Home | History | Annotate | Line # | Download | only in kern
kern_resource.c revision 1.58
      1 /*	$NetBSD: kern_resource.c,v 1.58 2000/06/27 17:41:25 mrg 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  * Resource controls and accounting.
     59  */
     60 
     61 int
     62 sys_getpriority(curp, v, retval)
     63 	struct proc *curp;
     64 	void *v;
     65 	register_t *retval;
     66 {
     67 	struct sys_getpriority_args /* {
     68 		syscallarg(int) which;
     69 		syscallarg(int) who;
     70 	} */ *uap = v;
     71 	struct proc *p;
     72 	int low = NZERO + PRIO_MAX + 1;
     73 
     74 	switch (SCARG(uap, which)) {
     75 
     76 	case PRIO_PROCESS:
     77 		if (SCARG(uap, who) == 0)
     78 			p = curp;
     79 		else
     80 			p = pfind(SCARG(uap, who));
     81 		if (p == 0)
     82 			break;
     83 		low = p->p_nice;
     84 		break;
     85 
     86 	case PRIO_PGRP: {
     87 		struct pgrp *pg;
     88 
     89 		if (SCARG(uap, who) == 0)
     90 			pg = curp->p_pgrp;
     91 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
     92 			break;
     93 		for (p = pg->pg_members.lh_first; p != 0;
     94 		     p = p->p_pglist.le_next) {
     95 			if (p->p_nice < low)
     96 				low = p->p_nice;
     97 		}
     98 		break;
     99 	}
    100 
    101 	case PRIO_USER:
    102 		if (SCARG(uap, who) == 0)
    103 			SCARG(uap, who) = curp->p_ucred->cr_uid;
    104 		proclist_lock_read();
    105 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
    106 			if (p->p_ucred->cr_uid == SCARG(uap, who) &&
    107 			    p->p_nice < low)
    108 				low = p->p_nice;
    109 		proclist_unlock_read();
    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(curp, v, retval)
    124 	struct proc *curp;
    125 	void *v;
    126 	register_t *retval;
    127 {
    128 	struct sys_setpriority_args /* {
    129 		syscallarg(int) which;
    130 		syscallarg(int) who;
    131 		syscallarg(int) prio;
    132 	} */ *uap = v;
    133 	struct proc *p;
    134 	int found = 0, error = 0;
    135 
    136 	switch (SCARG(uap, which)) {
    137 
    138 	case PRIO_PROCESS:
    139 		if (SCARG(uap, who) == 0)
    140 			p = curp;
    141 		else
    142 			p = pfind(SCARG(uap, who));
    143 		if (p == 0)
    144 			break;
    145 		error = donice(curp, p, 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_pgrp;
    154 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
    155 			break;
    156 		for (p = pg->pg_members.lh_first; p != 0;
    157 		    p = p->p_pglist.le_next) {
    158 			error = donice(curp, p, SCARG(uap, prio));
    159 			found++;
    160 		}
    161 		break;
    162 	}
    163 
    164 	case PRIO_USER:
    165 		if (SCARG(uap, who) == 0)
    166 			SCARG(uap, who) = curp->p_ucred->cr_uid;
    167 		proclist_lock_read();
    168 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
    169 			if (p->p_ucred->cr_uid == SCARG(uap, who)) {
    170 				error = donice(curp, p, SCARG(uap, prio));
    171 				found++;
    172 			}
    173 		proclist_unlock_read();
    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 	struct proc *curp, *chgp;
    187 	int n;
    188 {
    189 	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 	struct sys_setrlimit_args /* {
    215 		syscallarg(int) which;
    216 		syscallarg(const struct rlimit *) rlp;
    217 	} */ *uap = v;
    218 	int which = SCARG(uap, which);
    219 	struct rlimit alim;
    220 	int error;
    221 
    222 	error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit));
    223 	if (error)
    224 		return (error);
    225 	return (dosetrlimit(p, p->p_cred, which, &alim));
    226 }
    227 
    228 int
    229 dosetrlimit(p, cred, which, limp)
    230 	struct proc *p;
    231 	struct  pcred *cred;
    232 	int which;
    233 	struct rlimit *limp;
    234 {
    235 	struct rlimit *alimp;
