Home | History | Annotate | Line # | Download | only in kern
kern_resource.c revision 1.53.2.2
      1 /*	$NetBSD: kern_resource.c,v 1.53.2.2 2001/02/11 19:16:47 bouyer 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 <uvm/uvm_extern.h>
     56 
     57 /*
     58  * Maximum process data and stack limits.
     59  * They are variables so they are patchable.
     60  *
     61  * XXXX Do we really need them to be patchable?
     62  */
     63 rlim_t maxdmap = MAXDSIZ;
     64 rlim_t maxsmap = MAXSSIZ;
     65 
     66 /*
     67  * Resource controls and accounting.
     68  */
     69 
     70 int
     71 sys_getpriority(curp, v, retval)
     72 	struct proc *curp;
     73 	void *v;
     74 	register_t *retval;
     75 {
     76 	struct sys_getpriority_args /* {
     77 		syscallarg(int) which;
     78 		syscallarg(int) who;
     79 	} */ *uap = v;
     80 	struct proc *p;
     81 	int low = NZERO + PRIO_MAX + 1;
     82 
     83 	switch (SCARG(uap, which)) {
     84 
     85 	case PRIO_PROCESS:
     86 		if (SCARG(uap, who) == 0)
     87 			p = curp;
     88 		else
     89 			p = pfind(SCARG(uap, who));
     90 		if (p == 0)
     91 			break;
     92 		low = p->p_nice;
     93 		break;
     94 
     95 	case PRIO_PGRP: {
     96 		struct pgrp *pg;
     97 
     98 		if (SCARG(uap, who) == 0)
     99 			pg = curp->p_pgrp;
    100 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
    101 			break;
    102 		for (p = pg->pg_members.lh_first; p != 0;
    103 		     p = p->p_pglist.le_next) {
    104 			if (p->p_nice < low)
    105 				low = p->p_nice;
    106 		}
    107 		break;
    108 	}
    109 
    110 	case PRIO_USER:
    111 		if (SCARG(uap, who) == 0)
    112 			SCARG(uap, who) = curp->p_ucred->cr_uid;
    113 		proclist_lock_read();
    114 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
    115 			if (p->p_ucred->cr_uid == SCARG(uap, who) &&
    116 			    p->p_nice < low)
    117 				low = p->p_nice;
    118 		proclist_unlock_read();
    119 		break;
    120 
    121 	default:
    122 		return (EINVAL);
    123 	}
    124 	if (low == NZERO + PRIO_MAX + 1)
    125 		return (ESRCH);
    126 	*retval = low - NZERO;
    127 	return (0);
    128 }
    129 
    130 /* ARGSUSED */
    131 int
    132 sys_setpriority(curp, v, retval)
    133 	struct proc *curp;
    134 	void *v;
    135 	register_t *retval;
    136 {
    137 	struct sys_setpriority_args /* {
    138 		syscallarg(int) which;
    139 		syscallarg(int) who;
    140 		syscallarg(int) prio;
    141 	} */ *uap = v;
    142 	struct proc *p;
    143 	int found = 0, error = 0;
    144 
    145 	switch (SCARG(uap, which)) {
    146 
    147 	case PRIO_PROCESS:
    148 		if (SCARG(uap, who) == 0)
    149 			p = curp;
    150 		else
    151 			p = pfind(SCARG(uap, who));
    152 		if (p == 0)
    153 			break;
    154 		error = donice(curp, p, SCARG(uap, prio));
    155 		found++;
    156 		break;
    157 
    158 	case PRIO_PGRP: {
    159 		struct pgrp *pg;
    160 
    161 		if (SCARG(uap, who) == 0)
    162 			pg = curp->p_pgrp;
    163 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
    164 			break;
    165 		for (p = pg->pg_members.lh_first; p != 0;
    166 		    p = p->p_pglist.le_next) {
    167 			error = donice(curp, p, SCARG(uap, prio));
    168 			found++;
    169 		}
    170 		break;
    171 	}
    172 
    173 	case PRIO_USER:
    174 		if (SCARG(uap, who) == 0)
    175 			SCARG(uap, who) = curp->p_ucred->cr_uid;
    176 		proclist_lock_read();
    177 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
