Home | History | Annotate | Line # | Download | only in procfs
procfs_subr.c revision 1.92
      1 /*	$NetBSD: procfs_subr.c,v 1.92 2008/09/05 14:01:11 skrll Exp $	*/
      2 
      3 /*-
      4  * Copyright (c) 2006, 2007, 2008 The NetBSD Foundation, Inc.
      5  * All rights reserved.
      6  *
      7  * This code is derived from software contributed to The NetBSD Foundation
      8  * by Andrew Doran.
      9  *
     10  * Redistribution and use in source and binary forms, with or without
     11  * modification, are permitted provided that the following conditions
     12  * are met:
     13  * 1. Redistributions of source code must retain the above copyright
     14  *    notice, this list of conditions and the following disclaimer.
     15  * 2. Redistributions in binary form must reproduce the above copyright
     16  *    notice, this list of conditions and the following disclaimer in the
     17  *    documentation and/or other materials provided with the distribution.
     18  *
     19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     29  * POSSIBILITY OF SUCH DAMAGE.
     30  */
     31 
     32 /*
     33  * Copyright (c) 1993
     34  *	The Regents of the University of California.  All rights reserved.
     35  *
     36  * This code is derived from software contributed to Berkeley by
     37  * Jan-Simon Pendry.
     38  *
     39  * Redistribution and use in source and binary forms, with or without
     40  * modification, are permitted provided that the following conditions
     41  * are met:
     42  * 1. Redistributions of source code must retain the above copyright
     43  *    notice, this list of conditions and the following disclaimer.
     44  * 2. Redistributions in binary form must reproduce the above copyright
     45  *    notice, this list of conditions and the following disclaimer in the
     46  *    documentation and/or other materials provided with the distribution.
     47  * 3. Neither the name of the University nor the names of its contributors
     48  *    may be used to endorse or promote products derived from this software
     49  *    without specific prior written permission.
     50  *
     51  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     52  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     53  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     54  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     55  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     56  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     57  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     58  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     59  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     60  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     61  * SUCH DAMAGE.
     62  *
     63  *	@(#)procfs_subr.c	8.6 (Berkeley) 5/14/95
     64  */
     65 
     66 /*
     67  * Copyright (c) 1994 Christopher G. Demetriou.  All rights reserved.
     68  * Copyright (c) 1993 Jan-Simon Pendry
     69  *
     70  * This code is derived from software contributed to Berkeley by
     71  * Jan-Simon Pendry.
     72  *
     73  * Redistribution and use in source and binary forms, with or without
     74  * modification, are permitted provided that the following conditions
     75  * are met:
     76  * 1. Redistributions of source code must retain the above copyright
     77  *    notice, this list of conditions and the following disclaimer.
     78  * 2. Redistributions in binary form must reproduce the above copyright
     79  *    notice, this list of conditions and the following disclaimer in the
     80  *    documentation and/or other materials provided with the distribution.
     81  * 3. All advertising materials mentioning features or use of this software
     82  *    must display the following acknowledgement:
     83  *	This product includes software developed by the University of
     84  *	California, Berkeley and its contributors.
     85  * 4. Neither the name of the University nor the names of its contributors
     86  *    may be used to endorse or promote products derived from this software
     87  *    without specific prior written permission.
     88  *
     89  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     90  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     91  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     92  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     93  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     94  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     95  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     96  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     97  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     98  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     99  * SUCH DAMAGE.
