Home | History | Annotate | Line # | Download | only in kern
kern_resource.c revision 1.103.4.5
      1 /*	$NetBSD: kern_resource.c,v 1.103.4.5 2006/11/18 21:39:22 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 1.103.4.5 2006/11/18 21:39:22 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_mutex);
    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_mutex);
    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_mutex);
    162 			error = donice(l, p, SCARG(uap, prio));
    163 			mutex_exit(&p->p_mutex);
    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_mutex);
    180 			error = donice(l, p, SCARG(uap, prio));
    181 			mutex_exit(&p->p_mutex);
    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_mutex);
    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_mutex);
    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 onice;
    222 
    223 	LOCK_ASSERT(mutex_owned(&chgp->p_mutex));
    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 
    235   again:
    236 	if (n < (onice = chgp->p_nice) && kauth_authorize_process(cred,
    237 	    KAUTH_PROCESS_RESOURCE, chgp, (void *)KAUTH_REQ_PROCESS_RESOURCE_NICE,
    238 	    (void *)(u_long)n, NULL))
    239 		return (EACCES);
    240 	mutex_enter(&chgp->p_smutex);
    241 	if (onice != chgp->p_nice) {
    242 		mutex_exit(&chgp->p_smutex);
    243 		goto again;
    244 	}
    245 	chgp->p_nice = n;
    246 	(void)resetprocpriority(chgp);
    247 	mutex_exit(&chgp->p_smutex);
    248 	return (0);
    249 }
    250 
    251 /* ARGSUSED */
    252 int
    253 sys_setrlimit(struct lwp *l, void *v, register_t *retval)
    254 {
    255 	struct sys_setrlimit_args /* {
    256 		syscallarg(int) which;
    257 		syscallarg(const struct rlimit *) rlp;
    258 	} */ *uap = v;
    259 	int which = SCARG(uap, which);
    260 	struct rlimit alim;
    261 	int error;
    262 
    263 	error = copyin(SCARG(uap, rlp), &alim, sizeof(struct rlimit));
    264 	if (error)
    265 		return (error);
    266 	return (dosetrlimit(l, l->l_proc, which, &alim));
    267 }
    268 
    269 int
    270 dosetrlimit(struct lwp *l, struct proc *p, int which, struct rlimit *limp)
    271 {
    272 	struct rlimit *alimp;
    273 	struct plimit *oldplim;
    274 	int error;
    275 
    276 	if ((u_int)which >= RLIM_NLIMITS)
    277 		return (EINVAL);
    278 
    279 	if (limp->rlim_cur < 0 || limp->rlim_max < 0)
    280 		return (EINVAL);
    281 
    282 	alimp = &p->p_rlimit[which];
    283 	/* if we don't change the value, no need to limcopy() */
    284 	if (limp->rlim_cur == alimp->rlim_cur &&
    285 	    limp->rlim_max == alimp->rlim_max)
    286 		return 0;
    287 
    288 	if (limp->rlim_cur > limp->rlim_max) {
    289 		/*
    290 		 * This is programming error. According to SUSv2, we should
    291 		 * return error in this case.
    292 		 */
    293 		return (EINVAL);
    294 	}
    295 	if (limp->rlim_max > alimp->rlim_max && (error =
    296 	    kauth_authorize_process(l->l_cred, KAUTH_PROCESS_RESOURCE,
    297 	    p, (void *)KAUTH_REQ_PROCESS_RESOURCE_RLIMIT, limp,
    298 	    (void *)(u_long)which)))
    299 			return (error);
    300 
    301 	if (p->p_limit->p_refcnt > 1 &&
    302 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
    303 		p->p_limit = limcopy(oldplim = p->p_limit);
    304 		limfree(oldplim);
    305 		alimp = &p->p_rlimit[which];
    306 	}
    307 
    308 	switch (which) {
    309 
    310 	case RLIMIT_DATA:
    311 		if (limp->rlim_cur > maxdmap)
    312 			limp->rlim_cur = maxdmap;
    313 		if (limp->rlim_max > maxdmap)
    314 			limp->rlim_max = maxdmap;
    315 		break;
    316 
    317 	case RLIMIT_STACK:
    318 		if (limp->rlim_cur > maxsmap)
    319 			limp->rlim_cur = maxsmap;
    320 		if (limp->rlim_max > maxsmap)
    321 			limp->rlim_max = maxsmap;
    322 
    323 		/*
    324 		 * Return EINVAL if the new stack size limit is lower than
    325 		 * current usage. Otherwise, the process would get SIGSEGV the
    326 		 * moment it would try to access anything on it's current stack.
