Home | History | Annotate | Line # | Download | only in kern
kern_resource.c revision 1.80
      1 /*	$NetBSD: kern_resource.c,v 1.80 2004/04/23 02:13:29 yamt 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. Neither the name of the University nor the names of its contributors
     21  *    may be used to endorse or promote products derived from this software
     22  *    without specific prior written permission.
     23  *
     24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     34  * SUCH DAMAGE.
     35  *
     36  *	@(#)kern_resource.c	8.8 (Berkeley) 2/14/95
     37  */
     38 
     39 #include <sys/cdefs.h>
     40 __KERNEL_RCSID(0, "$NetBSD: kern_resource.c,v 1.80 2004/04/23 02:13:29 yamt Exp $");
     41 
     42 #include <sys/param.h>
     43 #include <sys/systm.h>
     44 #include <sys/kernel.h>
     45 #include <sys/file.h>
     46 #include <sys/resourcevar.h>
     47 #include <sys/malloc.h>
     48 #include <sys/pool.h>
     49 #include <sys/proc.h>
     50 #include <sys/sysctl.h>
     51 
     52 #include <sys/mount.h>
     53 #include <sys/sa.h>
     54 #include <sys/syscallargs.h>
     55 
     56 #include <uvm/uvm_extern.h>
     57 
     58 /*
     59  * Maximum process data and stack limits.
     60  * They are variables so they are patchable.
     61  */
     62 rlim_t maxdmap = MAXDSIZ;
     63 rlim_t maxsmap = MAXSSIZ;
     64 
     65 static struct uidinfo *getuidinfo(uid_t);
     66 static void freeuidinfo(struct uidinfo *);
     67 static struct uidinfo *allocuidinfo(uid_t);
     68 
     69 /*
     70  * Resource controls and accounting.
     71  */
     72 
     73 int
     74 sys_getpriority(l, v, retval)
     75 	struct lwp *l;
     76 	void *v;
     77 	register_t *retval;
     78 {
     79 	struct sys_getpriority_args /* {
     80 		syscallarg(int) which;
     81 		syscallarg(int) who;
     82 	} */ *uap = v;
     83 	struct proc *curp = l->l_proc, *p;
     84 	int low = NZERO + PRIO_MAX + 1;
     85 
     86 	switch (SCARG(uap, which)) {
     87 
     88 	case PRIO_PROCESS:
     89 		if (SCARG(uap, who) == 0)
     90 			p = curp;
     91 		else
     92 			p = pfind(SCARG(uap, who));
     93 		if (p == 0)
     94 			break;
     95 		low = p->p_nice;
     96 		break;
     97 
     98 	case PRIO_PGRP: {
     99 		struct pgrp *pg;
    100 
    101 		if (SCARG(uap, who) == 0)
    102 			pg = curp->p_pgrp;
    103 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
    104 			break;
    105 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
    106 			if (p->p_nice < low)
    107 				low = p->p_nice;
    108 		}
    109 		break;
    110 	}
    111 
    112 	case PRIO_USER:
    113 		if (SCARG(uap, who) == 0)
    114 			SCARG(uap, who) = curp->p_ucred->cr_uid;
    115 		proclist_lock_read();
    116 		LIST_FOREACH(p, &allproc, p_list) {
    117 			if (p->p_ucred->cr_uid == (uid_t) SCARG(uap, who) &&
    118 			    p->p_nice < low)
    119 				low = p->p_nice;
    120 		}
    121 		proclist_unlock_read();
    122 		break;
    123 
    124 	default:
    125 		return (EINVAL);
    126 	}
    127 	if (low == NZERO + PRIO_MAX + 1)
    128 		return (ESRCH);
    129 	*retval = low - NZERO;
    130 	return (0);
    131 }
    132 
    133 /* ARGSUSED */
    134 int
    135 sys_setpriority(l, v, retval)
    136 	struct lwp *l;
    137 	void *v;
    138 	register_t *retval;
    139 {
    140 	struct sys_setpriority_args /* {
    141 		syscallarg(int) which;
    142 		syscallarg(int) who;
    143 		syscallarg(int) prio;
    144 	} */ *uap = v;
    145 	struct proc *curp = l->l_proc, *p;