    100  *
    101  *	@(#)procfs_subr.c	8.6 (Berkeley) 5/14/95
    102  */
    103 
    104 #include <sys/cdefs.h>
    105 __KERNEL_RCSID(0, "$NetBSD: procfs_subr.c,v 1.92 2008/09/05 14:01:11 skrll Exp $");
    106 
    107 #include <sys/param.h>
    108 #include <sys/systm.h>
    109 #include <sys/time.h>
    110 #include <sys/kernel.h>
    111 #include <sys/proc.h>
    112 #include <sys/vnode.h>
    113 #include <sys/malloc.h>
    114 #include <sys/stat.h>
    115 #include <sys/file.h>
    116 #include <sys/filedesc.h>
    117 #include <sys/kauth.h>
    118 
    119 #include <miscfs/procfs/procfs.h>
    120 
    121 void procfs_hashins(struct pfsnode *);
    122 void procfs_hashrem(struct pfsnode *);
    123 struct vnode *procfs_hashget(pid_t, pfstype, int, struct mount *, int);
    124 
    125 LIST_HEAD(pfs_hashhead, pfsnode) *pfs_hashtbl;
    126 u_long	pfs_ihash;	/* size of hash table - 1 */
    127 #define PFSPIDHASH(pid)	((pid) & pfs_ihash)
    128 
    129 kmutex_t pfs_hashlock;
    130 kmutex_t pfs_ihash_lock;
    131 
    132 #define	ISSET(t, f)	((t) & (f))
    133 
    134 /*
    135  * allocate a pfsnode/vnode pair.  the vnode is
    136  * referenced, and locked.
    137  *
    138  * the pid, pfs_type, and mount point uniquely
    139  * identify a pfsnode.  the mount point is needed
    140  * because someone might mount this filesystem
    141  * twice.
    142  *
    143  * all pfsnodes are maintained on a singly-linked
    144  * list.  new nodes are only allocated when they cannot
    145  * be found on this list.  entries on the list are
    146  * removed when the vfs reclaim entry is called.
    147  *
    148  * a single lock is kept for the entire list.  this is
    149  * needed because the getnewvnode() function can block
    150  * waiting for a vnode to become free, in which case there
    151  * may be more than one process trying to get the same
    152  * vnode.  this lock is only taken if we are going to
    153  * call getnewvnode, since the kernel itself is single-threaded.
    154  *
    155  * if an entry is found on the list, then call vget() to
    156  * take a reference.  this is done because there may be
    157  * zero references to it and so it needs to removed from
    158  * the vnode free list.
    159  */
    160 int
    161 procfs_allocvp(mp, vpp, pid, pfs_type, fd, p)
    162 	struct mount *mp;
    163 	struct vnode **vpp;
    164 	pid_t pid;
    165 	pfstype pfs_type;
    166 	int fd;
    167 	struct proc *p;
    168 {
    169 	struct pfsnode *pfs;
    170 	struct vnode *vp;
    171 	int error;
    172 
    173  retry:
    174 	*vpp = procfs_hashget(pid, pfs_type, fd, mp, LK_EXCLUSIVE);
    175 	if (*vpp != NULL)
    176 		return (0);
    177 
    178 	if ((error = getnewvnode(VT_PROCFS, mp, procfs_vnodeop_p, &vp)) != 0) {
    179 		*vpp = NULL;
    180 		return (error);
    181 	}
    182 	MALLOC(pfs, void *, sizeof(struct pfsnode), M_TEMP, M_WAITOK);
    183 
    184 	mutex_enter(&pfs_hashlock);
    185 	if ((*vpp = procfs_hashget(pid, pfs_type, fd, mp, 0)) != NULL) {