    236 	extern unsigned maxdmap, maxsmap;
    237 	struct plimit *newplim;
    238 	int error;
    239 
    240 	if ((u_int)which >= RLIM_NLIMITS)
    241 		return (EINVAL);
    242 
    243 	if (limp->rlim_cur < 0 || limp->rlim_max < 0)
    244 		return (EINVAL);
    245 
    246 	alimp = &p->p_rlimit[which];
    247 	/* if we don't change the value, no need to limcopy() */
    248 	if (limp->rlim_cur == alimp->rlim_cur &&
    249 	    limp->rlim_max == alimp->rlim_max)
    250 		return 0;
    251 
    252 	if (limp->rlim_cur > alimp->rlim_max ||
    253 	    limp->rlim_max > alimp->rlim_max)
    254 		if ((error = suser(cred->pc_ucred, &p->p_acflag)) != 0)
    255 			return (error);
    256 	if (limp->rlim_cur > limp->rlim_max)
    257 		limp->rlim_cur = limp->rlim_max;
    258 	if (p->p_limit->p_refcnt > 1 &&
    259 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
    260 		newplim = limcopy(p->p_limit);
    261 		limfree(p->p_limit);
    262 		p->p_limit = newplim;
    263 		alimp = &p->p_rlimit[which];
    264 	}
    265 
    266 	switch (which) {
    267 
    268 	case RLIMIT_DATA:
    269 		if (limp->rlim_cur > maxdmap)
    270 			limp->rlim_cur = maxdmap;
    271 		if (limp->rlim_max > maxdmap)
    272 			limp->rlim_max = maxdmap;
    273 		break;
    274 
    275 	case RLIMIT_STACK:
    276 		if (limp->rlim_cur > maxsmap)
    277 			limp->rlim_cur = maxsmap;
    278 		if (limp->rlim_max > maxsmap)
    279 			limp->rlim_max = maxsmap;
    280 
    281 		/*
    282 		 * Stack is allocated to the max at exec time with
    283 		 * only "rlim_cur" bytes accessible (In other words,
    284 		 * allocates stack dividing two contiguous regions at
    285 		 * "rlim_cur" bytes boundary).
    286 		 *
    287 		 * Since allocation is done in terms of page, roundup
    288 		 * "rlim_cur" (otherwise, contiguous regions
    289 		 * overlap).  If stack limit is going up make more
    290 		 * accessible, if going down make inaccessible.
    291 		 */
    292 		limp->rlim_cur = round_page(limp->rlim_cur);
    293 		if (limp->rlim_cur != alimp->rlim_cur) {
    294 			vaddr_t addr;
    295 			vsize_t size;
    296 			vm_prot_t prot;
    297 
    298 			if (limp->rlim_cur > alimp->rlim_cur) {
    299 				prot = VM_PROT_ALL;
    300 				size = limp->rlim_cur - alimp->rlim_cur;
    301 				addr = USRSTACK - limp->rlim_cur;
    302 			} else {
    303 				prot = VM_PROT_NONE;
    304 				size = alimp->rlim_cur - limp->rlim_cur;
    305 				addr = USRSTACK - alimp->rlim_cur;
    306 			}
    307 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
    308 					      addr, addr+size, prot, FALSE);
    309 		}
    310 		break;
    311 
    312 	case RLIMIT_NOFILE:
    313 		if (limp->rlim_cur > maxfiles)
    314 			limp->rlim_cur = maxfiles;
    315 		if (limp->rlim_max > maxfiles)
    316 			limp->rlim_max = maxfiles;
    317 		break;
    318 
    319 	case RLIMIT_NPROC:
    320 		if (limp->rlim_cur > maxproc)
    321 			limp->rlim_cur = maxproc;
    322 		if (limp->rlim_max > maxproc)
    323 			limp->rlim_max = maxproc;
    324 		break;
    325 	}
    326 	*alimp = *limp;
    327 	return (0);
    328 }
    329 
    330 /* ARGSUSED */
    331 int
    332 sys_getrlimit(p, v, retval)
    333 	struct proc *p;
    334 	void *v;
    335 	register_t *retval;
    336 {
    337 	struct sys_getrlimit_args /* {
    338 		syscallarg(int) which;
    339 		syscallarg(struct rlimit *) rlp;
    340 	} */ *uap = v;
    341 	int which = SCARG(uap, which);
    342 
    343 	if ((u_int)which >= RLIM_NLIMITS)
    344 		return (EINVAL);
    345 	return (copyout(&p->p_rlimit[which], SCARG(uap, rlp),
    346 	    sizeof(struct rlimit)));
    347 }
    348 
    349 /*
    350  * Transform the running time and tick information in proc p into user,
    351  * system, and interrupt time usage.
