Home | History | Annotate | Line # | Download | only in kern
kern_resource.c revision 1.59
      1 /*	$NetBSD: kern_resource.c,v 1.59 2000/08/20 21:50:11 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 <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 	int s;
    191 
    192 	if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
    193 	    pcred->pc_ucred->cr_uid != chgp->p_ucred->cr_uid &&
    194 	    pcred->p_ruid != chgp->p_ucred->cr_uid)
    195 		return (EPERM);
    196 	if (n > PRIO_MAX)
    197 		n = PRIO_MAX;
    198 	if (n < PRIO_MIN)
    199 		n = PRIO_MIN;
    200 	n += NZERO;
    201 	if (n < chgp->p_nice && suser(pcred->pc_ucred, &curp->p_acflag))
    202 		return (EACCES);
    203 	chgp->p_nice = n;
    204 	SCHED_LOCK(s);
    205 	(void)resetpriority(chgp);
    206 	SCHED_UNLOCK(s);
    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 	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, p->p_cred, which, &alim));
    229 }
    230 
    231 int
    232 dosetrlimit(p, cred, which, limp)
    233 	struct proc *p;
    234 	struct  pcred *cred;
    235 	int which;
    236 	struct rlimit *limp;
    237 {
    238 	struct rlimit *alimp;
    239 	extern unsigned maxdmap, maxsmap;
    240 	struct plimit *newplim;
    241 	int error;
    242 
    243 	if ((u_int)which >= RLIM_NLIMITS)
    244 		return (EINVAL);
    245 
    246 	if (limp->rlim_cur < 0 || limp->rlim_max < 0)
    247 		return (EINVAL);
    248 
    249 	alimp = &p->p_rlimit[which];
    250 	/* if we don't change the value, no need to limcopy() */
    251 	if (limp->rlim_cur == alimp->rlim_cur &&
    252 	    limp->rlim_max == alimp->rlim_max)
    253 		return 0;
    254 
    255 	if (limp->rlim_cur > alimp->rlim_max ||
    256 	    limp->rlim_max > alimp->rlim_max)
    257 		if ((error = suser(cred->pc_ucred, &p->p_acflag)) != 0)
    258 			return (error);
    259 	if (limp->rlim_cur > limp->rlim_max)
    260 		limp->rlim_cur = limp->rlim_max;
    261 	if (p->p_limit->p_refcnt > 1 &&
    262 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
    263 		newplim = limcopy(p->p_limit);
    264 		limfree(p->p_limit);
    265 		p->p_limit = newplim;
    266 		alimp = &p->p_rlimit[which];
    267 	}
    268 
    269 	switch (which) {
    270 
    271 	case RLIMIT_DATA:
    272 		if (limp->rlim_cur > maxdmap)
    273 			limp->rlim_cur = maxdmap;
    274 		if (limp->rlim_max > maxdmap)
    275 			limp->rlim_max = maxdmap;
    276 		break;
    277 
    278 	case RLIMIT_STACK:
    279 		if (limp->rlim_cur > maxsmap)
    280 			limp->rlim_cur = maxsmap;
    281 		if (limp->rlim_max > maxsmap)
    282 			limp->rlim_max = maxsmap;
    283 
    284 		/*
    285 		 * Stack is allocated to the max at exec time with
    286 		 * only "rlim_cur" bytes accessible (In other words,
    287 		 * allocates stack dividing two contiguous regions at
    288 		 * "rlim_cur" bytes boundary).
    289 		 *
    290 		 * Since allocation is done in terms of page, roundup
    291 		 * "rlim_cur" (otherwise, contiguous regions
    292 		 * overlap).  If stack limit is going up make more
    293 		 * accessible, if going down make inaccessible.
    294 		 */
    295 		limp->rlim_cur = round_page(limp->rlim_cur);
    296 		if (limp->rlim_cur != alimp->rlim_cur) {
    297 			vaddr_t addr;
    298 			vsize_t size;
    299 			vm_prot_t prot;
    300 
    301 			if (limp->rlim_cur > alimp->rlim_cur) {
    302 				prot = VM_PROT_ALL;
    303 				size = limp->rlim_cur - alimp->rlim_cur;
    304 				addr = USRSTACK - limp->rlim_cur;
    305 			} else {
    306 				prot = VM_PROT_NONE;
    307 				size = alimp->rlim_cur - limp->rlim_cur;
    308 				addr = USRSTACK - alimp->rlim_cur;
    309 			}
    310 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
    311 					      addr, addr+size, prot, FALSE);
    312 		}
    313 		break;
    314 
    315 	case RLIMIT_NOFILE:
    316 		if (limp->rlim_cur > maxfiles)
    317 			limp->rlim_cur = maxfiles;
    318 		if (limp->rlim_max > maxfiles)
    319 			limp->rlim_max = maxfiles;
    320 		break;
    321 
    322 	case RLIMIT_NPROC:
    323 		if (limp->rlim_cur > maxproc)
    324 			limp->rlim_cur = maxproc;
    325 		if (limp->rlim_max > maxproc)
    326 			limp->rlim_max = maxproc;
    327 		break;
    328 	}
    329 	*alimp = *limp;
    330 	return (0);
    331 }
    332 
    333 /* ARGSUSED */
    334 int
    335 sys_getrlimit(p, v, retval)
    336 	struct proc *p;
    337 	void *v;
    338 	register_t *retval;
    339 {
    340 	struct sys_getrlimit_args /* {
    341 		syscallarg(int) which;
    342 		syscallarg(struct rlimit *) rlp;
    343 	} */ *uap = v;
    344 	int which = SCARG(uap, which);
    345 
    346 	if ((u_int)which >= RLIM_NLIMITS)
    347 		return (EINVAL);
    348 	return (copyout(&p->p_rlimit[which], SCARG(uap, rlp),
    349 	    sizeof(struct rlimit)));
    350 }
    351 
    352 /*
    353  * Transform the running time and tick information in proc p into user,
    354  * system, and interrupt time usage.
