Home | History | Annotate | Line # | Download | only in kern
kern_resource.c revision 1.60.2.4
      1 /*	$NetBSD: kern_resource.c,v 1.60.2.4 2002/05/29 21:33:12 nathanw 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/cdefs.h>
     44 __KERNEL_RCSID(0, "$NetBSD: kern_resource.c,v 1.60.2.4 2002/05/29 21:33:12 nathanw Exp $");
     45 
     46 #include <sys/param.h>
     47 #include <sys/systm.h>
     48 #include <sys/kernel.h>
     49 #include <sys/file.h>
     50 #include <sys/resourcevar.h>
     51 #include <sys/malloc.h>
     52 #include <sys/pool.h>
     53 #include <sys/lwp.h>
     54 #include <sys/proc.h>
     55 
     56 #include <sys/mount.h>
     57 #include <sys/sa.h>
     58 #include <sys/syscallargs.h>
     59 
     60 #include <uvm/uvm_extern.h>
     61 
     62 /*
     63  * Maximum process data and stack limits.
     64  * They are variables so they are patchable.
     65  *
     66  * XXXX Do we really need them to be patchable?
     67  */
     68 rlim_t maxdmap = MAXDSIZ;
     69 rlim_t maxsmap = MAXSSIZ;
     70 
     71 /*
     72  * Resource controls and accounting.
     73  */
     74 
     75 int
     76 sys_getpriority(l, v, retval)
     77 	struct lwp *l;
     78 	void *v;
     79 	register_t *retval;
     80 {
     81 	struct sys_getpriority_args /* {
     82 		syscallarg(int) which;
     83 		syscallarg(int) who;
     84 	} */ *uap = v;
     85 	struct proc *curp = l->l_proc, *p;
     86 	int low = NZERO + PRIO_MAX + 1;
     87 
     88 	switch (SCARG(uap, which)) {
     89 
     90 	case PRIO_PROCESS:
     91 		if (SCARG(uap, who) == 0)
     92 			p = curp;
     93 		else
     94 			p = pfind(SCARG(uap, who));
     95 		if (p == 0)
     96 			break;
     97 		low = p->p_nice;
     98 		break;
     99 
    100 	case PRIO_PGRP: {
    101 		struct pgrp *pg;
    102 
    103 		if (SCARG(uap, who) == 0)
    104 			pg = curp->p_pgrp;
    105 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
    106 			break;
    107 		for (p = pg->pg_members.lh_first; p != 0;
    108 		     p = p->p_pglist.le_next) {
    109 			if (p->p_nice < low)
    110 				low = p->p_nice;
    111 		}
    112 		break;
    113 	}
    114 
    115 	case PRIO_USER:
    116 		if (SCARG(uap, who) == 0)
    117 			SCARG(uap, who) = curp->p_ucred->cr_uid;
    118 		proclist_lock_read();
    119 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
    120 			if (p->p_ucred->cr_uid == SCARG(uap, who) &&
    121 			    p->p_nice < low)
    122 				low = p->p_nice;
    123 		proclist_unlock_read();
    124 		break;
    125 
    126 	default:
    127 		return (EINVAL);
    128 	}
    129 	if (low == NZERO + PRIO_MAX + 1)
    130 		return (ESRCH);
    131 	*retval = low - NZERO;
    132 	return (0);
    133 }
    134 
    135 /* ARGSUSED */
    136 int
    137 sys_setpriority(l, v, retval)
    138 	struct lwp *l;
    139 	void *v;
    140 	register_t *retval;
    141 {
    142 	struct sys_setpriority_args /* {
    143 		syscallarg(int) which;
    144 		syscallarg(int) who;
    145 		syscallarg(int) prio;
    146 	} */ *uap = v;
    147 	struct proc *curp = l->l_proc, *p;
    148 	int found = 0, error = 0;
    149 
    150 	switch (SCARG(uap, which)) {
    151 
    152 	case PRIO_PROCESS:
    153 		if (SCARG(uap, who) == 0)
    154 			p = curp;
    155 		else
    156 			p = pfind(SCARG(uap, who));
    157 		if (p == 0)
    158 			break;
    159 		error = donice(curp, p, SCARG(uap, prio));
    160 		found++;
    161 		break;
    162 
    163 	case PRIO_PGRP: {
    164 		struct pgrp *pg;
    165 
    166 		if (SCARG(uap, who) == 0)
    167 			pg = curp->p_pgrp;
    168 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
    169 			break;
    170 		for (p = pg->pg_members.lh_first; p != 0;
    171 		    p = p->p_pglist.le_next) {
    172 			error = donice(curp, p, SCARG(uap, prio));
    173 			found++;
    174 		}
    175 		break;
    176 	}
    177 
