Home | History | Annotate | Line # | Download | only in kern
kern_resource.c revision 1.103.4.1
      1 /*	$NetBSD: kern_resource.c,v 1.103.4.1 2006/09/11 18:07:25 ad 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 1.103.4.1 2006/09/11 18:07:25 ad 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/namei.h>
     49 #include <sys/pool.h>
     50 #include <sys/proc.h>
     51 #include <sys/sysctl.h>
     52 #include <sys/kauth.h>
     53 
     54 #include <sys/mount.h>
     55 #include <sys/sa.h>
     56 #include <sys/syscallargs.h>
     57 
     58 #include <uvm/uvm_extern.h>
     59 
     60 /*
     61  * Maximum process data and stack limits.
     62  * They are variables so they are patchable.
     63  */
     64 rlim_t maxdmap = MAXDSIZ;
     65 rlim_t maxsmap = MAXSSIZ;
     66 
     67 struct uihashhead *uihashtbl;
     68 u_long uihash;		/* size of hash table - 1 */
     69 struct simplelock uihashtbl_slock = SIMPLELOCK_INITIALIZER;
     70 
     71 
     72 /*
     73  * Resource controls and accounting.
     74  */
     75 
     76 int
     77 sys_getpriority(struct lwp *l, void *v, register_t *retval)
     78 {
     79 	struct sys_getpriority_args /* {
     80 		syscallarg(int) which;
     81 		syscallarg(id_t) who;
     82 	} */ *uap = v;
     83 	struct proc *curp = l->l_proc, *p;
     84 	int low = NZERO + PRIO_MAX + 1;
     85 	int who = SCARG(uap, who);
     86 
     87 	switch (SCARG(uap, which)) {
     88 
     89 	case PRIO_PROCESS:
     90 		if (who == 0)
     91 			p = curp;
     92 		else
     93 			p = p_find(who, 0);
     94 		if (p != NULL)
     95 			low = p->p_nice;
     96 		if (who != 0)
     97 			rw_exit(&proclist_lock);
     98 		break;
     99 
    100 	case PRIO_PGRP: {
    101 		struct pgrp *pg;
    102 
    103 		rw_enter(&proclist_lock, RW_READER);
    104 		if (who == 0)
    105 			pg = curp->p_pgrp;
    106 		else if ((pg = pg_find(who, PFIND_LOCKED | PFIND_UNLOCK_FAIL))
    107 		    == NULL)
    108 			break;
    109 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
    110 			if (p->p_nice < low)
    111 				low = p->p_nice;
    112 		}
    113 		rw_exit(&proclist_lock);
    114 		break;
    115 	}
    116 
    117 	case PRIO_USER:
    118 		if (who == 0)
    119 			who = (int)kauth_cred_geteuid(l->l_cred);
    120 		rw_enter(&proclist_lock, RW_READER);
    121 		PROCLIST_FOREACH(p, &allproc) {
    122 			mutex_enter(&p->p_crmutex);
    123 			if (kauth_cred_geteuid(p->p_cred) ==
    124 			    (uid_t)who && p->p_nice < low)
    125 				low = p->p_nice;
    126 			mutex_exit(&p->p_crmutex);
    127 		}
    128 		rw_exit(&proclist_lock);
    129 		break;
    130 
    131 	default:
    132 		return (EINVAL);
    133 	}
    134 	if (low == NZERO + PRIO_MAX + 1)
    135 		return (ESRCH);
    136 	*retval = low - NZERO;
    137 	return (0);
    138 }
    139 
    140 /* ARGSUSED */
    141 int
    142 sys_setpriority(struct lwp *l, void *v, register_t *retval)
    143 {
    144 	struct sys_setpriority_args /* {
    145 		syscallarg(int) which;
    146 		syscallarg(id_t) who;
    147 		syscallarg(int) prio;
    148 	} */ *uap = v;
    149 	struct proc *curp = l->l_proc, *p;
    150 	int found = 0, error = 0;
    151 	int who = SCARG(uap, who);
    152 
    153 	switch (SCARG(uap, which)) {
    154 
    155 	case PRIO_PROCESS:
    156 		if (who == 0)
    157 			p = curp;
    158 		else
    159 			p = p_find(who, 0);
    160 		if (p != 0) {
    161 			mutex_enter(&p->p_crmutex);
    162 			error = donice(l, p, SCARG(uap, prio));
    163 			mutex_exit(&p->p_crmutex);
    164 		}
    165 		if (who != 0)