    146 	int found = 0, error = 0;
    147 
    148 	switch (SCARG(uap, which)) {
    149 
    150 	case PRIO_PROCESS:
    151 		if (SCARG(uap, who) == 0)
    152 			p = curp;
    153 		else
    154 			p = pfind(SCARG(uap, who));
    155 		if (p == 0)
    156 			break;
    157 		error = donice(curp, p, SCARG(uap, prio));
    158 		found++;
    159 		break;
    160 
    161 	case PRIO_PGRP: {
    162 		struct pgrp *pg;
    163 
    164 		if (SCARG(uap, who) == 0)
    165 			pg = curp->p_pgrp;
    166 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
    167 			break;
    168 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
    169 			error = donice(curp, p, SCARG(uap, prio));
    170 			found++;
    171 		}
    172 		break;
    173 	}
    174 
    175 	case PRIO_USER:
    176 		if (SCARG(uap, who) == 0)
    177 			SCARG(uap, who) = curp->p_ucred->cr_uid;
    178 		proclist_lock_read();
    179 		LIST_FOREACH(p, &allproc, p_list) {
    180 			if (p->p_ucred->cr_uid == (uid_t) SCARG(uap, who)) {
    181 				error = donice(curp, p, SCARG(uap, prio));
    182 				found++;
    183 			}
    184 		}
    185 		proclist_unlock_read();
    186 		break;
    187 
    188 	default:
    189 		return (EINVAL);
    190 	}
    191 	if (found == 0)
    192 		return (ESRCH);
    193 	return (error);
    194 }
    195 
    196 int
    197 donice(curp, chgp, n)
    198 	struct proc *curp, *chgp;
    199 	int n;
    200 {
    201 	struct pcred *pcred = curp->p_cred;
    202 	int s;
    203 
    204 	if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
    205 	    pcred->pc_ucred->cr_uid != chgp->p_ucred->cr_uid &&
    206 	    pcred->p_ruid != chgp->p_ucred->cr_uid)
    207 		return (EPERM);
    208 	if (n > PRIO_MAX)
    209 		n = PRIO_MAX;
    210 	if (n < PRIO_MIN)
    211 		n = PRIO_MIN;
    212 	n += NZERO;
    213 	if (n < chgp->p_nice && suser(pcred->pc_ucred, &curp->p_acflag))
    214 		return (EACCES);
    215 	chgp->p_nice = n;
    216 	SCHED_LOCK(s);
    217 	(void)resetprocpriority(chgp);
    218 	SCHED_UNLOCK(s);
    219 	return (0);
    220 }
    221 
    222 /* ARGSUSED */
    223 int
    224 sys_setrlimit(l, v, retval)
    225 	struct lwp *l;
    226 	void *v;
    227 	register_t *retval;
    228 {
    229 	struct sys_setrlimit_args /* {
    230 		syscallarg(int) which;
    231 		syscallarg(const struct rlimit *) rlp;
    232 	} */ *uap = v;
    233 	struct proc *p = l->l_proc;
    234 	int which = SCARG(uap, which);
    235 	struct rlimit alim;
    236 	int error;
    237 
    238 	error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit));
    239 	if (error)
    240 		return (error);
    241 	return (dosetrlimit(p, p->p_cred, which, &alim));
    242 }
    243 
    244 int
    245 dosetrlimit(p, cred, which, limp)
    246 	struct proc *p;
    247 	struct  pcred *cred;
    248 	int which;
    249 	struct rlimit *limp;
    250 {
    251 	struct rlimit *alimp;
    252 	struct plimit *newplim;
    253 	int error;
    254 
    255 	if ((u_int)which >= RLIM_NLIMITS)
    256 		return (EINVAL);
    257 
    258 	if (limp->rlim_cur < 0 || limp->rlim_max < 0)
    259 		return (EINVAL);
    260 
    261 	alimp = &p->p_rlimit[which];
    262 	/* if we don't change the value, no need to limcopy() */
    263 	if (limp->rlim_cur == alimp->rlim_cur &&
    264 	    limp->rlim_max == alimp->rlim_max)
    265 		return 0;
    266 
    267 	if (limp->rlim_cur > limp->rlim_max) {
    268 		/*
    269 		 * This is programming error. According to SUSv2, we should
    270 		 * return error in this case.
