Home | History | Annotate | Line # | Download | only in kern
kern_resource.c revision 1.87.2.1
      1 /*	$NetBSD: kern_resource.c,v 1.87.2.1 2005/09/18 20:09:50 tron 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 1.87.2.1 2005/09/18 20:09:50 tron 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 struct uihashhead *uihashtbl;
     66 u_long uihash;		/* size of hash table - 1 */
     67 
     68 static struct uidinfo *getuidinfo(uid_t);
     69 static void freeuidinfo(struct uidinfo *);
     70 static struct uidinfo *allocuidinfo(uid_t);
     71 
     72 /*
     73  * Resource controls and accounting.
     74  */
     75 
     76 int
     77 sys_getpriority(l, v, retval)
     78 	struct lwp *l;
     79 	void *v;
     80 	register_t *retval;
     81 {
     82 	struct sys_getpriority_args /* {
     83 		syscallarg(int) which;
     84 		syscallarg(id_t) who;
     85 	} */ *uap = v;
     86 	struct proc *curp = l->l_proc, *p;
     87 	int low = NZERO + PRIO_MAX + 1;
     88 
     89 	switch (SCARG(uap, which)) {
     90 
     91 	case PRIO_PROCESS:
     92 		if (SCARG(uap, who) == 0)
     93 			p = curp;
     94 		else
     95 			p = pfind(SCARG(uap, who));
     96 		if (p == 0)
     97 			break;
     98 		low = p->p_nice;
     99 		break;
    100 
    101 	case PRIO_PGRP: {
    102 		struct pgrp *pg;
    103 
    104 		if (SCARG(uap, who) == 0)
    105 			pg = curp->p_pgrp;
    106 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
    107 			break;
    108 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
    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 		PROCLIST_FOREACH(p, &allproc) {
    120 			if (p->p_ucred->cr_uid == (uid_t) SCARG(uap, who) &&
    121 			    p->p_nice < low)
    122 				low = p->p_nice;
    123 		}
    124 		proclist_unlock_read();
    125 		break;
    126 
    127 	default:
    128 		return (EINVAL);
    129 	}
    130 	if (low == NZERO + PRIO_MAX + 1)
    131 		return (ESRCH);
    132 	*retval = low - NZERO;
    133 	return (0);
    134 }
    135 
    136 /* ARGSUSED */
    137 int
    138 sys_setpriority(l, v, retval)
    139 	struct lwp *l;
    140 	void *v;
    141 	register_t *retval;
    142 {
    143 	struct sys_setpriority_args /* {
    144 		syscallarg(int) which;
    145 		syscallarg(id_t) who;
    146 		syscallarg(int) prio;
    147 	} */ *uap = v;
    148 	struct proc *curp = l->l_proc, *p;
    149 	int found = 0, error = 0;
    150 
    151 	switch (SCARG(uap, which)) {
    152 
    153 	case PRIO_PROCESS:
    154 		if (SCARG(uap, who) == 0)
    155 			p = curp;
    156 		else
    157 			p = pfind(SCARG(uap, who));
    158 		if (p == 0)
    159 			break;
    160 		error = donice(curp, p, SCARG(uap, prio));
    161 		found++;
    162 		break;
    163 
    164 	case PRIO_PGRP: {
    165 		struct pgrp *pg;
    166 
    167 		if (SCARG(uap, who) == 0)
    168 			pg = curp->p_pgrp;
    169 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
    170 			break;
    171 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
    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 		PROCLIST_FOREACH(p, &allproc) {
    183 			if (p->p_ucred->cr_uid == (uid_t) SCARG(uap, who)) {
    184 				error = donice(curp, p, SCARG(uap, prio));
    185 				found++;
    186 			}
    187 		}
    188 		proclist_unlock_read();
    189 		break;
    190 
    191 	default:
    192 		return (EINVAL);
    193 	}
    194 	if (found == 0)
    195 		return (ESRCH);
    196 	return (error);
    197 }
    198 
    199 int
    200 donice(curp, chgp, n)
    201 	struct proc *curp, *chgp;
    202 	int n;
    203 {
    204 	struct pcred *pcred = curp->p_cred;
    205 	int s;
    206 
    207 	if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
    208 	    pcred->pc_ucred->cr_uid != chgp->p_ucred->cr_uid &&
    209 	    pcred->p_ruid != chgp->p_ucred->cr_uid)
    210 		return (EPERM);
    211 	if (n > PRIO_MAX)
    212 		n = PRIO_MAX;