    166 			rw_exit(&proclist_lock);
    167 		found++;
    168 		break;
    169 
    170 	case PRIO_PGRP: {
    171 		struct pgrp *pg;
    172 
    173 		rw_enter(&proclist_lock, RW_READER);
    174 		if (who == 0)
    175 			pg = curp->p_pgrp;
    176 		else if ((pg = pg_find(who, PFIND_LOCKED | PFIND_UNLOCK_FAIL)) == NULL)
    177 			break;
    178 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
    179 			mutex_enter(&p->p_crmutex);
    180 			error = donice(l, p, SCARG(uap, prio));
    181 			mutex_exit(&p->p_crmutex);
    182 			found++;
    183 		}
    184 		rw_exit(&proclist_lock);
    185 		break;
    186 	}
    187 
    188 	case PRIO_USER:
    189 		if (who == 0)
    190 			who = (int)kauth_cred_geteuid(l->l_cred);
    191 		rw_enter(&proclist_lock, RW_READER);
    192 		PROCLIST_FOREACH(p, &allproc) {
    193 			mutex_enter(&p->p_crmutex);
    194 			if (kauth_cred_geteuid(p->p_cred) ==
    195 			    (uid_t)SCARG(uap, who)) {
    196 				error = donice(l, p, SCARG(uap, prio));
    197 				found++;
    198 			}
    199 			mutex_exit(&p->p_crmutex);
    200 		}
    201 		rw_exit(&proclist_lock);
    202 		break;
    203 
    204 	default:
    205 		return (EINVAL);
    206 	}
    207 	if (found == 0)
    208 		return (ESRCH);
    209 	return (error);
    210 }
    211 
    212 /*
    213  * Renice a process.
    214  *
    215  * Call with the target process' credentials locked.
    216  */
    217 int
    218 donice(struct lwp *l, struct proc *chgp, int n)
    219 {
    220 	kauth_cred_t cred = l->l_cred;
    221 	int s;
    222 
    223 	LOCK_ASSERT(mutex_owned(&chgp->p_crmutex));
    224 
    225 	if (kauth_cred_geteuid(cred) && kauth_cred_getuid(cred) &&
    226 	    kauth_cred_geteuid(cred) != kauth_cred_geteuid(chgp->p_cred) &&
    227 	    kauth_cred_getuid(cred) != kauth_cred_geteuid(chgp->p_cred))
    228 		return (EPERM);
    229 	if (n > PRIO_MAX)
    230 		n = PRIO_MAX;
    231 	if (n < PRIO_MIN)
    232 		n = PRIO_MIN;
    233 	n += NZERO;
    234 	if (n < chgp->p_nice && kauth_authorize_generic(cred,
    235 	    KAUTH_GENERIC_ISSUSER, &l->l_acflag))
    236 		return (EACCES);
    237 	chgp->p_nice = n;
    238 	SCHED_LOCK(s);
    239 	(void)resetprocpriority(chgp);
    240 	SCHED_UNLOCK(s);
    241 	return (0);
    242 }
    243 
    244 /* ARGSUSED */
    245 int
    246 sys_setrlimit(struct lwp *l, void *v, register_t *retval)
    247 {
    248 	struct sys_setrlimit_args /* {
    249 		syscallarg(int) which;
    250 		syscallarg(const struct rlimit *) rlp;
    251 	} */ *uap = v;
    252 	int which = SCARG(uap, which);
    253 	struct rlimit alim;
    254 	int error;
    255 
    256 	error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit));
    257 	if (error)
    258 		return (error);
    259 	return (dosetrlimit(l, l->l_proc, which, &alim));
    260 }
    261 
    262 int
    263 dosetrlimit(struct lwp *l, struct proc *p, int which, struct rlimit *limp)
    264 {
    265 	struct rlimit *alimp;
    266 	struct plimit *oldplim;
    267 	int error;
    268 
    269 	if ((u_int)which >= RLIM_NLIMITS)
    270 		return (EINVAL);
    271 
    272 	if (limp->rlim_cur < 0 || limp->rlim_max < 0)
    273 		return (EINVAL);
    274 
    275 	alimp = &p->p_rlimit[which];
    276 	/* if we don't change the value, no need to limcopy() */
    277 	if (limp->rlim_cur == alimp->rlim_cur &&
    278 	    limp->rlim_max == alimp->rlim_max)
    279 		return 0;
    280 
    281 	if (limp->rlim_cur > limp->rlim_max) {
    282 		/*
    283 		 * This is programming error. According to SUSv2, we should
    284 		 * return error in this case.
