Home | History | Annotate | Line # | Download | only in procfs
procfs_subr.c revision 1.82
      1 /*	$NetBSD: procfs_subr.c,v 1.82 2007/11/07 00:23:38 ad Exp $	*/
      2 
      3 /*-
      4  * Copyright (c) 2006, 2007 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  * 3. All advertising materials mentioning features or use of this software
     19  *    must display the following acknowledgement:
     20  *	This product includes software developed by the NetBSD
     21  *	Foundation, Inc. and its contributors.
     22  * 4. Neither the name of The NetBSD Foundation nor the names of its
     23  *    contributors may be used to endorse or promote products derived
     24  *    from this software without specific prior written permission.
     25  *
     26  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     27  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     28  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     29  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     30  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     31  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     32  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     33  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     34  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     35  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     36  * POSSIBILITY OF SUCH DAMAGE.
     37  */
     38 
     39 /*
     40  * Copyright (c) 1993
     41  *	The Regents of the University of California.  All rights reserved.
     42  *
     43  * This code is derived from software contributed to Berkeley by
     44  * Jan-Simon Pendry.
     45  *
     46  * Redistribution and use in source and binary forms, with or without
     47  * modification, are permitted provided that the following conditions
     48  * are met:
     49  * 1. Redistributions of source code must retain the above copyright
     50  *    notice, this list of conditions and the following disclaimer.
     51  * 2. Redistributions in binary form must reproduce the above copyright
     52  *    notice, this list of conditions and the following disclaimer in the
     53  *    documentation and/or other materials provided with the distribution.
     54  * 3. Neither the name of the University nor the names of its contributors
     55  *    may be used to endorse or promote products derived from this software
     56  *    without specific prior written permission.
     57  *
     58  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     59  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     60  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     61  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     62  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     63  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     64  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     65  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     66  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     67  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     68  * SUCH DAMAGE.
     69  *
     70  *	@(#)procfs_subr.c	8.6 (Berkeley) 5/14/95
     71  */
     72 
     73 /*
     74  * Copyright (c) 1994 Christopher G. Demetriou.  All rights reserved.
     75  * Copyright (c) 1993 Jan-Simon Pendry
     76  *
     77  * This code is derived from software contributed to Berkeley by
     78  * Jan-Simon Pendry.
     79  *
     80  * Redistribution and use in source and binary forms, with or without
     81  * modification, are permitted provided that the following conditions
     82  * are met:
     83  * 1. Redistributions of source code must retain the above copyright
     84  *    notice, this list of conditions and the following disclaimer.
     85  * 2. Redistributions in binary form must reproduce the above copyright
     86  *    notice, this list of conditions and the following disclaimer in the
     87  *    documentation and/or other materials provided with the distribution.
     88  * 3. All advertising materials mentioning features or use of this software
     89  *    must display the following acknowledgement:
     90  *	This product includes software developed by the University of
     91  *	California, Berkeley and its contributors.
     92  * 4. Neither the name of the University nor the names of its contributors
     93  *    may be used to endorse or promote products derived from this software
     94  *    without specific prior written permission.
     95  *
     96  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     97  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     98  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     99  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
    100  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
    101  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
    102  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
    103  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
    104  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
    105  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
    106  * SUCH DAMAGE.