    213 	if (n < PRIO_MIN)
    214 		n = PRIO_MIN;
    215 	n += NZERO;
    216 	if (n < chgp->p_nice && suser(pcred->pc_ucred, &curp->p_acflag))
    217 		return (EACCES);
    218 	chgp->p_nice = n;
    219 	SCHED_LOCK(s);
    220 	(void)resetprocpriority(chgp);
    221 	SCHED_UNLOCK(s);
    222 	return (0);
    223 }
    224 
    225 /* ARGSUSED */
    226 int
    227 sys_setrlimit(l, v, retval)
    228 	struct lwp *l;
    229 	void *v;
    230 	register_t *retval;
    231 {
    232 	struct sys_setrlimit_args /* {
    233 		syscallarg(int) which;
    234 		syscallarg(const struct rlimit *) rlp;
    235 	} */ *uap = v;
    236 	struct proc *p = l->l_proc;
    237 	int which = SCARG(uap, which);
    238 	struct rlimit alim;
    239 	int error;
    240 
    241 	error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit));
    242 	if (error)
    243 		return (error);
    244 	return (dosetrlimit(p, p->p_cred, which, &alim));
    245 }
    246 
    247 int
    248 dosetrlimit(p, cred, which, limp)
    249 	struct proc *p;
    250 	struct  pcred *cred;
    251 	int which;
    252 	struct rlimit *limp;
    253 {
    254 	struct rlimit *alimp;
    255 	struct plimit *oldplim;
    256 	int error;
    257 
    258 	if ((u_int)which >= RLIM_NLIMITS)
    259 		return (EINVAL);
    260 
    261 	if (limp->rlim_cur < 0 || limp->rlim_max < 0)
    262 		return (EINVAL);
    263 
    264 	alimp = &p->p_rlimit[which];
    265 	/* if we don't change the value, no need to limcopy() */
    266 	if (limp->rlim_cur == alimp->rlim_cur &&
    267 	    limp->rlim_max == alimp->rlim_max)
    268 		return 0;
    269 
    270 	if (limp->rlim_cur > limp->rlim_max) {
    271 		/*
    272 		 * This is programming error. According to SUSv2, we should
    273 		 * return error in this case.
    274 		 */
    275 		return (EINVAL);
    276 	}
    277 	if (limp->rlim_max > alimp->rlim_max
    278 	    && (error = suser(cred->pc_ucred, &p->p_acflag)) != 0)
    279 			return (error);
    280 
    281 	if (p->p_limit->p_refcnt > 1 &&
    282 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
    283 		p->p_limit = limcopy(oldplim = p->p_limit);
    284 		limfree(oldplim);
    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_READ | VM_PROT_WRITE;
    332 				size = limp->rlim_cur - alimp->rlim_cur;
    333 				addr = (vaddr_t)p->p_vmspace->vm_minsaddr -
    334 				    limp->rlim_cur;
    335 			} else {
    336 				prot = VM_PROT_NONE;
    337 				size = alimp->rlim_cur - limp->rlim_cur;
    338 				addr = (vaddr_t)p->p_vmspace->vm_minsaddr -
    339 				     alimp->rlim_cur;
    340 			}
    341 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
    342 					      addr, addr+size, prot, FALSE);
    343 		}
    344 		break;
    345 
    346 	case RLIMIT_NOFILE:
    347 		if (limp->rlim_cur > maxfiles)
    348 			limp->rlim_cur = maxfiles;
    349 		if (limp->rlim_max > maxfiles)
    350 			limp->rlim_max = maxfiles;
    351 		break;
    352 
    353 	case RLIMIT_NPROC:
    354 		if (limp->rlim_cur > maxproc)
    355 			limp->rlim_cur = maxproc;
    356 		if (limp->rlim_max > maxproc)
    357 			limp->rlim_max = maxproc;
    358 		break;
    359 	}
    360 	*alimp = *limp;
    361 	return (0);
    362 }
    363 
    364 /* ARGSUSED */
    365 int
    366 sys_getrlimit(l, v, retval)
    367 	struct lwp *l;
    368 	void *v;
    369 	register_t *retval;
    370 {
    371 	struct sys_getrlimit_args /* {
    372 		syscallarg(int) which;
    373 		syscallarg(struct rlimit *) rlp;
    374 	} */ *uap = v;
    375 	struct proc *p = l->l_proc;
    376 	int which = SCARG(uap, which);
    377 
    378 	if ((u_int)which >= RLIM_NLIMITS)
    379 		return (EINVAL);
    380 	return (copyout(&p->p_rlimit[which], SCARG(uap, rlp),
    381 	    sizeof(struct rlimit)));
    382 }
    383 
    384 /*
    385  * Transform the running time and tick information in proc p into user,
    386  * system, and interrupt time usage.