    271 		 */
    272 		return (EINVAL);
    273 	}
    274 	if (limp->rlim_max > alimp->rlim_max
    275 	    && (error = suser(cred->pc_ucred, &p->p_acflag)) != 0)
    276 			return (error);
    277 
    278 	if (p->p_limit->p_refcnt > 1 &&
    279 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
    280 		newplim = limcopy(p->p_limit);
    281 		limfree(p->p_limit);
    282 		p->p_limit = newplim;
    283 		alimp = &p->p_rlimit[which];
    284 	}
    285 
    286 	switch (which) {
    287 
    288 	case RLIMIT_DATA:
    289 		if (limp->rlim_cur > maxdmap)
    290 			limp->rlim_cur = maxdmap;
    291 		if (limp->rlim_max > maxdmap)
    292 			limp->rlim_max = maxdmap;
    293 		break;
    294 
    295 	case RLIMIT_STACK:
    296 		if (limp->rlim_cur > maxsmap)
    297 			limp->rlim_cur = maxsmap;
    298 		if (limp->rlim_max > maxsmap)
    299 			limp->rlim_max = maxsmap;
    300 
    301 		/*
    302 		 * Return EINVAL if the new stack size limit is lower than
    303 		 * current usage. Otherwise, the process would get SIGSEGV the
    304 		 * moment it would try to access anything on it's current stack.
    305 		 * This conforms to SUSv2.
    306 		 */
    307 		if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE
    308 		    || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE)
    309 			return (EINVAL);
    310 
    311 		/*
    312 		 * Stack is allocated to the max at exec time with
    313 		 * only "rlim_cur" bytes accessible (In other words,
    314 		 * allocates stack dividing two contiguous regions at
    315 		 * "rlim_cur" bytes boundary).
    316 		 *
    317 		 * Since allocation is done in terms of page, roundup
    318 		 * "rlim_cur" (otherwise, contiguous regions
    319 		 * overlap).  If stack limit is going up make more
    320 		 * accessible, if going down make inaccessible.
    321 		 */
    322 		limp->rlim_cur = round_page(limp->rlim_cur);
    323 		if (limp->rlim_cur != alimp->rlim_cur) {
    324 			vaddr_t addr;
    325 			vsize_t size;
    326 			vm_prot_t prot;
    327 
    328 			if (limp->rlim_cur > alimp->rlim_cur) {
    329 				prot = VM_PROT_READ | VM_PROT_WRITE;
    330 				size = limp->rlim_cur - alimp->rlim_cur;
    331 				addr = USRSTACK - limp->rlim_cur;
    332 			} else {
    333 				prot = VM_PROT_NONE;
    334 				size = alimp->rlim_cur - limp->rlim_cur;
    335 				addr = USRSTACK - alimp->rlim_cur;
    336 			}
    337 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
    338 					      addr, addr+size, prot, FALSE);
    339 		}
    340 		break;
    341 
    342 	case RLIMIT_NOFILE:
    343 		if (limp->rlim_cur > maxfiles)
    344 			limp->rlim_cur = maxfiles;
    345 		if (limp->rlim_max > maxfiles)
    346 			limp->rlim_max = maxfiles;
    347 		break;
    348 
    349 	case RLIMIT_NPROC:
    350 		if (limp->rlim_cur > maxproc)
    351 			limp->rlim_cur = maxproc;
    352 		if (limp->rlim_max > maxproc)
    353 			limp->rlim_max = maxproc;
    354 		break;
    355 	}
    356 	*alimp = *limp;
    357 	return (0);
    358 }
    359 
    360 /* ARGSUSED */
    361 int
    362 sys_getrlimit(l, v, retval)
    363 	struct lwp *l;
    364 	void *v;
    365 	register_t *retval;
    366 {
    367 	struct sys_getrlimit_args /* {
    368 		syscallarg(int) which;
    369 		syscallarg(struct rlimit *) rlp;
    370 	} */ *uap = v;
    371 	struct proc *p = l->l_proc;
    372 	int which = SCARG(uap, which);
    373 
    374 	if ((u_int)which >= RLIM_NLIMITS)
    375 		return (EINVAL);
    376 	return (copyout(&p->p_rlimit[which], SCARG(uap, rlp),
    377 	    sizeof(struct rlimit)));
    378 }
    379 
    380 /*
    381  * Transform the running time and tick information in proc p into user,
    382  * system, and interrupt time usage.
