Home | History | Annotate | Line # | Download | only in procfs
procfs_subr.c revision 1.99
      1 /*	$NetBSD: procfs_subr.c,v 1.99 2011/06/12 03:35:58 rmind 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 1.99 2011/06/12 03:35:58 rmind 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(struct mount *mp, struct vnode **vpp, pid_t pid,
    162     pfstype pfs_type, int fd, struct proc *p)
    163 {
    164 	struct pfsnode *pfs;
    165 	struct vnode *vp;
    166 	int error;
    167 
    168  retry:
    169 	*vpp = procfs_hashget(pid, pfs_type, fd, mp, LK_EXCLUSIVE);
    170 	if (*vpp != NULL)
    171 		return (0);
    172 
    173 	error = getnewvnode(VT_PROCFS, mp, procfs_vnodeop_p, NULL, &vp);
    174 	if (error) {
    175 		*vpp = NULL;
    176 		return (error);
    177 	}
    178 	pfs = malloc(sizeof(struct pfsnode), M_TEMP, M_WAITOK);
    179 
    180 	mutex_enter(&pfs_hashlock);
    181 	if ((*vpp = procfs_hashget(pid, pfs_type, fd, mp, 0)) != NULL) {
    182 		mutex_exit(&pfs_hashlock);
    183 		ungetnewvnode(vp);
    184 		free(pfs, M_TEMP);
    185 		goto retry;
    186 	}
    187 
    188 	vp->v_data = pfs;
    189 	pfs->pfs_pid = pid;
    190 	pfs->pfs_type = pfs_type;
    191 	pfs->pfs_vnode = vp;
    192 	pfs->pfs_flags = 0;
    193 	pfs->pfs_fileno = PROCFS_FILENO(pid, pfs_type, fd);
    194 	pfs->pfs_fd = fd;
    195 
    196 	switch (pfs_type) {
    197 	case PFSroot:	/* /proc = dr-xr-xr-x */
    198 		vp->v_vflag |= VV_ROOT;
    199 		/*FALLTHROUGH*/
    200 	case PFSproc:	/* /proc/N = dr-xr-xr-x */
    201 		pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
    202 		vp->v_type = VDIR;
    203 		break;
    204 
    205 	case PFScurproc:	/* /proc/curproc = lr-xr-xr-x */
    206 	case PFSself:	/* /proc/self    = lr-xr-xr-x */
    207 	case PFScwd:	/* /proc/N/cwd = lr-xr-xr-x */
    208 	case PFSchroot:	/* /proc/N/chroot = lr-xr-xr-x */
    209 	case PFSexe:	/* /proc/N/exe = lr-xr-xr-x */
    210 		pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
    211 		vp->v_type = VLNK;
    212 		break;
    213 
    214 	case PFSfd:
    215 		if (fd == -1) {	/* /proc/N/fd = dr-xr-xr-x */
    216 			pfs->pfs_mode = S_IRUSR|S_IXUSR;
    217 			vp->v_type = VDIR;
    218 		} else {	/* /proc/N/fd/M = [ps-]rw------- */
    219 			file_t *fp;
    220 			vnode_t *vxp;
    221 
    222 			if ((fp = fd_getfile2(p, pfs->pfs_fd)) == NULL) {
    223 				error = EBADF;
    224 				goto bad;
    225 			}
    226 
    227 			pfs->pfs_mode = S_IRUSR|S_IWUSR;
    228 			switch (fp->f_type) {
    229 			case DTYPE_VNODE:
    230 				vxp = fp->f_data;
    231 
    232 				/*
    233 				 * We make symlinks for directories
    234 				 * to avoid cycles.