    387  */
    388 void
    389 calcru(p, up, sp, ip)
    390 	struct proc *p;
    391 	struct timeval *up;
    392 	struct timeval *sp;
    393 	struct timeval *ip;
    394 {
    395 	u_quad_t u, st, ut, it, tot;
    396 	unsigned long sec;
    397 	long usec;
    398 	int s;
    399 	struct timeval tv;
    400 	struct lwp *l;
    401 
    402 	s = splstatclock();
    403 	st = p->p_sticks;
    404 	ut = p->p_uticks;
    405 	it = p->p_iticks;
    406 	splx(s);
    407 
    408 	sec = p->p_rtime.tv_sec;
    409 	usec = p->p_rtime.tv_usec;
    410 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
    411 		if (l->l_stat == LSONPROC) {
    412 			struct schedstate_percpu *spc;
    413 
    414 			KDASSERT(l->l_cpu != NULL);
    415 			spc = &l->l_cpu->ci_schedstate;
    416 
    417 			/*
    418 			 * Adjust for the current time slice.  This is
    419 			 * actually fairly important since the error
    420 			 * here is on the order of a time quantum,
    421 			 * which is much greater than the sampling
    422 			 * error.
    423 			 */
    424 			microtime(&tv);
    425 			sec += tv.tv_sec - spc->spc_runtime.tv_sec;
    426 			usec += tv.tv_usec - spc->spc_runtime.tv_usec;
    427 		}
    428 	}
    429 
    430 	tot = st + ut + it;
    431 	u = sec * 1000000ull + usec;
    432 
    433 	if (tot == 0) {
    434 		/* No ticks, so can't use to share time out, split 50-50 */
    435 		st = ut = u / 2;
    436 	} else {
    437 		st = (u * st) / tot;
    438 		ut = (u * ut) / tot;
    439 	}
    440 	sp->tv_sec = st / 1000000;
    441 	sp->tv_usec = st % 1000000;
    442 	up->tv_sec = ut / 1000000;
    443 	up->tv_usec = ut % 1000000;
    444 	if (ip != NULL) {
    445 		if (it != 0)
    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 	size_t l = 0;
    510 
    511 	simple_lock(&lim->p_slock);
    512 	if (lim->pl_corename != defcorename)
    513 		l = strlen(lim->pl_corename) + 1;
    514 	simple_unlock(&lim->p_slock);
    515 
    516 	newlim = pool_get(&plimit_pool, PR_WAITOK);
    517 	simple_lock_init(&newlim->p_slock);
    518 	newlim->p_lflags = 0;
    519 	newlim->p_refcnt = 1;
    520 	newlim->pl_corename = (l != 0)
    521 		? malloc(l, M_TEMP, M_WAITOK)
    522 		: defcorename;
    523 
    524 	simple_lock(&lim->p_slock);
    525 	memcpy(newlim->pl_rlimit, lim->pl_rlimit,
    526 	    sizeof(struct rlimit) * RLIM_NLIMITS);
    527 
    528 	if (l != 0)
    529 		strlcpy(newlim->pl_corename, lim->pl_corename, l);
    530 	simple_unlock(&lim->p_slock);
    531 
    532 	return (newlim);
    533 }
    534 
    535 void
    536 limfree(lim)
    537 	struct plimit *lim;
    538 {
    539 	int n;
    540 
    541 	simple_lock(&lim->p_slock);
    542 	n = --lim->p_refcnt;
    543 	simple_unlock(&lim->p_slock);
    544 	if (n > 0)
    545 		return;
    546 #ifdef DIAGNOSTIC
    547 	if (n < 0)
    548 		panic("limfree");
    549 #endif
    550 	if (lim->pl_corename != defcorename)
    551 		free(lim->pl_corename, M_TEMP);
    552 	pool_put(&plimit_pool, lim);
    553 }
    554 
    555 struct pstats *
    556 pstatscopy(ps)
    557 	struct pstats *ps;
    558 {
    559 
    560 	struct pstats *newps;
    561 
    562 	newps = pool_get(&pstats_pool, PR_WAITOK);
    563 