    107  *
    108  *	@(#)procfs_subr.c	8.6 (Berkeley) 5/14/95
    109  */
    110 
    111 #include <sys/cdefs.h>
    112 __KERNEL_RCSID(0, "$NetBSD: procfs_subr.c,v 1.82 2007/11/07 00:23:38 ad Exp $");
    113 
    114 #include <sys/param.h>
    115 #include <sys/systm.h>
    116 #include <sys/time.h>
    117 #include <sys/kernel.h>
    118 #include <sys/proc.h>
    119 #include <sys/vnode.h>
    120 #include <sys/malloc.h>
    121 #include <sys/stat.h>
    122 #include <sys/file.h>
    123 #include <sys/filedesc.h>
    124 #include <sys/kauth.h>
    125 
    126 #include <miscfs/procfs/procfs.h>
    127 
    128 void procfs_hashins(struct pfsnode *);
    129 void procfs_hashrem(struct pfsnode *);
    130 struct vnode *procfs_hashget(pid_t, pfstype, int, struct mount *, int);
    131 
    132 LIST_HEAD(pfs_hashhead, pfsnode) *pfs_hashtbl;
    133 u_long	pfs_ihash;	/* size of hash table - 1 */
    134 #define PFSPIDHASH(pid)	((pid) & pfs_ihash)
    135 
    136 kmutex_t pfs_hashlock;
    137 kmutex_t pfs_ihash_lock;
    138 
    139 #define	ISSET(t, f)	((t) & (f))
    140 
    141 /*
    142  * allocate a pfsnode/vnode pair.  the vnode is
    143  * referenced, and locked.
    144  *
    145  * the pid, pfs_type, and mount point uniquely
    146  * identify a pfsnode.  the mount point is needed
    147  * because someone might mount this filesystem
    148  * twice.
    149  *
    150  * all pfsnodes are maintained on a singly-linked
    151  * list.  new nodes are only allocated when they cannot
    152  * be found on this list.  entries on the list are
    153  * removed when the vfs reclaim entry is called.
    154  *
    155  * a single lock is kept for the entire list.  this is
    156  * needed because the getnewvnode() function can block
    157  * waiting for a vnode to become free, in which case there
    158  * may be more than one process trying to get the same
    159  * vnode.  this lock is only taken if we are going to
    160  * call getnewvnode, since the kernel itself is single-threaded.
    161  *
    162  * if an entry is found on the list, then call vget() to
    163  * take a reference.  this is done because there may be
    164  * zero references to it and so it needs to removed from
    165  * the vnode free list.
    166  */
    167 int
    168 procfs_allocvp(mp, vpp, pid, pfs_type, fd, p)
    169 	struct mount *mp;
    170 	struct vnode **vpp;
    171 	pid_t pid;
    172 	pfstype pfs_type;
    173 	int fd;
    174 	struct proc *p;
    175 {
    176 	struct pfsnode *pfs;
    177 	struct vnode *vp;
    178 	int error;
    179 
    180  retry:
    181 	*vpp = procfs_hashget(pid, pfs_type, fd, mp, LK_EXCLUSIVE);
    182 	if (*vpp != NULL)
    183 		return (0);
    184 
    185 	if ((error = getnewvnode(VT_PROCFS, mp, procfs_vnodeop_p, &vp)) != 0) {
    186 		*vpp = NULL;
    187 		return (error);
    188 	}
    189 	MALLOC(pfs, void *, sizeof(struct pfsnode), M_TEMP, M_WAITOK);
    190 
    191 	mutex_enter(&pfs_hashlock);
    192 	if ((*vpp = procfs_hashget(pid, pfs_type, fd, mp, 0)) != NULL) {