    383  */
    384 void
    385 calcru(p, up, sp, ip)
    386 	struct proc *p;
    387 	struct timeval *up;
    388 	struct timeval *sp;
    389 	struct timeval *ip;
    390 {
    391 	u_quad_t u, st, ut, it, tot;
    392 	unsigned long sec;
    393 	long usec;
    394 	int s;
    395 	struct timeval tv;
    396 	struct lwp *l;
    397 
    398 	s = splstatclock();
    399 	st = p->p_sticks;
    400 	ut = p->p_uticks;
    401 	it = p->p_iticks;
    402 	splx(s);
    403 
    404 	sec = p->p_rtime.tv_sec;
    405 	usec = p->p_rtime.tv_usec;
    406 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
    407 		if (l->l_stat == LSONPROC) {
    408 			struct schedstate_percpu *spc;
    409 
    410 			KDASSERT(l->l_cpu != NULL);
    411 			spc = &l->l_cpu->ci_schedstate;
    412 
    413 			/*
    414 			 * Adjust for the current time slice.  This is
    415 			 * actually fairly important since the error
    416 			 * here is on the order of a time quantum,
    417 			 * which is much greater than the sampling
    418 			 * error.
    419 			 */
    420 			microtime(&tv);
    421 			sec += tv.tv_sec - spc->spc_runtime.tv_sec;
    422 			usec += tv.tv_usec - spc->spc_runtime.tv_usec;
    423 		}
    424 	}
    425 
    426 	tot = st + ut + it;
    427 	u = sec * 1000000ull + usec;
    428 
    429 	if (tot == 0) {
    430 		/* No ticks, so can't use to share time out, split 50-50 */
    431 		st = ut = u / 2;
    432 	} else {
    433 		st = (u * st) / tot;
    434 		ut = (u * ut) / tot;
    435 	}
    436 	sp->tv_sec = st / 1000000;
    437 	sp->tv_usec = st % 1000000;
    438 	up->tv_sec = ut / 1000000;
    439 	up->tv_usec = ut % 1000000;
    440 	if (ip != NULL) {
    441 		if (it != 0)
    442 			it = (u * it) / tot;
    443 		ip->tv_sec = it / 1000000;
    444 		ip->tv_usec = it % 1000000;
    445 	}
    446 }
    447 
    448 /* ARGSUSED */
    449 int
    450 sys_getrusage(l, v, retval)
    451 	struct lwp *l;
    452 	void *v;
    453 	register_t *retval;
    454 {
    455 	struct sys_getrusage_args /* {
    456 		syscallarg(int) who;
    457 		syscallarg(struct rusage *) rusage;
    458 	} */ *uap = v;
    459 	struct rusage *rup;
    460 	struct proc *p = l->l_proc;
    461 
    462 	switch (SCARG(uap, who)) {
    463 
    464 	case RUSAGE_SELF:
    465 		rup = &p->p_stats->p_ru;
    466 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
    467 		break;
    468 
    469 	case RUSAGE_CHILDREN:
    470 		rup = &p->p_stats->p_cru;
    471 		break;
    472 
    473 	default:
    474 		return (EINVAL);
    475 	}
    476 	return (copyout(rup, SCARG(uap, rusage), sizeof(struct rusage)));
    477 }
    478 
    479 void
    480 ruadd(ru, ru2)
    481 	struct rusage *ru, *ru2;
    482 {
    483 	long *ip, *ip2;
    484 	int i;
    485 
    486 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
    487 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
    488 	if (ru->ru_maxrss < ru2->ru_maxrss)
    489 		ru->ru_maxrss = ru2->ru_maxrss;
    490 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
    491 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
    492 		*ip++ += *ip2++;
    493 }
    494 
    495 /*
    496  * Make a copy of the plimit structure.
    497  * We share these structures copy-on-write after fork,
    498  * and copy when a limit is changed.