    327 		 * This conforms to SUSv2.
    328 		 */
    329 		if (limp->rlim_cur < p->p_vmspace->vm_ssize * PAGE_SIZE
    330 		    || limp->rlim_max < p->p_vmspace->vm_ssize * PAGE_SIZE)
    331 			return (EINVAL);
    332 
    333 		/*
    334 		 * Stack is allocated to the max at exec time with
    335 		 * only "rlim_cur" bytes accessible (In other words,
    336 		 * allocates stack dividing two contiguous regions at
    337 		 * "rlim_cur" bytes boundary).
    338 		 *
    339 		 * Since allocation is done in terms of page, roundup
    340 		 * "rlim_cur" (otherwise, contiguous regions
    341 		 * overlap).  If stack limit is going up make more
    342 		 * accessible, if going down make inaccessible.
    343 		 */
    344 		limp->rlim_cur = round_page(limp->rlim_cur);
    345 		if (limp->rlim_cur != alimp->rlim_cur) {
    346 			vaddr_t addr;
    347 			vsize_t size;
    348 			vm_prot_t prot;
    349 
    350 			if (limp->rlim_cur > alimp->rlim_cur) {
    351 				prot = VM_PROT_READ | VM_PROT_WRITE;
    352 				size = limp->rlim_cur - alimp->rlim_cur;
    353 				addr = (vaddr_t)p->p_vmspace->vm_minsaddr -
    354 				    limp->rlim_cur;
    355 			} else {
    356 				prot = VM_PROT_NONE;
    357 				size = alimp->rlim_cur - limp->rlim_cur;
    358 				addr = (vaddr_t)p->p_vmspace->vm_minsaddr -
    359 				     alimp->rlim_cur;
    360 			}
    361 			(void) uvm_map_protect(&p->p_vmspace->vm_map,
    362 			    addr, addr+size, prot, FALSE);
    363 		}
    364 		break;
    365 
    366 	case RLIMIT_NOFILE:
    367 		if (limp->rlim_cur > maxfiles)
    368 			limp->rlim_cur = maxfiles;
    369 		if (limp->rlim_max > maxfiles)
    370 			limp->rlim_max = maxfiles;
    371 		break;
    372 
    373 	case RLIMIT_NPROC:
    374 		if (limp->rlim_cur > maxproc)
    375 			limp->rlim_cur = maxproc;
    376 		if (limp->rlim_max > maxproc)
    377 			limp->rlim_max = maxproc;
    378 		break;
    379 	}
    380 	*alimp = *limp;
    381 	return (0);
    382 }
    383 
    384 /* ARGSUSED */
    385 int
    386 sys_getrlimit(struct lwp *l, void *v, register_t *retval)
    387 {
    388 	struct sys_getrlimit_args /* {
    389 		syscallarg(int) which;
    390 		syscallarg(struct rlimit *) rlp;
    391 	} */ *uap = v;
    392 	struct proc *p = l->l_proc;
    393 	int which = SCARG(uap, which);
    394 
    395 	if ((u_int)which >= RLIM_NLIMITS)
    396 		return (EINVAL);
    397 	return (copyout(&p->p_rlimit[which], SCARG(uap, rlp),
    398 	    sizeof(struct rlimit)));
    399 }
    400 
    401 /*
    402  * Transform the running time and tick information in proc p into user,
    403  * system, and interrupt time usage.
