Home | History | Annotate | Line # | Download | only in kern
kern_resource.c revision 1.78
      1 /*	$NetBSD: kern_resource.c,v 1.78 2004/04/08 06:20:30 atatat 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 1.78 2004/04/08 06:20:30 atatat 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 /*
     66  * Resource controls and accounting.
     67  */
     68 
     69 int
     70 sys_getpriority(l, v, retval)
     71 	struct lwp *l;
     72 	void *v;
     73 	register_t *retval;
     74 {
     75 	struct sys_getpriority_args /* {
     76 		syscallarg(int) which;
     77 		syscallarg(int) who;
     78 	} */ *uap = v;
     79 	struct proc *curp = l->l_proc, *p;
     80 	int low = NZERO + PRIO_MAX + 1;
     81 
     82 	switch (SCARG(uap, which)) {
     83 
     84 	case PRIO_PROCESS:
     85 		if (SCARG(uap, who) == 0)
     86 			p = curp;
     87 		else
     88 			p = pfind(SCARG(uap, who));
     89 		if (p == 0)
     90 			break;
     91 		low = p->p_nice;
     92 		break;
     93 
     94 	case PRIO_PGRP: {
     95 		struct pgrp *pg;
     96 
     97 		if (SCARG(uap, who) == 0)
     98 			pg = curp->p_pgrp;
     99 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
    100 			break;
    101 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
    102 			if (p->p_nice < low)
    103 				low = p->p_nice;
    104 		}
    105 		break;
    106 	}
    107 
    108 	case PRIO_USER:
    109 		if (SCARG(uap, who) == 0)
    110 			SCARG(uap, who) = curp->p_ucred->cr_uid;
    111 		proclist_lock_read();
    112 		LIST_FOREACH(p, &allproc, p_list) {
    113 			if (p->p_ucred->cr_uid == (uid_t) SCARG(uap, who) &&
    114 			    p->p_nice < low)
    115 				low = p->p_nice;
    116 		}
    117 		proclist_unlock_read();
    118 		break;
    119 
    120 	default:
    121 		return (EINVAL);
    122 	}
    123 	if (low == NZERO + PRIO_MAX + 1)
    124 		return (ESRCH);
    125 	*retval = low - NZERO;
    126 	return (0);
    127 }
    128 
    129 /* ARGSUSED */
    130 int
    131 sys_setpriority(l, v, retval)
    132 	struct lwp *l;
    133 	void *v;
    134 	register_t *retval;
    135 {
    136 	struct sys_setpriority_args /* {
    137 		syscallarg(int) which;
    138 		syscallarg(int) who;
    139 		syscallarg(int) prio;
    140 	} */ *uap = v;
    141 	struct proc *curp = l->l_proc, *p;
    142 	int found = 0, error = 0;
    143 
    144 	switch (SCARG(uap, which)) {
    145 
    146 	case PRIO_PROCESS:
    147 		if (SCARG(uap, who) == 0)
    148 			p = curp;
    149 		else
    150 			p = pfind(SCARG(uap, who));
    151 		if (p == 0)
    152 			break;
    153 		error = donice(curp, p, SCARG(uap, prio));
    154 		found++;
    155 		break;
    156 
    157 	case PRIO_PGRP: {
    158 		struct pgrp *pg;
    159 
    160 		if (SCARG(uap, who) == 0)
    161 			pg = curp->p_pgrp;
    162 		else if ((pg = pgfind(SCARG(uap, who))) == NULL)
    163 			break;
    164 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
    165 			error = donice(curp, p, SCARG(uap, prio));
    166 			found++;
    167 		}
    168 		break;
    169 	}
    170 
    171 	case PRIO_USER:
    172 		if (SCARG(uap, who) == 0)
    173 			SCARG(uap, who) = curp->p_ucred->cr_uid;
    174 		proclist_lock_read();
    175 		LIST_FOREACH(p, &allproc, p_list) {
    176 			if (p->p_ucred->cr_uid == (uid_t) SCARG(uap, who)) {
    177 				error = donice(curp, p, SCARG(uap, prio));
    178 				found++;
    179 			}
    180 		}
    181 		proclist_unlock_read();
    182 		break;
    183 
    184 	default:
