Home | History | Annotate | Line # | Download | only in kern
kern_resource.c revision 1.149
      1 /*	$NetBSD: kern_resource.c,v 1.149 2009/01/29 22:27:23 drochner 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 1.149 2009/01/29 22:27:23 drochner 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/kmem.h>
     49 #include <sys/namei.h>
     50 #include <sys/pool.h>
     51 #include <sys/proc.h>
     52 #include <sys/sysctl.h>
     53 #include <sys/timevar.h>
     54 #include <sys/kauth.h>
     55 #include <sys/atomic.h>
     56 #include <sys/mount.h>
     57 #include <sys/syscallargs.h>
     58 #include <sys/atomic.h>
     59 
     60 #include <uvm/uvm_extern.h>
     61 
     62 /*
     63  * Maximum process data and stack limits.
     64  * They are variables so they are patchable.
     65  */
     66 rlim_t maxdmap = MAXDSIZ;
     67 rlim_t maxsmap = MAXSSIZ;
     68 
     69 static pool_cache_t	plimit_cache;
     70 static pool_cache_t	pstats_cache;
     71 
     72 void
     73 resource_init(void)
     74 {
     75 
     76 	plimit_cache = pool_cache_init(sizeof(struct plimit), 0, 0, 0,
     77 	    "plimitpl", NULL, IPL_NONE, NULL, NULL, NULL);
     78 	pstats_cache = pool_cache_init(sizeof(struct pstats), 0, 0, 0,
     79 	    "pstatspl", NULL, IPL_NONE, NULL, NULL, NULL);
     80 }
     81 
     82 /*
     83  * Resource controls and accounting.
     84  */
     85 
     86 int
     87 sys_getpriority(struct lwp *l, const struct sys_getpriority_args *uap,
     88     register_t *retval)
     89 {
     90 	/* {
     91 		syscallarg(int) which;
     92 		syscallarg(id_t) who;
     93 	} */
     94 	struct proc *curp = l->l_proc, *p;
     95 	int low = NZERO + PRIO_MAX + 1;
     96 	int who = SCARG(uap, who);
     97 
     98 	mutex_enter(proc_lock);
     99 	switch (SCARG(uap, which)) {
    100 	case PRIO_PROCESS:
    101 		if (who == 0)
    102 			p = curp;
    103 		else
    104 			p = p_find(who, PFIND_LOCKED);
    105 		if (p != NULL)
    106 			low = p->p_nice;
    107 		break;
    108 
    109 	case PRIO_PGRP: {
    110 		struct pgrp *pg;
    111 
    112 		if (who == 0)
    113 			pg = curp->p_pgrp;
    114 		else if ((pg = pg_find(who, PFIND_LOCKED)) == NULL)
    115 			break;
    116 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
    117 			if (p->p_nice < low)
    118 				low = p->p_nice;
    119 		}
    120 		break;
    121 	}
    122 
    123 	case PRIO_USER:
    124 		if (who == 0)
    125 			who = (int)kauth_cred_geteuid(l->l_cred);
    126 		PROCLIST_FOREACH(p, &allproc) {
    127 			if ((p->p_flag & PK_MARKER) != 0)
    128 				continue;
    129 			mutex_enter(p->p_lock);
    130 			if (kauth_cred_geteuid(p->p_cred) ==
    131 			    (uid_t)who && p->p_nice < low)
    132 				low = p->p_nice;
    133 			mutex_exit(p->p_lock);
    134 		}
    135 		break;
    136 
    137 	default:
    138 		mutex_exit(proc_lock);
    139 		return (EINVAL);
    140 	}
    141 	mutex_exit(proc_lock);
    142 
    143 	if (low == NZERO + PRIO_MAX + 1)
    144 		return (ESRCH);
    145 	*retval = low - NZERO;
    146 	return (0);
    147 }
    148 
    149 /* ARGSUSED */
    150 int
    151 sys_setpriority(struct lwp *l, const struct sys_setpriority_args *uap,
    152     register_t *retval)
    153 {
    154 	/* {
    155 		syscallarg(int) which;
    156 		syscallarg(id_t) who;
    157 		syscallarg(int) prio;
    158 	} */
    159 	struct proc *curp = l->l_proc, *p;
    160 	int found = 0, error = 0;
    161 	int who = SCARG(uap, who);
    162 
    163 	mutex_enter(proc_lock);
    164 	switch (SCARG(uap, which)) {
    165 	case PRIO_PROCESS:
    166 		if (who == 0)
    167 			p = curp;
    168 		else
    169 			p = p_find(who, PFIND_LOCKED);
    170 		if (p != 0) {
    171 			mutex_enter(p->p_lock);
    172 			error = donice(l, p, SCARG(uap, prio));
    173 			mutex_exit(p->p_lock);
    174 			found++;
    175 		}
    176 		break;
    177 
    178 	case PRIO_PGRP: {