    193 		mutex_exit(&pfs_hashlock);
    194 		ungetnewvnode(vp);
    195 		FREE(pfs, M_TEMP);
    196 		goto retry;
    197 	}
    198 
    199 	vp->v_data = pfs;
    200 	pfs->pfs_pid = pid;
    201 	pfs->pfs_type = pfs_type;
    202 	pfs->pfs_vnode = vp;
    203 	pfs->pfs_flags = 0;
    204 	pfs->pfs_fileno = PROCFS_FILENO(pid, pfs_type, fd);
    205 	pfs->pfs_fd = fd;
    206 
    207 	switch (pfs_type) {
    208 	case PFSroot:	/* /proc = dr-xr-xr-x */
    209 		vp->v_vflag |= VV_ROOT;
    210 		/*FALLTHROUGH*/
    211 	case PFSproc:	/* /proc/N = dr-xr-xr-x */
    212 		pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
    213 		vp->v_type = VDIR;
    214 		break;
    215 
    216 	case PFScurproc:	/* /proc/curproc = lr-xr-xr-x */
    217 	case PFSself:	/* /proc/self    = lr-xr-xr-x */
    218 	case PFScwd:	/* /proc/N/cwd = lr-xr-xr-x */
    219 	case PFSchroot:	/* /proc/N/chroot = lr-xr-xr-x */
    220 	case PFSexe:	/* /proc/N/exe = lr-xr-xr-x */
    221 		pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
    222 		vp->v_type = VLNK;
    223 		break;
    224 
    225 	case PFSfd:
    226 		if (fd == -1) {	/* /proc/N/fd = dr-xr-xr-x */
    227 			pfs->pfs_mode = S_IRUSR|S_IXUSR;
    228 			vp->v_type = VDIR;
    229 		} else {	/* /proc/N/fd/M = [ps-]rw------- */
    230 			struct file *fp;
    231 			struct vnode *vxp;
    232 
    233 			fp = fd_getfile(p->p_fd, pfs->pfs_fd);
    234 			if (fp == NULL) {
    235 				error = EBADF;
    236 				goto bad;
    237 			}
    238 			FILE_USE(fp);
    239 
    240 			pfs->pfs_mode = S_IRUSR|S_IWUSR;
    241 			switch (fp->f_type) {
    242 			case DTYPE_VNODE:
    243 				vxp = (struct vnode *)fp->f_data;
    244 
    245 				/*
    246 				 * We make symlinks for directories
    247 				 * to avoid cycles.
    248 				 */
    249 				if (vxp->v_type == VDIR)
    250 					goto symlink;
    251 				vp->v_type = vxp->v_type;
    252 				break;
    253 			case DTYPE_PIPE:
    254 				vp->v_type = VFIFO;
    255 				break;
    256 			case DTYPE_SOCKET:
    257 				vp->v_type = VSOCK;
    258 				break;
    259 			case DTYPE_KQUEUE:
    260 			case DTYPE_MISC:
    261 			symlink:
    262 				pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|
    263 				    S_IXGRP|S_IROTH|S_IXOTH;
    264 				vp->v_type = VLNK;
    265 				break;
    266 			default:
    267 				error = EOPNOTSUPP;
    268 				FILE_UNUSE(fp, curlwp);
    269 				goto bad;
    270 			}
    271 			FILE_UNUSE(fp, curlwp);
    272 		}
    273 		break;
    274 
    275 	case PFSfile:	/* /proc/N/file = -rw------- */
    276 	case PFSmem:	/* /proc/N/mem = -rw------- */
    277 	case PFSregs:	/* /proc/N/regs = -rw------- */
    278 	case PFSfpregs:	/* /proc/N/fpregs = -rw------- */
    279 		pfs->pfs_mode = S_IRUSR|S_IWUSR;
    280 		vp->v_type = VREG;
    281 		break;
    282 
    283 	case PFSctl:	/* /proc/N/ctl = --w------ */
    284 	case PFSnote:	/* /proc/N/note = --w------ */
    285 	case PFSnotepg:	/* /proc/N/notepg = --w------ */
