
    161 			error = donice(curp, p, SCARG(uap, prio));
    162 			found++;
    163 		}
    164 		break;
    165 	}
    166 
    167 	case PRIO_USER:
    168 		if (SCARG(uap, who) == 0)
    169 			SCARG(uap, who) = curp->p_ucred->cr_uid;
    170 		proclist_lock_read();
    171 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
    172 			if (p->p_ucred->cr_uid == SCARG(uap, who)) {
    173 				error = donice(curp, p, SCARG(uap, prio));
    174 				found++;
    175 			}
    176 		proclist_unlock_read();
    177 		break;
    178 
    179 	default:
    180 		return (EINVAL);
    181 	}
    182 	if (found == 0)
    183 		return (ESRCH);
    184 	return (error);
    185 }
    186 
    187 int
    188 donice(curp, chgp, n)
    189 	register struct proc *curp, *chgp;
    190 	register int n;
    191 {
    192 	register struct pcred *pcred = curp->p_cred;
    193 
    194 	if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
    195 	    pcred->pc_ucred->cr_uid != chgp->p_ucred->cr_uid &&
    196 	    pcred->p_ruid != chgp->p_ucred->cr_uid)
    197 		return (EPERM);
    198 	if (n > PRIO_MAX)
    199 		n = PRIO_MAX;
    200 	if (n < PRIO_MIN)
    201 		n = PRIO_MIN;
    202 	n += NZERO;
    203 	if (n < chgp->p_nice && suser(pcred->pc_ucred, &curp->p_acflag))
    204 		return (EACCES);
    205 	chgp->p_nice = n;
    206 	(void)resetpriority(chgp);
    207 	return (0);
    208 }
    209 
    210 /* ARGSUSED */
    211 int
    212 sys_setrlimit(p, v, retval)
    213 	struct proc *p;
    214 	void *v;
    215 	register_t *retval;
    216 {
    217 	register struct sys_setrlimit_args /* {
    218 		syscallarg(int) which;
    219 		syscallarg(const struct rlimit *) rlp;
    220 	} */ *uap = v;
    221 	int which = SCARG(uap, which);
    222 	struct rlimit alim;
    223 	int error;
    224 
    225 	error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit));
    226 	if (error)
    227 		return (error);
    228 	return (dosetrlimit(p, which, &alim));
    229 }
    230 
    231 int
    232 dosetrlimit(p, which, limp)
    233 	struct proc *p;
    234 	int which;
    235 	struct rlimit *limp;
    236 {
    237 	register struct rlimit *alimp;
    238 	extern unsigned maxdmap, maxsmap;
    239 	int error;
    240 
    241 	if ((u_int)which >= RLIM_NLIMITS)
    242 		return (EINVAL);
    243 
    244 	if (limp->rlim_cur < 0 || limp->rlim_max < 0)
    245 		return (EINVAL);
    246 
    247 	alimp = &p->p_rlimit[which];
    248 	if (limp->rlim_cur > alimp->rlim_max ||
    249 	    limp->rlim_max > alimp->rlim_max)
    250 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
    251 			return (error);
    252 	if (limp->rlim_cur > limp->rlim_max)
    253 		limp->rlim_cur = limp->rlim_max;
    254 	if (p->p_limit->p_refcnt > 1 &&
    255 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
    256 		p->p_limit->p_refcnt--;
    257 		p->p_limit = limcopy(p->p_limit);
    258 		alimp = &p->p_rlimit[which];
    259 	}
    260 
    261 	switch (which) {
    262 
    263 	case RLIMIT_DATA:
    264 		if (limp->rlim_cur > maxdmap)
    265 			limp->rlim_cur = maxdmap;
    266 		if (limp->rlim_max > maxdmap)
    267 			limp->rlim_max = maxdmap;
    268 		break;
    269 
    270 	case RLIMIT_STACK:
    271 		if (limp->rlim_cur > maxsmap)
    272 			limp->rlim_cur = maxsmap;
    273 		if (limp->rlim_max > maxsmap)
    274 			limp->rlim_max = maxsmap;
    275 
    276 		/*
    277 		 * Stack is allocated to the max at exec time with
    278 		 * only "rlim_cur" bytes accessible (In other words,
    279 		 * allocates stack dividing two contiguous regions at
    280 		 * "rlim_cur" bytes boundary).
    281 		 *
    282 		 * Since allocation is done in terms of page, roundup
    283 		 * "rlim_cur" (otherwise, contiguous regions
    284 		 * overlap).  If stack limit is going up make more
    285 		 * accessible, if going down make inaccessible.
