Home | History | Annotate | Line # | Download | only in procfs
procfs_subr.c revision 1.100.6.1
      1 /*	$NetBSD: procfs_subr.c,v 1.100.6.1 2012/06/02 11:09:36 mrg 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 1.100.6.1 2012/06/02 11:09:36 mrg 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 PFStask:	/* /proc/N/task = dr-xr-xr-x */
    206 		if (fd == -1) {
    207 			pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|
    208 			    S_IROTH|S_IXOTH;
    209 			vp->v_type = VDIR;
    210 			break;
    211 		}
    212 		/*FALLTHROUGH*/
    213 	case PFScurproc:	/* /proc/curproc = lr-xr-xr-x */
    214 	case PFSself:	/* /proc/self    = lr-xr-xr-x */
    215 	case PFScwd:	/* /proc/N/cwd = lr-xr-xr-x */
    216 	case PFSchroot:	/* /proc/N/chroot = lr-xr-xr-x */
    217 	case PFSexe:	/* /proc/N/exe = lr-xr-xr-x */
    218 		pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
    219 		vp->v_type = VLNK;
    220 		break;
    221 
    222 	case PFSfd:
    223 		if (fd == -1) {	/* /proc/N/fd = dr-x------ */
    224 			pfs->pfs_mode = S_IRUSR|S_IXUSR;
    225 			vp->v_type = VDIR;
    226 		} else {	/* /proc/N/fd/M = [ps-]rw------- */
    227 			file_t *fp;
    228 			vnode_t *vxp;
    229 
    230 			if ((fp = fd_getfile2(p, pfs->pfs_fd)) == NULL) {
    231 				error = EBADF;
    232 				goto bad;
    233 			}
    234 
    235 			pfs->pfs_mode = S_IRUSR|S_IWUSR;
    236 			switch (fp->f_type) {
    237 			case DTYPE_VNODE:
    238 				vxp = fp->f_data;
    239 
    240 				/*
    241 				 * We make symlinks for directories
    242 				 * to avoid cycles.
    243 				 */
    244 				if (vxp->v_type == VDIR)
    245 					goto symlink;
    246 				vp->v_type = vxp->v_type;
    247 				break;
    248 			case DTYPE_PIPE:
    249 				vp->v_type = VFIFO;
    250 				break;
    251 			case DTYPE_SOCKET:
    252 				vp->v_type = VSOCK;
    253 				break;
    254 			case DTYPE_KQUEUE:
    255 			case DTYPE_MISC:
    256 			symlink:
    257 				pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|
    258 				    S_IXGRP|S_IROTH|S_IXOTH;
    259 				vp->v_type = VLNK;
    260 				break;
    261 			default:
    262 				error = EOPNOTSUPP;
    263 				closef(fp);
    264 				goto bad;
    265 			}
    266 			closef(fp);
    267 		}
    268 		break;
    269 
    270 	case PFSfile:	/* /proc/N/file = -rw------- */
    271 	case PFSmem:	/* /proc/N/mem = -rw------- */
    272 	case PFSregs:	/* /proc/N/regs = -rw------- */
    273 	case PFSfpregs:	/* /proc/N/fpregs = -rw------- */
    274 		pfs->pfs_mode = S_IRUSR|S_IWUSR;
    275 		vp->v_type = VREG;
    276 		break;
    277 
    278 	case PFSctl:	/* /proc/N/ctl = --w------ */