    285 		 */
    286 		return (EINVAL);
    287 	}
    288 	if (limp->rlim_max > alimp->rlim_max && (error =
    289 	    kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER,
    290 	    &l->l_acflag)) != 0)
    291 			return (error);
    292 
    293 	if (p->p_limit->p_refcnt > 1 &&
    294 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
    295 		p->p_limit = limcopy(oldplim = p->p_limit);
    296 		limfree(oldplim);
    297 		alimp = &p->p_rlimit[which];
    298 	}
    299 
    300 	switch (which) {
    301 
    302 	case RLIMIT_DATA:
    303 		if (limp->rlim_cur > maxdmap)
    304 			limp->rlim_cur = maxdmap;
    305 		if (limp->rlim_max > maxdmap)
    306 			limp->rlim_max = maxdmap;
    307 		break;
    308 
    309 	case RLIMIT_STACK:
    310 		if (limp->rlim_cur > maxsmap)
    311 			limp->rlim_cur = maxsmap;
    312 		if (limp->rlim_max > maxsmap)
    313 			limp->rlim_max = maxsmap;
    314 
    315 		/*
    316 		 * Return EINVAL if the new stack size limit is lower than
    317 		 * current usage. Otherwise, the process would get SIGSEGV the
    318 		 * moment it would try to access anything on it's current stack.
    319 		 * This conforms to SUSv2.
    320 		 */
    321 		if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE
    322 		    || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE)
    323 			return (EINVAL);
    324 
    325 		/*
    326 		 * Stack is allocated to the max at exec time with
    327 		 * only "rlim_cur" bytes accessible (In other words,
    328 		 * allocates stack dividing two contiguous regions at
    329 		 * "rlim_cur" bytes boundary).
    330 		 *
    331 		 * Since allocation is done in terms of page, roundup
    332 		 * "rlim_cur" (otherwise, contiguous regions
    333 		 * overlap).  If stack limit is going up make more
    334 		 * accessible, if going down make inaccessible.
    335 		 */
    336 		limp->rlim_cur = round_page(limp->rlim_cur);
    337 		if (limp->rlim_cur != alimp->rlim_cur) {
    338 			vaddr_t addr;
    339 			vsize_t size;
    340 			vm_prot_t prot;
    341 
    342 			if (limp->rlim_cur > alimp->rlim_cur) {
    343 				prot = VM_PROT_READ | VM_PROT_WRITE;
    344 				size = limp->rlim_cur - alimp->rlim_cur;
    345 				addr = (vaddr_t)p->p_vmspace->vm_minsaddr -
    346 				    limp->rlim_cur;
    347 			} else {
    348 				prot = VM_PROT_NONE;
    349 				size = alimp->rlim_cur - limp->rlim_cur;
    350 				addr = (vaddr_t)p->p_vmspace->vm_minsaddr -
    351 				     alimp->rlim_cur;
    352 			}
    353 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
    354 			    addr, addr+size, prot, FALSE);
    355 		}
    356 		break;
    357 
    358 	case RLIMIT_NOFILE:
    359 		if (limp->rlim_cur > maxfiles)
    360 			limp->rlim_cur = maxfiles;
    361 		if (limp->rlim_max > maxfiles)
    362 			limp->rlim_max = maxfiles;
    363 		break;
    364 
    365 	case RLIMIT_NPROC:
    366 		if (limp->rlim_cur > maxproc)
    367 			limp->rlim_cur = maxproc;
    368 		if (limp->rlim_max > maxproc)
    369 			limp->rlim_max = maxproc;
    370 		break;
    371 	}
    372 	*alimp = *limp;
    373 	return (0);
    374 }
    375 
    376 /* ARGSUSED */
    377 int
    378 sys_getrlimit(struct lwp *l, void *v, register_t *retval)
    379 {
    380 	struct sys_getrlimit_args /* {
    381 		syscallarg(int) which;
    382 		syscallarg(struct rlimit *) rlp;
    383 	} */ *uap = v;
    384 	struct proc *p = l->l_proc;
    385 	int which = SCARG(uap, which);
    386 
    387 	if ((u_int)which >= RLIM_NLIMITS)
    388 		return (EINVAL);
    389 	return (copyout(&p->p_rlimit[which], SCARG(uap, rlp),
    390 	    sizeof(struct rlimit)));
    391 }
    392 
    393 /*
    394  * Transform the running time and tick information in proc p into user,
    395  * system, and interrupt time usage.