    404  */
    405 void
    406 calcru(struct proc *p, struct timeval *up, struct timeval *sp,
    407     struct timeval *ip, struct timeval *rp)
    408 {
    409 	u_quad_t u, st, ut, it, tot;
    410 	unsigned long sec;
    411 	long usec;
    412 	int s;
    413 	struct timeval tv;
    414 	struct lwp *l;
    415 
    416 	LOCK_ASSERT(mutex_owned(&p->p_smutex));
    417 
    418 	s = splstatclock();
    419 	st = p->p_sticks;
    420 	ut = p->p_uticks;
    421 	it = p->p_iticks;
    422 	splx(s);
    423 
    424 	sec = 0;
    425 	usec = 0;
    426 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
    427 		lwp_lock(l);
    428 		sec += l->l_rtime.tv_sec;
    429 		usec += l->l_rtime.tv_usec;
    430 		if (l->l_stat == LSONPROC) {
    431 			struct schedstate_percpu *spc;
    432 
    433 			KDASSERT(l->l_cpu != NULL);
    434 			spc = &l->l_cpu->ci_schedstate;
    435 
    436 			/*
    437 			 * Adjust for the current time slice.  This is
    438 			 * actually fairly important since the error
    439 			 * here is on the order of a time quantum,
    440 			 * which is much greater than the sampling
    441 			 * error.
    442 			 */
    443 			microtime(&tv);
    444 			sec += tv.tv_sec - spc->spc_runtime.tv_sec;
    445 			usec += tv.tv_usec - spc->spc_runtime.tv_usec;
    446 		}
    447 		lwp_unlock(l);
    448 	}
    449 
    450 	tot = st + ut + it;
    451 	u = sec * 1000000ull + usec;
    452 
    453 	if (tot == 0) {
    454 		/* No ticks, so can't use to share time out, split 50-50 */
    455 		st = ut = u / 2;
    456 	} else {
    457 		st = (u * st) / tot;
    458 		ut = (u * ut) / tot;
    459 	}
    460 	if (sp != NULL) {
    461 		if (tot == 0) {
    462 			/* No ticks, so can't use to share time out, split 50-50 */
    463 			st = u / 2;
    464 		} else
    465 			st = (u * st) / tot;
    466 		sp->tv_sec = st / 1000000;
    467 		sp->tv_usec = st % 1000000;
    468 	}
    469 	if (up != NULL) {
    470 		if (tot == 0) {
    471 			/* No ticks, so can't use to share time out, split 50-50 */
    472 			ut = u / 2;
    473 		} else
    474 			ut = (u * ut) / tot;
    475 		up->tv_sec = ut / 1000000;
    476 		up->tv_usec = ut % 1000000;
    477 	}
    478 	if (ip != NULL) {
    479 		if (it != 0)
    480 			it = (u * it) / tot;
    481 		ip->tv_sec = it / 1000000;
    482 		ip->tv_usec = it % 1000000;
    483 	}
    484 	if (rp != NULL) {
    485 		rp->tv_sec = sec;
    486 		rp->tv_usec = usec;
    487 	}
    488 }
    489 
    490 /* ARGSUSED */
    491 int
    492 sys_getrusage(struct lwp *l, void *v, register_t *retval)
    493 {
    494 	struct sys_getrusage_args /* {
    495 		syscallarg(int) who;
    496 		syscallarg(struct rusage *) rusage;
    497 	} */ *uap = v;
    498 	struct rusage *rup;
    499 	struct proc *p = l->l_proc;
    500 
    501 	switch (SCARG(uap, who)) {
    502 
    503 	case RUSAGE_SELF:
    504 		rup = &p->p_stats->p_ru;
    505 		mutex_enter(&p->p_smutex);
    506 		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL, NULL);
    507 		mutex_exit(&p->p_smutex);
    508 		break;
    509 
    510 	case RUSAGE_CHILDREN:
    511 		rup = &p->p_stats->p_cru;
    512 		break;
    513 
    514 	default:
    515 		return (EINVAL);
    516 	}
    517 	return (copyout(rup, SCARG(uap, rusage), sizeof(struct rusage)));
    518 }
    519 
    520 void
    521 ruadd(struct rusage *ru, struct rusage *ru2)
    522 {
    523 	long *ip, *ip2;
    524 	int i;
    525 
    526 	timeradd(&ru->ru_utime, &ru2->ru_utime, &ru->ru_utime);
    527 	timeradd(&ru->ru_stime, &ru2->ru_stime, &ru->ru_stime);
    528 	if (ru->ru_maxrss < ru2->ru_maxrss)
    529 		ru->ru_maxrss = ru2->ru_maxrss;
    530 	ip = &ru->ru_first; ip2 = &ru2->ru_first;
    531 	for (i = &ru->ru_last - &ru->ru_first; i >= 0; i--)
    532 		*ip++ += *ip2++;
    533 }
    534 
    535 /*
    536  * Make a copy of the plimit structure.