    352  */
    353 void
    354 calcru(p, up, sp, ip)
    355 	struct proc *p;
    356 	struct timeval *up;
    357 	struct timeval *sp;
    358 	struct timeval *ip;
    359 {
    360 	u_quad_t u, st, ut, it, tot;
    361 	long sec, usec;
    362 	int s;
    363 	struct timeval tv;
    364 
    365 	s = splstatclock();
    366 	st = p->p_sticks;
    367 	ut = p->p_uticks;
    368 	it = p->p_iticks;
    369 	splx(s);
    370 
    371 	tot = st + ut + it;
    372 	if (tot == 0) {
    373 		up->tv_sec = up->tv_usec = 0;
    374 		sp->tv_sec = sp->tv_usec = 0;
    375 		if (ip != NULL)
    376 			ip->tv_sec = ip->tv_usec = 0;
    377 		return;
    378 	}
    379 
    380 	sec = p->p_rtime.tv_sec;
    381 	usec = p->p_rtime.tv_usec;
    382 	if (p->p_stat == SONPROC) {
    383 		struct schedstate_percpu *spc;
    384 
    385 		KDASSERT(p->p_cpu != NULL);
    386 		spc = &p->p_cpu->ci_schedstate;
    387 
    388 		/*
    389 		 * Adjust for the current time slice.  This is actually fairly
    390 		 * important since the error here is on the order of a time
    391 		 * quantum, which is much greater than the sampling error.
    392 		 */
    393 		microtime(&tv);
    394 		sec += tv.tv_sec - spc->spc_runtime.tv_sec;
    395 		usec += tv.tv_usec - spc->spc_runtime.tv_usec;
    396 	}
    397 	u = (u_quad_t) sec * 1000000 + usec;
    398 	st = (u * st) / tot;
    399 	sp->tv_sec = st / 1000000;
    400 	sp->tv_usec = st % 1000000;
    401 	ut = (u * ut) / tot;
    402 	up->tv_sec = ut / 1000000;
    403 	up->tv_usec = ut % 1000000;
    404 	if (ip != NULL) {
    405 		it = (u * it) / tot;
    406 		ip->tv_sec = it / 1000000;
    407 		ip->tv_usec = it % 1000000;
    408 	}
    409 }
    410 
    411 /* ARGSUSED */
    412 int
    413 sys_getrusage(p, v, retval)
    414 	struct proc *p;
    415 	void *v;
    416 	register_t *retval;
    417 {
    418 	struct sys_getrusage_args /* {
    419 		syscallarg(int) who;
    420 		syscallarg(struct rusage *) rusage;
    421 	} */ *uap = v;
    422 	struct rusage *rup;
    423 
    424 	switch (SCARG(uap, who)) {
    425 
    426 	case RUSAGE_SELF:
    427 		rup = &p->p_stats->p_ru;
    428 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
    429 		break;
    430 
    431 	case RUSAGE_CHILDREN:
    432 		rup = &p->p_stats->p_cru;
    433 		break;
    434 
    435 	default:
    436 		return (EINVAL);
    437 	}
    438 	return (copyout(rup, SCARG(uap, rusage), sizeof(struct rusage)));
    439 }
    440 
    441 void
    442 ruadd(ru, ru2)
    443 	struct rusage *ru, *ru2;
    444 {
    445 	long *ip, *ip2;
    446 	int i;
    447 
    448 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
    449 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
    450 	if (ru->ru_maxrss < ru2->ru_maxrss)
    451 		ru->ru_maxrss = ru2->ru_maxrss;
    452 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
    453 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
    454 		*ip++ += *ip2++;
    455 }
    456 
    457 /*
    458  * Make a copy of the plimit structure.
    459  * We share these structures copy-on-write after fork,
    460  * and copy when a limit is changed.
    461  */
    462 struct plimit *
    463 limcopy(lim)
    464 	struct plimit *lim;
    465 {
    466 	struct plimit *newlim;
    467 
    468 	newlim = pool_get(&plimit_pool, PR_WAITOK);
    469 	memcpy(newlim->pl_rlimit, lim->pl_rlimit,
    470 	    sizeof(struct rlimit) * RLIM_NLIMITS);
    471 	if (lim->pl_corename == defcorename) {
    472 		newlim->pl_corename = defcorename;
    473 	} else {
    474 		newlim->pl_corename = malloc(strlen(lim->pl_corename)+1,
    475 		    M_TEMP, M_WAITOK);
    476 		strcpy(newlim->pl_corename, lim->pl_corename);
    477 	}
    478 	newlim->p_lflags = 0;
    479 	newlim->p_refcnt = 1;
    480 	return (newlim);
    481 }
    482 
    483 void
    484 limfree(lim)
    485 	struct plimit *lim;
    486 {
    487 
    488 	if (--lim->p_refcnt > 0)
    489 		return;
    490 #ifdef DIAGNOSTIC
    491 	if (lim->p_refcnt < 0)
    492 		panic("limfree");
    493 #endif
    494 	if (lim->pl_corename != defcorename)
    495 		free(lim->pl_corename, M_TEMP);
    496 	pool_put(&plimit_pool, lim);
    497 }
    498