Home | History | Annotate | Line # | Download | only in kern
kern_resource.c revision 1.56.2.1
      1 /*	$NetBSD: kern_resource.c,v 1.56.2.1 2000/06/22 17:09:09 minoura 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 /*
     60  * Resource controls and accounting.
     61  */
     62 
     63 int
     64 sys_getpriority(curp, v, retval)
     65 	struct proc *curp;
     66 	void *v;
     67 	register_t *retval;
     68 {
     69 	struct sys_getpriority_args /* {
     70 		syscallarg(int) which;
     71 		syscallarg(int) who;
     72 	} */ *uap = v;
     73 	struct proc *p;
     74 	int low = NZERO + PRIO_MAX + 1;
     75 
     76 	switch (SCARG(uap, which)) {
     77 
     78 	case PRIO_PROCESS:
     79 		if (SCARG(uap, who) == 0)
     80 			p = curp;
     81 		else
     82 			p = pfind(SCARG(uap, who));
     83 		if (p == 0)
     84 			break;
     85 		low = p->p_nice;
     86 		break;
     87 
     88 	case PRIO_PGRP: {
     89 		struct pgrp *pg;
     90 
     91 		if (SCARG(uap, who) == 0)
     92 			pg = curp->p_pgrp;
     93 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
     94 			break;
     95 		for (p = pg->pg_members.lh_first; p != 0;
     96 		     p = p->p_pglist.le_next) {
     97 			if (p->p_nice < low)
     98 				low = p->p_nice;
     99 		}
    100 		break;
    101 	}
    102 
    103 	case PRIO_USER:
    104 		if (SCARG(uap, who) == 0)
    105 			SCARG(uap, who) = curp->p_ucred->cr_uid;
    106 		proclist_lock_read();
    107 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
    108 			if (p->p_ucred->cr_uid == SCARG(uap, who) &&
    109 			    p->p_nice < low)
    110 				low = p->p_nice;
    111 		proclist_unlock_read();
    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 	struct sys_setpriority_args /* {
    131 		syscallarg(int) which;
    132 		syscallarg(int) who;
    133 		syscallarg(int) prio;
    134 	} */ *uap = v;
    135 	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 		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 		proclist_lock_read();
    170 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
    171 			if (p->p_ucred->cr_uid == SCARG(uap, who)) {
    172 				error = donice(curp, p, SCARG(uap, prio));
    173 				found++;
    174 			}
    175 		proclist_unlock_read();
    176 		break;
    177 
    178 	default:
    179 		return (EINVAL);
    180 	}
    181 	if (found == 0)
    182 		return (ESRCH);
    183 	return (error);
    184 }
    185 
    186 int
    187 donice(curp, chgp, n)
    188 	struct proc *curp, *chgp;
    189 	int n;
    190 {
    191 	struct pcred *pcred = curp->p_cred;
    192 
    193 	if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
    194 	    pcred->pc_ucred->cr_uid != chgp->p_ucred->cr_uid &&
    195 	    pcred->p_ruid != chgp->p_ucred->cr_uid)
    196 		return (EPERM);
    197 	if (n > PRIO_MAX)
    198 		n = PRIO_MAX;
    199 	if (n < PRIO_MIN)
    200 		n = PRIO_MIN;
    201 	n += NZERO;
    202 	if (n < chgp->p_nice && suser(pcred->pc_ucred, &curp->p_acflag))
    203 		return (EACCES);
    204 	chgp->p_nice = n;
    205 	(void)resetpriority(chgp);
    206 	return (0);
    207 }
    208 
    209 /* ARGSUSED */
    210 int
    211 sys_setrlimit(p, v, retval)
    212 	struct proc *p;
    213 	void *v;
    214 	register_t *retval;
    215 {
    216 	struct sys_setrlimit_args /* {
    217 		syscallarg(int) which;
    218 		syscallarg(const struct rlimit *) rlp;
    219 	} */ *uap = v;
    220 	int which = SCARG(uap, which);
    221 	struct rlimit alim;
    222 	int error;
    223 
    224 	error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit));
    225 	if (error)
    226 		return (error);
    227 	return (dosetrlimit(p, p->p_cred, which, &alim));
    228 }
    229 
    230 int
    231 dosetrlimit(p, cred, which, limp)
    232 	struct proc *p;
    233 	struct  pcred *cred;
    234 	int which;
    235 	struct rlimit *limp;