    179 		struct pgrp *pg;
    180 
    181 		if (who == 0)
    182 			pg = curp->p_pgrp;
    183 		else if ((pg = pg_find(who, PFIND_LOCKED)) == NULL)
    184 			break;
    185 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
    186 			mutex_enter(p->p_lock);
    187 			error = donice(l, p, SCARG(uap, prio));
    188 			mutex_exit(p->p_lock);
    189 			found++;
    190 		}
    191 		break;
    192 	}
    193 
    194 	case PRIO_USER:
    195 		if (who == 0)
    196 			who = (int)kauth_cred_geteuid(l->l_cred);
    197 		PROCLIST_FOREACH(p, &allproc) {
    198 			if ((p->p_flag & PK_MARKER) != 0)
    199 				continue;
    200 			mutex_enter(p->p_lock);
    201 			if (kauth_cred_geteuid(p->p_cred) ==
    202 			    (uid_t)SCARG(uap, who)) {
    203 				error = donice(l, p, SCARG(uap, prio));
    204 				found++;
    205 			}
    206 			mutex_exit(p->p_lock);
    207 		}
    208 		break;
    209 
    210 	default:
    211 		mutex_exit(proc_lock);
    212 		return EINVAL;
    213 	}
    214 	mutex_exit(proc_lock);
    215 	if (found == 0)
    216 		return (ESRCH);
    217 	return (error);
    218 }
    219 
    220 /*
    221  * Renice a process.
    222  *
    223  * Call with the target process' credentials locked.
    224  */
    225 int
    226 donice(struct lwp *l, struct proc *chgp, int n)
    227 {
    228 	kauth_cred_t cred = l->l_cred;
    229 
    230 	KASSERT(mutex_owned(chgp->p_lock));
    231 
    232 	if (n > PRIO_MAX)
    233 		n = PRIO_MAX;
    234 	if (n < PRIO_MIN)
    235 		n = PRIO_MIN;
    236 	n += NZERO;
    237 	if (kauth_authorize_process(cred, KAUTH_PROCESS_NICE, chgp,
    238 	    KAUTH_ARG(n), NULL, NULL))
    239 		return (EACCES);
    240 	sched_nice(chgp, n);
    241 	return (0);
    242 }
    243 
    244 /* ARGSUSED */
    245 int
    246 sys_setrlimit(struct lwp *l, const struct sys_setrlimit_args *uap,
    247     register_t *retval)
    248 {
    249 	/* {
    250 		syscallarg(int) which;
    251 		syscallarg(const struct rlimit *) rlp;
    252 	} */
    253 	int which = SCARG(uap, which);
    254 	struct rlimit alim;
    255 	int error;
    256 
    257 	error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit));
    258 	if (error)
    259 		return (error);
    260 	return (dosetrlimit(l, l->l_proc, which, &alim));
    261 }
    262 
    263 int
    264 dosetrlimit(struct lwp *l, struct proc *p, int which, struct rlimit *limp)
    265 {
    266 	struct rlimit *alimp;
    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 	if (limp->rlim_cur > limp->rlim_max) {
    276 		/*
    277 		 * This is programming error. According to SUSv2, we should
    278 		 * return error in this case.
    279 		 */
    280 		return (EINVAL);
    281 	}
    282 
    283 	alimp = &p->p_rlimit[which];
    284 	/* if we don't change the value, no need to limcopy() */
    285 	if (limp->rlim_cur == alimp->rlim_cur &&
    286 	    limp->rlim_max == alimp->rlim_max)
    287 		return 0;
    288 
    289 	error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT,
    290 	    p, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_SET), limp, KAUTH_ARG(which));
    291 	if (error)
    292 		return (error);
    293 
    294 	lim_privatise(p, false);
    295 	/* p->p_limit is now unchangeable */
    296 	alimp = &p->p_rlimit[which];
    297 
    298 	switch (which) {
    299 
    300 	case RLIMIT_DATA:
    301 		if (limp->rlim_cur > maxdmap)
    302 			limp->rlim_cur = maxdmap;
    303 		if (limp->rlim_max > maxdmap)
    304 			limp->rlim_max = maxdmap;
    305 		break;
    306 
    307 	case RLIMIT_STACK:
    308 		if (limp->rlim_cur > maxsmap)
    309 			limp->rlim_cur = maxsmap;
    310 		if (limp->rlim_max > maxsmap)
    311 			limp->rlim_max = maxsmap;
    312 
    313 		/*
    314 		 * Return EINVAL if the new stack size limit is lower than
    315 		 * current usage. Otherwise, the process would get SIGSEGV the
    316 		 * moment it would try to access anything on it's current stack.