    355  */
    356 void
    357 calcru(p, up, sp, ip)
    358 	struct proc *p;
    359 	struct timeval *up;
    360 	struct timeval *sp;
    361 	struct timeval *ip;
    362 {
    363 	u_quad_t u, st, ut, it, tot;
    364 	long sec, usec;
    365 	int s;
    366 	struct timeval tv;
    367 
    368 	s = splstatclock();
    369 	st = p->p_sticks;
    370 	ut = p->p_uticks;
    371 	it = p->p_iticks;
    372 	splx(s);
    373 
    374 	tot = st + ut + it;
    375 	if (tot == 0) {
    376 		up->tv_sec = up->tv_usec = 0;
    377 		sp->tv_sec = sp->tv_usec = 0;
    378 		if (ip != NULL)
    379 			ip->tv_sec = ip->tv_usec = 0;
    380 		return;
    381 	}
    382 
    383 	sec = p->p_rtime.tv_sec;
    384 	usec = p->p_rtime.tv_usec;
    385 	if (p->p_stat == SONPROC) {
    386 		struct schedstate_percpu *spc;
    387 
    388 		KDASSERT(p->p_cpu != NULL);
    389 		spc = &p->p_cpu->ci_schedstate;
    390 
    391 		/*
    392 		 * Adjust for the current time slice.  This is actually fairly
    393 		 * important since the error here is on the order of a time
    394 		 * quantum, which is much greater than the sampling error.
    395 		 */
    396 		microtime(&tv);
    397 		sec += tv.tv_sec - spc->spc_runtime.tv_sec;
    398 		usec += tv.tv_usec - spc->spc_runtime.tv_usec;
    399 	}
    400 	u = (u_quad_t) sec * 1000000 + usec;
    401 	st = (u * st) / tot;
    402 	sp->tv_sec = st / 1000000;
    403 	sp->tv_usec = st % 1000000;
    404 	ut = (u * ut) / tot;
    405 	up->tv_sec = ut / 1000000;
    406 	up->tv_usec = ut % 1000000;
    407 	if (ip != NULL) {
    408 		it = (u * it) / tot;
    409 		ip->tv_sec = it / 1000000;
    410 		ip->tv_usec = it % 1000000;
    411 	}
    412 }
    413 
    414 /* ARGSUSED */
    415 int
    416 sys_getrusage(p, v, retval)
    417 	struct proc *p;
    418 	void *v;
    419 	register_t *retval;
    420 {
    421 	struct sys_getrusage_args /* {
    422 		syscallarg(int) who;
    423 		syscallarg(struct rusage *) rusage;
    424 	} */ *uap = v;
    425 	struct rusage *rup;
    426 
    427 	switch (SCARG(uap, who)) {
    428 
    429 	case RUSAGE_SELF:
    430 		rup = &p->p_stats->p_ru;
    431 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
    432 		break;
    433 
    434 	case RUSAGE_CHILDREN:
    435 		rup = &p->p_stats->p_cru;
    436 		break;
    437 
    438 	default:
    439 		return (EINVAL);
    440 	}
    441 	return (copyout(rup, SCARG(uap, rusage), sizeof(struct rusage)));
    442 }
    443 
    444 void
    445 ruadd(ru, ru2)
    446 	struct rusage *ru, *ru2;
    447 {
    448 	long *ip, *ip2;
    449 	int i;
    450 
    451 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
    452 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
    453 	if (ru->ru_maxrss < ru2->ru_maxrss)
    454 		ru->ru_maxrss = ru2->ru_maxrss;
    455 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
    456 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
    457 		*ip++ += *ip2++;
    458 }
    459 
    460 /*
    461  * Make a copy of the plimit structure.
    462  * We share these structures copy-on-write after fork,
    463  * and copy when a limit is changed.
    464  */
    465 struct plimit *
    466 limcopy(lim)
    467 	struct plimit *lim;
    468 {
    469 	struct plimit *newlim;
    470 
    471 	newlim = pool_get(&plimit_pool, PR_WAITOK);
    472 	memcpy(newlim->pl_rlimit, lim->pl_rlimit,
    473 	    sizeof(struct rlimit) * RLIM_NLIMITS);
    474 	if (lim->pl_corename == defcorename) {
    475 		newlim->pl_corename = defcorename;
    476 	} else {
    477 		newlim->pl_corename = malloc(strlen(lim->pl_corename)+1,
    478 		    M_TEMP, M_WAITOK);
    479 		strcpy(newlim->pl_corename, lim->pl_corename);
    480 	}
    481 	newlim->p_lflags = 0;
    482 	newlim->p_refcnt = 1;
    483 	return (newlim);
    484 }
    485 
    486 void
    487 limfree(lim)
    488 	struct plimit *lim;
    489 {
    490 
    491 	if (--lim->p_refcnt > 0)
    492 		return;
    493 #ifdef DIAGNOSTIC
    494 	if (lim->p_refcnt < 0)
    495 		panic("limfree");
    496 #endif
    497 	if (lim->pl_corename != defcorename)
    498 		free(lim->pl_corename, M_TEMP);
    499 	pool_put(&plimit_pool, lim);
    500 }
    501