    286 		 */
    287 		limp->rlim_cur = round_page(limp->rlim_cur);
    288 		if (limp->rlim_cur != alimp->rlim_cur) {
    289 			vaddr_t addr;
    290 			vsize_t size;
    291 			vm_prot_t prot;
    292 
    293 			if (limp->rlim_cur > alimp->rlim_cur) {
    294 				prot = VM_PROT_ALL;
    295 				size = limp->rlim_cur - alimp->rlim_cur;
    296 				addr = USRSTACK - limp->rlim_cur;
    297 			} else {
    298 				prot = VM_PROT_NONE;
    299 				size = alimp->rlim_cur - limp->rlim_cur;
    300 				addr = USRSTACK - alimp->rlim_cur;
    301 			}
    302 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
    303 					      addr, addr+size, prot, FALSE);
    304 		}
    305 		break;
    306 
    307 	case RLIMIT_NOFILE:
    308 		if (limp->rlim_cur > maxfiles)
    309 			limp->rlim_cur = maxfiles;
    310 		if (limp->rlim_max > maxfiles)
    311 			limp->rlim_max = maxfiles;
    312 		break;
    313 
    314 	case RLIMIT_NPROC:
    315 		if (limp->rlim_cur > maxproc)
    316 			limp->rlim_cur = maxproc;
    317 		if (limp->rlim_max > maxproc)
    318 			limp->rlim_max = maxproc;
    319 		break;
    320 	}
    321 	*alimp = *limp;
    322 	return (0);
    323 }
    324 
    325 /* ARGSUSED */
    326 int
    327 sys_getrlimit(p, v, retval)
    328 	struct proc *p;
    329 	void *v;
    330 	register_t *retval;
    331 {
    332 	register struct sys_getrlimit_args /* {
    333 		syscallarg(int) which;
    334 		syscallarg(struct rlimit *) rlp;
    335 	} */ *uap = v;
    336 	int which = SCARG(uap, which);
    337 
    338 	if ((u_int)which >= RLIM_NLIMITS)
    339 		return (EINVAL);
    340 	return (copyout(&p->p_rlimit[which], SCARG(uap, rlp),
    341 	    sizeof(struct rlimit)));
    342 }
    343 
    344 /*
    345  * Transform the running time and tick information in proc p into user,
    346  * system, and interrupt time usage.
    347  */
    348 void
    349 calcru(p, up, sp, ip)
    350 	register struct proc *p;
    351 	register struct timeval *up;
    352 	register struct timeval *sp;
    353 	register struct timeval *ip;
    354 {
    355 	register u_quad_t u, st, ut, it, tot;
    356 	register long sec, usec;
    357 	register int s;
    358 	struct timeval tv;
    359 
    360 	s = splstatclock();
    361 	st = p->p_sticks;
    362 	ut = p->p_uticks;
    363 	it = p->p_iticks;
    364 	splx(s);
    365 
    366 	tot = st + ut + it;
    367 	if (tot == 0) {
    368 		up->tv_sec = up->tv_usec = 0;
    369 		sp->tv_sec = sp->tv_usec = 0;
    370 		if (ip != NULL)
    371 			ip->tv_sec = ip->tv_usec = 0;
    372 		return;
    373 	}
    374 
    375 	sec = p->p_rtime.tv_sec;
    376 	usec = p->p_rtime.tv_usec;
    377 	if (p == curproc) {
    378 		/*
    379 		 * Adjust for the current time slice.  This is actually fairly
    380 		 * important since the error here is on the order of a time
    381 		 * quantum, which is much greater than the sampling error.
    382 		 */
    383 		microtime(&tv);
    384 		sec += tv.tv_sec - runtime.tv_sec;
    385 		usec += tv.tv_usec - runtime.tv_usec;
    386 	}
    387 	u = (u_quad_t) sec * 1000000 + usec;
    388 	st = (u * st) / tot;
    389 	sp->tv_sec = st / 1000000;
    390 	sp->tv_usec = st % 1000000;
    391 	ut = (u * ut) / tot;
    392 	up->tv_sec = ut / 1000000;
    393 	up->tv_usec = ut % 1000000;
    394 	if (ip != NULL) {
    395 		it = (u * it) / tot;
    396 		ip->tv_sec = it / 1000000;
    397 		ip->tv_usec = it % 1000000;
    398 	}
    399 }
    400 
    401 /* ARGSUSED */
    402 int
    403 sys_getrusage(p, v, retval)
    404 	register struct proc *p;
    405 	void *v;
    406 	register_t *retval;
    407 {
    408 	register struct sys_getrusage_args /* {
    409 		syscallarg(int) who;
    410 		syscallarg(struct rusage *) rusage;
    411 	} */ *uap = v;
    412 	register struct rusage *rup;
    413 
    414 	switch (SCARG(uap, who)) {
    415 
    416 	case RUSAGE_SELF:
    417 		rup = &p->p_stats->p_ru;
    418 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
    419 		break;
    420 
    421 	case RUSAGE_CHILDREN:
    422 		rup = &p->p_stats->p_cru;
    423 		break;
    424 
    425 	default:
    426 		return (EINVAL);
    427 	}
    428 	return (copyout(rup, SCARG(uap, rusage), sizeof(struct rusage)));
    429 }
    430 
    431 void
    432 ruadd(ru, ru2)
    433 	register struct rusage *ru, *ru2;
    434 {
    435 	register long *ip, *ip2;
    436 	register int i;
    437 
    438 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
    439 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
    440 	if (ru->ru_maxrss < ru2->ru_maxrss)
    441 		ru->ru_maxrss = ru2->ru_maxrss;
    442 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
    443 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
    444 		*ip++ += *ip2++;
    445 }
    446 
    447 /*
    448  * Make a copy of the plimit structure.
    449  * We share these structures copy-on-write after fork,
    450  * and copy when a limit is changed.
    451  */
    452 struct plimit *
    453 limcopy(lim)
    454 	struct plimit *lim;
    455 {
    456 	register struct plimit *newlim;
    457 
    458 	newlim = pool_get(&plimit_pool, PR_WAITOK);
    459 	memcpy(newlim->pl_rlimit, lim->pl_rlimit,
    460 	    sizeof(struct rlimit) * RLIM_NLIMITS);
    461 	newlim->p_lflags = 0;
    462 	newlim->p_refcnt = 1;
    463 	return (newlim);
    464 }
    465 
    466 void
    467 limfree(lim)
    468 	struct plimit *lim;
    469 {
    470 
    471 	if (--lim->p_refcnt > 0)
    472 		return;
    473 	pool_put(&plimit_pool, lim);
    474 }
    475