Home | History | Annotate | Line # | Download | only in kern
kern_resource.c revision 1.44
      1 /*	$NetBSD: kern_resource.c,v 1.44 1998/02/10 14:09:37 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	8.5 (Berkeley) 1/21/94
     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; p = p->p_pglist.le_next) {
    100 			if (p->p_nice < low)
    101 				low = p->p_nice;
    102 		}
    103 		break;
    104 	}
    105 
    106 	case PRIO_USER:
    107 		if (SCARG(uap, who) == 0)
    108 			SCARG(uap, who) = curp->p_ucred->cr_uid;
    109 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
    110 			if (p->p_ucred->cr_uid == SCARG(uap, who) &&
    111 			    p->p_nice < low)
    112 				low = p->p_nice;
    113 		break;
    114 
    115 	default:
    116 		return (EINVAL);
    117 	}
    118 	if (low == NZERO + PRIO_MAX + 1)
    119 		return (ESRCH);
    120 	*retval = low - NZERO;
    121 	return (0);
    122 }
    123 
    124 /* ARGSUSED */
    125 int
    126 sys_setpriority(curp, v, retval)
    127 	struct proc *curp;
    128 	void *v;
    129 	register_t *retval;
    130 {
    131 	register struct sys_setpriority_args /* {
    132 		syscallarg(int) which;
    133 		syscallarg(int) who;
    134 		syscallarg(int) prio;
    135 	} */ *uap = v;
    136 	register struct proc *p;
    137 	int found = 0, error = 0;
    138 
    139 	switch (SCARG(uap, which)) {
    140 
    141 	case PRIO_PROCESS:
    142 		if (SCARG(uap, who) == 0)
    143 			p = curp;
    144 		else
    145 			p = pfind(SCARG(uap, who));
    146 		if (p == 0)
    147 			break;
    148 		error = donice(curp, p, SCARG(uap, prio));
    149 		found++;
    150 		break;
    151 
    152 	case PRIO_PGRP: {
    153 		register struct pgrp *pg;
    154 
    155 		if (SCARG(uap, who) == 0)
    156 			pg = curp->p_pgrp;
    157 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
    158 			break;
    159 		for (p = pg->pg_members.lh_first; p != 0;
    160 		    p = p->p_pglist.le_next) {
    161 			error = donice(curp, p, SCARG(uap, prio));
    162 			found++;
    163 		}
    164 		break;
    165 	}
    166 
    167 	case PRIO_USER:
    168 		if (SCARG(uap, who) == 0)
    169 			SCARG(uap, who) = curp->p_ucred->cr_uid;
    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 		break;
    176 
    177 	default:
    178 		return (EINVAL);
    179 	}
    180 	if (found == 0)
    181 		return (ESRCH);
    182 	return (error);
    183 }
    184 
    185 int
    186 donice(curp, chgp, n)
    187 	register struct proc *curp, *chgp;
    188 	register int n;
    189 {
    190 	register struct pcred *pcred = curp->p_cred;
    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 	(void)resetpriority(chgp);
    205 	return (0);
    206 }
    207 
    208 /* ARGSUSED */
    209 int
    210 sys_setrlimit(p, v, retval)
    211 	struct proc *p;
    212 	void *v;
    213 	register_t *retval;
    214 {
    215 	register struct sys_setrlimit_args /* {
    216 		syscallarg(int) which;
    217 		syscallarg(const struct rlimit *) rlp;
    218 	} */ *uap = v;
    219 	int which = SCARG(uap, which);
    220 	struct rlimit alim;
    221 	int error;
    222 
    223 	error = copyin(SCARG(uap, rlp), &alim, sizeof (struct rlimit));
    224 	if (error)
    225 		return (error);
    226 	return (dosetrlimit(p, which, &alim));
    227 }
    228 
    229 int
    230 dosetrlimit(p, which, limp)
    231 	struct proc *p;
    232 	int which;
    233 	struct rlimit *limp;
    234 {
    235 	register struct rlimit *alimp;
    236 	extern unsigned maxdmap, maxsmap;
    237 	int error;
    238 
    239 	if ((u_int)which >= RLIM_NLIMITS)