    537  * We share these structures copy-on-write after fork,
    538  * and copy when a limit is changed.
    539  */
    540 struct plimit *
    541 limcopy(struct plimit *lim)
    542 {
    543 	struct plimit *newlim;
    544 	size_t l = 0;
    545 
    546 	simple_lock(&lim->p_slock);
    547 	if (lim->pl_corename != defcorename)
    548 		l = strlen(lim->pl_corename) + 1;
    549 	simple_unlock(&lim->p_slock);
    550 
    551 	newlim = pool_get(&plimit_pool, PR_WAITOK);
    552 	simple_lock_init(&newlim->p_slock);
    553 	newlim->p_lflags = 0;
    554 	newlim->p_refcnt = 1;
    555 	newlim->pl_corename = (l != 0)
    556 		? malloc(l, M_TEMP, M_WAITOK)
    557 		: defcorename;
    558 
    559 	simple_lock(&lim->p_slock);
    560 	memcpy(newlim->pl_rlimit, lim->pl_rlimit,
    561 	    sizeof(struct rlimit) * RLIM_NLIMITS);
    562 
    563 	if (l != 0)
    564 		strlcpy(newlim->pl_corename, lim->pl_corename, l);
    565 	simple_unlock(&lim->p_slock);
    566 
    567 	return (newlim);
    568 }
    569 
    570 void
    571 limfree(struct plimit *lim)
    572 {
    573 	int n;
    574 
    575 	simple_lock(&lim->p_slock);
    576 	n = --lim->p_refcnt;
    577 	simple_unlock(&lim->p_slock);
    578 	if (n > 0)
    579 		return;
    580 #ifdef DIAGNOSTIC
    581 	if (n < 0)
    582 		panic("limfree");
    583 #endif
    584 	if (lim->pl_corename != defcorename)
    585 		free(lim->pl_corename, M_TEMP);
    586 	pool_put(&plimit_pool, lim);
    587 }
    588 
    589 struct pstats *
    590 pstatscopy(struct pstats *ps)
    591 {
    592 
    593 	struct pstats *newps;
    594 
    595 	newps = pool_get(&pstats_pool, PR_WAITOK);
    596 
    597 	memset(&newps->pstat_startzero, 0,
    598 	(unsigned) ((caddr_t)&newps->pstat_endzero -
    599 		    (caddr_t)&newps->pstat_startzero));
    600 	memcpy(&newps->pstat_startcopy, &ps->pstat_startcopy,
    601 	((caddr_t)&newps->pstat_endcopy -
    602 	 (caddr_t)&newps->pstat_startcopy));
    603 
    604 	return (newps);
    605 
    606 }
    607 
    608 void
    609 pstatsfree(struct pstats *ps)
    610 {
    611 
    612 	pool_put(&pstats_pool, ps);
    613 }
    614 
    615 /*
    616  * sysctl interface in five parts
    617  */
    618 
    619 /*
    620  * a routine for sysctl proc subtree helpers that need to pick a valid
    621  * process by pid.
