2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1989, 1993
5 * The Regents of the University of California. All rights reserved.
6 * (c) UNIX System Laboratories, Inc.
7 * All or some portions of this file are derived from material licensed
8 * to the University of California by American Telephone and Telegraph
9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10 * the permission of UNIX System Laboratories, Inc.
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37 #include <sys/cdefs.h>
39 #include "opt_inet6.h"
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/dirent.h>
45 #include <sys/kernel.h>
47 #include <sys/malloc.h>
49 #include <sys/mount.h>
50 #include <sys/mutex.h>
52 #include <sys/rmlock.h>
53 #include <sys/refcount.h>
54 #include <sys/signalvar.h>
55 #include <sys/socket.h>
56 #include <sys/vnode.h>
58 #include <netinet/in.h>
59 #include <net/radix.h>
61 #include <rpc/types.h>
64 static MALLOC_DEFINE(M_NETADDR, "export_host", "Export host address structure");
66 #if defined(INET) || defined(INET6)
67 static struct radix_node_head *vfs_create_addrlist_af(
68 struct radix_node_head **prnh, int off);
70 static int vfs_free_netcred(struct radix_node *rn, void *w);
71 static void vfs_free_addrlist_af(struct radix_node_head **prnh);
72 static int vfs_hang_addrlist(struct mount *mp, struct netexport *nep,
73 struct export_args *argp);
74 static struct netcred *vfs_export_lookup(struct mount *, struct sockaddr *);
77 * Network address lookup element
80 struct radix_node netc_rnodes[2];
81 uint64_t netc_exflags;
82 struct ucred *netc_anon;
83 int netc_numsecflavors;
84 int netc_secflavors[MAXSECFLAVORS];
88 * Network export information
91 struct netcred ne_defexported; /* Default export */
92 struct radix_node_head *ne4;
93 struct radix_node_head *ne6;
97 * Build hash lists of net addresses and hang them off the mount point.
98 * Called by vfs_export() to set up the lists of export addresses.
101 vfs_hang_addrlist(struct mount *mp, struct netexport *nep,
102 struct export_args *argp)
105 struct radix_node_head *rnh;
107 struct radix_node *rn;
108 struct sockaddr *saddr, *smask = NULL;
109 #if defined(INET6) || defined(INET)
114 KASSERT(argp->ex_numsecflavors > 0,
115 ("%s: numsecflavors <= 0", __func__));
116 KASSERT(argp->ex_numsecflavors < MAXSECFLAVORS,
117 ("%s: numsecflavors >= MAXSECFLAVORS", __func__));
120 * XXX: This routine converts from a uid plus gid list
121 * to a `struct ucred' (np->netc_anon). This
122 * operation is questionable; for example, what should be done
123 * with fields like cr_uidinfo and cr_prison? Currently, this
124 * routine does not touch them (leaves them as NULL).
126 if (argp->ex_addrlen == 0) {
127 if (mp->mnt_flag & MNT_DEFEXPORTED) {
129 "MNT_DEFEXPORTED already set for mount %p", mp);
132 np = &nep->ne_defexported;
133 np->netc_exflags = argp->ex_flags;
134 np->netc_anon = crget();
135 np->netc_anon->cr_uid = argp->ex_uid;
136 crsetgroups(np->netc_anon, argp->ex_ngroups,
138 np->netc_anon->cr_prison = &prison0;
139 prison_hold(np->netc_anon->cr_prison);
140 np->netc_numsecflavors = argp->ex_numsecflavors;
141 bcopy(argp->ex_secflavors, np->netc_secflavors,
142 sizeof(np->netc_secflavors));
144 mp->mnt_flag |= MNT_DEFEXPORTED;
150 if (argp->ex_addrlen > MLEN) {
151 vfs_mount_error(mp, "ex_addrlen %d is greater than %d",