    186 		mutex_exit(&pfs_hashlock);
    187 		ungetnewvnode(vp);
    188 		FREE(pfs, M_TEMP);
    189 		goto retry;
    190 	}
    191 
    192 	vp->v_data = pfs;
    193 	pfs->pfs_pid = pid;
    194 	pfs->pfs_type = pfs_type;
    195 	pfs->pfs_vnode = vp;
    196 	pfs->pfs_flags = 0;
    197 	pfs->pfs_fileno = PROCFS_FILENO(pid, pfs_type, fd);
    198 	pfs->pfs_fd = fd;
    199 
    200 	switch (pfs_type) {
    201 	case PFSroot:	/* /proc = dr-xr-xr-x */
    202 		vp->v_vflag |= VV_ROOT;
    203 		/*FALLTHROUGH*/
    204 	case PFSproc:	/* /proc/N = dr-xr-xr-x */
    205 		pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
    206 		vp->v_type = VDIR;
    207 		break;
    208 
    209 	case PFScurproc:	/* /proc/curproc = lr-xr-xr-x */
    210 	case PFSself:	/* /proc/self    = lr-xr-xr-x */
    211 	case PFScwd:	/* /proc/N/cwd = lr-xr-xr-x */
    212 	case PFSchroot:	/* /proc/N/chroot = lr-xr-xr-x */
    213 	case PFSexe:	/* /proc/N/exe = lr-xr-xr-x */
    214 		pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
    215 		vp->v_type = VLNK;
    216 		break;
    217 
    218 	case PFSfd:
    219 		if (fd == -1) {	/* /proc/N/fd = dr-xr-xr-x */
    220 			pfs->pfs_mode = S_IRUSR|S_IXUSR;
    221 			vp->v_type = VDIR;
    222 		} else {	/* /proc/N/fd/M = [ps-]rw------- */
    223 			file_t *fp;
    224 			vnode_t *vxp;
    225 
    226 			if ((fp = fd_getfile2(p, pfs->pfs_fd)) == NULL) {
    227 				error = EBADF;
    228 				goto bad;
    229 			}
    230 
    231 			pfs->pfs_mode = S_IRUSR|S_IWUSR;
    232 			switch (fp->f_type) {
    233 			case DTYPE_VNODE:
    234 				vxp = fp->f_data;
    235 
    236 				/*
    237 				 * We make symlinks for directories
    238 				 * to avoid cycles.
    239 				 */
    240 				if (vxp->v_type == VDIR)
    241 					goto symlink;
    242 				vp->v_type = vxp->v_type;
    243 				break;
    244 			case DTYPE_PIPE:
    245 				vp->v_type = VFIFO;
    246 				break;
    247 			case DTYPE_SOCKET:
    248 				vp->v_type = VSOCK;
    249 				break;
    250 			case DTYPE_KQUEUE:
    251 			case DTYPE_MISC:
    252 			symlink:
    253 				pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|
    254 				    S_IXGRP|S_IROTH|S_IXOTH;
    255 				vp->v_type = VLNK;
    256 				break;
    257 			default:
    258 				error = EOPNOTSUPP;
    259 				closef(fp);
    260 				goto bad;
    261 			}
    262 			closef(fp);
    263 		}
    264 		break;
    265 
    266 	case PFSfile:	/* /proc/N/file = -rw------- */
    267 	case PFSmem:	/* /proc/N/mem = -rw------- */
    268 	case PFSregs:	/* /proc/N/regs = -rw------- */
    269 	case PFSfpregs:	/* /proc/N/fpregs = -rw------- */
    270 		pfs->pfs_mode = S_IRUSR|S_IWUSR;
    271 		vp->v_type = VREG;
    272 		break;
    273 
    274 	case PFSctl:	/* /proc/N/ctl = --w------ */
    275 	case PFSnote:	/* /proc/N/note = --w------ */
    276 	case PFSnotepg:	/* /proc/N/notepg = --w------ */
    277 		pfs->pfs_mode = S_IWUSR;
    278 		vp->v_type = VREG;
    279 		break;
    280 
    281 	case PFSmap:	/* /proc/N/map = -r--r--r-- */
    282 	case PFSmaps:	/* /proc/N/maps = -r--r--r-- */