    279 	case PFSnote:	/* /proc/N/note = --w------ */
    280 	case PFSnotepg:	/* /proc/N/notepg = --w------ */
    281 		pfs->pfs_mode = S_IWUSR;
    282 		vp->v_type = VREG;
    283 		break;
    284 
    285 	case PFSmap:	/* /proc/N/map = -r--r--r-- */
    286 	case PFSmaps:	/* /proc/N/maps = -r--r--r-- */
    287 	case PFSstatus:	/* /proc/N/status = -r--r--r-- */
    288 	case PFSstat:	/* /proc/N/stat = -r--r--r-- */
    289 	case PFScmdline:	/* /proc/N/cmdline = -r--r--r-- */
    290 	case PFSemul:	/* /proc/N/emul = -r--r--r-- */
    291 	case PFSmeminfo:	/* /proc/meminfo = -r--r--r-- */
    292 	case PFScpustat:	/* /proc/stat = -r--r--r-- */
    293 	case PFSdevices:	/* /proc/devices = -r--r--r-- */
    294 	case PFScpuinfo:	/* /proc/cpuinfo = -r--r--r-- */
    295 	case PFSuptime:	/* /proc/uptime = -r--r--r-- */
    296 	case PFSmounts:	/* /proc/mounts = -r--r--r-- */
    297 	case PFSloadavg:	/* /proc/loadavg = -r--r--r-- */
    298 	case PFSstatm:	/* /proc/N/statm = -r--r--r-- */
    299 	case PFSversion:	/* /proc/version = -r--r--r-- */
    300 		pfs->pfs_mode = S_IRUSR|S_IRGRP|S_IROTH;
    301 		vp->v_type = VREG;
    302 		break;
    303 
    304 #ifdef __HAVE_PROCFS_MACHDEP
    305 	PROCFS_MACHDEP_NODETYPE_CASES
    306 		procfs_machdep_allocvp(vp);
    307 		break;
    308 #endif
    309 
    310 	default:
    311 		panic("procfs_allocvp");
    312 	}
    313 
    314 	procfs_hashins(pfs);
    315 	uvm_vnp_setsize(vp, 0);
    316 	mutex_exit(&pfs_hashlock);
    317 
    318 	*vpp = vp;
    319 	return (0);
    320 
    321  bad:
    322 	mutex_exit(&pfs_hashlock);
    323 	free(pfs, M_TEMP);
    324 	vp->v_data = NULL;
    325 	ungetnewvnode(vp);
    326 	return (error);
    327 }
    328 
    329 int
    330 procfs_freevp(struct vnode *vp)
    331 {
    332 	struct pfsnode *pfs = VTOPFS(vp);
    333 
    334 	procfs_hashrem(pfs);
    335 
    336 	free(vp->v_data, M_TEMP);
    337 	vp->v_data = NULL;
    338 	return (0);
    339 }
    340 
    341 int
    342 procfs_rw(void *v)
    343 {
    344 	struct vop_read_args *ap = v;
    345 	struct vnode *vp = ap->a_vp;
    346 	struct uio *uio = ap->a_uio;
    347 	struct lwp *curl;
    348 	struct lwp *l;
    349 	struct pfsnode *pfs = VTOPFS(vp);
    350 	struct proc *p;
    351 	int error;
    352 
    353 	if (uio->uio_offset < 0)
    354 		return EINVAL;
    355 
    356 	if ((error = procfs_proc_lock(pfs->pfs_pid, &p, ESRCH)) != 0)
    357 		return error;
    358 
    359 	curl = curlwp;
    360 
    361 	/*
    362 	 * Do not allow init to be modified while in secure mode; it
    363 	 * could be duped into changing the security level.