    317 		 * This conforms to SUSv2.
    318 		 */
    319 		if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE
    320 		    || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE) {
    321 			return (EINVAL);
    322 		}
    323 
    324 		/*
    325 		 * Stack is allocated to the max at exec time with
    326 		 * only "rlim_cur" bytes accessible (In other words,
    327 		 * allocates stack dividing two contiguous regions at
    328 		 * "rlim_cur" bytes boundary).
    329 		 *
    330 		 * Since allocation is done in terms of page, roundup
    331 		 * "rlim_cur" (otherwise, contiguous regions
    332 		 * overlap).  If stack limit is going up make more
    333 		 * accessible, if going down make inaccessible.
    334 		 */
    335 		limp->rlim_cur = round_page(limp->rlim_cur);
    336 		if (limp->rlim_cur != alimp->rlim_cur) {
    337 			vaddr_t addr;
    338 			vsize_t size;
    339 			vm_prot_t prot;
    340 
    341 			if (limp->rlim_cur > alimp->rlim_cur) {
    342 				prot = VM_PROT_READ | VM_PROT_WRITE;
    343 				size = limp->rlim_cur - alimp->rlim_cur;
    344 				addr = (vaddr_t)p->p_vmspace->vm_minsaddr -
    345 				    limp->rlim_cur;
    346 			} else {
    347 				prot = VM_PROT_NONE;
    348 				size = alimp->rlim_cur - limp->rlim_cur;
    349 				addr = (vaddr_t)p->p_vmspace->vm_minsaddr -
    350 				     alimp->rlim_cur;
    351 			}
    352 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
    353 			    addr, addr+size, prot, false);
    354 		}
    355 		break;
    356 
    357 	case RLIMIT_NOFILE:
    358 		if (limp->rlim_cur > maxfiles)
    359 			limp->rlim_cur = maxfiles;
    360 		if (limp->rlim_max > maxfiles)
    361 			limp->rlim_max = maxfiles;
    362 		break;
    363 
    364 	case RLIMIT_NPROC:
    365 		if (limp->rlim_cur > maxproc)
    366 			limp->rlim_cur = maxproc;
    367 		if (limp->rlim_max > maxproc)
    368 			limp->rlim_max = maxproc;
    369 		break;
    370 	}
    371 
    372 	mutex_enter(&p->p_limit->pl_lock);
    373 	*alimp = *limp;
    374 	mutex_exit(&p->p_limit->pl_lock);
    375 	return (0);
    376 }
    377 
    378 /* ARGSUSED */
    379 int
    380 sys_getrlimit(struct lwp *l, const struct sys_getrlimit_args *uap,
    381     register_t *retval)
    382 {
    383 	/* {
    384 		syscallarg(int) which;
    385 		syscallarg(struct rlimit *) rlp;
    386 	} */
    387 	struct proc *p = l->l_proc;
    388 	int which = SCARG(uap, which);
    389 	struct rlimit rl;
    390 
    391 	if ((u_int)which >= RLIM_NLIMITS)
    392 		return (EINVAL);
    393 
    394 	mutex_enter(p->p_lock);
    395 	memcpy(&rl, &p->p_rlimit[which], sizeof(rl));
    396 	mutex_exit(p->p_lock);
    397 
    398 	return copyout(&rl, SCARG(uap, rlp), sizeof(rl));
    399 }
    400 
    401 /*
    402  * Transform the running time and tick information in proc p into user,
    403  * system, and interrupt time usage.
    404  *
    405  * Should be called with p->p_lock held unless called from exit1().