    283 	case PFSstatus:	/* /proc/N/status = -r--r--r-- */
    284 	case PFSstat:	/* /proc/N/stat = -r--r--r-- */
    285 	case PFScmdline:	/* /proc/N/cmdline = -r--r--r-- */
    286 	case PFSemul:	/* /proc/N/emul = -r--r--r-- */
    287 	case PFSmeminfo:	/* /proc/meminfo = -r--r--r-- */
    288 	case PFScpustat:	/* /proc/stat = -r--r--r-- */
    289 	case PFSdevices:	/* /proc/devices = -r--r--r-- */
    290 	case PFScpuinfo:	/* /proc/cpuinfo = -r--r--r-- */
    291 	case PFSuptime:	/* /proc/uptime = -r--r--r-- */
    292 	case PFSmounts:	/* /proc/mounts = -r--r--r-- */
    293 	case PFSloadavg:	/* /proc/loadavg = -r--r--r-- */
    294 	case PFSstatm:	/* /proc/N/statm = -r--r--r-- */
    295 		pfs->pfs_mode = S_IRUSR|S_IRGRP|S_IROTH;
    296 		vp->v_type = VREG;
    297 		break;
    298 
    299 #ifdef __HAVE_PROCFS_MACHDEP
    300 	PROCFS_MACHDEP_NODETYPE_CASES
    301 		procfs_machdep_allocvp(vp);
    302 		break;
    303 #endif
    304 
    305 	default:
    306 		panic("procfs_allocvp");
    307 	}
    308 
    309 	procfs_hashins(pfs);
    310 	uvm_vnp_setsize(vp, 0);
    311 	mutex_exit(&pfs_hashlock);
    312 
    313 	*vpp = vp;
    314 	return (0);
    315 
    316  bad:
    317 	mutex_exit(&pfs_hashlock);
    318 	FREE(pfs, M_TEMP);
    319 	vp->v_data = NULL;
    320 	ungetnewvnode(vp);
    321 	return (error);
    322 }
    323 
    324 int
    325 procfs_freevp(vp)
    326 	struct vnode *vp;
    327 {
    328 	struct pfsnode *pfs = VTOPFS(vp);
    329 
    330 	procfs_hashrem(pfs);
    331 
    332 	FREE(vp->v_data, M_TEMP);
    333 	vp->v_data = 0;
    334 	return (0);
    335 }
    336 
    337 int
    338 procfs_rw(v)
    339 	void *v;
    340 {
    341 	struct vop_read_args *ap = v;
    342 	struct vnode *vp = ap->a_vp;
    343 	struct uio *uio = ap->a_uio;
    344 	struct lwp *curl;
    345 	struct lwp *l;
    346 	struct pfsnode *pfs = VTOPFS(vp);
    347 	struct proc *p;
    348 	int error;
    349 
    350 	if (uio->uio_offset < 0)
    351 		return EINVAL;
    352 
    353 	if ((error = procfs_proc_lock(pfs->pfs_pid, &p, ESRCH)) != 0)
    354 		return error;
    355 
    356 	curl = curlwp;
    357 
    358 	/*
    359 	 * Do not allow init to be modified while in secure mode; it
    360 	 * could be duped into changing the security level.
    361 	 */
    362 #define	M2K(m)	((m) == UIO_READ ? KAUTH_REQ_PROCESS_PROCFS_READ : \
    363 		 KAUTH_REQ_PROCESS_PROCFS_WRITE)
    364 	mutex_enter(p->p_lock);
    365 	error = kauth_authorize_process(curl->l_cred, KAUTH_PROCESS_PROCFS,
    366 	    p, pfs, KAUTH_ARG(M2K(uio->uio_rw)), NULL);
    367 	mutex_exit(p->p_lock);
    368 	if (error) {
    369 		procfs_proc_unlock(p);
    370 		return (error);
    371 	}
    372 #undef	M2K
    373 
    374 	mutex_enter(p->p_lock);
    375 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
    376 		if (l->l_stat != LSZOMB)
    377 			break;
    378 	}
    379 	/* Process is exiting if no-LWPS or all LWPs are LSZOMB */
    380 	if (l == NULL) {
    381 		mutex_exit(p->p_lock);
    382 		procfs_proc_unlock(p);