    622  */
    623 static int
    624 sysctl_proc_findproc(struct lwp *l, struct proc **p2, pid_t pid)
    625 {
    626 	struct proc *ptmp;
    627 	int error = 0;
    628 
    629 	if (pid == PROC_CURPROC)
    630 		ptmp = l->l_proc;
    631 	else if ((ptmp = pfind(pid)) == NULL)
    632 		error = ESRCH;
    633 	else {
    634 		boolean_t isroot = kauth_authorize_generic(l->l_cred,
    635 		    KAUTH_GENERIC_ISSUSER, NULL);
    636 		/*
    637 		 * suid proc of ours or proc not ours
    638 		 */
    639 		if (kauth_cred_getuid(l->l_cred) !=
    640 		    kauth_cred_getuid(ptmp->p_cred) ||
    641 		    kauth_cred_getuid(l->l_cred) !=
    642 		    kauth_cred_getsvuid(ptmp->p_cred))
    643 			error = isroot ? 0 : EPERM;
    644 
    645 		/*
    646 		 * sgid proc has sgid back to us temporarily
    647 		 */
    648 		else if (kauth_cred_getgid(ptmp->p_cred) !=
    649 		    kauth_cred_getsvgid(ptmp->p_cred))
    650 			error = isroot ? 0 : EPERM;
    651 
    652 		/*
    653 		 * our rgid must be in target's group list (ie,
    654 		 * sub-processes started by a sgid process)
    655 		 */
    656 		else {
    657 			int ismember = 0;
    658 
    659 			if (kauth_cred_ismember_gid(l->l_cred,
    660 			    kauth_cred_getgid(ptmp->p_cred), &ismember) != 0 ||
    661 			    !ismember) {
    662 				error = isroot ? 0 : EPERM;
    663 			}
    664 		}
    665 	}
    666 
    667 	*p2 = ptmp;
    668 	return (error);
    669 }
    670 
    671 /*
    672  * sysctl helper routine for setting a process's specific corefile
    673  * name.  picks the process based on the given pid and checks the
    674  * correctness of the new value.
    675  */
    676 static int
    677 sysctl_proc_corename(SYSCTLFN_ARGS)
    678 {
    679 	struct proc *ptmp;
    680 	struct plimit *lim;
    681 	int error = 0, len;
    682 	char *cname;
    683 	char *tmp;
    684 	struct sysctlnode node;
    685 
    686 	/*
    687 	 * is this all correct?
    688 	 */
    689 	if (namelen != 0)
    690 		return (EINVAL);
    691 	if (name[-1] != PROC_PID_CORENAME)
    692 		return (EINVAL);
    693 
    694 	/*
    695 	 * whom are we tweaking?
    696 	 */
    697 	error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]);
    698 	if (error)
    699 		return (error);
    700 
    701 	cname = PNBUF_GET();
    702 	/*
    703 	 * let them modify a temporary copy of the core name
    704 	 */
    705 	node = *rnode;
    706 	strlcpy(cname, ptmp->p_limit->pl_corename, MAXPATHLEN);
    707 	node.sysctl_data = cname;
    708 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
    709 
    710 	/*
    711 	 * if that failed, or they have nothing new to say, or we've
    712 	 * heard it before...
    713 	 */
    714 	if (error || newp == NULL ||
    715 	    strcmp(cname, ptmp->p_limit->pl_corename) == 0) {
    716 		goto done;
    717 	}
    718 
    719 	if (kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CORENAME,
    720 	    l->l_proc, NULL, NULL, NULL) != 0)
    721 		return (EPERM);
    722 
    723 	/*
    724 	 * no error yet and cname now has the new core name in it.
    725 	 * let's see if it looks acceptable.  it must be either "core"
    726 	 * or end in ".core" or "/core".
    727 	 */
    728 	len = strlen(cname);
    729 	if (len < 4) {
    730 		error = EINVAL;
    731 	} else if (strcmp(cname + len - 4, "core") != 0) {
    732 		error = EINVAL;
    733 	} else if (len > 4 && cname[len - 5] != '/' && cname[len - 5] != '.') {
    734 		error = EINVAL;
    735 	}
    736 	if (error != 0) {
    737 		goto done;
    738 	}
    739 
    740 	/*
    741 	 * hmm...looks good.  now...where do we put it?