    406  */
    407 void
    408 calcru(struct proc *p, struct timeval *up, struct timeval *sp,
    409     struct timeval *ip, struct timeval *rp)
    410 {
    411 	uint64_t u, st, ut, it, tot;
    412 	struct lwp *l;
    413 	struct bintime tm;
    414 	struct timeval tv;
    415 
    416 	mutex_spin_enter(&p->p_stmutex);
    417 	st = p->p_sticks;
    418 	ut = p->p_uticks;
    419 	it = p->p_iticks;
    420 	mutex_spin_exit(&p->p_stmutex);
    421 
    422 	tm = p->p_rtime;
    423 
    424 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
    425 		lwp_lock(l);
    426 		bintime_add(&tm, &l->l_rtime);
    427 		if ((l->l_pflag & LP_RUNNING) != 0) {
    428 			struct bintime diff;
    429 			/*
    430 			 * Adjust for the current time slice.  This is
    431 			 * actually fairly important since the error
    432 			 * here is on the order of a time quantum,
    433 			 * which is much greater than the sampling
    434 			 * error.
    435 			 */
    436 			binuptime(&diff);
    437 			bintime_sub(&diff, &l->l_stime);
    438 			bintime_add(&tm, &diff);
    439 		}
    440 		lwp_unlock(l);
    441 	}
    442 
    443 	tot = st + ut + it;
    444 	bintime2timeval(&tm, &tv);
    445 	u = (uint64_t)tv.tv_sec * 1000000ul + tv.tv_usec;
    446 
    447 	if (tot == 0) {
    448 		/* No ticks, so can't use to share time out, split 50-50 */
    449 		st = ut = u / 2;
    450 	} else {
    451 		st = (u * st) / tot;
    452 		ut = (u * ut) / tot;
    453 	}
    454 	if (sp != NULL) {
    455 		sp->tv_sec = st / 1000000;
    456 		sp->tv_usec = st % 1000000;
    457 	}
    458 	if (up != NULL) {
    459 		up->tv_sec = ut / 1000000;
    460 		up->tv_usec = ut % 1000000;
    461 	}
    462 	if (ip != NULL) {
    463 		if (it != 0)
    464 			it = (u * it) / tot;
    465 		ip->tv_sec = it / 1000000;
    466 		ip->tv_usec = it % 1000000;
    467 	}
    468 	if (rp != NULL) {
    469 		*rp = tv;
    470 	}
    471 }
    472 
    473 /* ARGSUSED */
    474 int
    475 sys___getrusage50(struct lwp *l, const struct sys___getrusage50_args *uap,
    476     register_t *retval)
    477 {
    478 	/* {
    479 		syscallarg(int) who;
    480 		syscallarg(struct rusage *) rusage;
    481 	} */
    482 	struct rusage ru;
    483 	struct proc *p = l->l_proc;
    484 
    485 	switch (SCARG(uap, who)) {
    486 	case RUSAGE_SELF:
    487 		mutex_enter(p->p_lock);
    488 		memcpy(&ru, &p->p_stats->p_ru, sizeof(ru));
    489 		calcru(p, &ru.ru_utime, &ru.ru_stime, NULL, NULL);
    490 		rulwps(p, &ru);
    491 		mutex_exit(p->p_lock);
    492 		break;
    493 
    494 	case RUSAGE_CHILDREN:
    495 		mutex_enter(p->p_lock);
    496 		memcpy(&ru, &p->p_stats->p_cru, sizeof(ru));
    497 		mutex_exit(p->p_lock);
    498 		break;
    499 
    500 	default:
    501 		return EINVAL;
    502 	}
    503 
    504 	return copyout(&ru, SCARG(uap, rusage), sizeof(ru));
    505 }
    506 
    507 void
    508 ruadd(struct rusage *ru, struct rusage *ru2)
    509 {
    510 	long *ip, *ip2;
    511 	int i;
    512 
    513 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
    514 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
    515 	if (ru->ru_maxrss < ru2->ru_maxrss)
    516 		ru->ru_maxrss = ru2->ru_maxrss;
    517 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
    518 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
    519 		*ip++ += *ip2++;
    520 }
    521 
    522 void
    523 rulwps(proc_t *p, struct rusage *ru)
    524 {
    525 	lwp_t *l;
    526 
    527 	KASSERT(mutex_owned(p->p_lock));
    528 
    529 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
    530 		ruadd(ru, &l->l_ru);
    531 		ru->ru_nvcsw += (l->l_ncsw - l->l_nivcsw);
    532 		ru->ru_nivcsw += l->l_nivcsw;
    533 	}
    534 }
    535 
    536 /*
    537  * Make a copy of the plimit structure.
    538  * We share these structures copy-on-write after fork,
    539  * and copy when a limit is changed.
    540  *
    541  * Unfortunately (due to PL_SHAREMOD) it is possibly for the structure
    542  * we are copying to change beneath our feet!