    383 		return ESRCH;
    384 	}
    385 
    386 	lwp_addref(l);
    387 	mutex_exit(p->p_lock);
    388 
    389 	switch (pfs->pfs_type) {
    390 	case PFSnote:
    391 	case PFSnotepg:
    392 		error = procfs_donote(curl, p, pfs, uio);
    393 		break;
    394 
    395 	case PFSregs:
    396 		error = procfs_doregs(curl, l, pfs, uio);
    397 		break;
    398 
    399 	case PFSfpregs:
    400 		error = procfs_dofpregs(curl, l, pfs, uio);
    401 		break;
    402 
    403 	case PFSctl:
    404 		error = procfs_doctl(curl, l, pfs, uio);
    405 		break;
    406 
    407 	case PFSstatus:
    408 		error = procfs_dostatus(curl, l, pfs, uio);
    409 		break;
    410 
    411 	case PFSstat:
    412 		error = procfs_do_pid_stat(curl, l, pfs, uio);
    413 		break;
    414 
    415 	case PFSmap:
    416 		error = procfs_domap(curl, p, pfs, uio, 0);
    417 		break;
    418 
    419 	case PFSmaps:
    420 		error = procfs_domap(curl, p, pfs, uio, 1);
    421 		break;
    422 
    423 	case PFSmem:
    424 		error = procfs_domem(curl, l, pfs, uio);
    425 		break;
    426 
    427 	case PFScmdline:
    428 		error = procfs_docmdline(curl, p, pfs, uio);
    429 		break;
    430 
    431 	case PFSmeminfo:
    432 		error = procfs_domeminfo(curl, p, pfs, uio);
    433 		break;
    434 
    435 	case PFSdevices:
    436 		error = procfs_dodevices(curl, p, pfs, uio);
    437 		break;
    438 
    439 	case PFScpuinfo:
    440 		error = procfs_docpuinfo(curl, p, pfs, uio);
    441 		break;
    442 
    443 	case PFScpustat:
    444 		error = procfs_docpustat(curl, p, pfs, uio);
    445 		break;
    446 
    447 	case PFSloadavg:
    448 		error = procfs_doloadavg(curl, p, pfs, uio);
    449 		break;
    450 
    451 	case PFSstatm:
    452 		error = procfs_do_pid_statm(curl, l, pfs, uio);
    453 		break;
    454 
    455 	case PFSfd:
    456 		error = procfs_dofd(curl, p, pfs, uio);
    457 		break;
    458 
    459 	case PFSuptime:
    460 		error = procfs_douptime(curl, p, pfs, uio);
    461 		break;
    462 
    463 	case PFSmounts:
    464 		error = procfs_domounts(curl, p, pfs, uio);
    465 		break;
    466 
    467 	case PFSemul:
    468 		error = procfs_doemul(curl, p, pfs, uio);
    469 		break;
    470 
    471 #ifdef __HAVE_PROCFS_MACHDEP
    472 	PROCFS_MACHDEP_NODETYPE_CASES
    473 		error = procfs_machdep_rw(curl, l, pfs, uio);
    474 		break;
    475 #endif
    476 
    477 	default:
    478 		error = EOPNOTSUPP;
    479 		break;
    480 	}
    481 
    482 	/*
    483 	 * Release the references that we acquired earlier.
    484 	 */
    485 	lwp_delref(l);
    486 	procfs_proc_unlock(p);
    487 
    488 	return (error);
    489 }
    490 
    491 /*
    492  * Get a string from userland into (bf).  Strip a trailing
    493  * nl character (to allow easy access from the shell).
    494  * The buffer should be *buflenp + 1 chars long.  vfs_getuserstr
    495  * will automatically add a nul char at the end.
    496  *
    497  * Returns 0 on success or the following errors
    498  *
    499  * EINVAL:    file offset is non-zero.