    564 	memset(&newps->pstat_startzero, 0,
    565 	(unsigned) ((caddr_t)&newps->pstat_endzero -
    566 		    (caddr_t)&newps->pstat_startzero));
    567 	memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy,
    568 	((caddr_t)&newps->pstat_endcopy -
    569 	 (caddr_t)&newps->pstat_startcopy));
    570 
    571 	return (newps);
    572 
    573 }
    574 
    575 void
    576 pstatsfree(ps)
    577 	struct pstats *ps;
    578 {
    579 
    580 	pool_put(&pstats_pool, ps);
    581 }
    582 
    583 /*
    584  * sysctl interface in five parts
    585  */
    586 
    587 /*
    588  * a routine for sysctl proc subtree helpers that need to pick a valid
    589  * process by pid.
    590  */
    591 static int
    592 sysctl_proc_findproc(struct proc *p, struct proc **p2, pid_t pid)
    593 {
    594 	struct proc *ptmp;
    595 	int i, error = 0;
    596 
    597 	if (pid == PROC_CURPROC)
    598 		ptmp = p;
    599 	else if ((ptmp = pfind(pid)) == NULL)
    600 		error = ESRCH;
    601 	else {
    602 		/*
    603 		 * suid proc of ours or proc not ours
    604 		 */
    605 		if (p->p_cred->p_ruid != ptmp->p_cred->p_ruid ||
    606 		    p->p_cred->p_ruid != ptmp->p_cred->p_svuid)
    607 			error = suser(p->p_ucred, &p->p_acflag);
    608 
    609 		/*
    610 		 * sgid proc has sgid back to us temporarily
    611 		 */
    612 		else if (ptmp->p_cred->p_rgid != ptmp->p_cred->p_svgid)
    613 			error = suser(p->p_ucred, &p->p_acflag);
    614 
    615 		/*
    616 		 * our rgid must be in target's group list (ie,
    617 		 * sub-processes started by a sgid process)
    618 		 */
    619 		else {
    620 			for (i = 0; i < p->p_ucred->cr_ngroups; i++) {
    621 				if (p->p_ucred->cr_groups[i] ==
    622 				    ptmp->p_cred->p_rgid)
    623 					break;
    624 			}
    625 			if (i == p->p_ucred->cr_ngroups)
    626 				error = suser(p->p_ucred, &p->p_acflag);
    627 		}
    628 	}
    629 
    630 	*p2 = ptmp;
    631 	return (error);
    632 }
    633 
    634 /*
    635  * sysctl helper routine for setting a process's specific corefile
    636  * name.  picks the process based on the given pid and checks the
    637  * correctness of the new value.
    638  */
    639 static int
    640 sysctl_proc_corename(SYSCTLFN_ARGS)
    641 {
    642 	struct proc *ptmp, *p;
    643 	struct plimit *lim;
    644 	int error = 0, len;
    645 	char cname[MAXPATHLEN], *tmp;
    646 	struct sysctlnode node;
    647 
    648 	/*
    649 	 * is this all correct?
    650 	 */
    651 	if (namelen != 0)
    652 		return (EINVAL);
    653 	if (name[-1] != PROC_PID_CORENAME)
    654 		return (EINVAL);
    655 
    656 	/*
    657 	 * whom are we tweaking?
    658 	 */
    659 	p = l->l_proc;
    660 	error = sysctl_proc_findproc(p, &ptmp, (pid_t)name[-2]);
    661 	if (error)
    662 		return (error);
    663 
    664 	/*
    665 	 * let them modify a temporary copy of the core name
    666 	 */
    667 	node = *rnode;
    668 	strlcpy(cname, ptmp->p_limit->pl_corename, sizeof(cname));
    669 	node.sysctl_data = cname;
    670 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
    671 
    672 	/*
    673 	 * if that failed, or they have nothing new to say, or we've
    674 	 * heard it before...