    543  */
    544 struct plimit *
    545 lim_copy(struct plimit *lim)
    546 {
    547 	struct plimit *newlim;
    548 	char *corename;
    549 	size_t alen, len;
    550 
    551 	newlim = pool_cache_get(plimit_cache, PR_WAITOK);
    552 	mutex_init(&newlim->pl_lock, MUTEX_DEFAULT, IPL_NONE);
    553 	newlim->pl_flags = 0;
    554 	newlim->pl_refcnt = 1;
    555 	newlim->pl_sv_limit = NULL;
    556 
    557 	mutex_enter(&lim->pl_lock);
    558 	memcpy(newlim->pl_rlimit, lim->pl_rlimit,
    559 	    sizeof(struct rlimit) * RLIM_NLIMITS);
    560 
    561 	alen = 0;
    562 	corename = NULL;
    563 	for (;;) {
    564 		if (lim->pl_corename == defcorename) {
    565 			newlim->pl_corename = defcorename;
    566 			break;
    567 		}
    568 		len = strlen(lim->pl_corename) + 1;
    569 		if (len <= alen) {
    570 			newlim->pl_corename = corename;
    571 			memcpy(corename, lim->pl_corename, len);
    572 			corename = NULL;
    573 			break;
    574 		}
    575 		mutex_exit(&lim->pl_lock);
    576 		if (corename != NULL)
    577 			free(corename, M_TEMP);
    578 		alen = len;
    579 		corename = malloc(alen, M_TEMP, M_WAITOK);
    580 		mutex_enter(&lim->pl_lock);
    581 	}
    582 	mutex_exit(&lim->pl_lock);
    583 	if (corename != NULL)
    584 		free(corename, M_TEMP);
    585 	return newlim;
    586 }
    587 
    588 void
    589 lim_addref(struct plimit *lim)
    590 {
    591 	atomic_inc_uint(&lim->pl_refcnt);
    592 }
    593 
    594 /*
    595  * Give a process it's own private plimit structure.
    596  * This will only be shared (in fork) if modifications are to be shared.
    597  */
    598 void
    599 lim_privatise(struct proc *p, bool set_shared)
    600 {
    601 	struct plimit *lim, *newlim;
    602 
    603 	lim = p->p_limit;
    604 	if (lim->pl_flags & PL_WRITEABLE) {
    605 		if (set_shared)
    606 			lim->pl_flags |= PL_SHAREMOD;
    607 		return;
    608 	}
    609 
    610 	if (set_shared && lim->pl_flags & PL_SHAREMOD)
    611 		return;
    612 
    613 	newlim = lim_copy(lim);
    614 
    615 	mutex_enter(p->p_lock);
    616 	if (p->p_limit->pl_flags & PL_WRITEABLE) {
    617 		/* Someone crept in while we were busy */
    618 		mutex_exit(p->p_lock);
    619 		limfree(newlim);
    620 		if (set_shared)
    621 			p->p_limit->pl_flags |= PL_SHAREMOD;
    622 		return;
    623 	}
    624 
    625 	/*
    626 	 * Since most accesses to p->p_limit aren't locked, we must not
    627 	 * delete the old limit structure yet.
    628 	 */
    629 	newlim->pl_sv_limit = p->p_limit;
    630 	newlim->pl_flags |= PL_WRITEABLE;
    631 	if (set_shared)
    632 		newlim->pl_flags |= PL_SHAREMOD;
    633 	p->p_limit = newlim;
    634 	mutex_exit(p->p_lock);
    635 }
    636 
    637 void
    638 limfree(struct plimit *lim)
    639 {
    640 	struct plimit *sv_lim;
    641 
    642 	do {
    643 		if (atomic_dec_uint_nv(&lim->pl_refcnt) > 0)
    644 			return;
    645 		if (lim->pl_corename != defcorename)
    646 			free(lim->pl_corename, M_TEMP);
    647 		sv_lim = lim->pl_sv_limit;
    648 		mutex_destroy(&lim->pl_lock);
    649 		pool_cache_put(plimit_cache, lim);
    650 	} while ((lim = sv_lim) != NULL);
    651 }
    652 
    653 struct pstats *
    654 pstatscopy(struct pstats *ps)
    655 {
    656 
    657 	struct pstats *newps;
    658 
    659 	newps = pool_cache_get(pstats_cache, PR_WAITOK);
    660 
    661 	memset(&newps->pstat_startzero, 0,
    662 	(unsigned) ((char *)&newps->pstat_endzero -
    663 		    (char *)&newps->pstat_startzero));
    664 	memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy,
    665 	((char *)&newps->pstat_endcopy -
    666 	 (char *)&newps->pstat_startcopy));
    667 
    668 	return (newps);
    669 
    670 }
    671 
    672 void
    673 pstatsfree(struct pstats *ps)
    674 {
    675 
    676 	pool_cache_put(pstats_cache, ps);
    677 }
    678 
    679 /*
    680  * sysctl interface in five parts
    681  */
    682 
    683 /*
    684  * a routine for sysctl proc subtree helpers that need to pick a valid
    685  * process by pid.
