Home | History | Annotate | Line # | Download | only in kern
kern_resource.c revision 1.46
      1 /*	$NetBSD: kern_resource.c,v 1.46 1998/07/31 22:50:50 perry 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 "opt_uvm.h"
     44 
     45 #include <sys/param.h>
     46 #include <sys/systm.h>
     47 #include <sys/kernel.h>
     48 #include <sys/file.h>
     49 #include <sys/resourcevar.h>
     50 #include <sys/malloc.h>
     51 #include <sys/proc.h>
     52 
     53 #include <sys/mount.h>
     54 #include <sys/syscallargs.h>
     55 
     56 #include <vm/vm.h>
     57 
     58 #if defined(UVM)
     59 #include <uvm/uvm_extern.h>
     60 #endif
     61 
     62 void limfree __P((struct plimit *));
     63 /*
     64  * Resource controls and accounting.
     65  */
     66 
     67 int
     68 sys_getpriority(curp, v, retval)
     69 	struct proc *curp;
     70 	void *v;
     71 	register_t *retval;
     72 {
     73 	register struct sys_getpriority_args /* {
     74 		syscallarg(int) which;
     75 		syscallarg(int) who;
     76 	} */ *uap = v;
     77 	register struct proc *p;
     78 	register int low = NZERO + PRIO_MAX + 1;
     79 
     80 	switch (SCARG(uap, which)) {
     81 
     82 	case PRIO_PROCESS:
     83 		if (SCARG(uap, who) == 0)
     84 			p = curp;
     85 		else
     86 			p = pfind(SCARG(uap, who));
     87 		if (p == 0)
     88 			break;
     89 		low = p->p_nice;
     90 		break;
     91 
     92 	case PRIO_PGRP: {
     93 		register struct pgrp *pg;
     94 
     95 		if (SCARG(uap, who) == 0)
     96 			pg = curp->p_pgrp;
     97 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
     98 			break;
     99 		for (p = pg->pg_members.lh_first; p != 0;
    100 		     p = p->p_pglist.le_next) {
    101 			if (p->p_nice < low)
    102 				low = p->p_nice;
    103 		}
    104 		break;
    105 	}
    106 
    107 	case PRIO_USER:
    108 		if (SCARG(uap, who) == 0)
    109 			SCARG(uap, who) = curp->p_ucred->cr_uid;
    110 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
    111 			if (p->p_ucred->cr_uid == SCARG(uap, who) &&
    112 			    p->p_nice < low)
    113 				low = p->p_nice;
    114 		break;
    115 
    116 	default:
    117 		return (EINVAL);
    118 	}
    119 	if (low == NZERO + PRIO_MAX + 1)
    120 		return (ESRCH);
    121 	*retval = low - NZERO;
    122 	return (0);
    123 }
    124 
    125 /* ARGSUSED */
    126 int
    127 sys_setpriority(curp, v, retval)
    128 	struct proc *curp;
    129 	void *v;
    130 	register_t *retval;
    131 {
    132 	register struct sys_setpriority_args /* {
    133 		syscallarg(int) which;
    134 		syscallarg(int) who;
    135 		syscallarg(int) prio;
    136 	} */ *uap = v;
    137 	register struct proc *p;
    138 	int found = 0, error = 0;
    139 
    140 	switch (SCARG(uap, which)) {
    141 
    142 	case PRIO_PROCESS:
    143 		if (SCARG(uap, who) == 0)
    144 			p = curp;
    145 		else
    146 			p = pfind(SCARG(uap, who));
    147 		if (p == 0)
    148 			break;
    149 		error = donice(curp, p, SCARG(uap, prio));
    150 		found++;
    151 		break;
    152 
    153 	case PRIO_PGRP: {
    154 		register struct pgrp *pg;
    155 
    156 		if (SCARG(uap, who) == 0)
    157 			pg = curp->p_pgrp;
    158 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
    159 			break;
    160 		for (p = pg->pg_members.lh_first; p != 0;
    161 		    p = p->p_pglist.le_next) {
    162 			error = donice(curp, p, SCARG(uap, prio));
    163 			found++;
    164 		}
    165 		break;
    166 	}
    167 
    168 	case PRIO_USER:
    169 		if (SCARG(uap, who) == 0)
    170 			SCARG(uap, who) = curp->p_ucred->cr_uid;
    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 		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 	register struct proc *curp, *chgp;
    189 	register int n;
    190 {