    675 	 */
    676 	if (error || newp == NULL ||
    677 	    strcmp(cname, ptmp->p_limit->pl_corename) == 0)
    678 		return (error);
    679 
    680 	/*
    681 	 * no error yet and cname now has the new core name in it.
    682 	 * let's see if it looks acceptable.  it must be either "core"
    683 	 * or end in ".core" or "/core".
    684 	 */
    685 	len = strlen(cname);
    686 	if (len < 4)
    687 		return (EINVAL);
    688 	if (strcmp(cname + len - 4, "core") != 0)
    689 		return (EINVAL);
    690 	if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.')
    691 		return (EINVAL);
    692 
    693 	/*
    694 	 * hmm...looks good.  now...where do we put it?
    695 	 */
    696 	tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL);
    697 	if (tmp == NULL)
    698 		return (ENOMEM);
    699 	strlcpy(tmp, cname, len + 1);
    700 
    701 	lim = ptmp->p_limit;
    702 	if (lim->p_refcnt > 1 && (lim->p_lflags & PL_SHAREMOD) == 0) {
    703 		ptmp->p_limit = limcopy(lim);
    704 		limfree(lim);
    705 		lim = ptmp->p_limit;
    706 	}
    707 	if (lim->pl_corename != defcorename)
    708 		free(lim->pl_corename, M_TEMP);
    709 	lim->pl_corename = tmp;
    710 
    711 	return (error);
    712 }
    713 
    714 /*
    715  * sysctl helper routine for checking/setting a process's stop flags,
    716  * one for fork and one for exec.
    717  */
    718 static int
    719 sysctl_proc_stop(SYSCTLFN_ARGS)
    720 {
    721 	struct proc *p, *ptmp;
    722 	int i, f, error = 0;
    723 	struct sysctlnode node;
    724 
    725 	if (namelen != 0)
    726 		return (EINVAL);
    727 
    728 	p = l->l_proc;
    729 	error = sysctl_proc_findproc(p, &ptmp, (pid_t)name[-2]);
    730 	if (error)
    731 		return (error);
    732 
    733 	switch (rnode->sysctl_num) {
    734 	case PROC_PID_STOPFORK:
    735 		f = P_STOPFORK;
    736 		break;
    737 	case PROC_PID_STOPEXEC:
    738 		f = P_STOPEXEC;
    739 		break;
    740 	case PROC_PID_STOPEXIT:
    741 		f = P_STOPEXIT;
    742 		break;
    743 	default:
    744 		return (EINVAL);
    745 	}
    746 
    747 	i = (ptmp->p_flag & f) ? 1 : 0;
    748 	node = *rnode;
    749 	node.sysctl_data = &i;
    750 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
    751 	if (error || newp == NULL)
    752 		return (error);
    753 
    754 	if (i)
    755 		ptmp->p_flag |= f;
    756 	else
    757 		ptmp->p_flag &= ~f;
    758 
    759 	return (0);
    760 }
    761 
    762 /*
    763  * sysctl helper routine for a process's rlimits as exposed by sysctl.