    686  */
    687 static int
    688 sysctl_proc_findproc(struct lwp *l, struct proc **p2, pid_t pid)
    689 {
    690 	struct proc *ptmp;
    691 	int error = 0;
    692 
    693 	if (pid == PROC_CURPROC)
    694 		ptmp = l->l_proc;
    695 	else if ((ptmp = pfind(pid)) == NULL)
    696 		error = ESRCH;
    697 
    698 	*p2 = ptmp;
    699 	return (error);
    700 }
    701 
    702 /*
    703  * sysctl helper routine for setting a process's specific corefile
    704  * name.  picks the process based on the given pid and checks the
    705  * correctness of the new value.
    706  */
    707 static int
    708 sysctl_proc_corename(SYSCTLFN_ARGS)
    709 {
    710 	struct proc *ptmp;
    711 	struct plimit *lim;
    712 	int error = 0, len;
    713 	char *cname;
    714 	char *ocore;
    715 	char *tmp;
    716 	struct sysctlnode node;
    717 
    718 	/*
    719 	 * is this all correct?
    720 	 */
    721 	if (namelen != 0)
    722 		return (EINVAL);
    723 	if (name[-1] != PROC_PID_CORENAME)
    724 		return (EINVAL);
    725 
    726 	/*
    727 	 * whom are we tweaking?
    728 	 */
    729 	error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]);
    730 	if (error)
    731 		return (error);
    732 
    733 	/* XXX-elad */
    734 	error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, ptmp,
    735 	    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL);
    736 	if (error)
    737 		return (error);
    738 
    739 	if (newp == NULL) {
    740 		error = kauth_authorize_process(l->l_cred,
    741 		    KAUTH_PROCESS_CORENAME, ptmp,
    742 		    KAUTH_ARG(KAUTH_REQ_PROCESS_CORENAME_GET), NULL, NULL);
    743 		if (error)
    744 			return (error);
    745 	}
    746 
    747 	/*
    748 	 * let them modify a temporary copy of the core name
    749 	 */
    750 	cname = PNBUF_GET();
    751 	lim = ptmp->p_limit;
    752 	mutex_enter(&lim->pl_lock);
    753 	strlcpy(cname, lim->pl_corename, MAXPATHLEN);
    754 	mutex_exit(&lim->pl_lock);
    755 
    756 	node = *rnode;
    757 	node.sysctl_data = cname;
    758 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
    759 
    760 	/*
    761 	 * if that failed, or they have nothing new to say, or we've
    762 	 * heard it before...
    763 	 */
    764 	if (error || newp == NULL)
    765 		goto done;
    766 	lim = ptmp->p_limit;
    767 	mutex_enter(&lim->pl_lock);
    768 	error = strcmp(cname, lim->pl_corename);
    769 	mutex_exit(&lim->pl_lock);
    770 	if (error == 0)
    771 		/* Unchanged */
    772 		goto done;
    773 
    774 	error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CORENAME,
    775 	    ptmp, KAUTH_ARG(KAUTH_REQ_PROCESS_CORENAME_SET), cname, NULL);
    776 	if (error)
    777 		return (error);
    778 
    779 	/*
    780 	 * no error yet and cname now has the new core name in it.
    781 	 * let's see if it looks acceptable.  it must be either "core"
    782 	 * or end in ".core" or "/core".
    783 	 */
    784 	len = strlen(cname);
    785 	if (len < 4) {
    786 		error = EINVAL;
    787 	} else if (strcmp(cname + len - 4, "core") != 0) {
    788 		error = EINVAL;
    789 	} else if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.') {
    790 		error = EINVAL;
    791 	}
    792 	if (error != 0) {
    793 		goto done;
    794 	}
    795 
    796 	/*
    797 	 * hmm...looks good.  now...where do we put it?