    191 	register 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 	register 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, which, &alim));
    228 }
    229 
    230 int
    231 dosetrlimit(p, which, limp)
    232 	struct proc *p;
    233 	int which;
    234 	struct rlimit *limp;
    235 {
    236 	register struct rlimit *alimp;
    237 	extern unsigned maxdmap, maxsmap;
    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 (limp->rlim_cur > alimp->rlim_max ||
    248 	    limp->rlim_max > alimp->rlim_max)
    249 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
    250 			return (error);
    251 	if (limp->rlim_cur > limp->rlim_max)
    252 		limp->rlim_cur = limp->rlim_max;
    253 	if (p->p_limit->p_refcnt > 1 &&
    254 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
    255 		p->p_limit->p_refcnt--;
    256 		p->p_limit = limcopy(p->p_limit);
    257 		alimp = &p->p_rlimit[which];
    258 	}
    259 
    260 	switch (which) {
    261 
    262 	case RLIMIT_DATA:
    263 		if (limp->rlim_cur > maxdmap)
    264 			limp->rlim_cur = maxdmap;
    265 		if (limp->rlim_max > maxdmap)
    266 			limp->rlim_max = maxdmap;
    267 		break;
    268 
    269 	case RLIMIT_STACK:
    270 		if (limp->rlim_cur > maxsmap)
    271 			limp->rlim_cur = maxsmap;
    272 		if (limp->rlim_max > maxsmap)
    273 			limp->rlim_max = maxsmap;
    274 
    275 		/*
    276 		 * Stack is allocated to the max at exec time with
    277 		 * only "rlim_cur" bytes accessible (In other words,
    278 		 * allocates stack dividing two contiguous regions at
    279 		 * "rlim_cur" bytes boundary).
    280 		 *
    281 		 * Since allocation is done in terms of page, roundup
    282 		 * "rlim_cur" (otherwise, contiguous regions
    283 		 * overlap).  If stack limit is going up make more
    284 		 * accessible, if going down make inaccessible.
    285 		 */
    286 		limp->rlim_cur = round_page(limp->rlim_cur);
    287 		if (limp->rlim_cur != alimp->rlim_cur) {
    288 			vm_offset_t addr;
    289 			vm_size_t size;
    290 			vm_prot_t prot;
    291 
    292 			if (limp->rlim_cur > alimp->rlim_cur) {
    293 				prot = VM_PROT_ALL;
    294 				size = limp->rlim_cur - alimp->rlim_cur;
    295 				addr = USRSTACK - limp->rlim_cur;
    296 			} else {
    297 				prot = VM_PROT_NONE;
    298 				size = alimp->rlim_cur - limp->rlim_cur;
    299 				addr = USRSTACK - alimp->rlim_cur;
    300 			}
    301 #if defined(UVM)
    302 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
    303 					      addr, addr+size, prot, FALSE);
    304 #else
    305 			(void) vm_map_protect(&p->p_vmspace->vm_map,
    306 					      addr, addr+size, prot, FALSE);
    307 #endif
    308 		}
    309 		break;
    310 
    311 	case RLIMIT_NOFILE:
    312 		if (limp->rlim_cur > maxfiles)
    313 			limp->rlim_cur = maxfiles;
    314 		if (limp->rlim_max > maxfiles)
    315 			limp->rlim_max = maxfiles;
    316 		break;
    317 
    318 	case RLIMIT_NPROC:
    319 		if (limp->rlim_cur > maxproc)
    320 			limp->rlim_cur = maxproc;
    321 		if (limp->rlim_max > maxproc)
    322 			limp->rlim_max = maxproc;
    323 		break;
    324 	}
    325 	*alimp = *limp;
    326 	return (0);
    327 }
    328 
    329 /* ARGSUSED */
    330 int
    331 sys_getrlimit(p, v, retval)
    332 	struct proc *p;
    333 	void *v;
    334 	register_t *retval;
    335 {
    336 	register struct sys_getrlimit_args /* {
    337 		syscallarg(int) which;
    338 		syscallarg(struct rlimit *) rlp;
    339 	} */ *uap = v;
    340 	int which = SCARG(uap, which);
    341 
    342 	if ((u_int)which >= RLIM_NLIMITS)
    343 		return (EINVAL);
    344 	return (copyout(&p->p_rlimit[which], SCARG(uap, rlp),
    345 	    sizeof(struct rlimit)));
    346 }
    347 
    348 /*
    349  * Transform the running time and tick information in proc p into user,
    350  * system, and interrupt time usage.