    742 	 */
    743 	tmp = malloc(len + 1, M_TEMP, M_WAITOK|M_CANFAIL);
    744 	if (tmp == NULL) {
    745 		error = ENOMEM;
    746 		goto done;
    747 	}
    748 	strlcpy(tmp, cname, len + 1);
    749 
    750 	lim = ptmp->p_limit;
    751 	if (lim->p_refcnt > 1 && (lim->p_lflags & PL_SHAREMOD) == 0) {
    752 		ptmp->p_limit = limcopy(lim);
    753 		limfree(lim);
    754 		lim = ptmp->p_limit;
    755 	}
    756 	if (lim->pl_corename != defcorename)
    757 		free(lim->pl_corename, M_TEMP);
    758 	lim->pl_corename = tmp;
    759 done:
    760 	PNBUF_PUT(cname);
    761 	return error;
    762 }
    763 
    764 /*
    765  * sysctl helper routine for checking/setting a process's stop flags,
    766  * one for fork and one for exec.
    767  */
    768 static int
    769 sysctl_proc_stop(SYSCTLFN_ARGS)
    770 {
    771 	struct proc *ptmp;
    772 	int i, f, error = 0;
    773 	struct sysctlnode node;
    774 
    775 	if (namelen != 0)
    776 		return (EINVAL);
    777 
    778 	error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-2]);
    779 	if (error)
    780 		return (error);
    781 
    782 	switch (rnode->sysctl_num) {
    783 	case PROC_PID_STOPFORK:
    784 		f = PS_STOPFORK;
    785 		break;
    786 	case PROC_PID_STOPEXEC:
    787 		f = PS_STOPEXEC;
    788 		break;
    789 	case PROC_PID_STOPEXIT:
    790 		f = PS_STOPEXIT;
    791 		break;
    792 	default:
    793 		return (EINVAL);
    794 	}
    795 
    796 	i = (ptmp->p_flag & f) ? 1 : 0;
    797 	node = *rnode;
    798 	node.sysctl_data = &i;
    799 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
    800 	if (error || newp == NULL)
    801 		return (error);
    802 
    803 	mutex_enter(&ptmp->p_smutex);
    804 	if (i)
    805 		ptmp->p_sflag |= f;
    806 	else
    807 		ptmp->p_sflag &= ~f;
    808 	mutex_exit(&ptmp->p_smutex);
    809 
    810 	return (0);
    811 }
    812 
    813 /*
    814  * sysctl helper routine for a process's rlimits as exposed by sysctl.
    815  */
    816 static int
    817 sysctl_proc_plimit(SYSCTLFN_ARGS)
    818 {
    819 	struct proc *ptmp;
    820 	u_int limitno;
    821 	int which, error = 0;
    822         struct rlimit alim;
    823 	struct sysctlnode node;
    824 
    825 	if (namelen != 0)
    826 		return (EINVAL);
    827 
    828 	which = name[-1];
    829 	if (which != PROC_PID_LIMIT_TYPE_SOFT &&
    830 	    which != PROC_PID_LIMIT_TYPE_HARD)
    831 		return (EINVAL);
    832 
    833 	limitno = name[-2] - 1;
    834 	if (limitno >= RLIM_NLIMITS)
    835 		return (EINVAL);
    836 
    837 	if (name[-3] != PROC_PID_LIMIT)
    838 		return (EINVAL);
    839 
    840 	error = sysctl_proc_findproc(l, &ptmp, (pid_t)name[-4]);
    841 	if (error)
    842 		return (error);
    843 
    844 	node = *rnode;
    845 	memcpy(&alim, &ptmp->p_rlimit[limitno], sizeof(alim));
    846 	if (which == PROC_PID_LIMIT_TYPE_HARD)
    847 		node.sysctl_data = &alim.rlim_max;
    848 	else
    849 		node.sysctl_data = &alim.rlim_cur;
    850 
    851 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
    852 	if (error || newp == NULL)