    500  * EMSGSIZE:  message is longer than kernel buffer
    501  * EFAULT:    user i/o buffer is not addressable
    502  */
    503 int
    504 vfs_getuserstr(uio, bf, buflenp)
    505 	struct uio *uio;
    506 	char *bf;
    507 	int *buflenp;
    508 {
    509 	int xlen;
    510 	int error;
    511 
    512 	if (uio->uio_offset != 0)
    513 		return (EINVAL);
    514 
    515 	xlen = *buflenp;
    516 
    517 	/* must be able to read the whole string in one go */
    518 	if (xlen < uio->uio_resid)
    519 		return (EMSGSIZE);
    520 	xlen = uio->uio_resid;
    521 
    522 	if ((error = uiomove(bf, xlen, uio)) != 0)
    523 		return (error);
    524 
    525 	/* allow multiple writes without seeks */
    526 	uio->uio_offset = 0;
    527 
    528 	/* cleanup string and remove trailing newline */
    529 	bf[xlen] = '\0';
    530 	xlen = strlen(bf);
    531 	if (xlen > 0 && bf[xlen-1] == '\n')
    532 		bf[--xlen] = '\0';
    533 	*buflenp = xlen;
    534 
    535 	return (0);
    536 }
    537 
    538 const vfs_namemap_t *
    539 vfs_findname(nm, bf, buflen)
    540 	const vfs_namemap_t *nm;
    541 	const char *bf;
    542 	int buflen;
    543 {
    544 
    545 	for (; nm->nm_name; nm++)
    546 		if (memcmp(bf, nm->nm_name, buflen+1) == 0)
    547 			return (nm);
    548 
    549 	return (0);
    550 }
    551 
    552 /*
    553  * Initialize pfsnode hash table.
    554  */
    555 void
    556 procfs_hashinit()
    557 {
    558 	mutex_init(&pfs_hashlock, MUTEX_DEFAULT, IPL_NONE);
    559 	mutex_init(&pfs_ihash_lock, MUTEX_DEFAULT, IPL_NONE);
    560 	pfs_hashtbl = hashinit(desiredvnodes / 4, HASH_LIST, true, &pfs_ihash);
    561 }
    562 
    563 void
    564 procfs_hashreinit()
    565 {
    566 	struct pfsnode *pp;
    567 	struct pfs_hashhead *oldhash, *hash;
    568 	u_long i, oldmask, mask, val;
    569 
    570 	hash = hashinit(desiredvnodes / 4, HASH_LIST, true, &mask);
    571 
    572 	mutex_enter(&pfs_ihash_lock);
    573 	oldhash = pfs_hashtbl;
    574 	oldmask = pfs_ihash;
    575 	pfs_hashtbl = hash;
    576 	pfs_ihash = mask;
    577 	for (i = 0; i <= oldmask; i++) {
    578 		while ((pp = LIST_FIRST(&oldhash[i])) != NULL) {
    579 			LIST_REMOVE(pp, pfs_hash);
    580 			val = PFSPIDHASH(pp->pfs_pid);
    581 			LIST_INSERT_HEAD(&hash[val], pp, pfs_hash);
    582 		}
    583 	}
    584 	mutex_exit(&pfs_ihash_lock);
    585 	hashdone(oldhash, HASH_LIST, oldmask);
    586 }
    587 
    588 /*
    589  * Free pfsnode hash table.
    590  */
    591 void
    592 procfs_hashdone()
    593 {
    594 	hashdone(pfs_hashtbl, HASH_LIST, pfs_ihash);
    595 	mutex_destroy(&pfs_hashlock);
    596 	mutex_destroy(&pfs_ihash_lock);
    597 }
    598 
    599 struct vnode *
    600 procfs_hashget(pid, type, fd, mp, flags)
    601 	pid_t pid;
    602 	pfstype type;
    603 	int fd;
    604 	struct mount *mp;
    605 	int flags;
    606 {
    607 	struct pfs_hashhead *ppp;
    608 	struct pfsnode *pp;
    609 	struct vnode *vp;
    610 
    611 loop:
    612 	mutex_enter(&pfs_ihash_lock);
    613 	ppp = &pfs_hashtbl[PFSPIDHASH(pid)];
    614 	LIST_FOREACH(pp, ppp, pfs_hash) {
    615 		vp = PFSTOV(pp);
    616 		if (pid == pp->pfs_pid && pp->pfs_type == type &&
    617 		    pp->pfs_fd == fd && vp->v_mount == mp) {
    618 		    	if (flags == 0) {
    619 				mutex_exit(&pfs_ihash_lock);
    620 			} else {
    621 				mutex_enter(&vp->v_interlock);
    622 				mutex_exit(&pfs_ihash_lock);
    623 				if (vget(vp, flags | LK_INTERLOCK))
    624 					goto loop;
    625 			}
    626 			return (vp);
    627 		}
    628 	}
    629 	mutex_exit(&pfs_ihash_lock);
    630 	return (NULL);
    631 }
    632 
    633 /*
    634  * Insert the pfsnode into the hash table and lock it.