    178 	case PRIO_USER:
    179 		if (SCARG(uap, who) == 0)
    180 			SCARG(uap, who) = curp->p_ucred->cr_uid;
    181 		proclist_lock_read();
    182 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next)
    183 			if (p->p_ucred->cr_uid == SCARG(uap, who)) {
    184 				error = donice(curp, p, SCARG(uap, prio));
    185 				found++;
    186 			}
    187 		proclist_unlock_read();
    188 		break;
    189 
    190 	default:
    191 		return (EINVAL);
    192 	}
    193 	if (found == 0)
    194 		return (ESRCH);
    195 	return (error);
    196 }
    197 
    198 int
    199 donice(curp, chgp, n)
    200 	struct proc *curp, *chgp;
    201 	int n;
    202 {
    203 	struct pcred *pcred = curp->p_cred;
    204 	int s;
    205 
    206 	if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
    207 	    pcred->pc_ucred->cr_uid != chgp->p_ucred->cr_uid &&
    208 	    pcred->p_ruid != chgp->p_ucred->cr_uid)
    209 		return (EPERM);
    210 	if (n > PRIO_MAX)
    211 		n = PRIO_MAX;
    212 	if (n < PRIO_MIN)
    213 		n = PRIO_MIN;
    214 	n += NZERO;
    215 	if (n < chgp->p_nice && suser(pcred->pc_ucred, &curp->p_acflag))
    216 		return (EACCES);
    217 	chgp->p_nice = n;
    218 	SCHED_LOCK(s);
    219 	(void)resetprocpriority(chgp);
    220 	SCHED_UNLOCK(s);
    221 	return (0);
    222 }
    223 
    224 /* ARGSUSED */
    225 int
    226 sys_setrlimit(l, v, retval)
    227 	struct lwp *l;
    228 	void *v;
    229 	register_t *retval;
    230 {
    231 	struct sys_setrlimit_args /* {
    232 		syscallarg(int) which;
    233 		syscallarg(const struct rlimit *) rlp;
    234 	} */ *uap = v;
    235 	struct proc *p = l->l_proc;
    236 	int which = SCARG(uap, which);
    237 	struct rlimit alim;
    238 	int error;
    239 
    240 	error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit));
    241 	if (error)
    242 		return (error);
    243 	return (dosetrlimit(p, p->p_cred, which, &alim));
    244 }
    245 
    246 int
    247 dosetrlimit(p, cred, which, limp)
    248 	struct proc *p;
    249 	struct  pcred *cred;
    250 	int which;
    251 	struct rlimit *limp;
    252 {
    253 	struct rlimit *alimp;
    254 	struct plimit *newplim;
    255 	int error;
    256 
    257 	if ((u_int)which >= RLIM_NLIMITS)
    258 		return (EINVAL);
    259 
    260 	if (limp->rlim_cur < 0 || limp->rlim_max < 0)
    261 		return (EINVAL);
    262 
    263 	alimp = &p->p_rlimit[which];
    264 	/* if we don't change the value, no need to limcopy() */
    265 	if (limp->rlim_cur == alimp->rlim_cur &&
    266 	    limp->rlim_max == alimp->rlim_max)
    267 		return 0;
    268 
    269 	if (limp->rlim_cur > limp->rlim_max) {
    270 		/*
    271 		 * This is programming error. According to SUSv2, we should
    272 		 * return error in this case.
    273 		 */
    274 		return (EINVAL);
    275 	}
    276 	if (limp->rlim_max > alimp->rlim_max
    277 	    && (error = suser(cred->pc_ucred, &p->p_acflag)) != 0)
    278 			return (error);
    279 
    280 	if (p->p_limit->p_refcnt > 1 &&
    281 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
    282 		newplim = limcopy(p->p_limit);
    283 		limfree(p->p_limit);
    284 		p->p_limit = newplim;
    285 		alimp = &p->p_rlimit[which];
    286 	}
    287 
    288 	switch (which) {
    289 
    290 	case RLIMIT_DATA:
    291 		if (limp->rlim_cur > maxdmap)
    292 			limp->rlim_cur = maxdmap;
    293 		if (limp->rlim_max > maxdmap)
    294 			limp->rlim_max = maxdmap;
    295 		break;
    296 
    297 	case RLIMIT_STACK:
    298 		if (limp->rlim_cur > maxsmap)
    299 			limp->rlim_cur = maxsmap;
    300 		if (limp->rlim_max > maxsmap)
    301 			limp->rlim_max = maxsmap;
    302 
    303 		/*
    304 		 * Return EINVAL if the new stack size limit is lower than
    305 		 * current usage. Otherwise, the process would get SIGSEGV the
    306 		 * moment it would try to access anything on it's current stack.