    235 				 */
    236 				if (vxp->v_type == VDIR)
    237 					goto symlink;
    238 				vp->v_type = vxp->v_type;
    239 				break;
    240 			case DTYPE_PIPE:
    241 				vp->v_type = VFIFO;
    242 				break;
    243 			case DTYPE_SOCKET:
    244 				vp->v_type = VSOCK;
    245 				break;
    246 			case DTYPE_KQUEUE:
    247 			case DTYPE_MISC:
    248 			symlink:
    249 				pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|
    250 				    S_IXGRP|S_IROTH|S_IXOTH;
    251 				vp->v_type = VLNK;
    252 				break;
    253 			default:
    254 				error = EOPNOTSUPP;
    255 				closef(fp);
    256 				goto bad;
    257 			}
    258 			closef(fp);
    259 		}
    260 		break;
    261 
    262 	case PFSfile:	/* /proc/N/file = -rw------- */
    263 	case PFSmem:	/* /proc/N/mem = -rw------- */
    264 	case PFSregs:	/* /proc/N/regs = -rw------- */
    265 	case PFSfpregs:	/* /proc/N/fpregs = -rw------- */
    266 		pfs->pfs_mode = S_IRUSR|S_IWUSR;
    267 		vp->v_type = VREG;
    268 		break;
    269 
    270 	case PFSctl:	/* /proc/N/ctl = --w------ */
    271 	case PFSnote:	/* /proc/N/note = --w------ */
    272 	case PFSnotepg:	/* /proc/N/notepg = --w------ */
    273 		pfs->pfs_mode = S_IWUSR;
    274 		vp->v_type = VREG;
    275 		break;
    276 
    277 	case PFSmap:	/* /proc/N/map = -r--r--r-- */
    278 	case PFSmaps:	/* /proc/N/maps = -r--r--r-- */
    279 	case PFSstatus:	/* /proc/N/status = -r--r--r-- */
    280 	case PFSstat:	/* /proc/N/stat = -r--r--r-- */
    281 	case PFScmdline:	/* /proc/N/cmdline = -r--r--r-- */
    282 	case PFSemul:	/* /proc/N/emul = -r--r--r-- */
    283 	case PFSmeminfo:	/* /proc/meminfo = -r--r--r-- */
    284 	case PFScpustat:	/* /proc/stat = -r--r--r-- */
    285 	case PFSdevices:	/* /proc/devices = -r--r--r-- */
    286 	case PFScpuinfo:	/* /proc/cpuinfo = -r--r--r-- */
    287 	case PFSuptime:	/* /proc/uptime = -r--r--r-- */
    288 	case PFSmounts:	/* /proc/mounts = -r--r--r-- */
    289 	case PFSloadavg:	/* /proc/loadavg = -r--r--r-- */
    290 	case PFSstatm:	/* /proc/N/statm = -r--r--r-- */
    291 		pfs->pfs_mode = S_IRUSR|S_IRGRP|S_IROTH;
    292 		vp->v_type = VREG;
    293 		break;
    294 
    295 #ifdef __HAVE_PROCFS_MACHDEP
    296 	PROCFS_MACHDEP_NODETYPE_CASES
    297 		procfs_machdep_allocvp(vp);
    298 		break;
    299 #endif
    300 
    301 	default:
    302 		panic("procfs_allocvp");
    303 	}
    304 
    305 	procfs_hashins(pfs);
    306 	uvm_vnp_setsize(vp, 0);
    307 	mutex_exit(&pfs_hashlock);
    308 
    309 	*vpp = vp;
    310 	return (0);
    311 
    312  bad:
    313 	mutex_exit(&pfs_hashlock);
    314 	free(pfs, M_TEMP);
    315 	vp->v_data = NULL;
    316 	ungetnewvnode(vp);
    317 	return (error);
    318 }
    319 
    320 int
    321 procfs_freevp(struct vnode *vp)
    322 {
    323 	struct pfsnode *pfs = VTOPFS(vp);
    324 
    325 	procfs_hashrem(pfs);
    326 
    327 	free(vp->v_data, M_TEMP);
    328 	vp->v_data = NULL;
    329 	return (0);
    330 }
    331 
    332 int
    333 procfs_rw(void *v)
    334 {
    335 	struct vop_read_args *ap = v;
    336 	struct vnode *vp = ap->a_vp;
    337 	struct uio *uio = ap->a_uio;
    338 	struct lwp *curl;
    339 	struct lwp *l;
    340 	struct pfsnode *pfs = VTOPFS(vp);
    341 	struct proc *p;
    342 	int error;
    343 
    344 	if (uio->uio_offset < 0)
    345 		return EINVAL;
    346 
    347 	if ((error = procfs_proc_lock(pfs->pfs_pid, &p, ESRCH)) != 0)
    348 		return error;
    349 
    350 	curl = curlwp;
    351 
    352 	/*
    353 	 * Do not allow init to be modified while in secure mode; it
    354 	 * could be duped into changing the security level.