    798 	 */
    799 	tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL);
    800 	if (tmp == NULL) {
    801 		error = ENOMEM;
    802 		goto done;
    803 	}
    804 	memcpy(tmp, cname, len + 1);
    805 
    806 	lim_privatise(ptmp, false);
    807 	lim = ptmp->p_limit;
    808 	mutex_enter(&lim->pl_lock);
    809 	ocore = lim->pl_corename;
    810 	lim->pl_corename = tmp;
    811 	mutex_exit(&lim->pl_lock);
    812 	if (ocore != defcorename)
    813 		free(ocore, M_TEMP);
    814 
    815 done:
    816 	PNBUF_PUT(cname);
    817 	return error;
    818 }
    819 
    820 /*
    821  * sysctl helper routine for checking/setting a process's stop flags,
    822  * one for fork and one for exec.
    823  */
    824 static int
    825 sysctl_proc_stop(SYSCTLFN_ARGS)
    826 {
    827 	struct proc *ptmp;
    828 	int i, f, error = 0;
    829 	struct sysctlnode node;
    830 
    831 	if (namelen != 0)
    832 		return (EINVAL);
    833 
    834 	error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]);
    835 	if (error)
    836 		return (error);
    837 
    838 	/* XXX-elad */
    839 	error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, ptmp,
    840 	    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL);
    841 	if (error)
    842 		return (error);
    843 
    844 	switch (rnode->sysctl_num) {
    845 	case PROC_PID_STOPFORK:
    846 		f = PS_STOPFORK;
    847 		break;
    848 	case PROC_PID_STOPEXEC:
    849 		f = PS_STOPEXEC;
    850 		break;
    851 	case PROC_PID_STOPEXIT:
    852 		f = PS_STOPEXIT;
    853 		break;
    854 	default:
    855 		return (EINVAL);
    856 	}
    857 
    858 	i = (ptmp->p_flag & f) ? 1 : 0;
    859 	node = *rnode;
    860 	node.sysctl_data = &i;
    861 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
    862 	if (error || newp == NULL)
    863 		return (error);
    864 
    865 	mutex_enter(ptmp->p_lock);
    866 	error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_STOPFLAG,
    867 	    ptmp, KAUTH_ARG(f), NULL, NULL);
    868 	if (!error) {
    869 		if (i) {
    870 			ptmp->p_sflag |= f;
    871 		} else {
    872 			ptmp->p_sflag &= ~f;
    873 		}
    874 	}
    875 	mutex_exit(ptmp->p_lock);
    876 
    877 	return error;
    878 }
    879 
    880 /*
    881  * sysctl helper routine for a process's rlimits as exposed by sysctl.
    882  */
    883 static int
    884 sysctl_proc_plimit(SYSCTLFN_ARGS)
    885 {
    886 	struct proc *ptmp;
    887 	u_int limitno;
    888 	int which, error = 0;
    889         struct rlimit alim;
    890 	struct sysctlnode node;
    891 
    892 	if (namelen != 0)
    893 		return (EINVAL);
    894 
    895 	which = name[-1];
    896 	if (which != PROC_PID_LIMIT_TYPE_SOFT &&
    897 	    which != PROC_PID_LIMIT_TYPE_HARD)
    898 		return (EINVAL);
    899 
    900 	limitno = name[-2] - 1;
    901 	if (limitno >= RLIM_NLIMITS)
    902 		return (EINVAL);
    903 
    904 	if (name[-3] != PROC_PID_LIMIT)
    905 		return (EINVAL);
    906 
    907 	error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-4]);
    908 	if (error)
    909 		return (error);
    910 
    911 	/* XXX-elad */
    912 	error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, ptmp,
    913 	    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL);
    914 	if (error)
    915 		return (error);
    916 
    917 	/* Check if we can view limits. */
    918 	if (newp == NULL) {
    919 		error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RLIMIT,
    920 		    ptmp, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_GET), &alim,
    921 		    KAUTH_ARG(which));
    922 		if (error)
    923 			return (error);
    924 	}
    925 
    926 	node = *rnode;
    927 	memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim));
    928 	if (which == PROC_PID_LIMIT_TYPE_HARD)
    929 		node.sysctl_data = &alim.rlim_max;
    930 	else
    931 		node.sysctl_data = &alim.rlim_cur;
    932 
    933 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
    934 	if (error || newp == NULL)
    935 		return (error);
    936 
    937 	return (dosetrlimit(l, ptmp, limitno, &alim));
    938 }
    939 
    940 /*