    364 	 */
    365 #define	M2K(m)	((m) == UIO_READ ? KAUTH_REQ_PROCESS_PROCFS_READ : \
    366 		 KAUTH_REQ_PROCESS_PROCFS_WRITE)
    367 	mutex_enter(p->p_lock);
    368 	error = kauth_authorize_process(curl->l_cred, KAUTH_PROCESS_PROCFS,
    369 	    p, pfs, KAUTH_ARG(M2K(uio->uio_rw)), NULL);
    370 	mutex_exit(p->p_lock);
    371 	if (error) {
    372 		procfs_proc_unlock(p);
    373 		return (error);
    374 	}
    375 #undef	M2K
    376 
    377 	mutex_enter(p->p_lock);
    378 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
    379 		if (l->l_stat != LSZOMB)
    380 			break;
    381 	}
    382 	/* Process is exiting if no-LWPS or all LWPs are LSZOMB */
    383 	if (l == NULL) {
    384 		mutex_exit(p->p_lock);
    385 		procfs_proc_unlock(p);
    386 		return ESRCH;
    387 	}
    388 
    389 	lwp_addref(l);
    390 	mutex_exit(p->p_lock);
    391 
    392 	switch (pfs->pfs_type) {
    393 	case PFSnote:
    394 	case PFSnotepg:
    395 		error = procfs_donote(curl, p, pfs, uio);
    396 		break;
    397 
    398 	case PFSregs:
    399 		error = procfs_doregs(curl, l, pfs, uio);
    400 		break;
    401 
    402 	case PFSfpregs:
    403 		error = procfs_dofpregs(curl, l, pfs, uio);
    404 		break;
    405 
    406 	case PFSctl:
    407 		error = procfs_doctl(curl, l, pfs, uio);
    408 		break;
    409 
    410 	case PFSstatus:
    411 		error = procfs_dostatus(curl, l, pfs, uio);
    412 		break;
    413 
    414 	case PFSstat:
    415 		error = procfs_do_pid_stat(curl, l, pfs, uio);
    416 		break;
    417 
    418 	case PFSmap:
    419 		error = procfs_domap(curl, p, pfs, uio, 0);
    420 		break;
    421 
    422 	case PFSmaps:
    423 		error = procfs_domap(curl, p, pfs, uio, 1);
    424 		break;
    425 
    426 	case PFSmem:
    427 		error = procfs_domem(curl, l, pfs, uio);
    428 		break;
    429 
    430 	case PFScmdline:
    431 		error = procfs_docmdline(curl, p, pfs, uio);
    432 		break;
    433 
    434 	case PFSmeminfo:
    435 		error = procfs_domeminfo(curl, p, pfs, uio);
    436 		break;
    437 
    438 	case PFSdevices:
    439 		error = procfs_dodevices(curl, p, pfs, uio);
    440 		break;
    441 
    442 	case PFScpuinfo:
    443 		error = procfs_docpuinfo(curl, p, pfs, uio);
    444 		break;
    445 
    446 	case PFScpustat:
    447 		error = procfs_docpustat(curl, p, pfs, uio);
    448 		break;
    449 
    450 	case PFSloadavg:
    451 		error = procfs_doloadavg(curl, p, pfs, uio);
    452 		break;
    453 
    454 	case PFSstatm:
    455 		error = procfs_do_pid_statm(curl, l, pfs, uio);
    456 		break;
    457 
    458 	case PFSfd:
    459 		error = procfs_dofd(curl, p, pfs, uio);
    460 		break;
    461 
    462 	case PFSuptime:
    463 		error = procfs_douptime(curl, p, pfs, uio);
    464 		break;
    465 
    466 	case PFSmounts:
    467 		error = procfs_domounts(curl, p, pfs, uio);
    468 		break;
    469 
    470 	case PFSemul:
    471 		error = procfs_doemul(curl, p, pfs, uio);
    472 		break;
    473 
    474 	case PFSversion:
    475 		error = procfs_doversion(curl, p, pfs, uio);
    476 		break;
    477 
    478 #ifdef __HAVE_PROCFS_MACHDEP
    479 	PROCFS_MACHDEP_NODETYPE_CASES
    480 		error = procfs_machdep_rw(curl, l, pfs, uio);
    481 		break;
    482 #endif
    483 
    484 	default:
    485 		error = EOPNOTSUPP;
    486 		break;
    487 	}
    488 
    489 	/*
    490 	 * Release the references that we acquired earlier.
    491 	 */
    492 	lwp_delref(l);
    493 	procfs_proc_unlock(p);
    494 
    495 	return (error);
    496 }
    497 
    498 /*
    499  * Get a string from userland into (bf).  Strip a trailing
    500  * nl character (to allow easy access from the shell).
    501  * The buffer should be *buflenp + 1 chars long.  vfs_getuserstr
    502  * will automatically add a nul char at the end.