    178 			if (p->p_ucred->cr_uid == SCARG(uap, who)) {
    179 				error = donice(curp, p, SCARG(uap, prio));
    180 				found++;
    181 			}
    182 		proclist_unlock_read();
    183 		break;
    184 
    185 	default:
    186 		return (EINVAL);
    187 	}
    188 	if (found == 0)
    189 		return (ESRCH);
    190 	return (error);
    191 }
    192 
    193 int
    194 donice(curp, chgp, n)
    195 	struct proc *curp, *chgp;
    196 	int n;
    197 {
    198 	struct pcred *pcred = curp->p_cred;
    199 	int s;
    200 
    201 	if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
    202 	    pcred->pc_ucred->cr_uid != chgp->p_ucred->cr_uid &&
    203 	    pcred->p_ruid != chgp->p_ucred->cr_uid)
    204 		return (EPERM);
    205 	if (n > PRIO_MAX)
    206 		n = PRIO_MAX;
    207 	if (n < PRIO_MIN)
    208 		n = PRIO_MIN;
    209 	n += NZERO;
    210 	if (n < chgp->p_nice && suser(pcred->pc_ucred, &curp->p_acflag))
    211 		return (EACCES);
    212 	chgp->p_nice = n;
    213 	SCHED_LOCK(s);
    214 	(void)resetpriority(chgp);
    215 	SCHED_UNLOCK(s);
    216 	return (0);
    217 }
    218 
    219 /* ARGSUSED */
    220 int
    221 sys_setrlimit(p, v, retval)
    222 	struct proc *p;
    223 	void *v;
    224 	register_t *retval;
    225 {
    226 	struct sys_setrlimit_args /* {
    227 		syscallarg(int) which;
    228 		syscallarg(const struct rlimit *) rlp;
    229 	} */ *uap = v;
    230 	int which = SCARG(uap, which);
    231 	struct rlimit alim;
    232 	int error;
    233 
    234 	error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit));
    235 	if (error)
    236 		return (error);
    237 	return (dosetrlimit(p, p->p_cred, which, &alim));
    238 }
    239 
    240 int
    241 dosetrlimit(p, cred, which, limp)
    242 	struct proc *p;
    243 	struct  pcred *cred;
    244 	int which;
    245 	struct rlimit *limp;
    246 {
    247 	struct rlimit *alimp;
    248 	struct plimit *newplim;
    249 	int error;
    250 
    251 	if ((u_int)which >= RLIM_NLIMITS)
    252 		return (EINVAL);
    253 
    254 	if (limp->rlim_cur < 0 || limp->rlim_max < 0)
    255 		return (EINVAL);
    256 
    257 	alimp = &p->p_rlimit[which];
    258 	/* if we don't change the value, no need to limcopy() */
    259 	if (limp->rlim_cur == alimp->rlim_cur &&
    260 	    limp->rlim_max == alimp->rlim_max)
    261 		return 0;
    262 
    263 	if (limp->rlim_cur > alimp->rlim_max ||
    264 	    limp->rlim_max > alimp->rlim_max)
    265 		if ((error = suser(cred->pc_ucred, &p->p_acflag)) != 0)
    266 			return (error);
    267 	if (limp->rlim_cur > limp->rlim_max)
    268 		limp->rlim_cur = limp->rlim_max;
    269 	if (p->p_limit->p_refcnt > 1 &&
    270 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
    271 		newplim = limcopy(p->p_limit);
    272 		limfree(p->p_limit);
    273 		p->p_limit = newplim;
    274 		alimp = &p->p_rlimit[which];
    275 	}
    276 
    277 	switch (which) {
    278 
    279 	case RLIMIT_DATA:
    280 		if (limp->rlim_cur > maxdmap)
    281 			limp->rlim_cur = maxdmap;
    282 		if (limp->rlim_max > maxdmap)
    283 			limp->rlim_max = maxdmap;
    284 		break;
    285 
    286 	case RLIMIT_STACK:
    287 		if (limp->rlim_cur > maxsmap)
    288 			limp->rlim_cur = maxsmap;
    289 		if (limp->rlim_max > maxsmap)
    290 			limp->rlim_max = maxsmap;
    291 
    292 		/*
    293 		 * Stack is allocated to the max at exec time with
    294 		 * only "rlim_cur" bytes accessible (In other words,
    295 		 * allocates stack dividing two contiguous regions at
    296 		 * "rlim_cur" bytes boundary).