    764  */
    765 static int
    766 sysctl_proc_plimit(SYSCTLFN_ARGS)
    767 {
    768 	struct proc *ptmp, *p;
    769 	u_int limitno;
    770 	int which, error = 0;
    771         struct rlimit alim;
    772 	struct sysctlnode node;
    773 
    774 	if (namelen != 0)
    775 		return (EINVAL);
    776 
    777 	which = name[-1];
    778 	if (which != PROC_PID_LIMIT_TYPE_SOFT &&
    779 	    which != PROC_PID_LIMIT_TYPE_HARD)
    780 		return (EINVAL);
    781 
    782 	limitno = name[-2] - 1;
    783 	if (limitno >= RLIM_NLIMITS)
    784 		return (EINVAL);
    785 
    786 	if (name[-3] != PROC_PID_LIMIT)
    787 		return (EINVAL);
    788 
    789 	p = l->l_proc;
    790 	error = sysctl_proc_findproc(p, &ptmp, (pid_t)name[-4]);
    791 	if (error)
    792 		return (error);
    793 
    794 	node = *rnode;
    795 	memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim));
    796 	if (which == PROC_PID_LIMIT_TYPE_HARD)
    797 		node.sysctl_data = &alim.rlim_max;
    798 	else
    799 		node.sysctl_data = &alim.rlim_cur;
    800 
    801 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
    802 	if (error || newp == NULL)
    803 		return (error);
    804 
    805 	return (dosetrlimit(ptmp, p->p_cred, limitno, &alim));
    806 }
    807 
    808 /*
    809  * and finally, the actually glue that sticks it to the tree
    810  */
    811 SYSCTL_SETUP(sysctl_proc_setup, "sysctl proc subtree setup")
    812 {
    813 
    814 	sysctl_createv(clog, 0, NULL, NULL,
    815 		       CTLFLAG_PERMANENT,
    816 		       CTLTYPE_NODE, "proc", NULL,
    817 		       NULL, 0, NULL, 0,
    818 		       CTL_PROC, CTL_EOL);
    819 	sysctl_createv(clog, 0, NULL, NULL,
    820 		       CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER,
    821 		       CTLTYPE_NODE, "curproc",
    822 		       SYSCTL_DESCR("Per-process settings"),
    823 		       NULL, 0, NULL, 0,
    824 		       CTL_PROC, PROC_CURPROC, CTL_EOL);
    825 
    826 	sysctl_createv(clog, 0, NULL, NULL,
    827 		       CTLFLAG_PERMANENT|CTLFLAG_READONLY2|CTLFLAG_ANYWRITE,
    828 		       CTLTYPE_STRING, "corename",
    829 		       SYSCTL_DESCR("Core file name"),
    830 		       sysctl_proc_corename, 0, NULL, MAXPATHLEN,
    831 		       CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL);
    832 	sysctl_createv(clog, 0, NULL, NULL,
    833 		       CTLFLAG_PERMANENT,
    834 		       CTLTYPE_NODE, "rlimit",
    835 		       SYSCTL_DESCR("Process limits"),
    836 		       NULL, 0, NULL, 0,
    837 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL);
    838 
    839 #define create_proc_plimit(s, n) do {					\
    840 	sysctl_createv(clog, 0, NULL, NULL,				\
    841 		       CTLFLAG_PERMANENT,				\
    842 		       CTLTYPE_NODE, s,					\
    843 		       SYSCTL_DESCR("Process " s " limits"),		\
    844 		       NULL, 0, NULL, 0,				\
    845 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
    846 		       CTL_EOL);					\
    847 	sysctl_createv(clog, 0, NULL, NULL,				\
    848 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \
    849 		       CTLTYPE_QUAD, "soft",				\
    850 		       SYSCTL_DESCR("Process soft " s " limit"),	\
    851 		       sysctl_proc_plimit, 0, NULL, 0,			\
    852 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
    853 		       PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL);		\
    854 	sysctl_createv(clog, 0, NULL, NULL,				\
    855 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \
    856 		       CTLTYPE_QUAD, "hard",				\
    857 		       SYSCTL_DESCR("Process hard " s " limit"),	\
    858 		       sysctl_proc_plimit, 0, NULL, 0,			\
    859 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
    860 		       PROC_PID_LIMIT_TYPE_HARD, CTL_EOL);		\
    861 	} while (0/*CONSTCOND*/)
    862 
    863 	create_proc_plimit("cputime",		PROC_PID_LIMIT_CPU);
    864 	create_proc_plimit("filesize",		PROC_PID_LIMIT_FSIZE);