    853 		return (error);
    854 
    855 	return (dosetrlimit(l, ptmp, limitno, &alim));
    856 }
    857 
    858 /*
    859  * and finally, the actually glue that sticks it to the tree
    860  */
    861 SYSCTL_SETUP(sysctl_proc_setup, "sysctl proc subtree setup")
    862 {
    863 
    864 	sysctl_createv(clog, 0, NULL, NULL,
    865 		       CTLFLAG_PERMANENT,
    866 		       CTLTYPE_NODE, "proc", NULL,
    867 		       NULL, 0, NULL, 0,
    868 		       CTL_PROC, CTL_EOL);
    869 	sysctl_createv(clog, 0, NULL, NULL,
    870 		       CTLFLAG_PERMANENT|CTLFLAG_ANYNUMBER,
    871 		       CTLTYPE_NODE, "curproc",
    872 		       SYSCTL_DESCR("Per-process settings"),
    873 		       NULL, 0, NULL, 0,
    874 		       CTL_PROC, PROC_CURPROC, CTL_EOL);
    875 
    876 	sysctl_createv(clog, 0, NULL, NULL,
    877 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
    878 		       CTLTYPE_STRING, "corename",
    879 		       SYSCTL_DESCR("Core file name"),
    880 		       sysctl_proc_corename, 0, NULL, MAXPATHLEN,
    881 		       CTL_PROC, PROC_CURPROC, PROC_PID_CORENAME, CTL_EOL);
    882 	sysctl_createv(clog, 0, NULL, NULL,
    883 		       CTLFLAG_PERMANENT,
    884 		       CTLTYPE_NODE, "rlimit",
    885 		       SYSCTL_DESCR("Process limits"),
    886 		       NULL, 0, NULL, 0,
    887 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, CTL_EOL);
    888 
    889 #define create_proc_plimit(s, n) do {					\
    890 	sysctl_createv(clog, 0, NULL, NULL,				\
    891 		       CTLFLAG_PERMANENT,				\
    892 		       CTLTYPE_NODE, s,					\
    893 		       SYSCTL_DESCR("Process " s " limits"),		\
    894 		       NULL, 0, NULL, 0,				\
    895 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
    896 		       CTL_EOL);					\
    897 	sysctl_createv(clog, 0, NULL, NULL,				\
    898 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \
    899 		       CTLTYPE_QUAD, "soft",				\
    900 		       SYSCTL_DESCR("Process soft " s " limit"),	\
    901 		       sysctl_proc_plimit, 0, NULL, 0,			\
    902 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
    903 		       PROC_PID_LIMIT_TYPE_SOFT, CTL_EOL);		\
    904 	sysctl_createv(clog, 0, NULL, NULL,				\
    905 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE, \
    906 		       CTLTYPE_QUAD, "hard",				\
    907 		       SYSCTL_DESCR("Process hard " s " limit"),	\
    908 		       sysctl_proc_plimit, 0, NULL, 0,			\
    909 		       CTL_PROC, PROC_CURPROC, PROC_PID_LIMIT, n,	\
    910 		       PROC_PID_LIMIT_TYPE_HARD, CTL_EOL);		\
    911 	} while (0/*CONSTCOND*/)
    912 
    913 	create_proc_plimit("cputime",		PROC_PID_LIMIT_CPU);
    914 	create_proc_plimit("filesize",		PROC_PID_LIMIT_FSIZE);
    915 	create_proc_plimit("datasize",		PROC_PID_LIMIT_DATA);
    916 	create_proc_plimit("stacksize",		PROC_PID_LIMIT_STACK);
    917 	create_proc_plimit("coredumpsize",	PROC_PID_LIMIT_CORE);
    918 	create_proc_plimit("memoryuse",		PROC_PID_LIMIT_RSS);
    919 	create_proc_plimit("memorylocked",	PROC_PID_LIMIT_MEMLOCK);
    920 	create_proc_plimit("maxproc",		PROC_PID_LIMIT_NPROC);