    297 		 *
    298 		 * Since allocation is done in terms of page, roundup
    299 		 * "rlim_cur" (otherwise, contiguous regions
    300 		 * overlap).  If stack limit is going up make more
    301 		 * accessible, if going down make inaccessible.
    302 		 */
    303 		limp->rlim_cur = round_page(limp->rlim_cur);
    304 		if (limp->rlim_cur != alimp->rlim_cur) {
    305 			vaddr_t addr;
    306 			vsize_t size;
    307 			vm_prot_t prot;
    308 
    309 			if (limp->rlim_cur > alimp->rlim_cur) {
    310 				prot = VM_PROT_ALL;
    311 				size = limp->rlim_cur - alimp->rlim_cur;
    312 				addr = USRSTACK - limp->rlim_cur;
    313 			} else {
    314 				prot = VM_PROT_NONE;
    315 				size = alimp->rlim_cur - limp->rlim_cur;
    316 				addr = USRSTACK - alimp->rlim_cur;
    317 			}
    318 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
    319 					      addr, addr+size, prot, FALSE);
    320 		}
    321 		break;
    322 
    323 	case RLIMIT_NOFILE:
    324 		if (limp->rlim_cur > maxfiles)
    325 			limp->rlim_cur = maxfiles;
    326 		if (limp->rlim_max > maxfiles)
    327 			limp->rlim_max = maxfiles;
    328 		break;
    329 
    330 	case RLIMIT_NPROC:
    331 		if (limp->rlim_cur > maxproc)
    332 			limp->rlim_cur = maxproc;
    333 		if (limp->rlim_max > maxproc)
    334 			limp->rlim_max = maxproc;
    335 		break;
    336 	}
    337 	*alimp = *limp;
    338 	return (0);
    339 }
    340 
    341 /* ARGSUSED */
    342 int
    343 sys_getrlimit(p, v, retval)
    344 	struct proc *p;
    345 	void *v;
    346 	register_t *retval;
    347 {
    348 	struct sys_getrlimit_args /* {
    349 		syscallarg(int) which;
    350 		syscallarg(struct rlimit *) rlp;
    351 	} */ *uap = v;
    352 	int which = SCARG(uap, which);
    353 
    354 	if ((u_int)which >= RLIM_NLIMITS)
    355 		return (EINVAL);
    356 	return (copyout(&p->p_rlimit[which], SCARG(uap, rlp),
    357 	    sizeof(struct rlimit)));
    358 }
    359 
    360 /*
    361  * Transform the running time and tick information in proc p into user,
    362  * system, and interrupt time usage.
    363  */
    364 void
    365 calcru(p, up, sp, ip)
    366 	struct proc *p;
    367 	struct timeval *up;
    368 	struct timeval *sp;
    369 	struct timeval *ip;
    370 {
    371 	u_quad_t u, st, ut, it, tot;
    372 	long sec, usec;
    373 	int s;
    374 	struct timeval tv;
    375 
    376 	s = splstatclock();
    377 	st = p->p_sticks;
    378 	ut = p->p_uticks;
    379 	it = p->p_iticks;
    380 	splx(s);
    381 
    382 	tot = st + ut + it;
    383 	if (tot == 0) {
    384 		up->tv_sec = up->tv_usec = 0;
    385 		sp->tv_sec = sp->tv_usec = 0;
    386 		if (ip != NULL)
    387 			ip->tv_sec = ip->tv_usec = 0;
    388 		return;
    389 	}
    390 
    391 	sec = p->p_rtime.tv_sec;
    392 	usec = p->p_rtime.tv_usec;
    393 	if (p->p_stat == SONPROC) {
    394 		struct schedstate_percpu *spc;
    395 
    396 		KDASSERT(p->p_cpu != NULL);
    397 		spc = &p->p_cpu->ci_schedstate;
    398 
    399 		/*
    400 		 * Adjust for the current time slice.  This is actually fairly
    401 		 * important since the error here is on the order of a time
    402 		 * quantum, which is much greater than the sampling error.