152 argp->ex_addrlen, MLEN);
157 i = sizeof(struct netcred) + argp->ex_addrlen + argp->ex_masklen;
158 np = (struct netcred *) malloc(i, M_NETADDR, M_WAITOK | M_ZERO);
159 saddr = (struct sockaddr *) (np + 1);
160 if ((error = copyin(argp->ex_addr, saddr, argp->ex_addrlen)))
162 if (saddr->sa_family == AF_UNSPEC || saddr->sa_family > AF_MAX) {
164 vfs_mount_error(mp, "Invalid saddr->sa_family: %d");
167 if (saddr->sa_len > argp->ex_addrlen)
168 saddr->sa_len = argp->ex_addrlen;
169 if (argp->ex_masklen) {
170 smask = (struct sockaddr *)((caddr_t)saddr + argp->ex_addrlen);
171 error = copyin(argp->ex_mask, smask, argp->ex_masklen);
174 if (smask->sa_len > argp->ex_masklen)
175 smask->sa_len = argp->ex_masklen;
178 switch (saddr->sa_family) {
181 if ((rnh = nep->ne4) == NULL) {
182 off = offsetof(struct sockaddr_in, sin_addr) << 3;
183 rnh = vfs_create_addrlist_af(&nep->ne4, off);
189 if ((rnh = nep->ne6) == NULL) {
190 off = offsetof(struct sockaddr_in6, sin6_addr) << 3;
191 rnh = vfs_create_addrlist_af(&nep->ne6, off);
198 vfs_mount_error(mp, "%s %s %d",
199 "Unable to initialize radix node head ",
200 "for address family", saddr->sa_family);
203 RADIX_NODE_HEAD_LOCK(rnh);
204 rn = (*rnh->rnh_addaddr)(saddr, smask, &rnh->rh, np->netc_rnodes);
205 RADIX_NODE_HEAD_UNLOCK(rnh);
206 if (rn == NULL || np != (struct netcred *)rn) { /* already exists */
209 "netcred already exists for given addr/mask");
212 np->netc_exflags = argp->ex_flags;
213 np->netc_anon = crget();
214 np->netc_anon->cr_uid = argp->ex_uid;
215 crsetgroups(np->netc_anon, argp->ex_ngroups,
217 np->netc_anon->cr_prison = &prison0;
218 prison_hold(np->netc_anon->cr_prison);
219 np->netc_numsecflavors = argp->ex_numsecflavors;
220 bcopy(argp->ex_secflavors, np->netc_secflavors,
221 sizeof(np->netc_secflavors));
228 /* Helper for vfs_free_addrlist. */
231 vfs_free_netcred(struct radix_node *rn, void *w)
233 struct radix_node_head *rnh = (struct radix_node_head *) w;
236 (*rnh->rnh_deladdr) (rn->rn_key, rn->rn_mask, &rnh->rh);
237 cred = ((struct netcred *)rn)->netc_anon;
244 #if defined(INET) || defined(INET6)
245 static struct radix_node_head *
246 vfs_create_addrlist_af(struct radix_node_head **prnh, int off)
249 if (rn_inithead((void **)prnh, off) == 0)
251 RADIX_NODE_HEAD_LOCK_INIT(*prnh);
257 vfs_free_addrlist_af(struct radix_node_head **prnh)
259 struct radix_node_head *rnh;
262 RADIX_NODE_HEAD_LOCK(rnh);
263 (*rnh->rnh_walktree)(&rnh->rh, vfs_free_netcred, rnh);
264 RADIX_NODE_HEAD_UNLOCK(rnh);
265 RADIX_NODE_HEAD_DESTROY(rnh);
266 rn_detachhead((void **)prnh);
271 * Free the net address hash lists that are hanging off the mount points.
274 vfs_free_addrlist(struct netexport *nep)
278 if (nep->ne4 != NULL)
279 vfs_free_addrlist_af(&nep->ne4);
280 if (nep->ne6 != NULL)
281 vfs_free_addrlist_af(&nep->ne6);
283 cred = nep->ne_defexported.netc_anon;
286 nep->ne_defexported.netc_anon = NULL;
292 * High level function to manipulate export options on a mount point
293 * and the passed in netexport.
294 * Struct export_args *argp is the variable used to twiddle options,
295 * the structure is described in sys/mount.h
296 * The do_exjail argument should be true if *mp is in the mountlist
297 * and false if not. It is not in the mountlist for the NFSv4 rootfs
298 * fake mount point just used for exports.