    286 		pfs->pfs_mode = S_IWUSR;
    287 		vp->v_type = VREG;
    288 		break;
    289 
    290 	case PFSmap:	/* /proc/N/map = -r--r--r-- */
    291 	case PFSmaps:	/* /proc/N/maps = -r--r--r-- */
    292 	case PFSstatus:	/* /proc/N/status = -r--r--r-- */
    293 	case PFSstat:	/* /proc/N/stat = -r--r--r-- */
    294 	case PFScmdline:	/* /proc/N/cmdline = -r--r--r-- */
    295 	case PFSemul:	/* /proc/N/emul = -r--r--r-- */
    296 	case PFSmeminfo:	/* /proc/meminfo = -r--r--r-- */
    297 	case PFScpustat:	/* /proc/stat = -r--r--r-- */
    298 	case PFSdevices:	/* /proc/devices = -r--r--r-- */
    299 	case PFScpuinfo:	/* /proc/cpuinfo = -r--r--r-- */
    300 	case PFSuptime:	/* /proc/uptime = -r--r--r-- */
    301 	case PFSmounts:	/* /proc/mounts = -r--r--r-- */
    302 	case PFSloadavg:	/* /proc/loadavg = -r--r--r-- */
    303 	case PFSstatm:	/* /proc/N/statm = -r--r--r-- */
    304 		pfs->pfs_mode = S_IRUSR|S_IRGRP|S_IROTH;
    305 		vp->v_type = VREG;
    306 		break;
    307 
    308 #ifdef __HAVE_PROCFS_MACHDEP
    309 	PROCFS_MACHDEP_NODETYPE_CASES
    310 		procfs_machdep_allocvp(vp);
    311 		break;
    312 #endif
    313 
    314 	default:
    315 		panic("procfs_allocvp");
    316 	}
    317 
    318 	procfs_hashins(pfs);
    319 	uvm_vnp_setsize(vp, 0);
    320 	mutex_exit(&pfs_hashlock);
    321 
    322 	*vpp = vp;
    323 	return (0);
    324 
    325  bad:
    326 	mutex_exit(&pfs_hashlock);
    327 	FREE(pfs, M_TEMP);
    328 	vp->v_data = NULL;
    329 	ungetnewvnode(vp);
    330 	return (error);
    331 }
    332 
    333 int
    334 procfs_freevp(vp)
    335 	struct vnode *vp;
    336 {
    337 	struct pfsnode *pfs = VTOPFS(vp);
    338 
    339 	procfs_hashrem(pfs);
    340 
    341 	FREE(vp->v_data, M_TEMP);
    342 	vp->v_data = 0;
    343 	return (0);
    344 }
    345 
    346 int
    347 procfs_rw(v)
    348 	void *v;
    349 {
    350 	struct vop_read_args *ap = v;
    351 	struct vnode *vp = ap->a_vp;
    352 	struct uio *uio = ap->a_uio;
    353 	struct lwp *curl;
    354 	struct lwp *l;
    355 	struct pfsnode *pfs = VTOPFS(vp);
    356 	struct proc *p;
    357 	int error;
    358 
    359 	if (uio->uio_offset < 0)
    360 		return EINVAL;
    361 
    362 	if ((error = procfs_proc_lock(pfs->pfs_pid, &p, ESRCH)) != 0)
    363 		return error;
    364 
    365 	curl = curlwp;
    366 
    367 	/*
    368 	 * Do not allow init to be modified while in secure mode; it
    369 	 * could be duped into changing the security level.
    370 	 */
    371 #define	M2K(m)	((m) == UIO_READ ? KAUTH_REQ_PROCESS_CANPROCFS_READ : \
    372 		 KAUTH_REQ_PROCESS_CANPROCFS_WRITE)
    373 	mutex_enter(&p->p_mutex);
    374 	error = kauth_authorize_process(curl->l_cred, KAUTH_PROCESS_CANPROCFS,
    375 	    p, pfs, KAUTH_ARG(M2K(uio->uio_rw)), NULL);
    376 	mutex_exit(&p->p_mutex);
    377 	if (error) {
    378 		procfs_proc_unlock(p);
    379 		return (error);
    380 	}
    381 #undef	M2K
    382 
    383 	mutex_enter(&p->p_smutex);
    384 	l = proc_representative_lwp(p, NULL, 1);