    185 		return (EINVAL);
    186 	}
    187 	if (found == 0)
    188 		return (ESRCH);
    189 	return (error);
    190 }
    191 
    192 int
    193 donice(curp, chgp, n)
    194 	struct proc *curp, *chgp;
    195 	int n;
    196 {
    197 	struct pcred *pcred = curp->p_cred;
    198 	int s;
    199 
    200 	if (pcred->pc_ucred->cr_uid && pcred->p_ruid &&
    201 	    pcred->pc_ucred->cr_uid != chgp->p_ucred->cr_uid &&
    202 	    pcred->p_ruid != chgp->p_ucred->cr_uid)
    203 		return (EPERM);
    204 	if (n > PRIO_MAX)
    205 		n = PRIO_MAX;
    206 	if (n < PRIO_MIN)
    207 		n = PRIO_MIN;
    208 	n += NZERO;
    209 	if (n < chgp->p_nice && suser(pcred->pc_ucred, &curp->p_acflag))
    210 		return (EACCES);
    211 	chgp->p_nice = n;
    212 	SCHED_LOCK(s);
    213 	(void)resetprocpriority(chgp);
    214 	SCHED_UNLOCK(s);
    215 	return (0);
    216 }
    217 
    218 /* ARGSUSED */
    219 int
    220 sys_setrlimit(l, v, retval)
    221 	struct lwp *l;
    222 	void *v;
    223 	register_t *retval;
    224 {
    225 	struct sys_setrlimit_args /* {
    226 		syscallarg(int) which;
    227 		syscallarg(const struct rlimit *) rlp;
    228 	} */ *uap = v;
    229 	struct proc *p = l->l_proc;
    230 	int which = SCARG(uap, which);
    231 	struct rlimit alim;
    232 	int error;
    233 
    234 	error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit));
    235 	if (error)
    236 		return (error);
    237 	return (dosetrlimit(p, p->p_cred, which, &alim));
    238 }
    239 
    240 int
    241 dosetrlimit(p, cred, which, limp)
    242 	struct proc *p;
    243 	struct  pcred *cred;
    244 	int which;
    245 	struct rlimit *limp;
    246 {
    247 	struct rlimit *alimp;
    248 	struct plimit *newplim;
    249 	int error;
    250 
    251 	if ((u_int)which >= RLIM_NLIMITS)
    252 		return (EINVAL);
    253 
    254 	if (limp->rlim_cur < 0 || limp->rlim_max < 0)
    255 		return (EINVAL);
    256 
    257 	alimp = &p->p_rlimit[which];
    258 	/* if we don't change the value, no need to limcopy() */
    259 	if (limp->rlim_cur == alimp->rlim_cur &&
    260 	    limp->rlim_max == alimp->rlim_max)
    261 		return 0;
    262 
    263 	if (limp->rlim_cur > limp->rlim_max) {
    264 		/*
    265 		 * This is programming error. According to SUSv2, we should
    266 		 * return error in this case.
    267 		 */
    268 		return (EINVAL);
    269 	}
    270 	if (limp->rlim_max > alimp->rlim_max
    271 	    && (error = suser(cred->pc_ucred, &p->p_acflag)) != 0)
    272 			return (error);
    273 
    274 	if (p->p_limit->p_refcnt > 1 &&
    275 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
    276 		newplim = limcopy(p->p_limit);
    277 		limfree(p->p_limit);
    278 		p->p_limit = newplim;
    279 		alimp = &p->p_rlimit[which];
    280 	}
    281 
    282 	switch (which) {
    283 
    284 	case RLIMIT_DATA:
    285 		if (limp->rlim_cur > maxdmap)
    286 			limp->rlim_cur = maxdmap;
    287 		if (limp->rlim_max > maxdmap)
    288 			limp->rlim_max = maxdmap;
    289 		break;
    290 
    291 	case RLIMIT_STACK:
    292 		if (limp->rlim_cur > maxsmap)
    293 			limp->rlim_cur = maxsmap;
    294 		if (limp->rlim_max > maxsmap)
    295 			limp->rlim_max = maxsmap;
    296 
    297 		/*
    298 		 * Return EINVAL if the new stack size limit is lower than
    299 		 * current usage. Otherwise, the process would get SIGSEGV the
    300 		 * moment it would try to access anything on it's current stack.
    301 		 * This conforms to SUSv2.