    503  *
    504  * Returns 0 on success or the following errors
    505  *
    506  * EINVAL:    file offset is non-zero.
    507  * EMSGSIZE:  message is longer than kernel buffer
    508  * EFAULT:    user i/o buffer is not addressable
    509  */
    510 int
    511 vfs_getuserstr(struct uio *uio, char *bf, int *buflenp)
    512 {
    513 	int xlen;
    514 	int error;
    515 
    516 	if (uio->uio_offset != 0)
    517 		return (EINVAL);
    518 
    519 	xlen = *buflenp;
    520 
    521 	/* must be able to read the whole string in one go */
    522 	if (xlen < uio->uio_resid)
    523 		return (EMSGSIZE);
    524 	xlen = uio->uio_resid;
    525 
    526 	if ((error = uiomove(bf, xlen, uio)) != 0)
    527 		return (error);
    528 
    529 	/* allow multiple writes without seeks */
    530 	uio->uio_offset = 0;
    531 
    532 	/* cleanup string and remove trailing newline */
    533 	bf[xlen] = '\0';
    534 	xlen = strlen(bf);
    535 	if (xlen > 0 && bf[xlen-1] == '\n')
    536 		bf[--xlen] = '\0';
    537 	*buflenp = xlen;
    538 
    539 	return (0);
    540 }
    541 
    542 const vfs_namemap_t *
    543 vfs_findname(const vfs_namemap_t *nm, const char *bf, int buflen)
    544 {
    545 
    546 	for (; nm->nm_name; nm++)
    547 		if (memcmp(bf, nm->nm_name, buflen+1) == 0)
    548 			return (nm);
    549 
    550 	return (0);
    551 }
    552 
    553 /*
    554  * Initialize pfsnode hash table.
    555  */
    556 void
    557 procfs_hashinit(void)
    558 {
    559 	mutex_init(&pfs_hashlock, MUTEX_DEFAULT, IPL_NONE);
    560 	mutex_init(&pfs_ihash_lock, MUTEX_DEFAULT, IPL_NONE);
    561 	pfs_hashtbl = hashinit(desiredvnodes / 4, HASH_LIST, true, &pfs_ihash);
    562 }
    563 
    564 void
    565 procfs_hashreinit(void)
    566 {
    567 	struct pfsnode *pp;
    568 	struct pfs_hashhead *oldhash, *hash;
    569 	u_long i, oldmask, mask, val;
    570 
    571 	hash = hashinit(desiredvnodes / 4, HASH_LIST, true, &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, HASH_LIST, oldmask);
    587 }
    588 
    589 /*
    590  * Free pfsnode hash table.
    591  */
    592 void
    593 procfs_hashdone(void)
    594 {
    595 	hashdone(pfs_hashtbl, HASH_LIST, pfs_ihash);
    596 	mutex_destroy(&pfs_hashlock);
    597 	mutex_destroy(&pfs_ihash_lock);
    598 }
    599 
    600 struct vnode *
    601 procfs_hashget(pid_t pid, pfstype type, int fd, struct mount *mp, int flags)
    602 {
    603 	struct pfs_hashhead *ppp;
    604 	struct pfsnode *pp;
    605 	struct vnode *vp;
    606 
    607 loop:
    608 	mutex_enter(&pfs_ihash_lock);
    609 	ppp = &pfs_hashtbl[PFSPIDHASH(pid)];
    610 	LIST_FOREACH(pp, ppp, pfs_hash) {
    611 		vp = PFSTOV(pp);
    612 		if (pid == pp->pfs_pid && pp->pfs_type == type &&
    613 		    pp->pfs_fd == fd && vp->v_mount == mp) {
    614 		    	if (flags == 0) {
    615 				mutex_exit(&pfs_ihash_lock);
    616 			} else {
    617 				mutex_enter(vp->v_interlock);
    618 				mutex_exit(&pfs_ihash_lock);
    619 				if (vget(vp, flags))
    620 					goto loop;
    621 			}
    622 			return (vp);
    623 		}
    624 	}
    625 	mutex_exit(&pfs_ihash_lock);
    626 	return (NULL);
    627 }
    628 
    629 /*
    630  * Insert the pfsnode into the hash table and lock it.