    307 		 * This conforms to SUSv2.
    308 		 */
    309 		if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE
    310 		    || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE)
    311 			return (EINVAL);
    312 
    313 		/*
    314 		 * Stack is allocated to the max at exec time with
    315 		 * only "rlim_cur" bytes accessible (In other words,
    316 		 * allocates stack dividing two contiguous regions at
    317 		 * "rlim_cur" bytes boundary).
    318 		 *
    319 		 * Since allocation is done in terms of page, roundup
    320 		 * "rlim_cur" (otherwise, contiguous regions
    321 		 * overlap).  If stack limit is going up make more
    322 		 * accessible, if going down make inaccessible.
    323 		 */
    324 		limp->rlim_cur = round_page(limp->rlim_cur);
    325 		if (limp->rlim_cur != alimp->rlim_cur) {
    326 			vaddr_t addr;
    327 			vsize_t size;
    328 			vm_prot_t prot;
    329 
    330 			if (limp->rlim_cur > alimp->rlim_cur) {
    331 				prot = VM_PROT_ALL;
    332 				size = limp->rlim_cur - alimp->rlim_cur;
    333 				addr = USRSTACK - limp->rlim_cur;
    334 			} else {
    335 				prot = VM_PROT_NONE;
    336 				size = alimp->rlim_cur - limp->rlim_cur;
    337 				addr = USRSTACK - alimp->rlim_cur;
    338 			}
    339 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
    340 					      addr, addr+size, prot, FALSE);
    341 		}
    342 		break;
    343 
    344 	case RLIMIT_NOFILE:
    345 		if (limp->rlim_cur > maxfiles)
    346 			limp->rlim_cur = maxfiles;
    347 		if (limp->rlim_max > maxfiles)
    348 			limp->rlim_max = maxfiles;
    349 		break;
    350 
    351 	case RLIMIT_NPROC:
    352 		if (limp->rlim_cur > maxproc)
    353 			limp->rlim_cur = maxproc;
    354 		if (limp->rlim_max > maxproc)
    355 			limp->rlim_max = maxproc;
    356 		break;
    357 	}
    358 	*alimp = *limp;
    359 	return (0);
    360 }
    361 
    362 /* ARGSUSED */
    363 int
    364 sys_getrlimit(l, v, retval)
    365 	struct lwp *l;
    366 	void *v;
    367 	register_t *retval;
    368 {
    369 	struct sys_getrlimit_args /* {
    370 		syscallarg(int) which;
    371 		syscallarg(struct rlimit *) rlp;
    372 	} */ *uap = v;
    373 	struct proc *p = l->l_proc;
    374 	int which = SCARG(uap, which);
    375 
    376 	if ((u_int)which >= RLIM_NLIMITS)
    377 		return (EINVAL);
    378 	return (copyout(&p->p_rlimit[which], SCARG(uap, rlp),
    379 	    sizeof(struct rlimit)));
    380 }
    381 
    382 /*
    383  * Transform the running time and tick information in proc p into user,
    384  * system, and interrupt time usage.
    385  */
    386 void
    387 calcru(p, up, sp, ip)
    388 	struct proc *p;
    389 	struct timeval *up;
    390 	struct timeval *sp;
    391 	struct timeval *ip;
    392 {
    393 	u_quad_t u, st, ut, it, tot;
    394 	long sec, usec;
    395 	int s;
    396 	struct timeval tv;
    397 	struct lwp *l;
    398 
    399 	s = splstatclock();
    400 	st = p->p_sticks;
    401 	ut = p->p_uticks;
    402 	it = p->p_iticks;
    403 	splx(s);
    404 
    405 	tot = st + ut + it;
    406 	if (tot == 0) {
    407 		up->tv_sec = up->tv_usec = 0;
    408 		sp->tv_sec = sp->tv_usec = 0;
    409 		if (ip != NULL)
    410 			ip->tv_sec = ip->tv_usec = 0;
    411 		return;
    412 	}
    413 
    414 	sec = p->p_rtime.tv_sec;
    415 	usec = p->p_rtime.tv_usec;
    416 	for (l = LIST_FIRST(&p->p_lwps); l != NULL;
    417 	     l = LIST_NEXT(l, l_sibling)) {
    418 		if (l->l_stat == LSONPROC) {
    419 			struct schedstate_percpu *spc;
    420 
    421 			KDASSERT(l->l_cpu != NULL);
    422 			spc = &l->l_cpu->ci_schedstate;
    423 
    424 			/*
    425 			 * Adjust for the current time slice.  This is
    426 			 * actually fairly important since the error
    427 			 * here is on the order of a time quantum,
    428 			 * which is much greater than the sampling
    429 			 * error.