    921 	create_proc_plimit("descriptors",	PROC_PID_LIMIT_NOFILE);
    922 	create_proc_plimit("sbsize",		PROC_PID_LIMIT_SBSIZE);
    923 
    924 #undef create_proc_plimit
    925 
    926 	sysctl_createv(clog, 0, NULL, NULL,
    927 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
    928 		       CTLTYPE_INT, "stopfork",
    929 		       SYSCTL_DESCR("Stop process at fork(2)"),
    930 		       sysctl_proc_stop, 0, NULL, 0,
    931 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPFORK, CTL_EOL);
    932 	sysctl_createv(clog, 0, NULL, NULL,
    933 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
    934 		       CTLTYPE_INT, "stopexec",
    935 		       SYSCTL_DESCR("Stop process at execve(2)"),
    936 		       sysctl_proc_stop, 0, NULL, 0,
    937 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXEC, CTL_EOL);
    938 	sysctl_createv(clog, 0, NULL, NULL,
    939 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE|CTLFLAG_ANYWRITE,
    940 		       CTLTYPE_INT, "stopexit",
    941 		       SYSCTL_DESCR("Stop process before completing exit"),
    942 		       sysctl_proc_stop, 0, NULL, 0,
    943 		       CTL_PROC, PROC_CURPROC, PROC_PID_STOPEXIT, CTL_EOL);
    944 }
    945 
    946 struct uidinfo *
    947 uid_find(uid_t uid)
    948 {
    949 	struct uidinfo *uip;
    950 	struct uidinfo *newuip = NULL;
    951 	struct uihashhead *uipp;
    952 
    953 	uipp = UIHASH(uid);
    954 
    955 again:
    956 	simple_lock(&uihashtbl_slock);
    957 	LIST_FOREACH(uip, uipp, ui_hash)
    958 		if (uip->ui_uid == uid) {
    959 			simple_unlock(&uihashtbl_slock);
    960 			if (newuip)
    961 				free(newuip, M_PROC);
    962 			return uip;
    963 		}
    964 
    965 	if (newuip == NULL) {
    966 		simple_unlock(&uihashtbl_slock);
    967 		newuip = malloc(sizeof(*uip), M_PROC, M_WAITOK | M_ZERO);
    968 		goto again;
    969 	}
    970 	uip = newuip;
    971 
    972 	LIST_INSERT_HEAD(uipp, uip, ui_hash);
    973 	uip->ui_uid = uid;
    974 	simple_lock_init(&uip->ui_slock);
    975 	simple_unlock(&uihashtbl_slock);
    976 
    977 	return uip;
    978 }
    979 
    980 /*
    981  * Change the count associated with number of processes
    982  * a given user is using.
    983  */
    984 int
    985 chgproccnt(uid_t uid, int diff)
    986 {
    987 	struct uidinfo *uip;
    988 	int s;
    989 
    990 	if (diff == 0)
    991 		return 0;
    992 
    993 	uip = uid_find(uid);
    994 	UILOCK(uip, s);
    995 	uip->ui_proccnt += diff;
    996 	KASSERT(uip->ui_proccnt >= 0);
    997 	UIUNLOCK(uip, s);
    998 	return uip->ui_proccnt;
    999 }
   1000 
   1001 int
   1002 chgsbsize(struct uidinfo *uip, u_long *hiwat, u_long to, rlim_t xmax)
   1003 {
   1004 	rlim_t nsb;
   1005 	int s;
   1006 
   1007 	UILOCK(uip, s);
   1008 	nsb = uip->ui_sbsize + to - *hiwat;
   1009 	if (to > *hiwat && nsb > xmax) {
   1010 		UIUNLOCK(uip, s);
   1011 		splx(s);
   1012 		return 0;
   1013 	}
   1014 	*hiwat = to;
   1015 	uip->ui_sbsize = nsb;
   1016 	KASSERT(uip->ui_sbsize >= 0);
   1017 	UIUNLOCK(uip, s);
   1018 	return 1;
   1019 }
   1020