    396  */
    397 void
    398 calcru(struct proc *p, struct timeval *up, struct timeval *sp,
    399     struct timeval *ip)
    400 {
    401 	u_quad_t u, st, ut, it, tot;
    402 	unsigned long sec;
    403 	long usec;
    404 	int s;
    405 	struct timeval tv;
    406 	struct lwp *l;
    407 
    408 	s = splstatclock();
    409 	st = p->p_sticks;
    410 	ut = p->p_uticks;
    411 	it = p->p_iticks;
    412 	splx(s);
    413 
    414 	sec = p->p_rtime.tv_sec;
    415 	usec = p->p_rtime.tv_usec;
    416 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
    417 		if (l->l_stat == LSONPROC) {
    418 			struct schedstate_percpu *spc;
    419 
    420 			KDASSERT(l->l_cpu != NULL);
    421 			spc = &l->l_cpu->ci_schedstate;
    422 
    423 			/*
    424 			 * Adjust for the current time slice.  This is
    425 			 * actually fairly important since the error
    426 			 * here is on the order of a time quantum,
    427 			 * which is much greater than the sampling
    428 			 * error.
    429 			 */
    430 			microtime(&tv);
    431 			sec += tv.tv_sec - spc->spc_runtime.tv_sec;
    432 			usec += tv.tv_usec - spc->spc_runtime.tv_usec;
    433 		}
    434 	}
    435 
    436 	tot = st + ut + it;
    437 	u = sec * 1000000ull + usec;
    438 
    439 	if (tot == 0) {
    440 		/* No ticks, so can't use to share time out, split 50-50 */
    441 		st = ut = u / 2;
    442 	} else {
    443 		st = (u * st) / tot;
    444 		ut = (u * ut) / tot;
    445 	}
    446 	sp->tv_sec = st / 1000000;
    447 	sp->tv_usec = st % 1000000;
    448 	up->tv_sec = ut / 1000000;
    449 	up->tv_usec = ut % 1000000;
    450 	if (ip != NULL) {
    451 		if (it != 0)
    452 			it = (u * it) / tot;
    453 		ip->tv_sec = it / 1000000;
    454 		ip->tv_usec = it % 1000000;
    455 	}
    456 }
    457 
    458 /* ARGSUSED */
    459 int
    460 sys_getrusage(struct lwp *l, void *v, register_t *retval)
    461 {
    462 	struct sys_getrusage_args /* {
    463 		syscallarg(int) who;
    464 		syscallarg(struct rusage *) rusage;
    465 	} */ *uap = v;
    466 	struct rusage *rup;
    467 	struct proc *p = l->l_proc;
    468 
    469 	switch (SCARG(uap, who)) {
    470 
    471 	case RUSAGE_SELF:
    472 		rup = &p->p_stats->p_ru;
    473 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
    474 		break;
    475 
    476 	case RUSAGE_CHILDREN:
    477 		rup = &p->p_stats->p_cru;
    478 		break;
    479 
    480 	default:
    481 		return (EINVAL);
    482 	}
    483 	return (copyout(rup, SCARG(uap, rusage), sizeof(struct rusage)));
    484 }
    485 
    486 void
    487 ruadd(struct rusage *ru, struct rusage *ru2)
    488 {
    489 	long *ip, *ip2;
    490 	int i;
    491 
    492 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
    493 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
    494 	if (ru->ru_maxrss < ru2->ru_maxrss)
    495 		ru->ru_maxrss = ru2->ru_maxrss;
    496 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
    497 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
    498 		*ip++ += *ip2++;
    499 }
    500 
    501 /*
    502  * Make a copy of the plimit structure.
    503  * We share these structures copy-on-write after fork,
    504  * and copy when a limit is changed.