    385 	lwp_addref(l);
    386 	mutex_exit(&p->p_smutex);
    387 
    388 	switch (pfs->pfs_type) {
    389 	case PFSnote:
    390 	case PFSnotepg:
    391 		error = procfs_donote(curl, p, pfs, uio);
    392 		break;
    393 
    394 	case PFSregs:
    395 		error = procfs_doregs(curl, l, pfs, uio);
    396 		break;
    397 
    398 	case PFSfpregs:
    399 		error = procfs_dofpregs(curl, l, pfs, uio);
    400 		break;
    401 
    402 	case PFSctl:
    403 		error = procfs_doctl(curl, l, pfs, uio);
    404 		break;
    405 
    406 	case PFSstatus:
    407 		error = procfs_dostatus(curl, l, pfs, uio);
    408 		break;
    409 
    410 	case PFSstat:
    411 		error = procfs_do_pid_stat(curl, l, pfs, uio);
    412 		break;
    413 
    414 	case PFSmap:
    415 		error = procfs_domap(curl, p, pfs, uio, 0);
    416 		break;
    417 
    418 	case PFSmaps:
    419 		error = procfs_domap(curl, p, pfs, uio, 1);
    420 		break;
    421 
    422 	case PFSmem:
    423 		error = procfs_domem(curl, l, pfs, uio);
    424 		break;
    425 
    426 	case PFScmdline:
    427 		error = procfs_docmdline(curl, p, pfs, uio);
    428 		break;
    429 
    430 	case PFSmeminfo:
    431 		error = procfs_domeminfo(curl, p, pfs, uio);
    432 		break;
    433 
    434 	case PFSdevices:
    435 		error = procfs_dodevices(curl, p, pfs, uio);
    436 		break;
    437 
    438 	case PFScpuinfo:
    439 		error = procfs_docpuinfo(curl, p, pfs, uio);
    440 		break;
    441 
    442 	case PFScpustat:
    443 		error = procfs_docpustat(curl, p, pfs, uio);
    444 		break;
    445 
    446 	case PFSloadavg:
    447 		error = procfs_doloadavg(curl, p, pfs, uio);
    448 		break;
    449 
    450 	case PFSstatm:
    451 		error = procfs_do_pid_statm(curl, l, pfs, uio);
    452 		break;
    453 
    454 	case PFSfd:
    455 		error = procfs_dofd(curl, p, pfs, uio);
    456 		break;
    457 
    458 	case PFSuptime:
    459 		error = procfs_douptime(curl, p, pfs, uio);
    460 		break;
    461 
    462 	case PFSmounts:
    463 		error = procfs_domounts(curl, p, pfs, uio);
    464 		break;
    465 
    466 	case PFSemul:
    467 		error = procfs_doemul(curl, p, pfs, uio);
    468 		break;
    469 
    470 #ifdef __HAVE_PROCFS_MACHDEP
    471 	PROCFS_MACHDEP_NODETYPE_CASES
    472 		error = procfs_machdep_rw(curl, l, pfs, uio);
    473 		break;
    474 #endif
    475 
    476 	default:
    477 		error = EOPNOTSUPP;
    478 		break;
    479 	}
    480 
    481 	/*
    482 	 * Release the references that we acquired earlier.
    483 	 */
    484 	lwp_delref(l);
    485 	procfs_proc_unlock(p);
    486 
    487 	return (error);
    488 }
    489 
    490 /*
    491  * Get a string from userland into (bf).  Strip a trailing
    492  * nl character (to allow easy access from the shell).
    493  * The buffer should be *buflenp + 1 chars long.  vfs_getuserstr
    494  * will automatically add a nul char at the end.
    495  *
    496  * Returns 0 on success or the following errors
    497  *
    498  * EINVAL:    file offset is non-zero.