    631  */
    632 void
    633 procfs_hashins(struct pfsnode *pp)
    634 {
    635 	struct pfs_hashhead *ppp;
    636 
    637 	/* lock the pfsnode, then put it on the appropriate hash list */
    638 	VOP_LOCK(PFSTOV(pp), LK_EXCLUSIVE);
    639 
    640 	mutex_enter(&pfs_ihash_lock);
    641 	ppp = &pfs_hashtbl[PFSPIDHASH(pp->pfs_pid)];
    642 	LIST_INSERT_HEAD(ppp, pp, pfs_hash);
    643 	mutex_exit(&pfs_ihash_lock);
    644 }
    645 
    646 /*
    647  * Remove the pfsnode from the hash table.
    648  */
    649 void
    650 procfs_hashrem(struct pfsnode *pp)
    651 {
    652 	mutex_enter(&pfs_ihash_lock);
    653 	LIST_REMOVE(pp, pfs_hash);
    654 	mutex_exit(&pfs_ihash_lock);
    655 }
    656 
    657 void
    658 procfs_revoke_vnodes(struct proc *p, void *arg)
    659 {
    660 	struct pfsnode *pfs, *pnext;
    661 	struct vnode *vp;
    662 	struct mount *mp = (struct mount *)arg;
    663 	struct pfs_hashhead *ppp;
    664 
    665 	if (!(p->p_flag & PK_SUGID))
    666 		return;
    667 
    668 	mutex_enter(&pfs_ihash_lock);
    669 	ppp = &pfs_hashtbl[PFSPIDHASH(p->p_pid)];
    670 	for (pfs = LIST_FIRST(ppp); pfs; pfs = pnext) {
    671 		vp = PFSTOV(pfs);
    672 		pnext = LIST_NEXT(pfs, pfs_hash);
    673 		mutex_enter(vp->v_interlock);
    674 		if (vp->v_usecount > 0 && pfs->pfs_pid == p->p_pid &&
    675 		    vp->v_mount == mp) {
    676 		    	vp->v_usecount++;
    677 		    	mutex_exit(vp->v_interlock);
    678 			mutex_exit(&pfs_ihash_lock);
    679 			VOP_REVOKE(vp, REVOKEALL);
    680 			vrele(vp);
    681 			mutex_enter(&pfs_ihash_lock);
    682 		} else {
    683 			mutex_exit(vp->v_interlock);
    684 		}
    685 	}
    686 	mutex_exit(&pfs_ihash_lock);
    687 }
    688 
    689 int
    690 procfs_proc_lock(int pid, struct proc **bunghole, int notfound)
    691 {
    692 	struct proc *tp;
    693 	int error = 0;
    694 
    695 	mutex_enter(proc_lock);
    696 
    697 	if (pid == 0)
    698 		tp = &proc0;
    699 	else if ((tp = proc_find(pid)) == NULL)
    700 		error = notfound;
    701 	if (tp != NULL && !rw_tryenter(&tp->p_reflock, RW_READER))
    702 		error = EBUSY;
    703 
    704 	mutex_exit(proc_lock);
    705 
    706 	*bunghole = tp;
    707 	return error;
    708 }
    709 
    710 void
    711 procfs_proc_unlock(struct proc *p)
    712 {
    713 
    714 	rw_exit(&p->p_reflock);
    715 }
    716 
    717 int
    718 procfs_doemul(struct lwp *curl, struct proc *p,
    719     struct pfsnode *pfs, struct uio *uio)
    720 {
    721 	const char *ename = p->p_emul->e_name;
    722 	return uiomove_frombuf(__UNCONST(ename), strlen(ename), uio);
    723 }
    724