    499  */
    500 struct plimit *
    501 limcopy(lim)
    502 	struct plimit *lim;
    503 {
    504 	struct plimit *newlim;
    505 	size_t l;
    506 
    507 	newlim = pool_get(&plimit_pool, PR_WAITOK);
    508 	memcpy(newlim->pl_rlimit, lim->pl_rlimit,
    509 	    sizeof(struct rlimit) * RLIM_NLIMITS);
    510 	if (lim->pl_corename == defcorename) {
    511 		newlim->pl_corename = defcorename;
    512 	} else {
    513 		l = strlen(lim->pl_corename) + 1;
    514 		newlim->pl_corename = malloc(l, M_TEMP, M_WAITOK);
    515 		strlcpy(newlim->pl_corename, lim->pl_corename, l);
    516 	}
    517 	newlim->p_lflags = 0;
    518 	newlim->p_refcnt = 1;
    519 	return (newlim);
    520 }
    521 
    522 void
    523 limfree(lim)
    524 	struct plimit *lim;
    525 {
    526 
    527 	if (--lim->p_refcnt > 0)
    528 		return;
    529 #ifdef DIAGNOSTIC
    530 	if (lim->p_refcnt < 0)
    531 		panic("limfree");
    532 #endif
    533 	if (lim->pl_corename != defcorename)
    534 		free(lim->pl_corename, M_TEMP);
    535 	pool_put(&plimit_pool, lim);
    536 }
    537 
    538 struct pstats *
    539 pstatscopy(ps)
    540 	struct pstats *ps;
    541 {
    542 
    543 	struct pstats *newps;
    544 
    545 	newps = pool_get(&pstats_pool, PR_WAITOK);
    546 
    547 	memset(&newps->pstat_startzero, 0,
    548 	(unsigned) ((caddr_t)&newps->pstat_endzero -
    549 		    (caddr_t)&newps->pstat_startzero));
    550 	memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy,
    551 	((caddr_t)&newps->pstat_endcopy -
    552 	 (caddr_t)&newps->pstat_startcopy));
    553 
    554 	return (newps);
    555 
    556 }
    557 
    558 void
    559 pstatsfree(ps)
    560 	struct pstats *ps;
    561 {
    562 
    563 	pool_put(&pstats_pool, ps);
    564 }
    565 
    566 /*
    567  * sysctl interface in five parts
    568  */
    569 
    570 /*
    571  * a routine for sysctl proc subtree helpers that need to pick a valid
    572  * process by pid.
    573  */
    574 static int
    575 sysctl_proc_findproc(struct proc *p, struct proc **p2, pid_t pid)
    576 {
    577 	struct proc *ptmp;
    578 	int i, error = 0;
    579 
    580 	if (pid == PROC_CURPROC)
    581 		ptmp = p;
    582 	else if ((ptmp = pfind(pid)) == NULL)
    583 		error = ESRCH;
    584 	else {
    585 		/*
    586 		 * suid proc of ours or proc not ours
    587 		 */
    588 		if (p->p_cred->p_ruid != ptmp->p_cred->p_ruid ||
    589 		    p->p_cred->p_ruid != ptmp->p_cred->p_svuid)
    590 			error = suser(p->p_ucred, &p->p_acflag);
    591 
    592 		/*
    593 		 * sgid proc has sgid back to us temporarily
    594 		 */
    595 		else if (ptmp->p_cred->p_rgid != ptmp->p_cred->p_svgid)
    596 			error = suser(p->p_ucred, &p->p_acflag);
    597 
    598 		/*
    599 		 * our rgid must be in target's group list (ie,
    600 		 * sub-processes started by a sgid process)
    601 		 */
    602 		else {
    603 			for (i = 0; i < p->p_ucred->cr_ngroups; i++) {
    604 				if (p->p_ucred->cr_groups[i] ==
    605 				    ptmp->p_cred->p_rgid)
    606 					break;
    607 			}
    608 			if (i == p->p_ucred->cr_ngroups)
    609 				error = suser(p->p_ucred, &p->p_acflag);
    610 		}
    611 	}
    612 
    613 	*p2 = ptmp;
    614 	return (error);
    615 }
    616 
    617 /*
    618  * sysctl helper routine for setting a process's specific corefile
    619  * name.  picks the process based on the given pid and checks the
    620  * correctness of the new value.
    621  */
    622 static int
    623 sysctl_proc_corename(SYSCTLFN_ARGS)
    624 {
    625 	struct proc *ptmp, *p;
    626 	struct plimit *newplim;
    627 	int error = 0, len;
    628 	char cname[MAXPATHLEN], *tmp;
    629 	struct sysctlnode node;
    630 
    631 	/*
    632 	 * is this all correct?
    633 	 */
    634 	if (namelen != 0)
    635 		return (EINVAL);
    636 	if (name[-1] != PROC_PID_CORENAME)
    637 		return (EINVAL);
    638 
    639 	/*
    640 	 * whom are we tweaking?