    499  * EMSGSIZE:  message is longer than kernel buffer
    500  * EFAULT:    user i/o buffer is not addressable
    501  */
    502 int
    503 vfs_getuserstr(uio, bf, buflenp)
    504 	struct uio *uio;
    505 	char *bf;
    506 	int *buflenp;
    507 {
    508 	int xlen;
    509 	int error;
    510 
    511 	if (uio->uio_offset != 0)
    512 		return (EINVAL);
    513 
    514 	xlen = *buflenp;
    515 
    516 	/* must be able to read the whole string in one go */
    517 	if (xlen < uio->uio_resid)
    518 		return (EMSGSIZE);
    519 	xlen = uio->uio_resid;
    520 
    521 	if ((error = uiomove(bf, xlen, uio)) != 0)
    522 		return (error);
    523 
    524 	/* allow multiple writes without seeks */
    525 	uio->uio_offset = 0;
    526 
    527 	/* cleanup string and remove trailing newline */
    528 	bf[xlen] = '\0';
    529 	xlen = strlen(bf);
    530 	if (xlen > 0 && bf[xlen-1] == '\n')
    531 		bf[--xlen] = '\0';
    532 	*buflenp = xlen;
    533 
    534 	return (0);
    535 }
    536 
    537 const vfs_namemap_t *
    538 vfs_findname(nm, bf, buflen)
    539 	const vfs_namemap_t *nm;
    540 	const char *bf;
    541 	int buflen;
    542 {
    543 
    544 	for (; nm->nm_name; nm++)
    545 		if (memcmp(bf, nm->nm_name, buflen+1) == 0)
    546 			return (nm);
    547 
    548 	return (0);
    549 }
    550 
    551 /*
    552  * Initialize pfsnode hash table.
    553  */
    554 void
    555 procfs_hashinit()
    556 {
    557 	mutex_init(&pfs_hashlock, MUTEX_DEFAULT, IPL_NONE);
    558 	mutex_init(&pfs_ihash_lock, MUTEX_DEFAULT, IPL_NONE);
    559 	pfs_hashtbl = hashinit(desiredvnodes / 4, HASH_LIST, M_UFSMNT,
    560 	    M_WAITOK, &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, M_UFSMNT, M_WAITOK,
    571 	    &mask);
    572 
    573 	mutex_enter(&pfs_ihash_lock);
    574 	oldhash = pfs_hashtbl;
    575 	oldmask = pfs_ihash;
    576 	pfs_hashtbl = hash;
    577 	pfs_ihash = mask;
    578 	for (i = 0; i <= oldmask; i++) {
    579 		while ((pp = LIST_FIRST(&oldhash[i])) != NULL) {
    580 			LIST_REMOVE(pp, pfs_hash);
    581 			val = PFSPIDHASH(pp->pfs_pid);
    582 			LIST_INSERT_HEAD(&hash[val], pp, pfs_hash);
    583 		}
    584 	}
    585 	mutex_exit(&pfs_ihash_lock);
    586 	hashdone(oldhash, M_UFSMNT);
    587 }
    588 
    589 /*
    590  * Free pfsnode hash table.
    591  */
    592 void
    593 procfs_hashdone()
    594 {
    595 	hashdone(pfs_hashtbl, M_UFSMNT);
    596 	mutex_destroy(&pfs_hashlock);
    597 	mutex_destroy(&pfs_ihash_lock);
    598 }
    599 
    600 struct vnode *
    601 procfs_hashget(pid, type, fd, mp, flags)
    602 	pid_t pid;
    603 	pfstype type;
    604 	int fd;
    605 	struct mount *mp;
    606 	int flags;
    607 {
    608 	struct pfs_hashhead *ppp;
    609 	struct pfsnode *pp;
    610 	struct vnode *vp;
    611 
    612 loop:
    613 	mutex_enter(&pfs_ihash_lock);
    614 	ppp = &pfs_hashtbl[PFSPIDHASH(pid)];
    615 	LIST_FOREACH(pp, ppp, pfs_hash) {
    616 		vp = PFSTOV(pp);
    617 		if (pid == pp->pfs_pid && pp->pfs_type == type &&
    618 		    pp->pfs_fd == fd && vp->v_mount == mp) {
    619 		    	if (flags == 0) {
    620 				mutex_exit(&pfs_ihash_lock);
    621 			} else {
    622 				simple_lock(&vp->v_interlock);
    623 				mutex_exit(&pfs_ihash_lock);
    624 				if (vget(vp, flags | LK_INTERLOCK))
    625 					goto loop;
    626 			}
    627 			return (vp);
    628 		}
    629 	}
    630 	mutex_exit(&pfs_ihash_lock);
    631 	return (NULL);
    632 }
    633 
    634 /*
    635  * Insert the pfsnode into the hash table and lock it.