    865 	create_proc_plimit("datasize",		PROC_PID_LIMIT_DATA);
    866 	create_proc_plimit("stacksize",		PROC_PID_LIMIT_STACK);
    867 	create_proc_plimit("coredumpsize",	PROC_PID_LIMIT_CORE);
    868 	create_proc_plimit("memoryuse",		PROC_PID_LIMIT_RSS);
    869 	create_proc_plimit("memorylocked",	PROC_PID_LIMIT_MEMLOCK);
    870 	create_proc_plimit("maxproc",		PROC_PID_LIMIT_NPROC);
    871 	create_proc_plimit("descriptors",	PROC_PID_LIMIT_NOFILE);
    872 	create_proc_plimit("sbsize",		PROC_PID_LIMIT_SBSIZE);
    873 
    874 #undef create_proc_plimit
    875 
    876 	sysctl_createv(clog, 0, NULL, NULL,
    877 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
    878 		       CTLTYPE_INT, "stopfork",
    879 		       SYSCTL_DESCR("Stop process at fork(2)"),
    880 		       sysctl_proc_stop, 0, NULL, 0,
    881 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL);
    882 	sysctl_createv(clog, 0, NULL, NULL,
    883 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
    884 		       CTLTYPE_INT, "stopexec",
    885 		       SYSCTL_DESCR("Stop process at execve(2)"),
    886 		       sysctl_proc_stop, 0, NULL, 0,
    887 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL);
    888 	sysctl_createv(clog, 0, NULL, NULL,
    889 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
    890 		       CTLTYPE_INT, "stopexit",
    891 		       SYSCTL_DESCR("Stop process before completing exit"),
    892 		       sysctl_proc_stop, 0, NULL, 0,
    893 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL);
    894 }
    895 
    896 static struct uidinfo *
    897 getuidinfo(uid_t uid)
    898 {
    899 	struct uidinfo *uip;
    900 	struct uihashhead *uipp;
    901 
    902 	uipp = UIHASH(uid);
    903 
    904 	LIST_FOREACH(uip, uipp, ui_hash)
    905 		if (uip->ui_uid == uid)
    906 			return uip;
    907 	return NULL;
    908 }
    909 
    910 static void
    911 freeuidinfo(struct uidinfo *uip)
    912 {
    913 	LIST_REMOVE(uip, ui_hash);
    914 	FREE(uip, M_PROC);
    915 }
    916 
    917 static struct uidinfo *
    918 allocuidinfo(uid_t uid)
    919 {
    920 	struct uidinfo *uip;
    921 	struct uihashhead *uipp;
    922 
    923 	uipp = UIHASH(uid);
    924 	MALLOC(uip, struct uidinfo *, sizeof(*uip), M_PROC, M_WAITOK);
    925 	LIST_INSERT_HEAD(uipp, uip, ui_hash);
    926 	uip->ui_uid = uid;
    927 	uip->ui_proccnt = 0;
    928 	uip->ui_sbsize = 0;
    929 	return uip;
    930 }
    931 
    932 /*
    933  * Change the count associated with number of processes
    934  * a given user is using.
    935  */
    936 int
    937 chgproccnt(uid_t uid, int diff)
    938 {
    939 	struct uidinfo *uip;
    940 
    941 	if (diff == 0)
    942 		return 0;
    943 
    944 	if ((uip = getuidinfo(uid)) != NULL) {
    945 		uip->ui_proccnt += diff;
    946 		KASSERT(uip->ui_proccnt >= 0);
    947 		if (uip->ui_proccnt > 0)
    948 			return uip->ui_proccnt;
    949 		else {
    950 			if (uip->ui_sbsize == 0)
    951 				freeuidinfo(uip);
    952 			return 0;
    953 		}
    954 	} else {
    955 		if (diff < 0)
    956 			panic("chgproccnt: lost user %lu", (unsigned long)uid);
    957 		uip = allocuidinfo(uid);
    958 		uip->ui_proccnt = diff;
    959 		return uip->ui_proccnt;
    960 	}
    961 }
    962 
    963 int
    964 chgsbsize(uid_t uid, u_long *hiwat, u_long to, rlim_t max)
    965 {
    966 	*hiwat = to;
    967 	return 1;
    968 #ifdef notyet
    969 	struct uidinfo *uip;
    970 	rlim_t nsb;
    971 	int rv = 0;
    972 
    973 	if ((uip = getuidinfo(uid)) == NULL)
    974 		uip = allocuidinfo(uid);
    975 	nsb = uip->ui_sbsize + to - *hiwat;
    976 	if (to > *hiwat && nsb > max)
    977 		goto done;
    978 	*hiwat = to;
    979 	uip->ui_sbsize = nsb;
    980 	rv = 1;
    981 	KASSERT(uip->ui_sbsize >= 0);
    982 done:
    983 	if (uip->ui_sbsize == 0 && uip->ui_proccnt == 0)
    984 		freeuidinfo(uip);
    985 	return rv;
    986 #endif
    987 }
    988