    505  */
    506 struct plimit *
    507 limcopy(struct plimit *lim)
    508 {
    509 	struct plimit *newlim;
    510 	size_t l = 0;
    511 
    512 	simple_lock(&lim->p_slock);
    513 	if (lim->pl_corename != defcorename)
    514 		l = strlen(lim->pl_corename) + 1;
    515 	simple_unlock(&lim->p_slock);
    516 
    517 	newlim = pool_get(&plimit_pool, PR_WAITOK);
    518 	simple_lock_init(&newlim->p_slock);
    519 	newlim->p_lflags = 0;
    520 	newlim->p_refcnt = 1;
    521 	newlim->pl_corename = (l != 0)
    522 		? malloc(l, M_TEMP, M_WAITOK)
    523 		: defcorename;
    524 
    525 	simple_lock(&lim->p_slock);
    526 	memcpy(newlim->pl_rlimit, lim->pl_rlimit,
    527 	    sizeof(struct rlimit) * RLIM_NLIMITS);
    528 
    529 	if (l != 0)
    530 		strlcpy(newlim->pl_corename, lim->pl_corename, l);
    531 	simple_unlock(&lim->p_slock);
    532 
    533 	return (newlim);
    534 }
    535 
    536 void
    537 limfree(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(struct pstats *ps)
    557 {
    558 
    559 	struct pstats *newps;
    560 
    561 	newps = pool_get(&pstats_pool, PR_WAITOK);
    562 
    563 	memset(&newps->pstat_startzero, 0,
    564 	(unsigned) ((caddr_t)&newps->pstat_endzero -
    565 		    (caddr_t)&newps->pstat_startzero));
    566 	memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy,
    567 	((caddr_t)&newps->pstat_endcopy -
    568 	 (caddr_t)&newps->pstat_startcopy));
    569 
    570 	return (newps);
    571 
    572 }
    573 
    574 void
    575 pstatsfree(struct pstats *ps)
    576 {
    577 
    578 	pool_put(&pstats_pool, ps);
    579 }
    580 
    581 /*
    582  * sysctl interface in five parts
    583  */
    584 
    585 /*
    586  * a routine for sysctl proc subtree helpers that need to pick a valid
    587  * process by pid.
    588  */
    589 static int
    590 sysctl_proc_findproc(struct lwp *l, struct proc **p2, pid_t pid)
    591 {
    592 	struct proc *ptmp;
    593 	int error = 0;
    594 
    595 	if (pid == PROC_CURPROC)
    596 		ptmp = l->l_proc;
    597 	else if ((ptmp = pfind(pid)) == NULL)
    598 		error = ESRCH;
    599 	else {
    600 		/*
    601 		 * suid proc of ours or proc not ours
    602 		 */
    603 		if (kauth_cred_getuid(l->l_cred) !=
    604 		    kauth_cred_getuid(ptmp->p_cred) ||
    605 		    kauth_cred_getuid(l->l_cred) !=
    606 		    kauth_cred_getsvuid(ptmp->p_cred))
    607 			error = kauth_authorize_generic(l->l_cred,
    608 			    KAUTH_GENERIC_ISSUSER, &l->l_acflag);
    609 
    610 		/*
    611 		 * sgid proc has sgid back to us temporarily
    612 		 */
    613 		else if (kauth_cred_getgid(ptmp->p_cred) !=
    614 		    kauth_cred_getsvgid(ptmp->p_cred))
    615 			error = kauth_authorize_generic(l->l_cred,
    616 			    KAUTH_GENERIC_ISSUSER, &l->l_acflag);
    617 
    618 		/*
    619 		 * our rgid must be in target's group list (ie,
    620 		 * sub-processes started by a sgid process)
    621 		 */
    622 		else {
    623 			int ismember = 0;
    624 
    625 			if (kauth_cred_ismember_gid(l->l_cred,
    626 			    kauth_cred_getgid(ptmp->p_cred), &ismember) != 0 ||
    627 			    !ismember) {
    628 				error = kauth_authorize_generic(l->l_cred,
    629 				    KAUTH_GENERIC_ISSUSER, &l->l_acflag);
    630 			}
    631 		}
    632 	}
    633 
    634 	*p2 = ptmp;
    635 	return (error);
    636 }
    637 
    638 /*
    639  * sysctl helper routine for setting a process's specific corefile
    640  * name.  picks the process based on the given pid and checks the
    641  * correctness of the new value.