    302 		 */
    303 		if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE
    304 		    || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE)
    305 			return (EINVAL);
    306 
    307 		/*
    308 		 * Stack is allocated to the max at exec time with
    309 		 * only "rlim_cur" bytes accessible (In other words,
    310 		 * allocates stack dividing two contiguous regions at
    311 		 * "rlim_cur" bytes boundary).
    312 		 *
    313 		 * Since allocation is done in terms of page, roundup
    314 		 * "rlim_cur" (otherwise, contiguous regions
    315 		 * overlap).  If stack limit is going up make more
    316 		 * accessible, if going down make inaccessible.
    317 		 */
    318 		limp->rlim_cur = round_page(limp->rlim_cur);
    319 		if (limp->rlim_cur != alimp->rlim_cur) {
    320 			vaddr_t addr;
    321 			vsize_t size;
    322 			vm_prot_t prot;
    323 
    324 			if (limp->rlim_cur > alimp->rlim_cur) {
    325 				prot = VM_PROT_READ | VM_PROT_WRITE;
    326 				size = limp->rlim_cur - alimp->rlim_cur;
    327 				addr = USRSTACK - limp->rlim_cur;
    328 			} else {
    329 				prot = VM_PROT_NONE;
    330 				size = alimp->rlim_cur - limp->rlim_cur;
    331 				addr = USRSTACK - alimp->rlim_cur;
    332 			}
    333 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
    334 					      addr, addr+size, prot, FALSE);
    335 		}
    336 		break;
    337 
    338 	case RLIMIT_NOFILE:
    339 		if (limp->rlim_cur > maxfiles)
    340 			limp->rlim_cur = maxfiles;
    341 		if (limp->rlim_max > maxfiles)
    342 			limp->rlim_max = maxfiles;
    343 		break;
    344 
    345 	case RLIMIT_NPROC:
    346 		if (limp->rlim_cur > maxproc)
    347 			limp->rlim_cur = maxproc;
    348 		if (limp->rlim_max > maxproc)
    349 			limp->rlim_max = maxproc;
    350 		break;
    351 	}
    352 	*alimp = *limp;
    353 	return (0);
    354 }
    355 
    356 /* ARGSUSED */
    357 int
    358 sys_getrlimit(l, v, retval)
    359 	struct lwp *l;
    360 	void *v;
    361 	register_t *retval;
    362 {
    363 	struct sys_getrlimit_args /* {
    364 		syscallarg(int) which;
    365 		syscallarg(struct rlimit *) rlp;
    366 	} */ *uap = v;
    367 	struct proc *p = l->l_proc;
    368 	int which = SCARG(uap, which);
    369 
    370 	if ((u_int)which >= RLIM_NLIMITS)
    371 		return (EINVAL);
    372 	return (copyout(&p->p_rlimit[which], SCARG(uap, rlp),
    373 	    sizeof(struct rlimit)));
    374 }
    375 
    376 /*
    377  * Transform the running time and tick information in proc p into user,
    378  * system, and interrupt time usage.
    379  */
    380 void
    381 calcru(p, up, sp, ip)
    382 	struct proc *p;
    383 	struct timeval *up;
    384 	struct timeval *sp;
    385 	struct timeval *ip;
    386 {
    387 	u_quad_t u, st, ut, it, tot;
    388 	unsigned long sec;
    389 	long usec;
    390 	int s;
    391 	struct timeval tv;
    392 	struct lwp *l;
    393 
    394 	s = splstatclock();
    395 	st = p->p_sticks;
    396 	ut = p->p_uticks;
    397 	it = p->p_iticks;
    398 	splx(s);
    399 
    400 	sec = p->p_rtime.tv_sec;
    401 	usec = p->p_rtime.tv_usec;
    402 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
    403 		if (l->l_stat == LSONPROC) {
    404 			struct schedstate_percpu *spc;
    405 
    406 			KDASSERT(l->l_cpu != NULL);
    407 			spc = &l->l_cpu->ci_schedstate;
    408 
    409 			/*
    410 			 * Adjust for the current time slice.  This is
    411 			 * actually fairly important since the error
    412 			 * here is on the order of a time quantum,
    413 			 * which is much greater than the sampling
    414 			 * error.