    941  * and finally, the actually glue that sticks it to the tree
    942  */
    943 SYSCTL_SETUP(sysctl_proc_setup, "sysctl proc subtree setup")
    944 {
    945 
    946 	sysctl_createv(clog, 0, NULL, NULL,
    947 		       CTLFLAG_PERMANENT,
    948 		       CTLTYPE_NODE, "proc", NULL,
    949 		       NULL, 0, NULL, 0,
    950 		       CTL_PROC, CTL_EOL);
    951 	sysctl_createv(clog, 0, NULL, NULL,
    952 		       CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER,
    953 		       CTLTYPE_NODE, "curproc",
    954 		       SYSCTL_DESCR("Per-process settings"),
    955 		       NULL, 0, NULL, 0,
    956 		       CTL_PROC, PROC_CURPROC, CTL_EOL);
    957 
    958 	sysctl_createv(clog, 0, NULL, NULL,
    959 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
    960 		       CTLTYPE_STRING, "corename",
    961 		       SYSCTL_DESCR("Core file name"),
    962 		       sysctl_proc_corename, 0, NULL, MAXPATHLEN,
    963 		       CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL);
    964 	sysctl_createv(clog, 0, NULL, NULL,
    965 		       CTLFLAG_PERMANENT,
    966 		       CTLTYPE_NODE, "rlimit",
    967 		       SYSCTL_DESCR("Process limits"),
    968 		       NULL, 0, NULL, 0,
    969 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL);
    970 
    971 #define create_proc_plimit(s, n) do {					\
    972 	sysctl_createv(clog, 0, NULL, NULL,				\
    973 		       CTLFLAG_PERMANENT,				\
    974 		       CTLTYPE_NODE, s,					\
    975 		       SYSCTL_DESCR("Process " s " limits"),		\
    976 		       NULL, 0, NULL, 0,				\
    977 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
    978 		       CTL_EOL);					\
    979 	sysctl_createv(clog, 0, NULL, NULL,				\
    980 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \
    981 		       CTLTYPE_QUAD, "soft",				\
    982 		       SYSCTL_DESCR("Process soft " s " limit"),	\
    983 		       sysctl_proc_plimit, 0, NULL, 0,			\
    984 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
    985 		       PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL);		\
    986 	sysctl_createv(clog, 0, NULL, NULL,				\
    987 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \
    988 		       CTLTYPE_QUAD, "hard",				\
    989 		       SYSCTL_DESCR("Process hard " s " limit"),	\
    990 		       sysctl_proc_plimit, 0, NULL, 0,			\
    991 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
    992 		       PROC_PID_LIMIT_TYPE_HARD, CTL_EOL);		\
    993 	} while (0/*CONSTCOND*/)
    994 
    995 	create_proc_plimit("cputime",		PROC_PID_LIMIT_CPU);
    996 	create_proc_plimit("filesize",		PROC_PID_LIMIT_FSIZE);
    997 	create_proc_plimit("datasize",		PROC_PID_LIMIT_DATA);
    998 	create_proc_plimit("stacksize",		PROC_PID_LIMIT_STACK);
    999 	create_proc_plimit("coredumpsize",	PROC_PID_LIMIT_CORE);
   1000 	create_proc_plimit("memoryuse",		PROC_PID_LIMIT_RSS);
   1001 	create_proc_plimit("memorylocked",	PROC_PID_LIMIT_MEMLOCK);
   1002 	create_proc_plimit("maxproc",		PROC_PID_LIMIT_NPROC);
   1003 	create_proc_plimit("descriptors",	PROC_PID_LIMIT_NOFILE);
   1004 	create_proc_plimit("sbsize",		PROC_PID_LIMIT_SBSIZE);
   1005 
   1006 #undef create_proc_plimit
   1007 
   1008 	sysctl_createv(clog, 0, NULL, NULL,
   1009 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
   1010 		       CTLTYPE_INT, "stopfork",
   1011 		       SYSCTL_DESCR("Stop process at fork(2)"),
   1012 		       sysctl_proc_stop, 0, NULL, 0,
   1013 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL);
   1014 	sysctl_createv(clog, 0, NULL, NULL,
   1015 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
   1016 		       CTLTYPE_INT, "stopexec",
   1017 		       SYSCTL_DESCR("Stop process at execve(2)"),
   1018 		       sysctl_proc_stop, 0, NULL, 0,
   1019 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL);
   1020 	sysctl_createv(clog, 0, NULL, NULL,
   1021 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
   1022 		       CTLTYPE_INT, "stopexit",
   1023 		       SYSCTL_DESCR("Stop process before completing exit"),
   1024 		       sysctl_proc_stop, 0, NULL, 0,
   1025 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL);
   1026 }
   1027