    642  */
    643 static int
    644 sysctl_proc_corename(SYSCTLFN_ARGS)
    645 {
    646 	struct proc *ptmp;
    647 	struct plimit *lim;
    648 	int error = 0, len;
    649 	char *cname;
    650 	char *tmp;
    651 	struct sysctlnode node;
    652 
    653 	/*
    654 	 * is this all correct?
    655 	 */
    656 	if (namelen != 0)
    657 		return (EINVAL);
    658 	if (name[-1] != PROC_PID_CORENAME)
    659 		return (EINVAL);
    660 
    661 	/*
    662 	 * whom are we tweaking?
    663 	 */
    664 	error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]);
    665 	if (error)
    666 		return (error);
    667 
    668 	cname = PNBUF_GET();
    669 	/*
    670 	 * let them modify a temporary copy of the core name
    671 	 */
    672 	node = *rnode;
    673 	strlcpy(cname, ptmp->p_limit->pl_corename, MAXPATHLEN);
    674 	node.sysctl_data = cname;
    675 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
    676 
    677 	/*
    678 	 * if that failed, or they have nothing new to say, or we've
    679 	 * heard it before...
    680 	 */
    681 	if (error || newp == NULL ||
    682 	    strcmp(cname, ptmp->p_limit->pl_corename) == 0) {
    683 		goto done;
    684 	}
    685 
    686 	if (securelevel > 1)
    687 		return (EPERM);
    688 
    689 	/*
    690 	 * no error yet and cname now has the new core name in it.
    691 	 * let's see if it looks acceptable.  it must be either "core"
    692 	 * or end in ".core" or "/core".
    693 	 */
    694 	len = strlen(cname);
    695 	if (len < 4) {
    696 		error = EINVAL;
    697 	} else if (strcmp(cname + len - 4, "core") != 0) {
    698 		error = EINVAL;
    699 	} else if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.') {
    700 		error = EINVAL;
    701 	}
    702 	if (error != 0) {
    703 		goto done;
    704 	}
    705 
    706 	/*
    707 	 * hmm...looks good.  now...where do we put it?
    708 	 */
    709 	tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL);
    710 	if (tmp == NULL) {
    711 		error = ENOMEM;
    712 		goto done;
    713 	}
    714 	strlcpy(tmp, cname, len + 1);
    715 
    716 	lim = ptmp->p_limit;
    717 	if (lim->p_refcnt > 1 && (lim->p_lflags & PL_SHAREMOD) == 0) {
    718 		ptmp->p_limit = limcopy(lim);
    719 		limfree(lim);
    720 		lim = ptmp->p_limit;
    721 	}
    722 	if (lim->pl_corename != defcorename)
    723 		free(lim->pl_corename, M_TEMP);
    724 	lim->pl_corename = tmp;
    725 done:
    726 	PNBUF_PUT(cname);
    727 	return error;
    728 }
    729 
    730 /*
    731  * sysctl helper routine for checking/setting a process's stop flags,
    732  * one for fork and one for exec.
    733  */
    734 static int
    735 sysctl_proc_stop(SYSCTLFN_ARGS)
    736 {
    737 	struct proc *ptmp;
    738 	int i, f, error = 0;
    739 	struct sysctlnode node;
    740 
    741 	if (namelen != 0)
    742 		return (EINVAL);
    743 
    744 	error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]);
    745 	if (error)
    746 		return (error);
    747 
    748 	switch (rnode->sysctl_num) {
    749 	case PROC_PID_STOPFORK:
    750 		f = P_STOPFORK;
    751 		break;
    752 	case PROC_PID_STOPEXEC:
    753 		f = P_STOPEXEC;
    754 		break;
    755 	case PROC_PID_STOPEXIT:
    756 		f = P_STOPEXIT;
    757 		break;
    758 	default:
    759 		return (EINVAL);
    760 	}
    761 
    762 	i = (ptmp->p_flag & f) ? 1 : 0;
    763 	node = *rnode;
    764 	node.sysctl_data = &i;
    765 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
    766 	if (error || newp == NULL)
    767 		return (error);
    768 
    769 	if (i)
    770 		ptmp->p_flag |= f;
    771 	else
    772 		ptmp->p_flag &= ~f;
    773 
    774 	return (0);
    775 }
    776 
    777 /*
    778  * sysctl helper routine for a process's rlimits as exposed by sysctl.