    355 	 */
    356 #define	M2K(m)	((m) == UIO_READ ? KAUTH_REQ_PROCESS_PROCFS_READ : \
    357 		 KAUTH_REQ_PROCESS_PROCFS_WRITE)
    358 	mutex_enter(p->p_lock);
    359 	error = kauth_authorize_process(curl->l_cred, KAUTH_PROCESS_PROCFS,
    360 	    p, pfs, KAUTH_ARG(M2K(uio->uio_rw)), NULL);
    361 	mutex_exit(p->p_lock);
    362 	if (error) {
    363 		procfs_proc_unlock(p);
    364 		return (error);
    365 	}
    366 #undef	M2K
    367 
    368 	mutex_enter(p->p_lock);
    369 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
    370 		if (l->l_stat != LSZOMB)
    371 			break;
    372 	}
    373 	/* Process is exiting if no-LWPS or all LWPs are LSZOMB */
    374 	if (l == NULL) {
    375 		mutex_exit(p->p_lock);
    376 		procfs_proc_unlock(p);
    377 		return ESRCH;
    378 	}
    379 
    380 	lwp_addref(l);
    381 	mutex_exit(p->p_lock);
    382 
    383 	switch (pfs->pfs_type) {
    384 	case PFSnote:
    385 	case PFSnotepg:
    386 		error = procfs_donote(curl, p, pfs, uio);
    387 		break;
    388 
    389 	case PFSregs:
    390 		error = procfs_doregs(curl, l, pfs, uio);
    391 		break;
    392 
    393 	case PFSfpregs:
    394 		error = procfs_dofpregs(curl, l, pfs, uio);
    395 		break;
    396 
    397 	case PFSctl:
    398 		error = procfs_doctl(curl, l, pfs, uio);
    399 		break;
    400 
    401 	case PFSstatus:
    402 		error = procfs_dostatus(curl, l, pfs, uio);
    403 		break;
    404 
    405 	case PFSstat:
    406 		error = procfs_do_pid_stat(curl, l, pfs, uio);
    407 		break;
    408 
    409 	case PFSmap:
    410 		error = procfs_domap(curl, p, pfs, uio, 0);
    411 		break;
    412 
    413 	case PFSmaps:
    414 		error = procfs_domap(curl, p, pfs, uio, 1);
    415 		break;
    416 
    417 	case PFSmem:
    418 		error = procfs_domem(curl, l, pfs, uio);
    419 		break;
    420 
    421 	case PFScmdline:
    422 		error = procfs_docmdline(curl, p, pfs, uio);
    423 		break;
    424 
    425 	case PFSmeminfo:
    426 		error = procfs_domeminfo(curl, p, pfs, uio);
    427 		break;
    428 
    429 	case PFSdevices:
    430 		error = procfs_dodevices(curl, p, pfs, uio);
    431 		break;
    432 
    433 	case PFScpuinfo:
    434 		error = procfs_docpuinfo(curl, p, pfs, uio);
    435 		break;
    436 
    437 	case PFScpustat:
    438 		error = procfs_docpustat(curl, p, pfs, uio);
    439 		break;
    440 
    441 	case PFSloadavg:
    442 		error = procfs_doloadavg(curl, p, pfs, uio);
    443 		break;
    444 
    445 	case PFSstatm:
    446 		error = procfs_do_pid_statm(curl, l, pfs, uio);
    447 		break;
    448 
    449 	case PFSfd:
    450 		error = procfs_dofd(curl, p, pfs, uio);
    451 		break;
    452 
    453 	case PFSuptime:
    454 		error = procfs_douptime(curl, p, pfs, uio);
    455 		break;
    456 
    457 	case PFSmounts:
    458 		error = procfs_domounts(curl, p, pfs, uio);
    459 		break;
    460 
    461 	case PFSemul:
    462 		error = procfs_doemul(curl, p, pfs, uio);
    463 		break;
    464 
    465 #ifdef __HAVE_PROCFS_MACHDEP
    466 	PROCFS_MACHDEP_NODETYPE_CASES
    467 		error = procfs_machdep_rw(curl, l, pfs, uio);
    468 		break;
    469 #endif
    470 
    471 	default:
    472 		error = EOPNOTSUPP;
    473 		break;
    474 	}
    475 
    476 	/*
    477 	 * Release the references that we acquired earlier.
    478 	 */
    479 	lwp_delref(l);
    480 	procfs_proc_unlock(p);
    481 
    482 	return (error);
    483 }
    484 
    485 /*
    486  * Get a string from userland into (bf).  Strip a trailing
    487  * nl character (to allow easy access from the shell).