    430 			 */
    431 			microtime(&tv);
    432 			sec += tv.tv_sec - spc->spc_runtime.tv_sec;
    433 			usec += tv.tv_usec - spc->spc_runtime.tv_usec;
    434 
    435 			break;
    436 		}
    437 	}
    438 	u = (u_quad_t) sec * 1000000 + usec;
    439 	st = (u * st) / tot;
    440 	sp->tv_sec = st / 1000000;
    441 	sp->tv_usec = st % 1000000;
    442 	ut = (u * ut) / tot;
    443 	up->tv_sec = ut / 1000000;
    444 	up->tv_usec = ut % 1000000;
    445 	if (ip != NULL) {
    446 		it = (u * it) / tot;
    447 		ip->tv_sec = it / 1000000;
    448 		ip->tv_usec = it % 1000000;
    449 	}
    450 }
    451 
    452 /* ARGSUSED */
    453 int
    454 sys_getrusage(l, v, retval)
    455 	struct lwp *l;
    456 	void *v;
    457 	register_t *retval;
    458 {
    459 	struct sys_getrusage_args /* {
    460 		syscallarg(int) who;
    461 		syscallarg(struct rusage *) rusage;
    462 	} */ *uap = v;
    463 	struct rusage *rup;
    464 	struct proc *p = l->l_proc;
    465 
    466 	switch (SCARG(uap, who)) {
    467 
    468 	case RUSAGE_SELF:
    469 		rup = &p->p_stats->p_ru;
    470 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
    471 		break;
    472 
    473 	case RUSAGE_CHILDREN:
    474 		rup = &p->p_stats->p_cru;
    475 		break;
    476 
    477 	default:
    478 		return (EINVAL);
    479 	}
    480 	return (copyout(rup, SCARG(uap, rusage), sizeof(struct rusage)));
    481 }
    482 
    483 void
    484 ruadd(ru, ru2)
    485 	struct rusage *ru, *ru2;
    486 {
    487 	long *ip, *ip2;
    488 	int i;
    489 
    490 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
    491 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
    492 	if (ru->ru_maxrss < ru2->ru_maxrss)
    493 		ru->ru_maxrss = ru2->ru_maxrss;
    494 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
    495 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
    496 		*ip++ += *ip2++;
    497 }
    498 
    499 /*
    500  * Make a copy of the plimit structure.
    501  * We share these structures copy-on-write after fork,
    502  * and copy when a limit is changed.
    503  */
    504 struct plimit *
    505 limcopy(lim)
    506 	struct plimit *lim;
    507 {
    508 	struct plimit *newlim;
    509 
    510 	newlim = pool_get(&plimit_pool, PR_WAITOK);
    511 	memcpy(newlim->pl_rlimit, lim->pl_rlimit,
    512 	    sizeof(struct rlimit) * RLIM_NLIMITS);
    513 	if (lim->pl_corename == defcorename) {
    514 		newlim->pl_corename = defcorename;
    515 	} else {
    516 		newlim->pl_corename = malloc(strlen(lim->pl_corename)+1,
    517 		    M_TEMP, M_WAITOK);
    518 		strcpy(newlim->pl_corename, lim->pl_corename);
    519 	}
    520 	newlim->p_lflags = 0;
    521 	newlim->p_refcnt = 1;
    522 	return (newlim);
    523 }
    524 
    525 void
    526 limfree(lim)
    527 	struct plimit *lim;
    528 {
    529 
    530 	if (--lim->p_refcnt > 0)
    531 		return;
    532 #ifdef DIAGNOSTIC
    533 	if (lim->p_refcnt < 0)
    534 		panic("limfree");
    535 #endif
    536 	if (lim->pl_corename != defcorename)
    537 		free(lim->pl_corename, M_TEMP);
    538 	pool_put(&plimit_pool, lim);
    539 }
    540 
    541 struct pstats *
    542 pstatscopy(ps)
    543 	struct pstats *ps;
    544 {
    545 
    546 	struct pstats *newps;
    547 
    548 	newps = pool_get(&pstats_pool, PR_WAITOK);
    549 
    550 	memset(&newps->pstat_startzero, 0,
    551 	(unsigned) ((caddr_t)&newps->pstat_endzero -
    552 		    (caddr_t)&newps->pstat_startzero));
    553 	memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy,
    554 	((caddr_t)&newps->pstat_endcopy -
    555 	 (caddr_t)&newps->pstat_startcopy));
    556 
    557 	return (newps);
    558 
    559 }
    560 
    561 void
    562 pstatsfree(ps)
    563 	struct pstats *ps;
    564 {
    565 
    566 	pool_put(&pstats_pool, ps);
    567 }
    568