    240 		return (EINVAL);
    241 
    242 	if (limp->rlim_cur < 0 || limp->rlim_max < 0)
    243 		return (EINVAL);
    244 
    245 	alimp = &p->p_rlimit[which];
    246 	if (limp->rlim_cur > alimp->rlim_max ||
    247 	    limp->rlim_max > alimp->rlim_max)
    248 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
    249 			return (error);
    250 	if (limp->rlim_cur > limp->rlim_max)
    251 		limp->rlim_cur = limp->rlim_max;
    252 	if (p->p_limit->p_refcnt > 1 &&
    253 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
    254 		p->p_limit->p_refcnt--;
    255 		p->p_limit = limcopy(p->p_limit);
    256 		alimp = &p->p_rlimit[which];
    257 	}
    258 
    259 	switch (which) {
    260 
    261 	case RLIMIT_DATA:
    262 		if (limp->rlim_cur > maxdmap)
    263 			limp->rlim_cur = maxdmap;
    264 		if (limp->rlim_max > maxdmap)
    265 			limp->rlim_max = maxdmap;
    266 		break;
    267 
    268 	case RLIMIT_STACK:
    269 		if (limp->rlim_cur > maxsmap)
    270 			limp->rlim_cur = maxsmap;
    271 		if (limp->rlim_max > maxsmap)
    272 			limp->rlim_max = maxsmap;
    273 
    274 		/*
    275 		 * Stack is allocated to the max at exec time with
    276 		 * only "rlim_cur" bytes accessible (In other words,
    277 		 * allocates stack dividing two contiguous regions at
    278 		 * "rlim_cur" bytes boundary).
    279 		 *
    280 		 * Since allocation is done in terms of page, roundup
    281 		 * "rlim_cur" (otherwise, contiguous regions
    282 		 * overlap).  If stack limit is going up make more
    283 		 * accessible, if going down make inaccessible.
    284 		 */
    285 		limp->rlim_cur = round_page(limp->rlim_cur);
    286 		if (limp->rlim_cur != alimp->rlim_cur) {
    287 			vm_offset_t addr;
    288 			vm_size_t size;
    289 			vm_prot_t prot;
    290 
    291 			if (limp->rlim_cur > alimp->rlim_cur) {
    292 				prot = VM_PROT_ALL;
    293 				size = limp->rlim_cur - alimp->rlim_cur;
    294 				addr = USRSTACK - limp->rlim_cur;
    295 			} else {
    296 				prot = VM_PROT_NONE;
    297 				size = alimp->rlim_cur - limp->rlim_cur;
    298 				addr = USRSTACK - alimp->rlim_cur;
    299 			}
    300 #if defined(UVM)
    301 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
    302 					      addr, addr+size, prot, FALSE);
    303 #else
    304 			(void) vm_map_protect(&p->p_vmspace->vm_map,
    305 					      addr, addr+size, prot, FALSE);
    306 #endif
    307 		}
    308 		break;
    309 
    310 	case RLIMIT_NOFILE:
    311 		if (limp->rlim_cur > maxfiles)
    312 			limp->rlim_cur = maxfiles;
    313 		if (limp->rlim_max > maxfiles)
    314 			limp->rlim_max = maxfiles;
    315 		break;
    316 
    317 	case RLIMIT_NPROC:
    318 		if (limp->rlim_cur > maxproc)
    319 			limp->rlim_cur = maxproc;
    320 		if (limp->rlim_max > maxproc)
    321 			limp->rlim_max = maxproc;
    322 		break;
    323 	}
    324 	*alimp = *limp;
    325 	return (0);
    326 }
    327 
    328 /* ARGSUSED */
    329 int
    330 sys_getrlimit(p, v, retval)
    331 	struct proc *p;
    332 	void *v;
    333 	register_t *retval;
    334 {
    335 	register struct sys_getrlimit_args /* {
    336 		syscallarg(int) which;
    337 		syscallarg(struct rlimit *) rlp;
    338 	} */ *uap = v;
    339 	int which = SCARG(uap, which);
    340 
    341 	if ((u_int)which >= RLIM_NLIMITS)
    342 		return (EINVAL);
    343 	return (copyout(&p->p_rlimit[which], SCARG(uap, rlp),
    344 	    sizeof (struct rlimit)));
    345 }
    346 
    347 /*
    348  * Transform the running time and tick information in proc p into user,
    349  * system, and interrupt time usage.