    636  */
    637 void
    638 procfs_hashins(pp)
    639 	struct pfsnode *pp;
    640 {
    641 	struct pfs_hashhead *ppp;
    642 
    643 	/* lock the pfsnode, then put it on the appropriate hash list */
    644 	lockmgr(&pp->pfs_vnode->v_lock, LK_EXCLUSIVE, NULL);
    645 
    646 	mutex_enter(&pfs_ihash_lock);
    647 	ppp = &pfs_hashtbl[PFSPIDHASH(pp->pfs_pid)];
    648 	LIST_INSERT_HEAD(ppp, pp, pfs_hash);
    649 	mutex_exit(&pfs_ihash_lock);
    650 }
    651 
    652 /*
    653  * Remove the pfsnode from the hash table.
    654  */
    655 void
    656 procfs_hashrem(pp)
    657 	struct pfsnode *pp;
    658 {
    659 	mutex_enter(&pfs_ihash_lock);
    660 	LIST_REMOVE(pp, pfs_hash);
    661 	mutex_exit(&pfs_ihash_lock);
    662 }
    663 
    664 void
    665 procfs_revoke_vnodes(p, arg)
    666 	struct proc *p;
    667 	void *arg;
    668 {
    669 	struct pfsnode *pfs, *pnext;
    670 	struct vnode *vp;
    671 	struct mount *mp = (struct mount *)arg;
    672 	struct pfs_hashhead *ppp;
    673 
    674 	if (!(p->p_flag & PK_SUGID))
    675 		return;
    676 
    677 	mutex_enter(&pfs_ihash_lock);
    678 	ppp = &pfs_hashtbl[PFSPIDHASH(p->p_pid)];
    679 	for (pfs = LIST_FIRST(ppp); pfs; pfs = pnext) {
    680 		vp = PFSTOV(pfs);
    681 		pnext = LIST_NEXT(pfs, pfs_hash);
    682 		simple_lock(&vp->v_interlock);
    683 		if (vp->v_usecount > 0 && pfs->pfs_pid == p->p_pid &&
    684 		    vp->v_mount == mp) {
    685 		    	vp->v_usecount++;
    686 		    	simple_unlock(&vp->v_interlock);
    687 			mutex_exit(&pfs_ihash_lock);
    688 			VOP_REVOKE(vp, REVOKEALL);
    689 			vrele(vp);
    690 			mutex_enter(&pfs_ihash_lock);
    691 		} else {
    692 			simple_unlock(&vp->v_interlock);
    693 		}
    694 	}
    695 	mutex_exit(&pfs_ihash_lock);
    696 }
    697 
    698 int
    699 procfs_proc_lock(int pid, struct proc **bunghole, int notfound)
    700 {
    701 	struct proc *tp;
    702 	int error = 0;
    703 
    704 	mutex_enter(&proclist_lock);
    705 
    706 	if (pid == 0)
    707 		tp = &proc0;
    708 	else if ((tp = p_find(pid, PFIND_LOCKED)) == NULL)
    709 		error = notfound;
    710 	if (tp != NULL && !rw_tryenter(&tp->p_reflock, RW_READER))
    711 		error = EBUSY;
    712 
    713 	mutex_exit(&proclist_lock);
    714 
    715 	*bunghole = tp;
    716 	return error;
    717 }
    718 
    719 void
    720 procfs_proc_unlock(struct proc *p)
    721 {
    722 
    723 	rw_exit(&p->p_reflock);
    724 }
    725 
    726 int
    727 procfs_doemul(struct lwp *curl, struct proc *p,
    728     struct pfsnode *pfs, struct uio *uio)
    729 {
    730 	const char *ename = p->p_emul->e_name;
    731 	return uiomove_frombuf(__UNCONST(ename), strlen(ename), uio);
    732 }
    733