    415 			 */
    416 			microtime(&tv);
    417 			sec += tv.tv_sec - spc->spc_runtime.tv_sec;
    418 			usec += tv.tv_usec - spc->spc_runtime.tv_usec;
    419 		}
    420 	}
    421 
    422 	tot = st + ut + it;
    423 	u = sec * 1000000ull + usec;
    424 
    425 	if (tot == 0) {
    426 		/* No ticks, so can't use to share time out, split 50-50 */
    427 		st = ut = u / 2;
    428 	} else {
    429 		st = (u * st) / tot;
    430 		ut = (u * ut) / tot;
    431 	}
    432 	sp->tv_sec = st / 1000000;
    433 	sp->tv_usec = st % 1000000;
    434 	up->tv_sec = ut / 1000000;
    435 	up->tv_usec = ut % 1000000;
    436 	if (ip != NULL) {
    437 		if (it != 0)
    438 			it = (u * it) / tot;
    439 		ip->tv_sec = it / 1000000;
    440 		ip->tv_usec = it % 1000000;
    441 	}
    442 }
    443 
    444 /* ARGSUSED */
    445 int
    446 sys_getrusage(l, v, retval)
    447 	struct lwp *l;
    448 	void *v;
    449 	register_t *retval;
    450 {
    451 	struct sys_getrusage_args /* {
    452 		syscallarg(int) who;
    453 		syscallarg(struct rusage *) rusage;
    454 	} */ *uap = v;
    455 	struct rusage *rup;
    456 	struct proc *p = l->l_proc;
    457 
    458 	switch (SCARG(uap, who)) {
    459 
    460 	case RUSAGE_SELF:
    461 		rup = &p->p_stats->p_ru;
    462 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
    463 		break;
    464 
    465 	case RUSAGE_CHILDREN:
    466 		rup = &p->p_stats->p_cru;
    467 		break;
    468 
    469 	default:
    470 		return (EINVAL);
    471 	}
    472 	return (copyout(rup, SCARG(uap, rusage), sizeof(struct rusage)));
    473 }
    474 
    475 void
    476 ruadd(ru, ru2)
    477 	struct rusage *ru, *ru2;
    478 {
    479 	long *ip, *ip2;
    480 	int i;
    481 
    482 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
    483 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
    484 	if (ru->ru_maxrss < ru2->ru_maxrss)
    485 		ru->ru_maxrss = ru2->ru_maxrss;
    486 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
    487 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
    488 		*ip++ += *ip2++;
    489 }
    490 
    491 /*
    492  * Make a copy of the plimit structure.
    493  * We share these structures copy-on-write after fork,
    494  * and copy when a limit is changed.
    495  */
    496 struct plimit *
    497 limcopy(lim)
    498 	struct plimit *lim;
    499 {
    500 	struct plimit *newlim;
    501 	size_t l;
    502 
    503 	newlim = pool_get(&plimit_pool, PR_WAITOK);
    504 	memcpy(newlim->pl_rlimit, lim->pl_rlimit,
    505 	    sizeof(struct rlimit) * RLIM_NLIMITS);
    506 	if (lim->pl_corename == defcorename) {
    507 		newlim->pl_corename = defcorename;
    508 	} else {
    509 		l = strlen(lim->pl_corename) + 1;
    510 		newlim->pl_corename = malloc(l, M_TEMP, M_WAITOK);
    511 		strlcpy(newlim->pl_corename, lim->pl_corename, l);
    512 	}
    513 	newlim->p_lflags = 0;
    514 	newlim->p_refcnt = 1;
    515 	return (newlim);
    516 }
    517 
    518 void
    519 limfree(lim)
    520 	struct plimit *lim;
    521 {
    522 
    523 	if (--lim->p_refcnt > 0)
    524 		return;
    525 #ifdef DIAGNOSTIC
    526 	if (lim->p_refcnt < 0)
    527 		panic("limfree");
    528 #endif
    529 	if (lim->pl_corename != defcorename)
    530 		free(lim->pl_corename, M_TEMP);
    531 	pool_put(&plimit_pool, lim);
    532 }
    533 
    534 struct pstats *
    535 pstatscopy(ps)
    536 	struct pstats *ps;
    537 {
    538 
    539 	struct pstats *newps;
    540 
    541 	newps = pool_get(&pstats_pool, PR_WAITOK);
    542 
    543 	memset(&newps->pstat_startzero, 0,
    544 	(unsigned) ((caddr_t)&newps->pstat_endzero -
    545 		    (caddr_t)&newps->pstat_startzero));
    546 	memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy,
    547 	((caddr_t)&newps->pstat_endcopy -
    548 	 (caddr_t)&newps->pstat_startcopy));
    549 
    550 	return (newps);
    551 
    552 }
    553 
    554 void
    555 pstatsfree(ps)
    556 	struct pstats *ps;
    557 {
    558 
    559 	pool_put(&pstats_pool, ps);
    560 }
    561 
    562 /*
    563  * sysctl interface in five parts
    564  */
    565 
    566 /*
    567  * a routine for sysctl proc subtree helpers that need to pick a valid
    568  * process by pid.