    641 	 */
    642 	p = l->l_proc;
    643 	error = sysctl_proc_findproc(p, &ptmp, (pid_t)name[-2]);
    644 	if (error)
    645 		return (error);
    646 
    647 	/*
    648 	 * let them modify a temporary copy of the core name
    649 	 */
    650 	node = *rnode;
    651 	strlcpy(cname, ptmp->p_limit->pl_corename, sizeof(cname));
    652 	node.sysctl_data = cname;
    653 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
    654 
    655 	/*
    656 	 * if that failed, or they have nothing new to say, or we've
    657 	 * heard it before...
    658 	 */
    659 	if (error || newp == NULL ||
    660 	    strcmp(cname, ptmp->p_limit->pl_corename) == 0)
    661 		return (error);
    662 
    663 	/*
    664 	 * no error yet and cname now has the new core name in it.
    665 	 * let's see if it looks acceptable.  it must be either "core"
    666 	 * or end in ".core" or "/core".
    667 	 */
    668 	len = strlen(cname);
    669 	if (len < 4)
    670 		return (EINVAL);
    671 	if (strcmp(cname + len - 4, "core") != 0)
    672 		return (EINVAL);
    673 	if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.')
    674 		return (EINVAL);
    675 
    676 	/*
    677 	 * hmm...looks good.  now...where do we put it?
    678 	 */
    679 	tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL);
    680 	if (tmp == NULL)
    681 		return (ENOMEM);
    682 	strlcpy(tmp, cname, len + 1);
    683 
    684 	if (ptmp->p_limit->p_refcnt > 1 &&
    685 	    (ptmp->p_limit->p_lflags & PL_SHAREMOD) == 0) {
    686 		newplim = limcopy(ptmp->p_limit);
    687 		limfree(ptmp->p_limit);
    688 		ptmp->p_limit = newplim;
    689 	}
    690 	if (ptmp->p_limit->pl_corename != defcorename)
    691 		FREE(ptmp->p_limit->pl_corename, M_SYSCTLDATA);
    692 	ptmp->p_limit->pl_corename = tmp;
    693 
    694 	return (error);
    695 }
    696 
    697 /*
    698  * sysctl helper routine for checking/setting a process's stop flags,
    699  * one for fork and one for exec.
    700  */
    701 static int
    702 sysctl_proc_stop(SYSCTLFN_ARGS)
    703 {
    704 	struct proc *p, *ptmp;
    705 	int i, f, error = 0;
    706 	struct sysctlnode node;
    707 
    708 	if (namelen != 0)
    709 		return (EINVAL);
    710 
    711 	p = l->l_proc;
    712 	error = sysctl_proc_findproc(p, &ptmp, (pid_t)name[-2]);
    713 	if (error)
    714 		return (error);
    715 
    716 	switch (rnode->sysctl_num) {
    717 	case PROC_PID_STOPFORK:
    718 		f = P_STOPFORK;
    719 		break;
    720 	case PROC_PID_STOPEXEC:
    721 		f = P_STOPEXEC;
    722 		break;
    723 	case PROC_PID_STOPEXIT:
    724 		f = P_STOPEXIT;
    725 		break;
    726 	default:
    727 		return (EINVAL);
    728 	}
    729 
    730 	i = (ptmp->p_flag & f) ? 1 : 0;
    731 	node = *rnode;
    732 	node.sysctl_data = &i;
    733 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
    734 	if (error || newp == NULL)
    735 		return (error);
    736 
    737 	if (i)
    738 		ptmp->p_flag |= f;
    739 	else
    740 		ptmp->p_flag &= ~f;
    741 
    742 	return (0);
    743 }
    744 
    745 /*
    746  * sysctl helper routine for a process's rlimits as exposed by sysctl.