    350  */
    351 void
    352 calcru(p, up, sp, ip)
    353 	register struct proc *p;
    354 	register struct timeval *up;
    355 	register struct timeval *sp;
    356 	register struct timeval *ip;
    357 {
    358 	register u_quad_t u, st, ut, it, tot;
    359 	register long sec, usec;
    360 	register int s;
    361 	struct timeval tv;
    362 
    363 	s = splstatclock();
    364 	st = p->p_sticks;
    365 	ut = p->p_uticks;
    366 	it = p->p_iticks;
    367 	splx(s);
    368 
    369 	tot = st + ut + it;
    370 	if (tot == 0) {
    371 		up->tv_sec = up->tv_usec = 0;
    372 		sp->tv_sec = sp->tv_usec = 0;
    373 		if (ip != NULL)
    374 			ip->tv_sec = ip->tv_usec = 0;
    375 		return;
    376 	}
    377 
    378 	sec = p->p_rtime.tv_sec;
    379 	usec = p->p_rtime.tv_usec;
    380 	if (p == curproc) {
    381 		/*
    382 		 * Adjust for the current time slice.  This is actually fairly
    383 		 * important since the error here is on the order of a time
    384 		 * quantum, which is much greater than the sampling error.
    385 		 */
    386 		microtime(&tv);
    387 		sec += tv.tv_sec - runtime.tv_sec;
    388 		usec += tv.tv_usec - runtime.tv_usec;
    389 	}
    390 	u = (u_quad_t) sec * 1000000 + usec;
    391 	st = (u * st) / tot;
    392 	sp->tv_sec = st / 1000000;
    393 	sp->tv_usec = st % 1000000;
    394 	ut = (u * ut) / tot;
    395 	up->tv_sec = ut / 1000000;
    396 	up->tv_usec = ut % 1000000;
    397 	if (ip != NULL) {
    398 		it = (u * it) / tot;
    399 		ip->tv_sec = it / 1000000;
    400 		ip->tv_usec = it % 1000000;
    401 	}
    402 }
    403 
    404 /* ARGSUSED */
    405 int
    406 sys_getrusage(p, v, retval)
    407 	register struct proc *p;
    408 	void *v;
    409 	register_t *retval;
    410 {
    411 	register struct sys_getrusage_args /* {
    412 		syscallarg(int) who;
    413 		syscallarg(struct rusage *) rusage;
    414 	} */ *uap = v;
    415 	register struct rusage *rup;
    416 
    417 	switch (SCARG(uap, who)) {
    418 
    419 	case RUSAGE_SELF:
    420 		rup = &p->p_stats->p_ru;
    421 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
    422 		break;
    423 
    424 	case RUSAGE_CHILDREN:
    425 		rup = &p->p_stats->p_cru;
    426 		break;
    427 
    428 	default:
    429 		return (EINVAL);
    430 	}
    431 	return (copyout(rup, SCARG(uap, rusage), sizeof (struct rusage)));
    432 }
    433 
    434 void
    435 ruadd(ru, ru2)
    436 	register struct rusage *ru, *ru2;
    437 {
    438 	register long *ip, *ip2;
    439 	register int i;
    440 
    441 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
    442 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
    443 	if (ru->ru_maxrss < ru2->ru_maxrss)
    444 		ru->ru_maxrss = ru2->ru_maxrss;
    445 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
    446 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
    447 		*ip++ += *ip2++;
    448 }
    449 
    450 /*
    451  * Make a copy of the plimit structure.
    452  * We share these structures copy-on-write after fork,
    453  * and copy when a limit is changed.
    454  */
    455 struct plimit *
    456 limcopy(lim)
    457 	struct plimit *lim;
    458 {
    459 	register struct plimit *newlim;
    460 
    461 	MALLOC(newlim, struct plimit *, sizeof(struct plimit),
    462 	    M_SUBPROC, M_WAITOK);
    463 	bcopy(lim->pl_rlimit, newlim->pl_rlimit,
    464 	    sizeof(struct rlimit) * RLIM_NLIMITS);
    465 	newlim->p_lflags = 0;
    466 	newlim->p_refcnt = 1;
    467 	return (newlim);
    468 }
    469 
    470 void
    471 limfree(lim)
    472 	struct plimit *lim;
    473 {
    474 
    475 	if (--lim->p_refcnt > 0)
    476 		return;
    477 	FREE(lim, M_SUBPROC);
    478 }
    479