    569  */
    570 static int
    571 sysctl_proc_findproc(struct proc *p, struct proc **p2, pid_t pid)
    572 {
    573 	struct proc *ptmp;
    574 	int i, error = 0;
    575 
    576 	if (pid == PROC_CURPROC)
    577 		ptmp = p;
    578 	else if ((ptmp = pfind(pid)) == NULL)
    579 		error = ESRCH;
    580 	else {
    581 		/*
    582 		 * suid proc of ours or proc not ours
    583 		 */
    584 		if (p->p_cred->p_ruid != ptmp->p_cred->p_ruid ||
    585 		    p->p_cred->p_ruid != ptmp->p_cred->p_svuid)
    586 			error = suser(p->p_ucred, &p->p_acflag);
    587 
    588 		/*
    589 		 * sgid proc has sgid back to us temporarily
    590 		 */
    591 		else if (ptmp->p_cred->p_rgid != ptmp->p_cred->p_svgid)
    592 			error = suser(p->p_ucred, &p->p_acflag);
    593 
    594 		/*
    595 		 * our rgid must be in target's group list (ie,
    596 		 * sub-processes started by a sgid process)
    597 		 */
    598 		else {
    599 			for (i = 0; i < p->p_ucred->cr_ngroups; i++) {
    600 				if (p->p_ucred->cr_groups[i] ==
    601 				    ptmp->p_cred->p_rgid)
    602 					break;
    603 			}
    604 			if (i == p->p_ucred->cr_ngroups)
    605 				error = suser(p->p_ucred, &p->p_acflag);
    606 		}
    607 	}
    608 
    609 	*p2 = ptmp;
    610 	return (error);
    611 }
    612 
    613 /*
    614  * sysctl helper routine for setting a process's specific corefile
    615  * name.  picks the process based on the given pid and checks the
    616  * correctness of the new value.
    617  */
    618 static int
    619 sysctl_proc_corename(SYSCTLFN_ARGS)
    620 {
    621 	struct proc *ptmp, *p;
    622 	struct plimit *newplim;
    623 	int error = 0, len;
    624 	char cname[MAXPATHLEN], *tmp;
    625 	struct sysctlnode node;
    626 
    627 	/*
    628 	 * is this all correct?
    629 	 */
    630 	if (namelen != 0)
    631 		return (EINVAL);
    632 	if (name[-1] != PROC_PID_CORENAME)
    633 		return (EINVAL);
    634 
    635 	/*
    636 	 * whom are we tweaking?
    637 	 */
    638 	p = l->l_proc;
    639 	error = sysctl_proc_findproc(p, &ptmp, (pid_t)name[-2]);
    640 	if (error)
    641 		return (error);
    642 
    643 	/*
    644 	 * let them modify a temporary copy of the core name
    645 	 */
    646 	node = *rnode;
    647 	strlcpy(cname, ptmp->p_limit->pl_corename, sizeof(cname));
    648 	node.sysctl_data = cname;
    649 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
    650 
    651 	/*
    652 	 * if that failed, or they have nothing new to say, or we've
    653 	 * heard it before...
    654 	 */
    655 	if (error || newp == NULL ||
    656 	    strcmp(cname, ptmp->p_limit->pl_corename) == 0)
    657 		return (error);
    658 
    659 	/*
    660 	 * no error yet and cname now has the new core name in it.
    661 	 * let's see if it looks acceptable.  it must be either "core"
    662 	 * or end in ".core" or "/core".
    663 	 */
    664 	len = strlen(cname);
    665 	if (len < 4)
    666 		return (EINVAL);
    667 	if (strcmp(cname + len - 4, "core") != 0)
    668 		return (EINVAL);
    669 	if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.')
    670 		return (EINVAL);
    671 
    672 	/*
    673 	 * hmm...looks good.  now...where do we put it?