    747  */
    748 static int
    749 sysctl_proc_plimit(SYSCTLFN_ARGS)
    750 {
    751 	struct proc *ptmp, *p;
    752 	u_int limitno;
    753 	int which, error = 0;
    754         struct rlimit alim;
    755 	struct sysctlnode node;
    756 
    757 	if (namelen != 0)
    758 		return (EINVAL);
    759 
    760 	which = name[-1];
    761 	if (which != PROC_PID_LIMIT_TYPE_SOFT &&
    762 	    which != PROC_PID_LIMIT_TYPE_HARD)
    763 		return (EINVAL);
    764 
    765 	limitno = name[-2] - 1;
    766 	if (limitno >= RLIM_NLIMITS)
    767 		return (EINVAL);
    768 
    769 	if (name[-3] != PROC_PID_LIMIT)
    770 		return (EINVAL);
    771 
    772 	p = l->l_proc;
    773 	error = sysctl_proc_findproc(p, &ptmp, (pid_t)name[-4]);
    774 	if (error)
    775 		return (error);
    776 
    777 	node = *rnode;
    778 	memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim));
    779 	if (which == PROC_PID_LIMIT_TYPE_HARD)
    780 		node.sysctl_data = &alim.rlim_max;
    781 	else
    782 		node.sysctl_data = &alim.rlim_cur;
    783 
    784 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
    785 	if (error || newp == NULL)
    786 		return (error);
    787 
    788 	return (dosetrlimit(ptmp, p->p_cred, limitno, &alim));
    789 }
    790 
    791 /*
    792  * and finally, the actually glue that sticks it to the tree
    793  */
    794 SYSCTL_SETUP(sysctl_proc_setup, "sysctl proc subtree setup")
    795 {
    796 
    797 	sysctl_createv(clog, 0, NULL, NULL,
    798 		       CTLFLAG_PERMANENT,
    799 		       CTLTYPE_NODE, "proc", NULL,
    800 		       NULL, 0, NULL, 0,
    801 		       CTL_PROC, CTL_EOL);
    802 	sysctl_createv(clog, 0, NULL, NULL,
    803 		       CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER,
    804 		       CTLTYPE_NODE, "curproc",
    805 		       SYSCTL_DESCR("Per-process settings"),
    806 		       NULL, 0, NULL, 0,
    807 		       CTL_PROC, PROC_CURPROC, CTL_EOL);
    808 
    809 	sysctl_createv(clog, 0, NULL, NULL,
    810 		       CTLFLAG_PERMANENT|CTLFLAG_READONLY2|CTLFLAG_ANYWRITE,
    811 		       CTLTYPE_STRING, "corename",
    812 		       SYSCTL_DESCR("Core file name"),
    813 		       sysctl_proc_corename, 0, NULL, MAXPATHLEN,
    814 		       CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL);
    815 	sysctl_createv(clog, 0, NULL, NULL,
    816 		       CTLFLAG_PERMANENT,
    817 		       CTLTYPE_NODE, "rlimit",
    818 		       SYSCTL_DESCR("Process limits"),
    819 		       NULL, 0, NULL, 0,
    820 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL);
    821 
    822 #define create_proc_plimit(s, n) do {					\
    823 	sysctl_createv(clog, 0, NULL, NULL,				\
    824 		       CTLFLAG_PERMANENT,				\
    825 		       CTLTYPE_NODE, s,					\
    826 		       SYSCTL_DESCR("Process " s " limits"),		\
    827 		       NULL, 0, NULL, 0,				\
    828 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
    829 		       CTL_EOL);					\
    830 	sysctl_createv(clog, 0, NULL, NULL,				\
    831 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \
    832 		       CTLTYPE_QUAD, "soft",				\
    833 		       SYSCTL_DESCR("Process soft " s " limit"),	\
    834 		       sysctl_proc_plimit, 0, NULL, 0,			\
    835 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
    836 		       PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL);		\
    837 	sysctl_createv(clog, 0, NULL, NULL,				\
    838 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \
    839 		       CTLTYPE_QUAD, "hard",				\
    840 		       SYSCTL_DESCR("Process hard " s " limit"),	\
    841 		       sysctl_proc_plimit, 0, NULL, 0,			\
    842 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
    843 		       PROC_PID_LIMIT_TYPE_HARD, CTL_EOL);		\
    844 	} while (0/*CONSTCOND*/)
    845 
    846 	create_proc_plimit("cputime",		PROC_PID_LIMIT_CPU);
    847 	create_proc_plimit("filesize",		PROC_PID_LIMIT_FSIZE);