    779  */
    780 static int
    781 sysctl_proc_plimit(SYSCTLFN_ARGS)
    782 {
    783 	struct proc *ptmp;
    784 	u_int limitno;
    785 	int which, error = 0;
    786         struct rlimit alim;
    787 	struct sysctlnode node;
    788 
    789 	if (namelen != 0)
    790 		return (EINVAL);
    791 
    792 	which = name[-1];
    793 	if (which != PROC_PID_LIMIT_TYPE_SOFT &&
    794 	    which != PROC_PID_LIMIT_TYPE_HARD)
    795 		return (EINVAL);
    796 
    797 	limitno = name[-2] - 1;
    798 	if (limitno >= RLIM_NLIMITS)
    799 		return (EINVAL);
    800 
    801 	if (name[-3] != PROC_PID_LIMIT)
    802 		return (EINVAL);
    803 
    804 	error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-4]);
    805 	if (error)
    806 		return (error);
    807 
    808 	node = *rnode;
    809 	memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim));
    810 	if (which == PROC_PID_LIMIT_TYPE_HARD)
    811 		node.sysctl_data = &alim.rlim_max;
    812 	else
    813 		node.sysctl_data = &alim.rlim_cur;
    814 
    815 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
    816 	if (error || newp == NULL)
    817 		return (error);
    818 
    819 	return (dosetrlimit(l, ptmp, limitno, &alim));
    820 }
    821 
    822 /*
    823  * and finally, the actually glue that sticks it to the tree
    824  */
    825 SYSCTL_SETUP(sysctl_proc_setup, "sysctl proc subtree setup")
    826 {
    827 
    828 	sysctl_createv(clog, 0, NULL, NULL,
    829 		       CTLFLAG_PERMANENT,
    830 		       CTLTYPE_NODE, "proc", NULL,
    831 		       NULL, 0, NULL, 0,
    832 		       CTL_PROC, CTL_EOL);
    833 	sysctl_createv(clog, 0, NULL, NULL,
    834 		       CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER,
    835 		       CTLTYPE_NODE, "curproc",
    836 		       SYSCTL_DESCR("Per-process settings"),
    837 		       NULL, 0, NULL, 0,
    838 		       CTL_PROC, PROC_CURPROC, CTL_EOL);
    839 
    840 	sysctl_createv(clog, 0, NULL, NULL,
    841 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
    842 		       CTLTYPE_STRING, "corename",
    843 		       SYSCTL_DESCR("Core file name"),
    844 		       sysctl_proc_corename, 0, NULL, MAXPATHLEN,
    845 		       CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL);
    846 	sysctl_createv(clog, 0, NULL, NULL,
    847 		       CTLFLAG_PERMANENT,
    848 		       CTLTYPE_NODE, "rlimit",
    849 		       SYSCTL_DESCR("Process limits"),
    850 		       NULL, 0, NULL, 0,
    851 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL);
    852 
    853 #define create_proc_plimit(s, n) do {					\
    854 	sysctl_createv(clog, 0, NULL, NULL,				\
    855 		       CTLFLAG_PERMANENT,				\
    856 		       CTLTYPE_NODE, s,					\
    857 		       SYSCTL_DESCR("Process " s " limits"),		\
    858 		       NULL, 0, NULL, 0,				\
    859 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
    860 		       CTL_EOL);					\
    861 	sysctl_createv(clog, 0, NULL, NULL,				\
    862 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \
    863 		       CTLTYPE_QUAD, "soft",				\
    864 		       SYSCTL_DESCR("Process soft " s " limit"),	\
    865 		       sysctl_proc_plimit, 0, NULL, 0,			\
    866 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
    867 		       PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL);		\
    868 	sysctl_createv(clog, 0, NULL, NULL,				\
    869 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \
    870 		       CTLTYPE_QUAD, "hard",				\
    871 		       SYSCTL_DESCR("Process hard " s " limit"),	\
    872 		       sysctl_proc_plimit, 0, NULL, 0,			\
    873 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