    674 	 */
    675 	tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL);
    676 	if (tmp == NULL)
    677 		return (ENOMEM);
    678 	strlcpy(tmp, cname, len + 1);
    679 
    680 	if (ptmp->p_limit->p_refcnt > 1 &&
    681 	    (ptmp->p_limit->p_lflags & PL_SHAREMOD) == 0) {
    682 		newplim = limcopy(ptmp->p_limit);
    683 		limfree(ptmp->p_limit);
    684 		ptmp->p_limit = newplim;
    685 	}
    686 	if (ptmp->p_limit->pl_corename != defcorename)
    687 		FREE(ptmp->p_limit->pl_corename, M_SYSCTLDATA);
    688 	ptmp->p_limit->pl_corename = tmp;
    689 
    690 	return (error);
    691 }
    692 
    693 /*
    694  * sysctl helper routine for checking/setting a process's stop flags,
    695  * one for fork and one for exec.
    696  */
    697 static int
    698 sysctl_proc_stop(SYSCTLFN_ARGS)
    699 {
    700 	struct proc *p, *ptmp;
    701 	int i, f, error = 0;
    702 	struct sysctlnode node;
    703 
    704 	if (namelen != 0)
    705 		return (EINVAL);
    706 
    707 	p = l->l_proc;
    708 	error = sysctl_proc_findproc(p, &ptmp, (pid_t)name[-2]);
    709 	if (error)
    710 		return (error);
    711 
    712 	switch (rnode->sysctl_num) {
    713 	case PROC_PID_STOPFORK:
    714 		f = P_STOPFORK;
    715 		break;
    716 	case PROC_PID_STOPEXEC:
    717 		f = P_STOPEXEC;
    718 		break;
    719 	case PROC_PID_STOPEXIT:
    720 		f = P_STOPEXIT;
    721 		break;
    722 	default:
    723 		return (EINVAL);
    724 	}
    725 
    726 	i = (ptmp->p_flag & f) ? 1 : 0;
    727 	node = *rnode;
    728 	node.sysctl_data = &i;
    729 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
    730 	if (error || newp == NULL)
    731 		return (error);
    732 
    733 	if (i)
    734 		ptmp->p_flag |= f;
    735 	else
    736 		ptmp->p_flag &= ~f;
    737 
    738 	return (0);
    739 }
    740 
    741 /*
    742  * sysctl helper routine for a process's rlimits as exposed by sysctl.
    743  */
    744 static int
    745 sysctl_proc_plimit(SYSCTLFN_ARGS)
    746 {
    747 	struct proc *ptmp, *p;
    748 	u_int limitno;
    749 	int which, error = 0;
    750         struct rlimit alim;
    751 	struct sysctlnode node;
    752 
    753 	if (namelen != 0)
    754 		return (EINVAL);
    755 
    756 	which = name[-1];
    757 	if (which != PROC_PID_LIMIT_TYPE_SOFT &&
    758 	    which != PROC_PID_LIMIT_TYPE_HARD)
    759 		return (EINVAL);
    760 
    761 	limitno = name[-2] - 1;
    762 	if (limitno >= RLIM_NLIMITS)
    763 		return (EINVAL);
    764 
    765 	if (name[-3] != PROC_PID_LIMIT)
    766 		return (EINVAL);
    767 
    768 	p = l->l_proc;
    769 	error = sysctl_proc_findproc(p, &ptmp, (pid_t)name[-4]);
    770 	if (error)
    771 		return (error);
    772 
    773 	node = *rnode;
    774 	memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim));
    775 	if (which == PROC_PID_LIMIT_TYPE_HARD)
    776 		node.sysctl_data = &alim.rlim_max;
    777 	else
    778 		node.sysctl_data = &alim.rlim_cur;
    779 
    780 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
    781 	if (error || newp == NULL)
    782 		return (error);
    783 
    784 	return (dosetrlimit(ptmp, p->p_cred, limitno, &alim));
    785 }
    786 
    787 /*
    788  * and finally, the actually glue that sticks it to the tree
    789  */
    790 SYSCTL_SETUP(sysctl_proc_setup, "sysctl proc subtree setup")
    791 {
    792 
    793 	sysctl_createv(clog, 0, NULL, NULL,
    794 		       CTLFLAG_PERMANENT,
    795 		       CTLTYPE_NODE, "proc", NULL,
    796 		       NULL, 0, NULL, 0,
    797 		       CTL_PROC, CTL_EOL);
    798 	sysctl_createv(clog, 0, NULL, NULL,
    799 		       CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER,