301 vfs_export(struct mount *mp, struct export_args *argp, bool do_exjail)
303 struct netexport *nep;
309 if ((argp->ex_flags & (MNT_DELEXPORT | MNT_EXPORTED)) == 0)
312 if ((argp->ex_flags & MNT_EXPORTED) != 0 &&
313 (argp->ex_numsecflavors < 0
314 || argp->ex_numsecflavors >= MAXSECFLAVORS))
318 pr = curthread->td_ucred->cr_prison;
319 lockmgr(&mp->mnt_explock, LK_EXCLUSIVE, NULL);
320 nep = mp->mnt_export;
321 if (argp->ex_flags & MNT_DELEXPORT) {
327 if (mp->mnt_exjail != NULL && mp->mnt_exjail->cr_prison != pr &&
330 /* EXDEV will not get logged by mountd(8). */
333 } else if (mp->mnt_exjail != NULL &&
334 mp->mnt_exjail->cr_prison != pr) {
336 /* EPERM will get logged by mountd(8). */
341 if (mp->mnt_flag & MNT_EXPUBLIC) {
342 vfs_setpublicfs(NULL, NULL, NULL);
344 mp->mnt_flag &= ~MNT_EXPUBLIC;
347 vfs_free_addrlist(nep);
348 mp->mnt_export = NULL;
353 mp->mnt_exjail = NULL;
354 mp->mnt_flag &= ~(MNT_EXPORTED | MNT_DEFEXPORTED);
357 atomic_subtract_int(&pr->pr_exportcnt, 1);
361 if (argp->ex_flags & MNT_EXPORTED) {
364 if (mp->mnt_exjail == NULL) {
366 if (do_exjail && nep != NULL) {
367 vfs_free_addrlist(nep);
368 memset(nep, 0, sizeof(*nep));
371 } else if (mp->mnt_exjail->cr_prison != pr) {
378 nep = malloc(sizeof(struct netexport), M_MOUNT,
380 mp->mnt_export = nep;
383 if (argp->ex_flags & MNT_EXPUBLIC) {
384 if ((error = vfs_setpublicfs(mp, nep, argp)) != 0) {
386 mp->mnt_export = NULL;
393 if (do_exjail && mp->mnt_exjail == NULL) {
394 mp->mnt_exjail = crhold(curthread->td_ucred);
395 atomic_add_int(&pr->pr_exportcnt, 1);
397 mp->mnt_flag |= MNT_EXPUBLIC;
400 if (argp->ex_numsecflavors == 0) {
401 argp->ex_numsecflavors = 1;
402 argp->ex_secflavors[0] = AUTH_SYS;
404 if ((error = vfs_hang_addrlist(mp, nep, argp))) {
406 mp->mnt_export = NULL;
412 if (do_exjail && mp->mnt_exjail == NULL) {
413 mp->mnt_exjail = crhold(curthread->td_ucred);
414 atomic_add_int(&pr->pr_exportcnt, 1);
416 mp->mnt_flag |= MNT_EXPORTED;
421 lockmgr(&mp->mnt_explock, LK_RELEASE, NULL);
423 * Once we have executed the vfs_export() command, we do
424 * not want to keep the "export" option around in the
425 * options list, since that will cause subsequent MNT_UPDATE
426 * calls to fail. The export information is saved in
427 * mp->mnt_export, so we can safely delete the "export" mount option
430 vfs_deleteopt(mp->mnt_optnew, "export");
431 vfs_deleteopt(mp->mnt_opt, "export");
436 * Get rid of credential references for this prison.
439 vfs_exjail_delete(struct prison *pr)
446 * Since this function is called from prison_cleanup() after
447 * all processes in the prison have exited, the value of
448 * pr_exportcnt can no longer increase. It is possible for
449 * a dismount of a file system exported within this prison
450 * to be in progress. In this case, the file system is no
451 * longer in the mountlist and the mnt_exjail will be free'd
452 * by vfs_mount_destroy() at some time. As such, pr_exportcnt
453 * and, therefore "i", is the upper bound on the number of
454 * mnt_exjail entries to be found by this function.
456 i = atomic_load_int(&pr->pr_exportcnt);
457 KASSERT(i >= 0, ("vfs_exjail_delete: pr_exportcnt negative"));
460 mtx_lock(&mountlist_mtx);
462 TAILQ_FOREACH(mp, &mountlist, mnt_list) {
464 if (mp->mnt_exjail != NULL &&
465 mp->mnt_exjail->cr_prison == pr) {
467 error = vfs_busy(mp, MBF_MNTLSTLOCK | MBF_NOWAIT);
470 * If the vfs_busy() fails, we still want to
471 * get rid of mnt_exjail for two reasons:
472 * - a credential reference will result in
473 * a prison not being removed
474 * - setting mnt_exjail NULL indicates that
475 * the exports are no longer valid
476 * The now invalid exports will be deleted
477 * when the file system is dismounted or
478 * the file system is re-exported by mountd.
482 if (mp->mnt_exjail != NULL &&
483 mp->mnt_exjail->cr_prison == pr) {
485 mp->mnt_exjail = NULL;
497 lockmgr(&mp->mnt_explock, LK_EXCLUSIVE, NULL);
499 if (mp->mnt_exjail != NULL &&
500 mp->mnt_exjail->cr_prison == pr) {
502 mp->mnt_exjail = NULL;
503 mp->mnt_flag &= ~(MNT_EXPORTED | MNT_DEFEXPORTED);
505 vfs_free_addrlist(mp->mnt_export);
506 free(mp->mnt_export, M_MOUNT);
507 mp->mnt_export = NULL;
510 lockmgr(&mp->mnt_explock, LK_RELEASE, NULL);
515 mtx_lock(&mountlist_mtx);
523 mtx_unlock(&mountlist_mtx);
527 * Set the publicly exported filesystem (WebNFS). Currently, only
528 * one public filesystem is possible in the spec (RFC 2054 and 2055)
531 vfs_setpublicfs(struct mount *mp, struct netexport *nep,
532 struct export_args *argp)
539 * mp == NULL -> invalidate the current info, the FS is
540 * no longer exported. May be called from either vfs_export
541 * or unmount, so check if it hasn't already been done.