    874 		       PROC_PID_LIMIT_TYPE_HARD, CTL_EOL);		\
    875 	} while (0/*CONSTCOND*/)
    876 
    877 	create_proc_plimit("cputime",		PROC_PID_LIMIT_CPU);
    878 	create_proc_plimit("filesize",		PROC_PID_LIMIT_FSIZE);
    879 	create_proc_plimit("datasize",		PROC_PID_LIMIT_DATA);
    880 	create_proc_plimit("stacksize",		PROC_PID_LIMIT_STACK);
    881 	create_proc_plimit("coredumpsize",	PROC_PID_LIMIT_CORE);
    882 	create_proc_plimit("memoryuse",		PROC_PID_LIMIT_RSS);
    883 	create_proc_plimit("memorylocked",	PROC_PID_LIMIT_MEMLOCK);
    884 	create_proc_plimit("maxproc",		PROC_PID_LIMIT_NPROC);
    885 	create_proc_plimit("descriptors",	PROC_PID_LIMIT_NOFILE);
    886 	create_proc_plimit("sbsize",		PROC_PID_LIMIT_SBSIZE);
    887 
    888 #undef create_proc_plimit
    889 
    890 	sysctl_createv(clog, 0, NULL, NULL,
    891 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
    892 		       CTLTYPE_INT, "stopfork",
    893 		       SYSCTL_DESCR("Stop process at fork(2)"),
    894 		       sysctl_proc_stop, 0, NULL, 0,
    895 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL);
    896 	sysctl_createv(clog, 0, NULL, NULL,
    897 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
    898 		       CTLTYPE_INT, "stopexec",
    899 		       SYSCTL_DESCR("Stop process at execve(2)"),
    900 		       sysctl_proc_stop, 0, NULL, 0,
    901 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL);
    902 	sysctl_createv(clog, 0, NULL, NULL,
    903 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
    904 		       CTLTYPE_INT, "stopexit",
    905 		       SYSCTL_DESCR("Stop process before completing exit"),
    906 		       sysctl_proc_stop, 0, NULL, 0,
    907 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL);
    908 }
    909 
    910 struct uidinfo *
    911 uid_find(uid_t uid)
    912 {
    913 	struct uidinfo *uip;
    914 	struct uidinfo *newuip = NULL;
    915 	struct uihashhead *uipp;
    916 
    917 	uipp = UIHASH(uid);
    918 
    919 again:
    920 	simple_lock(&uihashtbl_slock);
    921 	LIST_FOREACH(uip, uipp, ui_hash)
    922 		if (uip->ui_uid == uid) {
    923 			simple_unlock(&uihashtbl_slock);
    924 			if (newuip)
    925 				free(newuip, M_PROC);
    926 			return uip;
    927 		}
    928 
    929 	if (newuip == NULL) {
    930 		simple_unlock(&uihashtbl_slock);
    931 		newuip = malloc(sizeof(*uip), M_PROC, M_WAITOK | M_ZERO);
    932 		goto again;
    933 	}
    934 	uip = newuip;
    935 
    936 	LIST_INSERT_HEAD(uipp, uip, ui_hash);
    937 	uip->ui_uid = uid;
    938 	simple_lock_init(&uip->ui_slock);
    939 	simple_unlock(&uihashtbl_slock);
    940 
    941 	return uip;
    942 }
    943 
    944 /*
    945  * Change the count associated with number of processes
    946  * a given user is using.
    947  */
    948 int
    949 chgproccnt(uid_t uid, int diff)
    950 {
    951 	struct uidinfo *uip;
    952 	int s;
    953 
    954 	if (diff == 0)
    955 		return 0;
    956 
    957 	uip = uid_find(uid);
    958 	UILOCK(uip, s);
    959 	uip->ui_proccnt += diff;
    960 	KASSERT(uip->ui_proccnt >= 0);
    961 	UIUNLOCK(uip, s);
    962 	return uip->ui_proccnt;
    963 }
    964 
    965 int
    966 chgsbsize(struct uidinfo *uip, u_long *hiwat, u_long to, rlim_t xmax)
    967 {
    968 	rlim_t nsb;
    969 	int s;
    970 
    971 	UILOCK(uip, s);
    972 	nsb = uip->ui_sbsize + to - *hiwat;
    973 	if (to > *hiwat && nsb > xmax) {
    974 		UIUNLOCK(uip, s);
    975 		splx(s);
    976 		return 0;
    977 	}
    978 	*hiwat = to;
    979 	uip->ui_sbsize = nsb;
    980 	KASSERT(uip->ui_sbsize >= 0);
    981 	UIUNLOCK(uip, s);
    982 	return 1;
    983 }
    984