    800 		       CTLTYPE_NODE, "curproc",
    801 		       SYSCTL_DESCR("Per-process settings"),
    802 		       NULL, 0, NULL, 0,
    803 		       CTL_PROC, PROC_CURPROC, CTL_EOL);
    804 
    805 	sysctl_createv(clog, 0, NULL, NULL,
    806 		       CTLFLAG_PERMANENT|CTLFLAG_READONLY2|CTLFLAG_ANYWRITE,
    807 		       CTLTYPE_STRING, "corename",
    808 		       SYSCTL_DESCR("Core file name"),
    809 		       sysctl_proc_corename, 0, NULL, MAXPATHLEN,
    810 		       CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL);
    811 	sysctl_createv(clog, 0, NULL, NULL,
    812 		       CTLFLAG_PERMANENT,
    813 		       CTLTYPE_NODE, "rlimit",
    814 		       SYSCTL_DESCR("Process limits"),
    815 		       NULL, 0, NULL, 0,
    816 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL);
    817 
    818 #define create_proc_plimit(s, n) do {					\
    819 	sysctl_createv(clog, 0, NULL, NULL,				\
    820 		       CTLFLAG_PERMANENT,				\
    821 		       CTLTYPE_NODE, s,					\
    822 		       SYSCTL_DESCR("Process " s " limits"),		\
    823 		       NULL, 0, NULL, 0,				\
    824 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
    825 		       CTL_EOL);					\
    826 	sysctl_createv(clog, 0, NULL, NULL,				\
    827 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \
    828 		       CTLTYPE_QUAD, "soft",				\
    829 		       SYSCTL_DESCR("Process soft " s " limit"),	\
    830 		       sysctl_proc_plimit, 0, NULL, 0,			\
    831 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
    832 		       PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL);		\
    833 	sysctl_createv(clog, 0, NULL, NULL,				\
    834 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \
    835 		       CTLTYPE_QUAD, "hard",				\
    836 		       SYSCTL_DESCR("Process hard " s " limit"),	\
    837 		       sysctl_proc_plimit, 0, NULL, 0,			\
    838 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
    839 		       PROC_PID_LIMIT_TYPE_HARD, CTL_EOL);		\
    840 	} while (0/*CONSTCOND*/)
    841 
    842 	create_proc_plimit("cputime",		PROC_PID_LIMIT_CPU);
    843 	create_proc_plimit("filesize",		PROC_PID_LIMIT_FSIZE);
    844 	create_proc_plimit("datasize",		PROC_PID_LIMIT_DATA);
    845 	create_proc_plimit("stacksize",		PROC_PID_LIMIT_STACK);
    846 	create_proc_plimit("coredumpsize",	PROC_PID_LIMIT_CORE);
    847 	create_proc_plimit("memoryuse",		PROC_PID_LIMIT_RSS);
    848 	create_proc_plimit("memorylocked",	PROC_PID_LIMIT_MEMLOCK);
    849 	create_proc_plimit("maxproc",		PROC_PID_LIMIT_NPROC);
    850 	create_proc_plimit("descriptors",	PROC_PID_LIMIT_NOFILE);
    851 
    852 #undef create_proc_plimit
    853 
    854 	sysctl_createv(clog, 0, NULL, NULL,
    855 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
    856 		       CTLTYPE_INT, "stopfork",
    857 		       SYSCTL_DESCR("Stop process at fork(2)"),
    858 		       sysctl_proc_stop, 0, NULL, 0,
    859 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL);
    860 	sysctl_createv(clog, 0, NULL, NULL,
    861 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
    862 		       CTLTYPE_INT, "stopexec",
    863 		       SYSCTL_DESCR("Stop process at execve(2)"),
    864 		       sysctl_proc_stop, 0, NULL, 0,
    865 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL);
    866 	sysctl_createv(clog, 0, NULL, NULL,
    867 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
    868 		       CTLTYPE_INT, "stopexit",
    869 		       SYSCTL_DESCR("Stop process before completing exit"),
    870 		       sysctl_proc_stop, 0, NULL, 0,
    871 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL);
    872 }
    873