544 if (nfs_pub.np_valid) {
545 nfs_pub.np_valid = 0;
546 if (nfs_pub.np_index != NULL) {
547 free(nfs_pub.np_index, M_TEMP);
548 nfs_pub.np_index = NULL;
555 * Only one allowed at a time.
557 if (nfs_pub.np_valid != 0 && mp != nfs_pub.np_mount)
561 * Get real filehandle for root of exported FS.
563 bzero(&nfs_pub.np_handle, sizeof(nfs_pub.np_handle));
564 nfs_pub.np_handle.fh_fsid = mp->mnt_stat.f_fsid;
566 if ((error = VFS_ROOT(mp, LK_EXCLUSIVE, &rvp)))
569 if ((error = VOP_VPTOFH(rvp, &nfs_pub.np_handle.fh_fid)))
575 * If an indexfile was specified, pull it in.
577 if (argp->ex_indexfile != NULL) {
578 if (nfs_pub.np_index == NULL)
579 nfs_pub.np_index = malloc(MAXNAMLEN + 1, M_TEMP,
581 error = copyinstr(argp->ex_indexfile, nfs_pub.np_index,
582 MAXNAMLEN, (size_t *)0);
585 * Check for illegal filenames.
587 for (cp = nfs_pub.np_index; *cp; cp++) {
595 free(nfs_pub.np_index, M_TEMP);
596 nfs_pub.np_index = NULL;
601 nfs_pub.np_mount = mp;
602 nfs_pub.np_valid = 1;
607 * Used by the filesystems to determine if a given network address
608 * (passed in 'nam') is present in their exports list, returns a pointer
609 * to struct netcred so that the filesystem can examine it for
610 * access rights (read/write/etc).
612 static struct netcred *
613 vfs_export_lookup(struct mount *mp, struct sockaddr *nam)
615 RADIX_NODE_HEAD_RLOCK_TRACKER;
616 struct netexport *nep;
617 struct netcred *np = NULL;
618 struct radix_node_head *rnh;
619 struct sockaddr *saddr;
621 nep = mp->mnt_export;
624 if ((mp->mnt_flag & MNT_EXPORTED) == 0)
628 * Lookup in the export list
633 switch (saddr->sa_family) {
642 RADIX_NODE_HEAD_RLOCK(rnh);
643 np = (struct netcred *) (*rnh->rnh_matchaddr)(saddr, &rnh->rh);
644 RADIX_NODE_HEAD_RUNLOCK(rnh);
645 if (np != NULL && (np->netc_rnodes->rn_flags & RNF_ROOT) != 0)
651 * If no address match, use the default if it exists.
653 if (np == NULL && (mp->mnt_flag & MNT_DEFEXPORTED) != 0)
654 return (&nep->ne_defexported);
660 * XXX: This comment comes from the deprecated ufs_check_export()
661 * XXX: and may not entirely apply, but lacking something better:
662 * This is the generic part of fhtovp called after the underlying
663 * filesystem has validated the file handle.
665 * Verify that a host should have access to a filesystem.
669 vfs_stdcheckexp(struct mount *mp, struct sockaddr *nam, uint64_t *extflagsp,
670 struct ucred **credanonp, int *numsecflavors, int *secflavors)
674 lockmgr(&mp->mnt_explock, LK_SHARED, NULL);
675 np = vfs_export_lookup(mp, nam);
677 lockmgr(&mp->mnt_explock, LK_RELEASE, NULL);
681 *extflagsp = np->netc_exflags;
682 if ((*credanonp = np->netc_anon) != NULL)
685 *numsecflavors = np->netc_numsecflavors;
686 KASSERT(*numsecflavors > 0,
687 ("%s: numsecflavors <= 0", __func__));
688 KASSERT(*numsecflavors < MAXSECFLAVORS,
689 ("%s: numsecflavors >= MAXSECFLAVORS", __func__));
691 if (secflavors && np->netc_numsecflavors > 0)
692 memcpy(secflavors, np->netc_secflavors, np->netc_numsecflavors *
694 lockmgr(&mp->mnt_explock, LK_RELEASE, NULL);