    848 	create_proc_plimit("datasize",		PROC_PID_LIMIT_DATA);
    849 	create_proc_plimit("stacksize",		PROC_PID_LIMIT_STACK);
    850 	create_proc_plimit("coredumpsize",	PROC_PID_LIMIT_CORE);
    851 	create_proc_plimit("memoryuse",		PROC_PID_LIMIT_RSS);
    852 	create_proc_plimit("memorylocked",	PROC_PID_LIMIT_MEMLOCK);
    853 	create_proc_plimit("maxproc",		PROC_PID_LIMIT_NPROC);
    854 	create_proc_plimit("descriptors",	PROC_PID_LIMIT_NOFILE);
    855 	create_proc_plimit("sbsize",		PROC_PID_LIMIT_SBSIZE);
    856 
    857 #undef create_proc_plimit
    858 
    859 	sysctl_createv(clog, 0, NULL, NULL,
    860 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
    861 		       CTLTYPE_INT, "stopfork",
    862 		       SYSCTL_DESCR("Stop process at fork(2)"),
    863 		       sysctl_proc_stop, 0, NULL, 0,
    864 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL);
    865 	sysctl_createv(clog, 0, NULL, NULL,
    866 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
    867 		       CTLTYPE_INT, "stopexec",
    868 		       SYSCTL_DESCR("Stop process at execve(2)"),
    869 		       sysctl_proc_stop, 0, NULL, 0,
    870 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL);
    871 	sysctl_createv(clog, 0, NULL, NULL,
    872 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
    873 		       CTLTYPE_INT, "stopexit",
    874 		       SYSCTL_DESCR("Stop process before completing exit"),
    875 		       sysctl_proc_stop, 0, NULL, 0,
    876 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL);
    877 }
    878 
    879 static struct uidinfo *
    880 getuidinfo(uid_t uid)
    881 {
    882 	struct uidinfo *uip;
    883 	struct uihashhead *uipp;
    884 
    885 	uipp = UIHASH(uid);
    886 
    887 	LIST_FOREACH(uip, uipp, ui_hash)
    888 		if (uip->ui_uid == uid)
    889 			return uip;
    890 	return NULL;
    891 }
    892 
    893 static void
    894 freeuidinfo(struct uidinfo *uip)
    895 {
    896 	LIST_REMOVE(uip, ui_hash);
    897 	FREE(uip, M_PROC);
    898 }
    899 
    900 static struct uidinfo *
    901 allocuidinfo(uid_t uid)
    902 {
    903 	struct uidinfo *uip;
    904 	struct uihashhead *uipp;
    905 
    906 	uipp = UIHASH(uid);
    907 	MALLOC(uip, struct uidinfo *, sizeof(*uip), M_PROC, M_WAITOK);
    908 	LIST_INSERT_HEAD(uipp, uip, ui_hash);
    909 	uip->ui_uid = uid;
    910 	uip->ui_proccnt = 0;
    911 	uip->ui_sbsize = 0;
    912 	return uip;
    913 }
    914 
    915 /*
    916  * Change the count associated with number of processes
    917  * a given user is using.
    918  */
    919 int
    920 chgproccnt(uid_t uid, int diff)
    921 {
    922 	struct uidinfo *uip;
    923 
    924 	if (diff == 0)
    925 		return 0;
    926 
    927 	if ((uip = getuidinfo(uid)) != NULL) {
    928 		uip->ui_proccnt += diff;
    929 		KASSERT(uip->ui_proccnt >= 0);
    930 		if (uip->ui_proccnt > 0)
    931 			return uip->ui_proccnt;
    932 		else {
    933 			if (uip->ui_sbsize == 0)
    934 				freeuidinfo(uip);
    935 			return 0;
    936 		}
    937 	} else {
    938 		if (diff < 0)
    939 			panic("chgproccnt: lost user %lu", (unsigned long)uid);
    940 		uip = allocuidinfo(uid);
    941 		uip->ui_proccnt = diff;
    942 		return uip->ui_proccnt;
    943 	}
    944 }
    945 
    946 int
    947 chgsbsize(uid_t uid, u_long *hiwat, u_long to, rlim_t max)
    948 {
    949 	struct uidinfo *uip;
    950 	rlim_t nsb;
    951 	int rv = 0;
    952 
    953 	if ((uip = getuidinfo(uid)) == NULL)
    954 		uip = allocuidinfo(uid);
    955 	nsb = uip->ui_sbsize + to - *hiwat;
    956 	if (to > *hiwat && nsb > max)
    957 		goto done;
    958 	*hiwat = to;
    959 	uip->ui_sbsize = nsb;
    960 	rv = 1;
    961 	KASSERT(uip->ui_sbsize >= 0);
    962 done:
    963 	if (uip->ui_sbsize == 0 && uip->ui_proccnt == 0)
    964 		freeuidinfo(uip);
    965 	return rv;
    966 }
    967