    403 		 */
    404 		microtime(&tv);
    405 		sec += tv.tv_sec - spc->spc_runtime.tv_sec;
    406 		usec += tv.tv_usec - spc->spc_runtime.tv_usec;
    407 	}
    408 	u = (u_quad_t) sec * 1000000 + usec;
    409 	st = (u * st) / tot;
    410 	sp->tv_sec = st / 1000000;
    411 	sp->tv_usec = st % 1000000;
    412 	ut = (u * ut) / tot;
    413 	up->tv_sec = ut / 1000000;
    414 	up->tv_usec = ut % 1000000;
    415 	if (ip != NULL) {
    416 		it = (u * it) / tot;
    417 		ip->tv_sec = it / 1000000;
    418 		ip->tv_usec = it % 1000000;
    419 	}
    420 }
    421 
    422 /* ARGSUSED */
    423 int
    424 sys_getrusage(p, v, retval)
    425 	struct proc *p;
    426 	void *v;
    427 	register_t *retval;
    428 {
    429 	struct sys_getrusage_args /* {
    430 		syscallarg(int) who;
    431 		syscallarg(struct rusage *) rusage;
    432 	} */ *uap = v;
    433 	struct rusage *rup;
    434 
    435 	switch (SCARG(uap, who)) {
    436 
    437 	case RUSAGE_SELF:
    438 		rup = &p->p_stats->p_ru;
    439 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
    440 		break;
    441 
    442 	case RUSAGE_CHILDREN:
    443 		rup = &p->p_stats->p_cru;
    444 		break;
    445 
    446 	default:
    447 		return (EINVAL);
    448 	}
    449 	return (copyout(rup, SCARG(uap, rusage), sizeof(struct rusage)));
    450 }
    451 
    452 void
    453 ruadd(ru, ru2)
    454 	struct rusage *ru, *ru2;
    455 {
    456 	long *ip, *ip2;
    457 	int i;
    458 
    459 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
    460 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
    461 	if (ru->ru_maxrss < ru2->ru_maxrss)
    462 		ru->ru_maxrss = ru2->ru_maxrss;
    463 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
    464 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
    465 		*ip++ += *ip2++;
    466 }
    467 
    468 /*
    469  * Make a copy of the plimit structure.
    470  * We share these structures copy-on-write after fork,
    471  * and copy when a limit is changed.
    472  */
    473 struct plimit *
    474 limcopy(lim)
    475 	struct plimit *lim;
    476 {
    477 	struct plimit *newlim;
    478 
    479 	newlim = pool_get(&plimit_pool, PR_WAITOK);
    480 	memcpy(newlim->pl_rlimit, lim->pl_rlimit,
    481 	    sizeof(struct rlimit) * RLIM_NLIMITS);
    482 	if (lim->pl_corename == defcorename) {
    483 		newlim->pl_corename = defcorename;
    484 	} else {
    485 		newlim->pl_corename = malloc(strlen(lim->pl_corename)+1,
    486 		    M_TEMP, M_WAITOK);
    487 		strcpy(newlim->pl_corename, lim->pl_corename);
    488 	}
    489 	newlim->p_lflags = 0;
    490 	newlim->p_refcnt = 1;
    491 	return (newlim);
    492 }
    493 
    494 void
    495 limfree(lim)
    496 	struct plimit *lim;
    497 {
    498 
    499 	if (--lim->p_refcnt > 0)
    500 		return;
    501 #ifdef DIAGNOSTIC
    502 	if (lim->p_refcnt < 0)
    503 		panic("limfree");
    504 #endif
    505 	if (lim->pl_corename != defcorename)
    506 		free(lim->pl_corename, M_TEMP);
    507 	pool_put(&plimit_pool, lim);
    508 }
    509