    236 {
    237 	struct rlimit *alimp;
    238 	extern unsigned maxdmap, maxsmap;
    239 	struct plimit *newplim;
    240 	int error;
    241 
    242 	if ((u_int)which >= RLIM_NLIMITS)
    243 		return (EINVAL);
    244 
    245 	if (limp->rlim_cur < 0 || limp->rlim_max < 0)
    246 		return (EINVAL);
    247 
    248 	alimp = &p->p_rlimit[which];
    249 	/* if we don't change the value, no need to limcopy() */
    250 	if (limp->rlim_cur == alimp->rlim_cur &&
    251 	    limp->rlim_max == alimp->rlim_max)
    252 		return 0;
    253 
    254 	if (limp->rlim_cur > alimp->rlim_max ||
    255 	    limp->rlim_max > alimp->rlim_max)
    256 		if ((error = suser(cred->pc_ucred, &p->p_acflag)) != 0)
    257 			return (error);
    258 	if (limp->rlim_cur > limp->rlim_max)
    259 		limp->rlim_cur = limp->rlim_max;
    260 	if (p->p_limit->p_refcnt > 1 &&
    261 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
    262 		newplim = limcopy(p->p_limit);
    263 		limfree(p->p_limit);
    264 		p->p_limit = newplim;
    265 		alimp = &p->p_rlimit[which];
    266 	}
    267 
    268 	switch (which) {
    269 
    270 	case RLIMIT_DATA:
    271 		if (limp->rlim_cur > maxdmap)
    272 			limp->rlim_cur = maxdmap;
    273 		if (limp->rlim_max > maxdmap)
    274 			limp->rlim_max = maxdmap;
    275 		break;
    276 
    277 	case RLIMIT_STACK:
    278 		if (limp->rlim_cur > maxsmap)
    279 			limp->rlim_cur = maxsmap;
    280 		if (limp->rlim_max > maxsmap)
    281 			limp->rlim_max = maxsmap;
    282 
    283 		/*
    284 		 * Stack is allocated to the max at exec time with
    285 		 * only "rlim_cur" bytes accessible (In other words,
    286 		 * allocates stack dividing two contiguous regions at
    287 		 * "rlim_cur" bytes boundary).
    288 		 *
    289 		 * Since allocation is done in terms of page, roundup
    290 		 * "rlim_cur" (otherwise, contiguous regions
    291 		 * overlap).  If stack limit is going up make more
    292 		 * accessible, if going down make inaccessible.
    293 		 */
    294 		limp->rlim_cur = round_page(limp->rlim_cur);
    295 		if (limp->rlim_cur != alimp->rlim_cur) {
    296 			vaddr_t addr;
    297 			vsize_t size;
    298 			vm_prot_t prot;
    299 
    300 			if (limp->rlim_cur > alimp->rlim_cur) {
    301 				prot = VM_PROT_ALL;
    302 				size = limp->rlim_cur - alimp->rlim_cur;
    303 				addr = USRSTACK - limp->rlim_cur;
    304 			} else {
    305 				prot = VM_PROT_NONE;
    306 				size = alimp->rlim_cur - limp->rlim_cur;
    307 				addr = USRSTACK - alimp->rlim_cur;
    308 			}
    309 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
    310 					      addr, addr+size, prot, FALSE);
    311 		}
    312 		break;
    313 
    314 	case RLIMIT_NOFILE:
    315 		if (limp->rlim_cur > maxfiles)
    316 			limp->rlim_cur = maxfiles;
    317 		if (limp->rlim_max > maxfiles)
    318 			limp->rlim_max = maxfiles;
    319 		break;
    320 
    321 	case RLIMIT_NPROC:
    322 		if (limp->rlim_cur > maxproc)
    323 			limp->rlim_cur = maxproc;
    324 		if (limp->rlim_max > maxproc)
    325 			limp->rlim_max = maxproc;
    326 		break;
    327 	}
    328 	*alimp = *limp;
    329 	return (0);
    330 }
    331 
    332 /* ARGSUSED */
    333 int
    334 sys_getrlimit(p, v, retval)
    335 	struct proc *p;
    336 	void *v;
    337 	register_t *retval;
    338 {
    339 	struct sys_getrlimit_args /* {
    340 		syscallarg(int) which;
    341 		syscallarg(struct rlimit *) rlp;
    342 	} */ *uap = v;
    343 	int which = SCARG(uap, which);
    344 
    345 	if ((u_int)which >= RLIM_NLIMITS)
    346 		return (EINVAL);
    347 	return (copyout(&p->p_rlimit[which], SCARG(uap, rlp),
    348 	    sizeof(struct rlimit)));
    349 }
    350 
    351 /*
    352  * Transform the running time and tick information in proc p into user,
    353  * system, and interrupt time usage.