    488  * The buffer should be *buflenp + 1 chars long.  vfs_getuserstr
    489  * will automatically add a nul char at the end.
    490  *
    491  * Returns 0 on success or the following errors
    492  *
    493  * EINVAL:    file offset is non-zero.
    494  * EMSGSIZE:  message is longer than kernel buffer
    495  * EFAULT:    user i/o buffer is not addressable
    496  */
    497 int
    498 vfs_getuserstr(struct uio *uio, char *bf, int *buflenp)
    499 {
    500 	int xlen;
    501 	int error;
    502 
    503 	if (uio->uio_offset != 0)
    504 		return (EINVAL);
    505 
    506 	xlen = *buflenp;
    507 
    508 	/* must be able to read the whole string in one go */
    509 	if (xlen < uio->uio_resid)
    510 		return (EMSGSIZE);
    511 	xlen = uio->uio_resid;
    512 
    513 	if ((error = uiomove(bf, xlen, uio)) != 0)
    514 		return (error);
    515 
    516 	/* allow multiple writes without seeks */
    517 	uio->uio_offset = 0;
    518 
    519 	/* cleanup string and remove trailing newline */
    520 	bf[xlen] = '\0';
    521 	xlen = strlen(bf);
    522 	if (xlen > 0 && bf[xlen-1] == '\n')
    523 		bf[--xlen] = '\0';
    524 	*buflenp = xlen;
    525 
    526 	return (0);
    527 }
    528 
    529 const vfs_namemap_t *
    530 vfs_findname(const vfs_namemap_t *nm, const char *bf, int buflen)
    531 {
    532 
    533 	for (; nm->nm_name; nm++)
    534 		if (memcmp(bf, nm->nm_name, buflen+1) == 0)
    535 			return (nm);
    536 
    537 	return (0);
    538 }
    539 
    540 /*
    541  * Initialize pfsnode hash table.
    542  */
    543 void
    544 procfs_hashinit(void)
    545 {
    546 	mutex_init(&pfs_hashlock, MUTEX_DEFAULT, IPL_NONE);
    547 	mutex_init(&pfs_ihash_lock, MUTEX_DEFAULT, IPL_NONE);
    548 	pfs_hashtbl = hashinit(desiredvnodes / 4, HASH_LIST, true, &pfs_ihash);
    549 }
    550 
    551 void
    552 procfs_hashreinit(void)
    553 {
    554 	struct pfsnode *pp;
    555 	struct pfs_hashhead *oldhash, *hash;
    556 	u_long i, oldmask, mask, val;
    557 
    558 	hash = hashinit(desiredvnodes / 4, HASH_LIST, true, &mask);
    559 
    560 	mutex_enter(&pfs_ihash_lock);
    561 	oldhash = pfs_hashtbl;
    562 	oldmask = pfs_ihash;
    563 	pfs_hashtbl = hash;
    564 	pfs_ihash = mask;
    565 	for (i = 0; i <= oldmask; i++) {
    566 		while ((pp = LIST_FIRST(&oldhash[i])) != NULL) {
    567 			LIST_REMOVE(pp, pfs_hash);
    568 			val = PFSPIDHASH(pp->pfs_pid);
    569 			LIST_INSERT_HEAD(&hash[val], pp, pfs_hash);
    570 		}
    571 	}
    572 	mutex_exit(&pfs_ihash_lock);
    573 	hashdone(oldhash, HASH_LIST, oldmask);
    574 }
    575 
    576 /*
    577  * Free pfsnode hash table.
    578  */
    579 void
    580 procfs_hashdone(void)
    581 {
    582 	hashdone(pfs_hashtbl, HASH_LIST, pfs_ihash);
    583 	mutex_destroy(&pfs_hashlock);
    584 	mutex_destroy(&pfs_ihash_lock);
    585 }
    586 
    587 struct vnode *
    588 procfs_hashget(pid_t pid, pfstype type, int fd, struct mount *mp, int flags)
    589 {
    590 	struct pfs_hashhead *ppp;
    591 	struct pfsnode *pp;
    592 	struct vnode *vp;
    593 
    594 loop:
    595 	mutex_enter(&pfs_ihash_lock);
    596 	ppp = &pfs_hashtbl[PFSPIDHASH(pid)];
    597 	LIST_FOREACH(pp, ppp, pfs_hash) {
    598 		vp = PFSTOV(pp);
    599 		if (pid == pp->pfs_pid && pp->pfs_type == type &&
    600 		    pp->pfs_fd == fd && vp->v_mount == mp) {
    601 		    	if (flags == 0) {
    602 				mutex_exit(&pfs_ihash_lock);
    603 			} else {
    604 				mutex_enter(vp->v_interlock);
    605 				mutex_exit(&pfs_ihash_lock);
    606 				if (vget(vp, flags))
    607 					goto loop;
    608 			}
    609 			return (vp);
    610 		}
    611 	}
    612 	mutex_exit(&pfs_ihash_lock);
    613 	return (NULL);
    614 }
    615 
    616 /*
    617  * Insert the pfsnode into the hash table and lock it.