    351  */
    352 void
    353 calcru(p, up, sp, ip)
    354 	register struct proc *p;
    355 	register struct timeval *up;
    356 	register struct timeval *sp;
    357 	register struct timeval *ip;
    358 {
    359 	register u_quad_t u, st, ut, it, tot;
    360 	register long sec, usec;
    361 	register int s;
    362 	struct timeval tv;
    363 
    364 	s = splstatclock();
    365 	st = p->p_sticks;
    366 	ut = p->p_uticks;
    367 	it = p->p_iticks;
    368 	splx(s);
    369 
    370 	tot = st + ut + it;
    371 	if (tot == 0) {
    372 		up->tv_sec = up->tv_usec = 0;
    373 		sp->tv_sec = sp->tv_usec = 0;
    374 		if (ip != NULL)
    375 			ip->tv_sec = ip->tv_usec = 0;
    376 		return;
    377 	}
    378 
    379 	sec = p->p_rtime.tv_sec;
    380 	usec = p->p_rtime.tv_usec;
    381 	if (p == curproc) {
    382 		/*
    383 		 * Adjust for the current time slice.  This is actually fairly
    384 		 * important since the error here is on the order of a time
    385 		 * quantum, which is much greater than the sampling error.
    386 		 */
    387 		microtime(&tv);
    388 		sec += tv.tv_sec - runtime.tv_sec;
    389 		usec += tv.tv_usec - runtime.tv_usec;
    390 	}
    391 	u = (u_quad_t) sec * 1000000 + usec;
    392 	st = (u * st) / tot;
    393 	sp->tv_sec = st / 1000000;
    394 	sp->tv_usec = st % 1000000;
    395 	ut = (u * ut) / tot;
    396 	up->tv_sec = ut / 1000000;
    397 	up->tv_usec = ut % 1000000;
    398 	if (ip != NULL) {
    399 		it = (u * it) / tot;
    400 		ip->tv_sec = it / 1000000;
    401 		ip->tv_usec = it % 1000000;
    402 	}
    403 }
    404 
    405 /* ARGSUSED */
    406 int
    407 sys_getrusage(p, v, retval)
    408 	register struct proc *p;
    409 	void *v;
    410 	register_t *retval;
    411 {
    412 	register struct sys_getrusage_args /* {
    413 		syscallarg(int) who;
    414 		syscallarg(struct rusage *) rusage;
    415 	} */ *uap = v;
    416 	register struct rusage *rup;
    417 
    418 	switch (SCARG(uap, who)) {
    419 
    420 	case RUSAGE_SELF:
    421 		rup = &p->p_stats->p_ru;
    422 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
    423 		break;
    424 
    425 	case RUSAGE_CHILDREN:
    426 		rup = &p->p_stats->p_cru;
    427 		break;
    428 
    429 	default:
    430 		return (EINVAL);
    431 	}
    432 	return (copyout(rup, SCARG(uap, rusage), sizeof(struct rusage)));
    433 }
    434 
    435 void
    436 ruadd(ru, ru2)
    437 	register struct rusage *ru, *ru2;
    438 {
    439 	register long *ip, *ip2;
    440 	register int i;
    441 
    442 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
    443 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
    444 	if (ru->ru_maxrss < ru2->ru_maxrss)
    445 		ru->ru_maxrss = ru2->ru_maxrss;
    446 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
    447 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
    448 		*ip++ += *ip2++;
    449 }
    450 
    451 /*
    452  * Make a copy of the plimit structure.
    453  * We share these structures copy-on-write after fork,
    454  * and copy when a limit is changed.
    455  */
    456 struct plimit *
    457 limcopy(lim)
    458 	struct plimit *lim;
    459 {
    460 	register struct plimit *newlim;
    461 
    462 	MALLOC(newlim, struct plimit *, sizeof(struct plimit),
    463 	    M_SUBPROC, M_WAITOK);
    464 	bcopy(lim->pl_rlimit, newlim->pl_rlimit,
    465 	    sizeof(struct rlimit) * RLIM_NLIMITS);
    466 	newlim->p_lflags = 0;
    467 	newlim->p_refcnt = 1;
    468 	return (newlim);
    469 }
    470 
    471 void
    472 limfree(lim)
    473 	struct plimit *lim;
    474 {
    475 
    476 	if (--lim->p_refcnt > 0)
    477 		return;
    478 	FREE(lim, M_SUBPROC);
    479 }
    480