    354  */
    355 void
    356 calcru(p, up, sp, ip)
    357 	struct proc *p;
    358 	struct timeval *up;
    359 	struct timeval *sp;
    360 	struct timeval *ip;
    361 {
    362 	u_quad_t u, st, ut, it, tot;
    363 	long sec, usec;
    364 	int s;
    365 	struct timeval tv;
    366 
    367 	s = splstatclock();
    368 	st = p->p_sticks;
    369 	ut = p->p_uticks;
    370 	it = p->p_iticks;
    371 	splx(s);
    372 
    373 	tot = st + ut + it;
    374 	if (tot == 0) {
    375 		up->tv_sec = up->tv_usec = 0;
    376 		sp->tv_sec = sp->tv_usec = 0;
    377 		if (ip != NULL)
    378 			ip->tv_sec = ip->tv_usec = 0;
    379 		return;
    380 	}
    381 
    382 	sec = p->p_rtime.tv_sec;
    383 	usec = p->p_rtime.tv_usec;
    384 	if (p->p_stat == SONPROC) {
    385 		struct schedstate_percpu *spc;
    386 
    387 		KDASSERT(p->p_cpu != NULL);
    388 		spc = &p->p_cpu->ci_schedstate;
    389 
    390 		/*
    391 		 * Adjust for the current time slice.  This is actually fairly
    392 		 * important since the error here is on the order of a time
    393 		 * quantum, which is much greater than the sampling error.
    394 		 */
    395 		microtime(&tv);
    396 		sec += tv.tv_sec - spc->spc_runtime.tv_sec;
    397 		usec += tv.tv_usec - spc->spc_runtime.tv_usec;
    398 	}
    399 	u = (u_quad_t) sec * 1000000 + usec;
    400 	st = (u * st) / tot;
    401 	sp->tv_sec = st / 1000000;
    402 	sp->tv_usec = st % 1000000;
    403 	ut = (u * ut) / tot;
    404 	up->tv_sec = ut / 1000000;
    405 	up->tv_usec = ut % 1000000;
    406 	if (ip != NULL) {
    407 		it = (u * it) / tot;
    408 		ip->tv_sec = it / 1000000;
    409 		ip->tv_usec = it % 1000000;
    410 	}
    411 }
    412 
    413 /* ARGSUSED */
    414 int
    415 sys_getrusage(p, v, retval)
    416 	struct proc *p;
    417 	void *v;
    418 	register_t *retval;
    419 {
    420 	struct sys_getrusage_args /* {
    421 		syscallarg(int) who;
    422 		syscallarg(struct rusage *) rusage;
    423 	} */ *uap = v;
    424 	struct rusage *rup;
    425 
    426 	switch (SCARG(uap, who)) {
    427 
    428 	case RUSAGE_SELF:
    429 		rup = &p->p_stats->p_ru;
    430 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
    431 		break;
    432 
    433 	case RUSAGE_CHILDREN:
    434 		rup = &p->p_stats->p_cru;
    435 		break;
    436 
    437 	default:
    438 		return (EINVAL);
    439 	}
    440 	return (copyout(rup, SCARG(uap, rusage), sizeof(struct rusage)));
    441 }
    442 
    443 void
    444 ruadd(ru, ru2)
    445 	struct rusage *ru, *ru2;
    446 {
    447 	long *ip, *ip2;
    448 	int i;
    449 
    450 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
    451 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
    452 	if (ru->ru_maxrss < ru2->ru_maxrss)
    453 		ru->ru_maxrss = ru2->ru_maxrss;
    454 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
    455 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
    456 		*ip++ += *ip2++;
    457 }
    458 
    459 /*
    460  * Make a copy of the plimit structure.
    461  * We share these structures copy-on-write after fork,
    462  * and copy when a limit is changed.
    463  */
    464 struct plimit *
    465 limcopy(lim)
    466 	struct plimit *lim;
    467 {
    468 	struct plimit *newlim;
    469 
    470 	newlim = pool_get(&plimit_pool, PR_WAITOK);
    471 	memcpy(newlim->pl_rlimit, lim->pl_rlimit,
    472 	    sizeof(struct rlimit) * RLIM_NLIMITS);
    473 	if (lim->pl_corename == defcorename) {
    474 		newlim->pl_corename = defcorename;
    475 	} else {
    476 		newlim->pl_corename = malloc(strlen(lim->pl_corename)+1,
    477 		    M_TEMP, M_WAITOK);
    478 		strcpy(newlim->pl_corename, lim->pl_corename);
    479 	}
    480 	newlim->p_lflags = 0;
    481 	newlim->p_refcnt = 1;
    482 	return (newlim);
    483 }
    484 
    485 void
    486 limfree(lim)
    487 	struct plimit *lim;
    488 {
    489 
    490 	if (--lim->p_refcnt > 0)
    491 		return;
    492 #ifdef DIAGNOSTIC
    493 	if (lim->p_refcnt < 0)
    494 		panic("limfree");
    495 #endif
    496 	if (lim->pl_corename != defcorename)
    497 		free(lim->pl_corename, M_TEMP);
    498 	pool_put(&plimit_pool, lim);
    499 }
    500