    618  */
    619 void
    620 procfs_hashins(struct pfsnode *pp)
    621 {
    622 	struct pfs_hashhead *ppp;
    623 
    624 	/* lock the pfsnode, then put it on the appropriate hash list */
    625 	VOP_LOCK(PFSTOV(pp), LK_EXCLUSIVE);
    626 
    627 	mutex_enter(&pfs_ihash_lock);
    628 	ppp = &pfs_hashtbl[PFSPIDHASH(pp->pfs_pid)];
    629 	LIST_INSERT_HEAD(ppp, pp, pfs_hash);
    630 	mutex_exit(&pfs_ihash_lock);
    631 }
    632 
    633 /*
    634  * Remove the pfsnode from the hash table.
    635  */
    636 void
    637 procfs_hashrem(struct pfsnode *pp)
    638 {
    639 	mutex_enter(&pfs_ihash_lock);
    640 	LIST_REMOVE(pp, pfs_hash);
    641 	mutex_exit(&pfs_ihash_lock);
    642 }
    643 
    644 void
    645 procfs_revoke_vnodes(struct proc *p, void *arg)
    646 {
    647 	struct pfsnode *pfs, *pnext;
    648 	struct vnode *vp;
    649 	struct mount *mp = (struct mount *)arg;
    650 	struct pfs_hashhead *ppp;
    651 
    652 	if (!(p->p_flag & PK_SUGID))
    653 		return;
    654 
    655 	mutex_enter(&pfs_ihash_lock);
    656 	ppp = &pfs_hashtbl[PFSPIDHASH(p->p_pid)];
    657 	for (pfs = LIST_FIRST(ppp); pfs; pfs = pnext) {
    658 		vp = PFSTOV(pfs);
    659 		pnext = LIST_NEXT(pfs, pfs_hash);
    660 		mutex_enter(vp->v_interlock);
    661 		if (vp->v_usecount > 0 && pfs->pfs_pid == p->p_pid &&
    662 		    vp->v_mount == mp) {
    663 		    	vp->v_usecount++;
    664 		    	mutex_exit(vp->v_interlock);
    665 			mutex_exit(&pfs_ihash_lock);
    666 			VOP_REVOKE(vp, REVOKEALL);
    667 			vrele(vp);
    668 			mutex_enter(&pfs_ihash_lock);
    669 		} else {
    670 			mutex_exit(vp->v_interlock);
    671 		}
    672 	}
    673 	mutex_exit(&pfs_ihash_lock);
    674 }
    675 
    676 int
    677 procfs_proc_lock(int pid, struct proc **bunghole, int notfound)
    678 {
    679 	struct proc *tp;
    680 	int error = 0;
    681 
    682 	mutex_enter(proc_lock);
    683 
    684 	if (pid == 0)
    685 		tp = &proc0;
    686 	else if ((tp = proc_find(pid)) == NULL)
    687 		error = notfound;
    688 	if (tp != NULL && !rw_tryenter(&tp->p_reflock, RW_READER))
    689 		error = EBUSY;
    690 
    691 	mutex_exit(proc_lock);
    692 
    693 	*bunghole = tp;
    694 	return error;
    695 }
    696 
    697 void
    698 procfs_proc_unlock(struct proc *p)
    699 {
    700 
    701 	rw_exit(&p->p_reflock);
    702 }
    703 
    704 int
    705 procfs_doemul(struct lwp *curl, struct proc *p,
    706     struct pfsnode *pfs, struct uio *uio)
    707 {
    708 	const char *ename = p->p_emul->e_name;
    709 	return uiomove_frombuf(__UNCONST(ename), strlen(ename), uio);
    710 }
    711