    635  */
    636 void
    637 procfs_hashins(pp)
    638 	struct pfsnode *pp;
    639 {
    640 	struct pfs_hashhead *ppp;
    641 
    642 	/* lock the pfsnode, then put it on the appropriate hash list */
    643 	vlockmgr(&pp->pfs_vnode->v_lock, LK_EXCLUSIVE);
    644 
    645 	mutex_enter(&pfs_ihash_lock);
    646 	ppp = &pfs_hashtbl[PFSPIDHASH(pp->pfs_pid)];
    647 	LIST_INSERT_HEAD(ppp, pp, pfs_hash);
    648 	mutex_exit(&pfs_ihash_lock);
    649 }
    650 
    651 /*
    652  * Remove the pfsnode from the hash table.
    653  */
    654 void
    655 procfs_hashrem(pp)
    656 	struct pfsnode *pp;
    657 {
    658 	mutex_enter(&pfs_ihash_lock);
    659 	LIST_REMOVE(pp, pfs_hash);
    660 	mutex_exit(&pfs_ihash_lock);
    661 }
    662 
    663 void
    664 procfs_revoke_vnodes(p, arg)
    665 	struct proc *p;
    666 	void *arg;
    667 {
    668 	struct pfsnode *pfs, *pnext;
    669 	struct vnode *vp;
    670 	struct mount *mp = (struct mount *)arg;
    671 	struct pfs_hashhead *ppp;
    672 
    673 	if (!(p->p_flag & PK_SUGID))
    674 		return;
    675 
    676 	mutex_enter(&pfs_ihash_lock);
    677 	ppp = &pfs_hashtbl[PFSPIDHASH(p->p_pid)];
    678 	for (pfs = LIST_FIRST(ppp); pfs; pfs = pnext) {
    679 		vp = PFSTOV(pfs);
    680 		pnext = LIST_NEXT(pfs, pfs_hash);
    681 		mutex_enter(&vp->v_interlock);
    682 		if (vp->v_usecount > 0 && pfs->pfs_pid == p->p_pid &&
    683 		    vp->v_mount == mp) {
    684 		    	vp->v_usecount++;
    685 		    	mutex_exit(&vp->v_interlock);
    686 			mutex_exit(&pfs_ihash_lock);
    687 			VOP_REVOKE(vp, REVOKEALL);
    688 			vrele(vp);
    689 			mutex_enter(&pfs_ihash_lock);
    690 		} else {
    691 			mutex_exit(&vp->v_interlock);
    692 		}
    693 	}
    694 	mutex_exit(&pfs_ihash_lock);
    695 }
    696 
    697 int
    698 procfs_proc_lock(int pid, struct proc **bunghole, int notfound)
    699 {
    700 	struct proc *tp;
    701 	int error = 0;
    702 
    703 	mutex_enter(proc_lock);
    704 
    705 	if (pid == 0)
    706 		tp = &proc0;
    707 	else if ((tp = p_find(pid, PFIND_LOCKED)) == NULL)
    708 		error = notfound;
    709 	if (tp != NULL && !rw_tryenter(&tp->p_reflock, RW_READER))
    710 		error = EBUSY;
    711 
    712 	mutex_exit(proc_lock);
    713 
    714 	*bunghole = tp;
    715 	return error;
    716 }
    717 
    718 void
    719 procfs_proc_unlock(struct proc *p)
    720 {
    721 
    722 	rw_exit(&p->p_reflock);
    723 }
    724 
    725 int
    726 procfs_doemul(struct lwp *curl, struct proc *p,
    727     struct pfsnode *pfs, struct uio *uio)
    728 {
    729 	const char *ename = p->p_emul->e_name;
    730 	return uiomove_frombuf(__UNCONST(ename), strlen(ename), uio);
    731 }
    732